JANUARY 2012

PAPERS

VOLUME 60

NUMBER 1

IETPAK

(ISSN 0018-926X)

Antennas Maximum Gain of a Lossy Antenna ...... ......... ........ ....... .. ......... ........ ......... . A. Arbabi and S. Safavi-Naeini . Experimental Validation of Performance Limits and Design Guidelines for Small Antennas .... ........ ......... ......... .. .. ........ ......... ........ D. F. Sievenpiper, D. C. Dawson, M. M. Jacob, T. Kanar, S. Kim, J. Long, and R. G. Quarfoth Substrate Integrated Waveguide (SIW) Leaky-Wave Antenna With Transverse Slots ... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ....... J. Liu, D. R. Jackson, and Y. Long Subwavelength Substrate-Integrated Fabry-Prot Cavity Antennas Using Articial Magnetic Conductor ....... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... Y. Sun, Z. N. Chen, Y. Zhang, H. Chen, and T. S. P. See A Recongurable Wideband and Multiband Antenna Using Dual-Patch Elements for Compact Wireless Devices ...... .. .. ........ ...... H. F. Abutarboush, R. Nilavalan, S. W. Cheung, K. M. Nasr, T. Peter, D. Budimir, and H. Al-Raweshidy Frequency-Recongurable Monopole Antennas .. ........ ......... ......... ........ ........ A. Tariq and H. Ghafouri-Shiraz Low Prole Fully Planar Folded Dipole Antenna on a High Impedance Surface ......... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ...... A. Vallecchi, J. R. De Luis, F. Capolino, and F. De Flaviis Crumpling of PIFA Textile Antenna ...... ......... ........ ......... ... ...... ........ ......... ......... .. Q. Bai and R. Langley . Higher Order Mode Excitation for High-Gain Broadside Radiation From Cylindrical Dielectric Resonator Antennas . .. .. ........ ......... ......... ........ ......... ......... ........ ......... .. D. Guha, A. Banerjee, C. Kumar, and Y. M. M. Antar On the Characteristics of the Highly Directive Resonant Cavity Antenna Having Metal Strip Grating Superstrate ...... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ......... ...... A. Foroozesh and L. Shafai Nature of Cross-Polarized Radiations from Probe-Fed Circular Microstrip Antennas and Their Suppression Using Different Geometries of Defected Ground Structure (DGS) ... ......... ........ ......... ......... . C. Kumar and D. Guha . Single, Dual and Tri-Band-Notched Ultrawideband (UWB) Antennas Using Capacitively Loaded Loop (CLL) Resonators ..... ......... ........ ......... ....... .. ........ ......... ......... ........ . C.-C. Lin, P. Jin, and R. W. Ziolkowski . Leaky Wave Enhanced Feeds for Multibeam Reectors to be Used for Telecom Satellite Based Links ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ....... A. Neto, M. Ettorre, G. Gerini, and P. De Maagt Arrays 94 GHz Substrate Integrated Monopulse Antenna Array ......... ......... ........ ........ Y. J. Cheng, W. Hong, and K. Wu The Planar Ultrawideband Modular Antenna (PUMA) Array .... ......... ........ ....... S. S. Holland and M. N. Vouvakis A Two-Channel Time Modulated Linear Array With Adaptive Beamforming .. ......... ......... . Y. Tong and A. Tennant Aperiodic Array Layout Optimization by the Constraint Relaxation Approach . ......... ......... ........ ......... ......... .. .. ........ ......... ......... ...... T. N. Kaifas, D. G. Babas, G. S. Miaris, K. Siakavara, E. E. Vaadis, and J. N. Sahalos Time Reversal Based Broadband Synthesis Method for Arbitrarily Structured Beam-Steering Arrays . ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ....... D. Zhao, Y. Jin, B.-Z. Wang, and R. Zang A Generalized Hybrid Approach for the Synthesis of Uniform Amplitude Pencil Beam Ring-Arrays .. ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ......... ... O. M. Bucci and D. Pinchera Polarimetry With Phased Array Antennas: Theoretical Framework and Denitions ..... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ....... K. F. Warnick, M. V. Ivashina, S. J. Wijnholds, and R. Maaskant An Amplifying Recongurable Reectarray Antenna ... ......... ......... . ....... ......... ..... K. K. Kishor and S. V. Hum Design of Retrodirective Antenna Arrays for Short-Range Wireless Power Transmission ..... .. .. Y. Li and V. Jandhyala .

2 8 20 30 36 44 51 63 71 78 92 102 110 121 130 141 148 164 174 184 197 206

(Contents Continued on p. 1)

(Contents Continued from Front Cover) Anisotropic Impedance Surfaces for Linear to Circular Polarization Conversion ........ ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ..... E. Doumanis, G. Goussetis, J. L. Gmez-Tornero, R. Cahill, and V. Fusco Imaging and Propagation Transmitting-Mode Time Reversal Imaging Using MUSIC Algorithm for Surveillance in Wireless Sensor Network ... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ... X.-F. Liu, B.-Z. Wang, and J. L.-W. Li UWB Microwave Imaging of Objects With Canonical Shape .... ......... ........ ......... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ...... N. Ghavami, G. Tiberi, D. J. Edwards, and A. Monorchio Experimental Characterization of an UWB Propagation Channel in Underground Mines ........ ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ .... Y. Rissa, L. Talbi, and M. Ghaddar Statistical Prediction of Site Diversity Gainon Earth-Space Paths Based on RadarMeasurements in the UK .... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ......... .... C. Nagaraja and I. E. Otung Wideband Characterization of Backscatter Channels: Derivations and Theoretical Background ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ...... D. Arnitz, U. Muehlmann, and K. Witrisal Calibration of Electric Field Sensors Onboard the Resonance Satellite .. ........ ......... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ........ M. Sampl, W. Macher, C. Gruber, T. Oswald, H. O. Rucker, and M. Mogilevsky Numerical and Analytical Techniques A Second-Order Asymptotic Approximation for the Sommerfeld Half-Space Problem . ......... ..... ... ......... .. W. Lihh . Direct Rational Function Fitting Method for Accurate Evaluation of Sommerfeld Integrals in Stratied Media ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ...... T. N. Kaifas Integral Equation Modeling of Doubly Periodic Structures With an Efcient PMCHWT Formulation . ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ... S. Nosal, P. Soudais, and J.-J. Greffet Reduced-Order Models of Finite Element Approximations of Electromagnetic Devices Exhibiting Statistical Variability ...... ......... ........ ......... ......... ........ ......... ........ P. Sumant, H. Wu, A. Cangellaris, and N. Aluru Spherical ADI FDTD Method With Application to Propagation in the Earth Ionosphere Cavity ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ......... ..... D. L. Paul and C. J. Railton Analysis of Directional Logging Tools in Anisotropic and Multieccentric Cylindrically-Layered Earth Formations .... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ...... G.-S. Liu, F. L. Teixeira, and G.-J. Zhang Analytic Transient Analysis of Radiation From Ellipsoidal Reector Antennas for Impulse-Radiating Antennas Applications ... ......... ........ ......... ......... ........ ......... ..... S.-C. Tuan, H.-T. Chou, K.-Y. Lu, and H.-H. Chou An Analytic Solution of Transient Scattering From Perfectly Conducting Ellipsoidal Surfaces Illuminated by an Electromagnetic Plane Wave .. ......... ......... ........ ......... ..... H.-T. Chou, S.-C. Tuan, K.-Y. Lu, and H.-H. Chou Greens Function Extraction for Interfaces With Impedance Boundary Conditions ...... ....... E. Slob and K. Wapenaar Electromagnetic Field of a Horizontal Innitely Long Wire Over the Dielectric-Coated Earth .. ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ ......... .. Y. J. Zhi, K. Li, and Y. T. Fang Decomposable Medium Conditions in Four-Dimensional Representation ...... I. V. Lindell, L. Bergamin, and A. Favaro
COMMUNICATIONS

212 220 231 240 247 257 267 274 282 292 301 310 318 328 340 351 360 367

Novel UHF RFID Tag Antenna for Metallic Foil Packages ...... ......... ........ ......... .. J. Ryoo, J. Choo, and H. Choo Design of a Broadband All-Textile Slotted PIFA . ..... ... .. P. J. Soh, G. A. E. Vandenbosch, S. L. Ooi, and N. H. M. Rais . An Ultrawideband (UWB) Slotline Antenna Under Multiple-Mode Resonance ......... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ . X. D. Huang, C. H. Cheng, and L. Zhu A Compact Hepta-Band Loop-Inverted F Recongurable Antenna for Mobile Phone ... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... . Y. Li, Z. Zhang, J. Zheng, Z. Feng, and M. F. Iskander Hybrid Monopole-DRAs Using Hemispherical/Conical-Shaped Dielectric Ring Resonators: Improved Ultrawideband Designs ......... ......... ........ ......... ......... ........ ......... ......... ........ D. Guha, B. Gupta, and Y. M. M. Antar A Half Maxwell Fish-Eye Lens Antenna Based on Gradient-Index Metamaterials ...... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ....... Z. L. Mei, J. Bai, T. M. Niu, and T. J. Cui Dual-Polarized Planar Feed for Low-Prole Hemispherical Luneburg Lens Antennas .. ...... A. R. Weily and N. Nikolic TM Scattering by Perfectly Conducting Polygonal Cross-Section Cylinders: A New Surface Current Density Expansion Retaining up to the Second-Order Edge Behavior ..... ......... ......... .... G. Coluccini, M. Lucido, and G. Panariello . A Modication of the Kummers Method for Efcient Computation of the 2-D and 3-D Greens Functions for 1-D Periodic Structures ...... ......... ........ ......... ......... ........ ......... ......... ........ ......... ......... ........ .... S. P. Skobelev A Spatial Beam Splitter Consisting of a Near-Zero Refractive Index Medium .. ......... ......... ........ ......... ......... .. .. ........ ......... ......... ........ ......... ......... ........ ......... ......... ....... R.-B. Hwang, N.-C. Hsu, and C.-Y. Chin Extended Mode-Based Bandwidth Analysis for Asymmetric Near-Field Communication Systems .... Y. Tak and S. Nam
ERRATA

377 379 385 389 393 398 402 407 412 417 421

Errata to Three-Dimensional Near-Field Microwave Holography Using Reected and Transmitted Signals . ......... .. .. ........ ......... ......... ........ ......... ......... ........ ... R. K. Amineh, M. Ravan, A. Khalatpour, and N. K. Nikolova List of Reviewers for 2011 ....... ......... ......... ........ ......... ......... ........ ......... ......... ........ ......... ......... .

425 426

IEEE ANTENNAS AND PROPAGATION SOCIETY

All members of the IEEE are eligible for membership in the Antennas and Propagation Society and will receive on-line access to this TRANSACTIONS through IEEE Xplore upon payment of the annual Society membership fee of $24.00. Print subscriptions to this TRANSACTIONS are available to Society members for an additional fee of $36.00. For information on joining, write to the IEEE at the address below. Member copies of Transactions/Journals are for personal use only. ADMINISTRATIVE COMMITTEE M. SALAZAR PALMA, President S. R. BEST, President Elect J. S. TYO, Secretary-Treasurer 2012 2013 2014 M. ANDO* R. D. NEVELS* Y. M. M. ANTAR D. B. DAVIDSON J. T. BERNHARD* M. EL-SHENAWEE S. MACI S. K. RAO M. MARTINEZ-VZQUEZ M. W. SHIELDS D. F. SIEVENPIPER Honorary Life Members: R. C. HANSEN, W. R. STONE *Past President Committee Chairs and Representatives EuRAAP Representative: W. ROSS STONE Fellows Nominations Committee: L. C. KEMPEL Finance: J. S. TYO Gold Representative: Historian: K. D. STEPHAN IEEE Press Liaison: R. J. MAILLOUX IEEE Public Relations Representative: W. R. STONE IEEE Social Implications of Technology: R. L. HAUPT Institutional Listings: T. S. BIRD Joint Committee on High-Power Electromagnetics: Long-Range Planning: J. L. VOLAKIS Magazine Editor-in-Chief: W. R. STONE Meetings Coordination: S. A. LONG Membership: Nano Technology Council: G. W. HANSON & A. AKYURTLU New Technology Directions: S. C. HAGNESS Nominations: R. D. NEVELS PACE: J. M. JOHNSON

Antenna Measurements (AMTA): S. SCHNEIDER Antennas & Wireless Propagation Letters Editor-in-Chief: G. LAZZI Applied Computational EM Society (ACES): A. F. PETERSON AP-S/URSI Joint Meetings Committee: J. L. YOUNG Awards: C. G. CHRISTODOULOU Awards Coordinator: C. A. BALANIS Best Paper Awards H. NAKANO Chapter Activities: Z. SHEN CCIR: P. MCKENNA Committee on Man and Radiation: K. ITO Constitution and Bylaws: O. KILIC Digital Archive Editor-in-Chief: A. Q. MARTIN Distinguished Lecturers: D. R. JACKSON Education: D. F. KELLY EAB Continuing Education: S. R. RENGARAJAN Electronic Design Automation Council: M. VOUVAKIS Electronic Publications Editor-in-Chief: S. R. BEST EuCap Representative: J. L. VOLAKIS

Publications: T. S. BIRD Region 8 Representative: B. ARBESSER-RASTBURG Region 9 Representative: S. BARBIN Region 10 Representative: J. L-W. LI Sensor Council: A. I. ZAGHOUL, T. S. BIRD, M. W. SHIELDS Standards CommitteeAntennas: M. H. FRANCIS Standards CommitteePropagation: D. V. THIEL TABARC Correspondent: C. A. BALANIS TAB Magazines Committee: W. R. STONE TAB New Technology Directions Committee: S. C. HAGNESS TAB Transactions Committee: M. A. JENSEN Transactions Editor-in-Chief: M. A. JENSEN Transnational Committee: USAB Committee on Information Policy: S. WEIN USAB R&D Committee: A. C. SCHELL USNC/URSI: J. T. BERNHARD Women in Engineering Representative: P. F. WAHID

AP Transactions website: http://ieeeaps.org/aps_trans/

Albuquerque: MICHAEL HARRISON Argentina: GUSTAVO FANO Atlanta: KRISHNA NAISHADHAM Australian Capital Territory: DAVID MURRAY Baltimore: JACK SCHOFIELD Bangalore, India: KALARICKAPARAMBIL VINOY Beijing: DONGLIN SU Benelux: GUY VANDENBOSCH Boston: GREGORY CHARVAT Bulgaria: KATYA ASPARUHOVA Calcutta: DEBATOSH GUHA Central & South Italy: LUCIANO TARRICONE Central Texas: JEREMY PRUITT Chicago: DANILO ERRICOLO Cleveland: MAX SCARDELLETTI College Station: GREG HUFF Columbus: DIMITRIS PSYCHOUDAKIS Connecticut: CHARLOTTE BLAIR Croatia: RADOVAN ZENTNER Czech-Slovakia: JIRI DRINOVSKY Dallas: MURALI IYENGAR Dayton: ANDREW TERZUOLI Denver-Boulder: MICHAEL JANEZIC East Ukraine: KONSTYANTYN V. ILYENKO Eastern North Carolina: TODD NICHOLS Egypt: HADIA EL-HENNAWY Finland: ARTTU LUUKANEN

AP Transactions Manuscript Central website: http://tap-ieee.manuscriptcentral.com Chapter Chairs Milwaukee: SHRINIVAS G. JOSHI Mohawk Valley: E. PAUL RATAZZI Montreal: KE WU Morocco: MOHAMED ESSAAIDI Moscow: DIMITRY M. SAZONOV Nagoya, Japan: NOBUYOSHI KIKUMA Nanjing: WEI HONG New Jersey Coast: FILOMENA CITARELLA New South Wales: KARU ESSELLE Nizhny Novgorod: GEORGE L. PAKHOMOV North Italy: GIUSEPPE VECCHI North Jersey: HAR DAYAL, KIRIT DIXIT Norway: KARL MARTIN GJERTSEN Orlando: SIAMAK EBADI Ottawa: QIUBO YE Philadelphia: JACK NACHAMKIN Phoenix: STEVE ROCKWELL Poland: JERZY PIOTROWSKI Portugal: NUNO BORGES DE CARVALHO Princeton-Central Jersey: ALLEN KATZ Queensland: ASHLEY ROBINSON Rio de Janeiro: JOSE RICARDO BERGMANN San Diego: JERRY TWOMEY Santa Clara Valley - San Francisco: PAUL HARMS Saratov/Penza: NIKITA M. RYSKIN Seattle: LIXIN CAI
Seoul: JAEHOON CHOI SE Michigan: TAYFUN OZDEMIR Siberia Section Tomsk: ROMAN V. MESCHERIAKOV Singapore: LING CHUEN ONG South Africa: RIANA GESCHKE South Australia: CHRISTOPHE FUMEAUX Southern Alberta: ELISE FEAR Spain: JOSE I. ALONSO Spokane DON MINFORD Springeld: PAUL R. SIQUEIRA St Petersburg: SVETLANA P. ZUBKO Sweden: ANDERS RYDBERG Switzerland: MICHAEL MATTES Syracuse: HAKAN PARTAL Tainan: WEN-SHAN CHEN Taipei: HSI-TSENG CHOU Thailand: TUPTIM ANGKAEW Toronto: GEORGE ELEFTHERIADES Turkey: IBRAHIM TEKIN Tucson: HAO XIN UK/RI: ALI REZAZADEH Vancouver: ALON NEWTON Victoria: KAMRAN GHORBANI Washington, DC: BRIAN RIELY West Ukraine: IRYNA IVASENKO Winnipeg: VLADIMIR OKHMATOVSKI

Florida West Coast: JING WANG Foothill Section: FRANK FREYNE Fort Worth: SAIBUN TJUATJA France: MHAMED DRISSI FSU Georgia: GIORGI GHVEDASHVILI Fukuoka: KAZUNORI UCHIDA Germany: GEORG BOCK Gujarat, India: SHRI RAJEEV JYOTI Harbin, China: QUN WU Hong Kong: KWOK CHUNG Houston: STUART A. LONG Hungary: LAJOS NAGY Huntsville: ERIC R. GRIGORIAN Hyderabad, India: LAKSHMI MERUGU Indonesia: FITRI YULI ZULKIFLI Israel: SHAMUEL AUSTER Japan Council: KEIZO CHO Kansai: TAKESHI MANABE Kharkov: OLEKSANDR M. DUMIN Lebanon: ELIAS NASSAR Lithuania: VYTAUTAS URBANAVICIUS Long Island: BRYAN TROPPER Los Angeles: JOHN GIANVITTORIO Macau: KAM WENG TAM Malaysia: MOHAMAD A. RAHIM Melbourne, FL: RICK BOTSFORD

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION

Is the leading international engineering journal on the general topics of electromagnetics, antennas and wave propagation. The journal is devoted to antennas, including analysis, design, development, measurement, and testing; radiation, propagation, and the interaction of electromagnetic waves with discrete and continuous media; and applications and systems pertinent to antennas, propagation, and sensing, such as applied optics, millimeter- and sub-millimeter-wave techniques, antenna signal processing and control, radio astronomy, and propagation and radiation aspects of terrestrial and space-based communication, including wireless, mobile, satellite, and telecommunications. Author contributions of relevant full length papers and shorter Communications are welcomed. See inside back cover for Editorial Board.

IEEE Ofcers

GORDON W. DAY, President PETER W. STAECKER, President-Elect CELIA L. DESMOND, Secretary HAROLD L. FLESCHER, Treasurer MOSHE KAM, Past President

MICHAEL R. LIGHTNER, Vice President, Educational Activities DAVID A. HODGES, Vice President, Publication Services and Products HOWARD E. MICHEL, Vice President, Member and Geographic Activities STEVE M. MILLS, President, Standards Association FREDERICK C. MINTZER, Vice President, Technical Activities JAMES M. HOWARD, President, IEEE-USA PETER N. CLOUT, Director, Division IVElectromagnetics and Radiation

IEEE Executive Staff

DR. E. JAMES PRENDERGAST, Executive Director & Chief Operating Ofcer ALEXANDER PASIK, Information Technology THOMAS SIEGERT, Business Administration PATRICK MAHONEY, Marketing MATTHEW LOEB, Corporate Activities CECELIA JANKOWSKI, Member and Geographic Activities DOUGLAS GORHAM, Educational Activities ANTHONY DURNIAK, Publications Activities EILEEN LACH, General Counsel & Corporate Compliance JUDITH GORMAN, Standards Activities BETSY DAVIS, SPHR, Human Resources MARY WARD-CALLAN, Technical Activities CHRIS BRANTLEY, IEEE-USA

IEEE Periodicals Transactions/Journals Department

Staff Director: FRAN ZAPPULLA Editorial Director: DAWN MELLEY Production Director: PETER M. TUOHY Senior Managing Editor: WILLIAM A. COLACCHIO Associate Editor: DAWN L. MENENDEZ

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION (ISSN 0018-926X) is published monthly by the Institute of Electrical and Electronics Engineers, Inc. Responsibility for the contents rests upon the authors and not upon the IEEE, the Society/Council, or its members. IEEE Corporate Ofce: 3 Park Avenue, 17th Floor, New York, NY 10016-5997. IEEE Operations Center: 445 Hoes Lane, Piscataway, NJ 08854-4141. NJ Telephone: +1 732 981 0060. Price/Publication Information: Individual copies: IEEE Members $20.00 (rst copy only), nonmembers $126.00. (Note: Postage and handling charge not included.) Member and nonmember subscription prices available upon request. Copyright and Reprint Permissions: Abstracting is permitted with credit to the source. Libraries are permitted to photocopy for private use of patrons, provided the per-copy fee indicated in the code at the bottom of the rst page is paid through the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923. For all other copying, reprint, or republication permission, write to Copyrights and Permissions Department, IEEE Publications Administration, 445 Hoes Lane, Piscataway, NJ 08854-4141. Copyright 2012 by The Institute of Electrical and Electronics Engineers, Inc. All rights reserved. Periodicals Postage Paid at New York, NY and at additional mailing ofces. Postmaster: Send address changes to IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, IEEE, 445 Hoes Lane, Piscataway, NJ 08854-4141. GST Registration No. 125634188. CPC Sales Agreement #40013087. Return undeliverable Canada addresses to: Pitney Bowes IMEX, P.O. Box 4332, Stanton Rd., Toronto, ON M5W 3J4, Canada. IEEE prohibits discrimination, harassment, and bullying. For more information visit http://www.ieee.org/nondiscrimination. Printed in U.S.A.

Digital Object Identier 10.1109/TAP.2011.2181917

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Maximum Gain of a Lossy Antenna

Amir Arbabi, Student Member, IEEE, and Saeddin Safavi-Naeini, Member, IEEE
AbstractAn upper bound on the achievable gain of a lossy antenna is derived. This limit depends on the antenna size and a newly dened loss merit factor, which is shown to be a measure of the antenna material loss. The derived limit extends the well-known upper limit on the ratio of antenna directivity to its quality factor to the lossy antennas. Optimal antenna current distribution for the maximum gain is found, and the radiation pattern and antenna efciency are also presented. Index TermsAntenna maximum gain, lossy antenna, quality factor, small antenna.

I. INTRODUCTION

N ANTENNA is an essential component of a wireless system. Small-size antennas with optimal gain and bandwidth are on high demand for compact power-efcient portable radios. However, there is a well-known tradeoff between the antenna gain, its size, and its achievable bandwidth. It is shown that arbitrary high gain can be achieved from an arbitrary small perfect electric conductor (PEC) sphere if the current distribution on the sphere is selected in a proper way [1]. Chu derived an upper limit on the antenna directivity and ratio of antenna directivity to quality factor for omnidirectional antennas with a nite number of vectorial spherical modes [2]. Indeed, regardless of size, there is no limit on the gain of a lossless antenna, but for achieving high gain, the current amplitude on the antenna surface should be large. Large current amplitude on the antenna generates large reactive stored energy around the antenna. This reactive energy increases the antenna quality factor. Following Chu, several researchers tried to quantify the tradeoff between the antenna directivity and the large reactive energy stored around the antenna [1], [3][6]. In [5] and [6], it is shown that there is a limit on the ratio of directivity to quality factor of a general lossless antenna. In all of these studies, the antenna is assumed to be lossless and its quality factor is regarded as a measure of the antenna input impedance bandwidth. Although a large quality factor represents high reactive energy and small bandwidth, the relation between the conventional quality factor and bandwidth becomes less useful for low and moderate values of quality factor [7]. In addition, low- antennas can be matched to a desired impedance over a bandwidth wider than what is determined by their quality factors by using lumped, quasilumped

or, as has recently been reported, by exploiting engineered materials as a part of the matching circuit [8]. Furthermore, in millimeter wave, Tetrahertz, and optical frequency antennas, radiation efciency reduction due to the antenna losses is more restrictive than the bandwidth. Based on these facts, the study of the material loss and optimizing the antenna current distribution in a way to minimize its material loss and maximize the gain allow us to optimize the antenna structure under more realistic conditions. To our knowledge, this investigation has not been performed yet. Although [9] considered the effect of energy loss due to surrounding medium of an antenna, the antenna itself was still assumed to be lossless and enclosed in a lossless sphere. Therefore, the antenna material loss that resulted from a high reactive near eld was not considered. In this paper, the effect of the antenna material loss on the maximum achievable gain is analyzed. To this end, the antenna is assumed to be enclosed in a virtual sphere and its elds outside the sphere are expanded in terms of orthogonal vectorial spherical modes whose coefcients are related to the current distribution inside the sphere. It is shown that each of the vectorial spherical modes has its own radiation efciency which depends on the antenna size and a newly dened material loss merit factor. It is observed that the mode efciencies, except for a few lowest order modes, are very small. This result is a manifestation of the nite rank of the free space Greens function whose more general form is discussed in [10]. II. PROBLEM DESCRIPTION It is assumed that the antenna radiates in free space and the smallest sphere that encloses the antenna has a radius of . The coordinate system is chosen in a way that the origin is the center of the sphere and its axis is oriented in the antennas maximum radiation direction. It is also assumed that the antenna is fed by a lumped source and, therefore, the direct radiation of the applied source is ignored. The electric and magnetic eld satisfy the Maxwells equations in the entire space (1a) (1b) where is the equivalent current density (2)

Manuscript received January 30, 2011; revised May 01, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported by the National Science and Engineering Council (NSERC) of Canada and Research In Motion (RIM). The authors are with the Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada (e-mail: aarbabi@maxwell.uwaterloo.ca; safavi@maxwell.uwaterloo.ca). Digital Object Identier 10.1109/TAP.2011.2167934

is position dependent. It should be noticed and that the conductivity of the material can be considered as a part of . With this assumption, the current is nonzero only over the region occupied by the antenna, which is enclosed by the virtual sphere. Outside of this sphere, elds are solution to source free

0018-926X/$26.00 2011 IEEE

ARBABI AND SAFAVI-NAEINI: MAXIMUM GAIN OF A LOSSY ANTENNA

Maxwells equations and can be expanded in terms of vectorial spherical harmonics [11]

and constitute a set of mutually noted that although orthogonal vectorial functions, they do not form a complete set as shown in (8a) and (8b) at the bottom of the page. The total radiated power by the antenna can be found using (3a) and (3b) with integration of the Poyntings vector (9) because of the orthogonality of the modes, the total radiated power is the superposition of the power radiated by each individual mode. Maximum radiation intensity, as was assumed at 0 direction and can be found from the beginning, is in the (3a) in the far-eld region as

(3a)

(3b) where order, is the spherical Hankel function of the second are the spherical harmonics (4) and is a differential operator dened as (5) and in (3a) and (3b) are coefcients of the The and modes, respectively. These coefcients can be found by projecting the volume current density on an orthogonal vectorial set. If represents the sphere volume, then (see Appendix A) (6a) (6b) and are given in (41a) and (41b) in Apwhere pendix A. It can be veried that and are mutually orthogonal (7a) (7b) (7c) where when is Kroneckers delta function which is nonzero only and, in this case, it is equal to , and are norms of and , respectively. These norms are given in (8a) and (8b) and are independent of . It should be

(10) Now the material loss will be considered. The dissipative loss of the antenna results from the antenna material loss and is given by

(11) It will be shown that for a given antenna size, the antenna gain only depends on a dimensionless parameter dened as (12) which is referred to as loss merit factor. Using this new parameter, material loss becomes (13) is an effective loss merit factor for the antenna and is dened according to (13). This parameter is equal to the loss merit factor of the antenna material if the antenna is made of only one type of material. For an inhomogeneous antenna, the effective loss merit factor is smaller than the largest merit factor of its different constituents.

(8a) (8b)

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

The volume current density can be written as the summation and of two parts: Its projection on the space spanned by and a part which is orthogonal to this space, that is (14) where is orthogonal to and (15) and and are projections of and are given by on normalized and

and by dening (20a) (20b) Equation (19) can be rewritten as

(21) (16a) (16b) and are efciencies of it can be seen from (21) that and modes. Finally, the antenna each individual gain can be found to be given by (22), shown at the bottom of the page. III. OPTIMIZATION OF THE GAIN In (22), the terms in the numerator and the denominator are all positive and is only present in the denominator. To maximize the fraction, the denominator can be minimized independently to zero. As was by setting the nonradiating current density does not radiate and according to mentioned in Section II, (18), it adds to the dissipated power; therefore, the optimum . Similar reasoning leads to antenna should have (23) By dening (24a) (24b) (18) (24c) (24d)

is the nonradiating part of the equivalent current density. The vanish identically at any point outside elds generated by the sphere volume . The nonradiating current density neinor modes outside the sphere. Further disther excites cussion on nonradiating current densities can be found in [12]. From (14) and (15), the norm of can be written as

(17) plugging the left-hand side of (17) in (13) gives

The antenna total input power is the summation of the radiated and dissipated power

Equation (22) can be rewritten as

(25)

(19)

and and Because of the symmetry of (25) with respect to and , equating partial derivatives of the with respect to

(22)

ARBABI AND SAFAVI-NAEINI: MAXIMUM GAIN OF A LOSSY ANTENNA

to zero will result in similar equations and identical optimal values for these coefcients

(26a) (26b) Furthermore, the numerator is maximized by requiring (27) using (26) and (27), (25) is simplied as

(28)
Fig. 1. Mode efciencies for TM ( of n for kR = 10 and M = 10 . ) and TE ( ) modes as a function

Finally,

and

give (29)

and (30a) (30b)

IV. RESULTS AND DISCUSSION Plugging in norms of and mode efciencies are found as from (8) into (20), the

(31a) (31b)

Fig. 2. Maximum antenna gain as a function of kR for different values of M for kR between 0.05 and 0.5. The dashed line curve shows Harringtons normal = (kR) + 2kR. gain G

These equations show that the mode efciencies and, therefore, and the antenna the maximum gain are only a function of size. For copper, silver, and gold, the loss merit factor ( ) at and for a dielectric microwave frequencies is on the order of 10 and a loss tangent of material with a permittivity of is about 5 . Fig. 1 shows the mode efciencies as a function of for 10 and . As can be seen and show step-like behavior and are from this gure, almost equal to each other for the same value of . This step-like behavior is reminiscent of the assumptions made by Chu and Harrington for omn-directional and directional antennas [1], [2]. For a directional antenna, based on the mode impedance for each of spherical modes, Harrington assumed that only modes

are propagating modes and other modes are in with cutoff. Our results show that the efciency of each individual spherical mode is independent of . The step-like dependency of efciencies on can be regarded as a cutoff, but the cutoff edge is not only a function of the antenna size but also antenna material (effective loss merit factor). For example, in Fig. 1, can be assumed to be propagating modes, modes with and other modes have small efciency and can be ignored. Figs. 2 and 3 show the maximum antenna gain for different . The dashed line curve shows Harringtons values of . It is clear normal gain dened as , the maximum gain that for a large and moderate value of is larger than what Harrington found using only modes with . This larger gain comes with higher quality factor and smaller input impedance bandwidth.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 3. Maximum antenna gain as a function of kR for different values of M for kR between 0.5 and 5. The dashed line curve shows Harringtons normal kR kR. gain G

= ( ) +2

Fig. 5. Efciency of the antenna with maximum gain as a function of kR for different values of the antenna effective loss merit factor (M ).

V. CONCLUSION In this paper, a new fundamental limit on antenna gain was introduced. It was shown that although a lossless antenna regardless of its size can have arbitrarily high gain, the antenna dissipative loss limits the gain for a real antenna. The maximum gain is a function of the antenna size and its material loss merit factor. General plots for maximum gain for different antenna size and loss merit factor were provided. APPENDIX SPHERICAL HARMONICS COEFFICIENTS OF THE FIELD OF A VOLUME CURRENT DENSITY Coefcients of eld expansion in terms of vectorial spherical harmonics can be found from volume current and charge density as [11]

Fig. 4. Radiation pattern of the antenna with maximum gain with R and M .

= 10

2 =
(33a)

Using (14), (16), and (30), the equivalent current distribution inside the sphere for an antenna with maximum gain is given by

(33b) where is the speed of light in vacuum and is the volume charge density related to current density by the continuity equation

(32) (34) Fig. 4 shows the radiation pattern of an antenna with the maximum gain for and . Since and are almost the same, the radiation patterns in different cutting planes containing the axis are similar. As can be seen from this gure, the side lobe level is about 20 dB. Fig. 5 shows the efciency of the antenna with maximum gain as a function of its size for different values of the effective loss merit factor. after substituting from (34) into (33a), the rst term in the right-hand side can be written as

ARBABI AND SAFAVI-NAEINI: MAXIMUM GAIN OF A LOSSY ANTENNA

(42b) (35) REFERENCES the surface integral is zero because is assumed to be enclosed in the volume . The right-hand side of (33b) can be simplied as
[1] R. F. Harrington, Time Harmonic Electromagnetic Fields. New York: McGraw-Hill, 1961. [2] L. J. Chu, Physical limitations of omni-directional antennas, J. Appl. Phys., vol. 19, pp. 11631175, Dec. 1948. [3] H. A. Wheeler, Small antennas, IEEE Trans. Antennas Propag., vol. AP-23, no. 4, pp. 462469, Jul. 1975. [4] R. C. Hansen, Fundamental limitations in antennas, Proc. IEEE, vol. 69, no. 2, pp. 170182, Feb. 1981. [5] R. L. Fante, Maximum possible gain for an arbitrary ideal antenna with specied quality factor, IEEE Trans. Antennas Propag., vol. AP-40, no. 12, pp. 15861588, Dec. 1992. [6] W. Geyi, Physical limitations of antenna, IEEE Trans. Antennas Propag., vol. 51, no. 8, pp. 21162123, Aug. 2003. [7] S. R. Best, The Foster reactance theorem and quality factor for antennas, IEEE Antennas Wireless Propag. Lett., vol. 3, pp. 306309, 2004. [8] R. W. Ziolkowski and A. Erentok, Metamaterial-based efcient electrically small antennas, IEEE Trans. Antennas Propag., vol. 54, no. 7, pp. 21132130, Jul. 2006. [9] A. Karlsson, Physical limitations of antennas in a lossy medium, IEEE Trans. Antennas Propag., vol. 52, no. 8, pp. 20272033, Aug. 2004. [10] D. A. B. Miller, Fundamental limit for optical componenets, J. Opt. Soc. Amer. B, vol. 24, no. 10, Oct. 2007. [11] J. D. Jackson, Classical Electrodynamics, 3rd ed. New York: Wiley, 1999. [12] A. J. Devaney and E. Wolf, Radiating and nonradiating classical current distributions and the elds they generate, Phys. Rev. D, vol. 8, no. 4, pp. 10441047, 1973. Amir Arbabi (S06) was born in Malayer, Iran, in 1984. He received the B.Sc. degree in electrical engineering from the University of Tehran, Tehran, Iran, in 2006, the M.Sc. degree in electrical engineering from the University of Waterloo, Waterloo, ON, Canada, in 2009, and is currently pursuing the Ph.D. degree in electrical engineering at the University of Illinois, Urbana-Champaign. He has been a Research Assistant with the Photonic Systems Laboratory, University of Illinois, Urbana-Champaign, since 2009. His research interests include plasmonic and nanophotonic devices, fast numerical methods for simulation and design of novel photonic devices, and fundamental features and limits in optics and electromagnetics. Mr. Arbabi was a recipient of the Ontario Graduate Scholarship and the Presidents Graduate Scholarship while at the University of Waterloo. He is a student member of the Optical Society of America and a reviewer for the IEEE PHOTONICS JOURNAL.

(36) and with similar reasoning, the surface integral has vanished. Using (35) and (36), (33a) and (33b) can be rewritten as

(37a) (37b) Using the following identities: (38) and (39) Equations (37a) and (37b) are further simplied as

(40a) (40b) and by dening

Saeddin Safavi-Naeini (M00) was born in Gachsaran, Iran, in 1951. He received the B.Sc. degree in electrical engineering from the University of Tehran, Tehran, Iran, in 1974 and the M.Sc. and Ph.D. degrees in electrical engineering from the University of Illinois, Urbana-Champaign, in 1975 and 1979, respectively. He joined the Department of Electrical and Computer Engineering, University of Tehran, as an Assistant Professor in 1980 and became an Associate Professor in 1988. Since 2002, he has been a Full Professor in the Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, ON, Canada. His research interests and activities include numerical electromagnetics applied to radio-frequency/microwave/millimeter-wave systems and circuits, antennas and propagation, wireless communication systems, very-high-speed digital circuits, and optical communication systems. He has been a scientic and technical consultant to a number of national and international telecommunication industrial and research organizations since 1980.

(41a) (41b) equations (40a) and (40b) become (42a)

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Experimental Validation of Performance Limits and Design Guidelines for Small Antennas
Daniel F. Sievenpiper, Fellow, IEEE, David C. Dawson, Member, IEEE, Minu M. Jacob, Student Member, IEEE, Tumay Kanar, Student Member, IEEE, Sanghoon Kim, Jiang Long, and Ryan G. Quarfoth, Student Member, IEEE

AbstractThe theoretical limit for small antenna performance that was derived decades ago by Wheeler and Chu governs design tradeoffs for size, bandwidth, and efciency. Theoretical guidelines have also been derived for other details of small antenna design such as permittivity, aspect ratio, and even the nature of the internal structure of the antenna. In this paper, we extract and analyze experimental performance data from a large body of published designs to establish several facts that have not previously been demonstrated: (1) The theoretical performance limit for size, bandwidth, and efciency are validated by all available experimental evidence. (2) Although derived for electrically small antennas, the same theoretical limit is also generally a good design rule for antennas that are not electrically small. (3) The theoretical predictions for the performance due to design factors such as permittivity, aspect ratio, and the internal structure of the antenna are also supported by the experimental evidence. The designs that have the highest performance are those that involve the lowest permittivity, have an aspect ratio close to unity, and for which the elds ll the minimum size enclosing sphere with the greatest uniformity. This work thus validates the established theoretical design guidelines. Index TermsBandwidth, dielectric resonator antenna, efciency, fractal, metamaterial, planar antenna, quality factor, slot antenna, small antenna.

I. INTRODUCTION MALL antennas have been an important topic of research for many decades, and interest in the eld is increasing with the development of new systems that require broadband antennas with a small form factor. The analysis of small antennas is generally considered to have begun with the work of Wheeler [1] and Chu [2], who established the theoretical limits that show how electrical size and bandwidth are related. Since this early work, numerous authors have revisited these theories, and have suggested further renements. Although slightly more accurate, all of these new theories share the same basic conclusions established in the 1940sthat size can only be reduced at the expense of bandwidth or efciency. Furthermore, the early

Manuscript received February 10, 2011; revised May 17, 2011; accepted July 12, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported by SPAWAR under Contract N66001-03-2-8938. D. F. Sievenpiper, M. M. Jacob, T. Kanar, S. Kim, J. Long, and R. G. Quarfoth are all with the University of California San Diego, La Jolla, CA 92093 USA (e-mail: dsievenpiper@eng.ucsd.edu). D. C. Dawson is with SPAWAR Systems Center Pacic, San Diego, CA 92152 USA (e-mail: david.c.dawson@navy.mil). Digital Object Identier 10.1109/TAP.2011.2167938

papers as well as others that followed have provided theoretical guidelines for other aspects of small antenna design. In general, the best performance will be achieved if the dielectric constant is as low as possible, if the aspect ratio is close to unity, and if the internal structure of the antenna is such that the elds ll the minimum size enclosing sphere with the greatest possible uniformity. Along with the work that has been done to develop theoretical limits, a large amount of effort has been put into developing specic antenna designs in an attempt to optimize the relationship between size and bandwidth. In the 64 years since Wheelers rst paper, thousands of new antenna designs have been published, and each year we continue to see new publications exploring every conceivable arrangement of metal shapes and dielectric regions. However, many of these designs have sub-optimum performance, and could have been predicted to perform poorly if the theoretical design guidelines were more clearly understood from the start. In addition, many antenna designs are proposed for which performance is overestimated, such as by ignoring losses or incorrectly calculating the true electrical size of the antenna. This challenges not only antenna engineers, who must address these unphysical performance claims, but also system engineers, who end up relying on performance metrics that are ultimately unachievable. These issues may be caused in part because the theoretical design guidelines are not widely understood, and in fact have never been rigorously validated experimentally. Unfortunately, it is impossible to experimentally prove a physical theoryit can only be disproven by contradictory experimental evidence. Nonetheless, a theory that has been tested extensively and found to be true in all tests is generally accepted as correct, at least until contradictory evidence is found. In the eld of small antennas, there have been many attempts to optimize antenna designs to get as close as possible to the theoretical limits. However, each of these antennas represents a local optimization, and in each case it is possible that the authors have simply not chosen the best design, and perhaps another one may be found that could exceed the theoretical limit. The purpose of this paper is to systematically extract experimental results from a sufciently large sample of existing designs to demonstrate that the Wheeler-Chu limit is valid and correct across a broad range of electrical sizes and bandwidths. We furthermore show that the design guidelines that have been established for permittivity, aspect ratio, and the internal structure of the antenna are also supported by the experimental evidence. This paper establishes several important facts: (1) When the size and efciency are correctly calculated, the measured band-

0018-926X/$26.00 2011 IEEE

SIEVENPIPER et al.: EXPERIMENTAL VALIDATION OF PERFORMANCE LIMITS AND DESIGN GUIDELINES FOR SMALL ANTENNAS

width of an electrically small antenna does not exceed the theoretical limit, regardless of the design. (2) The theoretical limit for small antennas is also a good design guideline even for antennas that are not electrically small. (3) The experimental evidence supports the theoretical predictions that performance of a small antenna is maximized with low permittivity, low aspect ratio antennas, in which the elds ll the smallest enclosing sphere as uniformly as possible. For electrically small antennas, a class of wire cage designs appears to have a performance advantage compared to other types, while some of the new and popular concepts, such as fractals and metamaterials, do not appear to provide a performance advantage compared to conventional designs. There have been other studies comparing various specic antenna types to the theoretical limit, such as Best and Hannas recent paper in which they compared several different designs [3], and Bests paper involving specically planar designs [4]. However, to date there has not yet been a study which has systematically examined the body of experimental data to validate the theoretical limits over a wide variety of antenna types. Thus, the value of this paper is to demonstrate that the Wheeler-Chu limit has been extensively tested using 64 years of small antenna performance data, and has been found to be valid in all cases. The results shown here are also consistent with the design guidelines established decades ago [1]. It is expected that this will provide useful guidance for future small antenna designers. II. BACKGROUND ON SMALL ANTENNA THEORY In this section we give a brief overview of the theoretical analysis of small antennas and the results that are relevant to this study. For a more detailed examination of these theories, see for example the rst chapter in either of the books by Hansen [5] or Volakis [6]. The rst author to establish the link between antenna, bandwidth, and efciency was Wheeler [1]. He studied two simple small antennas, a cylindrical parallel plate capacitor and a cylindrical coil inductor. He calculated the radiation power factor for the capacitive antenna as (1) and for the inductive antenna as (2) where C or L is the capacitance or inductance, and G or R is the radiation shunt conductance or series resistance. He showed that the maximum power factor for a cylindrical antenna of either type, with circular area A and height b, is (3) , and denotes a shape factor that multiplies where the area A to obtain the effective area, as augmented by the eld outside the cylindrical volume. The shape factor approaches unity for thin, at capacitive antennas or long, thin inductors,

although it can be much larger for other shapes. For more details on the use of the shape factor, please refer to the original work [1]. Note that we have changed some variable names in order to be consistent throughout this paper. Wheeler also introduced an ideal spherical wire coil antenna [7], [8] that has a power factor of

(4) where a is the minimum radius of a sphere enclosing the antenna. Note that if the sphere is lled with a material having innite permeability, as Wheeler explains [9], expelling the avoidable stored energy from inside the antenna, then p can be increased by up to a factor of 3 compared to the air lled case. Its maximum value is fundamentally limited by the unavoidable stored energy outside the antenna, to (5) Wheeler also illustrated the relationship between power factor and bandwidth [9] which depends on the matching circuit and the allowable reection coefcient, as described by Fano [10]. We can recognize Wheelers denitions for the power factor in (1) and (2), as the inverse of the quality factor Q of an RC or RL circuit. By inverting (5) we nd a good approximation to the expressions for the minimum Q that are derived by other authors using more rigorous methods. In addition to this limitation on bandwidth, Wheeler also identied in his initial paper on this subject [1] several important guidelines for small antenna design which are relevant to the present study: (1) that the addition of an electrically large ground plane can potentially double p for a given volume, (2) that increasing the permittivity inside antenna decreases p roughly in proportion to , and (3) that increasing the permeability can increase p by up to a factor of 3. (4) He also later explained [9] that for non-spherical antennas, the power factor is reduced because the spherical volume is only partially utilized. Chu was the rst to derive the minimum quality factor Q of a small antenna based on an expansion of the eld in terms of spherical modes [2]. Q is generally dened [11], [12] by the ratio of stored energy W to radiated power P at a particular frequency for an otherwise lossless antenna, (6) where W is dened as (7) and and are the time-average, non-propagating, stored magnetic and electric energy. Unfortunately, Chu does not explicitly state a formula relating Q and size, thus requiring some further work by the reader to apply his results to antenna design. However, he does provide a plot showing a small dipole with an

10

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

ideal matching circuit as described by Fano [10] that has a bandwidth approximately proportional to . Hansen [13] and later McLean [12] followed Chus analysis to derive an expression for the Q of the lowest order mode in terms of the antennas electrical size

Q through the maximum allowable voltage standing wave ratio VSWR, or s

(11) In most cases we are concerned with matching the rst one or two modes, however Villalobos [25] has also derived limits for matching higher order modes. Additional studies have focused on the limitations for specic types of antennas. Examples that are relevant to this study are as follows. Sten [26] studied antennas near a conducting plane and determined that the proper measure of antenna size is a sphere that encloses both the antenna and its image currents. Ida [27] studied dielectric loaded monopole antennas, showing that the efciency-bandwidth product is reduced for large values of permittivity. Thal studied spherical wire antennas [28], and loop antennas [29], and concluded that the Q values are at least 3 times the theoretical limits for TE mode antennas, or at least 1.5 times the theoretical limit for TM mode antennas. Gustafsson [30] examined various shapes and determined the theoretical limits on Q, showing that it increases for any shape that deviates signicantly from a sphere, and that the maximum performance appears at an aspect ratio falling in the range between 1 and 2. Ghorbani [31] studied microstrip antennas and concluded that although the addition of resonant structures within the microstrip pattern are usually added to increase bandwidth, they actually reduce the maximum bandwidth that could be achieved with an ideal matching circuit. This is consistent with the guidance given by Wheeler [9], because the elds associated with these resonant structures are conned to a subset of the overall antenna volume. Finally, Stuart et al. [32] studied multi-resonant antennas. The authors showed through example that although the Q given by (6) does not deviate from the fundamental limit, the Q implied by the impedance of the antenna, (12) can be signicantly different if the antenna has two closely spaced resonances. Furthermore, they found that neither quantity is a good predictor of the half-power VSWR bandwidth for multiresonant antennas. Although the Q of an electrically small antenna can never be lower than the limits of (9) and (10), the bandwidth given by (11) is an approximation which assumes a matched antenna, and it is most accurate when the bandwidth is dened in terms of a low VSWR. Furthermore, as Fano shows [10], maximizing the reection coefcient toward a given tolerance increases the available bandwidth within that tolerance. Thus, by designing the antenna to meet a sufciently high VSWR tolerance over the band of interest, it is possible to exceed the bandwidth predicted by (11). In summary, the equations for minimum Q are always correct, but the equation for bandwidth based on a given Q can be exceeded by making the antenna or matching circuit multi-resonant, and by maximizing the reection coefcient within that band. This is illustrated in Fig. 1.

(8) Hansen also showed that loss can be represented as an additional resistance in series with the radiation resistance, so Q can be reduced at the expense of efciency. Therefore a more useful quantity for comparing antenna performance is the quality factor divided by efciency, . Collin and Rothschild [11] approached the problem by subtracting the energy associated with radiation from the total energy to nd simple expressions for the Q of each mode. They give the value for the lowest order spherical mode as (9) McLean [12] derived the propagating and non-propagating elds, and calculated Q from the ratio of these terms, arriving at the same result as Collin, above. He also found that the Q for circularly polarized antennas involving both TM and TE modes together is (10) Although this result is often associated with circularly polarized antennas, Pozar [14] claried that this formula for Q is simply a result of using two modes, and is not specically a function of the polarization of the antenna. Other published papers have provided various other expressions for the radiation Q, including Fante [15], Geyi [16], Hansen and Collin [17], Thal [18], and Vandenbosch [19]. However, (9) and (10) above are generally accepted today as correct. Furthermore, the other variations that have been explored generally deviate from the expressions above by only a small amount. There has also been work to simultaneously optimize gain and Q, such as the preliminary work by Fante [20] which includes numerical results for maximum G/Q. This work was later contested by Thal [18]. Geyi [21] provides an analytical result, and nds that it is possible to simultaneously minimize Q to a value given in (10), while maximizing G/Q, giving a maximum gain of 1.5 for an omnidirectional antenna, or 3 for a directive antenna. Although most studies have focused on Q, the quantity that is of most interest to antenna engineers is frequency bandwidth, B. Among others, Geyi [22] addressed this issue, concluding that B and 1/Q are equivalent for antennas with , however, this claim has been disputed recently by Best [23]. In any case, when attempting to match a given load impedance there is a tradeoff between bandwidth and acceptable reection coefcient [10]. Yaghjian and Best [24] derived the relationship between B and

SIEVENPIPER et al.: EXPERIMENTAL VALIDATION OF PERFORMANCE LIMITS AND DESIGN GUIDELINES FOR SMALL ANTENNAS

11

Fig. 1. Example Smith chart plot for a single-resonant, matched antenna (solid line) consistent with (11) compared to a multi-resonant, unmatched antenna (dashed line). These curves illustrate that for a given reection coefcient toler, designing an antenna with multiple resonances and avoiding a perfect ance match can improve bandwidth, within the Bode limit.

This last discussion would seem to suggest that the fundamental limits on Q would have minimal utility for antenna design, because the quantity that system designers care about is bandwidth, not Q. However, multiple resonant modes in a single antenna must be orthogonal, either in polarization or space. For electrically small antennas, the two lowest order modes can be orthogonal in polarization, and can be designed to be close in frequency. This can indeed reduce Q and improve bandwidth, as illustrated by the fact that (10) for the case of two modes provides a Q that is one-half that of (9) for . However, higher spatial modes will generally occur at higher frequencies. If the structure is loaded with reactance elements to lower the frequency of the higher spatial modes, they will occupy a subset of the total antenna volume, thus further raising their Q. Finally, even if multiple tuned circuits are included in the matching network, the maximum bandwidth is ultimately governed by the Bode limit [33] (13) Lopez [34], [35] and Hansen [36], [37] have explained the potential bandwidth improvement for various numbers of tuned circuits. For example, for a VSWR of 2, one additional tuned circuit can improve the bandwidth by a factor of 2.3 and an innite number of tuned circuits would provide a theoretical bandwidth improvement of 3.8. However, in practice most of the benet is obtained with one or two tuned circuits, and an excessively complicated matching circuit would contain substantial losses. III. ANALYSIS PROCEDURE To compare various electrically small antenna designs to the theoretical performance limits described above, we extracted

experimental data from the published literature. A search for small and antenna using the online search engine IEEE Xplore at the end of 2010 yields 7484 papers, which is far too many to analyze. We limited our search to only publications in IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION. This is based on the rationale that this is the premier journal for antennas, so any antenna design that has lasting impact would eventually be published here in some form. We also included the IRE Transactions, the predecessor to this journal, although there were no papers published in IRE Transactions that met the search criteria and that included sufcient data to quantify the antenna performance. This limit resulted in 763 papers, which is a reasonable number. Our approach obviously cannot nd every possible small antenna design, because there are many that only appear in conference papers or other journals. However, we needed to use consistent criteria for inclusion of papers, and this sample size is sufcient for us to draw meaningful conclusions. Upon examining each of the 763 papers, we determined that many could be eliminated based on title alone. Papers that discussed small reectors, small arrays, and other topics not related to electrically small antennas were excluded. Papers that focused on antennas embedded in materials other than free space, such as water, the human body, or other lossy media, were also excluded. Furthermore, papers that described ultrawideband (UWB) antennas were excluded, because designs aimed at multi-octave bandwidth generally are not electrically small, and usually involve different design approaches than electrically small antennas. For a similar reason, antennas focused on high frequency bands, such as millimeter wave antennas, or antennas integrated on a semiconductor substrate were excluded because those papers are generally focused on goals other than optimizing size or bandwidth. Nonetheless, published antennas that passed through our manual lter but were found to not be electrically small were still left in the data set because they provide some insight into the range of applicability of the fundamental limits as design guidelines. In addition to the criteria described above, multiband antennas were also excluded because they are typically designed to optimize the number and spacing of bands, rather than the width of a single band. Diversity antennas were excluded for a similar reason. Tunable antennas were excluded unless we could identify one tuning point as representative of that design. In general, if a paper did not include sufcient information to evaluate the antenna, such as frequency, bandwidth, size, efciency, gain, or plots from which this data can be extracted, then it was not included. Although active matching techniques such as non-Foster circuits can potentially exceed the limits discussed here, we considered only passive structures. Finally, since our goal was to validate the theoretical limit, we only included papers with measured data. By manually ltering the papers as described above, we obtained 112 published antenna designs that contained sufcient experimental data for us to analyze. For each paper, we recorded the center frequency and fractional bandwidth, or extracted these from the frequencies of the band edges, or from plots of . For papers that included efciency data, we used the values provided by the authors. For those that did not, we

12

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

used radiation patterns and gain when these were available. We estimated the directivity using (14) where and are the 3 dB beamwidths in radians in the two orthogonal planes. Although this approximation is most correct for directive antennas, it is still sufcient for our needs here, where we aim to keep the errors to within a few tens of percent or less. We then estimated the efciency using the quoted gain and calculated directivity (15) We used this approach for 19 of the papers. For published articles that provided neither efciency nor a radiation pattern, but still quoted gain, we estimated the efciency by using (16) where the ideal gain depended on the type of antenna. Dipolelike antennas without a ground plane were assigned an ideal gain of 1.5, monopole-like antennas on a large ground plane that were vertically polarized with a null toward zenith were given an ideal gain of 3, and patch-like antennas on a large ground plane where there is one central lobe that rolls off rapidly toward the horizon were given an ideal gain of 6. We used this approach for 11 of the papers. For three of the papers involving moderately high Q designs on lossy dielectrics, the efciency was approximated using the measured antenna Q and the loss tangent of the dielectric,

Fig. 2. The method for determining the radius, a, of the smallest enclosing sphere. (a) For an antenna with no ground plane, it is the smallest sphere to enclose the entire antenna. (b) For antennas on a small ground plane with less radius, or closer than from an edge, the sphere encloses the entire than ground plane. (c) For antennas on an electrically large ground plane, the sphere encloses the antenna and its image currents.

(17) For these cases we calculated an implied Q from the bandwidth using (11). This approach was only used when efciency could not be estimated using any other means. For 39 of the papers, the efciency was either quoted as nearly 100% by the author, or was assumed to be nearly 100% based on the design. That assumption was only applied if there was no other data from which to extract efciency, and when such an assumption was considered reasonable, such as for low-Q antennas built using entirely low loss metals and dielectrics. While these methods are approximate, we expect that they will be accurate to within a few tens of percent or less, and errors of this magnitude will not have a signicant effect on the overall conclusions of this paper. The electrical size of the antenna was calculated as the radius of the smallest sphere which encloses the entire antenna, as shown in Fig. 2. For antennas that do not include a ground plane, this is straightforward. For those that include an electrically large ground plane, with a radius at the center of the operating band, then the sphere includes the antenna and the image currents, so the radius a equals the distance to the farthest

point on the antenna from the bottom center. For antennas on an electrically small ground plane, or closer than to the edge of the ground plane, the entire ground plane was included in the size. The value for k was taken at the center of the operating band. A maximum bandwidth efciency product, , was calculated using (9) for linearly polarized antennas, or (10) for circularly polarized antennas, and applying (11). We standardized all designs to a VSWR of 2, which is consistent with the requirements of many applications, and the vast majority of published papers. These equations were also applied as appropriate when the polarization was not stated or was indeterminate, but where a judgment of whether it involved one or two modes could be made from the symmetry of the antenna and the feed. In addition to recording size, bandwidth, and efciency data, we also analyzed the effects of design type, permittivity, and aspect ratio, to compare to the design guidelines discussed above. For antennas containing multiple dielectric materials, the permittivity of the material lling most of the resonant portion of the antenna was used. The aspect ratio was taken as the ratio of the largest to smallest dimension of the outside boundary of the antenna. For antennas on electrically large ground planes, the dimension normal to the ground plane was doubled in calculating the aspect ratio, so as to include the effect of the image currents, as shown in Fig. 2(c). We grouped the antennas into several categories depending on the style of design. Antennas such as patches, planar inverted F antennas (PIFAs) and other similar designs which included a ground plane and had a width greater than their height were designated as Planar [38][89]. Dipoles and other such structures were designated as Linear [90][105] as well as designs that involve dipole-like modes on metal sheets, regardless of their aspect ratio. This category also included any vertically polarized antenna on an electrically large ground plane that had a height

SIEVENPIPER et al.: EXPERIMENTAL VALIDATION OF PERFORMANCE LIMITS AND DESIGN GUIDELINES FOR SMALL ANTENNAS

13

Fig. 3. The measured product for 110 antenna designs published in the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION by the end of the year 2010. . Specic references on the outer edge of the performance limit are noted. The theoretical limits are derived by applying (11) to (9) and (10) using a VSWR of

exceeding its width, and therefore behaved as a monopole. A sub-category of the linear type was a class that we called Feed antennas [106][119]. These were antennas on small ground planes that were typically shaped as a mobile phone or other such object, in which a small resonant structure actually serves as an exciter or feed for a mode which involves the entire ground plane. This is a category for which under-reporting the true antenna electrical size by neglecting the ground plane is common. Dielectric resonator antennas (DRA) [120][123] also formed a separate category, as well as antennas that involved materials with relative permeability , which were designated as Magnetic [124][126]. One category that performed particularly well was called the Wire Cage [127][137] type. These antennas were generally complicated wire structures with roughly spherical shape, or having an aspect ratio close to 1. The nal categories included designs which take advantage of popular trends in antenna design, and attribute their performance to Fractal [138][141] or Metamaterial [142][147] features. However, antennas that only included a single period or unit of such concepts, such as a solitary split ring resonator, were generally lumped into one of the other categories as appropriate. For papers that included multiple designs, we chose the best performing design, or the smallest. There were two papers where the measured results are so far above the theoretical limits that they were labeled as Problematic [148], [149] and not included in the data set. In both of those cases, the issues with the results can be traced back to problems in how the measurements were performed. IV. RESULTS AND DISCUSSION The centerpiece of this work is shown in Fig. 3. We plot the bandwidth efciency product versus the electrical size for the 110 relevant small antenna papers published in the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, and compare the measured results to the theoretical limits. The curves

representing the theoretical limits are derived by applying (11) to (9) and (10) using a VSWR of , and including efciency to obtain

(18) where for linearly polarized or single-mode antennas, and for circularly polarized or dual-mode antennas. Note that (9) and (10) for the minimum Q are inviolable. However, direct data for Q is not readily available for most published antennas, so we are using measured bandwidth as a proxy for Q through (11). This is based on the assumption of a self-matched antenna without additional matching circuits, which is consistent with the vast majority of published antennas. It is possible to exceed this bandwidth-Q relationship of (11) by using additional matching circuits. However, it is never possible to exceed the Bode limit of (13) for a given Q [33]. Thus, although the Friedman antenna [48] stands out as exceeding the theoretical limits curves plotted in Fig. 3, it does not actually exceed the Wheeler-Chu limit. Furthermore, the potential benets of each additional matching circuit are well established [34][37]. The double-tuned matching circuit of the Friedman antenna can be expected to provide a bandwidth improvement factor of up to 2.8, [34][37] and the measured product of this antenna is well within this limit. It should also be noted that the Friedman antenna is similar in concept to the widely cited Goubau antenna [150]. The two are identical in electrical size and bandwidth performance. The exceptional performance of the Goubau antenna is often credited to its multi-mode design. In fact, both antennas include complicated matching networks. While Friedman explicitly uses an effective circuit approach to design his antenna, Goubau achieves the same result with wire loops, slots, and other features integrated throughout his antenna.

14

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

There are several important observations to note in Fig. 3. (1) No electrically small antennas have been published in this journal that exceed the theoretical limit. (2) That limit also provides a good guideline for the maximum bandwidth even for antennas of moderate to large electrical size. (3) For electrically small antennas, wire cage designs appear to have a performance advantage compared to other types. (4) For moderate electrical size there are many standard planar or linear designs that can come close to the theoretical limit. (5) Dielectric resonator antennas perform relatively poorly because they are based on high dielectric materials, as predicted by the design guidelines discussed above. (6) Magnetic antennas are expected to have a performance advantage of up to a factor of three compared to other types. However very few antennas based on magnetic materials appeared in our sample of papers. There is one design that does achieve good performance, so this may be a promising area for future research if low-loss magnetic materials can be realized at frequencies of interest. (7) Fractal and metamaterial designs, although popular in recent years, do not appear to provide any performance advantage compared to more conventional antenna types. From these observations, we can draw several important conclusions about the relative merits of different antenna types, and the relationship between these performance variations and the established theoretical design guidelines. In general, the high-performance wire cage designs include low permittivity, low aspect ratio, and they have their elds evenly distributed throughout their volume or surface, so they are consistent with established design guidelines discussed above. Furthermore, the dielectric resonator designs involve high permittivity materials, so their poor bandwidth performance relative to their size is expected. Finally, the poor performance of metamaterial and fractal designs is consistent with the idea that the elds associated with the antenna should ll the smallest enclosing sphere as uniformly as possible. These designs typically involve highly resonant structures embedded within the antenna which tend to concentrate the elds to those regions, so they effectively use a subset of the available volume. In addition to examining the relative performance of different antenna classes, we also specically examined the effect of permittivity and shape. Fig. 4 shows the measured product divided by the theoretical limit given by (18) for the ka value of each antenna as a function of permittivity. For antennas that include multiple materials, the permittivity corresponds to the material lling the main resonant structure of the antenna. Note that the maximum measured performance decreases with increasing dielectric constant over a wide range of antenna designs. Wheelers original paper [1] instructed that for a capacitive antenna with a shape factor , the power factor is reduced by (19) Thus, for a shape factor of unity, corresponding to a thin, at capacitor, the performance is reduced approximately by the inverse of the permittivity. This performance roll-off is less severe for higher shape factors, so a roll-off is not a strict rule, but

Fig. 4. The measured product divided by the theoretical limit, compared to the relative permittivity of the material lling the antenna. The performance , is reduced with increasing permittivity. The dashed line shows a trend of an approximation from Wheelers paper, for a shape factor of 1. Most of the designs that lie above the dashed line actually contained multiple materials, and we have recorded the highest permittivity value, thus overestimating the effective permittivity in these cases.

it is still a good overall design guideline for many types of antennas. Also, from the spread of points in Fig. 4 it is obviously possible to achieve even worse performance, such as by inefciently using the antenna volume. Although an estimate of the shape factor for each antenna would allow us to more directly compare the data with Wheelers formula, we must remember that these antennas include a wide variety of internal structures, and many deviate signicantly from an ideal small capacitor. Thus, the concept of the shape factor would be difcult to apply directly to this entire data set. It is worth noting that many of the dielectric resonator antennas lie very close to Wheelers prediction. There are also several antennas that lie above the theoretical curve, and in most of these cases the discrepancy from the theory has to do with the choice of permittivity. These designs either contain multiple dielectric materials [66], [72][74], [89], or the elds extend partially into air regions within or around the antenna [85], [103]. In these cases we have recorded the highest permittivity among several materials that make up the resonant portion of the antenna, and therefore have clearly overestimated the effective permittivity. Accurately determining the true effective permittivity would be difcult, so these data points are left as exceptions. There are two additional cases that either involve magnetic materials [124] which would be expected to have higher performance, or complex matching circuits [48] which have been discussed above. We can observe that as a general rule, the maximum measured performance is reduced with increasing permittivity, and that assuming a performance reduction of roughly as a useful guideline for many designs, particularly if the effective permittivity can be accurately calculated. The maximum performance also decreases with increasing aspect ratio, as shown in Fig. 5. For this plot, we have used the ratio of the largest to smallest exterior dimensions of the antenna, regardless of orientation, so our aspect ratio is always greater than one. For antennas that involved an electrically large

SIEVENPIPER et al.: EXPERIMENTAL VALIDATION OF PERFORMANCE LIMITS AND DESIGN GUIDELINES FOR SMALL ANTENNAS

15

Fig. 5. The measured product divided by the theoretical limit, compared to the aspect ratio of the antenna. The performance is reduced with increasing aspect ratio. The dashed line shows a good linear approximation to Gustaffsons , curve for vertically polarized cylindrical antennas with roughly corresponding to some of the antennas designated here as the Planar , corresponding approxitype. The dotted curve is for mately to our Linear type.

ground plane, the image currents were also included in these dimensions. Gustaffson [30], calculated the performance reduction versus aspect ratio for various ideal shapes. He showed that for most shapes the maximum performance is achieved with an aspect ratio between 1 and 2, and the performance diminished with aspect ratio at various rates depending on the antenna shape. Note that his denitions for antenna shapes do not correspond exactly to our categories. Nonetheless, some comparisons can still be made. In Fig. 5 we have added theoretical curves which are linear approximations to Gustaffsons curve for vertically antennas having a cylindrical shape. For cylindrical antennas with a , this curve can be compared with some of the antennas in our Planar category. The theoretical performance decreases by approximately a factor of 10 for each decade increase in the diameter/height ratio. In other words, the performance is inversely proportional to the aspect ratio. Not all of these designs have a circular cross section [76], so it is difcult to compare the aspect ratio directly with Gustaffsons ideal curves. A separate line is included for cylindrical antennas with , which describes some of the antennas in our Linear category. A well-designed example is Noguchis [101] dual mode helix antenna. Although a perfect match to ideal antenna types is not to be expected for so many different designs, the trends in measured antenna performance are still generally consistent with the theoretical predictions. For many designs, and particularly for planar type antennas, it is a useful guideline to assume that performance may be reduced at least in proportion to the inverse of the aspect ratio. V. CONCLUSION We have demonstrated that the Wheeler-Chu limit for electrically small antenna performance is supported by experimental evidence for all papers containing measured data that have been published in the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION. We argue that this is a sufciently large data set to validate the theoretical limit. We have also shown that

the limit serves as a good design guideline even for antennas that are not electrically small. We have further shown that the design guidelines for performance reduction with permittivity and aspect ratio agree with the theoretical predictions, to the extent that such a comparison can be made for a wide variety of antenna types. The expected performance for most antennas degrades approximately with the inverse of permittivity, and for planar type antennas it degrades approximately with the inverse of aspect ratio. Finally, we have shown that antenna types in which the resonant structure is restricted to a subset of the overall antenna volume perform poorly relative to those in which the elds are evenly distributed within the minimum size enclosing sphere. All of these conclusions are consistent with the existing theoretical design guidelines. In general, to design an electrically small antenna with the largest possible bandwidth efciency product, the antenna should have a low permittivity, an aspect ratio close to unity, and should have the stored elds distributed evenly throughout its volume. For this reason, in the electrically small regime, wire cage designs appear to have an advantage compared to other types. However, at larger scales there are many other designs that perform well compared to the theoretical limits. ACKNOWLEDGMENT The authors would like to thank S. Best for very helpful discussions about the relationship between Q and bandwidth, and multiresonant antennas. REFERENCES
[1] H. A. Wheeler, Fundamental limitations of small antennas, in Proc. IRE, 1947, vol. 35, pp. 14791484. [2] L. J. Chu, Physical limitations of omni-directional antennas, J. Appl. Phys., vol. 19, pp. 11631175, 1948. [3] S. R. Best and D. L. Hanna, A performance comparison of fundamental small-antenna designs, IEEE Antennas Propag. Mag., vol. 52, pp. 4770, 2010. [4] S. Best, Optimization of the bandwidth of electrically small planar antennas, presented at the Antenna Applications Symp., Monticello, IL, 2009. [5] R. C. Hansen, Electrically Small, Superdirective, And Superconducting Antennas. Hoboken, NJ: Wiley-Interscience, 2006. [6] J. L. Volakis, C.-C. Chen, and K. Fujimoto, Small Antennas: Miniaturization Techniques and Applications. New York: McGraw-Hill, 2010. [7] H. A. Wheeler, The spherical coil as an inductor, shield, or antenna, in Proc. IRE, 1958, vol. 46, pp. 15951602. [8] H. A. Wheeler, The radiansphere around a small antenna, in Proc. IRE, 1959, vol. 47, pp. 13251331. [9] H. Wheeler, Small antennas, IEEE Trans. Antennas Propag., vol. 23, pp. 462469, 1975. [10] R. M. Fano, Theoretical limitations on the broadband matching of arbitrary impedances R.L.E. Tech. Rep. 41, Jan. 2, 1948. [11] R. Collin and S. Rothschild, Evaluation of antenna Q, IEEE Trans. Antennas Propag., vol. 12, pp. 2327, 1964. [12] J. S. McLean, A re-examination of the fundamental limits on the radiation Q of electrically small antennas, IEEE Trans. Antennas Propag., vol. 44, p. 672, 1996. [13] R. C. Hansen, Fundamental limitations in antennas, Proc. IEEE, vol. 69, pp. 170182, 1981. [14] D. M. Pozar, New results for minimum Q, maximum gain, and polarization properties of electrically small arbitrary antennas, in Proc. EuCAP 3rd Eur. Conf. on Antennas and Propagation, 2009, pp. 19931996. [15] R. Fante, Quality factor of general ideal antennas, IEEE Trans. Antennas Propag., vol. 17, pp. 151155, 1969. [16] W. Geyi, A method for the evaluation of small antenna Q, IEEE Trans. Antennas Propag., vol. 51, pp. 21242129, 2003.

16

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[17] R. C. Hansen and R. E. Collin, A new Chu formula for Q, IEEE Antennas Propag. Mag., vol. 51, pp. 3841, 2009. [18] H. L. Thal, Gain and Q bounds for coupled TM-TE modes, IEEE Trans. Antennas Propag., vol. 57, pp. 18791885, 2009. [19] G. A. E. Vandenbosch, Reactive energies, impedance, and Q factor of radiating structures, IEEE Trans. Antennas Propag., vol. 58, pp. 11121127, 2010. [20] R. L. Fante, Maximum possible gain for an arbitrary ideal antenna with specied quality factor, IEEE Trans. Antennas Propag., vol. 40, pp. 15861588, 1992. [21] W. Geyi, Physical limitations of antenna, IEEE Trans. Antennas Propag., vol. 51, pp. 21162123, 2003. [22] W. Geyi, P. Jarmuszewski, and Y. Qi, The Foster reactance theorem for antennas and radiation Q, IEEE Trans. Antennas Propag., vol. 48, pp. 401408, 2000. [23] S. R. Best, The Foster reactance theorem and quality factor for antennas, IEEE Antennas Wireless Propag. Lett., vol. 3, pp. 306309, 2004. [24] A. D. Yaghjian and S. R. Best, Impedance, bandwidth, and Q of antennas, IEEE Trans. Antennas Propag., vol. 53, pp. 12981324, 2005. [25] M. C. Villalobos, H. D. Foltz, and J. S. McLean, Broadband matching limitations for higher order spherical modes, IEEE Trans. Antennas Propag., vol. 57, pp. 10181026, 2009. [26] J. C. E. Sten, A. Hujanen, and P. K. Koivisto, Quality factor of an electrically small antenna radiating close to a conducting plane, IEEE Trans. Antennas Propag., vol. 49, pp. 829837, 2001. [27] I. Ida, J. Sato, T. Sekizawa, H. Yoshimura, and K. Ito, Dependence of the efciency-bandwidth product on electrical volume of small dielectric loaded antennas, IEEE Trans. Antennas Propag., vol. 50, pp. 821826, 2002. [28] H. L. Thal, New radiation Q limits for spherical wire antennas, IEEE Trans. Antennas Propag., vol. 54, pp. 27572763, 2006. [29] H. L. Thal, A reevaluation of the radiation Q bounds for loop antennas, IEEE Antennas Propag. Mag., vol. 51, pp. 4752, 2009. [30] M. Gustafsson, C. Sohl, and G. Kristensson, Illustrations of new physical bounds on linearly polarized antennas, IEEE Trans. Antennas Propag., vol. 57, pp. 13191327, 2009. [31] A. Ghorbani, M. Ansarizadeh, and R. A. Abd-alhameed, Bandwidth limitations on linearly polarized microstrip antennas, IEEE Trans. Antennas Propag., vol. 58, pp. 250257, 2010. [32] H. R. Stuart, S. R. Best, and A. D. Yaghjian, Limitations in relating quality factor to bandwidth in a double resonance small antenna, IEEE Antennas Wireless Propag. Lett., vol. 6, pp. 460463, 2007. [33] H. W. Bode, Network Analysis and Feedback Amplier Design. New York: D. Van Nostrand, 1945. [34] A. R. Lopez, Review of narrowband impedance-matching limitations, IEEE Antennas Propag. Mag., vol. 46, pp. 8890, 2004. [35] A. R. Lopez, Rebuttal to Fano limits on matching bandwidth, IEEE Antennas Propag. Mag., vol. 47, pp. 128129, 2005. [36] R. C. Hansen, Fano limits on matching bandwidth, IEEE Antennas Propag. Mag., vol. 47, pp. 8990, 2005. [37] R. C. Hansen, Fano limits again, IEEE Antennas Propag. Mag., vol. 48, p. 75, 2006. [38] J. Anguera, L. Boada, C. Puente, C. Borja, and J. Soler, Stacked H-shaped microstrip patch antenna, IEEE Trans. Antennas Propag., vol. 52, pp. 983993, 2004. [39] J. M. Baracco, M. Paquay, and P. de Maagt, An electromagnetic bandgap curl antenna for phased array applications, IEEE Trans. Antennas Propag., vol. 53, pp. 173180, 2005. [40] W.-S. Chen, C.-K. Wu, and K.-L. Wong, Square-ring microstrip antenna with a cross strip for compact circular polarization operation, IEEE Trans. Antennas Propag., vol. 47, pp. 15661568, 1999. [41] K. L. Chung, A wideband circularly polarized H-shaped patch antenna, IEEE Trans. Antennas Propag., vol. 58, pp. 33793383, 2010. [42] J. S. Colburn and Y. Rahmat-Samii, Patch antennas on externally perforated high dielectric constant substrates, IEEE Trans. Antennas Propag., vol. 47, pp. 17851794, 1999. [43] J. S. Colburn, J. J. Lynch, A. Obatoyinbo, and T. Talty, Spirapole antenna for communication systems utilizing both satellite and terrestrial assets, IEEE Trans. Antennas Propag., vol. 53, pp. 866869, 2005. [44] M. Crews and G. Thiele, On the design of shallow depth T-bar fed slot antennas, IEEE Trans. Antennas Propag., vol. 25, pp. 833836, 1977. [45] L. M. Feldner, C. T. Rodenbeck, C. G. Christodoulou, and N. Kinzie, Electrically small frequency-agile PIFA-as-a-package for portable wireless devices, IEEE Trans. Antennas Propag., vol. 55, pp. 33103319, 2007.

[46] R. Fenwick, A new class of electrically small antennas, IEEE Trans. Antennas Propag., vol. 13, pp. 379383, 1965. [47] H. D. Foltz, J. S. McLean, and G. Crook, Disk-loaded monopoles with parallel strip elements, IEEE Trans. Antennas Propag., vol. 46, pp. 18941896, 1998. [48] C. Friedman, Wide-band matching of a small disk-loaded monopole, IEEE Trans. Antennas Propag., vol. 33, pp. 11421148, 1985. [49] V. F. Fusco, V. Srinivasan, and J. I. Syed, Switchable circular disk antenna for RCS and mobile communications applications, IEEE Trans. Antennas Propag., vol. 52, pp. 994998, 2004. [50] Y. Ge, K. P. Esselle, and T. S. Bird, A compact E-shaped patch antenna with corrugated wings, IEEE Trans. Antennas Propag., vol. 54, pp. 24112413, 2006. [51] W. Hong, N. Behdad, and K. Sarabandi, Size reduction of cavitybacked slot antennas, IEEE Trans. Antennas Propag., vol. 54, pp. 14611466, 2006. [52] W. Hong and K. Sarabandi, Low prole miniaturized planar antenna with omnidirectional vertically polarized radiation, IEEE Trans. Antennas Propag., vol. 56, pp. 15331540, 2008. [53] W. Hong and K. Sarabandi, Low-prole, multi-element, miniaturized monopole antenna, IEEE Trans. Antennas Propag., vol. 57, pp. 7280, 2009. [54] W. Hong and K. Sarabandi, Platform embedded slot antenna backed by shielded parallel plate resonator, IEEE Trans. Antennas Propag., vol. 58, pp. 28502857, 2010. [55] A. Hoorfar and A. Perrotta, An experimental study of microstrip antennas on very high permittivity ceramic substrates and very small ground planes, IEEE Trans. Antennas Propag., vol. 49, pp. 838840, 2001. [56] H. Iwasaki, A circularly polarized small-size microstrip antenna with a cross slot, IEEE Trans. Antennas Propag., vol. 44, pp. 13991401, 1996. [57] H. K. Kan and R. B. Waterhouse, A small CP-printed antenna using 120 sequential rotation, IEEE Trans. Antennas Propag., vol. 50, pp. 398399, 2002. [58] H. K. Kan and R. B. Waterhouse, Shorted spiral-like printed antennas, IEEE Trans. Antennas Propag., vol. 50, pp. 396397, 2002. [59] T. Kashiwa, T. Onishi, and I. Fukai, Analysis of microstrip antennas on a curved surface using the conformal grids FD-TD method, IEEE Trans. Antennas Propag., vol. 42, pp. 423427, 1994. [60] A. A. Kishk, L. K. Fong, W. C. Mok, and L. Kwai-Man, A wideband small size microstrip antenna proximately coupled to a hook shape probe, IEEE Trans. Antennas Propag., vol. 52, pp. 5965, 2004. [61] K.-L. Lau, K.-M. Luk, and K.-F. Lee, Design of a circularly-polarized vertical patch antenna, IEEE Trans. Antennas Propag., vol. 54, pp. 13321335, 2006. [62] B. Lee and F. J. Harackiewicz, Miniature microstrip antenna with a partially lled high-permittivity substrate, IEEE Trans. Antennas Propag., vol. 50, pp. 11601162, 2002. [63] R. Li, G. DeJean, M. M. Tentzeris, and J. Laskar, Development and analysis of a folded shorted-patch antenna with reduced size, IEEE Trans. Antennas Propag., vol. 52, pp. 555562, 2004. [64] F. Manshadi, End-loaded crossed-slot radiating elements, IEEE Trans. Antennas Propag., vol. 39, pp. 12371240, 1991. [65] G. Mayhew-Ridgers, J. W. Odendaal, and J. Joubert, Single-layer capacitive feed for wideband probe-fed microstrip antenna elements, IEEE Trans. Antennas Propag., vol. 51, pp. 14051407, 2003. [66] H. Mosallaei and K. Sarabandi, Antenna miniaturization and bandwidth enhancement using a reactive impedance substrate, IEEE Trans. Antennas Propag., vol. 52, pp. 24032414, 2004. [67] H. Nakano, H. Takeda, T. Honma, H. Mimaki, and J. Yamauchi, Extremely low-prole helix radiating a circularly polarized wave, IEEE Trans. Antennas Propag., vol. 39, pp. 754757, 1991. [68] H. Nakano, N. Suzuki, T. Ishii, and J. Yamauchi, Mesh antennas for dual polarization, IEEE Trans. Antennas Propag., vol. 49, pp. 715723, 2001. [69] H. Nakano, H. Iwaoka, K. Morishita, and J. Yamauchi, A wideband low-prole antenna composed of a conducting body of revolution and a shorted parasitic ring, IEEE Trans. Antennas Propag., vol. 56, pp. 11871192, 2008. [70] H. Nakano, Y. Ogino, and J. Yamauchi, Bent two-leaf antenna radiating a tilted, linearly polarized, wide beam, IEEE Trans. Antennas Propag., vol. 58, pp. 37213725, 2010. [71] N. Nasimuddin, Z. Chen, and X. Qing, Asymmetric-circular shaped slotted microstrip antennas for circular polarization and RFID applications, IEEE Trans. Antennas Propag., pp. 11, 2010, vol..

SIEVENPIPER et al.: EXPERIMENTAL VALIDATION OF PERFORMANCE LIMITS AND DESIGN GUIDELINES FOR SMALL ANTENNAS

17

[72] D. Psychoudakis, Y. H. Koh, J. L. Volakis, and J. H. Halloran, Design method for aperture-coupled microstrip patch antennas on textured dielectric substrates, IEEE Trans. Antennas Propag., vol. 52, pp. 27632765, Oct. 2004. [73] D. Psychoudakis, J. L. Volakis, Z. Wing, and J. W. Halloran, Cavity-backed miniature wideband UHF circular polarized antenna with textured dielectrics, IEEE Trans. Antennas Propag., vol. 54, pp. 35863592, Dec. 2006. [74] D. Psychoudakis, J. L. Volakis, Z. N. Wing, S. K. Pillai, and J. W. Halloran, Enhancing UHF antenna functionality through dielectric inclusions and texturization, IEEE Trans. Antennas Propag., vol. 54, pp. 317329, Feb. 2006. [75] R. R. Ramirez, F. De Flaviis, and N. G. Alexopoules, Single-feed circularly polarized microstrip ring antenna and arrays, IEEE Trans. Antennas Propag., vol. 48, pp. 10401047, 2000. [76] K. P. Ray, G. Kumar, and H. C. Lodwal, Hybrid-coupled broadband triangular microstrip antennas, IEEE Trans. Antennas Propag., vol. 51, pp. 139141, 2003. [77] W. S. T. Rowe and R. B. Waterhouse, Reduction of backward radiation for CPW fed aperture stacked patch antennas on small ground planes, IEEE Trans. Antennas Propag., vol. 51, pp. 14111413, 2003. [78] W. S. T. Rowe and R. B. Waterhouse, Integratable wide-band dual polarized antennas with rear eld cancellation, IEEE Trans. Antennas Propag., vol. 51, pp. 469477, 2003. [79] M. A. Saed, Efcient method for analysis and design of aperture-coupled rectangular microstrip antennas, IEEE Trans. Antennas Propag., vol. 41, pp. 986988, 1993. [80] D. Singh, C. Kalialakis, P. Gardner, and P. S. Hall, Small H-shaped antennas for MMIC applications, IEEE Trans. Antennas Propag., vol. 48, pp. 11341141, 2000. [81] A. Thumvichit, T. Takano, and Y. Kamata, Characteristics verication of a half-wave dipole very close to a conducting plane with excellent impedance matching, IEEE Trans. Antennas Propag., vol. 55, pp. 5358, 2007. [82] G. A. E. Vandenbosch and A. R. Van de Capelle, Study of the capacitively fed microstrip antenna element, IEEE Trans. Antennas Propag., vol. 42, pp. 16481652, 1994. [83] R. B. Waterhouse, S. D. Targonski, and D. M. Kokotoff, Design and performance of small printed antennas, IEEE Trans. Antennas Propag., vol. 46, pp. 16291633, 1998. [84] S.-H. Wi, Y.-S. Lee, and J.-G. Yook, Wideband microstrip patch antenna with U-shaped parasitic elements, IEEE Trans. Antennas Propag., vol. 55, pp. 11961199, 2007. [85] K.-L. Wong and T.-W. Chiou, Broad-band single-patch circularly polarized microstrip antenna with dual capacitively coupled feeds, IEEE Trans. Antennas Propag., vol. 49, pp. 4144, 2001. [86] K.-L. Wong, C.-L. Tang, and J.-Y. Chiou, Broadband probe-fed patch antenna with a W-shaped ground plane, IEEE Trans. Antennas Propag., vol. 50, pp. 827831, 2002. [87] S. H. Yeung, K. F. Man, K. M. Luk, and C. H. Chan, A trapeizform U-slot folded patch feed antenna design optimized with jumping genes evolutionary algorithm, IEEE Trans. Antennas Propag., vol. 56, pp. 571577, 2008. [88] B. Y. Zheng and Z. X. Shen, Effect of a nite ground plane on microstrip-fed cavity-backed slot antennas, IEEE Trans. Antennas Propag., vol. 53, pp. 862865, Feb. 2005. [89] Y. J. Zhou, C. C. Chen, and J. L. Volakis, Single-fed circularly polarized antenna element with reduced coupling for GPS arrays, IEEE Trans. Antennas Propag., vol. 56, pp. 14691472, May 2008. [90] E. Arnieri, L. Boccia, G. Amendola, and G. D. Massa, A compact high gain antenna for small satellite applications, IEEE Trans. Antennas Propag., vol. 55, pp. 277282, 2007. [91] R. Azadegan and K. Sarabandi, A novel approach for miniaturization of slot antennas, IEEE Trans. Antennas Propag., vol. 51, pp. 421429, 2003. [92] R. Azadegan and K. Sarabandi, Bandwidth enhancement of miniaturized slot antennas using folded, complementary, and self-complementary realizations, IEEE Trans. Antennas Propag., vol. 55, pp. 24352444, 2007. [93] N. Behdad and K. Sarabandi, Bandwidth enhancement and further size reduction of a class of miniaturized slot antennas, IEEE Trans. Antennas Propag., vol. 52, pp. 19281935, 2004. [94] A. Dastranj, A. Imani, and M. Naser-Moghaddasi, Printed wide-slot antenna for wideband applications, IEEE Trans. Antennas Propag., vol. 56, pp. 30973102, 2008.

[95] A. Erentok and R. W. Ziolkowski, Metamaterial-inspired efcient electrically small antennas, IEEE Trans. Antennas Propag., vol. 56, pp. 691707, 2008. [96] B. Ghosh, S. M. Haque, D. Mitra, and S. Ghosh, A loop loading technique for the miniaturization of non-planar and planar antennas, IEEE Trans. Antennas Propag., vol. 58, pp. 21162121, 2010. [97] F.-R. Hsiao and K.-L. Wong, Omnidirectional planar folded dipole antenna, IEEE Trans. Antennas Propag., vol. 52, pp. 18981902, 2004. [98] P. Jin and R. W. Ziolkowski, Low-Q, electrically small, efcient neareld resonant parasitic antennas, IEEE Trans. Antennas Propag., vol. 57, pp. 25482563, 2009. [99] J. M. Johnson and Y. Rahmat-Samii, The tab monopole, IEEE Trans. Antennas Propag., vol. 45, pp. 187188, 1997. [100] D. H. Kwon and Y. Kim, Small low-prole loop wideband antennas with unidirectional radiation characteristics, IEEE Trans. Antennas Propag., vol. 55, pp. 7277, 2007. [101] K. Noguchi, S. I. Betsudan, T. Katagi, and M. Mizusawa, A compact broad-band helical antenna with two-wire helix, IEEE Trans. Antennas Propag., vol. 51, pp. 21762181, 2003. [102] K. Sarabandi and R. Azadegan, Design of an efcient miniaturized UHF planar antenna, IEEE Trans. Antennas Propag., vol. 51, pp. 12701276, 2003. [103] T. G. Spence and D. H. Werner, A novel miniature broadband/multiband antenna based on an end-loaded planar open-sleeve dipole, IEEE Trans. Antennas Propag., vol. 54, pp. 36143620, 2006. [104] K. G. Thomas, N. Lenin, and M. Sreenivasan, Wide-band dual sleeve antenna, IEEE Trans. Antennas Propag., vol. 54, pp. 10341037, 2006. [105] R. B. Waterhouse and D. Novak, Uni-planar folded meander line slot antenna with short circuit, IEEE Trans. Antennas Propag., vol. 54, pp. 35493551, 2006. [106] Y.-W. Chi and K.-L. Wong, Compact multiband folded loop chip antenna for small-size mobile phone, IEEE Trans. Antennas Propag., vol. 56, pp. 37973803, 2008. [107] S. I. Latif, L. Shafai, and S. K. Sharma, Bandwidth enhancement and size reduction of microstrip slot antennas, IEEE Trans. Antennas Propag., vol. 53, pp. 9941003, 2005. [108] C.-T. Lee and K.-L. Wong, Internal WWAN clamshell mobile phone antenna using a current trap for reduced ground plane effects, IEEE Trans. Antennas Propag., vol. 57, pp. 33033308, 2009. [109] C.-W. Ling, C.-Y. Lee, C.-L. Tang, and S.-J. Chung, Analysis and application of an on-package planar inverted-f antenna, IEEE Trans. Antennas Propag., vol. 55, pp. 17741780, 2007. [110] C. T. Rodenbeck, Planar miniature RFID antennas suitable for integration with batteries, IEEE Trans. Antennas Propag., vol. 54, pp. 37003706, 2006. [111] G. Ruvio and M. J. Ammann, A novel wideband semi-planar miniaturized antenna, IEEE Trans. Antennas Propag., vol. 55, pp. 26792685, 2007. [112] K. M. Z. Shams and M. Ali, Study and design of a capacitively coupled polymeric internal antenna, IEEE Trans. Antennas Propag., vol. 53, pp. 985993, 2005. [113] S. K. Sharma, L. Shafai, and N. Jacob, Investigation of wide-band microstrip slot antenna, IEEE Trans. Antennas Propag., vol. 52, pp. 865872, 2004. [114] C.-M. Su, K.-L. Wong, C.-L. Tang, and S.-H. Yeh, EMC internal patch antenna for UMTS operation in a mobile device, IEEE Trans. Antennas Propag., vol. 53, pp. 38363839, 2005. [115] G. F. Tsachtsiris, C. F. Soras, M. P. Karaboikis, and V. T. Makios, Analysis of a modied Sierpinski gasket monopole antenna printed on dual band wireless devices, IEEE Trans. Antennas Propag., vol. 52, pp. 25712579, 2004. [116] Y.-S. Wang and S.-J. Chung, A short open-end slot antenna with equivalent circuit analysis, IEEE Trans. Antennas Propag., vol. 58, pp. 17711775, 2010. [117] K.-L. Wong and C.-H. Chang, WLAN chip antenna mountable above the system ground plane of a mobile device, IEEE Trans. Antennas Propag., vol. 53, pp. 34963499, 2005. [118] K.-L. Wong and C.-H. Chang, Surface-mountable EMC monopole chip antenna for WLAN operation, IEEE Trans. Antennas Propag., vol. 54, pp. 11001104, 2006. [119] K.-L. Wong, C.-H. Chang, and Y.-C. Lin, Printed PIFA EM compatible with nearby conducting elements, IEEE Trans. Antennas Propag., vol. 55, pp. 29192922, 2007. [120] K. Y. Hui and K. M. Luk, A miniature dielectric resonator loaded patch antenna, IEEE Trans. Antennas Propag., vol. 53, pp. 21182122, 2005.

18

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[121] K. Y. Hui and K. M. Luk, Bandwidth enhancement of small dielectric resonator loaded patch antenna, IEEE Trans. Antennas Propag., vol. 54, pp. 18821885, 2006. [122] M. T. Lee, K. M. Luk, K. W. Leung, and M. K. Leung, A small dielectric resonator antenna, IEEE Trans. Antennas Propag., vol. 50, pp. 14851487, 2002. [123] A. Rashidian and D. M. Klymyshyn, On the two segmented and high aspect ratio rectangular dielectric resonator antennas for bandwidth enhancement and miniaturization, IEEE Trans. Antennas Propag., vol. 57, pp. 27752780, 2009. [124] A. Buerkle and K. Sarabandi, A wide-band, circularly polarized, magnetodielectric resonator antenna, IEEE Trans. Antennas Propag., vol. 53, pp. 34363442, 2005. [125] G.-M. Yang, X. Xing, A. Daigle, M. Liu, O. Obi, S. Stoute, K. Naishadham, and N. X. Sun, Tunable miniaturized patch antennas with self-biased multilayer magnetic lms, IEEE Trans. Antennas Propag., vol. 57, pp. 21902193, 2009. [126] G.-M. Yang, X. Xing, A. Daigle, O. O. , M. Liu, J. Lou, S. Stoute, K. Naishadham, and N. X. Sun, Planar annular ring antennas with multilayer self-biased nico-ferrite lms loading, IEEE Trans. Antennas Propag., vol. 58, pp. 648655, 2010. [127] E. E. Altshuler, Electrically small self-resonant wire antennas optimized using a genetic algorithm, IEEE Trans. Antennas Propag., vol. 50, pp. 297300, 2002. [128] E. E. Altshuler, A method for matching an antenna having a small radiation resistance to a 50-ohm coaxial line, IEEE Trans. Antennas Propag., vol. 53, pp. 30863089, 2005. [129] S. R. Best, The radiation properties of electrically small folded spherical helix antennas, IEEE Trans. Antennas Propag., vol. 52, pp. 953960, 2004. [130] S. R. Best, Low Q electrically small linear and elliptical polarized spherical dipole antennas, IEEE Trans. Antennas Propag., vol. 53, pp. 10471053, 2005. [131] H. Choo, R. L. Rogers, and L. Hao, Design of electrically small wire antennas using a pareto genetic algorithm, IEEE Trans. Antennas Propag., vol. 53, pp. 10381046, 2005. [132] J. A. Dobbins and R. L. Rogers, Folded conical helix antenna, IEEE Trans. Antennas Propag., vol. 49, pp. 17771781, 2001. [133] P. Hallbjorner, Electrically small unbalanced four-arm wire antenna, IEEE Trans. Antennas Propag., vol. 52, pp. 14241428, 2004. [134] O. S. Kim, Low-Q electrically small spherical magnetic dipole antennas, IEEE Trans. Antennas Propag., vol. 58, pp. 22102217, 2010. [135] A. Krall, J. McCorkle, J. Scarzello, and A. Syeles, The omni microstrip antenna: A new small antenna, IEEE Trans. Antennas Propag., vol. 27, pp. 850853, 1979. [136] R.-L. Li and V. F. Fusco, Circularly polarized twisted loop antenna, IEEE Trans. Antennas Propag., vol. 50, pp. 13771381, 2002. [137] S. Lim, R. L. Rogers, and L. Hao, A tunable electrically small antenna for ground wave transmission, IEEE Trans. Antennas Propag., vol. 54, pp. 417421, 2006. [138] C. P. Baliarda, J. Romeu, and A. Cardama, The Koch monopole: A small fractal antenna, IEEE Trans. Antennas Propag., vol. 48, pp. 17731781, 2000. [139] S. R. Best, On the performance properties of the Koch fractal and other bent wire monopoles, IEEE Trans. Antennas Propag., vol. 51, pp. 12921300, 2003. [140] M. Naghshvarian-Jahromi, Novel wideband planar fractal monopole antenna, IEEE Trans. Antennas Propag., vol. 56, pp. 38443849, 2008. [141] M. F. Pantoja, F. G. Ruiz, A. R. Bretones, S. G. Garcia, R. G. Martin, J. M. G. Arbesu, J. Romeu, J. M. Rius, P. L. Werner, and D. H. Werner, GA design of small thin-wire antennas: Comparison with Sierpinskytype prefractal antennas, IEEE Trans. Antennas Propag., vol. 54, pp. 18791882, 2006. [142] Z.-G. Dong, H. Liu, J.-X. Cao, T. Li, S.-M. Wang, S.-N. Zhu, and X. Zhang, Enhanced sensing performance by the plasmonic analog of electromagnetically induced transparency in active metamaterials, Appl. Phys. Lett., vol. 97, p. 114101, 2010. [143] T. Kokkinos and A. P. Feresidis, Low-prole folded monopoles with embedded planar metamaterial phase-shifting lines, IEEE Trans. Antennas Propag., vol. 57, pp. 29973008, 2009. [144] C.-J. Lee, L. K. M. K. H. , and T. Itoh, Composite right/left-handed transmission line based compact resonant antennas for RF module integration, IEEE Trans. Antennas Propag., vol. 54, pp. 22832291, 2006. [145] M. Palandoken, A. Grede, and H. Henke, Broadband microstrip antenna with left-handed metamaterials, IEEE Trans. Antennas Propag., vol. 57, pp. 331338, 2009.

[146] J.-H. Park, Y.-H. Ryu, and J.-H. Lee, Mu-zero resonance antenna, IEEE Trans. Antennas Propag., vol. 58, pp. 18651875, 2010. [147] S. Yarga, K. Sertel, and J. L. Volakis, A directive resonator antenna using degenerate band edge crystals, IEEE Trans. Antennas Propag., vol. 57, pp. 799803, Mar. 2009. [148] R. DeVore and P. Bohley, The electrically small magnetically loaded multiturn loop antenna, IEEE Trans. Antennas Propag., vol. 25, pp. 496505, 1977. [149] T. J. Warnagiris and T. J. Minardo, Performance of a meandered line as an electrically small transmitting antenna, IEEE Trans. Antennas Propag., vol. 46, pp. 17971801, 1998. [150] G. Goubau, Multi-element monopole antennas, in Proc. Workshop on Electrically Small Antennas, ECOM, Ft. Monmouth, NJ, 1976, pp. 6367.

Daniel F. Sievenpiper (S94M99SM04F09) received the B.S. and Ph.D. degrees in electrical engineering from the University of California, Los Angeles, in 1994 and 1999, respectively. Since 2010, he has been a Professor at the University of California, San Diego, where his research focuses on antennas and electromagnetic structures. Previously, he was the Director of the Applied Electromagnetics Laboratory, HRL Laboratories in Malibu, CA, where his research included articial impedance surfaces, conformal antennas, tunable and wearable antennas, and beam steering methods. He has more than 60 issued patents and more than 50 technical publications. Prof. Sievenpiper was named a Fellow of the IEEE in 2009. He was awarded the URSI Issac Koga Gold Medal in 2008, and in 2010, he was elected to the Antennas and Propagation Society Administrative Committee.

David C. Dawson (M03) received the B.S. and M.S. degrees in electrical engineering from the San Diego State University, San Diego, CA, in 2003 and 2007, respectively. He served in the US Navy from 1989 to 1997. Since 2003, he has been an Engineer at the Applied Electromagnetics Branch, Space and Naval Warfare Systems Center (SPAWAR), in San Diego, CA, where his work focuses on antennas, electromagnetic structures, and EMC. He has one patent awarded and several others pending. His interests include electrically small antennas.

Minu M. Jacob (S11) received the B.Tech. degree in electronics and communication engineering from Amrita School of Engineering, Coimbatore, India, in 2009. She is currently working toward the M.S. degree at the University of California, San Diego. Her research interests include electrically small antennas, non-Foster circuits and RF/analog communication circuit design.

Tumay Kanar (S10) received the B.S. degree in electrical engineering from Cornell University, NY, in 2010. He is currently working toward the Ph.D. degree at the University of California San Diego. Mr. Kanar is the recipient of UCSD Powell Foundation Fellowship for 20102013.

SIEVENPIPER et al.: EXPERIMENTAL VALIDATION OF PERFORMANCE LIMITS AND DESIGN GUIDELINES FOR SMALL ANTENNAS

19

Sanghoon Kim received the M.S. and B.S. degrees in physics from the Konkuk University, Seoul, Korea, in 2008 and 2005, respectively, and is currently working toward Ph.D. degree at the University of California at San Diego. His recent research activities have focused on an attenuation of materials and active nonlinear circuits for self-tuning metamaterial surfaces.

Ryan G. Quarfoth (S11) received the B.S. degree in engineering (with high distinction) from Harvey Mudd College, Claremont, CA, in 2009. He is currently working toward the Ph.D. degree at the University of California, San Diego (UCSD). In 2010, he joined the Applied Electromagnetics Group at UCSD where he is studying anisotropic articial impedance surfaces as a graduate student researcher. Mr. Quarfoth is a member of Tau Beta Pi and Sigma Xi.

Jiang Long received the B.S. and M.S. degrees from Zhejiang University, Hangzhou, China, in 2007 and 2010, respectively. He is currently working toward the Ph.D. degree at the University of California, San Diego. His current research area is non-Foster loaded articial impedance surface metamaterials.

20

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Substrate Integrated Waveguide (SIW) Leaky-Wave Antenna With Transverse Slots

Juhua Liu, David R. Jackson, Fellow, IEEE, and Yunliang Long, Senior Member, IEEE
AbstractA novel slotted substrate integrated waveguide (SIW) leaky-wave antenna is proposed. This antenna works in the TE mode of the SIW. Leakage is obtained by introducing a periodic set of transverse slots on the top of the SIW, which interrupt the current ow on the top wall. It is seen that three modes (a leaky mode, a proper waveguide mode, and a surface-wave-like mode) can all propagate on this structure. The wavenumbers of the modes are calculated theoretically and are numerically evaluated by HFSS simulation. The leakage loss, dielectric loss, and conductor loss are also analyzed. A uniform slotted SIW leaky-wave antenna is designed that has good beam scanning from near broadside (though not exactly at broadside) to forward endre. This type of SIW leaky-wave antenna has a wide impedance bandwidth and a narrow beam that scans with frequency. Measured results are consistent with the simulation and the theoretical analysis. Index TermsLeaky wave, leaky-wave antenna, periodic structure, SIW, substrate integrated waveguide, surface wave.

Fig. 1. Geometry of the slotted SIW leaky-wave antenna. The SIW has mm, mm, mm, and mm. The slots mm, mm, and mm. have

I. INTRODUCTION HE TRANSVERSE slotted rectangular waveguide [1] is a simple structure that works as a leaky-wave antenna having frequency beam-scanning capability, with an orthogonal polarization from the conventional traveling-wave slotted antenna [2]. Because of the polarization, the transverse slotted rectangular waveguide can scan from near broadside to endre if the waveguide is lled with a dielectric material [3]. However, rectangular waveguide is costly, heavy, and bulky. Substrate integrated waveguide (SIW) has been recently investigated [4][9] for its signicant advantages such as low cost, low loss, and easy integration with planar circuits. It consists of a wide microstrip line that is shorted at the edges with conductive vias, acting as a rectangular waveguide. The SIW has been widely developed for integrated microwave and millimeter-wave components and antennas. Several types of SIW leaky-wave antennas have been proposed recently [10][16]. The SIW leaky-wave antennas published in [10] and [11] utilize leakage due to the periodic gaps
Manuscript received August 16, 2010; revised March 17, 2011; accepted April 04, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported by the Natural Science Foundation of China (61172026), the Research Program of Guangzhou (2010Y1-C401), and NSFC-Guangdong (U0935002). J. Liu and Y. Long are with the Department of Electronics and Communication Engineering, Sun Yat-sen University, Guangzhou 510275, China (e-mail: liujuhua_2000@hotmail.com; isslyl@mail.sysu.edu.cn). D. R. Jackson is with the Department of Electrical and Computer Engineering, University of Houston, Houston, TX 77204-4005 USA (e-mail: djackson@uh.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167910

between the vias with sufciently large period. This leaky-wave antenna radiates from the side-walls of the SIW. The half-mode SIW leaky-wave antenna discussed in [12] is similar to the halfwidth Menzel leaky-wave antenna in microstrip line [17], which has leakage from the open edges of the microstrip line. An SIW leaky-wave antenna with a long slot was investigated in [13], and low sidelobes were obtained. In these cases [10][13], it is difcult to achieve endre radiation due to the polarization of the radiating equivalent magnetic currents. The composite right/left-handed SIW leaky-wave antennas proposed in [14] and [15] have beam scanning from the backward to the forward region. An SIW leaky-wave antenna with a series of inclined slots [16] can produce different polarizations depending on the input port. The half-mode SIW transverse slot array antenna in [18] is not used as a leaky-wave antenna and radiates mainly at broadside. The SIW slot array antenna in [19] has a wide bandwidth, a broad beamwidth, and a high gain, but this antenna is also not a leaky-wave antenna and does not have a scanning ability. In this paper, we propose a novel leaky-wave antenna based on SIW. The proposed SIW leaky-wave antenna, shown in Fig. 1, not only maintains the advantages of transverse slotted rectangular waveguide, but also has the advantages of SIW including low cost and ease of fabrication. This antenna has a simple structure and is easy to feed with a microstrip line. This antenna works in the TE mode of the SIW. Leakage is obtained by introducing a periodic set of centered transverse slots on the top of the SIW, which interrupt the current ow on the top wall. This antenna scans in the forward quadrant as frequency changes. Because of the polarization, it can radiate well at endre when an innite (or large) ground plane is present. (However, it is difcult to obtain a beam radiating at exactly endre if a nite ground plane is used due to nite ground diffraction effects.) Since the transverse slotted SIW antenna works similarly as a transverse slotted rectangular waveguide, we analyze this

0018-926X/$26.00 2011 IEEE

LIU et al.: SIW LEAKY-WAVE ANTENNA WITH TRANSVERSE SLOTS

21

structure based on the theory of a transverse slotted rectangular waveguide [3]. A leaky (improper) waveguide mode, a proper waveguide mode, and a surface-wave-like mode can all propagate in the transverse slotted SIW antenna just as in the transverse slotted rectangular waveguide. In a leaky-wave antenna, the wavenumber is the key to determine the radiation characteristics. We calculate the wavenumber theoretically and by using HFSS simulation. The leakage loss, dielectric loss, and conductor loss are also analyzed. The radiation efciency is then found theoretically and compared to HFSS simulation. A uniform slotted SIW leaky-wave antenna is designed that has beam scanning from near broadside (though not exactly at broadside) to forward endre (not exactly at endre for a nite ground plane). This type of SIW leaky-wave antenna has a wide impedance bandwidth and a narrow beam that scans with frequency. When the antenna radiates not too close to endre, it works as a leaky-wave antenna, and the radiation is mainly due to the leaky wave, which is fast wave. However, results show that when this antenna radiates near endre, it becomes a surface-wave antenna since the radiation is mostly produced by a surface wave, which is slow wave. A prototype is fabricated, and measured results are consistent with the simulation and the theoretical results. II. THEORY We show the geometry of the slotted SIW leaky-wave antenna in Fig. 1. Leakage is obtained by introducing a periodic set of transverse slots on the top of the SIW (between via centers), which interrupt the current ow on the top wall. The width of the SIW is . The dielectric constant of the substrate is , and the thickness is . The diameter of the vias is , and the distance between adjacent vias is . The length and width of the slots are and , and the period of the slots is . For the slotted SIW investigated here, the structure has a small period compared to the wavelength so that the antenna radiates from the dominant space harmonic. To ensure a pure polarization, the slot width is assumed small compared to the slot length . The SIW can be taken as a conventional dielectric-lled rectangular waveguide with an effective width [8]. The effective width can be approximately calculated from empirical formulas [5], [7], [8] and can also be accurately calculated from HFSS [6] and other numerical calculations [9]. For example, the effective width of the SIW shown in Fig. 1 is mm, which is calculated from the empirical formulas in [7]. Since SIW can be taken as a conventional dielectric-lled rectangular waveguide, the slotted SIW can be analyzed as a dielectric-lled rectangular waveguide with transverse slots. Fortunately, the dispersion relations and modes for the transverse slotted rectangular waveguide with dielectric lling have been well investigated in [3]. Reference [3] gives equations for analyzing transverse slotted rectangular waveguide with an arbitrary dielectric lling, while [1] gives equations only for transverse slotted rectangular waveguide with air lling. As assumed in [3], the tangential electric eld in the zeroth slot on the SIW is assumed to be (1)

By enforcing an aperture magnetic eld integral equation using a space-domain approach, the propagation wavenumber of the slotted rectangular waveguide can then be calculated from the following transcendental equation [3]:

(2) where

(3) and (4) (5) (6) In (6) is the wavenumber in the substrate. If the substrate is a lossy dielectric with a loss tangent tan is (7) Note that in (5) has two possible values due to the square root function. The choice corresponds to an improper wave and causes the wave to be exponentially increasing with radial distance from the -axis. The other choice corresponds to a proper wave and causes the wave to be exponentially decaying with radial distance. For a conventional rectangular waveguide with transverse slots, three types of modal solutions are found: a leaky improper waveguide mode, a bound (nonleaky) proper waveguide mode, and a surface-wave type of mode [3]. The leaky waveguide mode is an attenuated mode that typically dominates above the cutoff frequency of the TE mode in the closed rectangular waveguide. The proper waveguide mode is a bound mode that has no attenuation for a lossless structure (where the conductor and the dielectric are lossless). Both the leaky waveguide mode and proper waveguide mode have most of their power stored inside the waveguide and represent a perturbation of the closed waveguide (the former dominates in the fast-wave region, the latter in the slow-wave region). The surface-wave mode is also a proper bound wave, but this mode has most of its power stored at the surface of the waveguide and not inside the waveguide, so

22

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Slotted SIW with rectangular waveguide feed. The SIW has the same dimensions as in Fig. 1. The width of the rectangular waveguide is mm. The slots at both ends are linearly tapered. The length mm, mm.

this mode does not represent a perturbation of the closed rectangular waveguide. For the improper leaky waveguide mode, is chosen so that the imaginary part of is positive for and negative for all other . For the proper waveguide mode or the surface-wave mode solution, is chosen so that the imaginary part of is negative for all . A more complete explanation for the modal analysis is given in [3]. III. METHOD TO EXTRACT WAVENUMBER FROM HFSS In order to validate the wavenumber calculated theoretically from (2), we can extract the wavenumber by simulating a slotted SIW with HFSS. Here, we rst simulate the structure shown in Fig. 2 and sample the near-eld electric eld along the uniform slotted SIW at a periodic distance (the period of the slots). The tapered ends help to minimize unwanted end reections, resulting in a more ideal eld in the uniform region. To simplify the simulation, the slotted SIW is fed by a rectangular waveguide with a width that is the effective width of the SIW. Suppose there are two traveling waves with wavenumbers and propagating inside the SIW, so that the eld can be modeled using (8) We use MATLAB to numerically t these two traveling waves to the sampled near-eld data obtained from HFSS, using a least-square criterion for determining the best t. For a bound-mode solution, we assume the wavenumber to be a real number. The procedure is illustrated in Fig. 3 at a frequency of 11.7 GHz, where we assume that there is a leaky mode with an unknown complex wavenumber and a surface-wave mode with a real wavenumber propagating inside the waveguide. Fig. 3 shows that the sum of the leaky mode and the surface-wave mode ts the near-eld data from HFSS very well. At this frequency, the amplitude of the physical leaky mode is much stronger than the amplitude of the surface-wave mode, though the surface-wave mode causes a noticeable oscillation in the eld plot. Since the surface-wave mode has most of the elds outside (but near the surface) of the SIW, it is more accurate to sample the near eld outside the SIW for determining the wavenumber of the surface-wave mode. For the leaky mode or proper waveguide mode, it is better to sample the electric eld inside the SIW because these modes have most of the elds inside the SIW. Once the wavenumbers are known, the amplitudes of the modes in (8) can be found by sampling along a given line (the amplitudes of the modes depend on the location of the line). Since two modes (leaky mode and surface-wave mode) can simultaneously propagate along the slotted SIW when the antenna has

Fig. 3. Electric eld inside the uniform slotted SIW. The antenna has the structure in Fig. 2. The slotted SIW has the same dimensions as in Fig. 1. The freGHz. The substrate and the conductor are lossless. quency is

not yet scanned to endre, in order to extract a more accurate result from HFSS, the length of the antenna shown in Fig. 2 should be relatively large. Here, we select the length of as 180 mm (about 7 wavelengths in free space). For the leaky waveguide mode or the proper waveguide mode, a value that is close to the theoretical wavenumber from (2), or even the wavenumber of the closed equivalent rectangular waveguide, can be used as the initial guess for determining the nal wavenumber in the least-squares solution. For the surface-wave mode, a value close to the wavenumber in free space can be used as the initial guess. If we t two traveling waves using expression (8) to the electric eld at the slots obtained from HFSS simulation, we can also estimate the percentage of the leaky mode that is present in the total near eld by using % (9)

where we suppose the leaky mode is mode 1 and the surfacewave mode is mode 2, so represents the amplitude of the leaky mode and represents the amplitude of the surfacewave mode. The tangential electric eld is sampled at the center of the slots in this calculation since this is the eld that is responsible for radiation. IV. NUMERICAL RESULTS A. Dispersion Relations We calculate the wavenumber for the equivalent rectangular waveguide ( mm) with transverse slots from (2) and show the normalized phase and attenuation constants in Figs. 4 and 5. From Fig. 4, we nd that when the frequency is less than 11.98 GHz, the phase constant of the leaky mode is less than , so the leaky mode is a fast wave and the solution is in the physical region. In this same frequency region, there is another solution that is proper with a zero attenuation constant. This proper mode is called a surface-wave type of mode because most of its power is outside the waveguide but close to

LIU et al.: SIW LEAKY-WAVE ANTENNA WITH TRANSVERSE SLOTS

23

Fig. 4. Normalized phase constant in slotted SIW. The slotted SIW has the dimensions given in Fig. 1. The substrate and the conductors are lossless.

Fig. 6. Percentage of the leaky mode present in the slotted SIW. The dimensions are the same as in Fig. 4.

Fig. 5. Normalized attenuation constant in slotted SIW. The slotted SIW has , and the dimensions shown in Fig. 1. The lossy dielectric has S/m. the lossy conductor has

the surface [3]. When the frequency is larger than 11.98 GHz, the phase constant of the leaky mode is larger than , so the leaky mode is a slow wave and it will lose physical signicance. Beyond 11.98 GHz, a proper waveguide mode now also exists instead of a surface-wave mode, which has most of its power stored inside the waveguide. We calculate the wavenumber of the modes in the slotted SIW from HFSS. By numerically tting two traveling waves to the near-eld electric eld obtained from the HFSS simulation, we can extract the wavenumber of leaky wave and the other proper wave (either the surface-wave mode or the proper waveguide mode) [3]. We show the normalized phase and attenuation constants of the leaky mode in slotted SIW from HFSS simulation in Figs. 4 and 5 and compare with theoretical results for the slotted rectangular waveguide [from (2)]. From the phase constant shown in Fig. 4, we nd that the simulated results for the slotted SIW

are very close to the theoretical results for the rectangular waveguide with transverse slots. In Fig. 5, the attenuation constant of the leaky mode in slotted SIW is very close to the theoretical results for slotted rectangular waveguide, so the leakage from the vias is very small compared to the leakage from the slots in the slotted SIW, and the slotted SIW can be taken as a slotted rectangular waveguide. By numerically tting two traveling waves to the near-eld electric eld on the slots obtained from the HFSS simulation, we can also estimate the percentage of the leaky mode that is present in the slotted SIW (the procedure is discussed in Section III). The percentage of the leaky mode is shown in Fig. 6. From Fig. 6, we nd that at low frequency (when the antenna has not yet scanned to endre), the leaky mode is the dominant mode. However, the percentage of the leaky mode decreases as the frequency increases because the surface-wave mode is excited more as the frequency increases. Also, even more importantly, as the frequency approaches toward the leaky-mode cutoff (11.98 GHz), the leaky mode loses physical signicance rapidly, so the percentage of the leaky mode becomes very small. B. Dielectric Loss When loss is present, all modes have a complex wavenumber. For calculating the dielectric loss in slotted rectangular waveguide, we use the wavenumber from (7) in (2) to solve for the complex wavenumber of the proper and improper modes in the lossy slotted rectangular waveguide. This yields the attenuation constant that accounts for leakage and dielectric loss. We then calculate the attenuation constant due to only leakage loss by using (2) with a real . By subtracting from , we obtain the attenuation constant due to dielectric loss. For the leaky-mode solution, the attenuation constant due to dielectric loss in slotted rectangular waveguide is included in Fig. 5. We also calculate the attenuation constant for the leaky mode in slotted SIW from HFSS simulation and show the results in Fig. 5. In the HFSS simulation, we follow a similar

24

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 8. Geometry of uniform slotted SIW leaky-wave antenna. The uniform portion of the slotted SIW has the same dimensions as in Fig. 1. The dielectric , and the conductor has a conductivity of has a tangent loss of S/m. The antenna is fed with a tapered microstrip line with a length of 15 mm and a width linearly varying from 4.7 to 3.1 mm. The length and width of the ground plane are 310.2 and 40 mm, respectively. (a) Geometry (not to scale). (b) Measured antenna. Fig. 7. Dielectric loss effect on the proper mode in slotted SIW. The slotted SIW has the dimensions shown in Fig. 1. The lossy dielectric has .

mode. The upper cutoff frequency of the leaky waveguide mode ( GHz) lies within this region. C. Conductor Loss To calculate the conductor loss of slotted SIW, we rst use HFSS to simulate a slotted SIW with a lossy conductor and calculate the attenuation constant that includes leakage and conductor loss. Second, we simulate a slotted SIW that has perfect conductors to determine that only includes leakage. Then, we subtract to get the attenuation constant that is due to conductor loss. We include the conductor loss of slotted SIW obtained from HFSS in Fig. 5. Since it is difcult to calculate theoretically the conductor loss in slotted rectangular waveguide or in slotted SIW, we only compare simulated results for slotted SIW with the conductor loss of a closed rectangular waveguide, calculated from [20]

procedure discussed above, in which we rst simulate a structure that uses a lossy dielectric substrate to obtain , and then subtract that is obtained from the simulation of a lossless structure. Fig. 5 shows that the theoretical dielectric loss of rectangular waveguide with transverse slots agrees very well with the dielectric loss of slotted SIW from HFSS. As shown in Fig. 7, the attenuation constant of the proper waveguide mode is not zero because of the dielectric loss. We show the dielectric loss in the slotted rectangular waveguide and compare to the dielectric loss in both the slotted SIW and the closed rectangular waveguide. Also included is the dielectric loss of the surface-wave mode, obtained theoretically. (The dielectric loss of the surface-wave mode is not easy to obtain from HFSS because the amplitude of this mode is very small in the region where the leaky mode is the dominant mode.) The dielectric loss of the surface-wave mode is very small because most of the power carried in this mode is outside the waveguide and not in the lossy substrate. When the surface-wave mode transitions to become the proper waveguide mode, the normalized attenuation constant is not close to zero any more, but becomes a little larger than the attenuation constant of the closed rectangular waveguide. Fig. 7 also shows that the results for the slotted SIW from HFSS are very close to the theoretical results for the slotted rectangular waveguide. Fig. 7 indicates that the surface-wave mode is evolving into the proper waveguide mode in the region (shown shaded in Fig. 7) where the frequency is between 11.82 and 12.28 GHz. At the same time that the surface-wave mode evolves into the proper waveguide mode, the leaky waveguide mode loses its physical signicance, and the waveguide mode is replaced by the proper waveguide mode [3] as the physical waveguide-like mode (though the improper leaky mode continues to exist). This corresponds to the region in Fig. 6 in which the percentage of leaky mode decreases sharply. The transition region from 11.82 to 12.28 GHz thus includes three modes: the leaky waveguide mode, the surface-wave mode, and the proper waveguide

(10) This comparison shows that the conductor loss of slotted SIW is larger than the conductor loss of a corresponding closed rectangular waveguide. This is due to the fact that the current must bend around the slots, increasing the current density surrounding the slots, resulting in increased conductor loss. In fact, the conductor loss of slotted rectangular waveguide (not shown) is also larger than the conductor loss of closed rectangular waveguide for this same reason. V. UNIFORM SLOTTED SIW LEAKY-WAVE ANTENNA A. Geometry The geometry of the uniform SIW leaky-wave antenna is shown in Fig. 8. The dominant part of this leaky-wave antenna is a uniform SIW with a series of uniform transverse slots, and this part has a length of 220 mm.

LIU et al.: SIW LEAKY-WAVE ANTENNA WITH TRANSVERSE SLOTS

25

Fig. 9. -parameters of the uniform slotted SIW leaky-wave antenna shown in Fig. 8. The antenna is with nite ground plane.

Fig. 10. Radiation efciency for the slotted SIW antenna shown in Fig. 8. The antenna is with nite ground plane.

Since the slots cause a perturbation to the TE mode of the SIW [3], to suppress the reection coefcient , we taper the slots linearly at both ends as shown in Fig. 8. The length of the tapered part is 22.5 mm. We taper the slot length from zero to 4.55 mm and keep the other parameters of the slots constant. To enhance the bandwidth, we taper the SIW width as a cosine function at both ends, as in [12]. The length of the tapered SIW section is 30 mm. We taper the width of the SIW from 10.5 to 12.9 mm and keep the other parameters of the SIW constant. We use microstrip line to feed the SIW leaky-wave antenna. In order to obtain a better match between the input SIW and the microstrip line, we also linearly taper the microstrip line using the method described in [21]. B. S-Parameters We use HFSS to simulate the structure shown in Fig. 8. The reection coefcients are shown in Fig. 9 and compared to the measured data. The measured results for are a little higher than the simulated results because the fabrication is not accurate enough and the loss from the SMA connectors is not included in the simulation. However, both results show that is mostly less than 10 dB in the band of interest (from 10.2 to 12 GHz). The measured data for are a little lower than the simulated results, which might result from the increased and probably the increased conductor loss in the measurement. A small shift of frequency is also observed between the simulated results and the measured data, probably due to a difference in the dielectric constant used in the experiment and some small errors in fabrication. Due to such errors, the propagation wavenumber in the fabricated antenna has a small shift compared to the simulated antenna. Nevertheless, the simulated results are close to the measured results. Both the measured and simulated results show that when the frequency is a little larger than 10.2 GHz, is relatively high because the attenuation constant is small and most of the power is transmitted to the output end. However, as the frequency increases, will decrease because the attenuation constant of the leaky mode increases, which leads to more power being radiated into space. Measured results show that is lower than 10 dB in the band from 11.15 to 11.9 GHz, which

indicates that most of the power is consumed (radiated and dissipated) by the antenna. However, as the frequency increases still further beyond the endre point, increases dramatically since the leaky mode is now becoming nonphysical and the proper waveguide mode is becoming the dominant mode, which has very little attenuation (due only to conductor and dielectric loss). C. Radiation Efciency We estimate the radiation efciency waveguide from of slotted rectangular (11) and the dielectric loss where the sum of the leakage loss can be calculated from (2) and the conductor loss can be calculated from (10). We show the radiation efciency of slotted rectangular waveguide calculated this way in Fig. 10. In Fig. 10, we also show the radiation efciency of the slotted SIW antenna obtained directly from HFSS by simulating the structure shown in Fig. 8. Note that the radiation efciency obtained from HFSS is not the total efciency because the HFSS radiation efciency calculated here does not take the mismatch or the power absorbed in the load port into account. This is done to agree with the denition used in (11). At a frequency that is close to the cutoff frequency of the SIW (10 GHz), the theoretical radiation efciency is lower than the simulated one from HFSS because there is another radiation mechanism due to the connection between the microstrip line and the SIW. When the frequency increases and the frequency is lower than the upper cutoff frequency of the leaky mode, the theoretical results from (11) are a little higher than the radiation efciency calculated from HFSS because the conductor loss of closed rectangular waveguide calculated from (10) is smaller than the conductor loss of the slotted SIW, and the material loss (conductor loss and dielectric loss) of the feeding structure is not included in the theoretical calculation. Nevertheless, the theoretical results are close to the radiation efciency from HFSS. When the frequency is larger than the upper cutoff frequency of the leaky mode, the leaky

26

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 11. Radiation patterns for the structure shown in Fig. 8 at 10.2 GHz.

Fig. 12. Radiation patterns for the structure shown in Fig. 8 at 10.8 GHz.

mode loses physical signicance, and the radiation efciency of the slotted SIW becomes lower than the theoretical results from (11). D. Radiation Patterns We show the normalized radiation patterns in Figs. 1114. It is seen that the measured results are very close to the simulated patterns for the antenna with a nite ground plane, although a small discrepancy for the radiation angle due to a shift of frequency is observed. The main beam of the antenna can scan from near broadside to near endre as frequency increases. When the antenna is radiating near broadside the radiation efciency is rather poor due to the smaller leakage attenuation and increased attenuation due to material loss. However, as the antenna scans in the forward quadrant, the radiation efciency increases. In the theoretical calculation, we use the theoretical wavenumber from (2) and calculate the normalized radiation pattern by (12) where is the length of the th slot in Fig. 8. In the theoretical calculation for the wavenumber , the leaky loss and the dielectric loss are accounted for, but the conductor loss is not included because it is difcult to calculate the conductor loss of slotted SIW theoretically. In the theoretical calculation, when the antenna has not yet scanned to endre, we use the theoretical wavenumber for the leaky mode to calculate the pattern from (12). Figs. 1113 show that the theoretical results that assume a leaky mode are very close to the main beam of the simulated results from HFSS, which means this antenna works as a leaky-wave antenna when the antenna has not yet scanned to endre. When the antenna has not yet scanned to endre, the radiation is mainly produced by the leaky mode and the phase constant of the leaky mode is less than , so the antenna works as a fast-wave (leaky-wave) radiator. In Fig. 12, the simulated pattern for the antenna with innite ground plane has a sidelobe at endre that is due to the

Fig. 13. Radiation patterns for the structure shown in Fig. 8 at 11.5 GHz.

surface-wave radiation. This sidelobe is not observed in the theoretical calculation because the surface-wave mode is not included in the theoretical calculation. The surface-wave mode has a wavenumber only slightly larger than the wavenumber of free space, so the surface-wave mode can only radiate at endre and contaminates the leaky-wave radiation pattern. For the antenna with a nite ground plane, the sidelobe due to the surface-wave mode is radiating at near endre because of nite ground plane diffraction effects. In the measured patterns, the sidelobe due to the surface wave is also observed near endre. The surface-wave mode is excited stronger as frequency increases. Hence, at low frequency, the sidelobe due to the surface-wave mode is not pronounced, as shown in Fig. 11. As frequency increases, the leaky mode scans very close to endre, and the main beam of the leaky mode becomes very close to the main beam of the surface-wave mode. Therefore, the main beam of the leaky mode becomes merged into that of the surface-wave mode, as shown in Fig. 13. Note that in Fig. 13, the radiation pattern calculated from theory only accounts for the leaky mode, so no merging for the leaky-mode pattern into a surface-wave pattern is observed. In Fig. 14, we show the theoretical radiation pattern that assumes a leaky mode and compare to the simulated results and

LIU et al.: SIW LEAKY-WAVE ANTENNA WITH TRANSVERSE SLOTS

27

Fig. 14. Radiation patterns for the structure shown in Fig. 8 at 11.7 GHz.

Fig. 15. Simulated and theoretically calculated radiation patterns for the antenna shown in Fig. 8 at 12 GHz.

measured results at a frequency of 11.7 GHz. It appears the theoretical leaky-wave pattern agrees well with the measured results. However, since a shift of frequency (about 0.15-GHz shift at a frequency of 11.7 GHz) exists between the measured result and the theoretical one, the theoretical frequency corresponding to the measured antenna at 11.7 GHz should be chosen to be about 11.85 GHz. As discussed before in connection with Figs. 6 and 7, the percentage of the leaky mode is not very large and the surface-wave mode is transitioning into the proper waveguide mode at 11.85 GHz, so the measured pattern should actually be composed of a leaky mode and a mode that is somewhat between the surface-wave mode and the proper waveguide mode. The measured antenna appears to suffer more from nite ground plane diffraction effects than does the antenna simulated in HFSS. Therefore, owing to the serious nite ground plane diffraction effects, the measured main beam is not so close to endre and by coincidence agrees well with the simulated antenna even though a small shift of frequency exists. Because of the nite ground plane diffraction effects, the practical antenna cannot scan to exactly endre. In Fig. 15, we show the simulated results for the antenna working as a slow-wave (surface-wave) antenna at 12 GHz. Also included is the radiation pattern calculated theoretically by (12). Note that in the theoretical calculation by (12), the wavenumber should be the one for the surface-wave or proper waveguide mode (the two are essentially coincident at this frequency) because at this frequency the leaky mode is not the dominant mode any more as shown in Fig. 6. Fig. 15 shows that the theoretical results agree very well with the simulated results for the antenna with an innite ground plane. In fact, the theoretical result that assumes a leaky mode (not shown) is not so close to the simulated results. Since the proper mode (surface-wave mode or proper waveguide mode) is a slow-wave mode, the antenna works as a slow-wave antenna at this frequency. Fig. 15 shows that the simulated radiation pattern for the antenna with a nite ground plane is not radiating exactly at endre due to nite ground plane diffraction. At endre, the pattern for the nite ground plane case is lower than the same pattern at the beam peak by about 6 dB, a result that is consistent with the geometrical theory of

Fig. 16. Directivity and gain of the uniform slotted SIW leaky-wave antenna shown in Fig. 8.

diffraction (GTD). (GTD predicts a 6-dB drop at endre for a nite ground plane, regardless of the ground plane size.) Using a larger ground plane should help to minimize the nite ground plane diffraction effects. Here we do not show the measured results because the gain of the measured radiation pattern is rather poor due to the shift of frequency and the more serious nite ground diffraction effects. E. Gain and Directivity Fig. 16 shows the measured gain for the practical antenna and the simulated gains for the antennas with a nite ground plane and with an innite ground plane. The shape of the measured result is very similar to the simulated ones (realized gains), although the measured gain is lower than the simulated gains. When the frequency is larger than the cutoff frequency of the SIW (10.2 GHz), the gain increases as the frequency increases because the leakage constant (attenuation constant due to radiation) increases as the frequency increases, and hence more power is radiated into space instead of being absorbed at the output termination. As the frequency increases even more, the measured gain and the simulated gain for the antenna with a

28

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

nite ground plane decrease a little bit, probably due to an increase of the beamwidth. When the frequency increases still further, the gains drop sharply because the main beam passes endre and more power is transmitted to the terminating end instead of being radiated into space. As the frequency approaches cutoff (about 10 GHz) the attenuation due to material loss increases dramatically (see Fig. 5). While material loss does not directly affect the directivity (only the gain), the increased material loss results in an increased overall attenuation constant for the leaky mode, and this in turn lowers the size of the effective radiating aperture, which lowers the directivity. The gain is lowered even more due to the decreased radiation efciency associated with the increased material loss. There are several reasons conjectured for the measured gain being lower than the simulated gain. First, the loss from the SMA connector that is not included in the simulation and some errors in fabrication result in a higher , which also decreases the measured gain. Second, the conductor loss and dielectric loss in the fabricated antenna are probably higher than that in the simulated one (especially the conductor loss). In the fabricated antenna, the vias are replaced by tinned pins, which might lead to an increased conductor loss. This also results in a lower . Third, the chamber for measuring the antenna is not ideally large enough for the length of the antenna measured here, which would lead to some inaccuracy for the gain measurement. As a matter of fact, a similar discrepancy between the simulated results and measured results for another SIW leaky-wave antenna is also observed in [15], and the reasons discussed above were also mentioned in [15]. In our case, another discrepancy is also observed. The measured antenna reaches the highest gain at about 11.4 GHz, while the simulated antenna with a nite ground plane reaches the highest gain at about 11.9 GHz. This is because a small shift of frequency exists due to fabrication, and the measured antenna also suffers a more serious nite ground plane diffraction effect than the simulated one. Comparing the simulated gains for the antenna with a nite ground plane and the antenna with an innite ground plane, it is also found that the nite ground plane diffraction effects would cause the gain of the antenna with a nite ground plane to drop at a lower frequency compared to the antenna with an innite ground plane. The simulated directivity and gain for the antenna with an innite ground plane is also shown in Fig. 16 as a function of frequency. Fig. 16 shows that when the frequency is larger than 10.2 GHz, the directivity decreases as a function of frequency because the beamwidth increases as the scan angle goes from broadside to endre (see [3]). However, when the frequency goes to about 11.7 GHz and the beam approaches endre, the radiation pattern will change from a conical beam to a pencil beam, so the directivity will increase. The beam reaches endre at about 11.9 GHz though the directivity continues to increase until about 12.2 GHz since the beam continues to narrow for a while as the beam scans beyond endre where the phase constant is a little larger than . At 12.2 GHz, the antenna works as a slow-wave antenna since . According to the HansenWoodyard maximum directivity condition [22], [23], the largest directivity will occur when for and , where is the antenna length. Although the HansenWoodyard condition assumes (a

uniform aperture illumination), this condition is approximately valid for narrow-beam leaky-wave antennas [24]. From HFSS simulation, we nd that at 12.2 GHz the phase constant of the uniform slotted SIW is , which is very close to the HansenWoodyard maximum directivity condition , corresponding to mm. Note that the directivity does not take into account the power lost at the terminating load port, and this explains why the shape of the gain and directivity curves are a bit different, and the fact that the gain curves peaks at a lower frequency than do the directivity curves. The simulated directivity and gain for the antenna with a nite ground plane are also shown in Fig. 16. When the antenna has not yet scanned to endre, the simulated directivity and gain for the nite ground plane antenna are very close to those for the antenna with an innite ground plane. However, when the antenna scans to endre, the simulated directivity and gain for the antenna with the nite ground plane start to drop at lower frequencies compared to the antenna with an innite ground plane (due to the nite ground plane diffraction effects). After the antenna reaches endre, the gain begins to drop sharply because the elds are evolving into the proper waveguide mode, and more power is transmitted into the terminating end instead of being radiated into space. VI. CONCLUSION A novel slotted SIW leaky-wave antenna is proposed that has the advantages of both rectangular waveguide with transverse slots and SIW. This structure supports a leaky mode, a proper waveguide mode, and a surface-wave mode, depending on the frequency region. The wavenumber of these waves was calculated theoretically and also evaluated by HFSS simulation. The leakage loss, dielectric loss, and conductor loss were evaluated by theory and from HFSS simulation. One of the advantages of this type of leaky-wave antenna is that it can scan to endre. When the beam is scanned not too close to endre, the pattern is due to the dominant leaky mode, as expected. However, as the beam is scanned to endre, the structure acts as a slow-wave type of antenna, with the radiation coming from a combination of the proper waveguide mode and the surface-wave mode. Beyond endre, the radiation mainly comes from the proper waveguide mode. Excellent agreement is obtained among the theoretical results, simulated results, and measured data. ACKNOWLEDGMENT The authors wish to acknowledge the China Scholarship Council (CSC) for awarding J. Liu a scholarship to study in the U.S. The authors also wish to acknowledge the following persons: V. R. Komanduri from the University of Houston, Houston, TX, for assistance in the theoretical research; Y. Li and X. Tang from Sun Yat-sen University, Guangzhou, China, for assistance in the experiment and measurement; and Y. Wu from Guangdong University of Technology, Guangzhou, China, for assistance with the measurements. The authors also wish to thank the reviewers for useful suggestions that have helped to improve this paper.

LIU et al.: SIW LEAKY-WAVE ANTENNA WITH TRANSVERSE SLOTS

29

REFERENCES
[1] R. F. Hyneman, Closely-spaced transverse slots in rectangular waveguide, IRE Trans. Antennas Propag., vol. AP-7, no. 4, pp. 335342, Oct. 1959. [2] V. H. Rumsey, Traveling wave slot antennas, J. Appl. Phys., vol. 24, pp. 13581365, Nov. 1953. [3] J. Liu, D. R. Jackson, and Y. Long, Modal analysis of dielectric-lled rectangular waveguide with transverse slots, IEEE Trans. Antennas Propag., vol. 59, no. 9, pp. 31943203, Sep. 2011. [4] D. Deslandes and K. Wu, Design consideration and performance analysis of substrate integrated waveguide components, in Proc. 32nd Eur. Microw. Conf., Milan, Italy, Sep. 2002, vol. 2, pp. 881884. [5] Y. Cassivi, L. Perregrini, P. Arcioni, M. Bressan, K. Wu, and G. Conciauro, Dispersion characteristics of substrate integrated rectangular waveguide, IEEE Microw. Wireless Compon. Lett., vol. 12, no. 9, pp. 333335, Sep. 2002. [6] F. Xu and K. Wu, Numerical multimode calibration technique for extraction of complex propagation constants of substrate integrated waveguide, in Proc. IEEE MTT-S Int. Symp., June 2004, vol. 2, pp. 12291232. [7] L. Yan, W. Hong, K. Wu, and T. J. Cui, Investigations on the propagation characteristics of the substrate integrated waveguide based on the method of lines, Proc. Inst. Elect. Eng., Microw. Antennas Propag., vol. 152, pp. 3542, Feb. 2005. [8] F. Xu and K. Wu, Guided-wave and leakage characteristics of substrate integrated waveguide, IEEE Trans. Microw. Theory Tech., vol. 53, no. 1, pp. 6673, Jan. 2005. [9] D. Deslandes and K. Wu, Accurate modeling, wave mechanisms, and design considerations of a substrate integrated waveguide, IEEE Trans. Microw. Theory Tech., vol. 54, no. 6, pp. 25162526, Jun. 2006. [10] D. Deslandes and K. Wu, Substrate integrated waveguide leaky-wave antenna: Concept and design considerations, in Proc. AsiaPacic Microw. Conf., Dec. 2005, pp. 346349. [11] F. Xu, K. Wu, and X. Zhang, Periodic leaky-wave antenna for millimeter wave applications based on substrate integrated waveguide, IEEE Trans. Antennas Propag., vol. 58, no. 2, pp. 340347, Feb. 2010. [12] J. Xu, W. Hong, H. Tang, Z. Kuai, and K. Wu, Half-mode substrate integrated waveguide (HMSIW) leaky-wave antenna for millimeterwave applications, IEEE Antenna Wireless Propag. Lett., vol. 7, pp. 8588, 2008. [13] Y. J. Cheng, W. Hong, K. Wu, and Y. Fan, Millimeter-wave substrate integrated waveguide long slot leaky-wave antennas and two-dimensional multibeam applications, IEEE Trans. Antennas Propag., vol. 59, no. 1, pp. 4047, Jan. 2011. [14] Y. Dong and T. Itoh, Composite right/left-handed substrate integrated waveguide leaky-wave antennas, in Proc. Eur. Microw. Conf., Rome, Italy, Sep. 2009, pp. 276279. [15] Y. Dong and T. Itoh, Composite right/left-handed substrate integrated waveguide and half mode substrate integrated waveguide leaky-wave structures, IEEE Trans. Antennas Propag., vol. 59, no. 3, pp. 767775, Mar. 2011. [16] Y. Cheng, W. Hong, and K. Wu, Millimeter-wave half mode substrate integrated waveguide frequency scanning antenna with quadri-polarization, IEEE Trans. Antennas Propag., vol. 58, no. 6, pp. 18481855, Jun. 2010. [17] G. M. Zelinski, G. A. Thiele, M. L. Hastriter, M. J. Havrilla, and A. J. Terzuoli, Half width leaky wave antennas, Microw. Antennas Propag., vol. 1, no. 2, pp. 341348, Apr. 2007. [18] Q. Lai, W. Hong, Z. Kuai, Y. Zhang, and K. Wu, Half-mode substrate integrated waveguide transverse slot array antennas, IEEE Trans. Antennas Propag., vol. 57, no. 4, pp. 10641072, Apr. 2009. [19] M. Ohira, A. Miura, and M. Ueba, 60-GHz wideband substrate-integrated-waveguide slot array using closely spaced elements for planar multisector antenna, IEEE Trans. Antennas Propag., vol. 58, no. 3, pp. 993998, Mar. 2009. [20] D. M. Pozar, Microwave Engineering, 2nd ed. New York: Wiley, 1998, p. 125. [21] D. Deslandes and K. Wu, Integrated microstrip and rectangular waveguide in planar form, IEEE Microw. Wireless Compon. Lett., vol. 11, no. 2, pp. 6870, Feb. 2001. [22] W. L. Stutzman and G. A. Thiele, Antenna Theory and Design, 2nd ed. New York: Wiley, 1998. [23] A. Sutinjo, M. Okoniewski, and R. H. Johnson, Radiation from fast and slow traveling waves, IEEE Antennas Propag. Mag., vol. 50, no. 4, pp. 175181, Aug. 2008.

[24] E. M. OConnor, D. R. Jackson, and S. A. Long, Extension of the HansenWoodyard condition for endre leaky-wave antennas, IEEE Antennas Wireless Propag. Lett., vol. 9, pp. 12021204, 2010.

Juhua Liu was born in Heyuan, China, in 1981. He received the B.S. and Ph.D. degrees in electrical engineering from Sun Yat-sen University, Guangzhou, China, in 2004 and 2011, respectively. From 2008 to 2009, he was a Visiting Scholar with the Department of Electrical and Computer Engineering, University of Houston, Houston, TX. His present research interests include microstrip antennas, substrate integrated waveguide antennas, leaky-wave antennas, periodic structures, and computational electromagnetics.

David R. Jackson (S83M84SM95F99) was born in St. Louis, MO, in 1957. He received the B.S.E.E. and M.S.E.E. degrees from the University of Missouri, Columbia, in 1979 and 1981, respectively, and the Ph.D. degree in electrical engineering from the University of California, Los Angeles, in 1985. From 1985 to 1991, he was an Assistant Professor with the Department of Electrical and Computer Engineering, University of Houston, Houston, TX. From 1991 to 1998, he was an Associate Professor in the same department, and since 1998 he has been a Professor with the department. His present research interests include microstrip antennas and circuits, leaky-wave antennas, leakage and radiation effects in microwave integrated circuits, periodic structures, and electromagnetic compatibility and interference. Prof. Jackson is presently serving as the Chair of the Distinguished Lecturer Committee of the IEEE AP-S Society and is a Member-at-Large for U.S. Commission B of URSI (the International Union of Radio Science). He also serves as the chair of the MTT-15 (Microwave Field Theory) Technical Committee. He is on the Editorial Board for the IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES. Previously, for the IEEE AP-S Society, he has been the Chair of the Transnational Committee, the Chapter Activities Coordinator, a Distinguished Lecturer, and a member of the AdCom. He was an Associate Editor for the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION. He also previously served as the Chair of U.S. Commission B of URSI. He has also served as an Associate Editor for Radio Science and the International Journal of RF and Microwave Computer-Aided Engineering.

Yunliang Long (M01SM02) was born in Chongqing, China, in 1963. He received the B.Sc., M.Eng., and Ph.D. degrees in electrical engineering from the University of Electronic Science and Technology of China (UESTC), Chengdu, China, in 1983, 1989, and 1992, respectively. From 1992 to 1994, he was a Postdoctoral Research Fellow, then employed as an Associate Professor, with the Department of Electronics, Sun Yat-sen University, Guangzhou, China. From 1998 to 1999, he was a Visiting Scholar with the IHF, RWTH University of Aachen, Aachen, Germany. From 2000 to 2001, he was a Research Fellow with the Department of Electronics Engineering, City University of Hong Kong, Hong Kong. Currently, he is a Professor and the Head of the Department of Electronics and Communication Engineering, Sun Yat-sen University. He has authored and coauthored over 130 academic papers. His research interests include antennas and propagation theory, EM theory in inhomogeneous lossy medium, computational electromagnetics, and wireless communication applications. Prof. Long is a member of the Committee of Microwave Society of CIE and is on the Editorial Board of the Chinese Journal of Radio Science. He is Vice Chairman of the Guangzhou Electronic Industrial Association. His name is listed in Whos Who in the World.

30

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Subwavelength Substrate-Integrated Fabry-Prot Cavity Antennas Using Articial Magnetic Conductor

Yong Sun, Zhi Ning Chen, Fellow, IEEE, Yewen Zhang, Hong Chen, and Terence S. P. See
AbstractThis paper presents the high-gain low-prole subwavelength substrate-integrated Fabry-Prot (FP) cavity antennas with articial magnetic conductor (AMC) sheets. A partially reective planar AMC sheet and a ground plane are used as the two reectors of the FP-type resonant cavity for an ultra-thin planar design. The cavity is fully lled with dielectric substrate for further reduction of thickness of the antenna and easy integration. A microstrip patch antenna is embedded into the cavity as a feed. As design examples, the antennas are designed to operate at 10 GHz with a xed overall thickness of (where is the . operating wavelength in free space) and an aperture of The losses caused by both dielectric and conductors are analyzed, which are critical for a fully dielectric substrate antenna design. The via-walls surrounding the radiating aperture are introduced to improve radiation patterns and gain by suppressing the surface waves, which are another critical loss for a thin fully dielectric substrate antenna design. Measured results show that such dielectric-integrated subwavelength cavity antennas feature the high gain of 12.5 dBi, low prole, easy integration into circuit board, and mechanical robustness, which makes them suitable for low-cost mass production. Index TermsFabry-Prot cavity, high gain antenna, substrate integrated.

In addition, periodic metallic structures, i.e. metallic patch arrays or their complements, which show partially reection property, have been utilized as superstrate layers in order to significantly enhance the directivity of primary radiating sources at boresight [5][7]. The structure is based on the formation of a resonant Fabry-Prot (FP) cavity between two metallic arrays which act as partially reective surfaces (PRS), or between one periodic metallic array and a ground plane. The basic radiating element, such as monopole, dipole, microstrip patch, and waveguide apertures can be positioned outside or inside the cavity as a feed [8]. Using a ray optics model, the FP cavity antennas were studied. The FP cavity antennas achieved a maximum directivity at the boresight direction at a given operating wavelength ( is the operating wavelength in free space when the cavity is fully air-lled) as follows [5][7]: (1) is the complex reection coefcient of PRS. The where thickness of the FP cavity, is determined by (2) where is the reection phase of the ground plane. Usually, of a highly reective PRS, namely, a the reection phase . Therefore, periodic metallic patch or mesh array is about . the minimum thickness of the cavity is about Recently, a two-layer metamaterial structure i.e. AMC has been applied as a ground plane and/or PRS to lower down the prole of FP cavity antennas [9][11]. With one layer AMC, the dispersive reection phase of AMC is around 0 , which leads to [9]. Next, the anthe thickness reduction of the cavity to tennas with much lower prole were proposed when the two reectors of the cavity both are with the AMC structures [11]. Furthermore, various electronically recongurable periodic structures have been introduced to realize beam-steering and electrically scanned antennas [12][14]. In this paper, the subwavelength fully substrate-integrated FP cavity antennas using AMC-PRS are proposed and designed at 10 GHz. The AMC-PRS reector consists of a patch array layer and a mesh array layer with electrically small spacing between the layers [11]. A feeding patch operating at 10 GHz is embedded into the FP cavity between the AMC-PRS and a ground plane. The FP cavity is fully lled with dielectric substrate for small thickness, easy integration with other circuits and mechanical robustness. The vital issues such as the realized gain, loss analysis, as well as truncation effects due to the nite lateral dimensions are taken into account. The prototypes with

I. INTRODUCTION

VER the last few years periodic metallic structures have attracted increasing interest in microwave and antenna engineering due to their unique properties such as electromagnetic band gap (EBG) and articial magnetic conductor (AMC) behaviors [1]. The presence of EBG similar to those observed from dielectric photonic crystals is able to suppress surface waves and enhance the performance of printed antennas and circuits, whereas the AMC surfaces can reect incident waves with a reection phase from to [1][4].

Manuscript received April 29, 2010; revised April 07, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the Agency for Science, Technology and Research (A*STAR), Singapore, Metamaterials Program under Grant # 0921540097, the Program for Key Basic Research of the Shanghai Science and Technology Committee under Grant # 08dj1400301, and in part by the National Basic Research Program of China under Grant # 2006CB921701. Y. Sun, Y. Zhang, and H. Chen are with the Department of Physics, Tongji University, Shanghai 200092, China. Z. N. Chen and T. S. P. See are with the Institute for Infocomm Research, Agency of Science, Technology, and Research, Singapore 138632, Singapore (e-mail: chenzn@i2r.a-star.edu.sg; spsee@i2r.a-star.edu.sg). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167902

0018-926X/$26.00 2011 IEEE

SUN et al.: SUBWAVELENGTH SUBSTRATE-INTEGRATED FABRY-PROT CAVITY ANTENNAS

31

Fig. 1. Geometry of the proposed antenna: (a) cross-section view, (b) 3D, , and (c) unit cell of the AMC-PRS with lled substrate, where .

Fig. 2. Simulated complex reection coefcient with varying geometrical parameters: (a) phase for varying , (b) amplitude for varying , (c) phase for varying , (d) amplitude for varying .

an aperture of and thickness of are designed using a full-wave electromagnetic (EM) simulation tool (CST Microwave Studio) and validated experimentally. II. FABRY-PROT CAVITY ANTENNAS A. Antenna Structure and Design Procedure Fig. 1 shows the geometry of the proposed FP cavity antenna operating at 10 GHz. perfectly electrically conducting (PEC) meshes and square patches are patterned onto the top and bottom surfaces of a 0.813-mm thick Rogers RO4003 substrate ( , ) to form an AMC-PRS layer as shown in Figs. 1(a) and (b). Fig. 1(c) zooms in the unit cell of the AMC-PRS. The reection phase of AMC-PRS, which is a critical parameter to form an FP cavity, is controlled by the width of the mesh strip and the patch as well as the lattice period of . The is set and xed in this design. The overall thickness of the structure is 3.35 mm or when the spacing is 2.54 mm. With the lled dielectric, the electrical thickness causes the phase variation is 72 instead of 180 which is required by a conventional air-lled FP cavity antenna. A 7.2 6.4-mm rectangular microstrip feeding patch is designed and embedded in the cavity at the center of the plane, which is fed by a 50- microstrip line of a width of 1.6 mm through a SubMiniature version A (SMA) connector, as shown in Fig. 1(b). The spacing between the feeding patch and ground plane is 0.813 mm. The antenna is designed as follows. First, the reection characteristic of the AMC-PRS is simulated by changing the size of a unit cell as shown in Fig. 1(c). The magnetic and electric wall boundary conditions are enforced along the and directions to form a waveguide, respectively. Next, the Port 1 of the waveguide and its reference plane are added at the bottom of the patch, and Port 2 of the waveguide is positioned at a distance away from the top mesh surface. Hence, the excitation of Port

1 is equivalent to an incident plane wave from the direction and having a linear polarization along the -direction. The FP cavity mode is excited at the desired frequency such that the calculated phase of is with the ground plane at the bottom. The effects of and on the complex reection coefcient are examined as shown in Fig. 2. It is found that is mainly affected by while the resonant frequency by . It should be noted that the unit cell of the AMC-PRS is square-shaped, and has a 90-degree rotational symmetry. Thus, the properties of the AMC-PRS are the same for the -polarization. Furthermore, such a FP cavity is also able to work for circular polarization at the same frequency. Last, the FP cavity antenna is simulated with the feeding patch. B. Effect of Reectivity on Antenna Gain The relationship (1) is valid for an ideal case with innite dimensions, point source excitation, and without any losses. In practical design, the radiation property of the feeding structure, nite-size PRS and ground plane and the losses caused by dielectric, conductors and surface waves should be taken into account using a full-wave EM simulator (CST Microwave Studio). The antennas are congured based on three sets of parameters, namely i) , , ii) , , and iii) , with resonance at 9.81, 9.86, and 9.88 GHz but with different amplitudes of the reection coefcients as shown in Fig. 3(a) and (b). The antennas are with a 160 mm 160 mm ground plane in the plane at and a 110 mm 110 mm AMC-PRS ( periods along both the and axes). The simulated , directivity and realized gain at the boresight are shown in Fig. 3(c) and (d). The simulated results showed that the higher the reectivity of AMC-PRS, the higher the boresight directivity of the FP cavity antenna. In order to explore this phenomena, the eld distributions, which is normalized by the two-dimensional (2-D) maximum at are plotted in Fig. 3(e). It is found that the effective aperture of the antenna increases as the reectivity

32

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 4. Directivity and realized gain with the AMC-PRS formed by cells.

unit

TABLE I COMPARISON OF GAIN

Dielectric loss:

Fig. 3. (a) Simulated reection phase: (b) amplitude of reection, (c) , eld distributions in the (d) directivity and realized gain at boresight, (e) plane at , normalized by 2-D maximum for the three types of AMC-PRS congurations.

In general, the boresight directivity of the FP antenna should be the product of (1) and the boresight directivity of the primary source [5]. D. Loss Analysis The overall thickness of the FP cavity antennas with the AMC-PRS is 3.35 mm or . For such low-prole antennas, any ohmic losses should be taken into account. Table I compares the directivity and gain of antenna of 60.5 mm 60.5 mm with and without losses. The radiation efciency excluding the impedance mismatch can be extracted by calculating the difference between and . The simulated radiation efciency is 2.1 dB (62%). It should be noted that the radiation efciency can be improved by reducing the reectivity of the AMC-PRS but the directivity is lowered. The radiation efciency is 1.8 dB (66%) for the lossless conductor. Without the dielectric loss, the radiation efciency reaches 0.5 dB (89%). III. EXPERIMENTAL RESULTS The subwavelength substrate-integrated FP cavity antennas were fabricated. An 11 11 copper mesh array and a square patch array are, respectively etched onto the top and bottom surfaces of a 0.813-mm thick Rogers RO4003 substrate, where the size of the AMC-PRS is 60.5 mm 60.5 mm. The three RO4003 substrates are assembled together with two 0.1-mm thick dielectric glue layers which have similar electrical properties to those of RO4003. The overall volume of the antenna is 80.5 mm 80.5 mm 3.35 mm. The predicted resonant frequency of the cavity mode is 9.81 GHz. Fig. 5(a) shows a photo of the antenna prototype. The and the realized gain are shown in Fig. 5(b). The resonant frequency shift of 1% between experiment and simulation is probably due to the fabrication tolerance. The measured maximum

increases. However, the 3-dB realized gain bandwidth and the radiation efciency decrease for an increased reectivity. This suggests that the realized gain of the FP cavity antenna is limited by the inevitable loss. Thus, the reectivity of the reector should be optimized for the maximum realized gain instead of the directivity. C. Effect of Physical Size on Antenna Gain The FP cavity antenna is also known as a special kind of leaky wave antenna but with a nite size. The antenna gain performance against size is examined by the simulation. The overall size of AMC-PRS with , varies from 60.5 mm 60.5 mm to 93.5 mm 93.5 mm , where the antenna is positioned at the center of the ground plane and with a 20-mm wide dielectric rim. The directivity and realized gain at the boresight are shown in Fig. 4. The predicted resonant frequency of the cavity mode is 9.81 GHz but the maximum directivity appears at 9.90 GHz for the AMC-PRS of , 9.86 GHz for , 9.84 GHz for , and 9.82 GHz for . The maximum directivity and gain are achieved at 9.84 GHz for . It should be noted that the maximum directivity of 17 dBi at 9.84 GHz has exceeded the predicted of 16.1 dBi using (1) because this prediction does not take into account the contribution of the directive source, namely the feeding patch to the achieved gain.

SUN et al.: SUBWAVELENGTH SUBSTRATE-INTEGRATED FABRY-PROT CAVITY ANTENNAS

33

Fig. 5. (a) Photo of antenna, (b)

and realized gain.

Fig. 7. (a) Photo of antenna with via-walls, (b)

and realized gain.

Fig. 8. Radiation patterns of antenna with via walls in (a) H-plane, (b) E-plane.

Fig. 6. Radiation patterns in (a) H-plane, (b) E-plane.

realized gain is about 12.5 dBi at 9.95 GHz, while the simulated gain is 12.1 dBi at 9.92 GHz. The radiation patterns of the antenna were measured in an anechoic chamber using an Orbit MiDAS far-eld measurement system and Agilent 8510C vector network analyzer. The normalized radiation patterns in the H-plane and E-plane are shown in Fig. 6. It is found that the measured maximum realized gain is not at the boresight. There is about a beam squint in the measured radiation patterns in the E-plane as observed in the simulations, and the beam squinting varies when the frequency deviates from 9.92 GHz. For example, it is at 9.84 GHz.

The beam squint is caused by the existence of the asymmetrical feeding structure, which includes the microstrip line and the SMA connector. Moreover, the sidelobe level is about 7 dB in the H-plane. In order to suppress the sidelobe levels, via-walls are introduced along the sides of the antenna to suppress the unwanted surface waves at the edges of the cavity. Fig. 7(a) and (b) show the photo of the antenna with the via-walls as well as the and realized gain. The measured and simulated maximum realized gain is about 12.2 dBi at 9.98 GHz and 12.3 dBi at 9.94 GHz, respectively. The radiation patterns in the H- and E-planes are shown in Fig. 8. A beam squint of is observed in the measured E-plane radiation patterns. However, the introduction of the via-walls reduces the sidelobe level by 5 dB in measurement and 7.5 dB in simulation.

34

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

IV. CONCLUSION The subwavelength substrate-integrated high-gain FabryPerot cavity antennas have been presented and studied. Using the conventional printed circuit board (PCB) technique, the antennas have been fully integrated into a grounded dielectric substrate for thin and robust design as well as easy fabrication and integration into other circuits. By applying the articial magnetic conductor as a partially reective surface (AMC-PRS), although the prole of the proposed antenna is kept at only 3.35 mm , the realized gain of 12.5 dBi has been achieved in the X-band (10 GHz). The effects of the reectivity and size of the AMC-PRS on the antenna gain have been studied. The study has shown that the reectivity can be optimized for the maximum gain, considering the nite lateral dimensions and the losses. The radiation efciency of the antennas has been mainly affected by the dielectric loss. Compared to conventional air-lled FP cavity antennas, the proposed FP cavity antennas have achieved the desired performance with a much lower prole, stronger mechanical robustness and easier integration, which makes them suitable for cost-effective mass production. REFERENCES
[1] D. Sievenpiper, L. Zhang, R. F. J. Broas, N. G. Alexopolous, and E. Yablonovitch, High-impedance electromagnetic surfaces with a forbidden frequency band, IEEE Trans. Microw. Theory Tech., vol. 47, no. 11, pp. 20592074, Nov. 1999. [2] R. Gonzalo, P. de Maagt, and M. Sorolla, Enhanced patch antenna performance by suppressing surface waves using photonic bandgap substrates, IEEE Trans. Microw. Theory Tech., vol. 47, no. 11, pp. 21312138, Nov. 1999. [3] R. Abhari and G. V. Eleftheriades, Metallo-dielectric electromagnetic bandgap structures for suppression and isolation of the parallel-plate noise in high-speed circuits, IEEE Trans. Microw. Theory Tech., vol. 51, no. 6, pp. 16291639, Jun. 2003. [4] G. Goussettis, A. P. Feresidis, and J. C. Vardaxoglou, Tailoring the AMC and EBG characteristics of periodic metallic arrays printed on grounded dielectric substrate, IEEE Trans. Antennas Propag., vol. 54, no. 1, pp. 8289, Jan. 2006. [5] G. V. Trentini, Partially reecting sheet arrays, IRE Trans. Antennas Propag., vol. AP-4, pp. 666671, 1956. [6] J. R. James, S. J. A. Kinany, P. D. Peel, and G. Andrasic, Leaky-wave multiple dichroic beamformers, Electron. Lett., vol. 25, 1989. [7] A. P. Feresidis and J. C. Vardaxoglou, High gain planar antenna using optimized partially reective surfaces, Proc. Inst. Elect. Eng. Microw. Antennas Propag., vol. 148, no. 6, pp. 345350, Dec. 2001. [8] T. Akalin, J. Danglot, O. Vanbesien, and D. Lippens, A highly directive dipole antenna embedded in a Fabry-Perot-type cavity, IEEE Microw. Wireless Compon Lett., vol. 12, no. 2, pp. 4850, Feb. 2002. [9] A. P. Feresidis, G. Goussetis, S. Wang, and J. C. Vardaxoglou, Articial magnetic conductor surfaces and their application to low-prole high gain planar antennas, IEEE Trans. Antennas Propag., vol. 53, no. 1, pp. 209215, Jan. 2005. [10] L. Zhou, H. Li, Y. Qin, Z. Wei, and C. T. Chan, Directive emissions from subwavelength metamaterial-based cavities, Appl. Phys. Lett., vol. 86, pp. 101101-1101101-3, Feb. 2005. [11] A. Ourir, A. de Lustrac, and J. M. Lourtioz, All-metamaterial-based for ultrathin directive antennas, Appl. subwavelength cavities Phys. Lett., vol. 88, pp. 084103-1084103-3, Feb. 2006. [12] A. Ourir, S. N. Burokur, and A. de Lustrac, Phase-varying metamaterial for compact steerable directive antennas, Electron. Lett., vol. 43, no. 9, pp. 493494, Apr. 2007. [13] A. Ourir, S. N. Burokur, and A. de Lustrac, Electronically recongurable metamaterial for compact directive cavity antennas, Electron. Lett., vol. 43, no. 13, pp. 698700, June 2007. [14] A. Ourir, S. N. Burokur, and A. de Lustrac, Electronic beam steering of an active metamaterial-based directive subwavelength cavity, presented at the 2nd Eur. Conf. Antennas Propag., Nov. 1116, 2007.

Yong Sun was born in Zhejiang, China in 1982. He received the B.Sc. degree in applied physics from Shanghai Tongji University, China, in 2005, and is currently working toward the Ph.D. degree in the Department of Physics, Tongji University, Shanghai, China. During the second half of 2009, he was an exchange student with the Institute for Infocomm Research, Agency of Science, Technology, and Research, Singapore. His research interests include photonic bandgap materials, metamaterials and their applications.

Zhi Ning Chen (M99SM05F07) received the B.Eng., M.Eng., and Ph.D. degrees from the Institute of Communications Engineering, China, and his second Ph.D. degree from the University of Tsukuba, Japan, in 1985, 1988, 1993, and 2003, all in electrical engineering. From 1988 to 1997, he worked at Institute of Communications Engineering, Southeast University, and City University of Hong Kong, China with teaching and research appointments. In 1997, he was awarded the JSPS Fellowship to join in University of Tsukuba, Japan. In 2001 and 2004, he visited University of Tsukuba under the JSPS Fellowship Program (senior level). In 2004, he worked at IBM T. J. Watson Research Center, USA as Academic Visitor. Since 1999, he has worked with Institute for Infocomm Research, Singapore and his current appointments are Principal Scientist and Department Head for RF & Optical. He is concurrently holding Adjunct/Guest Professorships at Southeast University, Nanjing University, Shanghai Jiao Tong University, Tongji University, University of Science and Technology, China, Dalian Maritime University, and National University of Singapore. He has published 310 journal and conference papers as well as authored and edited the books entitled Broadband Planar Antennas, UWB Wireless Communication, Antennas for Portable Devices, and Antennas for Base Stations in Wireless Communications. He also contributed chapters to the books entitled UWB Antennas and Propagation for Communications, Radar, and Imaging as well as Antenna Engineering Handbook. He is holding 28 granted and led patents with 21 licensed deals with industry. His current research interest includes applied electromagnetic engineering, RF transmission over bio-channels, and antennas for wireless systems, in particular at mmW, submmW, and THz for medical and healthcare applications. Dr .Chen is a Fellow of the IEEE for his contribution to small and broadband antennas for wireless applications. He is serving as an IEEE Antennas and Propagation Society Distinguished Lecturer and Associate Editor of IEEE TRANSACTION ON ANTENNAS AND PROPAGATION. He has organized many international technical events as key organizer. As General Chairs, he established International Workshop on Antenna Technology (iWAT) and International Symposium of Info-Comm and Media Technology in Bio-Medical and Healthcare Applications (3T-in-3A). He is the recipient of the Best Paper Award of International Symposium of Antennas and Propagation 2010, the CST University Publication Award 2008, IEEE AP-S Honorable Mention Student Paper ConQuartest 2008, IES Prestigious Engineering Achievement Award 2006, terly Best Paper Award 2004, and IEEE iWAT 2005 Best Poster Award.

Yewen Zhang was born in Zhejiang, China, in 1955. He received the B.Eng., M.Sc., and Ph.D. degrees in electrical engineering from Xian Jiaotong University, Shaanxi, China, in 1982, 1984, and 1989, respectively. In 1989, he joined Ecole Suprieur de Physique et de Chimie Industrielles de la Ville de Paris (ESPCI) as a Postdoctoral Research Fellow and become a Research Fellow subsequently. In 1996, he moved to the Department of Applied Physics, Hong Kong Polytechnic University, as a Research Fellow. In 1997, he joined the Department of Physics, Tongji University, as a Full Professor. His research interests lie in wide range of dielectric phenomena since 1984. Recently, he has paid attention to photonic crystals and metamaterials for microwave.

SUN et al.: SUBWAVELENGTH SUBSTRATE-INTEGRATED FABRY-PROT CAVITY ANTENNAS

35

Hong Chen received the B.Sc. degree in physics from Shanghai Fudan University, Shanghai, China, in 1982, and the Ph.D. degree in condensed matter physics from Shanghai Jiaotong University, Shaanxi, China, in 1986. He has been with the Pohl Institute of Solid State Physics at Shanghai Tongji University since 1986, where he is currently a Full Professor. His current research interests include photonic bandgap materials and metamaterials.

Terence S. P. See received the B.Eng. and M.Eng. degrees in electrical engineering from the National University of Singapore in 2002 and 2004, respectively. In 2004, he joined the Institute for Infocomm Research, Singapore. He is currently holding the position of Senior Research Engineer in the Antenna Lab under the RF and Optical Department. His main research interests include antenna design and theory, particularly in small and broadband antennas and arrays, diversity antennas, antennas for portable devices, and antennas for on-body communications and bio-medical applications.

36

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

A Recongurable Wideband and Multiband Antenna Using Dual-Patch Elements for Compact Wireless Devices
Hattan F. Abutarboush, Member, IEEE, R. Nilavalan, Senior Member, IEEE, S. W. Cheung, Senior Member, IEEE, Karim M. Nasr, Senior Member, IEEE, Thomas Peter, Djuradj Budimir, Senior Member, IEEE, and Hamed Al-Raweshidy, Senior Member, IEEE

AbstractA recongurable wideband and multiband C-Slot patch antenna with dual-patch elements is proposed and studied. It occupies a compact volume of 50 50 1.57 (3925 mm3 ), including the ground plane. The antenna can operate in two dual-band modes and a wideband mode from 5 to 7 GHz. Two parallel C-Slots on the patch elements are employed to perturb the surface current paths for excitation of the dual-band and the wideband modes. Two switches, implemented using PIN diodes, are placed on the connecting lines of a simple feed network to the patch elements. Dual-band modes are achieved by switching ON either one of the two patch elements, while the wideband mode with an impedance bandwidth of 33.52% is obtained by switching ON both patch elements. The frequencies in the dual-band modes can be independently controlled using positions and dimensions of the C-Slots without affecting the wideband mode. The advantage of the proposed antenna is that two dual-band operations and one wideband operation can be achieved using the same dimensions. This overcomes the need for increasing the surface area normally incurred when designing wideband patch antennas. Simulation results are validated experimentally through prototypes. The measured radiation patterns and peak gains show stable responses and are in good agreements. Coupling between the two patch elements plays a major role for achieving the wide bandwidth and the effects of mutual coupling between the patch elements are also studied. Index TermsC-slot, cognitive radio, patch antenna, recongurable antenna, slot antenna, small antenna, switched antenna, wideband antenna.

I. INTRODUCTION

ATCH antennas suffer from narrow bandwidth which can limit their uses in some modern wireless applications [1], [2]; therefore, there is an increasing demand for low-prole,
Manuscript received January 10, 2011; revised April 15, 2011; accepted July 12, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The measurements at the NPL SMART chamber were supported by the Measurements for Innovators (MFI) program and the National Measurement Ofce, an Executive Agency of the Department for Business, Innovation and Skills. H. F. Abutarboush, R. Nilavalan, T. Peter, and H. Al-Raweshidy are with the Wireless Networks and Communications Centre (WNCC), School of Engineering and Design, Brunel University, West London UB8 3PH, U.K. (e-mail: hattan.abutarboush@ieee.org). S. W. Cheung is with the Department of Electrical and Electronics Engineering, Hong Kong University, Hong Kong, China. K. M. Nasr is with the National Physical Laboratory (NPL), Teddington TW11 0LW, U.K. D. Budimir is with the Wireless Communications Research Group, Department of Electronics and Computer Science, Westminster University, London W1W 6UW, U.K. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167925

easy to manufacture, and multiband/wideband antennas which can be easily integrated within communication systems. A variety of studies have come up with different techniques to achieve wideband operation for printed antennas. Some of the techniques employed are changing the physical size of the antenna, modifying the radiator shape to allow current paths to travel at longer distances (which sometime increases the antenna size), and adding additional parts such as multi layers or gaps (which again makes the antenna larger and of a higher prole). Other techniques include using U-slot array [3], shorting wall [4], folded shorting wall [5], Y-V Slot [6], slots form [7], stacked patch [8], pair of slits on the patch (with total size of the antenna 150 150 14.3 mm ) [9], E-shaped patch on thick substrates with ground plane size of 140 210 mm [10] and using circular arc shaped slot on thick substrate [11]. The designs in [3][11] can achieve wide impedance bandwidths. However, these antennas are large in sizes and difcult to t into small and slim devices. A recongurable antenna is another solution to achieve a wide impedance bandwidth by switching ON and OFF some parts of the antenna. To allow the operating frequencies and the bandwidths to be recongurable, switching components are normally used. PIN diodes, varactor diodes or MEMS switches are the most frequently used components in the design of recongurable antennas [12][15]. In literature, few papers reported the approach of switching between wideband and narrowband operations. For example in [16][18], studies were done on switching between dual ports, one port for Ultra Wideband (UWB) and the other port for a single narrowband. However, the use of an UWB antenna for multiband applications could result in unwanted emissions in the transmission mode. In this paper, a single-feed recongurable wideband and multiband antenna using two patch elements on a planar structure is proposed. Two C-Slots are employed on the patch elements for excitation of the dual-band and wideband modes. Two PIN diode switches are placed on the connecting lines of a simple feed network to the patch elements. Dual-band and wideband-mode operations are obtained by switching ON/OFF the two patch elements. The antenna can be used for narrowband applications such as the WLAN and WiMAX and wideband operations in the frequency range from 5 to 7 GHz for other wireless standards. The design eliminates the need for using two ports as proposed in [17], [18] and increases the number of possible frequency bands from one to four. The

0018-926X/$26.00 2011 IEEE

ABUTARBOUSH et al.: RECONFIGURABLE WIDEBAND AND MULTIBAND ANTENNA USING DUAL-PATCH ELEMENTS

37

Fig. 1. Conguration of proposed antenna with dc biasing networks. TABLE I DETAILED DIMENSIONS OF PROPOSED ANTENNA (UNITS IN mm)

line impedance of 63 . Two PIN diodes are placed on the connecting lines to the patch elements and used as switches. Just to prove our design concept, we have used the practical PIN diodes, SMP1320-079 from Skyworks Solutions Inc. with a size of 1.5 0.7 mm , as the switches. In computer simulation, these two diodes are modeled using the resistance, inductance, and as the capacitance (RLC) boundary sheet which gives 0.9 impedance value of the PIN diode in the ON state and 0.3 pF as the capacitance value in the OFF state. These PIN diodes are turned ON/OFF using a dc biased signal, so two coupling chip capacitors ( and ) each with 10 pF are used to prevent the dc signal from owing to the main feed line but allow the RF current to pass through. The biasing networks for the two PIN diodes are also shown in Fig. 1, where the inductors , , and , all with 12 nH, are used as radio-frequency (RF) chokes to provide high impedance for the RF signals. The resistors, , , and , each with 10 , are used to control the dc biasing current to (or dc biasing voltage of about 0.7 V across) the PIN diodes. These lumped components will have insignicant effects on the antenna performance because the impedances of the RL circuits are much higher than the impedance of the antenna, allowing very little currents to ow through. III. SIMULATION AND MEASUREMENTS RESULTS A. Impedance Bandwidth for dB

volume of the proposed antenna, including the ground plane and the substrate, is 50 mm 50 mm 1.57 mm. The design can also be used for cognitive radio applications as described in [19]. II. ANTENNA CONFIGURATION AND DESIGN PROCEDURE Fig. 1 shows the schematic diagram of our proposed recongurable antenna which consists of two patch elements (patch elements #1 and #2) with a simple feed network, two PIN-diode switches and two chip capacitors all on one side of the substrate and a ground plane on the other side of the substrate. The complete antenna is designed using the EM simulator, HFSS V.11.2, based on nite elements modeling (FEM) and fabricated on an FR-4 substrate with thickness of 1.57 mm and a relative permit. The key antenna parameters of the antenna tivity of are shown in Table I. The dimensions of the patch elements of the antenna are optimized to operate in the 5.5-GHz WLAN band. Since cutting a slot on the radiator can change the current distribution and the current path, and hence improve the impedance matching especially at higher frequencies, as discussed in [20] and [21], in our design, we use two C-Slots on the two patch elements, as shown in Fig. 1, to generate a wide impedance bandwidth and to create multiple resonant frequencies. The feed network has a main 50- feed line and two connecting lines which have been optimized, in terms of impedance bandwidth, to have a

The two PIN diodes provide three possible and useful switching states, i.e., ON-OFF, OFF-ON, and ON-ON states (note that the OFF-OFF state has no practical use). Simulation tests using the HFSS have been carried out on the impedance dB) of the bandwidth (for reection coefcient antenna in different states. In the OFF-ON state, only patch element #2 is ON and functioning. Simulation results in Fig. 2(a) show that a dual-band is obtained at 5.6 and 6.2 GHz, with the respective bandwidths of 5.2% and 4.85%. In the ON-OFF state, only patch element #1 of the antenna is ON and radiating. The results in Fig. 2(b) show that another dual-band mode is obtained at 5 and 5.7 GHz, with the corresponding impedance bandwidth of 4.2% and 2.4%. In the ON-ON state, both patch elements are radiating. A wide bandwidth of 33.52%, covering the frequency range from 4.99 to 7 GHz, is obtained as shown in Fig. 2(c). To validate the simulation results, the proposed in the ON-OFF, antenna has also been fabricated and the OFF-ON, and ON-ON states have been measured using Agilent N5230A vector network analyzer. Results are shown in Fig. 2(a)(c) for comparison. It can be seen that the simulated and measured results are in good agreements. The small discrepancies between the simulated and measured results could be attributed to the fabrication accuracy of the prototype. B. Effects of C-Slots Multiband operation of the antenna is achieved mainly by the C-Slots on the patch elements. Simulation tests have been carof the antenna without the C-Slots in the ried out to study the with and ON-OFF and OFF-ON states. Fig. 3 compares the without the C-Slots in the patch elements. In the ON-OFF state when patch element #1 is active and patch element #2 is OFF,

38

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Simulated and measured S

of proposed antenna in (a) OFF-ON, (b) ON-OFF, and (c) ON-ON states.

Fig. 3. Effects of C-Slots on S11 in (a) ON-OFF and (b) OFF-ON states.

Fig. 3(a) shows that, without the C-Slots, the antenna has a resonant frequency at 5.6 GHz. While with the C-Slots, a dual-band is generated at 5 and 5.7 GHz. In the OFF-ON state when patch element #2 is turned ON and patch element #2 is OFF, Fig. 3(b) shows that the antenna without the C-Slots has a single band at 5.55 GHz. While with the C-Slots on the patch elements, a dual-band is generated at 5.6 and 6.2 GHz. Therefore, the C-Slots on the patch elements help generate a dual band. Moreover, simulation results have also shown that the widths and the lengths of the patch elements determine the centre frequencies

in the single-band cases. While the positions and dimensions of the C-Slots on the patch elements determine the centre frequencies in the dual-band cases. The simulated-frequency bands generated with and without the C-Slots in our design are summarized in Table II. C. Current Distributions Cutting slots on the radiator of an antenna can change the current path and so can be used to generate dual-band or even multiple-bands operations. In our proposed design, if the slots

ABUTARBOUSH et al.: RECONFIGURABLE WIDEBAND AND MULTIBAND ANTENNA USING DUAL-PATCH ELEMENTS

39

TABLE II GENERATED BANDS WITH AND WITHOUT C-SLOTS

In the OFF-ON and ON-OFF states, the radiation patterns at the pair-resonant frequencies of 5.6 and 6.2 GHz, and 5 and 5.7 GHz, respectively, are used for comparison. In the ON-ON state where the antenna has an operation bandwidth from 5 to 7 GHz, the radiation patterns at the extreme frequencies of 5 and 7 GHz and the middle frequency of 6 GHz are studied. From the current distributions shown in Fig. 4, it can be seen that the directions of the dominant currents at the frequencies studied are mainly in the -direction, so the - and - planes are the - and -planes, respectively. As a result, the co-polarization patterns in Figs. 57 are all relatively unidirectional toward the -direction with small back radiation due to the nite ground-plane size. The high cross polarizations at 5.6 and 5.7 GHz in the OFF-ON and ON-OFF states are due to high current concentration in the -Directions. Some minor discrepancies occur between the simulated and measured results, which could be due to the effect of the coaxial cable connected to the antenna during measurements. The peak gains of the antenna at different frequencies are between 3 and 5 dBi in different switching states and summarized in Table III. The simulated radiation efciency of the antenna in the ON-ON state ranges from 60% to 70%. IV. INDEPENDENT CONTROL OF EACH BANDS

Fig. 4. Simulated current distributions in (a) OFF-ON and (b) ON-OFF states.

are absent, the antenna will have only one major current path on each of the patch elements. However, in the presence of the two C-slots, the current path on each of the patch elements is disturbed, hence creating the dual-band operation. Fig. 4 shows the simulated current distributions on the patch elements in different switching states. In the OFF-ON state, patch element #2 is radiating and patch element #1 is OFF. Fig. 4(a) shows that the current travels around the C-Slot on patch element #2, as expected, generating the resonant frequencies at 5.6 and 6.2 GHz as shown in Fig. 2(a) for the WLAN 802.11a/h/j/n applications. In the ON-OFF state, only patch element #1 is ON and radiating and patch element #2 is OFF. Fig. 4(b) shows that the current travels around the C-Slot on patch element #1, generating the dual band at 5 and 5.7 GHz as shown in Fig. 2(b), for the WLAN 802.11a/h/j/n applications. The dominate current paths for OFF-ON and ON-OFF states are shown in Fig. 4. , where is the These paths correspond to approximately wavelength at the resonant frequency of the respective band and , with being the free space given by wavelength. The 6.2-GHz band in the OFF-ON case is generated from a higher order mode. In the ON-ON state where both patch elements are ON, the wideband operation is obtained by coupling which will be explained later. D. Measured Radiation Patterns and Gains The radiation patterns of the antenna have been measured using the Small Antenna Radiated Testing Range (SMART) at the National Physical Laboratory (NPL), with results normalized to the maximum values. Figs. 68 show the measured and simulated co- and cross-polarization patterns of the antenna in different switch states at several frequencies across the operating bandwidth.

To design antennas with multiple-band operations, it is desirable to have independent-frequency controls on the frequencies. Achieving this option is very challenging. Very often, when one parameter is changed, all the frequency bands are affected [22], [23] and the antenna needs to be completely re-optimized. Sometimes, the shape of the designed antenna has to be signicantly changed, causing a lot of inconvenience in designing wireless devices. Results in previous sections have shown that, in the ON-OFF and OFF-ON states, the C-Slots on the patch elements of the antenna can be used to generate two frequency bands for dualband operation. Here, we show how to use the C-Slots to independently control the frequency bands for dual-band operation without affecting the wideband operation. In the ON-OFF state, the simulation results in Fig. 8(a) shows the effects of in the slot of patch element #1 on the changing the length lower band of the dual band. It can be seen that increasing moves the 5-GHz band lower but keeps the 5.7-GHz band xed is kept xed here). Fig. 8(b) shows that the effect of (Note, and together on the higher band of the dual changing and together moves the 5.7-GHz band. Here, reducing band to a higher frequency band, yet the 5-GHz band remains unchanged. In the OFF-ON state, Fig. 9(a) shows the effects of can be changing the size of , indicating that the length of used to move the lower band (at around 5.6 GHz), yet keeping the 6.2-GHz band xed. Finally, Fig. 9(b) shows that changing (i.e., moving and together closer or further the distance while keeping the same) can shift the higher band from (at around 6.2 GHz) to a higher or lower frequency band, yet maintaining the 5.6-GHz band. It should be noted that, in some cases, after xing the frequency of the band, we may need to optimize the other parameters of the antenna to achieve the de. These results show that we can independently consirable trol the frequencies of the dual bands by using the C-Slots on

40

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 5. Simulated and measured Co and X-pol in E and H-planes in OFF-ON state at (a) 5.6 GHz and (b) 6.2 GHz.

Fig. 6. Simulated and measured Co and X-pol in E and H-planes in ON-OFF state at (a) 5 GHz and (b) 5.7 GHz.

Fig. 7. Simulated and measured Co and X-pol in E and H-planes in ON-ON state at (a) 5 GHz, (b) 6 GHz, and (c) 7 GHz.

Fig. 8. In ON-OFF state: (a) effects of S on lower band of dual band and (b) effects of W and S together on higher band of dual band. TABLE III MEASURED PEAK GAINS

the patch elements. Simulation results have also shown that the wideband performance in the ON-ON state is not affected when these narrow bands are moved to other frequencies. This degree of freedom further enhances the antenna capability. More simu-

lation tests have shown that the maximum frequency separations that can be achieved between the dual-band in the ON-OFF and OFF-ON states are 130 MHz (5.455.58 GHz) and 150 MHz (5.946.09 GHz), respectively. Based on the above discussions, we propose the following simple guidelines to design an antenna for dual-band operation at desirable frequencies: 1) Optimize the dimensions of the main radiator (the patches without C-Slots) to operate in the 5.6 GHz band. 2) Set the dimensions and positions of the C-Slots in conjunction with the discussion provided in Section III-C.

ABUTARBOUSH et al.: RECONFIGURABLE WIDEBAND AND MULTIBAND ANTENNA USING DUAL-PATCH ELEMENTS

41

Fig. 9. In OFF-ON state, (a) effects of S on lower band of dual band and (b) effects of G on higher band of dual band.

Fig. 10. Antenna with absorber to remove coupling effects (a) Top view and (b) side view.

Fig. 11. Reection coefcient (S ) with absorber in (a) ON-OFF (b) OFF-ON, and (c) ON-ON states.

3) Adjust the locations and dimensions of the C-Slots on each patch elements to achieve the desirable dual bands. Since they can be independently controlled, this can be easily achieved. 4) Optimize the dimensions of the feed network to the patch elements, which is essential for wideband operation. 5) Attach the switches and coupling capacitors at the locations given in Fig. 1. V. EFFECTS OF COUPLING Previous results have shown that when both patch elements are ON, the antenna has a wide bandwidth. This must be the results of mutual coupling between the patch elements, which is examined here. To study the coupling effects between the 2 patch elements, we place an EM wave absorber (high lost material) between the 2 patch elements in the simulation model as shown in Fig. 10(a)(b) to remove the coupling effects and

in the ON-OFF, simulate the impedance bandwidth using OFF-ON and ON-ON states. With patch element #1 turned ON, patch element #2 turned OFF, and the coupling effect from element #2 to element #1 minis shown in Fig. 11(a). imized by the absorber, the simulated without the absorber, i.e., For comparison, the simulated with coupling, is also shown in the same gure. It can be seen , particularly near the dual frequency that the differences in bands, are quite insignicant, indicating that the coupling between the two patch elements is very small for this case. With patch element #1 turned OFF and patch element #2 turned ON, with and without the absorber are shown in the simulated near the dual-frequency bands Fig. 11(b). The differences in are slightly noticeable. Nevertheless, it is insignicant and so the coupling between the two patch elements is still very small. With both patch elements #1 and #2 are ON, i.e., in the ON-ON with and without the absorber are shown state, the simulated in Fig. 11(c). It can be seen that the mutual coupling between

42

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

the 2 elements is much stronger and signicantly reduces the to less than dB across the whole frequency band. As a result, the operation bandwidth of the antenna is much wider. VI. CONCLUSION A recongurable multiband and wideband patch antenna, employing dual-patch elements and C-Slots with a compact volume of 50 50 1.57 mm , has been presented and studied using simulation and measurement. Two PIN diode switches are used to switch ON and OFF two patch elements to operate the antenna in two different dual-band modes or a wideband mode (with a bandwidth of 33.52%). The frequencies in the dual-band modes can be independently control using the C-Slots without affecting the wideband performance. Simulation results have shown that the wideband performance is achieved by the coupling effects between the patch elements. The measured and simulated results have shown that radiation patterns across 57 GHz are stable in different modes. The main advantages of the proposed antenna include low prole, lightweight and easy to fabricate simple structure targeting future smaller wireless communication devices. ACKNOWLEDGMENT The authors would like to thank Skyworks Solutions, Inc., for providing samples used in this work. REFERENCES
[1] G. Kumar and K. P. Ray, Broadband Microstrip Antennas. Boston, MA: Artech House, 2003, pp. 1823. [2] D. M. Pozar and D. H. Schaubert, Microstrip Antennas. New York: IEEE Press, 1995. [3] H. Wang, X. B. Huang, and D. G. Fang, A single layer wideband U-slot microstrip patch antenna array, IEEE Antennas Wireless Propag. Lett., vol. 7, pp. 912, 2008. [4] C. Mak, R. Chair, K. Lee, K. Luk, and A. Kishk, Half U-slot patch antenna with shorting wall, Elect. Lett., vol. 39, pp. 17791780, 2003. [5] Y. Li, R. Chair, K. M. Luk, and K. F. Lee, Broadband triangular patch antenna with a folded shorting wall, IEEE Antennas Wireless Propag. Lett., vol. 3, no. 1, pp. 189192, Dec. 2004. [6] S. Qu and Q. Xue, A Y-shaped stub proximity coupled V-slot microstrip patch antenna, IEEE Antennas Wireless Propag. Lett., vol. 6, pp. 4042, 2007. [7] Y. Lee and J. Sun, A new printed antenna for multiband wireless applications, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 402405, 2009. [8] J. Anguera, C. Puente, C. Borja, and J. Soler, Dual-frequency broadband-stacked microstrip antenna using a reactive loading and a fractalshaped radiating edge, IEEE Antennas Wireless Propag. Lett., vol. 6, pp. 309312, 2007. [9] K.-L. Wong and W.-H. Hsu, A broad-band rectangular patch antenna with a pair of wide slits, IEEE Trans. Antennas Propag., vol. 49, no. 9, pp. 13451347, Sep. 2001. [10] F. Yang, X. Zhang, X. Ye, and Y. Rahmat-Samii, Wide-band E-shaped patch antennas for wireless communications, IEEE Trans. Antennas Propag., vol. 49, no. 7, pp. 10941100, Jul. 2001. [11] R. Bhalla and L. Shafai, Broadband patch antenna with a circular arc shaped slot, in Proc. IEEE Antennas Propag. Soc. Int. Symp., 2002, vol. 1, pp. 394397. [12] A. Sheta and S. Mahmoud, A widely tunable compact patch antenna, IEEE Antennas Wireless Propag. Lett., vol. 7, pp. 4042, 2008. [13] A. Mak, C. Rowell, R. Murch, and C. Mak, Recongurable multiband antenna designs for wireless communication devices, IEEE Trans. Antennas Propag., vol. 55, no. 7, pp. 19191928, Jul. 2007.

[14] S. Yang, C. Zhang, H. Pan, A. Fathy, and V. Nair, Frequency-recongurable antennas for multiradio wireless platforms, IEEE Microw. Mag., vol. 10, no. 1, pp. 6683, Feb. 2009. [15] Y. Huang and K. Boyle, Antennas: From Theory to Practice. Hoboken, NJ: Wiley, 2008, ch. 8. [16] E. Ebrahimi and P. S. Hall, A dual port wide-narrowband antenna for cognitive radio, in Proc. 3rd Eur. Conf. Antennas Propag. (EuCAP), 2009, pp. 809812. [17] F. Ghanem, P. S. Hall, and J. R. Kelly, Two port frequency recongurable antenna for cognitive radios, Elect. Lett., vol. 45, pp. 534536, 2009. [18] R. Kelly, P. S. Hall, and P. Gardner, Integrated wide-narrow band antenna for switched operation, in Proc. 3rd Eur. Conf. Antennas Propag., 2009, pp. 37573760. [19] H. F. AbuTarboush, S. Khan, R. Nilavalan, H. S. Al-Raweshidy, and D. Budimir, Recongurable wideband patch antenna for cognitive radio, in Loughborough Antennas Propag. Conf., 2009, pp. 141144. [20] C. A. Balanis, Antenna Theory, 2nd ed. New York: Wiley, 1997. [21] J. Volakis, Antenna Engineering Handbook. New York: McGrawHill, 2007. [22] R. Sujith, V. Deepu, D. Laila, C. Aanandan, K. Vasudevan, and P. Mohanan, A compact dual-band modied T-shaped CPW-fed monopole antenna, Microw. Opt. Technol. Lett., vol. 51, no. 4, pp. 937939, 2009. [23] S. Lee, H. Park, S. Hong, and J. Choi, Design of a multiband antenna using a planner inverted-F structure, in Proc. 9th Int. Conf. Adv. Commun. Technol., 2007, vol. 3, pp. 16651668.

Hattan F. Abutarboush (M07) received the B.Sc. (Eng) Honors degree in electrical communications and electronics engineering from Greenwich and MSA University, London, U.K., in 2005, the M.Sc. degree in mobile personal and satellite communications from the Department of Electrical and Electronics Engineering, Westminster University, London, in 2007, and the Ph.D. degree in antennas and propagations from the Department of Electronics and Computer Engineering, Brunel University, West London, U.K., in July 2011. His Ph.D. research work was mainly on xed and recongurable multiband antennas. He was a Research Visitor at Hong Kong University in December 2010 and worked as a Research Associate for the American University in Cairo (AUC), Cairo, Egypt, from April to July 2011 where he worked on the packaging of novel millimeter-wave antennas. He has published several journal articles and conference papers. He was invited for special session on multiband antennas at ICEAA 2009 (Italy). His current research interests lie in the design of recongurable antennas, antennas for mobile phones, miniaturized antennas, multiple antennas, smart antennas, antenna arrays, EBG, RF/microwave circuit design, and millimeter-wave antennas. Dr. Abutarboush is a member of IET and a reviewer for several journals and international conferences.

R. Nilavalan (M05SM10) received the B.Sc. Eng. degree in electrical and electronics engineering from the University of Peradeniya, Peradeniya, Sri Lanka, in 1995 and the Ph.D. degree in radio frequency systems from the University of Bristol, Bristol, U.K., in 2001. From 1999 to 2005, he was a Researcher at the Centre for Communications Research (CCR), University of Bristol. At Bristol, his research involved theoretical and practical analyses of post-reception synthetic focusing concepts for near-eld imaging and research on numerical FDTD techniques. Since 2005, he has been with the electronics and computer engineering subject area, Brunel University, where he is currently a Lecturer in wireless communications. His main research interests include antennas and propagation, microwave circuit designs, numerical electromagnetic modeling, and digital video broadcast techniques. He has published over 70 papers and articles in international conferences and journals in his research area. Dr. Nilavalan was a member of the European commission, Network of Excellence on Antennas (20022005) and a member of the IET.

ABUTARBOUSH et al.: RECONFIGURABLE WIDEBAND AND MULTIBAND ANTENNA USING DUAL-PATCH ELEMENTS

43

S. W. Cheung (M82SM02) received the B.Sc. degree (First Class Honors) in electrical and electronic engineering from Middlesex University, Middlesex, U.K., in 1982 and the Ph.D. degree from Loughborough University, Loughborough, U.K., in 1986. From 1982 to 1986, he was a Research Assistant in the Department of Electronic and Electrical Engineering, Loughborough University of Technology, where he collaborated with Rutherford Appleton Laboratory and many U.K. universities to work a project for new generations of satellite systems. From 1986 to 1988, he was a Post-Doctorate Research Assistant with the Communications Research Group of Kings College, London University, working on research for future generations of satellite systems. In 1988, he joined the Radio and Satellite Communications Division in British Telecom Research Laboratories as an Assistant Executive Engineer. He is an Associate Professor at the University of Hong Kong. His current research interests include antenna designs, 2G, 3G, and 4G mobile communications systems, MIMO systems and satellite communications, predistortion of high-power ampliers and e-learning. He has published over 130 technical papers in international journals and conferences. He also has served as reviewer for different international journals and conferences in the areas of antennas and propagation and mobile communications. Dr. Cheung has been serving the IEEE in Hong Kong for the past 20 years. In 2009 and 2010, he was the Chairman of the IEEE Hong Kong Joint Chapter on Circuits and Systems and Communications. Currently, he is the Treasurer of the IEEE Hong Kong Section.

University of Hong Kong to develop transparent green antennas for UWB applications as part of a collaborative research effort. He is currently involved also on collaborative research works with Queen Mary University of London and University of Cambridge. Mr. Thomas was awarded a VCs travel prize by the Graduate School of Brunel University in November 2010 to present his research paper on the development of a Green UWB antenna at the ISAP2010 conference in Macau.

Karim M. Nasr (M05SM11) received the Ph.D. degree in smart antenna systems for indoor wireless networks from the University of Manchester, Manchester, U.K., in 2005. He previously held postdoctoral research positions at the University of Manchester, Brunel University, and BBC Research investigating future wireless and broadcast communication systems through a number of U.K. and European research projects. He was also a Visiting Researcher at the Antennas and Propagation Division of Aalborg University. He is currently a Higher Research Scientist at the National Physical Laboratory (NPL), Teddington, U.K., investigating advanced wireless communication systems and high-precision large-volume laserbased metrology. His research interests include propagation measurements and modeling, DSP and metrology for broadband wireless and broadcast systems, smart antennas and multiuser MIMO systems, UWB, joint Physical/MAC layers optimization and coexistence of wireless systems, advanced antenna metrology, and laser-based coordinate metrology. Dr. Nasr is a reviewer for several IEEE Transactions and a member of TPC of several wireless international conferences. He is a senior member of the IET and European COST Actions 273 and 2100 on wireless communication systems.

Djuradj Budimir (M93SM02) received the Dipl. Ing. and M.Sc. degrees in electronic engineering from the University of Belgrade, Belgrade, Serbia, and the Ph.D. degree in electronic and electrical engineering from the University of Leeds, Leeds, U.K. In March 1994, he joined the Department of Electronic and Electrical Engineering, Kings College London, University of London. Since January 1997, he has been with the School of Electronics and Computer Science, University of Westminster, London, U.K., where he is now a Reader of wireless communications and leads the Wireless Communications Research Group. He has authored and coauthored over 240 journal and conference papers in the eld of RF, microwave and millimeter-wave systems. He is the author of the books Generalized Filter Design by Computer Optimization (Artech House, 1998) and Software and Users Manual EPFIL-Waveguide E -plane Filter Design (Artech House, 2000), and a chapter in the book Encyclopaedia of RF and Microwave Engineering (Wiley, 2005). His research interests include analysis and design of hybrid and MMIC, design of ampliers, lters and multiplexing networks for RF, microwave and millimeter-wave applications and RF, and microwave wireless system design. Dr. Budimir is a Member of the EPSRC Peer Review College and a Charter Engineer.

Thomas Peter received the M.Eng. degree in electrical engineering from the University Technology Malaysia (UTM), Johor Bahru, Malaysia, in 2007. He is currently pursuing the Ph.D. degree in electronics and computer engineering at Brunel University, West London, U.K. His current research interest includes UWB antennas and communications, transparent antennas for green technology, energy harvesting and low detection antennas for stealth. From January to March 2011, he was a Visiting Researcher at the

Hamed Al-Raweshidy (M92SM97) received the Ph.D. degree from Strathclyde University, Glasgow, U.K., in 1991. He was with Space and Astronomy Research Centre (Iraq), PerkinElmer (USA), Carl Zeiss (Germany), British Telecom (Oxford), Manchester Met., and Kent University. He is currently the Director of the Wireless Networks and Communications Centre (WNCC), Brunel University, London, U.K. He published over 250 papers in international journals and referred conferences. He is the editor of the rst book in Radio over Fibre Technologies for Mobile Communications Networks and contributed chapters for six books. He is Editor-in-Chief of Communication Networks Journal (USA). He has acted as Guest Editor for the International Journal of Wireless Personal Communications. He is a member of several journal editorial boards such as the Journal of Communications and Mobile Computing and Journal of Wireless Personal Communications. He act as a consultant and involved in projects with several companies and operators such as Vodafone (U.K.), Ericsson (Sweden), Andrew (USA), NEC (Japan), Nokia (Finland), Siemens (Germany), Franc Telecom (France), Thales (U.K. and France), and Tekmar (Italy). He has been able to attract over 3,000,000 from research projects.

44

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Frequency-Recongurable Monopole Antennas

Abubakar Tariq, Student Member, IEEE, and Hooshang Ghafouri-Shiraz, Senior Member, IEEE
AbstractA novel coplanar-waveguide (CPW) bandpass lter using short-circuit slotlines and varactor diodes is introduced. This bandpass lter, when integrated with a CPW wideband antenna, produces frequency agility with a wideband mode and a continuous narrowband mode. The design of another CPW lter based on a square-ring resonator with switches is also shown and applied to a wideband antenna, making it recongurable. Both lters are based on controlling two stop bands far enough apart so that there is a passband between them. When the stop band frequencies are altered using switches or varactors, the passband is altered. Complete working and design principles along with simulated S-parameter results of the lters are presented. The simulated and measured reection coefcients of the antennas incorporating the lters are also shown. Good monopole-like radiation characteristics are observed for both antennas. The lters are small in size and can be incorporated in any CPW antenna design to make it recongurable. The benets of the novel varactor lter antenna over the switch lter antenna are also described. Index TermsBandpass lter, coplanar waveguide (CPW), ltering antenna, frequency recongurability, wideband antenna.

I. INTRODUCTION HERE HAS been rapid progress in wireless communication technologies during recent years. Communication devices these days can operate with different standards and their associated frequencies (e.g., Global System for Mobile Communications (GSM), Universal Mobile Telecommunication System (UMTS), WiFi, personal communications services (PCS), and Bluetooth). This has led to an increased interest in the efcient use of electromagnetic spectrum and the development of frequency-agile antennas. Conventionally, the problem of frequency recongurability was solved by switching between elements [1]. This method has the disadvantages of increased size, complexity, and undesirable coupling between elements. Thus, frequency-agile antennas need to be designed which incorporate lters or slots within the antenna structure to obtain agility so that the aforementioned disadvantages can be overcome. Frequency-agile antennas, which can scan the spectrum in a wideband mode and choose an available narrowband mode, also provide the best solution for software-dened and cognitive radios [2], [3]. Considering the advantages and uses of frequency-recongurable antennas, signicant research has been reported on them

Manuscript received January 20, 2011; revised May 11, 2011; accepted June 21, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported by the University of Birmingham and Overseas Research Student Award Scheme. The authors are with the Electronic, Electrical and Computer Engineering Department, University of Birmingham, Birmingham B15 2TT, U.K. (e-mail: Axt769@bham.ac.uk; ghafourh@bham.ac.uk). Digital Object Identier 10.1109/TAP.2011.2167929

[1][9]. Reference [4] gave a comprehensive review of various methods used to achieve frequency agility using patch, wire, planar-inverted F antennas, etc. They also provided a study of the pros and cons associated with existing solutions for wireless mobile platforms. Compact slot antenna recongurability was achieved using switches [5] and varactors [6], providing good tuning ranges along with maintaining similar radiation characteristics at all frequencies. A very wide tuning ratio of 3.52:1 was achieved in a slot antenna using only a PIN diode and a varactor diode while maintaining a consistent radiation pattern throughout the tuning range [1]. Vivaldi antenna recongurability was achieved in [7], providing three narrowband (low, mid, high) and a wideband operation. Dynamic control of band rejection along with a wideband state were achieved in [8] using PIN diodes. A conceptual model for the design was also introduced. The components usually used to achieve recongurability are switches and varicap diodes and will be utilized by the authors. Switches can either be PIN diodes or microelectromechanical systems (MEMS). MEMS have the advantage of low power consumption, because the application of voltage can turn the switch on or off without any current ow but the actuation voltage required is quiet high. Switches obtain agility (pattern, polarization, frequency) by physically changing the connectivity or length of certain elements (resonators, lters, stubs) within an antenna. Varicap diodes provide recongurability by capacitively loading antennas or by electrically changing the length of antennas or lters. In this paper, two methods of introducing frequency agility to wideband monopole antennas are presented, one employing PIN diodes and the other using varactors. Both methods use the same principle of combining a wideband antenna with a frequency-recongurable bandpass lter incorporated into the structure of the antenna. The passband of the rst lter is controlled by actuslots by placing switches ally changing the length of two slots produce two along the length of the slots. The two stopbands with a passband resulting in between them. Thus, changing the length, changes the stopband frequencies, causing a change in the passband frequency. For the second design, a novel lter is introduced which varies two stopbands produced resonator and its rst harmonic at . by a fundamental This variation of the stopbands is achieved by introducing the varactor at a suitable point along the resonator that affects the fundamental and harmonic resonance. This design using varactors instead of switches provides the additional advantages of low power consumption, smaller area, simpler feed network, fewer lumped elements, low cost, and a continuous passband range rather than a discrete one. Section II describes the lter design procedure and validates the working principle. Section III presents the simulated and measured results of the fabricated antennas using the lters de-

0018-926X/$26.00 2011 IEEE

TARIQ AND GHAFOURI-SHIRAZ: FREQUENCY-RECONFIGURABLE MONOPOLE ANTENNAS

45

Fig. 1. Topology of the CPW switched Filter. 0.5 mm, 1 mm, 8 mm, 2 mm, 2 mm, 3.35 mm, 0.5 mm, 3 mm.

Fig. 2. S-parameter results of the CPW switched lter shown in Fig. 1 with resulting in resonators of lengths 18.35 mm and 9 mm.

scribed in Section II. Then, there are the conclusions in Section IV. II. ANTENNA FILTER DESIGN A. Switched Filter Design The design principle for the lter using switches is based on utilizing resonators to create two stopbands far apart enough to make a passband possible between them. This design is modied from the unequal resonator lter design of [9]. The lter in [9] was used to switch between single, double, and 2 triple passbands. The proposed lter using switches in this paper is designed to achieve recongurability of a single passband by placing switches along the rings. This lter, when integrated with a wideband antenna, results in a frequency-agile system with a wideband operation when the lter is disconnected by placing switches in the ON state at the entrance of the square rings and narrowband operations when different diode switch combinations along the square ring are biased ON/OFF. The basic structure for the lter was a coplanar waveguide (CPW) with a square-ring resonator as shown in Fig. 1. One permanent copper strip is present to avoid the use of an RF choke inductor in the square-ring resonator because of its effect on the S-parameter response at the frequencies used. This permanent copper strip divides the square ring into two unequal stopband resonators. Along the larger resonator, three RF diode switches are placed which change the length of this resonator, resulting in shifting of the stopband frequency at which it resonates. Although the shorter resonators length remains constant, its resonant frequency varies very slightly. When the signal at point X in Fig. 1 enters the square-ring resonator, it sees two unequal resonators. These resonators produce two transmission zeroes resulting in two stopbands. By controlling the length of these resonators, the stopband positions can be controlled and, hence, the passband can be controlled, resulting in a frequency-agile passband lter. This will rst be demonstrated in simulation using Computer Simulation Technology Microwave Studio (CST MWS). In the initial study, switches were modeled as 0.3-mm-wide copper wires to demonstrate the design principle so that the effect of the nonideal nature of diode is not taken into account. In the antenna
Fig. 3. Electric-eld distribution at (a) 3.56 GHz and (b) 6.698 GHz clearly resonator resonating. showing larger and smaller

design stage, actual diode parameters provided by the relevant manufacturers will be used in simulations. In Fig. 1, when the switch combination is ON, the effective length of the larger and smaller resonators is 18.35 mm and 9 mm which corresponds to frequencies of 3.32 GHz and 6.58 GHz, respectively, calculated using TXLINE from AWR. The simulated S parameter results of this combination of resonators lter are shown in Fig. 2. Two stopbands at 3.56 and 6.698 GHz can be clearly observed with a passband at 5.432 GHz in between them. This agrees well with the calculated values obtained from TXLINE. Fig. 3(a) and (b) shows the electric-eld distribution at 3.56 GHz and 6.698 GHz, respectively. It can be clearly seen that slot lengths of 18.35 mm and 9 mm are resonating at 3.56 GHz and 6.698 GHz, respectively. For the other switch combinations in Fig. 1, the simulation of the stopband and passband frequencies is shown in Fig. 4(a) and (b) respectively. The lower stopband frequency moves considerably as different switch combinations are used and the higher stopband frequency remains almost constant. This higher stopband frequency can be moved as well if more switches are placed in the path of the smaller 9 mm resonator. But in this paper, obtaining the most passbands with the least number of switches was desired. A tuning range of 1.7 (3.63 to 6.18 GHz) is achieved with this lter. B. Varactor Filter This section will introduce a novel lter with a continuous frequency tuning range. It employs varactor diodes as the tuning element. The basic idea is a slot which resonates at and (1st harmonic), producing two transmission zeros resulting in two stopbands. Fig. 5 shows the varactor lter along

46

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 6. Simulated S-parameters of the CPW lter in Fig. 5 without the varactor.

Fig. 7. Normalized voltage wave in the short-circuit slotline at the fundamental frequency and rst harmonic. Fig. 4. Simulated (a) and (b) results of the lter in Fig. 1 with different switch combinations. Resonant frequencies of 3.63 GHz, 4.95 GHz, and 6.18 GHz are achieved in (b).

Fig. 8. Simulated reection coefcient of the short-circuit slotline with the varactor at B (null point of the rst harmonic). Fig. 5. Topology of the CPW varactor lter: the dimensions are 2.1 7 mm, 3 mm, 4 mm, and 1 mm. The CPW mm, dimensions are the same as Fig. 1.

with its dimensions. This slot of length 13 mm without any varactors resonates at and , which is equal to 4.62 and 13.3 GHz, respectively, calculated using TXLINE form AWR. The S-parameter simulation of the lter shown in Fig. 5 without any varactors is shown in Fig. 6. The calculated and simulated values agree quite well. Fig. 7 shows the normalized voltage wave of the resonator shorted at one end and open ended at the other. The rst harmonic at is also shown. Point S and O represent the short- and open-circuit points, respectively. Point B represents the null point of the rst harmonic and point A represents the position at which the varactor will be placed in our lter design. The varactor was modeled in CST MWS as a capacitor with a forward resistance of 2.5 , which was the forward resistance of the varactor diode obtained from manufacturers data sheet. In order to verify the working of the lter, a varactor was placed at point B. The resulting simulation is shown in Fig. 8. It can be clearly observed as expected that the resonators

resonance does not change from a value of 12.7 GHz since the varactor is placed at its null point but the resonance of the resonator changes. When the varactor position is optimized and it is placed at point A, it can affect the fundamental and rst harmonic resonators, resulting in S-parameters shown in Figs. 9 and 10. Thus, a continuous passband from 2.57 to 5.54 GHz is obtained, giving a tuning range of 2.16. This design has several advantages over the switched lter design and the lter in [9], including simpler structure, less lumped elements, low cost, low power consumption, continuous passband, and smaller size. III. ANTENNA DESIGN The two lters described in the previous section, when combined with any wideband CPW antenna, provide frequency-recongurable operation along with wideband operation when the entrance to the square resonator or short-circuit slot is blocked. A. Wideband Antenna The CPW antenna chosen here to demonstrate the working of the two lters is an elliptical monopole antenna, which is

TARIQ AND GHAFOURI-SHIRAZ: FREQUENCY-RECONFIGURABLE MONOPOLE ANTENNAS

47

Fig. 9. Simulated

result of the short-circuit slotline with the varactor at A.

Fig. 12. Switched monopole antenna along with the biasing circuit, the lumped elements connecting with the bias lines are the RF chokes (inductors), the lumped elements connecting cut portions of copper are dc blockers (capacitors), and the rest are diode switches.

Fig. 10. Simulated reection coefcient of the short-circuit slotline with the varactor at A. Resonant frequencies vary from 2.57 to 5.54 GHz.

Fig. 11. Topology of the wideband CPW antenna. The dimensions are 44 24 mm, 0.5 mm, 5 mm, 6 mm}, 2 mm, mm, 17.5 mm, and 25 mm.

shown in Fig. 11 along with its dimensions. The antenna was fabricated on a TLC-30 substrate with 3 and a thickness of 1.56718 mm. B. Switched Monopole Antenna The switched monopole antenna will combine the switched lter from Section II-A and the wideband monopole antenna introduced in the previous section. The diodes used as switches are BAR50-02V from Inneon. They were chosen because of

their wide frequency range (10 MHz to 6 GHz), low capacitance 0.15 pF at zero bias voltage above frequencies of 1 GHz, and low forward resistance. Accu-P-0603-22pf-25V capacitors from AVX were used as dc blockers, and 0604HQ-2N6XJLB inductors from coilcraft served as RF chokes. Fig. 12 shows the antenna along with the biasing circuit. The total size of the antenna is 51 71 mm, including the biasing circuit. This switched monopole antenna has ve modes of operation When the diode switches (switch combination 1) placed at the entrance of the ring are on, a wideband is obtained because no current ows into the square slot effectively disconnecting the square-ring resonator, resulting in wideband operation. The measured and simulated results for wideband operation are shown in Fig. 13(a). The other four narrowband modes of operation are obtained by switching on the three switches along the longer resonator as described in Section II-A one at a time or keeping them all off. The different switch combinations producing the two short-circuit slots along with the measured and simulated reection coefcients are shown in Fig. 13(b)(d). Fig. 13(e) shows the combined measured results for all switch combinations. As can be seen, a tuning range from 3 to 4.45 GHz is observed. For simulations, actual diode parameters were used for both the ON and OFF state of the diode obtained from Inneon. Both simulation and measurements show good agreement with each other. All of the errors can be attributed to fabrication problems and the fact that the feed network along with the feeding wires was not included in the simulation. Also, the capacitor and inductors used are imperfect and introduce errors of their own. As shown in Section II-A, the switch 2 combination produced results when a perfect copper strip was used, but in the actual design when the off state capacitance of the diode was taken into account, the results were unsatisfactory but some agreement was still observed between the simulated and measured results. Only the measured result is shown in Fig. 13(e). C. Varactor Antenna For the varactor-tuned antenna, the varactor lter design from Section II-B is combined with the wideband monopole antenna from Section III-A. The varactor diodes used were SMV1231 from Skyworks which have a tuning range from 2.2 to 0.466 pF when the voltage is changed from 0 to 8 V. It has a

48

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 14. Varactor monopole antenna along with the biasing circuit; the lumped elements connecting with the bias lines are the RF chokes (inductors), the lumped elements connecting cut portions of copper are dc blockers (capacitors), and the lumped element at the entrance of the slot consists of switches, and varactors are along the slot.

along with its biasing circuit, is shown in Fig. 14. The total size of the antenna, including the feed structure, is 51 51 mm. The varactor monopole antenna has a wideband operation when diodes placed at the entrance of the short-circuit slot are turned on. Besides this, it has a continuous narrowband mode of operation ranging from 2.88 to 4.62 GHz when the voltage is changed from 0 to 8 V. Simulations took into account the 2.5forward resistance of the varactor along with the offstate capacitance of the diode. Fig. 15 shows the measured and simulated reection coefcients for the narrowband and wideband operation. Both simulation and measured results show good agreement with one another but the range of the narrowband operation in simulation is higher compared to measured results. The range difference could be accounted for the actual capacitive range of the varactor being smaller compared to the datasheet. All of the other differences can be attributed to fabrication errors, feed network and wires, and imperfections of the components (varactor diode, PIN diode RF choke, and dc blockers). The slots which make the lters are demonstrated on CPW-fed antennas in this paper, but when these slots are applied to tapered slot antennas, such as Vivaldi antennas, similar results are obtained and the same principles apply. So this design is applicable to tapered slot antennas as well. D. Radiation Pattern The radiation pattern of the switched monopole antenna and the varactor-tuned antenna was measured and simulated. In simulations, it was discovered that there was very little difference between the radiation pattern of the antenna without any squarering resonator or slotline and the antennas with these elements. The reason is that the slots are weakly excited in the passband. Only the maximum and minimum passband frequency radiation patterns of both antennas will be presented here because the results in other frequency bands and in the wideband mode are similar. Fig. 16 shows the measured and simulated radiation patterns in the E and H plane of the switched monopole antenna at 3 and 4.45 GHz. Fig. 17 shows the measured and simulated radiation patterns in the E and H plane of the varactor-tuned monopole antenna at 2.88 and 4.62 GHz. There are differences between the simulated and measured results in the copolarization and crosspolarization patterns which can be accounted to

Fig. 13. (a) Wideband measured and simulated reection coefcient when switch combination 1 is on. (b) Narrowband measured and simulated reection coefcient when switch combination 3 is on. Measured resonance at 4.45 GHz. (c) Narrowband measured and simulated reection coefcient when switch combination 4 is on. Measured resonance at 3.8 GHz. (d) Narrowband measured and simulated reection coefcient when no switch is on. Measured resonance at 3 GHz. (e) Reection coefcient measured results for all switch combinations.

forward resistance of 2.5 . Accu-P-0603-22pf-25V capacitors from AVX were used as dc blockers and 0604HQ-2N6XJLB inductors from coilcraft served as RF chokes. The antenna,

TARIQ AND GHAFOURI-SHIRAZ: FREQUENCY-RECONFIGURABLE MONOPOLE ANTENNAS

49

Fig. 15. (a) Measured and (b) simulated reection coefcients of the varactor monopole showing wideband mode and narrowband modes when the capacitance is changed. Fig. 16. Measured and simulated radiation patterns at 3 and 4.45 GHz for the switched monopole antenna.

the fact that the dc feed wires and feed network were not used in simulations. The differences in the switched monopole antenna are more pronounced since it has more dc feeding wires. When an antenna with the square-ring slot using copper strip instead of PIN diodes, without feeding wires, was tested, the obtained results were much closer to simulated results, showing the wires are causing the major degradation. This observation is also made in [8]. The radiation pattern degradation can be improved by using a system containing chips which can all be controlled by a single dc feed line[10] or by using a dc bias network with a coin battery, which makes the length of the dc bias lines minimal [11]. The switches and varactors used have an effect on the overall efciency and gain of the antenna as they introduce losses. When actual switch parameters instead of ideal switch parameters are used, the efciency reduces from 89.13% to 88.72% at 3.2 GHz and from 96.83% to 93.97% at 4.4 GHz for the switch monopole antenna according to simulation results. When gain measurements are done in an anechoic chamber for the antenna with all of the RF elements, wires and feed network compared to an antenna with 0.3-mm copper strips as switches, the gain reduces from 0.19 to 0.06 dB at 3 GHz and from 2.87 to 0.04 dB at 4.4 GHz, respectively. For the varactor antenna, the results are better as fewer RF elements, wires, and a smaller feed network are used. When actual switch and varactor parameters instead of ideal switch and varactors are used, the efciency reduces from 92.47% to 90.57% at 2.8 GHz and from 96.83% to 93.76% at 4.62 GHz for the switch monopole antenna according to simulation results. When gain measurements are done in an anechoic chamber for the antenna with all of the RF elements, wires, and feed network compared to an antenna with 0.3-mm copper strips as switches and a xed capacitor, the gain reduces from 0.34 to 0.21 dB at 2.8 GHz

Fig. 17. Measured and simulated radiation patterns at 2.88 GHz and 4.62 GHz for the varactor monopole antenna.

and from 2.58 to 1.2 dB at 4.62 GHz. It is observed that there is a greater reduction in gain at higher frequencies compared to lower frequencies for both types of antennas. This can be

50

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

attributed to the fact that at higher frequencies, the insertion loss of the RF components used is higher. This is also conrmed from the component datasheets provided by the manufacturers. IV. CONCLUSION A novel CPW varactor lter is introduced which can achieve a wide tuning range by changing the resonant frequency of and resonators using varactors. Another CPW lter using switches to control two resonators is also designed. These CPW lters are very compact and can be integrated within the structure of an antenna without taking any extra space. The integration of these lters with CPW wideband antennas results in frequency agility. Good agreements between simulated and measured results are obtained for the S-parameters of the lters and their respective antennas. The working principles of both lters are validated by using slotline equations and simulations. Since wideband and narrowband operation are obtained, these CPW wideband antennas with integrated lters are excellent for cognitive radio applications. The radiation pattern of both antennas is simulated and measured and ways to improve the pattern degradation have been suggested. Future works include using these lters to include a passband and stopband operation along with making multiple bands recongurable. REFERENCES
[1] H. Li, J. Xiong, Y. Yu, and S. He, A simple compact recongurable slot antenna with a very wide tuning range, IEEE Trans. Antennas Propag., vol. 58, no. 11, pp. 37253728, Nov. 2010. [2] P. S. Hall, P. Gardner, J. Kelly, E. Ebrahimi, M. R. Hamid, F. Ghanem, F. J. Herraiz-Martinez, and D. Segovia-Vargas, Recongurable antenna challenges for future radio systems, in Proc. 3rd Eur. Conf. Antennas Propag., Berlin, Germany, 2009, pp. 949955. [3] S.-H. Oh, J. T. Aberle, S. Anantharaman, K. Arai, H. L. Chong, and S. C. Koay, Electronically tunable antenna pair and novel RF front-end architecture for software-dened radios, EURASIP J. Appl. Signal Process., vol. 2005, pp. 27012707, 2005. [4] S. Yang, C. Zhang, H. K. Pan, A. E. Fathy, and V. K. Nair, Frequency recongurable antennas for multiradio wireless platforms, IEEE Microw. Mag., vol. 10, no. 1, pp. 6683, Feb. 2009. [5] D. Peroulis, K. Sarabandi, and L. P. B. Katehi, Design of recongurable slot antennas, IEEE Trans. Antennas Propag., vol. 53, no. 2, pp. 645654, Feb. 2005. [6] N. Behdad and K. Sarabandi, A varactor-tuned dual-band slot antenna, IEEE Trans. Antennas Propag., vol. 54, no. 2, pp. 401408, Feb. 2006. [7] M. R. Hamid, P. Gardner, P. S. Hall, and F. Ghanem, Recongurable vivaldi antenna, Microw. Opt. Technol. Lett., vol. 52, no. 4, pp. 785787, Apr. 2010.

[8] J. Perruisseau-Carrier, P. Pardo-Carrera, and P. Miskovsky, Modeling, design and characterization of a very wideband slot antenna with recongurable band rejection, IEEE Trans. Antennas Propag., vol. 58, no. 7, pp. 22182226, Jul. 2010. [9] H. B. El-Shaarawy, F. Coccetti, R. Plana, M. El-Said, and E. A. Hashish, Novel recongurable defected ground structure resonator on coplanar waveguide, IEEE Trans. Antennas Propag., vol. 58, no. 11, pp. 36223628, Nov. 2010. [10] T. H. Hand and S. A. Cummer, Controllable magnetic metamaterial using digitally addressable split-ring resonators, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 262265, 2009. [11] S. J. Wu and T. G. Ma, A wideband slotted bow-tie antenna with recongurable CPW-to-slotline transition for pattern diversity, IEEE Trans. Antennas Propag., vol. 56, no. 2, pp. 327334, Feb. 2008. Abubakar Tariq (S10) was born in Lahore, Pakistan, in 1984. He received the B.Sc. (Hons.) degree from the University of Engineering and Technology, Lahore, Pakistan, in 2007 and the M.Sc. degree from the University of Birmingham, Birmingham, U.K., in 2008, where he is currently pursuing the Ph.D. degree in electronic, electrical and computer engineering. His research interests include biomedical applications of radio-frequency/microwave sensors, digital signal processing, and frequency- and pattern-recongurable antenna design. Mr. Tariq is a student member of the Institute of Engineering and Technology. He was the recipient of the Best Student Award in 2008. He is also a recipient of a number of scholarships, including the Overseas Research Student Award Scheme 20082011, University of Birmingham Scholarship 20082011, TI group postgraduate scholarship 2007, and School of Engineering High Flyer Award 2007.

Hooshang Ghafouri-Shiraz (S85M86SM88) received the B.Sc. and M.Sc. degrees in electronic and electrical engineering from Shiraz University, Shiraz, Iran, in 1973 and 1978, respectively, and the D.Eng. degree from the University of Tokyo, Tokyo, Japan, in 1985. From 1985 to 1987, he was a Senior Research Fellow in the Second Research Laboratory, Optoelectronic Development Division, Anritsu Corporation, Japan. Since 1987, he has been a member of the academic staff at the Department of Electronic, Electrical and Computer Engineering, University of Birmingham, Birmingham, U.K. Currently, is a Reader in Optical and Microwave Communications and the Head of Postgraduate Research (Ph.D.) at the University of Birmingham. He has published more than 230 journal and conference papers, three chapters contributions and the following ve books: 1) Optical CDMA Networks: Principles, Analysis and Applications (Wiley, 2012); 2) The Principles of Semiconductor Laser Diodes and Optical Ampliers: Analysis and Transmission Line Laser Modelling (Imperial College Press, 2004); 3) Distributed Feedback Laser Diodes and Optical Wavelength Tunable Filters (Wiley, 2003); 4) Fundamentals of Laser Diode Ampliers (Wiley, 1996); and 5) Distributed Feedback Laser Diodes: Principles and Physical Modelling (Wiley, 1996). His research interests are in optical devices, optical code-division multiple access, and antennae for vital signs monitoring.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

51

Low Prole Fully Planar Folded Dipole Antenna on a High Impedance Surface
Andrea Vallecchi, Javier R. De Luis, Member, IEEE, Filippo Capolino, Senior Member, IEEE, and Franco De Flaviis, Senior Member, IEEE

AbstractA fully planar antenna design incorporating a high impedance surface (HIS) is presented. The HIS is composed by a periodic array of subwavelength dogbone-shaped conductors printed on top of a thin dielectric substrate and backed by a metallic ground plane. First, the characteristics of a dipole over PEC or PMC layers, a dielectric slab, and the HIS are compared and studied in detail, highlighting the advantages provided by the use of the HIS. Then, the design of a low prole folded dipole antenna working at 5.5 GHz on top of the HIS is described. The surface provides close to 6% antenna impedance bandwidth and increased gain up to 7 dBi, while shielding the lower half space from radiation. The antenna structure comprises three metal layers without any vias between them, and its overall thickness is . The dipole is fed by a balanced twin lead line through a balun transformer integrated in the same antenna layer. A prototype has been built and measurements conrming simulation results are provided. Index TermsAMC ground plane, articial magnetic conductor, articial magnetism, articial substrate, impedance surface, low prole antenna, metamaterials, reection phase.

I. INTRODUCTION INCE the advent of electromagnetic bandgap (EBG) materials and metamaterials, researchers have tried to realize articial magnetic conductors (AMCs) [1][5] that can provide an effective shielding of radiation without requiring the vanishing of the tangential electric eld. One important application of AMCs is, for example, as the ground plane of an electric dipole for a low prole design with increased forward directivity [6][13]. In this context, according to different terminologies in use, the metamaterial surface is denoted either as EBG ground plane [7], [8], [10], [14], reactive impedance substrate [9], magnetic ground plane, or high impedance surface (HIS) [5].
Manuscript received September 03, 2010; revised May 17, 2011; accepted June 21, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the Semiconductor Research Cooperation (SRC)-GRC under Grant 2009-VJ-1962. A. Vallecchi was with the Department of Electrical Engineering and Computer Science, University of California-Irvine, Irvine, CA 92697 USA. He is now with the Department of Information Engineering, University of Siena, 53100 Siena, Italy (e-mail: andrea.vallecchi@unisi.it). J. R. De Luis, F. Capolino, and F. De Flaviis are with the Department of Electrical Engineering and Computer Science, University of California-Irvine, Irvine, CA 92697 USA (e-mail: jrodrig2@uci.edu; f.capolino@uci.edu; franco@uci.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167912

High impedance surfaces have been initially realized by using mushroom type structures [1], [7], [14], [15] requiring a ground plane and multiple metallic vias. However, some authors have suggested that the presence of vias could produce distortion of the radiation pattern within the operating band [10]. In this work an articial reactive impedance surface is realized using a periodic structure printed on a planar grounded substrate without any via. This structure is similar to the one presented in [5] though its modeling is substantially different since in this case a subwavelength thickness is used for the dielectric layer. Recently, a planar metamaterial consisting of paired metallic conductors with dogbone shape was shown to support symmetric (electric) and antisymmetric (magnetic) resonance modes. The antisymmetric mode, also referred to as transmission line mode, has been demonstrated to be associated with articial magnetism [11][13], [16][18]. In accordance with image theory, the magnetic resonance can be obtained by using a single layer of dogbones over a perfect electric conductor (PEC) plane, therefore reducing by half the required thickness of the HIS with respect to the basic conguration in [13]. The proposed structure does not require any via between its constituent layers and can be entirely fabricated in planar technology [19], [20]. Along the lines of the design reported in [19], the HIS is used in this work to enhance the bandwidth and directivity performance of a printed dipole, located very close to an underneath metallic ground plane. The proposed antenna consists of a three metal layer structure. The top layer allows the placement of any planar radiating element or array that can benet from the properties of the HIS. The intermediate layer accommodates the periodic metallic dogbone-like patterned surface that constitutes the metamaterial, whereas the bottom layer serves as a sheltering ground plane. The design is centered within the 5-GHz band for WiFi/WLAN applications. The organization of the paper is the following. In Section II, different designs of a low prole dipole antenna placed either on top of PEC and PMC layers or a dielectric slab are analyzed in order to highlight the advantages of using a HIS. Then, in Section III, the fully planar HIS with subwavelength thickness and its effective transmission line model are presented. Section IV describes the design of a low prole folded dipole on top of the proposed HIS. Simulations and measurements are presented in Section V. Finally, the inuence of the HIS size on the dipole antenna performance are numerically investigated in Section VI.

0018-926X/$26.00 2011 IEEE

52

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 1. 3D radiation patterns of a strip planar dipole operating at 5.5 GHz at a from (a) PEC and (b) PMC layers with size . Red distance and green colors indicate strong and weak eld intensity, respectively.

II. ANALYSIS

OF

DIFFERENT LOW PROFILE ANTENNA DESIGN APPROACHES

In order to understand the benets of using a HIS, different designs of a low prole dipole antenna are compared in a comprehensive manner. We rst consider the cases when a dipole is placed at a certain small electrical distance above a PEC or PMC plane, as shown in Fig. 1(a) and (b), respectively. Although PMC materials do not exist in nature, this design is considered for comparative purposes. The enhancement of the antenna directivity for the dipole over the PEC surface is larger than for the PMC case, as shown in Fig. 1. However, other important antenna parameters are differently affected by the presence of the PEC or PMC reecting surfaces as discussed next, and compared to the performance of a dipole over a HIS, discussed in the next sections. A simulation experiment is conducted with HFSS to understand the behavior of a strip dipole located close to a PEC or PMC layer. In Fig. 2 the distance between a dipole with length and a PEC plane with area is increased from to , where denotes the free space wavelength at 5.5 GHz. Fig. 2(a) shows the input impedance as a function of the dipole distance from the PEC or PMC plane. When the dipole approaches the plane, the input resistance decreases considerably for the PEC case (the dipole and image currents cancel each other), whereas it reaches the largest value for the PMC case (the dipole and image currents add up in phase). On the other hand, when the distance of the dipole from the reecting surface is larger than the input impedance starts to be comparable with the free space values for both the PEC and PMC cases. The directivity of the dipole at small distances from the conducting plane, shown in Fig. 2(b), is at least 3 dBi higher for the PEC surface compared to the PMC case. This is also conrmed by the shape of the radiation patterns shown in Fig. 1 when the dipole is just above the PEC or PMC plane. From this experiment we can outline some important considerations upon the design aspects of a low prole dipole above a PEC plane:

Fig. 2. (a) Input impedance and (b) forward directivity of a dipole working at 5.5 GHz versus the dipole separation distance from a PEC or PMC nite ground . plane with size of

(a) The low radiation resistance (comparable in some cases to the loss resistance) implies that in a system with losses, the antenna radiation efciency decreases considerably. Besides, impedance matching the antenna to a 50 Ohm system will become increasingly more difcult as the distance of the dipole from the reecting surface is reduced. (b) The energy stored in the near eld of the dipole will be large, producing an antenna with a high (dened as the ratio of energy stored to energy accepted by the antenna [21]) and, correspondingly, with a very narrow impedance bandwidth. The last point is ultimately the main bottleneck of using a dipole over a PEC plane for most practical applications. On the other hand, the impedance value for the low prole dipole over the PMC layer seems more favorable from the antenna matching point of view, but in this case the radiation pattern does not show the desired directivity enhancement. A folded dipole could be used over the PEC plane in place of the regular dipole, as suggested in [21], to increase the input resistance of the antenna and achieve a better impedance matching. The reection coefcient and input impedance for a folded dipole with length of located above a PEC plane are compared with the corresponding quantities for the regular dipole over PEC and PMC planes in Fig. 3 and Fig. 4, respectively. The reection coefcient (dashed line) in this case is below 12 dB at 5.45 GHz. Even though matching can be improved by using a folded dipole, the disadvantage of a narrow impedance bandwidth (2.75% at the level of 10 dB) is still present, as apparent from Fig. 3.

VALLECCHI et al.: LOW PROFILE FULLY PLANAR FOLDED DIPOLE ANTENNA ON A HIGH IMPEDANCE SURFACE

53

Fig. 3. Reection coefcient for different dipole antenna congurations: reg(5.5 GHz) in front of a PEC layer (solid black ular dipole of length line); regular dipole in front of a PMC layer (dashed blue line); folded dipole (5.5 GHz) in front of a PEC plate (solid red line). In all with length away from the cases, the antenna is placed at a distance of 3.22 mm . PEC plate, whose size

Fig. 5. Magnitude of the reection coefcient of a dipole of length on , thickness top of a grounded dielectric slab with dielectric permittivity and area .

Fig. 4. Input resistance and reactance corresponding to the three different antenna congurations considered in Fig. 3.

Fig. 6. 3D radiation pattern at 5.44 GHz (a), 5.58 GHz. of the dipole with reection coefcient shown in Fig. 5. Red and green colors indicate strong and weak eld respectively. The dipole is placed (printed) on top of a dielectric stub . of thickness

In order to design a low prole dipole one could also think of placing the dipole on top of a high permittivity dielectric slab backed by a metallic conductor plane. Besides yielding to small thicknesses because of a smaller effective wavelength in the dielectric substrate, this results also in a shorter dipole length . The drawback is that the high permittivity substrate supports surface waves that can trap a signicant amount of power emitted by the antenna and become the main radiation mechanism, which distorts the dipole radiation pattern. In this scenario, the patterns become of the endre type, as the surface waves will diffract at the substrate edges and radiated into space. In addition, if the waves are propagating in a lossy substrate, the radiation efciency will drop resulting in a low gain antenna. Fig. 5 shows the reection coefcient of a dipole of length on top of a grounded dielectric slab with relative permittivity , thickness and size . The effect of the dielectric slab is to provide two different resonances at 5.44 GHz and 5.58 GHz. The pattern associated with these resonances corresponds to a surface wave radiated in the (Fig. 6(a)) and direction (Fig. 6(b)), respectively. Therefore, the patterns become of the endre type, and the dipole merely acts as the source for the surface wave propagation that will be diffracted at the substrate edges and radiated into space.

III. HIGH IMPEDANCE SURFACE From the discussion in the previous section, it is clear that the PEC, PMC and dielectric slab cases are complementary in terms of advantages. While the PEC ground plane provides a directive broadside pattern but suffers from high quality factor and narrow impedance bandwidth, the ideal PMC surface shows improved bandwidth but lower directivity. On the other hand, a high dielectric permittivity substrate results in most of the power emitted by the dipole being trapped in the substrate. A HIS can combine the advantages of both PEC and PMC ground planes. In fact, a proper HIS design can achieve an improved bandwidth with respect to the PEC case and higher forward directivity with respect to the PMC case. Fig. 7 shows the developed HIS design, that is formed by a doubly periodic array of dogbone-shaped conductors printed on a grounded dielectric substrate with thickness (dielectric plus copper thickness), relative dielectric constant , and (e.g., RT/Duroid 5880). The various geometrical parameters characterizing the unit cell of this metamaterial (Fig. 7(b)) are as follows (in mm): , , , , , , . The thickness of the copper metal layers is 35 . The total

54

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 7. (a) HIS composed of a doubly-periodic array of dogbone-shaped conductors printed on a PEC-backed dielectric material. (b) Unit cell of the HIS with indication of geometrical parameters.

Fig. 8. Phase of the reection coefcient calculated from CST full-wave simulations, de-embedded at the level of the top dielectric layer, (where the dipole away from the bottom ground plane, for antenna will be located), a plane wave normally incident as shown in Fig. 7(a).

number of dogbones constituting the surface is 30 (5 6 unit cells) which gives a total surface area of . The presence of an additional dielectric layer (1.61 mm thick) covering the dogbone metallic array, which is required to support the dipole is taken into account in the design and simulations. The principle of operation of HISs can be understood from the interpretation of the following two important characteristics. i) The phase of the reection coefcient (Fig. 8) that the surface presents at normal incidence, expressed as (1) ii) The value of normalized surface reactance versus frequency, dened as . The magnitude of the reection coefcient is related to the dielectric and conductor losses, and can be signicantly smaller than unity for certain patterned conductor geometries and for thin substrates. However, for the present application, the HIS conguration involves minimal losses and . Therefore, either the phase of or can be used to characterize the HIS response. The behavior of these types of surfaces has been analyzed at oblique incidence [5] and different incident polarizations [7]. For the application considered here, we will limit our analysis to the case of normal incidence and linear polarization. Fig. 8 shows the phase of the reection coefcient for this surface (assumed with innite transverse extent) where the reference plane has been de-embedded up to the dielectric cover at to the point where the antenna will be located.

The so called magnetic resonance frequency [13] is very close to the frequency where the reection phase value reaches zero degrees and the surface behaves as an AMC. For this specic dogbone geometry, the magnetic frequency is designed to occur at . Below and above this frequency the reactance becomes inductive and capacitive, respectively. When the asymptote of the phase tends to 180 degrees, the surface substantially behaves as a PEC. The position of the magnetic frequency can be controlled through different design parameters, such as the substrate thickness and permittivity, and the shape of dogbone conductors, as described in [12], [13], [18]. Approximate formulas for its evaluation can be found in [13]. In the intermediate states between the AMC and PEC behaviors, the surface exhibits a reactance that is either capacitive or inductive, and near the resonance frequency presents a high impedance value. Therefore, the surface can be more generally referred to as a HIS [9]. As a result of this behavior, placing a dipole on top of a HIS can provide improved impedance matching, bandwidth enhancement, and increased forward directivity due to the total reection properties of the reactive surface (the reection coefcient has a unity magnitude, as previously pointed out). In addition, below the magnetic resonance the HIS exhibits an inductive reactance that can compensate for the capacitive reactance of a dipole below its natural resonant frequency. As a consequence, the resonance of the dipole can be achieved at a lower frequency than its resonant frequency in free space, resulting in some degree of antenna miniaturization, as stated in [9]. An important question arises. Which is the frequency range where a HIS can be more effectively used? Some authors argue that it could be advantageous in terms of antenna impedance matching to work in the region where the phase of the reection coefcient is 90 45 degrees (i.e., the inductive region) [7], [9]. The authors of [7] conclude with a parametric analysis where it is shown that the operative HIS bandwidth increases with the smoothness of the phase reection in the transition region. In general one could classify and rank the performance of any HIS by using this factor as a gure of merit. However, other authors have used different operation regimes beyond the magnetic frequency to realize low prole antennas with total thickness below [10], [15] without apparent pattern degradation and within the operative band. One of the most relevant conclusion of these studies is that the HIS operative bandwidth must be considered together with the antenna design and not as an independent property valid for all antenna cases. After all, [7], [10], [15] seem to agree that the operative bandwidth denition for HISs needs to consider radiation pattern bandwidth as well as impedance bandwidth characteristics simultaneously. IV. LOW PROFILE ANTENNA OVER IMPEDANCE SUBSTRATE A. Dipole over HIS A copper planar dipole is now placed on top of the HIS to compare with the scenarios previously presented. The length of the printed dipole antenna is assumed as a variable parameter, whereas the dipole width is xed to 1 mm
A

HIGH

VALLECCHI et al.: LOW PROFILE FULLY PLANAR FOLDED DIPOLE ANTENNA ON A HIGH IMPEDANCE SURFACE

55

Fig. 9. Planar dipole antenna placed at 1.61 mm above the HIS. The total thickness of the antenna (including the dipole, the copper thickness . and the dogbone substrate) is 3.22 mm

. Fig. 9 shows the dipole over the HIS, which is excited using an ideal lumped port source between its arms. The height of the dipole above the HIS including copper thickness (i.e., the dogbone metal layer) is 1.61 mm and thus, the total thickness of the antenna (including the dipole, the copper thickness and the dogbone grounded substrate) is 3.22 mm . For a practical and planar realization of the feed, a twin line with characteristic impedance of approximately 112 Ohm will be used (a lower resistance would require the use of very wide twin line conductors, whose spurious radiation could deteriorate the antenna radiation pattern, and/or a gap between the line conductors smaller than 0.2 mm which would be difcult to be reliably realized by simple lithographic techniques). Therefore, the dipole needs to exhibit a similar input impedance at its terminals to achieve a good impedance matching and low reection coefcient. The simulated input resistance and reactance of the dipole for several dipole lengths are shown in Fig. 10(a) and (b), respectively. The length of the dipole is expressed in terms of the wavelength in the dielectric substrate, , as , with varying over the values 0.48, 0.5, 0.52, 0.54. The dipole input resistance exhibits much lower values than 112 Ohm at the desired frequency of operation (5.5 GHz). It is however higher than the 8 Ohm of the regular dipole at the same distance from a PEC plane as seen in Section II. Increasing the dipole length, we can obtain slightly higher resistance values at the expense of a higher reactance, which makes the matching more critical. Different center locations of the dipole were also simulated to test whether this could have any inuence on the input resistance. In particular, instead that placing the dipole between two rows of dogbones, i.e. above a non-patterned portion of the HIS as in Fig. 9, the dipole aligned with one dogbone row has been simulated, with its center either between two adjacent dogbones, or right over a dogbone center. It was found that both of these alternative dipole center locations indeed yield a higher input resistance, but unfortunately also a steeper variation of the input reactance, so that the matching to the twin line impedance can be only accomplished over a very narrow frequency band. Therefore, if the overall antenna prole is not to be increased (and the target is to keep a low prole), another type of antenna able to provide a higher input resistance is desirable. It is well

Fig. 10. (a) Input resistance and (b) input reactance for several dipole lengths , , 0.5, 0.52, 0.54. The dipole is placed (1.61 mm) above the HIS, which is printed on a dielectric substrate with permittivity and .

known that folded dipole antennas resonate with an input resistance about four times larger than that of a regular dipole [22]. At the same time, a folded dipole presents a larger impedance bandwidth (17% simulated in free-space) than a single dipole (typically 10% in free-space) due to the fact that the folded arm helps to smooth the behavior of the input reactance with frequency. A larger resistance than in Fig. 10(a) could be by itself a good starting point for a low prole antenna to achieve good matching to a twin line. Therefore this design approach is developed next. B. Printed Folded Dipole Over HIS A folded dipole antenna with overall length of 17.25 mm , gap between arms of 1 mm , and strip width of 1 mm, as depicted in Fig. 11, has been designed to work at 5.5 GHz. The dipole is excited by a twin line with conductor width of 1 mm, gap between conductors of 0.2 mm, and a simulated characteristic impedance of 112 Ohm. Further miniaturization of the surface can be accomplished (cf. Section VI), though not attempted here in this rst fully planar realization. The folded dipole is placed on top of the designed HIS surface. Its reection coefcient is plotted in Fig. 12 and compared to that obtained when the dogbone array is removed (i.e., dipole on top of a 3.22 mm thick RT/Duroid-5880 dielectric substrate backed by a copper ground plane only). The folded dipole over the HIS fed by the twin line presents a good

56

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 11. A folded dipole antenna with length of 17.25 mm , gap , and strip width of 1 mm. The antenna is between arms of 1 mm 1.61 mm away from the HIS and is designed to behave as AMC at 5.5 GHz.

Fig. 13. (a) Wilkinson balun designed to transform the unbalance microstrip transmission line (port 1) to the balance twin lead transmission line (ports 2 and and 3). The structure is printed on RT/Duroid 5880 with . (b) Magnitude of the reection coefcient, common mode rejection and differential mode transmission of the proposed Wilkinson balun.

Fig. 12. Reection coefcients of the folded dipole with length over a copper ground plane (solid black line) and the HIS (dashed blue line). The . overall thickness for both structures is equal to

impedance matching condition at the desired frequency and a fractional bandwidth of about 6%. This value is higher than the 2.75% of the folded dipole at the same distance from a PEC plane as seen in Section II. If gain-bandwidth enhancement is the primary design objective, the choice of the nal antenna shape and overall stack up prole may be helped by the use of optimization algorithms as shown in [8]. C. Printed Balun Design When a balanced dipole antenna is fed by a twin line and directly connected to an unbalanced transmission line such as a coaxial cable, currents will ow on the coaxial outer conductor producing pattern distortion due to spurious radiation from the feed. A balance to unbalanced transformer or balun is typically used to mitigate this problem. A balun suppresses the common mode current in the dipole arms. Some types of balun also provide input/output impedance transformation. Many approaches for balun realizations are reported in the literature [23]. In this paper, due to the desired low-prole characteristic, a microstrip-to-coplanar strips transition consisting of a symmetric T-junction [24] followed by a

180 delay line [25] is printed on the same layer as the folded dipole, as shown in Fig. 13(a). This balun structure was shown to feature broadband, low reection coefcient. The ground plane for the balun is placed on the intermediate dogbone layer (Fig. 14). Though this balun is quite large, due to the low permittivity of the substrate (which, for simplicity, is the same of the dipole) and tends to affect the antenna radiation performance, it is needed here for antenna measurements via coaxial cables. Nevertheless, it can be considered as an external module rather than part of the antenna, and it could be realized in a more compact form using another type of substrate or different technology. In Fig. 13, the input port 1 is the microstrip unbalanced transmission line, with characteristic impedance , where the SMA connector will be soldered. At the output port, we have three separated conductors, and thus two different and not degenerated quasi-TEM modes can be supported. The even and odd mode of the coupled output microstrips represent these two physical modes that can travel along the line. Both these modes are considered in the simulation. The even mode has a characteristic impedance , whereas the common mode has a characteristic impedance . The power coming from port 1 is split into two different paths. The balun is designed to introduce 180 degrees of phase delay between the two microstrip line branches by adjusting their

VALLECCHI et al.: LOW PROFILE FULLY PLANAR FOLDED DIPOLE ANTENNA ON A HIGH IMPEDANCE SURFACE

57

Fig. 15. Simulated and measured reection coefcient at the antenna connector (see Fig. 14). Simulations are performed with CST Microwave Studio and IE3D, ). In IE3D the taking into account the nite thickness of metal layers (35 dielectric layers are assumed to be of innite extent, whereas both the cases of innite and nite-size ground plane have been simulated. The IE3D mesh is made by approximately 20 subdomain divisions per wavelength (at a frequency of 7 GHz).

transmission is close to 0 dB. Moreover, the operative bandwidth of the balun ( 36%) is much larger than the simulated impedance bandwidth of the folded dipole. V. SIMULATED AND MEASURED RESULTS OF DIPOLE WITH BALUN OVER HIS
A

FOLDED

Fig. 14. Fully planar realization of a low prole dipole antenna on top of a HIS. (a) Top metal layer with a folded dipole and a balun from the unbalanced microstrip feed line to the balanced twin line. (b) Intermediate metal layer (on the opposite side of the dipole substrate) showing the ground plane for the balun and the dogbone layer under the folded dipole. The dogbones are then located on top of a grounded substrate, as shown in (c), the cross-sectional view of the layer stackup.

lengths so that , where are the lengths of the two branches as in Fig. 13(a), and is the guided wavelength in the microstrip. By doing this, the propagation mode in the coupled output microstrips will be dominantly the odd mode, which is tightly bound to the pair of line conductors and can be easily transferred into the differential twin line mode after the ground plane in the intermediate layer of the dielectric stack is truncated. To reduce undesired reection and radiation losses at the 90-degree microstrip bends, optimal miters have been employed [26]. As a result, the good reection coefcient and common mode rejection characteristics shown in Fig. 13(b) have been obtained. Indeed, the common mode rejection is larger than 20 dB, while the differential mode

Once the folded dipole is connected to the balun, the design is complete, and ready for fabrication and measurements. For the fabrication of the prototype we have used RT/Duroid 5880 laminates, with and , and 35- -thick copper foils. The two faces of the antenna top layer, containing the dipole with its feed, on the upper side, and the ground of the dipole feed as well as the dogbone array, on the lower side, are illustrated in Fig. 14(a) and (b), respectively, whereas a sketch of the antenna cross-section is depicted in Fig. 14(c). The reection coefcient of the prototype was measured using an HP8510C vector network analyzer and is compared with simulation results in Fig. 15. Good agreement is observed between the numerical results obtained by using two completely different EM simulators, one based on the nite integration technique in time domain (CST Microwave Studio), and the other on the method of moments (IE3D from Zeland). The simulations with IE3D are carried out using an innite as well as a nite ground plane (in Fig. 15, dashed blue and red curves, respectively), whereas the dielectric layers are assumed to be of innite extent. Differently, the simulations with CST also take into account the niteness of the dielectric substrates. In both CST and IE3D, metal layers are modelled with their actual nite thickness. All simulations and measurements show that at the center frequency the magnitude of the reection coefcient is as low as 20 dB. Anyway, it is noted that the prediction by IE3D is closer to the measured data in terms of both the resonance frequency and reection coefcient value. The small frequency shift between simulations and measurements may be due to fabrication tolerances. 35 thick (1 oz) foil rolled copper (RT/Duroid 5880) was used and a 40 thick photoresist laminate was placed on top to make the board photosensitive. However, the space between the twin lines is 200

58

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 16. Measured reection coefcient response for the complete antenna structure including the HIS (Complete antenna), and the structure where the dogbone layer is replaced by a solid ground plane (STD GND plane).

and between contiguous dogbones is just 170 , and therefore is comparable with the thickness of the metal plus photoresist layer stack (75 ). This high aspect ratio entails poor control on the gaps prole producing over-etching and widening the spacing between dogbones. Indeed, as shown in [16], the magnetic resonance tends to increase slightly when the gap between dogbones is enlarged, which gives support to our hypothesis. Fig. 16 shows the measured magnitude of the reection coefcient for the complete antenna conguration including the HIS (Complete antenna), and the structure where the dogbone layer is replaced by a standard solid copper ground plane (STD GND plane). As apparent, the HIS contributes to signicantly improve impedance matching and bandwidth performance. Fig. 17 shows the normalized total eld radiation pattern in the principal and planes of the dipole simulated using IE3D (with the nite size ground plane) versus measurements performed in our anechoic chamber facility. The simulated peak gain for each specic plane is 7.1 dBi in the -plane, and 5.6 dBi in the -plane, including all losses in the feed network, whereas maximum gain values of 6.4 dBi and 5.3 dBi were measured at 5.6 GHz (i.e. the measured resonance frequency) in the plane and -plane, respectively. A greater gain is obtained in the plane because the maximum of radiation occurs at a slightly tilted angle with respect to the boresight direction. This small angle tilt in the -plane is likely due to the coupling between the balun and the dipole and it is also observed in measurements. The patterns show that radiation in the forward direction is considerably increased with respect to the free-space dipole (providing a maximum directivity of 2.14 dBi). The radiation efciency for this antenna was simulated to be 90% (IE3D), and measured to be 86% at 5.6 GHz by using the cylindrical Wheeler cap of diameter 90 mm and height 150 mm according to the measurement technique described in [27], [28]. VI. HIS SIZE MINIATURIZATION STUDY In order to study the effect of the HIS size on the antenna performance, simulation results relevant to the four structures shown in Fig. 18, differing for the number of unit cells the HIS is comprised of, namely 5 4, 5 2, 4 4 and 3 4 dogbone

Fig. 17. Simulated and measured normalized total realized gain radiation patterns: (a) - and (b) -planes. Both copolar and cross-polar components are shown.

Fig. 18. Different sizes of HIS columns and rows, respectively. (a) 5

, where and 4, (b) 5 2, (c) 4

denote dogbones 4, and (d) 3 4.

unit cells, were compared to those obtained for the antenna with the original HIS (Fig. 11), formed by 5 6 unit cells. First we will consider the case that the number of unit cells is reduced while retaining an innite ground plane underneath the dogbone layer. This case may apply when the antenna is placed on top of a large ground plane (e.g. when the antenna shares the ground plane with other system modules). For this case, the printed balun and all the other antenna dimensions are kept constant. Secondly, we will assume that the ground plane has the same nite size as the dogbone layer. In this case the balun structure could be larger than the relative ground, having the same size as the HIS, and therefore is not considered in the simulations, and the antenna is directly fed through the twin line. The presented results for the input resistance are referred to the folded dipole terminals by appropriate de-embedding. To our knowledge the HIS size reduction consideration has been only briey commented on [8], [10] with two different HIS cases in terms of reection coefcient. In this paper, a complete investigation

VALLECCHI et al.: LOW PROFILE FULLY PLANAR FOLDED DIPOLE ANTENNA ON A HIGH IMPEDANCE SURFACE

59

TABLE I RELATIVE BANDWIDTH AND PEAK GAINS FOR DIFFERENT HIS SIZES WITH INFINITE GROUND PLANE AT 5.5 GHz

Fig. 19. Simulated reection coefcient using Zeland IE3D for different HIS sizes.

Fig. 21. Reection coefcient of the folded dipole input impedance simulated with CST Microwave Studio for different HIS sizes. The reference impedance is 112 ohms (characteristic impedance twin line) due to the fact that the balun structure has been removed.

case, very small changes are observed in Fig. 19 for the bandwidth and resonant frequency, with respect to the 5 6 original HIS size. However, when the HIS size is further reduced, the resonant frequency shifts towards higher frequencies while the relative bandwidth reduces. The simulated relative bandwidth and peak gains for all considered congurations are summarized in Table I. For the HISs smaller than the original size, a slight reduction of the maximum gain is observed in both the E- and H-planes. With respect to the radiation pattern shape very little variation was observed between the different HIS sizes. This experiment shows that the impact of reducing the HIS size over the frequency and antenna gain is limited if a large ground plane is available. B. Finite Ground Plane We now consider the case when the ground plane has the same nite size as the dogbone layer. As mentioned above, in this case the balun structure is not considered in the simulations. The reection coefcient and input impedance de-embedded to the folded dipole terminals for variable HIS and ground plane sizes are shown in Fig. 21 and Fig. 22, respectively. In this case, the antenna impedance matching varies signicantly with the HIS size, contrarily to the innite ground plane case. More precisely, when the size of the HIS is reduced from the original 5 6 unit cells to 5 4 and 5 2 unit cells, the reection coefcient changes slightly and the dipole remains impedance matched to the twin line. Instead, the matching noticeably deteriorates when the HIS is reduced to 4 4 and 3 4 unit cells. Indeed, in these latter cases the dipole input resistance signicantly decreases, and not even by adjusting the dipole length the antenna can any longer be matched to the twin line impedance. In fact, varying the dipole length changes

Fig. 20. (a) Real and (b) imaginary parts of the folded dipole input impedance simulated with Zeland IE3D for different HIS sizes.

of impedance, reection coefcient, gain and radiation pattern is offered for multiple HIS sizes for nite and innite ground plane cases. A. Innite Ground Plane Fig. 19 shows the reection coefcient of the ve antenna congurations from Fig. 18 simulated under an innite ground plane assumption, while in Fig. 20 are plotted the corresponding real and imaginary parts of the input impedance. Table I summarizes the relative 10 dB impedance bandwidths and peak gains for all considered HIS sizes. For the 5 4 dogbone array

60

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 22. (a) Real and (b) imaginary parts of the folded dipole input impedance simulated with CST Microwave Studio for different HIS sizes. Fig. 23. Simulated normalized total gain radiation patterns 5.5 GHz of the antennas in Fig. 18 with nite ground plane: (a) - and (b) -planes for different HIS sizes. TABLE II RELATIVE BANDWIDTHS, PEAK GAINS AND RADIATION EFFICIENCIES, FOR DIFFERENT HIS SIZES WITH FINITE GROUND PLANE AT 5.5 GHz

the imaginary part of the input impedance and the frequency where it goes to zero (resonance), but it does not provide the larger values of the input resistance at the resonance required to impedance matching the antenna with the smaller HISs. On the other hand, due to the absence of the balun structure, the additional resonance shown in Fig. 20(b) around 4.5 GHz does not appear in Fig. 22(b). Next we examine the radiation performance of those nite size HIS antenna congurations exhibiting a good impedance matching and low reection coefcient at the desired operating frequency. Fig. 23 shows the radiation patterns simulated by CST at 5.5 GHz for the folded dipole with the 5 6, 5 4 and 5 2 HISs, assuming the relative ground planes to have the same nite size as the different dogbone layers above it. Some reduction of the gain is observed when the HIS size decreases and back radiation level increases. Having removed the balun structure, pattern tilting appears to be negligible. Peak gains, relative bandwidths, and radiation efciencies associated to the selected antenna congurations are summarized in Table II. Though for increasingly smaller HISs obviously the gain and radiation efciency progressively reduces, it can be noted that even the smallest HIS provides considerably improved performance both in terms of peak gain with respect a free-space dipole (2.14 dBi) and impedance bandwidth with respect to both congurations examined in Section II that include a PEC, PMC (Fig. 3), or a grounded dielectric slab (Fig. 5) as reecting surfaces.

As a nal remark, from the above simulation results we conclude that too small HISs tend to give input resistance values much lower than required to match the dipole to the twin line impedance ( 112 Ohms), and increasingly smaller gains. Practically, to achieve good impedance matching and gain performance, the HIS is required to extend all around the dipole for at least one row/column of unit cells. VII. CONCLUSIONS In this work, the advantages of using high impedance surfaces over other potential solutions for the design of low prole dipole antennas have been remarked. It has been shown that by using a HIS, a low prole antenna (in the direction), with directional pattern and higher impedance bandwidth than a dipole in front of a PEC can be achieved. We have presented simulation and measurement results of a fully planar realization of a low prole dipole antenna and

VALLECCHI et al.: LOW PROFILE FULLY PLANAR FOLDED DIPOLE ANTENNA ON A HIGH IMPEDANCE SURFACE

61

a comprehensive explanation of the HIS characteristics. The paper shows a working prototype and we show that the total surface area realized here could be miniaturized up to some extent. The measured magnitude of the reection coefcient ( 15 dB) and measured gain (6.4 dBi) conrm the good performance of our prototype and prove that the developed concept can be used to design antennas with miniaturized thickness in fully planar technology. This kind of low prole antennas also offer viable solutions for vehicle, naval, and aerospace platforms, for integrating a conformal efcient antenna system on top of pre-existing metal structure. ACKNOWLEDGMENT The authors acknowledge Rogers Corporation for providing them with the substrate material used to fabricate the antenna. They also thank Mentor Graphics Corporation (IE3D), Computer Simulation Technology (CST) and Ansoft HFSS for providing the simulation tools that were instrumental in the design process. Final thanks go to C. Guclu and J. Sloan, at the University of California Irvine, for their valuable input and discussions. REFERENCES
[1] D. Sievenpiper, L. Zhang, R. F. Jimene Broas, N. G. Alexopolous, and E. Yablonovitch, High-impedance electromagnetic surfaces with a forbidden frequency band, IEEE Trans. Microwave Theory Tech., vol. 47, no. 11, pp. 20592074, Nov. 1999. [2] N. G. Alexopoulos, H. Contopanagos, and C. Kyriazidou, Method of Generating a Magnetic Interface, U.S. Patent 7109947B2, 2000. [3] N. G. Alexopoulos, H. Contopanagos, and C. Kyriazidou, Inductor Circuit With a Magnetic Interface, U.S. Patent 7116202B2, 2000. [4] N. G. Alexopoulos, F. De Flaviis, and Y. Liu, KCA elements in electromagnetically metamorphic objects and interfaces, in Proc. Union Radio-Scientique Internationale (URSI) Symp. on Electromagn. Theory, Pisa, Italy, 2004, pp. 688690. [5] C. R. Simovski, P. D. Maagt, and I. V. Melchakova, High-impedance surfaces having stable resonance with respect to polarization and incidence angle, IEEE Trans. Antennas Propag., vol. 53, no. 3, pp. 908914, Mar. 2005. [6] N. G. Alexopoulos, H. Contopanagos, and C. Kyriazidou, Antenna With a Magnetic Interface, U.S. Patent 6853350B2, 2000. [7] F. Yang and Y. Rahmat-Samii, Reection phase characterizations of the EBG ground plane for low prole wire antenna applications, IEEE Trans. Antennas Propag., vol. 51, no. 10, pp. 26912703, Oct. 2003. [8] L. Akhoondzadeh-Asl, D. J. Kern, P. S. Hall, and D. H. Werner, Wideband dipoles on electromagnetic bandgap ground planes, IEEE Trans. Antennas Propag., vol. 55, no. 9, pp. 24262434, Sept. 2007. [9] H. Mosallaei and K. Sarabandi, Antenna miniaturization and bandwidth enhancement using a reactive impedance substrate, IEEE Trans. Antennas Propag., vol. 52, no. 9, pp. 24032414, Sep. 2004. [10] S. R. Best and D. L. Hanna, Design of broadband dipole in close proximity to an EBG ground plane, IEEE Antennas Propag. Mag., vol. 50, no. 6, pp. 5264, 2008. [11] A. Vallecchi and F. Capolino, Thin high-impedance metamaterial substrate and its use in low prole antennas suitable for system integration, in Proc. 59th Electronic Components and Technology Conf. (ECTC), May 2009, pp. 777783. [12] A. Vallecchi, M. Albani, and F. Capolino, Planar metamaterial transverse equivalent network and its application to low-prole antenna designs, in Proc. Eur. Conf. on Antennas and Propagation (EUCAP), Berlin, Germany, Mar. 2327, 2009, pp. 861865. [13] G. Donzelli, A. Vallecchi, F. Capolino, and A. Schuchinsky, Anisotropic metamaterial made of paired planar conductors: Particle resonances, phenomena and properties, Metamaterials (Elsevier), vol. 3, no. 1, pp. 1027, 2009. [14] F. Yang, V. Demir, D. A. Elsherbeni, A. Z. Elsherbeni, and A. A. Eldek, Planar dipole antennas near the edge of an EBG ground plane, in Proc. Antennas and Propag Symp., Jul. 2005, vol. 1A, pp. 750753.

[15] M. Z. Azad and M. Ali, Novel wideband directional dipole antenna on a mushroom like EBG structure, IEEE Trans. Antennas Propag., vol. 56, no. 5, May. 2008. [16] A. Vallecchi, F. Capolino, and A. Schuchinsky, 2-D Isotropic effective negative refractive index metamaterial in planar technology, IEEE Microw. Wireless Comp. Lett., vol. 19, no. 5, pp. 269271, 2009. [17] A. Vallecchi and F. Capolino, Metamaterials based on pairs of tightlycoupled scatterers, in Theory and Phenomena of Metamaterials. Boca Raton, FL: CRC Press, 2009, ch. 19. [18] A. Vallecchi and F. Capolino, Tightly coupled tripole conductor pairs as constituents for a planar 2D-isotropic negative refractive index metamaterial, Opt. Express, vol. 17, no. 17, pp. 1521615227, 2009. [19] A. Vallecchi, F. Capolino, J. De Luis, and F. De Flaviis, A low prole folded dipole antenna on a reactive high impedance substrate, presented at the Int. Conf. on Electromagnetics in Advanced Applications (ICEAA 2009), Torino, Italy, Sep. 1418, 2009. [20] A. Vallecchi, F. Capolino, and M. Albani, Metamaterial made of pairs of conductors in planar technology: A Z-transmission line approach, presented at the Metamaterials Conf., Pamplona, Spain, Sep. 2126, 2008. [21] S. R Best, Improving the performance properties of a dipole element closely spaced to a PEC ground plane, IEEE Antennas Wireless Propag. Lett., vol. 3, 2004. [22] G. A. Thiele, E. P. Ekelman, Jr, and L. W. Henderson, On the accuracy of the transmission line model for the folded dipole, IEEE Trans. Antennas Propag., vol. 28, no. 5, pp. 700703, Sep. 1980. [23] J. S. Lim, H. S. Yang, Y. T. Lee, S. Kim, K. S. Seo, and S. Nam, E-Band Wilkinson balun using CPW MMIC technology, Electron. Lett., vol. 40, no. 14, Jul. 2004. [24] E. J. Wilkinson, An N-way hybrid power divider, IRE Trans. Microw. Theory Technol., vol. MTT-8, pp. 116118, 1960. [25] Y. Qian and T. Itoh, A broadband microstrip-to-CPS transition, in Asia Pacic Microwave Conf. Dig., Hong Kong, 1997, pp. 609612. [26] R. J. P. Douville and D. S. James, Experimental study of symmetric microstrip bends and their compensation, IEEE Trans. Microw. Theory Tech., vol. MTT-26, pp. 175181, Mar. 1978. [27] R. H. Johnston and J. G. McRory, An improved small antenna radiation-efciency measurement method, IEEE Antennas Propag. Mag., vol. 40, pp. 4048, 1998. [28] C. Mendes and C. Peixeiro, Radiation efciency of several handset antennas obtained with a modied wheeler cap method, in Proc. 2nd Eur. Conf. on Antennas and Propagation, EuCAP, 2007, pp. 14.

Andrea Vallecchi received the Laurea (M.Sc) degree (summa cum laude) in electronic engineering from the University of Florence, Florence, Italy, and the Ph.D. degree in information engineering, applied electromagnetics, and telecommunications from the University of Salerno, Salerno, Italy. He then worked at the Laboratory of Antennas and Microwaves. University of Florence, as a Research Associate, and subsequently, since 2007, has been a Postdoctoral Research Fellow at the University of Siena. In 2009, he spent some months working as an Assistant Specialist at the University of California, Irvine. In 2009 and 2010, he was a Visiting Researcher at Queens University of Belfast, Belfast, U.K. He is currently with the Department of Information Engineering, University of Siena. His present research interests are in the theoretical characterization and design of metamaterials for applications at microwaves and visible frequencies.

Javier R. De Luis (S08M11) was born in Elche, Spain, in 1982. He received the Telecommunication Engineer degree from the Universitat Politcnica de Catalunya (UPC), Barcelona, Spain and the M.S. degree from the University of California at Irvine (UCI) in 2008, where he is currently working toward the Ph.D. degree. His current research is focused on recongurable antennas for tunable RF handset front end and smart systems.

62

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Filippo Capolino (S94M97SM04) received the Laurea (cum laude) and Ph.D. degrees in electrical engineering from the University of Florence, Italy, in 1993 and 1997, respectively. He is presently employed as an Assistant Professor at the Department of Electrical Engineering and Computer Science, University of California, Irvine. He has been an Assistant Professor at the Department of Information Engineering, University of Siena, Italy. During 19971999, he was a Postdoctoral Fellow with the Department of Aerospace and Mech. Engineering, Boston University, MA. From 2000 to 2001 and in 2006, he was a Research Assistant Visiting Professor with the Department of Electrical and Comp. Engineering, University of Houston, TX. His research interests include antennas, metamaterials and their applications, sensors in both microwave and optical ranges, wireless systems, chip-integrated antennas. He has been the EU Coordinator of the EU Doctoral Programs on Metamaterials (20042009). Dr. Capolino received several Young and Senior Scientist Travel Grants to attend international conferences (IEEE and URSI) and two Student and Young Scientist Paper Competition Awards. In 2000, he received the R.W. P. King Prize Paper Award from the IEEE Antennas and Propagation Society for the Best Paper by an author under 36. He is a coauthor of Fast Breaking Papers, Oct. 2007 in EE and CS, about metamaterials (paper that had the highest percentage increase in citations in Essential Science Indicators, ESI). In 20022008, he served as an Associate Editor for the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION. Since 2007, he is a founder and an Editor of the new Elsevier journal Metamaterials. He is the Editor of the Metamaterials Handbook, CRC-Press, 2009.

Franco De Flaviis (SM07) was born in Teramo, Italy, in 1963. He received the Laurea degree in electronics engineering from the University of Ancona (Italy), in 1990, and the M.S. and Ph.D. degrees in electrical engineering from the University of California at Los Angeles (UCLA), in 1994 and 1997, respectively. In 1991, he was an Engineer at Alcatel, where his research specialized in the area of microwave mixer design. In 1992, he was Visiting Researcher at UCLA working on low intermodulation mixers. He is currently a Professor with the Department of Electrical Engineering and Computer Science, University of California, Irvine. He has authored and coauthored over 100 papers in refereed journals and conference proceedings, led several international patents and authored one book and three book chapters. His research interests include the development of microelectromechanical systems (MEMS) for RF applications fabricated on unconventional substrates, such as printed circuit board and microwave laminates with particular emphasis on recongurable antenna systems. He is also active in the research eld of highly integrated packaging for RF and wireless applications. Dr. De Flaviis is a member of URSI Commission B.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

63

Crumpling of PIFA Textile Antenna

Qiang Bai and Richard Langley, Member, IEEE
AbstractThe performance of a PIFA textile antenna under different crumpling conditions is presented. The PIFA was designed for integration into clothing and other textile applications and the bending and crumpling conditions studied are typical of those found under normal use conditions both on and off body. Input impedance, efciency and radiation patterns are investigated based on numerical and experimental methods at 2.4 GHz. Crumpling can have a serious effect on the resonant frequency, bandwidth and radiation from textile antennas. Index TermsBody worn antenna, crumpled antenna, PIFA antenna, textile antenna.

I. INTRODUCTION EARABLE antennas for wireless communication systems have attracted more and more attention in recent years [1][7]. Applications for textile and body worn systems range include emergency service radio at 400 MHz, mobile telephone bands from 800 MHz to 2200 MHz and wireless networks at the 2.4 GHz and 5 GHz bands. A exible textile substrate material and planar structure are normally used by designers in order to seamlessly integrate antennas into a garment. In [1][7], wearable antennas for GPS, WLAN and Bluetooth applications were studied including one for MIMO systems [8]. One of the main challenges of wearable antenna design is the uncertain form of the garment surface because it is difcult to keep the garment and consequently the wearable antennas at at all times, especially for elements made of textile materials. However there are many applications other than body worn systems where antennas will be located on textiles and other exible items such as parachutes, tents, luggage, etc. where the general properties of antenna crumpling will be a potential problem. It is therefore necessary to evaluate the textile antennas performance under bending and crumpling conditions. Previous research has concentrated on wearable antenna bending effects [9][12] rather than the more complex issue of crumpling. In [13] and [14] the authors briey described the performance of a coplanar patch antenna alone and over a high impedance surface when crumpled in one plane. In this paper the results of a comprehensive study of the performance of a different crumpled antenna, a PIFA antenna, manufactured on felt material and operating at the 2.4 GHz ISM band are presented. The antenna performance under realistic and quite severe crumpling conditions
Manuscript received March 03, 2011; revised May 20, 2011; accepted June 21, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the Electronic and Electrical Engineering Department, University of Shefeld, Shefeld S1 3JD, U.K. (e-mail: Q.Bai@shefeld.ac.uk; R.J.Langley@shefeld.ac.uk). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167944

Fig. 1. Geometry of PIFA antenna.

is studied based on both simulation and measurement results. Simulations were carried out in CST Microwave Studio, which is a 3D full eld solver implementing a nite integral technique (FIT). A network analyzer was used to measure the reection coefcients while the antenna radiation patterns were measured in a fully anechoic chamber at the University. In Section II the geometry of the PIFA is described and the general performance of the crumpled PIFA antenna is presented in Section III and this is followed by the performance when the antenna is close to the human body in Section IV. II. ANTENNA STRUCTURE The antenna and ground plane, Fig. 1, were manufactured by cutting them from Zelt conducting material and attaching the components to a layer of thin felt material 1.1 mm thick with using a very thin layer of adhesive. and The material Zelt is a high quality nylon based substrate plated S/m. Its with copper and tin with a conductivity of thickness is 0.06 mm with a manufacturers surface resistivity specication lower than 0.01 ohm/square. The planar inverted-f antenna (PIFA) shown in Fig. 1 was 28.5 mm long fed by a 50 ohm microstrip line and at its left hand end there is a shorting pin connected to the ground on the opposite side of the substrate. The ground covers the area from the bottom of the microstrip line up to the dotted horizontal line shown in Fig. 1. The antenna input impedance can be adjusted by changing the separation between the shorting stub and the

0018-926X/$26.00 2011 IEEE

64

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Photo 1. Crumpling Rohacell formers.

Fig. 2. PIFA antenna crumpled in Y-Z plane.

microstrip line. The antenna dimensions measure 35 mm by 40 mm. More geometry details are presented in Fig. 1. III. ANTENNA PERFORMANCE UNDER CRUMPLING CONDITIONS IN FREE SPACE In practical situations for a textile antenna integrated into a textile or exible substrate or into clothing on the human body the antenna will be subject to crumpling. For example as a person takes up various positions a exible antenna will not only bend but also crumple particularly near the joints. A study of antenna performance under different crumpling conditions is therefore necessary. The crumpling was achieved experimentally by machining two complementary Rohacell formers between which the antennas were compressed to form the crumple. Several of these formers are shown in Photo 1. A. Antenna Crumpling in Y-Z Plane In a practical environment, many types of crumpling may take place. Fig. 2 presents three typical crumpling cases which are chosen for the following investigation for crumpling in the Y-Z plane of the antenna. In the rst case, the antenna aperture is reduced from 35 mm 40 mm when at to 25 mm 40 mm with the crumpling depth set at 5 mm. In Case 2 the crumpling depth was increased to 12 mm where the aperture further reduces to just 16 mm 40 mm. On the human body, Cases 1 and 2 are most likely to appear near the knee or elbow joints but fabric antennas for any application might crumple in this manner. However, from observation of various textile based antennas the crumpling illustrated in Case 3 is possible, where the crumpling peak-to-peak distance is increased to 24 mm and the crumpling depth is 20 mm with an aperture of 21 mm 40 mm. The dimensions of antenna for each crumpling case are described in Fig. 3 and Table I. Fig. 4 shows the simulated and measured reection coefcients. These are seen to be in very good agreement. Case 1 shows a shift in resonant frequency upwards of 40 MHz with the dB bandwidth halved while for Case 2 the reection coefcient diminishes to a maximum of about dB indicating a loss in match and radiated power. The reason may be the separation between the top radiating element and the ground is changed signicantly in this crumpling case as a relatively small peak-to-peak distance, mm, is used (Figs. 3 and
Fig. 3. Dimension of crumpled antenna in Y-Z plane.

TABLE I DIMENSIONS OF THE CRUMPLED ANTENNA IN Y-Z PLANE

5). In crumpling Case 3, the distance l is 24 mm and the antenna is bent almost 90 degrees, but the reection coefcient is only slightly changed with the resonance increasing by just 10 MHz. The current distributions on the PIFA for the antenna at and crumpling Case 2 are shown in Fig. 5. The distribution on the PIFA element is similar to the case when it is at but there is current excited on the ground plane due to the bending and giving rise to different radiation patterns and impedance match. The radiation patterns change under crumpling compared with the at antenna as plotted in Fig. 6 where measured and simulated patterns are compared for all three crumpling cases.

BAI AND LANGLEY: CRUMPLING OF PIFA TEXTILE ANTENNA

65

Fig. 4. Reection coefcient simulated and (b) measured.

for PIFA antenna crumpling in Y-Z plane (a)

Fig. 5. Simulated PIFA current distribution for (a) at and (b) Case 2.

The radiation patterns are normalized to the highest value of the at antenna. For the Y-Z measurement plane, Fig. 6(a), all three crumpled antennas show some distortions but primarily rotations when compared with the at PIFA radiation pattern, e.g., for Case 2 the pattern was rotated by 30 . The simulated radiation patterns in this plane, Fig. 6(a), show reasonable agreement with measurements particularly for the pattern rotation angles. For the X-Z plane, Fig. 6(b), the crumpling

Fig. 6. Measured and simulated co-polar radiation patterns of the PIFA antenna crumpled in Y-Z plane: (a) Y-Z plane pattern and (b) X-Z plane pattern.

reduces the radiated power levels for all three cases by up to 15 dB at some angles. The origin of the asymmetry for the

66

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE II DIMENSIONS OF THE CRUMPLED ANTENNA IN X-Z PLANE

Fig. 7. PIFA antenna crumpled in X-Z plane.

Fig. 8. Dimension of crumpled antenna in X-Z plane.

crumpled antenna measurements at about 120 is not clear. Simulation radiation patterns show power reductions in the radiation patterns for the different crumpling geometries of up to 8 dB compared to the at antenna although these are less than that found for the measurements. The Rohacell material used for the moulds which is 25 mm thick can give up to 2 dB loss for the measurements and this will be a contributing factor. B. Antenna Crumpling in X-Z Plane In X-Z plane, the antenna crumpling starts from the center of the micro-strip line (Fig. 7). As shown in Fig. 8, the antenna is also studied for three cases where the depths were the same as for the Y-Z plane crumpling. The antenna aperture is reduced from 35 40 mm to 35 26 mm in the rst crumpling Case 1, then further reduced to 35 16 mm in Case 2. As the crumpling depth increases to 20 mm in the third Case 3, the PIFA antenna is sharply bent with only 35 15 mm effective aperture and in this case there is effectively a single bend. The dimensions of antenna for each crumpling case are described in Table II. For this plane measurements were only possible for the case of the at antenna and Case 3 as the rigid coaxial SMA feed connector was 8 mm wide and it was not possible to crumple the antenna correctly. However the good agreement between simulation and measurement obtained for the Y-Z plane gives condence that the computed performance will be indicative of the effects of crumpling in practice for the other cases.
Fig. 9. Reection coefcient for PIFA antenna crumpled in X-Z plane. (a) Simulated, (b) Case 3 simulated and measurement.

The simulated reection coefcient of the X-Z plane crumpled antenna is presented in Fig. 9(a) where the antenna resonant frequencies increase in all three crumpling cases. The shifts are much more signicant in this plane than the Y-Z plane discussed earlier. For the antenna in Case 3 the antenna is bent by nearly 180 and the resonant frequency increases from 2.4 GHz for the at antenna to 2.58 GHz. For Case 1 the resonant frequency moves by 420 MHz from 2.4 GHz to 2.72 GHz while when the crumpling depth further increases to 12 mm, the resonance shifts to 2.92 GHz with a reduced maximum reection coefcient of dB. The measured results in Fig. 9(b) showed a shift in resonant frequency of just 200 MHz. The resonant frequency resulting from the crumpling is not related to the projected aperture as Cases 2 and 3 have the same aperture width but different resonant frequencies while Case 1 has the widest

BAI AND LANGLEY: CRUMPLING OF PIFA TEXTILE ANTENNA

67

Fig. 10. Simulated current distribution for Case 3.

aperture but a resonance between Cases 2 and 3. Overall crumpling of the PIFA antenna particularly in the X-Z plane degrades the reection coefcient and detunes the antenna so that the resonance peak shifts out of the ISM band. In Fig. 9(b) the simulated reection coefcient is compared with that measured for the antenna at and for crumpling Case 3. The resonant frequency for Case 3 was 2.58 GHz simulated and 2.55 GHz measured which is good agreement. The current distribution for Case 3 is plotted in Fig. 10 where it can be seen that this is very similar to the at case shown in Fig. 5(a) although the effective radiating aperture is two dimensional due to the bending. In addition the bending redistributes the current more into the 90 and 270 directions. Similar current distributions are found for the other two crumpling cases. The increase in resonant frequency is due to redistributed current across the aperture rather than the effective aperture. Fig. 11 shows the simulated radiation patterns for the X-Z plane crumpled PIFA antennas compared to the at antenna. In addition the measured radiation patterns for the at antenna and Case 3 are also plotted. The radiation patterns are normalized to the highest value for that group of results allowing the differences in levels for each individual case to be readily seen. The simulated radiation patterns in the X-Z measurement plane are virtually identical (not normalized) in all cases whether at or crumpled. In the Y-Z plane the radiation pattern levels are similar but the pattern was rotated by up to 40 compared to the at case for case 2, see Fig. 11(b). Measurements show some deviation from the simulations. This may be partly due to the fact that the antenna shape in Case 3 was not exactly the same as the simulations due to the rigidity of the connector. In Fig. 11(a) the crumpled antenna radiation patterns are slightly asymmetric at 90 compared to 270 . In fact due to the mould used there was about 1 cm more Rohacell at 90 than at 270 and this may be one reason why the power level at 90 is lower than 270 . In Fig. 11(b) the measured and simulated patterns are in better agreement showing that crum-

Fig. 11. Simulated and measured radiation patterns of PIFA crumpled in X-Z plane. Measurements for at and case 3 only: (a) X-Z plane pattern and (b) Y-Z plane pattern.

pling the antenna squints the pattern and reduces the forward radiation levels.

68

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE III SIMULATED ANTENNA EFFICIENCY

The gain of the at PIFA antenna was 1.47 dBi and the radiation patterns plotted in the two planes show the relative power levels with respect to the at antennas. It was not possible to measure the antennas 3D radiation patterns but using CST the radiation and total (overall) efciencies of the antennas were computed and are presented in Table III in free space and on body, the latter will be discussed later. These efciencies include conductive and dielectric losses. The total efciency included the reection coefcient as the crumpling changed the impedance. The free space PIFA antenna efciencies when at were both 83%. However Table III shows how both the radiation and total efciencies vary as the antenna was crumpled along each plane. For the Y-Z crumpling plane Case 2 shows a signicant dip in the radiation efciency due to the shape of the antenna and hence current distribution, down from 83% for the at case to just 51%. The other two cases retain good radiation efciency in this plane. Case 2 also had a poor reection coefcient again reected in the total efciency of just 34%. For the X-Z crumpling plane the radiation efciencies remain high but the shift in the resonant frequency in this plane reduces the overall efciencies in each case, Case 2 suffering most with a total efciency of just 26%. The authors have conducted limited crumpling studies on two other antenna geometries at 2.4 GHz, a coplanar patch antenna and a microstrip patch antenna. Under similar crumpling conditions measurements on both antennas produced frequency shifts similar to those presented here for the PIFA. IV. ANTENNA PERFORMANCE UNDER CRUMPLING CONDITIONS ON BODY The previous section described the performance of a crumpled antenna in free space. Measured and simulated results

Fig. 12. Simulated reection coefcients of crumpled PIFA antenna on body. (a) Crumpled in Y-Z plane. (b) Crumpled in X-Z plane.

were in good agreement giving condence in the use of CST Microwave Studio to simulate these conditions. Although this paper deals with the general problem of the crumpling of exible antennas for many applications body worn systems are important. The human body will have an important effect on the PIFA antenna and in this section the crumpled antennas were simulated when placed on a simple single layer human tissue model 40 mm deep, which has a permittivity of 56.575 and a conductivity of 0.8166 S/m. The antenna was spaced 5 mm above the surface of the body. The antenna reection coefcients, radiation patterns and efciencies were recalculated for the same three cases of crumpling as in Section III. Fig. 12 shows the reection coefcients for the antenna on body crumpled in the Y-Z and X-Z planes. Comparing them with Figs. 4 and 9 for the antenna in free space, in the Y-Z crumpling plane the depth of the nulls for the at case and Case 3 are reduced but they remain well matched, for the other two cases, 1 and 2, the reection coefcients are reduced by 2 dB to a best value of dB and dB respectively. There was a slight shift downward in the resonant frequency of around 2030 MHz in all four cases. For the X-Z crumpling plane apart from the at case crumpling leads to an improved reection coefcient to about dB at resonance but a downwards shift in frequency of between 40 MHz and 130 MHz resulted moving the resonance outside the ISM band. Measurements for the antenna in free space showed a smaller shift in the resonant frequency than the simulations. The PIFA was situated just 5 mm from the body

BAI AND LANGLEY: CRUMPLING OF PIFA TEXTILE ANTENNA

69

Fig. 13. Simulated co-polar radiation patterns of the PIFA antenna crumpled in Y-Z plane Case 2: (a) Y-Z plane pattern and (b) X-Z plane pattern.

Fig. 14. Simulated radiation patterns of PIFA crumpled in X-Z plane for Case 1: (a) X-Z plane pattern and (b) Y-Z plane pattern.

at its nearest point in this study and increasing the distance of the antenna from the body generally improved the reection coefcient but had little effect on the resonant frequency as the physical distortion introduced by the crumpling was the predominant effect as demonstrated in Section III. Principal plane radiation patterns are compared for the free space and on body situations in Fig. 13 for crumpling in the Y-Z plane Case 2 and in Fig. 14 for crumpling in the X-Z plane Case 1. The radiation patterns are normalized to the highest value of the two plots. In both Figures the body absorbs the back radiation while also increasing the forward radiation by about 3 dB for Case 2, Y-Z crumpling plane Fig. 13, and 0.4 dB for Case 1, X-Z crumpling plane Fig. 14. The change in forward gain was very dependent on the exact crumpling pattern. The radiation efciency fell from 51% to 30% and 79% to 26% respectively. Simulations in all cases of crumpling generally show an increase in the forward radiation with a signicant amount of absorption from the body in the backward direction. Table III summarizes the free space and on body efciencies for all cases. For the at antenna the efciency falls from 83%

to just 16% on body due to absorption by the body and there was also a corresponding loss in the forward gain of 1.4 dB. However once the antenna was crumpled in most cases the forward gain improved due to different interaction with the body and again there was signicant absorption of the back radiation. Radiation efciencies varied considerably on body from 68% at best (Case 3 Y-Z plane) to 26% (Case 2 X-Z plane) the latter still being 10% better than the at PIFA indicating that crumpling actually helps the overall radiation efciency. Taking the impedance mismatches into account for the total efciency further signicant reductions in efciency are found. V. CONCLUSION The performance of a exible PIFA antenna operating at 2.4 GHz and manufactured on the textile material felt was investigated under crumpled conditions likely to be encountered when the antenna is integrated within textile materials either on or off body. Crumpling was investigated in two perpendicular planes for three cases of crumpling in each plane. In each case the

70

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

crumpling was signicant. Simulation results were compared with measurements where practical. Overall the simulated results were veried by measurements although the measurements showed a smaller shift in the resonant frequency in the X-Z plane than the simulations. In all cases in free space the reection coefcient and hence impedance of the antenna changed when crumpled, in one case the value was reduced to just dB and in other cases a signicant shift in resonant frequency resulted, the resonant frequency shifting out of the designed ISM band. Radiation patterns showed squinting and a loss of radiated power in some directions. The efciencies of the antennas varied considerably with the type and plane of crumpling, both in terms of radiation efciency which suffered in some cases and particularly the shift in resonant frequency reducing the overall efciency to 26% in one case. Placing the antenna onto a simple human body model resulted in signicant falls in efciency for all cases especially the at PIFA, absorption of the backward radiation by the body being the problem. However crumpling increased the overall forward gain as there was less interaction with the body but overall antenna efciency levels were poor at 68% at best and just 16% at worst. The input impedance match improved for X-Z plane crumpling but there were signicant shifts in frequency of up to 130 MHz. The frequency shift was primarily caused by the crumpling of the antenna, reducing the interaction with the human body had little effect. Similar results were found for other antenna geometries under similar crumpling conditions. In general the performance of a textile antenna when crumpled can be substantially degraded with signicant shifts in frequency and radiation performance. REFERENCES
[1] Antennas and Propagation for Body-Centric Wireless Communications, P. S. Hall and Y. Hao, Eds.. Norwood, MA: Artech House, 2006. [2] S. Zhu and R. J. Langley, Dual-band wearable textile antennas on an EBG substrate, IEEE Trans. Antennas Propag., vol. 57, no. 4, pt. 1, pp. 926935, Apr. 2009. [3] L. Liu, S. Zhu, and R. J. Langley, Dual-band triangular patch antenna with modied ground plane, IET Electron. Lett., vol. 43, pp. 140141, Feb. 2007. [4] M. Tanaka and J. Jang, Wearable microstrip antenna, in Proc. IEEE AP-S, 2003, vol. 2, pp. 704707. [5] S. L. Cotton and W. G. Scanlon, Measurements, modeling and simulation of the off-body radio channel for the implementation of bodyworn antenna diversity at 868 MHz, IEEE Trans. Antennas Propag., vol. 57, no. 12, pp. 39513961, Dec. 2009. [6] P. Salonen, Y. Rahmat-Samii, M. Schaffrath, and M. Kivikoski, Effect of textile materials on wearable antenna performance: A case study of GPS antennas, in Proc. IEEE AP-S, 2004, vol. 1, pp. 459462. [7] T. F. Kennedy, P. W. Fink, A. W. Chu, N. J. Champagne, G. Y. Lin, and M. A. Khayat, Body-worn E-textile antennas: The good, the lowmass, and the conformal, IEEE Trans. Antennas Propag., vol. 57, no. 4, pt. 1, pp. 910918, Apr. 2009.

[8] Y. Ouyang, D. J. Love, and W. J. Chappell, Body-worn distributed MIMO system, IEEE Trans. Veh. Technol., vol. 58, no. 4, pp. 17521765, May 2009. [9] P. Salonen and Y. Rahmat-Samii, Textile antennas: Effects of antenna bending on input matching and impedance bandwidth, IEEE Aerosp. Electron. Syst. Mag., vol. 22, no. 12, pp. 1822, Dec. 2007. [10] I. Locher, M. Klemm, T. Kirstein, and G. Troster, Design and characterization of purely textile patch antennas, IEEE Trans. Adv. Packag., vol. 29, no. 4, pp. 777788, Nov. 2006. [11] C. Hertleer, H. Rogier, L. Vallozzi, and L. Van Langenhove, A textile antenna for off-body communication integrated into protective clothing for reghters, IEEE Trans. Antennas Propag., vol. 57, no. 4, pt. 1, pp. 919925, 2009. [12] Y. Bayram, Y. Zhou, B. S. Shim, S. Xu, J. Zhu, N. A. Kotov, and J. L. Volakis, E-textile conductors and polymer composites for conformal lightweight antennas, IEEE Trans. Antennas Propag., vol. 58, no. 8, pp. 27302736, 2010. [13] Q. Bai and R. Langley, Crumpled textile antennas, Electron. Lett., vol. 45, no. 9, pp. 436438, Apr. 23, 2009. [14] Q. Bai and R. Langley, Crumpled integrated AMC antenna, Electron. Lett., vol. 45, no. 13, pp. 662663, Jun. 18, 2009.

Qiang Bai received the M.S. degree in electronics engineering from the University of Kent, Canterbury, U.K., in 2006 and the M.Phil. degree in body centric communications from the University of Birmingham, Birmingham, U.K., in 2009. He is currently working toward the Ph.D. degree at the University of Shefeld, Shefeld, UK. His research interests include on-body communications and textile antenna for wearable applications, electromagnetic bandgap structures and compact UWB antennas.

Richard Langley (M79) received the B.Sc. and Ph.D. degrees from the University of Kent, Canterbury, U.K. After spending some time working on communications satellites at Marconi Space Systems in the 1970s he became a Lecturer at the University of Kent in 1979. He was promoted to a personal Chair in Antenna Systems in 1994. In 1997, he founded the European Technology Centre for Harada Industries Japan, the worlds largest supplier of automotive antennas. The Centre researches and develops advanced hidden antenna systems for the global automotive market including radio, telephone and navigation systems. After successfully building up the technology and business he rejoined academic life in 2003. He is now Head of the Communications Research Group at the University of Shefeld. His main research is in the elds of automotive antennas, propagation in the built environment, frequency selective surfaces, electromagnetic band gap materials and applications, multi-function antenna systems and recongurable antennas. He has published over 250 papers in international journals and conferences. Prof. Langley was Honorary Editor of the Institute of Electrical Engineering ProceedingsMicrowaves, Antennas and Propagation from 19952003. He initiated the setting up of the Wireless Friendly Building Forum in 2009 to address the problems of wireless signal propagation in buildings and the built environment. He is a past Chair of the IET Antennas and Propagation Professional Network.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

71

Higher Order Mode Excitation for High-Gain Broadside Radiation From Cylindrical Dielectric Resonator Antennas
Debatosh Guha, Senior Member, IEEE, Archita Banerjee, Student Member, IEEE, Chandrakanta Kumar, Member, IEEE, and Yahia M. M. Antar, Fellow, IEEE
AbstractA resonant mode ( ), other than those ( and ) conventionally excited and used in a cylindrical dielectric resonator antenna (CDRA) has been examined with a view for using it as another radiating mode with broadside radiation patterns. Excitation of the mode, being the most challenging aspect, has been discussed and resolved by employing an innovative technique. The proposed concept has been successfully veried and experimentally demonstrated for the rst time. More than 8-dBi peak gain with excellent broadside radiation has been obtained from a prototype shaped from a commercially available low-loss dielectric material with relative permittivity 10. Index TermsCylindrical dielectric resonator antenna, dielectric resonator antenna, radiation modes in cylindrical dielectric resonators.
Fig. 1. Electric-eld distributions due to the rst few resonant modes in an isomode. (b) lated CDR [3]: Top and Cross-sectional views. (a) mode. (c) mode.

I. INTRODUCTION YLINDICAL dielectric resonators (DR) were rst reported as antennas by Long et al. [1] and since then, cylindrical dielectric resonator antennas (CDRAs) are probably the most celebrated ones in terms of investigations and applications. Long [1] rst used the hybrid electromagnetic mode having the lowest resonant frequency to excite a CDRA placed on a at ground plane and fed by a vertical coaxial probe. The broadside radiation patterns in either principal plane were similar to those due to a circular or a rectangular microstrip patch excited with their respective dominant mode. Apart from , another mode in a cylindrical DR results in radiation. This is the mode having symmetrically oriented monopole-like radiated elds around the DRA axis [2]. First, a few modes in an isolated cylindrical DR, their eld distributions, and properties were studied in [3] and [4]. A cylinManuscript received November 16, 2010; revised March 06, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported by the University of Calcutta under RFSMS/ UGC Grant. D. Guha and A. Banerjee are with the Institute of Radio Physics and Electronics, University of Calcutta, Kolkata 700 009, India (e-mail: dguha@ieee. org; archita@ieee.org). C. Kumar is with the Department of Space, Government of India, Communication Systems Group, Bangalore 560017, India (e-mail: kumarchk@ieee.org). Y. M. M. Antar is with the Electrical Engineering Department, Royal Military College of Canada, Kingston, ON K7K 7B4 Canada (e-mail: antar-y@rmc.ca). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167922

drical DR mounted on grounded substrate was analyzed in [5]. Modes in bisected or split CDRA were also theoretically examined [6]. In a conventional CDRA, the mode is of lowest ( for [4]). The value for the mode is quite high ( [4]), although the mode is of relatively low ( [4]). In this consideration, the mode should be equally good for radiation. Its eld congurations [Fig. 1(c)] also promise excellent linearly polarized radiation along the broadside, as does the mode. But interestingly, this mode is not common in the DRA community. Very recently, the present authors have proposed an idea to excite a CDRA with the mode [7]. This is systematically investigated in this paper in a more elaborate way and experimentally studied. The most challenging part of this study, that is, the excitation of the mode has been discussed and resolved using an innovative technique. A prototype has been physically realized. Successful excitation of the mode with high gain broadside radiations has been experimentally demonstrated for the rst time. II. PROPOSED TECHNIQUE TO EXCITE THE A. Concept Fig. 1 shows the electric-eld distributions for three resonant modes, as discussed here for an isolated CDR. We know that a metallic ground plane is needed for a practical DRA to hold and support the feed structure. Now if one examines the electric elds in Fig. 1, it would be evident in Fig. 1(a) and (b) that a metal boundary can be placed across the dotted line GG without affecting the modal eld distributions. The electric plane of symmetry across GG actually supports an electric image. This makes its effective height twice the physical height. This enMODE

0018-926X/$26.00 2011 IEEE

72

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Schematic diagram of a CDRA symmetrically placed on a nonresonant microstrip patch (NMP), fed by a coaxial SMA probe: Top and cross-sectional views.

Fig. 3. Simulated versus frequency response of a CDR excited using an (7.23 GHz). NMP: NMP (Fig. 2) and simulated electric elds at minimum 2.33, 1.575 mm, 40 mm, 5 mm, 2 mm, probe 1.3 mm; DRA: 38, 5.25 mm, 4.6 mm.

ables these two modes ( and ) to be excited in a CDRA by placing them on a metal sheet or a ground plane. Now let us consider Fig. 1(c). Like the aforementioned discussion, if a metal sheet or electric boundary is introduced across KK , the horizontal electric elds around KK vanish immediately, destroying the necessary boundary condition to excite the mode. Hence, a conventional ground plane is to be avoided for exciting the mode. To satisfy these horizontal eld components undisturbed, we have intuitively conceived a linearly polarized current sheet in place of a ground plane, which can excite the required electric elds. This means that a CDRA would be placed on a current sheet, not on a ground plane.

C. Verication For verication purposes, a previously studied CDRA element [3], [4] has been chosen and its theoretical data have been employed for comparison. The whole structure, shown in Fig. 2, is simulated using [8] and the parameters along with the optimum dimensions of the NMP are provided in Fig. 3. The quantities , , and determine the resonant frequency of an isolated DRA, which, in the present feeding mechanism, is inuenced by the feed parameters, such as , , , and . The parameter has no direct effect on the excited modes in the DRA. Simulated characteristics, shown in Fig. 3, indicate one prominent resonance around 7.23 GHz with 22 dB. This is compared with the theoretically determined value [4] in Table I. To earn condence in our comparison, two more known modes have been simulated and compared in Table I. The theory [4] considered isolated resonators, but in here, a metal probe and a ground plane are incorporated, resulting in deviations from the theory. This causes deviation of simulated data from the theoretical values. Resonances for and modes shift by about 9% with reference to the theoretical predictions. But that for the mode ( ) shows a relative deviation by only 1.6%. Indeed, to excite the mode, the probe is placed by just touching the CDRA surface, but not protruding through the DRA body. The feed mechanism for the or the mode perturbs the ideal isolated conguration [4] and, thus, results in deviation from the theoretical values [4]. Modal electric elds due to the newly excited mode are also examined in Fig. 3. Simulated eld portrays, viewed from the top and the side, are used. They closely resemble the schematic

B. Proposed Feed Conguration To verify our concept, we have chosen a CDRA sample, which was theoretically analyzed in [3] and [4]. The current sheet has been realized using a nonresonant microstrip patch (NMP). The patch may be etched on a thin PTFE substrate. Initially, we are restricted to the circular shape of the patch which is activated by a conventional SMA probe as shown in Fig. 2. Other geometries of current sheets involving purely planar circuits are under investigation. The dimensions of the feed structure (NMP tted with SMA probe) need optimization and those for the present study have been obtained by using [8]. The parameters are presented in the caption of Fig. 3.

GUHA et al.: HIGHER ORDER MODE EXCITATION FOR HIGH-GAIN BROADSIDE RADIATION FROM CYLINDRICAL DRAs

73

TABLE I RESONANT FREQUENCIES OF THE FIRST FEW MODES IN A CYLINDRICAL DRA PARAMETERS AS IN FIG. 3

Fig. 5. Measured and simulated ters are shown as in Fig. 4.

versus frequency responses. The parame-

. The value of can be calculated in terms of the DRA parameters using a closed-form expression, readily available in any DRA book [9], [10]. For the present prototype (parameters provided in Fig. 4), it was estimated as 5.7 GHz. The NMP has been designed accordingly, optimized for efcient excitation of the desired mode and, nally, etched on an RT Duroid 5870 microwave laminate.
Fig. 4. Photograph of the fabricated NMP used as feed and fabricated DRA. 2.33, 1.575 mm, 60 mm, 5 mm, 5 mm, NMP: 1.3 mm; DRA: 10, 10 mm, 10 mm.

B. Excitation and Resonances The NMP-fed CDRA (Fig. 2) has been tested for input using Agilents 8363B network analyzer. The measured data are compared with the simulated values in Fig. 5 for a wide range of frequency. Two resonances are apparent around 5.98 GHz and 7.4 GHz, respectively. The nature of the resonant modes has been identied with the help of simulated electric elds at respective frequencies. Electric-eld distributions at 7.38 GHz resemble the mode as shown in Fig. 6. Simulated portrays are documented as evidence. The top and side views are shown. In the colored diagram, the highest intensity is represented by red and the lowest by blue, and those in gray scale, appear as deep black and faint gray, respectively. The size of a eld vector also carries information about its magnitude. For clarity and better understanding, schematic diagrams representing respective cross-sectional views are also presented in Fig. 6. In Fig. 6(b), simulated elds are directed toward the left-hand side, which actually reveals elds near the dielectric boundary. In the cross-sectional view, the elds are directed toward the right-hand side. They indeed reveal the eld lying over the plane passing through the center of the DRA. The eld patterns due to resonance at 6.0 GHz look neither like the nor the mode. This is examined in Fig. 7, which appears very similar to the mode. However, strong elds near the boundary of the NMP are present, which inuences horizontal elds in the DRA and deviates from its ideal orientation. The remaining modal features resemble the mode and, thus, we prefer calling it the -like mode. This is further veried in the following section by studying the radiation characteristics due to this mode. This conjecture is veried using the results in Fig. 8. versus

diagrams shown in Fig. 1(c). But so far, its radiation characteristics are absolutely unknown. No previous work had addressed this, either theoretically or experimentally. The simulation study gives an idea, which appears very similar to those produced by the mode. III. PROTOTYPES AND EXPERIMENTAL STUDIES A. Prototypes It would be excellent if a prototype made up of the same material ( ) [4] could be realized and measured. Unfortunately, machinable 38 material is uncommon and, as such, we have used Emerson & Cumings Eccostock HiK ( 10) material readily available in the laboratory. The prototype is shown in Fig. 4. The photograph in Fig. 4 also includes an NMP etched on a 62-mil-thick RT Duroid 5870 substrate. In practical congurations, the DRA is symmetrically placed on top of the patch and glued to the substrate. Fabrication of the feed, especially the soldering of the probe with the patch, is to be done very carefully so that the nished patch turns to be an absolutely at surface causing almost no air gap in between the DRA and the feed surface. The DRA parameters have been determined on the basis of the dielectric permittivity of the material available in the laboratory. Since no computer-aided design (CAD) formula is available to predict the resonant frequency due to the mode ( ), this has been primarily estimated, following Table I, as

74

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 8. Simulated versus frequency response of a CDRA for different radii of the feeding patch. Parameters as in Fig. 4.

Fig. 6. Portray of electric elds in the DRA examined in Fig. 5. (a) Top view. (b) Side view.

7.4 GHz:

Fig. 9. Measured and simulated versus frequency responses of the NMP, used as a feed to the CDRA. Inset: photographs of the SMA-fed prototype viewed from the back and front sides. The parameters appear as they do in Fig. 4.

of interest. The feed structure, viewed from the front and back, is shown as the inset. C. Radiation Patterns
Fig. 7. Portray of the electric elds in the DRA examined in Fig. 5. GHz: (a) Top view. (b) Side view. 6.0

frequency of the same DRA is examined by varying the radius of the NMP. Almost no change at 7.4 GHz is revealed, but that near the rst resonance is affected to some extent by the change in NMP diameter. The -like mode, therefore, should radiate like an mode with some changed or additional features, as examined later. The feed itself should not take part in the resonance or radiation mechanism. The nonresonant nature of the proposed feed (i.e., a stand-alone NMP excited by an SMA probe) is veried in Fig. 9. Both measured and simulated data are presented. The values of vary from 1.5 to 0 dB, indicating absolutely no resonance occurring in the feeding patch over the frequencies

Radiation properties due to both the modes are experimentally studied using automated anechoic chamber and are presented in Figs. 10 and 11. Measured results are also compared with the corresponding simulated values. Fig. 10 shows the radiation patterns obtained at 7.4 GHz. Very close agreement between simulation and measurement is revealed. The radiation pattern in -plane is more symmetric than that in -plane. Measured nulls in -plane show an asymmetry. This may be attributed to the vertical probe, which is used to excite the NMP. In the present prototype, that probe is optimally located near the patch-edge and, hence, causes this asymmetry. Measured cross-polarized elds in the -plane show more than 20-dB isolation from the peak gain. This value becomes more than 25 dB in the -plane. Both simulated and measured data indicate more than 8.5-dBi gain. This, indeed, is improved by 12 dB compared to that caused by the mode. The increment

GUHA et al.: HIGHER ORDER MODE EXCITATION FOR HIGH-GAIN BROADSIDE RADIATION FROM CYLINDRICAL DRAs

75

Fig. 10. Measured and simulated radiation characteristics due to the mode in a CDRA having characteristics as in Fig. 5. 7.4 GHz. (a) -plane. (b) -plane.

Fig. 11. Measured and simulated radiation characteristics due to the mode in a CDRA having characteristics as in Fig. 5. 6.0 GHz. (a) -plane. (b) -plane.

in gain is quite justied as the radiating one is a higher order mode. The radiation patterns due to the -like mode, portrayed in Fig. 7, are studied in Fig. 11. Very good broadside patterns are revealed and like Fig. 10, measurements closely agree with the simulated data. Compared to Fig. 10, the peak gain value is low by 0.7 dB, and the beam is wider by about 25 in the -planes. A comparative study based on measured results is presented in Table II. This result corroborates our conjecture that the rst resonance near 6 GHz is due to the -like mode with useful broadside radiation patterns. D. Radiation Efciencies Radiation efciency at each resonant mode has been measured by the Wheeler cap technique [11], which indeed uses input resistances of an antenna with and without the Wheeler cap. This Wheeler cap is actually a metallic cavity used to shield or cover the radiating element Input resistance at resonance its Radiation resistance of an antenna Loss resistance

TABLE II LIKE AND RADIATION CHARACTERISTICS COMPARED FOR MODES IN THE PROTOTYPE EXAMINED IN FIGS. 10 AND 11

Indeed, accounts for the radiated elds only and is for the unavoidable leakage taking place in the antenna structure. Under the capped condition, no radiation takes place and, thus, with cap . When the cap is removed, the total input resistance is obtained as without cap . The quantities and can be determined by using a network analyzer and manipulated to calculate radiation efciency as [12]:

76

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE III RADIATION EFFICIENCY OF THE PROTOTYPE MEASURED AT THE INDIVIDUAL RESONANT FREQUENCIES. PARAMETERS AS IN FIG. 4

Radiated power Total Input power (1) The measured is very sensitive to the cap and its placement on the ground plane. Measured data are shown in Table III and those are obtained at frequencies corresponding to measured minima occurring in Fig. 5. The radiation efciencies at both frequencies appear almost comparable.

[6] A. A. Kishk and A. W. Glisson, Bandwidth enhancement for split cylindrical dielectric resonator antennas, Progress Electromagn. Res., vol. PIER 33, pp. 97118, 2001. [7] D. Guha, A. Banerjee, and Y. M. M. Antar, New radiating mode in a cylindrical DRA to produce broadside high gain radiation, presented at the IEEE Antennas Propag. Soc. Int. Symp., Toronto, ON, Canada, 2010. [8] High Frequency Structure Simulator (HFSS)11.1 ed. Ansoft Corp., 2009. [9] K. M. Luk and K. W. Leung, Eds., Dielectric resonator antennas, Baldock, U.K., Research Studies Press, 2003. [10] A. Petosa, Dielectric Resonator Antenna Handbook, 1st ed. Norwood, MA: Artech House, 2007. [11] H. A. Wheeler, The radian sphere around a small antenna, Proc. IRE, vol. 47, no. 8, pp. 13251331, 1959. [12] R. Chair, A. A. Kishk, K. F. Lee, and D. Kajfez, Performance comparisons between dielectric resonator antennas and printed microstrip patch antennas at X-band, Microw. J., Jan. 2006.

IV. CONCLUSION This paper presents a new approach of exciting a cylindrical DRA using an NMP. Here, we have experimentally established for the rst time that the mode in a CDRA can be successfully excited, resulting in high gain with broadside radiation. The possibility of exciting a secondary -like mode is also experimentally shown. Thorough investigations are needed to verify its consistency for different values of DRA dimensions, NMP radius, and the permittivity of the dielectric materials. The NMP is fed by a coaxial probe. The development of a fully planar feed using the microstrip line will alleviate a major lacuna of integrating a DRA with a fully planar circuit. In our view, many unknown issues related to these studies need to be unveiled through proper investigations in the near future. These are important to improve the use of the new mode and the new feeding technique.

REFERENCES
[1] S. A. Long, M. W. McAllister, and L. C. Shen, The resonant cylindrical cavity antenna, IEEE Trans. Antennas Propag., vol. AP-31, no. 3, pp. 406412, May 1983. [2] A. A. Kishk and Y. M. M. Antar, Dielectric resonator antennas, in Antenna Engineering Handbook, J. Volakis, Ed., 4th ed. New York: McGraw-Hill, 2007, ch. 17. [3] A. W. Glisson, D. Kajfez, and J. James, Evaluation of modes in dielectric resonators using a surface integral equation formulation, IEEE Trans. Microw. Theory Tech., vol. MTT-31, no. 12, pp. 10231029, Dec. 1983. [4] D. Kajfez, A. W. Glission, and J. James, Computed modal eld distributions for isolated dielectric resonators, IEEE Trans. Microw. Theory Tech., vol. MTT-32, no. 12, pp. 16091616, Dec. 1984. [5] G. J. Karonis, D. I. Kaklamani, and N. K. Uzunoglu, Accurate analysis of a cylindrical dielectric resonator mounted on a grounded dielectric substrate, Progress Electromagn. Res., vol. PIER 23, pp. 187219, 1999.

Debatosh Guha (SM03) received the B.Tech., M.Tech., and Ph.D. degrees in radio physics and electronics from the University of Calcutta, Kolkata, India, in 1986, 1988, and 1994, respectively. He began his professional career as an Engineer in Webel Telecommunication Industries Ltd., India, in 1989. In 1990, he joined the Institute of Radio Physics and Electronics, as a Senior Research Fellow of the Council of Scientic and Industrial Research (CSIR), India. In 1994, he joined the same university as a Lecturer in Radio Physics and Electronics. He is a Professor in the Institute of Radio Physics and Electronics, University of Calcutta. At different periods, he was a Visiting Research Professor and Visiting Professor in the Electrical and Computer Engineering Department, Royal Military College of Canada, Kingston, ON, Canada. He also visited several foreign universities and research institutes which include the University of Houston, Houston, TX; University of Bath, Bath, U.K.; Queen Mary College University of London; and Communication Research Centre of Canada, Ottawa, ON, Canada. His current research interests include the application of defected ground structure to low-prole antennas, analysis and characterization of printed antennas for wireless communications, ultrawideband dielectric resonator antennas (DRAs), and uncommon DRA modes for microwave integrated circuit applications. Dr. Guha is the Founder Chair of the IEEE AP-MTT Calcutta Chapter, and served various international symposia and conferences in his eld as General Chair/Program Chair. He is a recipient of the Jawaharlal Nehru Memorial Fund Prize and URSI Young Scientist Award in 1996. He is on the board of reviewers of several international journals, including IEEE TRANSACTION ON ANTENNAS AND PROPAGATION, IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, Institute of Engineering and Technology Microwave, Antennas and Propagation, Electronic Letters, and Radio Science. He is the author/editor of a book Microstrip and Printed Antennas: New Trends, Techniques and Applications (Wiley, 2010). He has more than 45 journal publications to his credit as well as about 75 conference articles.

Archita Banerjee (S10) was born in West Bengal, India. She received the B.Sc. degree in physics (Hons.) from Burdwan University, Burdwan, India, in 2005, the M.Sc. degree in electronics from Pt. Ravishankar Shukla University, Raipur, India, in 2007, and is currently working toward the Ph.D. degree in radio physics and electronics at the Institute of Radio Physics and Electronics, University of Calcutta, Kolkata, India. Her research interests include dielectric resonator antennas and printed antennas with defected ground structures and hybrid DRAs.

GUHA et al.: HIGHER ORDER MODE EXCITATION FOR HIGH-GAIN BROADSIDE RADIATION FROM CYLINDRICAL DRAs

77

Chandrakanta Kumar (M09) was born in Shibpur, West Bengal, India, in 1976. He received the M.Tech. degree in radio physics and electronics from the University of Calcutta, Kolkata, India, in 2001. He subsequently joined the Communication Systems Group of ISRO Satellite Centre, Bangalore, India, as an Engineer. Since then, he is actively involved in the design and development of antenna systems of the Indian space programme and related ground stations. He served as Project Manager, Antenna Systems, for the rst Indian mission to the moon: Chandrayaan-1. Currently, he has a similar responsibility for the GSAT-12, ASTROSAT and Chandrayaan-2 missions. He has about 20 publications in international journals and conferences to his credit. His research interests are defected ground structure integrated microstrip patch antennas, DRAs, lightweight antennas for spacecraft, and microelectromechanical systems. Mr. Kumar is a recipient of the Young Scientist Award-2009 from the Indian Space Research Organization (ISRO). He is also a member of the team that received the Team Excellence Award-2008 from ISRO for his contribution to the Chandrayaan-1 antenna system.

Yahia M. M. Antar (S73M76SM85F00) received the B.Sc. (Hons.) degree in electrical engineering from Alexandria University, Alexandria, Egypt, in 1966, and the M.Sc. and Ph.D. degrees in electrical engineering from the University of Manitoba, Winnipeg, MB, Canada, in 1971 and 1975, respectively. In 1979, he joined the Division of Electrical Engineering, National Research Council of Canada, Ottawa, ON, Canada, where he worked on polarization radar applications in remote sensing of precipitation, radio-wave propagation, electromagnetic scattering, and radar

cross-section investigations. In 1987, he joined the staff of the Department of Electrical and Computer Engineering, Royal Military College of Canada, Kingston, ON, Canada, where he has held the position of Professor since 1990. He has authored or coauthored more than 170 journal papers and 300 refereed conference papers, holds several patents, chaired several national and international conferences, and has given plenary talks at conferences in many countries. He has supervised or co-supervised many Ph.D. and M.Sc. theses at the Royal Military College of Canada, and at Queens University, Kingston, where several have received the Governor General of Canada Gold Medal, the outstanding Ph.D. thesis of the Division of Applied Science, as well as many best paper awards in major symposia. He is an Associate Editor (Features) of the IEEE Antennas and Propagation Magazine, an Associate Editor of the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, and a member of the Editorial Board of the RFMiCAE Journal. Dr. Antar is a Fellow of the Engineering Institute of Canada (FEIC) and a Fellow of the Electromagnetic Academy. He was awarded a Tier 1 Canada Research Chair in Electromagnetic Engineering in 2002, which was renewed in 2009. In 2003, he was awarded the 2003 Royal Military College Excellence in Research Prize. He was elected by the Council of the International Union of Radio Science (URSI) to the Board as Vice President in 2008, and to the IEEE Antennas and Propagation Society Administration Committee in 2009. On January 31, 2011, he was appointed Member of the Defence Science Advisory Board (DSAB). In 1977, he was awarded a Government of Canada Visiting Fellowship at the Communications Research Centre, Ottawa, where he worked with the Space Technology Directorate on communications antennas for satellite systems. He was elected and served as the Chairman of the Canadian National Commission for Radio Science (CNC, URSI,19992008) and Commission B National Chair (19931999). He holds an adjunct appointment at the University of Manitoba and has a cross appointment at Queens University. He has also served, since 2008, as Associate Director of the Defence and Security Research Institute (DSRI). He served on the NSERC grants selection and strategic grants committees, Ontario Early Research Awards (ERA) panels, and the National Science Foundation Electrical, Communications, and Cyber Systems (ECCS) review panel.

78

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

On the Characteristics of the Highly Directive Resonant Cavity Antenna Having Metal Strip Grating Superstrate
Alireza Foroozesh, Member, IEEE, and Lotfollah Shafai, Life Fellow, IEEE
AbstractVarious resonant cavity antennas (RCAs) having different metal strip gratings (MSGs) as their superstrates are studied in terms of their directivity and scanning properties using transverse equivalent network (TEN) model in conjunction with the periodic method of moments (MoM). It is shown that radiation patterns in E- and H-planes coincide over a wide angle range in RCAs when their MSG superstrates are free-standing and highly reective, as reported in the previous literature. However, using less reective MSG or employing dielectric support in the MSG superstrate degrades the equality of E- and H-plane radiation patterns. Moreover, as the scan angle increases, radiation patterns in E- and H-planes become noticeably different. It is elucidated that this is due to the angular dependence of the reection coefcient phase of the MSG superstrates. A comparative study is also performed on a few different RCAs based on both simulation and measurement results. It is illustrated that in practice when the excitation source of the RCA is a probe-fed microstrip antenna and the RCA is nite in size, cross-polarization increases considerably as opposed to the RCAs having ideal sources and, innite ground plane and MSG superstrate in the transverse direction. Index TermsFrequency selective surface, metal strip gratings, reection coefcient, resonance cavity antennas, transmission coefcient.

Fig. 1. Resonance cavity antenna (RCA) consisting of free-standing metal strip grating (MSG) superstrate and an electric Hertzian dipole on the grounded dielectric slab substrate.

I. INTRODUCTION

HE resonant cavity antennas (RCAs) can have high directivity both in broadside [1] and scan angles [2]. The mechanism of the gain enhancement in such antennas can be described in different ways, for example, the ray-tracing of the multiple reections between the ground plane and superstrate [1], or leaky waves [2]. Regardless of how the phenomenon is described, the key role in the directivity enhancement is played by the superstrate layer which needs to be highly reective and only allow a small portion of the wave to leak out. Various types of highly reective surfaces have been exploited as the RCA superstrates, such as high contrast layers [3][7], and frequency selective surfaces (FSSs) [8][12]. Some advantages of employing FSSs are pointed out in [8], however FSSs appear in different shapes [13] and [14]. Among those, and indeed, one of the canonical forms is the metal strip grating (MSG) which

Manuscript received January 28, 2011; revised June 06, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the Department of Electrical and Computer Engineering, University of Manitoba, Winnipeg, Canada (e-mail: aforoozesh@ieee.org). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167933

possesses unique impedance surface characteristics that makes it attractive for certain antenna applications [15] and [16]. A geometry of an RCA having MSG superstrate is depicted in Fig. 1. Its excitation source is a y-polarized Hertzian dipole placed at origin. It has been shown a free-standing version of the MSG superstrate can produce azimuthally omnidirectional pencil beams pointing at broadside or azimuthally omnidirectional conical beams scanned off broadside [16]. This remarkable property has elegantly been described in [16] and attributed to the shunt equivalent susceptance of the MSG which was later used in the transverse equivalent network (TEN) model of the RCA. As well, polarization purity of such an antenna has been studied in [16] through TEN modeling for the ideal cases of innite ground plane and MSG sizes. It was shown that this antenna is highly polarized as well [16]. In this paper, rst, the effects of the strip width of the MSG superstrate as well as the effects of adding dielectric slabs as the support of the MSG superstrate are studied. This comparative study points out the need to use free-standing MSG as the RCAs superstrate to obtain equal E- and H-plane radiation patterns. This study is carried out using the TEN models of the RCAs, which has been proved to be a useful, fast and adequately accurate computational tool in multilayered antenna design [8][12], [15][17], [19] and [20]. Second, a full-wave study on the polarization purity of the RCAs having MSG superstrate is performed. To this end, actual antenna is considered which is excited by a microstrip patch and its FSS superstrate is truncated. As a comparative study, its

0018-926X/$26.00 2011 IEEE

FOROOZESH AND SHAFAI: ON THE CHARACTERISTICS OF THE HIGHLY DIRECTIVE RESONANT CAVITY ANTENNA

79

cross-polar radiation patterns are compared against those obtained from RCAs having different FSS superstrates. The other RCAs have microstrip patch and aperture FSSs which were successfully designed, analyzed and experimentally studied in the previous work [7] and [8]. Both simulation and measurement results are presented and discussed. The paper is organized as follows. In Section II, rst, different MSGs are characterized using periodic MoM employing entire-domain basis functions. Then, directivities of different RCAs versus different resonant lengths are obtained using the TEN modeling. Next, scanning properties of the radiation patterns are investigated for different RCAs. In Section III, four different RCAs having four different superstrates are analyzed using a MoM-CAD software package. Their radiation patterns are plotted in different planes and their polarization purity is studied. Experimental results of the four fabricated RCAs are presented in Section IV. Finally, conclusions are drawn in Section V. II. RCA ANALYSIS USING TEN The RCA having free-standing MSG superstrate is depicted in Fig. 1. This may not be the case in all applications. Side view of the RCAs having dielectric support for the MSG superstrate are shown in Fig. 2(a) and (b). When a support is employed for the MSG superstrate, two congurations can be realized. In one, MSG is on the bottom of the dielectric and in the other, MSG is placed on top of it, as shown in Fig. 2(a) and (b), respectively. Simple ray-tracing method based on the multiple reections was originally used to describe the mechanism of the gain enhancement in this type of antennas [1]. Such a phenomenology is shown in Fig. 2(c). The TEN modeling of this antenna is depicted in Fig. 2(d). Different MSGs, either free-standing or etched on dielectric support, are characterized in this section. The analysis method and derivations are briey explained and well addressed in Appendix A. The operating frequency of 15 GHz is considered for all cases. Then, these MSGs are employed as the RCAs superstrates. Directivity versus RCA length is plotted for these cases. Finally, the scanned-angle radiation patterns of these RCAs are studied. A. MSG Characterization Here, four different MSGs are considered. All MSGs have mm. Two of them are backed by the same periodicity of dielectric support having relative permittivity of 2.5. The other two are free-standing. In each group, one has the strip-width of mm and the other has the strip-width of mm. Reection and transmission coefcients of these MSGs are plotted versus frequency for the normal incidence in Figs. 3 and 4. One can observe that for the identical strip width, the reectivity of the free-standing MSG is higher than that of its counterpart at a xed frequency. This effects is more pronounced in the case of mm wherein the strip is narrow with respect to the periodicity. This observation can be viewed from epsilon-effective perspective. Adding a dielectric slab as the MSG support increases the effective permittivity of the media and shifts the frequency downward. On the other hand, the reection coefcient magnitude of any MSG, functioning in its inductive mode (which is the case in this study), is monotonically decreasing.

Fig. 2. (a) Side-view of an RCA having a superstrate consisting of a periodic surface along with the dielectric support on top of it. (b) Side-view of an RCA having a superstrate consisting of a periodic surface along with the dielectric support underneath it. (c) Illustration of the phenomenology of the gain enhancement in the RCAs using multiple reections and wave leakage. (d) Transverse-equivalent-network (TEN) modeling of an RCA.

Therefore, as long as the MSG functions in its inductive mode, its reectivity is less when it is a free-standing one as compared to the ones backed by a dielectric slab. In order to estimate the radiation patterns and directivity of the RCAs using TEN, the MSG properties versus incident angle is required in both E- and H-planes. Therefore, when the source is y-polarized, one needs to nd the reected and transmitted elds for the and polarizations, in (E-plane) and (H-plane), respectively. Reection and transmission properties of the aforementioned MSGs are displayed versus incident angle for the frequency of 15 GHz in Figs. 5 and 6. Here, too, it is observed that the reectivity of the free-standing MSGs is higher than that of their counterparts having dielectric support. B. RCA Directivity The TEN modeling of the RCA is depicted in Fig. 2(d). One should note that, since the reection and transmission properties of the MSGs have only been determined in the E-plane and the H-plane in the previous section, TEN is only able to estimate the radiation patterns in the E- and H-planes. Consequently, the directivity is obtained based on those two radiation patterns as addressed in [12]. In other words, one needs to compute the reection and transmission coefcients versus incident angle in other planes as well, then take them into account in the TEN modeling in order to obtain a more accurate estimate. In those cases, cross-polarized reection and transmission coefcients need to be calculated in different planes, too.

80

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 3. Reection coefcients vs. frequency of four different MSGs having different strip width and dielectric substrate and superstrates. (a) Magnitude and (b) phase.

Fig. 4. Transmission coefcients vs. frequency of four different MSGs having different strip width and dielectric substrate and superstrates. (a) Magnitude and (b) phase.

This makes the TEN modeling more complicated, as and polarizations are no longer decoupled in ports 1 and 2 as shown in Fig. 2(d). However, since the evaluation of the polarization purity of these antennas are not carried out in this section, and the previous TEN modeling has proven to provide adequate accuracy [3][12], simple TEN modeling sufces, as presented in Fig. 2(d). The derivation of the scattering parameters of the two-port network shown in Fig. 2(d) (MSG superstrate) is briey explained in Appendix B. Directivities of different RCAs versus their cavity length are plotted in Fig. 7. Expectedly, RCAs having superstrate with thicker conducting strips possess higher directivity than their corresponding counterparts, since their superstrate layers exhibit higher reectivity. As shown in Fig. 7, there is also a comparison between the directivity of the two different RCA congurations, shown in Fig. 2(a) and (b), for which, the MSGs are placed on bottom and top of the support dielectric slab, respectively. It is demonstrated in Fig. 7 that those RCAs having the MSG on the bottom of the dielectric superstrate resonate at the larger cavity length and have higher gain as compared to the RCAs having the MSG on the top. This observation is consistent with [7], which indicates that using different superstrates

with identical reectivity does not guarantee to achieve RCAs having identical directivity. Reection phase of the superstrates needs to be considered, too, as it plays an important role in the RCA directivity. The reection phase determines the resonant length of the RCA. C. Scanning Properties of the Radiation Patterns By increasing the cavity resonant length and keeping the operating frequency xed or alternatively, increasing the frequency and keeping the cavity resonance length xed, one can scan the beam in resonant cavity leaky-wave antennas [9], [10], and [17]. Various types of FSS superstrate such as microstrip patch or slot can be employed in the scanning RCAs as addressed in [9] and [10], respectively. However, it is asserted in [15] and [16] that far-eld radiation patterns of an RCA having free-standing MSG superstrate are equal in E- and H-planes due to the unique surface impedance characteristics of the MSGs. This unique feature of MSG only exists when it is suspended in the air, i.e., free-standing MSG. An expression for the surface susceptance of the free-standing MSG has been presented in [18] which has been used in the TEN modeling to obtain some of the radiation patterns in [15] and

FOROOZESH AND SHAFAI: ON THE CHARACTERISTICS OF THE HIGHLY DIRECTIVE RESONANT CAVITY ANTENNA

81

Fig. 5. Reection coefcients vs. incident angle of four different MSGs having different strip width and dielectric substrate and superstrates. (a) Magnitude and (b) phase.

Fig. 6. Transmission coefcients vs. incident angle of four different MSGs having different strip width and dielectric substrate and superstrates. (a) Magnitude and (b) phase.

[16]. In this paper, periodic spectral-domain moment method as described in Appendix A is exploited to obtain scattering parameters of the two-port network in Fig. 2(d). These parameters are then applied to the TEN modeling of the antenna as was carried out in [8]. The TEN modeling is based on reciprocity and inherently employs it as discussed in the literature, e.g., [19] and [20]. Therefore, the TEN method used in this work is equivalent to that explained in [9] and [10]. Fundamental concepts and useful expressions in designing a scanned-beam RCA, are well presented in [9], [10], and [17] when the substrate consists of only one dielectric. In such cases, the relationship between the resonance cavity length and the scan angle is expressed as
Fig. 7. Directivity of the RCAs having different dielectric substrates and MSG superstrates versus cavity length at 15 GHz.

(1)

82

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I SCAN ANGLE AND CORRESPONDING REFLECTION PHASES OF THE GROUNDED DIELECTRIC SLAB AND THE MSG SUPERSTRATE AND RESONANT LENGTHS IN MM, FOR BOTH TE AND TM POLARIZATIONS. MM,
AND MM

TABLE III SCAN ANGLE AND CORRESPONDING REFLECTION PHASES OF THE GROUNDED DIELECTRIC SLAB AND THE MSG SUPERSTRATE AND RESONANT LENGTHS IN MM, FOR BOTH TE AND TM POLARIZATIONS. MM, AND MM

TABLE II SCAN ANGLE AND CORRESPONDING REFLECTION PHASES OF THE GROUNDED DIELECTRIC SLAB AND THE MSG SUPERSTRATE AND RESONANT LENGTHS IN MM, FOR BOTH TE AND TM POLARIZATIONS. MM,
AND MM

TABLE IV SCAN ANGLE AND CORRESPONDING REFLECTION PHASES OF THE GROUNDED DIELECTRIC SLAB AND THE MSG SUPERSTRATE AND RESONANT LENGTHS IN MM, FOR BOTH TE AND TM POLARIZATIONS. MM, AND MM

where is the corresponding wavelength in the free space, and is the substrate relative permittivity. However, in many actual RCAs, the excitation source is placed on the grounded dielectric slab as shown in Fig. 2. In these cases, the substrate does not consist of a uniform dielectric material but of two or more layers including the air-gap. Therefore, following the formulas and derivation procedures presented in [1] and [8], the relationship between the RCA air-gap height and the scan angle is expressed as

(2) where and are the reection coefcient phases of the grounded dielectric slab and the superstrate layer, respectively, and is an integer. One should note that and are both angularly and polarization ( and ) dependent. This can clearly be seen in Fig. 5(b) for the case of . When the dielectric substrate losses are negligible, can be calculated as

scan angles of 0 , 15 , 30 , 45 , 60 , 75 of the aforementioned four RCAs are summarized in Tables I, II, III and IV. As can be seen in Tables I and II, when the dielectric support is air the TE and TM resonant heights are almost the same for a fairly large range of scan angles. This is even more pronounced when the strip widths are wide, i.e., the reectivity of the MSG is higher (Table II). In the case of the use of dielectric support for the MSG superstrate, it is observed that TE and TM resonant heights are considerably different for large scan angles. Radiation patterns of the RCAs using aforementioned congurations are plotted in Figs. 8 and 9 for different scan angles. Resonant length, here in this work, is considered as the mean value of the TE and TM resonant air-gap heights

(4)

(3) where and are intrinsic impedances of the wave in dielectric and free-space, respectively. One should note that and are both angularly and polarization dependent, and is angularly dependent ([9] and [19]). Therefore, for any scan angle two different heights are obtained, one due to the TE and the other due to TM polarization. The RCA conguration which is shown in Fig. 2(a) is considered. Two different strip widths and two different superstrate dielectric supports are considered which makes four different combinations of the RCAs in total. Corresponding reection phases and air-gap heights for various

Therefore, it is expected in the cases of using dielectric slabs in the antenna structure or less-reective superstrates, the radiation patterns in the E- and H-planes become noticeably different. The best equality between the two radiation patterns belongs to the case that the MSG is free-standing and thicker (more highly-reective), and the substrate dielectric is air as shown in Fig. 8(b). Another interesting observation is the appearance of the grating lobes other than the scanned main beam when . For example, in the case of , free-standing MSG and air substrate, the broadside beam appears as well, and the air-gap height is almost twice that of the case with only the broadside main-beam . In these cases, second parallel-plate waveguide mode is excited. This is consistent with the observation made in [9]. As scan angle increases, the difference between and grows, especially in the RCAs having dielectric slab supports. Also, the difference between and increases. Therefore, resonant air-gap heights for

FOROOZESH AND SHAFAI: ON THE CHARACTERISTICS OF THE HIGHLY DIRECTIVE RESONANT CAVITY ANTENNA

83

Fig. 8. Radiation patterns of an RCA having free-standing MSG superstrate mm and and PEC ground plane for different scan angles, (a) case of mm. mm for both cases. (b) case of

Fig. 9. Radiation patterns of an RCA having MSG superstrate with dielectric slab on top as the support and grounded dielectric slab as the substrate for difmm and (b) case of mm. ferent scan angles. (a) Case of mm for both cases. Relative permittivity of both dielectrics used in the substrate and superstrate is 2.5 and its thickness is 1.59 mm.

TE and TM are signicantly different at the large scan angles and consequently for a xed resonant length, , radiation patterns in E- and H-planes scan different angles. In these cases, the difference between E- and H-planes are more pronounced. This comparative study restates that the unique features of the surface impedance of the MSGs belongs to the cases of the RCAs that use air as substrate-superstrate dielectric as described in [16]. However, one can see even in the RCAs having free-standing MSG, reectivity of the MSG effectively has impact on the equality of radiation patterns in E- and H-planes, as illustrated in Fig. 8. III. RCA ANALYSIS USING A MOM-CAD ANALYZER One of the recent applications of the RCAs is for reector feeds [21] and [22]. Several advantages of such feeds in millimeter-wave as well as very low frequency applications have been pointed out in [22]. Low cross-polarization in the radiation patterns of the reectors can be achieved when the feed has equal E- and H-plane radiation patterns as explained in [23, p. 222]. On the other hand, as asserted in [16], RCAs exhibit

equal E-plane and H-plane radiation patterns when their superstrates consist of MSGs. However, ideal RCAs have innite superstrate, ground planes and ideally polarized excitation source (Hertzian dipole) as were considered in [16]. In practice, actual RCAs have nite superstrates and waveguide aperture or microstrip antenna can be used as their excitation sources. Furthermore, having low cross-polarization is another feature of an efcient reector feed. In this section, the polarization purity of the actual RCAs having MSG superstrates along with their other characteristics are investigated. A comparative study is also carried out with the previously designed RCAs reported in [7] and [8]. In this work, microstrip patch antenna is employed as the excitation source in the design of the actual RCAs as shown in Fig. 10. A MoM-CAD software package (Ansoft Designer) is employed to analyze the RCAs in this section. In all simulations of this section, ground plane conductor and dielectric layers employed in both superstrate and substrate are assumed innite in the transverse direction. However, the conducting periodic surfaces of the superstrates have the dimensions of 15 15 cm in

84

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 10. (a) Conguration of an RCA having FSS superstrate and probe-fed microstrip antenna excitation. (b) Source microstrip patch antenna having dimm, mm, mm. is the distance mensions of between the microstrip patch center to the feeding probe. One should note that superstrate dielectric layer is a support sheet for a highly reective surface which can be designed in any FSS topology, e.g., microstrip patch, aperture or MSG.

Fig. 11. Unit cells of the different periodic surfaces used in this paper as the mm and mm, (b) superstrate. (a) Microstrip patch FSS having mm and mm, (c) thick MSG rectangular aperture FSS having mm and (d) mm, and thin MSG having having mm, and mm.

Fig. 12. Four different FSSs employed as the RCA superstrates. Each has the dimensions of 15 15 cm . (a) Actual aperture (left) and microstrip patch (right) FSSs. (b) Right: Actual thick MSG. Left: Actual thin MSG.

all designs. As well, conducting surfaces such as ground plane, microstrip patch and the metallic parts of the FSS layer are all PEC. Superstrate and substrate dielectrics are both made of Arlon Diclad 522 having relative permittivity of 2.5, loss tangent of 0.001 and thickness of 1.59 mm. Four different RCAs having four different FSSs namely patch, aperture, thick MSG and thin MSG are analyzed here. Their unit cells are shown in Fig. 11. The actual fabricated FSSs, whose dimensions are 15 15 cm , are portrayed in Figs. 12. They are designed so that they exhibit identical reection coefcient magnitude at the frequency of 8.28 GHz. Their reection coefcients versus frequency are shown in Fig. 13. At the frequency of 8.28 GHz, their corresponding reection coefcients versus incident angle, and , are plotted in Figs. 14 and 15, respectively. This has also been done in order to preserve consistency with our previous work in [7] and [8], and carry out a comparative study on the cross-polarization characteristics of the RCAs. As can be seen in Fig. 13, the aperture FSS, thick and thin MSGs are all inductive. Interestingly, all the inductive reection coefcients are all almost identical over the frequency range. On the other hand, microstrip patch FSS exhibit capacitive behavior according to its reection coefcient characteristics [13]. This has been illustrated using transmission line analogy in conjunction

with full-wave analyses in [24]. The TM-TM reection coefcients of the aforementioned FSSs show signicant angular variation as plotted in Fig. 14. They also differ especially in the grazing angles as can be seen in Fig. 14. However, the TE-TE reection coefcients of the aforementioned FSSs exhibit less angular dependence as plotted in Fig. 15. As well, the TE-TE reection coefcient magnitudes of all the aforementioned FSSs are almost identical over the incident angle range. Co- and cross-polarized radiation patterns of the four aforementioned RCAs are plotted in different planes in Figs. 16, 17, 18 and 19. In order to perform a fair comparison between the polarization purity of the aforementioned RCAs some criteria are developed here. Since the rst null in the radiation patterns of the most RCAs occur around and 30 , the maximum value of the cross-polarization level is considered in this range for all cases. Also, since the peak gain of the co-polar radiation pattern of each RCA is different from that of the others, relative cross-polarization level needs to be considered. To this end, in the range of is adopted as a gauge for polarization purity in accordance with the denition in the available literature [25]. These values are listed in Table V. One should note that and denote the co- and cross-polarized electric elds in the far zone, respectively. Also, one should note that the Ludwigs third denition

FOROOZESH AND SHAFAI: ON THE CHARACTERISTICS OF THE HIGHLY DIRECTIVE RESONANT CAVITY ANTENNA

85

Fig. 13. Reection coefcients of the FSSs shown in Figs. 11 and 12 versus frequency for normal incidence.

Fig. 15. of the FSSs shown in Figs. 11 and 12 versus incident angle for incident eld having TE polarization.

Fig. 14. of the FSSs shown in Figs. 11 and 12 versus incident angle for incident eld having TM polarization.

is employed to dene and [26]. Since the source antenna is y-polarized in this work, the corresponding co-polarization and cross-polarizations far-elds are respectively dened as (5) (6) As can be observed in Figs. 16, 17, 18 and 19 as well as Table V, the highest directivity belongs to the RCA having microstrip patch (capacitive) FSS superstrate, which is 21.7 dBi. The aperture FSS and both the thick and thin MSGs show inductive behavior in this RCA conguration. The RCAs having these inductive superstrates exhibit about 3 dB less directivity than that of the microstrip patch RCA. It can be seen that the polarization purity of the microstrip patch RCA is the highest

Fig. 16. Calculated far-eld radiation patterns of the RCA having microstrip patch FSS superstrate in different planes using a MoM-CAD commercial software package (Ansoft Designer). (a) co-polar and (b) cross-polar radiation patterns.

whereas that of the aperture RCA is the lowest. According to the calculations presented in [16], MSG superstrates are expected

86

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 17. Calculated far-eld radiation patterns of the RCA having aperture FSS superstrate in different planes using a MoM-CAD commercial software package (Ansoft Designer). (a) co-polar and (b) cross-polar radiation patterns.

Fig. 18. Calculated far-eld radiation patterns of the RCA having thick MSG superstrate in different planes using a MoM-CAD commercial software package (Ansoft Designer). (a) co-polar and (b) cross-polar radiation patterns.

to have the highest polarization purity. However, those calculations in [16] correspond to the use of ideal Hertzian dipole excitation and innitely long free-standing MSGs. Here, the excitation source is a probe-fed microstrip antenna which obviously introduces considerable amount of cross-polarized elds. As well, truncation of the MSGs causes diffraction of the eld at the edges and intensify cross-polarization. Still, RCAs with MSG superstrates demonstrate better polarization purity as compared to their counterpart employing inductive aperture FSS superstrate. IV. MEASUREMENT RESULTS OF THE FABRICATED RCAS Four fabricated superstrates are shown in Fig. 12. These superstrates are employed in different RCAs having identical microstrip patch antenna as the excitation source. Co- and crosspolarized radiation patterns of the four aforementioned RCAs are plotted in different planes in Figs. 2023. Input impedance characteristics of these antennas are shown in Fig. 24. As can be seen, the antennas are well-matched over the frequency range of 8.2-to-9 GHz. Because of the existence of the double-loops shown in Fig. 24(b), these antennas are potentially wide-band in the input impedance sense. Some of the characteristics of

the fabricated RCAs are listed in Table VI. Peak gains at obtained using both simulation and measurement, listed in Tables V and VI, respectively, are in good agreement. D-plane is the inter-cardinal plane considered as . These measurement results also demonstrate that the polarization purity of the RCAs having MSG superstrates is better than their counterpart having aperture superstrate. Also, the RCA having microstrip patch FSS has the highest polarization purity. One should note that in the fabricated antennas the ground plane as well as the dielectric slabs used as the superstrate or substrate are nite in size and have the dimensions of 15 15 cm . Therefore, this can be one of reasons of the slight discrepancies between the simulation and measurement results. V. CONCLUSION Various characteristics of the RCAs having MSG superstrates were investigated. As such are the beam scanning, ability of producing equal radiation patterns in E- and H-planes and cross-polarization. First, using TEN modeling, it was shown that the equality of E- and H-plane radiation patterns can be better achieved in the cases of free-standing MSG superstrates. This is consistent with the analysis provided in [16]. When

FOROOZESH AND SHAFAI: ON THE CHARACTERISTICS OF THE HIGHLY DIRECTIVE RESONANT CAVITY ANTENNA

87

Fig. 20. Measured co- and cross-polar far-eld radiation patterns of the RCA having microstrip patch FSS superstrate in the E-, H- and D-planes.

Fig. 19. Calculated far-eld radiation patterns of the RCA having thin MSG superstrate in different planes using a MoM-CAD commercial software package (Ansoft Designer). (a) co-polar and (b) cross-polar radiation patterns.

Fig. 21. Measured co- and cross-polar far-eld radiation patterns of the RCA having aperture FSS superstrate in the E-, H- and D-planes.

TABLE V CHARACTERISTICS OF THE FOUR RCAS HAVING FOUR DIFFERENT FSS SUPERSTRATES STUDIED IN THIS PAPER. (SIMULATION RESULTS)

the dielectric support is added to the MSG structure, radiation patterns in E- and H-planes become noticeably different. Furthermore, through a few case studies, it was found out that the higher the reectivity of the MSGs are, a better overlap is achieved between E- and H-plane radiation patterns. It was also shown that as the scan angle increases the corresponding TE and TM resonant heights diverges considerably because of the differences in the reection coefcient phases of the MSGs, and that as well causes E- and H-planes to coincide less. Since the angular dependence of the MSGs is higher when they have dielectric support, consequently the inequality of E- and H-plane radiation patterns are more noticeable for these cases as well.

Fig. 22. Measured co- and cross-polar far-eld radiation patterns of the RCA having thick MSG superstrate in the E-, H- and D-planes.

The other interesting observation was the effect of the difference between locating the MSG on the top or bottom of a superstrate dielectric has on the resonance length and directivity of the RCAs. It was revealed that when the MSG is placed on top of the dielectric superstrate, the shorter resonant length and less directive antenna is obtained as compared to the case of placing

88

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE VI CHARACTERISTICS OF THE FOUR RCAS HAVING FOUR DIFFERENT FSS SUPERSTRATES STUDIED IN THIS PAPER. (MEASUREMENT RESULTS)

Fig. 23. Measured co- and cross-polar far-eld radiation patterns of the RCA having thin MSG superstrate in the E-, H- and D-planes.

An investigation was also carried out on the polarization purity of the RCAs. Both MoM-CAD analyses and measurement results were presented. They were in very good agreement and conrmative. Previously in [16], under the assumption of using ideal Hertzian dipole as the RCA excitation, inniteness of the ground plane, and free-standing MSGs, high polarization purity of the RCAs having MSG superstrates had been elucidated based on the TEN modeling. In this paper, it is revealed that in practice, when microstrip patch antenna is used as the excitation, and ground plane, dielectric substrate, and superstrates, as well as the FSSs and MSGs are all nite in size, polarization purity of the RCAs having MSGs are not necessarily the highest. Probe-fed microstrip patch introduces cross-polarization to the far-eld pattern. Moreover, truncated ground plane and superstrate increase the cross-polarization. In the studied cases, it was found out that the RCA having microstrip patch FSS superstrate exhibit the highest polarization purity. However, selection of a superstrate that can exhibit less cross-polarized far-elds is not conclusive and depends on other factors such as the size of the superstrate and ground plane, too, rather than only the type of the employed FSS in the superstrate.

APPENDIX A REFLECTION AND TRANSMISSION CALCULATIONS FOR MSGS BACKED BY A DIELECTRIC SLAB Conguration of a metal strip grating backed by a dielectric slab which is illuminated by an arbitrary plane wave incidence is depicted in Fig. 25. The electric surface current densities on the zeroth strip in the and directions are expanded using entire-domain basis functions as

Fig. 24. Input impedance characteristics of the fabricated RCAs having difof the RCAs and (b) representaferent superstrates shown in Fig. 12. (a) of the RCAs on the Smith chart. tion of the

(7)

MSG underneath of the superstrate dielectric support. Although the reection magnitudes in both cases are the same, the reection phases for the two cases are different, which causes the RCAs in the two cases to have different resonant length and directivities.

(8) Exploiting either modal analysis presented in [27] and [28], or periodic MoM presented in [28] and [29], the reected and

FOROOZESH AND SHAFAI: ON THE CHARACTERISTICS OF THE HIGHLY DIRECTIVE RESONANT CAVITY ANTENNA

89

transmitted electric elds in the x and y direction are calculated as

Fig. 25. Side view of the metal strip gratings on a dielectric slab illuminated by an arbitrary plane wave incidence.

elds due to the grounded dielectric slab as well as the induced current elds on the scatterers (periodic conducting gratings) are distinctly expressed. In (9) and (10), , and are calculated as (11) (12) (13) (14) (9) All other parameters have been dened or described in [27] and [28], and not brought here for the sake of brevity. Other eld components can be found in a similar fashion as presented in [28]. Eventually, co- and cross-polarization reection coefcients, in this work, are calculated as (15) (16) (17) (18)

where corresponding, and components of the pertinent electric and magnetic elds can be calculated by using the following expressions. (19) (10) and are the components of the elecwhere tric eld dyadic Greens function for the grounded dielectric slab when excitation source is a horizontal Hertzian dipole placed at the air-dielectric interface. Also, and denote transmitted eld in region (3) and reected eld in region (1), respectively. One should note that in (9) and (10), the contributions of the TE and TM polarizations of the reected (20)

APPENDIX B CALCULATION OF THE SCATTERING PARAMETERS TWO-PORT NETWORK SUPERSTRATE

OF THE

Once the transmitted and reected eld of the zeroth Floquets mode is obtained, they can be considered as the and

90

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

of the two-port network shown in Fig. 2. This owes to the fact that all other elds associated with other Floquets modes are evanescent, since in this study [9]. Since neither anisotropic material nor any active element is used in the FSS, it is reciprocal. Therefore, . On the other hand, if it is assumed that both dielectric and ohmic losses are negligible, we may write [30]

(21) Hence, one can obtain: .

ACKNOWLEDGMENT The authors would like to thank B. Tabachnik for his assistance in antenna measurement and C. Smit for fabricating the FSSs.

REFERENCES
[1] G. Von Trentini, Partially reecting sheet arrays, IRE Trans. Antennas Propag., vol. 4, pp. 666671, 1956. [2] D. R. Jackson and A. A. Oliner, A leaky-wave analysis of the highgain printed antenna conguration, IEEE Trans. Antennas Propag., vol. 36, pp. 905910, Jul. 1988. [3] D. R. Jackson and N. G. Alexopoulos, Gain enhancement methods for printed circuit antennas, IEEE Trans. Antennas Propag., vol. 33, pp. 976987, Sep. 1985. [4] N. G. Alexopoulos and D. R. Jackson, Fundamental superstrate (cover) effects on printed circuit antennas, IEEE Trans. Antennas Propag., vol. 32, pp. 807816, Aug. 1984. [5] A. Foroozesh and L. Shafai, Effects of articial magnetic conductors in the design of low-prole high-gain antennas with high-permittivity dielectric superstrate, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 1013, 2009. [6] A. Foroozesh, M. N. M. Kehn, and L. Shafai, Application of articial ground planes in dual-band orthogonally-polarized low-prole highgain planar antenna design, Progr. Electromagn. Res., vol. PIER-84, pp. 407436, 2008. [7] A. Foroozesh and L. Shafai, Investigation into the effects of the reection phase characteristics of highly-reective superstrates on resonant cavity antennas, IEEE Trans. Antennas Propag., vol. 58, pp. 33923396, Oct. 2010. [8] A. Foroozesh and L. Shafai, Investigation into the effects of the patchtype FSS superstrate on the high gain cavity resonance antenna, IEEE Trans. Antennas Propag., vol. 58, pp. 258270, Feb. 2010. [9] T. Zhao, D. R. Jackson, J. T. Williams, H. Y. Yang, and A. A. Oliner, 2-D periodic leaky-wave antennasPart I: Metal patch design, IEEE Trans. Antennas Propag., vol. 53, pp. 35053514, Nov. 2005. [10] T. Zhao, D. R. Jackson, and J. T. Williams, 2-D periodic leaky-wave antennasPart II: Slot design, IEEE Trans. Antennas Propag., vol. 53, pp. 35153524, Nov. 2005. [11] S. Maci, M. Caiazzo, A. Cucini, and M. Casaletti, A pole-zero matching method for EBG surfaces composed of a dipole FSS printed on a grounded dielectric slab, IEEE Trans. Antennas Propag., vol. 53, pp. 7081, Jan. 2005. [12] A. Foroozesh, L. Shafai, and M. Ng Mou Kehn, Application of polarization and angular dependent articial ground planes in compact planar high-gain antenna design, Radio Sci., vol. 43, 2008, RS6S03. [13] B. A. Munk, Frequency Selective Surfaces: Theory and Design. New York: Wiley, 2000. [14] T. K. Wu, Frequency Selective Surface and Grid Arrays. New York: Wiley, 1995. [15] G. Lovat, P. Burghignoli, F. Capolino, D. R. Jackson, and D. R. Wilton, High-gain omnidirectional radiation patterns from a metal strip grating leaky-wave antenna, in Proc. IEEE AP-S Int. Symp., Honolulu, HI, Jun. 1015, 2007, pp. 57975800.

[16] P. Burghignoli, G. Lovat, F. Capolino, D. R. Jackson, and D. R. Wilton, Highly polarized, directive radiation from a Fabry-Prot cavity leakywave antenna based on metal strip grating, IEEE Trans. Antennas Propag., vol. 58, pp. 38733883, Dec. 2010. [17] T. Zhao, D. R. Jackson, J. T. Williams, and A. A. Oliner, General formulas for 2-D leaky-wave antennas, IEEE Trans. Antennas Propag., vol. 53, pp. 35253533, Nov. 2005. [18] N. Marcuvitz, Waveguide Handbook. Stevenage, U.K.: PeregrinusIEE, 1986. [19] X. H. Wu, A. A. Kishk, and A. W. Glisson, A transmission line method to compute the far-eld radiation of arbitrarily directed Hertzian dipoles in multilayer dielectric structure: Theory and applications, IEEE Trans. Antennas Propag., vol. 54, pp. 27312741, Oct. 2006. [20] X. H. Wu, A. A. Kishk, and A. W. Glisson, A transmission line method to compute the far-eld radiation of arbitrary Hertzian dipoles in a multilayer structure embedded with PEC strip interfaces, IEEE Trans. Antennas Propag., vol. 55, pp. 31913198, Nov. 2007. [21] C. Menudier, R. Chantalat, M. Thevenot, T. Monediere, P. Dumon, and B. Jecko, Phase center study of the electromagnetic band gap antenna: Application to reector antennas, IEEE Antennas Wireless Propag. Lett., vol. 6, pp. 227231, 2007. [22] A. Neto, N. Llombart, G. Gerini, D. M. Bonnedal, and P. de Maagt, EBG enhanced feeds for the improvement of the aperture efciency of reector antennas, IEEE Trans. Antennas Propag., vol. 55, pp. 21852193, Nov. 2007. [23] R. E. Collin, Antennas and Radiowave Propagation. New York: McGraw-Hill, 1985. [24] A. Foroozesh and L. Shafai, Investigation of reection properties of reactive impedance substrate, in Proc. Canadian Conf. on Electrical and Computer Engineering CCECE-2005, May 2005, pp. 18981901. [25] K. F. Lee, K. M. Luk, and P. Y. Tam, Cross polarisation characteristics of circular patch antennas, IEE Electron. Lett., vol. 28, no. 6, pp. 587589, 1992. [26] A. C. Ludwig, The denition of cross polarization, IEEE Trans. Antennas Propag., vol. 21, no. 1, pp. 116119, Jan. 1973. [27] A. Foroozesh and L. Shafai, Electromagnetic scattering by periodic conducting strip grating on grounded dielectric slab: Study of the angular and polarization dependence in different incident planes, presented at the 14th Int. Symp. on Antenna Technology and Applied Electromagnetics the American Electromagnetics Conf. (ANTEM-AMEREM), Ottawa, Jul. 2010. [28] A. Foroozesh and L. Shafai, On the analysis methods of and reection characteristics for an AMC/AEC surface, IEEE Antennas Propag. Mag., vol. 52, no. 5, pp. 7190, 2010. [29] A. K. Bhatacharyya, On the convergence of MoM and mode matching solutions for innite array and waveguide problems, IEEE Trans. Antennas Propag., vol. 51, pp. 15991606, Jul. 2003. [30] D. M. Pozar, Microwave Engineering. New York: Addison-Wesley Inc., 1990.

Alireza Foroozesh (M08) received the B.Sc. degree from Tehran Polytechnic, Tehran, Iran, in 1996, the M.Sc. degree from the Iran University of Science and Technology, Tehran, in 1999, and the Ph.D. degree from the University of Manitoba, Winnipeg, Manitoba, Canada, in 2007, all in electrical engineering. From May 2000 to July 2002, he was a Researcher with the Antenna Laboratory, Iran Telecommunication Research Center (ITRC), where he was involved in projects related to antenna design and measurement. He is currently a Postdoctoral Fellow with the Department of Electrical and Computer Engineering, Concordia University, Montreal, Canada. His main research interest is the analysis and modeling of periodic structures and their applications to antennas and microwave systems. Dr. Foroozesh was the recipient of the Best Student Paper Award at the International Symposium on Antennas and Propagation (ISAP 2007) in Niigata, Japan. He received a Young Scientist Travel Grant at ISAP 2007, Niigata, Japan, and a Young Scientist Award at the Electromagnetic Theory Symposium (EMTS 2007), Ottawa, Canada.

FOROOZESH AND SHAFAI: ON THE CHARACTERISTICS OF THE HIGHLY DIRECTIVE RESONANT CAVITY ANTENNA

91

Lotfollah Shafai (LF07) received the B.Sc. degree from the University of Tehran, Tehran, Iran, in 1963 and the M.Sc. and Ph.D. degrees from the University of Toronto, Toronto, ON, Canada, in 1966 and 1969, all in electrical engineering. In November 1969, he joined the Department of Electrical and Computer Engineering, University of Manitoba as a Sessional Lecturer, becoming an Assistant Professor in 1970, Associate Professor in 1973, and Professor in 1979. Since 1975, he has made special efforts to link the University research to the industrial development, by assisting industries in the development of new products or establishing new technologies. To enhance the University of Manitoba contact with industry, in 1985 he assisted in establishing The Institute for Technology Development and was its Director until 1987, when he became the Head of the Electrical Engineering Department. His assistance to industry was instrumental in establishing an Industrial Research Chair in Applied Electromagnetics at the University of Manitoba in 1989, which he held until July 1994. Dr. Shafai has been a participant in nearly all Antennas and Propagation Symposia and participates on the Review Committees. He is a member of URSI Commission B and was its Chairman during 1985-88. In 1986, he established the

Symposium on Antenna Technology and Applied Electromagnetics, ANTEM, at the University of Manitoba that is currently held every two years. He has been the recipient of numerous awards. In 1978, his contribution to the design of a small ground station for the Hermus satellite was selected as the 3rd Meritorious Industrial Design. In 1984, he received the Professional Engineers Merit Award and in 1985, The Thinker Award from Canadian Patents and Development Corporation. From the University of Manitoba, he received the Research Awards in 1983, 1987, and 1989, the Outreach Award in 1987 and the Sigma Xi, Senior Scientist Award in 1989. In 1990, he received the Maxwell Premium Award from the Institution of Engineering and Technology (London) and in 1993 and 1994, the Distinguished Achievement Awards from Corporate Higher Education Forum. In 1998, he received the Winnipeg RH Institute Foundation Medal for Excellence in Research. In 1999 and 2000, he received the University of Manitoba, Faculty Association Research Award. He is an elected Fellow of the IEEE since 1987 and was elected Life Fellow in 2007. He is a Fellow of The Royal Society of Canada in 1998. He was a recipient of the IEEE Third Millennium Medal in 2000 and in 2002 was elected a Fellow of The Canadian Academy of Engineering and Distinguished Professor at The University of Manitoba. He holds a Canada Research Chair in Applied Electromagnetics and is International Chair of URSI Commission B.

92

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Nature of Cross-Polarized Radiations from Probe-Fed Circular Microstrip Antennas and Their Suppression Using Different Geometries of Defected Ground Structure (DGS)
Chandrakanta Kumar, Member, IEEE, and Debatosh Guha, Senior Member, IEEE
AbstractExperiments with probe-fed circular patches using conventional and defected ground planes ashed some interesting features relating to cross-polarized (XP) electric elds and associated radiations before the present authors. Those led to a series of new investigations for understanding the nature of XP elds and to deal with them using defected ground structure (DGS) for improved XP performance. In the rst phase of investigation, the XP radiations of a probe-fed circular patch with conventional ground plane have been critically studied as a function of the radial probe location. Remarkably signicant effect is experimentally demonstrated. New information about orthogonal resonant elds and its importance in designing an antenna is provided. In the second phase of investigation, limitations of dot-shaped DGS in reducing XP level are experimentally studied. As its improved variants, two new DGS geometries such as annular ring and circular arcs have been explored. The arc-DGS appears to be highly efcient in terms of suppressing XP elds. Suppression by 1012 dB has been experimentally demonstrated. Each design has been experimented in both C- and X-bands to earn condence on the measured data. Index TermsCross-polarized radiation, defected ground structure (DGS), microstrip antenna.

I. INTRODUCTION IRCULAR microstrip patch excited with dominant mode radiates linearly polarized elds along the broadside and the plane of polarization is determined by an imaginary vertical plane (xz-plane) passing through the line joining the feed and the centre of the patch. Therefore, ideally no y-component (orthogonal to xz-plane) of electric eld (E-eld) is supposed to be present. But orthogonally polarized E-elds have been evident since the early experiments, e.g., [1] and are designated as the cross-polarized (XP) radiations. It is typically about 25 dB down relative to the co-polarized (CoP) gain in E-plane, but increases considerably in H-plane,

Fig. 1. Probe-fed circular microstrip patch with circular dot-shaped DGS. (a) View from ground plane side and (b) cross-sectional view.

Manuscript received November 15, 2010; revised June 05, 2011; accepted July 15, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. C. Kumar is with Communication Systems Group, ISRO Satellite Centre, Department of Space, Government of India, Bangalore 560017, India (e-mail: kumarchk@ieee.org). D. Guha is with the Institute of Radio Physics and Electronics, University of Calcutta, Kolkata, India (e-mail: dguha@ieee.org). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167921

particularly showing peaks around 30 40 on either side of bore sight. The source of XP elds was theoretically analyzed in [2] and rst higher order mode was attributed to this phenomenon. The theory considered innite ground plane (GP) and showed absolutely no XP component in E-plane. This does not agree with a practical antenna. Recently, Guha et al. [3] have successfully applied defected ground structure (DGS) to suppress the XP radiation from a circular patch. They have indicated orthogonal component of dominant mode as one of the XP producing factors, also discussed in Section III. They employed a pair of circular dotshaped defects on y-axis with their centers aligned to the patch boundary (Fig. 1) and conjectured that the defects would affect boundary conditions against setting in any orthogonal resonant elds. They demonstrated about 5 dB reduction in XP level and reported measured results in S-band. The authors addressed this problem with twofold objective: (i) to verify whether dot-DGS works consistently in other frequency bands [4] and (ii) whether the size, shape and location of the DGS could be modied to achieve improved XP characteristics. But it is interesting to note that new results show some

0018-926X/$26.00 2011 IEEE

KUMAR AND GUHA: NATURE OF CROSS-POLARIZED RADIATIONS FROM PROBE-FED CIRCULAR MICROSTRIP ANTENNAS

93

additional features over and above those obtained in [3]. The XP elds are found to react with the same DGS in different ways for different feed positions. Minor change in feed position is identied as major controlling parameter for XP radiating elds. Limitation of the dot-DGS in improving CoP to XP isolation has also been established experimentally. This, at the same time, provoked us to explore newer DGS geometries to achieve improved suppression in H-plane. This paper actually presents thorough investigations having two important objectives. First one is to understand the XP behavior in probe-fed circular patches though a series of experiments. The change in XP producing modes with the change in probe location has been individually studied and discussed. To the best of our knowledge, such comprehensive experimental evidence leading to understanding the source of XP elds is documented here for the rst time. The second objective is to alleviate the limitations of dot-DGS using other geometries. Two DGS geometries have been conceived with the idea of introducing relatively extended area of interaction between the defects and the XP elds. Small circular dot, therefore, has been replaced by an annular ring-shaped DGS resulting in much reduced XP elds. Further reduction has been explored and experimentally achieved replacing annular ring by circular arcs. Using arc-DGS, peak XP level in H-plane pattern can be changed from 13 dB to 24 dB, indicating relative suppression by 11 dB. Physical insights into its working principle are also discussed. The DGS congurations have been experimented with two sets of prototypes operating in C- and X-bands, respectively. More condence on applicability of arc-DGS has been earned by testing the same for elliptic patch geometry, operating at identical frequency band. Signicant results have been documented. II. PROBE LOCATION AND ITS SIGNIFICANCE The nature of electromagnetic (EM) elds resulting in XP radiations from a conventional circular patch is investigated in this section with the help of a series of experiments along with simulated eld portrays obtained using [5]. The location of the feeding probe has been discovered as an important factor. A. The Prototypes Three 18 mm diameter identical patches etched on RT duroid 5870 substrates measuring have been fed by R125.464.270W SMA probes with central conductor of 1.3 mm diameter. The location of the feed is determined by radial distance from the center of the patch and in here, three different values have been chosen on the basis of the impedance matching. Those were computed accurately using formulations in [6] and veried using simulated results [5]. The value appears to be optimally matched location. Two more offset values ( and ) are chosen on either side of the matched location, maintaining considerable matching with minima near 15 dB. B. Simulated and Measured Results The prototypes have been measured using Agilents E8363B network analyzer and an automated anechoic chamber. Mea-

Fig. 2. Measured and simulated characteristics of conventional circular , , patches with different probe locations. Parameters: , , ground plane: .

Fig. 3. Simulated radiation patterns of conventional circular patches with different feed locations examined in Fig. 2. Parameters as in Fig. 2. (a) E-plane, (b) H-plane.

sured and simulated characteristics with respect to three different feed locations are shown in Fig. 2. Measurement with indicates considerably good matching of input impedance. But or 3.2 mm are identical in terms of input impedance showing minima around 15 dB. All three are found to resonate around 5.9 GHz. The feeds are, therefore, located at about , , and distance from the centre of each patch, being wavelength corresponding to the operating frequency. The radiation characteristics are examined near 5.9 GHz. The simulated results are depicted in Fig. 3. The co-polarized ra-

94

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 4. Measured radiation patterns of the prototypes with different feed locations examined in Figs. 2 and 3. Parameters as in Fig. 2. (a) E-plane, (b) H-plane.

diations in either principal plane are identical for all values. Let us look at the XP levels, which in H-plane, show minor change, particularly in terms of symmetry around for particular probe locations. Only E-plane XP levels appear to be greatly affected by varying feed positions. Optimally matched feed results in the lowest XP level, which considerably increases as the feed moves on either side. This variation in E-plane is qualitatively corroborated by the measurements as shown in Fig. 4. Measured E-plane XP values show an increase along bore sight by 7 dB as the feed moves from optimally matched location towards the patch edge , and by 12 dB as it moves towards the patch centre . But no major change is revealed in H-plane, excepting asymmetry around the bore sight. Further investigations are needed to justify the above phenomenon. It is quite apparent that elds producing XP radiations behave in different ways for different values. This is examined in the following section. C. Orthogonal Resonant Fields Simulated electric elds in the substrate are portrayed in Fig. 5 for three different feed locations. These emphasize distribution of weak elds over yz-plane particularly, i.e., orthogonal to the principal plane of resonance. Fig. 5(a) is for optimally matched location. Signicant E-elds are vertically unipolar near patch edges. These resemble rst higher order mode [7]. To easily visualize the fact, modal eld and current distributions in a circular patch resonator are shown in

Fig. 5. Simulated electric eld in the substrate indicating the variation along : elds vertically unipolar; (b) : elds yz-plane: (a) : elds of opposite polarity. Other paof opposite polarity; (c) rameters as in Fig. 2.

Fig. 6. The co-polarized nature of weak elds around the patch edge effectively results no radiation along broadside. Rather, they are responsible to contribute to oblique radiation in H-plane. The experiments presented in Fig. 4, exactly corroborate the same. E-plane XP level is nearly 40 dB down with respect to the CoP peak value, which in H-plane is only 20 dB. This observation fully agrees with Lees theoretical approach [2]. But Fig. 5 (b) and (c) obtained for and 2.25 mm, respectively, reveal a different prole of orthogonal elds. The E-elds concentrating near patch edges are vertically polarized with opposite polarity and resemble weak mode on H-plane. This corresponds to Fig. 6(c). We, therefore, prefer to call them orthogonal component of dominant mode (OCDM), which should contribute to radiation along broad side. This is exactly reected from both the Figs. 3 and 4. The OCDM is prone to be excited where the feed is weakly coupled in terms

KUMAR AND GUHA: NATURE OF CROSS-POLARIZED RADIATIONS FROM PROBE-FED CIRCULAR MICROSTRIP ANTENNAS

95

Fig. 6. Electric eld and current distributions in isolated circular patches: mode (rst higher order mode), (b) mode (dominant mode), (a) surface current, (c) Orthogonal component of dominant mode (OCDM). and : Z-polarized electric elds lying in between patch and ground plane; white dot on patch indicates feed location.

of launching energy fully to the primary resonant mode. Thus a fraction of the input energy tends to excite similar weak resonance orthogonally. This OCDM, which is not usually accounted for, cannot be neglected in calculating XP radiations from a circular patch. Oblique XP radiation in H-plane looses symmetrical nature as the feed shifts from optimally matched to slightly offset position (Figs. 3(b)). Indeed weak OCDM then gets excited and corresponding E-eld distribution over H-plane is not symmetrically uniform. This is clearly observed from Fig. 5(b) and (c) where electric eld vectors are marked as vertically upward and vertically downward, which indeed cause asymmetry in H-plane XP patterns. The rst higher mode appears to be a constant factor in any probe fed circular patch. This happens in case of and 3.2 mm; presence of mode over and above OCDM is not apparent in Fig. 5. For comparative study, all the eld portrays in Fig. 5 are captured in same scale of intensity. Highest intensity is represented by red and the lowest by blue , and those in gray scale, appear as deep black and faint gray, respectively. The size of the eld vectors also carries information about the magnitude and as such one can easily surmise that the electric elds in Fig. 5(c) are stronger than those in Fig. 5(b). This exactly agrees with simulated and measured data shown in Figs. 3 and 4. III. DOT- SHAPED DGS INVESTIGATED WITH C-BAND PATCH FOR VARYING PROBE LOCATIONS The DGS was rst introduced to suppress XP radiation from a circular microstrip patch by the same research group [3]. They used circular dot-shaped defects as shown in Fig. 1. The idea was to perturb the boundary conditions, which weakens any resonance setting in orthogonally, but hardly affects the primary radiating elds. In [3], a single prototype operating at S-band and fed at was examined and about 5 dB reduction in XP value was reported. In here, same DGS geometry has

Fig. 7. Photographs of one prototype with circular dot-DGS. (a) Radiating side, , , , (b) ground plane side. Parameters: , , , ground plane: .

been investigated in C-band under varying condition of feed locations in light of the observations in Section II. A. The Prototypes The patches, investigated in Section II, use normal GP with feeding probe at , 2.8 mm, and 3.2 mm, respectively. The same congurations have been investigated here with dot-DGSs as schematically shown in Fig. 1. The defect diameter has been optimized using simulated results, where the basis of optimization was achieving optimum suppression in XP values keeping the primary radiating elds as well as the matching of input impedance unaffected. The optimum defect diameter appears to be , which is about , being the operating wavelength of radiating patch. The dimension in terms of is very close to that examined in [3] with . Photographs of the prototype with is shown in Fig. 7. B. Measured and Simulated Results . The investigation starts with the patch with Fig. 8(a) shows measured radiation patterns with and without DGS. Simulated data have been incorporated to verify the measurements. The co-polarized radiations with and without DGS are identical in either plane. The XP values in H-plane also do not show signicant change. But a considerable reduction in E-plane is evident. Similar observation for has

96

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 8. Measured and simulated radiation characteristics of dot-DGS integrated patches fed at different radial distances. , (c) , Other parameters as in Fig. 7. (b)

: (a)

also been documented in Fig. 8(b). The order of suppression is higher in Fig. 8(b) than that in 8(a). The studies in Section II have indicated that for both and 2.25 mm, OCDM results in high XP radiations in E-plane. The defects indeed weaken the OCDM and thus reduce the XP eld intensity. Measured suppressions also corroborate the observations in Section II. The rst higher mode is hardly affected by the geometry of dot-DGS showing almost no effect in H-plane. But it is interesting to note that the same DGS responds in a different way to a patch with (optimally matched). The results are shown in Fig. 8(c). E-plane XP increases with the dot-DGS. Indeed, no OCDM is apparent in Fig. 5(a) for and the rst higher mode predominates. This increased XP values with DGS may be attributed to eld components, which are predominantly distributed around patch edges surrounding the dot-defects and get diffracted. It is thus

revealed that, small sized defect like dot cannot stop or weaken mode, predominantly occurring for optimally matched feed. This, in turn, indicates the need of relatively larger defect to take care of mode and to suppress XP elds in H-plane. Above studies indicate the limitations of dot-DGS and as such new extended geometries have been explored in the following sections. IV. ANNULAR RING-SHAPED DGS To alleviate above limitation of dot-DGS, annular ring-shaped defect has been explored for a circular radiating patch. Fringing electric elds under a patch follow the patch curvature and hence an annular ring has been conceived as an effective tool, which compared to small dot, should provide larger interaction area for mode and can help in reducing XP radiation in H-plane. Annular ring-DGS was employed by this group to reduce mutual coupling between two

KUMAR AND GUHA: NATURE OF CROSS-POLARIZED RADIATIONS FROM PROBE-FED CIRCULAR MICROSTRIP ANTENNAS

97

Fig. 10. Prototype with annular ring-shaped DGS operating in X-band. , , , , , , , ground plane: .

Fig. 9. Circular microstrip patch with annular ring-DGS. (a) Cross-sectional view and (b) bottom view.

dielectric resonator antennas [8], and has recently been tried as a XP suppressing structure [9]. Extensive measurements supported by simulated data are presented here for the rst time. A. The Geometry and Prototypes A schematic diagram of ring-DGS integrated circular microstrip patch is shown in Fig. 9. The ring-DGS is characterized by two parameters and . Inner radius of the ring is . The ring is located concentrically with respect to the patch. Measurements presented in Section II are in C-band. To use those data for comparison purpose, ring-DGS integrated patch has also been studied at the same frequency in C-band. This is also extended to X-band to conrm its operation in different frequencies. In C-band design, the patch is optimally matched keeping other parameters as in previous sections. The ring parameters have been optimized keeping in view that it considerably reduces H-plane XP, as well as minimally affects the primary radiation elds. This is one of the primary considerations in the design and the parameters are determined as , , using simulated results [5]. This indicates that the inner diameter of the ring is larger than that of the patch. The X-band prototype uses the same PTFE substrate as in C-band and hence changes in terms of operating wavelength. For operating frequency , the patch radius is 5 mm. The patch is fed by a standard SMA probe with at the optimally matched location . The GP size is , i.e., about . Optimized DGS parameters are determined as and . Photograph of X-band prototype is shown in Fig. 10.

Fig. 11. Measured and simulated versus frequency of two prototypes using , ring-DGS. Probes located at optimally matched point. (a) C-band: , , , , , , ground plane: ; (b) X-band: Parameters as in Fig. 10.

B. Measured and Simulated Results Each set of prototypes with and without DGS has been measured using the same experimental setup as mentioned in Section II-B. Fig. 11 shows versus frequency plots of C-band and X-band prototypes with and without annular-DGS. Presence of DGS shows no considerable change in resonant frequency. Actually a DGS is designed such a way that it does

98

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 13. Measured and simulated radiation patterns for X-band antennas with and without ring-DGS. Feed located at optimally matched point. Parameters as . (a) E-plane, (b) H-plane. in Fig. 11(b); Fig. 12. Measured radiation patterns for C-band Prototypes with and without ; parameters: as in ring-DGS. Feed optimally matched with . (a) E-plane, (b) H-plane. Fig. 11(a),

not perturb the primary resonant elds. Thus it should have minimal effect on the input impedance, reection coefcient and the primary radiation characteristics. At rst, C-band results are discussed. Its radiation characteristics obtained at 5.9 GHz are shown in Fig. 12. Those due to identical patch with normal GP are also compared. Presence of ring-DGS shows no change in co-polarized patterns around bore sight in either plane. But in contrary to the dot-DGS, considerable suppression in H-plane XP level is evident. Suppression by about 5 dB has been experimentally demonstrated and this results up to 25 dB isolation between CoP and XP levels. E-plane XP level normally appears below 30 dB; neither any improvement, nor degradation is caused by the ring-DGS. The defect in ground plane shows some unavoidable back radiation, which is more pronounced in H-plane and appears beyond . The radiation patterns obtained with X-band prototypes are presented in Fig. 13. Gross observation is similar to those discussed above for C-band antenna. Here, CoP to XP isolation is about 17 dB. Figs. 12 and 13 examine the radiations near the center of impedance bandwidth. The performance of the DGS over the entire band has been experimentally studied. A set of representative results obtained at X-band is shown in Fig. 14 indicating consistently suppressed XP values over the entire bandwidth.

Fig. 14. Measured XP radiations in H-plane circular patches with and without ring-DGS measured in X-band.

V. CIRCULAR ARC-SHAPED DGS During optimizing the size and location of annular ring, it is observed that its smaller radius increases proximity of the defect to the patch boundary, which in turn improves the XP performance. This proximity increases interaction between the defect and the fringing-elds. This favors reducing XP and at the same time, disturbs primary co-polarized radiating elds causing considerable back radiation through defects as revealed from Fig. 12. A trade off is therefore conceived. The ring has been truncated over small regions around E-plane so that the primary radiating elds get free from any detrimental effect by the DGS.

KUMAR AND GUHA: NATURE OF CROSS-POLARIZED RADIATIONS FROM PROBE-FED CIRCULAR MICROSTRIP ANTENNAS

99

Fig. 17. Simulated and Measured characteristics of arc-DGS integrated circular patch operating in X-band. Parameters as in Fig. 16, feed located at optimally matched point.

Fig. 15. Circular microstrip patch with arc-shaped DGS. (a) Cross-sectional view and (b) bottom view.

Fig. 16. Prototypes with arc-shaped DGS operating in X band. , , , , , , ground plane: .

, ,

Thus the shape has been evolved as a pair of arcs as shown in Fig. 15. Now a designer has a liberty to change value without affecting E-plane fringing elds. A. The Geometry and Prototypes Circular arc-DGS, conceived as an improved variant of annular ring-DGS, incorporates an additional parameter depicted in Fig. 15. The parameter denes angular width of the arc. Like ring-DGS, the new structure is also examined at two identical frequencies in C- and X-bands. Identical substrate has been used. Feed locations are same as Section IV. The arc-parameters g, , and have been optimized for C-band as , , and , with , and for X-band as , , and , with . It is interesting to note that, optimum values are very close

Fig. 18. Measured and simulated radiation patterns for X-band antennas with and without arc-DGS. Feed located at optimally matched point. Parameters as . in Fig. 16. (a) E-plane, (b) H-plane.

to each other. So this value of seems to be optimum for any circular patch. Photograph of X-band prototype is shown in Fig. 16. B. Measured and Simulated Results Possibility of achieving reduced XP using arc-DGS has been indicated by the authors in a recent letter [10]. We, therefore, refrain from repeating those results in here. New results obtained in X-band are discussed with a view to conrming its applicability over a wide frequency range. Fig. 17 shows the operating

100

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 19. Measured XP radiations in H-plane of circular patches at X-band with and without arc-DGS at different frequencies over the impedance bandwidth.

Fig. 21. Measured and simulated results obtained with the elliptic patch shown versus frequency with normal and defected ground in Fig. 20: (a) Measured . planes; (b) H-plane radiation patterns obtained at

Fig. 20. Prototype of Elliptic patch integrated with arc-shaped DGS. (a) Viewed from top, (b) viewed from ground plane side. , , , , , inner boundary of defect 0.25 mm away from patch boundary, substrate and ground plane parameters as in Fig. 16.

frequency of the X-band prototype and its radiation characteristics near 10 GHz are shown in Fig. 18. Remarkable improvement in reducing H-plane XP level is revealed. Peak level of XP using a conventional GP is about 13 dB down the CoP peak value, which in presence of arc-DGS comes down to 24 dB. Reduction by about 11 dB is experimentally evident. The valley region showing more than 30 dB isolation in H-plane pattern is found to be widened by additional 20 on either side of bore sight. E-plane XP level is normally 30 dB down and the DGS causes no further change. Copolarized patterns with and without

DGS are very similar. No inuence of the DGS is apparent. Signicant back radiation, as evident in Fig. 12 is absent in Fig. 18. The H-plane XP values obtained experimentally with C-band patch indicate suppression by about 12 dB, although not shown here. Fig. 19 examines the XP patterns in particular at different frequencies over the impedance bandwidth. Arc-DGS consistently performs over the entire bandwidth. Further investigation has been executed to verify its effectiveness for different patch geometry. Elliptic patch is chosen as the closest variant of circular geometry [11] and identical arc-DGSs have been examined. Major to minor axis ratio is considered as 1.3. Minor axis is aligned to the plane of resonance, i.e., E-plane, and its dimension is set equal to obtain resonance near 10.05 GHz. Arc-DGS of identical value (=1.5 mm) with have been used. The prototype operating in X-band is shown in Fig. 20. Measured reection coefcients of the elliptic patch with and without DGS are shown in Fig. 21(a). Presence of DGS shifts the resonant frequency from 10.05 GHz to 10.02 GHz. H-plane radiation patterns with and without DGS are shown in Fig. 21(b). Simulated data are also compared with the measurements. DGS shows expected improvements in XP levels. Reduction by about 11 dB is revealed from both measurements and simulated results.

KUMAR AND GUHA: NATURE OF CROSS-POLARIZED RADIATIONS FROM PROBE-FED CIRCULAR MICROSTRIP ANTENNAS

101

VI. CONCLUSION After such an extensive investigation, the concluding remarks may be divided into two parts, rstly on conventional circular patch and its XP behavior and secondly, on the new DGS geometries suitable for reducing XP levels. A designer should be very careful to calculate the probe location in view of the best possible matching of input impedance. This always gives the lowest XP values in either principal plane. If anyone asks about the tolerance in choosing the feed point, the answer after above study would be almost zero. Even 0.5 mm in C-band makes a signicant change in both the nature and magnitude of XP radiations. Annular ring-DGS is denitely a better choice compared to dot-DGS, but arc-DGS appears to be the best of all in all respects. Arc-DGS suppresses XP level in H-plane by 1012 dB and at the same time does not affect the copolarized elds. Increase in back radiation is also insignicant. The performance of the proposed DGSs has been veried in two different frequency bands and hence they are expected to be equally effective for all circular patches. These DGSs should nd several applications where polarization purity of antenna is an important aspect, e.g., wireless base station network, high quality EMI sensors, linearly polarized array etc. ACKNOWLEDGMENT The authors are extremely thankful to the reviewers for their constructive criticism, suggestions, and help in improving the quality of the paper. REFERENCES
[1] Y. T. Lo, D. Solomon, and W. F. Richards, Theory and experiments on microstrip antennas, IEEE Trans. Antennas Propag., vol. 27, pp. 137145, 1979. [2] K. F. Lee, K. M. Luk, and P. Y. Tam, Cross-polarization characteristics of circular patch antenna, Electron. Lett., vol. 28, no. 6, pp. 587589, Mar. 12, 1992. [3] D. Guha, M. Biswas, and Y. M. M. Antar, Microstrip patch antenna with defected ground structure for cross polarization suppression, IEEE Antennas Wireless Propag. Lett., vol. 4, pp. 455458, 2005. [4] C. Kumar and D. Guha, A new look into the cross-polarized radiation form a circular microstrip antenna and suppression using dot-shaped DGS, presented at the IEEE Antennas and Propagation Symp., 2010. [5] High Frequency Structure Simulator (HFSS) Ansoft, ver. 11.1. [6] D. Guha, Y. M. M. Antar, J. Y. Siddiqui, and M. Biswas, Resonant resistance of probe-and microstrip-line-fed circular microstrip patches, IEE Proc. Microwave Antenna Propag., pp. 481484, 2005. [7] R. Garg, P. Bhartia, I. Bahl, and A. Ittipiboon, Microstrip Antenna Design Handbook. Boston: Artech House, 2001, ch. 5. [8] D. Guha, S. Biswas, T. Joseph, and M. T. Sebastian, Defected ground structure to reduce mutual coupling between cylindrical dielectric resonator antennas, Electron. Lett., vol. 44, no. 14, pp. 836837, Jul. 3, 2008.

[9] C. Kumar and D. Guha, New defected ground structures (DGSs) to reduce cross-polarized radiation of circular microstrip antennas, presented at the IEEE Applied Electromagnetic Conf., Kolkata, India, Dec 1416, 2009. [10] D. Guha, C. Kumar, and S. Pal, Improved cross-polarization characteristics of circular microstrip antenna employing arc-shaped defected ground structure (DGS), IEEE Antennas Wireless Propag. Lett., vol. 08, pp. 13671369, 2009. [11] M. C. Bailey and M. D. Deshpande, Analysis of elliptical and circular microstrip antennas using moment method, IEEE Trans. Antennas Propag., vol. 33, pp. 954959, 1985. Chandrakanta Kumar (M09) was born in Shibpur, W.B., India in 1976. He received the M.Tech. degree in radio physics and electronics from the University of Calcutta, India, in 2001. He then joined Communication Systems Group of ISRO Satellite Centre, Bangalore, India, as an Engineer. Since then he is actively involved in design and development of antenna systems of the Indian space programme and related ground stations. He served as project manager, antenna systems, for rst Indian mission to the Moon; Chandrayaan-1. Presently he is holding similar responsibility for GSAT-12, ASTROSAT and Chandrayaan-2 missions. He has about 20 publications in international journals and conferences to his credit. His areas of interest are DGS integrated microstrip patch antennas, DR antennas, lightweight antennas for spacecraft, and MEMS. Mr. Kumar is a recipient of the Young Scientist Award2009 from the Indian Space Research Organization (ISRO). He is also a member of the team that received the Team Excellence Award2008 of ISRO for his contribution in Chandrayaan-1 antenna systems.

Debatosh Guha (SM03) received the B. Tech., M. Tech., and Ph.D. degrees in radio physics and electronics from the University of Calcutta, in 1986, 1988 and 1994, respectively. He started his professional career as an Engineer at Webel Telecommunication Industries Limited, in 1989. In 1990, he joined the Institute of Radio Physics and Electronics, as a Senior Research Fellow of the Council of Scientic and Industrial Research (CSIR), India. In 1994, he joined the same University as a Lecturer in Radio Physics and Electronics. Currently, he is a Professor in the Institute of Radio Physics and Electronics, University of Calcutta, India. He was a Visiting Research Professor and Visiting Professor in the Electrical and Computer Engineering Department, Royal Military College of Canada, Kingston, ON, Canada, for different periods. He also visited several foreign universities and research institutes which include the University of Houston, University of Bath, Queen Mary College University of London, Communication Research Centre of Canada, etc. He has published more than 125 technical papers and a book entitled Microstrip and Printed Antennas: New Trends, Techniques and Applications (Wiley, 2010). His current research interest include application of defected ground structures (DGS) printed and dielectric resonator antennas, analysis and characterization of printed antennas for wireless communications, UWB dielectric resonator antennas, and uncommon DRA modes for MIC applications. Dr. Guha is a recipient of Jawaharlal Nehru Memorial Fund Prize and URSI Young Scientist Award, 1996. He is the Founding Chair of the IEEE AP-MTT Calcutta Chapter, and also served as General Chairs/Program Chairs at various International Symposia and Conferences in his eld. He is on the Board of Reviewers of several international journals including the IEEE TRANSACTION ON ANTENNAS AND PROPAGATION, IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, IET Microwave, Antennas and Propagation, Electronic Letters, Radio Science, etc.

102

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Single, Dual and Tri-Band-Notched Ultrawideband (UWB) Antennas Using Capacitively Loaded Loop (CLL) Resonators
Chia-Ching Lin, Associate Member, IEEE, Peng Jin, Member, IEEE, and Richard W. Ziolkowski, Fellow, IEEE
AbstractTwo compact, printed, ultrawideband (UWB) monopole antennas with tri-band notched characteristics are reported. The notched lters are achieved by introducing printed, electrically small, capacitively-loaded loop (CLL) resonators. The directly driven elements consist of printed top-loaded CLL-based microstrip feed lines. By adding three CLL monopoles and 50 elements close to the feed line, band-notch properties in the WiMAX (3.33.6GHz), lower WLAN (5.155.35GHz) and higher WLAN (5.7255.825GHz) bands are achieved. Each antenna system is contained on a sheet of Rogers Duroid 5880 substrate. One is designed with three additional CLL elements; the other is achieved with only two. Comparisons between the simulation and measurement results show that these UWB antennas have broadband matched impedance values and stable radiation patterns for all radiating frequencies. Index TermsAntenna efciency, antenna pattern, antennas, metamaterials, ultrawideband (UWB).

I. INTRODUCTION LTRAWIDEBAND (UWB) communication systems have become an attractive wireless topic since the Federal Communication Commission (FCC) rst approved in 2002 the rules for the utilization of the 3.110.6 GHz unlicensed band for commercial UWB communications [1]. The high demands on such communication systems have stimulated research into many UWB antenna designs. Nonetheless, antenna designs for UWB applications face many challenges including their impedance matching, radiation stability, compact size, low manufacturing cost and electromagnetic interference (EMI) problems. The EMI problems are quite serious for UWB systems since there are several other existing narrowband services which occupy frequency bands within the designated UWB

Manuscript received March 17, 2011; revised June 08, 2011; accepted July 15, 2011. Date of publication September 15, 2011; date of current version January 05, 2012 This work was supported in part by DARPA Contract HR0011-05-C-0068 and in part by ONR Contract H940030920902. C.-C. Lin was with the Department of Electrical and Computer Engineering, University of Arizona, Tucson, AZ 85721-0104 USA. He is now with Ruckus Wireless, Sunnyvale, CA 94085 USA. P. Jin was with the Department of Electrical and Computer Engineering, University of Arizona, Tucson, AZ 85721-0104 USA. He is now with Broadcom Corporation, Irvine, CA 92617-3038 USA. R. W. Ziolkowski is with the Department of Electrical and Computer Engineering, University of Arizona, Tucson, AZ 85721-0104 USA (e-mail: ziolkowski@email.arizona.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167947

bandwidth. These include world interoperability for microwave access (WiMAX) service from 3.3 to 3.6 GHz; wireless local area network (WLAN) services such as IEEE 802.11a in the USA (5.15 to 5.35 GHz, 5.725 to 5.825 GHz) and HIPERLAN/2 in Europe (5.15 to 5.35 GHz, 5.47 to 5.725 GHz). To mitigate any interference with these coexisting systems, it is necessary to introduce a UWB antenna that has intrinsic ltering properties at their service frequencies. To minimize the footprint of the antenna system, the signal processing requirements, and the cost, it is highly desirable for those lters to be intrinsically handled rather than through any additional external band-stop lter devices. There are several methods with which one can achieve a band-notched UWB antenna. The most popular approach is to embed different shaped slots in the radiating element or in its ground plane. Examples include U-shaped, H-shaped or C-shaped slots [2][6]. However, most UWB antennas have been designed with only one or two notched-frequency bands. On the other hand, the complimentary split ring resonator (CSRR) structure, for instance, was proposed recently to reject more than one of the unwanted frequency bands [7][9]. Although those designs are low prole, achieve stable radiation patterns, and have constant gain, the lower WLAN band (5.155.35 GHz) was not rejected successfully. Two tri-band-notched UWB antennas are proposed in this paper. One design utilizes three electrically small resonators, each being realized as a capacitively-loaded loop (CLL). The CLL element was introduced originally to achieve a mu-negative metamaterial that acts as an articial magnetic conductor for low prole antenna applications [10]. Each CLL element has a high-Q characteristic and a compact size, making it a very suitable candidate for a band-stop lter function. Like the split ring resonator (SRR) element [11], the CLL element is self-resonant and has a resonance frequency that is determined primarily by its loop inductance and the capacitances resulting from the cuts which open the loop. In contrast, the CLL element has a much simpler, more compact design. It will be demonstrated that by placing one, two or three CLL elements near the feedline and tuning their sizes, one can control the band-notched frequencies of the radiator, while minimizing their space requirements, to achieve single, dual, and tri-band notched-lter UWB antennas. The approach is analogous to the lter methods that introduce various types of metamaterial unit cell elements into transmission lines [12]. A second design employs a CLL-based slot element in the top-loaded monopole element and two CLL elements located along the feedline. It overcomes the tuning

0018-926X/$26.00 2011 IEEE

LIN et al.: SINGLE, DUAL AND TRI-BAND-NOTCHED UWB ANTENNAS USING CLL RESONATORS

103

Fig. 2. Comparisons of the measured and simulated VSWR values for the baseline UWB antenna shown in Fig. 1.

Fig. 3. Geometry of the single band-notched UWB antenna. All dimensions , , , , are in mm: , , and .

Fig. 1. Geometry of the baseline UWB antenna. (a) top view; (b) bottom view. , , , , All dimensions are in mm: , , , , , , , , , , and .

sensitivities associated with the rst design. Comparisons between the ANSOFT nite element method (FEM) based high frequency structure simulation (HFSS, v. 12.1) results and experiments for both designs show very good agreement. Initial results from these investigations were reported in [13]. II. UWB ANTENNAS WITH AND WITHOUT BAND-NOTCHED FILTERS A. Baseline UWB Design Fig. 1(a) and 1(b) show the top and bottom views of the baseline design. It is a top-loaded CLL-based UWB antenna that is implemented with Rogers Duroid 5880 board material. The latter has a relative permittivity , loss tangent , 0.5 oz electrodeposited copper (17 thickness) and 31 mil thickness. Its overall size is ; it is fed by a microstrip line to achieve a 50 input impedance when connected by an SMA

connector to the source. The SMA connector is included in the HFSS model. The radiating portion of the antenna associated with the microstrip feed line is on one side (front) of the board; the conducting ground plane is on the other (back) side. Both the radiating monopole and the ground plane have curved shapes to help tailor the input impedance to a 50 source, to increase the radiation efciency and to help produce smooth transitions from one resonant mode to another. This characteristic ensures a good impedance match (i.e., ) over the entire UWB frequency range, 3.110.6 GHz. The copper thickness and realistic values of its conductivity were included. The nal optimization parameters and the measured and simulated VSWR values of the proposed UWB antenna are shown in Fig. 1 and Fig. 2, respectively. It is below 2 not only over the entire UWB band, but also at even higher frequencies. B. Single Band-Notched UWB Antenna Design In order to reduce the EMI with the WiMAX band, a band-notched function covering the interval 3.33.6 GHz is desired. The proposed single-notched UWB antenna is illustrated in Fig. 3. Notice that the gap region of the element has an augmented design to enhance its capacitance. By placing the element close to the feed line, this resonator is strongly coupled to it. It captures and stores all of the input energy at

104

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I COMPARISONS BETWEEN SIMULATIONS AND THEORETICAL PREDICTIONS OF THE BAND-NOTCHED FREQUENCY

its resonance frequency and thus creates a single band-notched frequency lter. Note that with this design, there is no need to change the dimensions of the original UWB antenna. Rather, one simply needs to tune the resonance of the element by adjusting its dimensions to achieve the desired band-notched function. Moreover, this electrically small resonator has a minimal impact on other frequencies. A parametric study of the element provides a necessary understanding of the interaction between the main radiator and the element. The band-notched frequency is given approximately by the expression (1) where is the total length of the element, is the effective dielectric constant, and is the speed of the light. Given a desired resonance frequency, one can use (1) to dene the initial total length of the element for an initial design. One can then use numerical simulations to adjust the dimensions of the element to obtain the nal design. Table I shows a comparison between the simulated and theoretical predictions of the resonance frequency corresponding to different total lengths of the element. Note that the ability to further tune the CLL element by modifying its gap region is not only advantageous for achieving a specic band-notched function, but it also allows one to shrink the overall size of the CLL element. This provides one with the ability to add extra parasitic elements in a limited space and, hence, leads to a smaller-sized antenna without a loss in the desired functionality. The HFSS-predicted VSWR values for the single-notched UWB antenna are shown in Figs. 4 and 5, respectively, for different lengths and sizes of the gap in the CLL element. Clearly, one can see that the length of the element and the size of its gap play critical roles in dening the band-notched frequency. For instance, the band-notched frequency decreases from 3.927 to 3.559 GHz as the length increases from 29.0 to 32.0 mm. Similarly, because the capacitance decreases as the CLL gap size, , increases, the band-notched frequency increases from 3.342 to 3.703 GHz as is varied from 0.5 mm to 2.0 mm. Note, however, that differences between the theoretical and simulated results in Table I also occur because of the coupling between the feedline and the CLL element. When the element is close to the feed line, the increased coupling produces extra capacitance, which then leads to a slightly lower band-notch frequency. A comparison of the VSWR values for several gap sizes between the element and the feed line, , is given in Fig. 6. One nds that as the element is moved closer to the feed

Fig. 4. HFSS-predicted VSWR characteristics for various total lengths of the CLL element. The gap size is xed at 0.5 mm.

Fig. 5. HFSS-predicted VSWR characteristics for various CLL element gap . The total length of the CLL element is xed at 32.0 mm. sizes,

Fig. 6. HFSS-predicted VSWR values for various gap sizes between the of the CLL element and CLL element and the feedline. The total length are xed at 32.0 mm and 0.5 mm, respectively. its gap size

line, a signicant increase in the coverage of the band-notched frequency response occurs. For instance, the coverage of the band-notched frequency increases from 0.216 GHz to 1.108 GHz as decreases from 0.45 mm to 0.15 mm. Consequently, the location of the element also provides a means to adjust the coverage of the band-notched lter. These results further illustrate that when changes, there is only a small effect on the band-notch frequency, e.g., the shift is only 0.1 GHz in Fig. 6. We note that having control of the coverage of the band-notch lter is very important to achieving the desired, practical performance characteristics of a band-notched UWB antenna. The nal dimensions of the element for the single band-notched frequency design, as shown in Fig. 3. The

LIN et al.: SINGLE, DUAL AND TRI-BAND-NOTCHED UWB ANTENNAS USING CLL RESONATORS

105

Fig. 7. HFSS-predicted surface current distributions at different frequencies. (a) 3 GHz, (b) 3.342 GHz, and (c) 10.0 GHz.

Fig. 9. The HFSS-predicted overall efciency and maximum realized gain for the single band-notched UWB antenna.

Fig. 8. Measured and HFSS-predicted VSWR values for the single band-notched UWB antenna.

HFSS-predicted current distributions at 3.0, 3.342 and 10.0 GHz are shown in Fig. 7. The currents ow mainly on the element at the notch frequency 3.342 GHz, while the currents on the radiating elements (the feedline and the monopole) at that frequency are very small. In contrast, the currents on the feedline and the monopole are large and negligible on the element away from the frequency band of the effective notched-lter. The measured and simulated VSWR values versus frequency of the single band-notched UWB antenna are compared in Fig. 8. The measured bandwidth of the band-notched is from 3.25 to 3.62 GHz. As expected, the desired band-notched frequency is introduced by the element. Additionally, when compared to the original UWB antenna, the single band-notched UWB antenna successfully blocks the WiMAX band (3.33.6 GHz) and still maintains good impedance matching at the rest of the UWB band. The HFSS-predicted overall efciency and maximum realized gain values are shown in Fig. 9 as a function of the frequency. As designed, signicant drop of these values is observed within the rejection band. C. Dual Band-Notched UWB Antenna Design In addition to the WiMAX services from 3.33.6 GHz, the lower WLAN (5.155.35 GHz) band may also cause interference within the UWB band. To avoid EMI in this band, the dual band-notched UWB antenna and the optimized design paramaters for the element shown in Fig. 10 was investigated. It is achieved simply by adding an additional CLL element near the feed line; i.e., the element is introduced on the other side of the feed line. To design a notch-band lter for the lower

Fig. 10. Geometry the dual band-notched UWB antenna. All dimensions are in , , , , , , mm: . and

WLAN frequency set, one can still use (1) as the initial design to predict the approximate length of the element. Even though the element is added near the feed line, it was found (rather attractively) that the dimensions of the main radiator and element remain the same. Therefore, each CLL element is acting independently and, consequently, one obtains two band-stop lters at the different, desired frequencies. The measured and simulated VSWR values versus frequency for the dual band-notched UWB antenna with two CLL elements (one on each side of the feedline) are compared in Fig. 11. Good agreement is observed. The measured dual band-notched frequencies cover 3.233.70 GHz and 5.095.58 GHz. The HFSS-predicted overall efciency and maximum realized gain values versus frequency are shown in Fig. 12. They show the anticipated dramatic drop in these values at both band-notched frequency bands. D. Tri-Band Notched UWB Antenna Design Besides the WiMAX and lower WLAN bands, the higher WLAN band also operates within the UWB band. Again, to re-

106

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 11. Comparison of the measured and simulated VSWR values of the dual band-notched UWB antenna.

Fig. 13. The tri-band notched UWB antenna. (a) Geometry, and (b) design di, , , mensions. All dimensions are in mm: , , , , , , , , , , .

Fig. 12. The HFSS-predicted overall efciency and maximum realized gain for the dual band-notched UWB antenna. Fig. 14. Comparison of the measured and simulated VSWR values of the triband notched UWB antenna.

duce the potential EMI effects, a tri-band notched design was obtained. As with the single and dual band cases, the tri-band design begins with (1) to obtain the approximate length of the additional CLL element. However, the available area near the feed line limits where one can put a third CLL element. Nevertheless, as mentioned above, the benet of the CLL element is its exible tuning ability, i.e., the size of the gap region in the CLL element provides an extra degree of freedom to obtain the band-notched frequency within a specic area. Therefore, we simply shorten the total length of the element and decrease the gap size simultaneously to not only maintain the desired band-notched frequency, but also provide the extra area to add the additional element to realize the third band-notched frequency lter. Additionally, since the higher frequency WLAN band is narrower than the lower one, the gap between the element and the feed line should also increase. Again, the size of the main radiator and the ground plane size remain the same. Moreover, the dimension of the element also remains the same. Thus, the only difference between the dual and tri band-notched designs is to change the dimensions of the element in order to add the element. The proposed tri-band-notched UWB antenna and the correspondingly optimized parameters are shown in Fig. 13. Good agreement between the HFSS-predicted and the measured VSWR values versus frequency is shown in Fig. 14 for the

Fig. 15. The HFSS-predicted overall efciency and maximum realized gain for the tri-band notched UWB antenna.

tri-band notched UWB antenna. The measured tri-band notches cover the bands: 3.293.67 GHz, 5.125.35 GHz and 5.675.83 GHz. As shown in Fig. 15, the antenna gain and radiation efciency signicantly decrease over these three band-notched frequency sets. The HFSS-predicted current distributions shown in Fig. 16 demonstrate that the currents ow around each CLL

LIN et al.: SINGLE, DUAL AND TRI-BAND-NOTCHED UWB ANTENNAS USING CLL RESONATORS

107

Fig. 16. HFSS-predicted current distributions at each band-notched frequency (a) 3.342 GHz (b) 5.279 GHz (c) 5.821 GHz.

Fig. 18. Geometry of the tri-band notched UWB antenna with one slot CLLbased element in the top-loaded monopole and two CLL elements near its feed, , , , line. All dimensions are in mm: , , , , , , , , , , , , , , , , , , , and .

The HFSS-predicted and measured far-eld radiation patterns for the proposed tri-band notched UWB antenna at 3.1, 5.5 and 10.6 GHz are shown in Fig. 17. Note that if the radiating element is assumed to be printed on the xy-plane, its radiation pattern is x-polarized because its monopole lies along the x-direction. Therefore, the E-plane is the xy-plane, while the H-plane is the yz-plane. The results shown in Fig. 17 demonstrate, as anticipated, that the H-plane patterns are omni-directional in all cases, while the E-plane patterns similarly exhibit the expected monopole-like behaviors. Distortions in the E-plane patterns begin to occur at higher frequencies because the radiating elements are no longer small relative to those wavelengths. III. ALTERNATIVE TRI-BAND NOTCHED UWB ANTENNA DESIGN
Fig. 17. The comparison between HFSS-predicted and measured radiation patterns (dB), E-Plane (left) and H-Plane (right), for the tri-band notched UWB antenna. (a) 3.1 GHz, (b) 5.5 GHz, (c) 10.6 GHz. Solid line: simulation; dashed line: measurement.

element only at the corresponding band-notch frequency. Additionally, one observes a small coupling between the and elements. Since the frequencies of the lower and higher WLAN bands are very close, less coupling would be desirable. One way to reduce this coupling is to increase the distance between the and elements. One could, for example, shorten the total length of each CLL element and decrease the gap size in order to have a much smaller sized CLL element and then increase the distance between the and elements. Unfortunately, this simple approach requires a much narrower gap in both the and elements; it leads to difculties due to limitations of the fabrication techniques.

Recall that while the coupling between the and elements in the tri-band notched design shown in Fig. 13 was small, the dimensions of those elements still had to be retuned to achieve the desired band-notched frequencies. As a consequence, the design becomes sensitive to the associated small gap regions needed to maintain their resonance frequencies while being smaller in size. Thus, it requires some effort to tune the entire system properly. To remove this sensitivity issue, we developed an alternative tri-band notched UWB antenna, shown in Fig. 18. It has one CLL-based slot element embedded in the top-loaded monopole and two additional CLL elements near the feedline. The idea is to tune the CLL-based slot element for the longer wavelength, lower frequency rejection band and then have the two, basically independent CLL elements near the feedline create the second and third band-notched lters. As the widths and lengths of this slot element are adjusted, its band-notch frequency can be tuned. This is illustrated with the single-notch version shown in Fig. 19. The measured and

108

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 19. Measured and simulated VSWR values for the single-notched UWB antenna. A CLL-based slot element is introduced into the top-loaded monopole to achieve the desired band-notched response at 3.35 GHz. The surface current distribution at this rejection frequency is localized near the slot element.

Fig. 21. The HFSS-predicted overall efciency and maximum realized gain for the tri-band notched UWB antenna displayed in Fig. 18.

IV. CONCLUSIONS Two CLL-based tri-band notched UWB antennas were introduced in this paper. They were obtained by using either three additional CLL elements or by embedding a CLL-based slot element into the top-loaded monopole and adding two CLL elements near the feed line. Single- and dual-notched band antennas were also considered to explain the performance characteristics of the tri-band designs. Parametric studies of all of these antennas provided guidelines on how to control not only the band-notched frequencies, but also the bandwidth of the rejected frequencies as well. The tri-band notched design with the three CLL elements had some sensitivities to achieve the band-notched characteristics for the closely adjacent frequencies associated with the lower and higher WLAN band that were overcome with the other design. By introducing the CLLbased slot element in the top-loaded monopole, the area limitations near feed line and the couplings introduced from the close proximity of two CLL elements were avoided. This design thus provided three independently tunable band-notch sets. The comparisons between the measured and simulated values for both tri-band notched UWB designs veried their predicted performance characteristics, including stable radiation patterns, high gain and radiation efciencies, and broadband matched impedance values for all radiating frequencies REFERENCES
[1] Federal Communications Commission Revision of Part 15 of the Commissions Rules Regarding Ultra-Wideband Transmission System from 3.1 to 10.6 GHz Federal Communications Commission. Washington, DC, ET-Docket, 2002, pp. 98153, FCC. [2] W. S. Lee, D. Z. Kim, K. J. Kim, and J. W. Yu, Wideband planar monopole antennas with dual band-notched characteristics, IEEE Trans. Microw. Theory Tech., vol. 54, pp. 28002806, Jun. 2006. [3] S. J. Hong, J. W. Shin, H. Park, and J. H. Choi, Analysis of the bandstop techniques for ultrawideband antenna, Microw. Opt. Technol. Lett., vol. 49, pp. 10581062, May 2007. [4] X. L. Bao and M. J. Ammann, Printed band-reject UWB antenna with H-shaped slot, in Proc. IEEE IWAT Workshop, Mar. 2007, pp. 319322. [5] Q. Chu and Y. Yang, A compact ultrawideband antenna with 3.4/5.5 GHz dual band-notched characteristics, IEEE Trans. Antennas Propag., vol. 56, pp. 36373644, Dec. 2008.

Fig. 20. Measured and simulated VSWR values versus frequency for the triband notched UWB antenna displayed in Fig. 18.

HFSS-predicted VSWR values are given along with the surface current distribution at 3.35 GHz, its band-notch frequency. Most of the currents are seen to ow around the gap region of the CLL-based slot element. This embedded CLL-based design provides a more exible tuning ability, particularly within the limited area close to the feed line. The two additional CLL elements are again located on opposite sides of the feedline. They were initially designed using (1). The nal optimization parameters are shown in Fig. 18. Fig. 20 gives a performance comparison between the measured and simulated values of the VSWR versus frequency for the fabricated antenna. The measured tri-band notches cover 3.29.3.72 GHz, 5.075.40 GHz and 5.625.93 GHz. As shown in Fig. 21, the antenna gain and radiation efciency signicantly decrease at these three band-notched frequencies and have high values away from them. We note that the measured upshifts in the lowest band shown in both Figs. 19 and 20 were consistent and found to be caused by fabrication errors. The fabricated slot widths were simply narrower than specied. We had emphasized the fabrication tolerances of the CLL elements with the vendor (Prototon Circuits, Tucson, AZ) anticipating that those of the larger slots would not be an issue. While this choice led to the very good agreement at the higher two-bands shown in Figs. 19 and 20, it clearly impacted the lower frequency band results.

LIN et al.: SINGLE, DUAL AND TRI-BAND-NOTCHED UWB ANTENNAS USING CLL RESONATORS

109

[6] H. Zhang, R. Zhou, Z. Wu, H. Xin, and R. W. Ziolkowski, Designs of ultra wideband (UWB) printed elliptical monopole antennas with slots, Microw. Opt. Technol. Lett., vol. 52, pp. 486471, Feb. 2010. [7] J. Kim, C. S. Cho, and J. W. Lee, 5.2 GHz notched ultra-wideband antenna using slot-type SRR, Electron. Lett., vol. 42, pp. 315316, Mar. 2006. [8] J. Liu, S. Gong, Y. Xu, X. Zhang, C. Feng, and N. Qi, Compact printed ultra-wideband monopole antenna with dual band-notched characteristics, Electron. Lett., vol. 44, pp. 710711, Jun. 2008. [9] J. Ding, Z. Lin, and Z. Ying, A compact ultra-wideband slot antenna with multiple notch frequency bands, Microw. Opt. Technol. Lett., vol. 49, pp. 30563060, 2007. [10] A. Erentok, P. Luljak, and R. W. Ziolkowski, Antenna performance near a volumetric metamaterial realization of an articial magnetic conductor, IEEE Trans. Antennas and Propag,, vol. 53, pp. 160172, Jan. 2005. [11] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, Magnetism from conductors, and enhanced non-linear phenomena, IEEE Trans. Microwave Theory Tech., vol. 47, pp. 20752084, Nov. 1999. [12] R. Marqus, F. Martn, and M. Sorolla, Metamaterials With Negative Parameters: Theory, Design and Microwave Applications. Hoboken, NJ: Wiley, 2008. [13] C.-C. Lin and R. W. Ziolkowski, Tri-band notched ultra-wideband antenna using capacitively loaded loops, presented at the Int. Symp. on Antennas and Propagation, Toronto, Canada, Jul. 2010, paper 408.4. Chia-Ching Lin (S07A11) received the B.Sc. and M.Sc. degrees in electronic engineering from the National Taiwan University of Science and Technology (NTUST), in 1999 and 2003, respectively, and the ECE Ph.D. degree from the University of Arizona, Tucson, in 2010. He is currently with Ruckus Wireless, Sunnyvale, CA. His research interests include electrically small antennas and metamaterial applications to antenna designs.

Peng Jin (S05M10) received the EE B.Sc. degree from the University of Science and Technology of China, Heifei, in 1999, the EE M.Sc. degree from the North Dakota State University, Fargo, in 2004, and the ECE Ph.D. degree from the University of Arizona, Tucson, in 2010. He is currently with the Signal Integrity Group, Broadcom Corporation, Irvine, CA. His research interests include electrically small antennas and metamaterial applications to antenna designs.

Richard W. Ziolkowski (M87SM91F94) received his Sc.B. degree in physics, magna cum laude with honors, from Brown University (1974), his M.S. (1975) and Ph.D. (1980) degrees in physics from the University of Illinois at Urbana-Champaign. He was a member of the Engineering Research Division at the Lawrence Livermore National Laboratory from 1981 to 1990 and served as the leader of the Computational Electronics and Electromagnetics Thrust Area for the Engineering Directorate. He currently is serving as the Litton Industries John M. Leonis Distinguished Professor in the Department of Electrical and Computer Engineering at the University of Arizona. He holds a joint appointment with the College of Optical Sciences. He is an IEEE Fellow and an OSA Fellow. He was the President of the IEEE Antennas and Propagation Society in 2005. He continues to be very active in the IEEE, OSA, and APS professional societies. He has served on the International Advisory Boards and Technical Program Committees of several international conferences, including iWAT, ISAP, Metamaterials, and META. His research interests include the application of new physics and engineering ideas to linear and nonlinear problems dealing with the interaction of electromagnetic waves with complex media, metamaterials, and realistic structures.

110

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Leaky Wave Enhanced Feeds for Multibeam Reectors to be Used for Telecom Satellite Based Links
Andrea Neto, Senior Member, IEEE, Mauro Ettorre, Member, IEEE, Giampiero Gerini, Senior Member, IEEE, and Peter De Maagt, Fellow, IEEE

AbstractThe use of dielectric super-layers for shaping the radiation pattern of focal plane feeds of a multibeam reector system is discussed. Using the super-layers, it is possible to reduce the spillover from the reectors without increasing the dimension of each aperture. The effect has been demonstrated using a realistic array conguration. The experimental demonstration is obtained with congurations which are typical for satellite based multi beam telecommunication links. Thanks to the adoption of a 4 reuse scheme, based on frequency and polarization orthogonal channels, the measured edge of coverage directivity has improved, with respect to a standard technology case, by 2 dB over an operational bandwidth of about 2%. Index TermsLeaky waves, leaky-wave antennas, reector antenna feeds, reector antennas.

I. INTRODUCTION

that radiate incrementally as they propagate between the ground plane and a super-layer. In particular, [4] demonstrated that an increase of edge of coverage gain of 1.3 dB over a bandwidth (BW) of about 6%, can be obtained for a dual reector system characterized by mod) in erate values of the Focal distance to Diameter ratios ( the order of 0.85. The achievement was demonstrated only theoretically since it was assuming the presence of a ltering feeding network to facilitate frequency and polarization orthogonality of neighboring beams. In this contribution the same strategy based on the use of super-layers is rst adapted to the design of reratios which are typical ector systems that use larger for satellite based communications links. Second, the new dedicated design is experimentally validated by hardware demonstrations of the dielectric super-layers, the feeds and the ltering structures. A. Requirements The starting point of this paper is a telecommunication satellite-based multi beam ( ) reector system for spot coverage of the earth. The beams will be arranged in an hexagin the order of onal lattice and separated by an angle . To achieve such beam separation and negligible performance degradation for the outer beams, the of the system should be in the order of 1.5 to 2. The separation between the (where is the wavelength at the feeds will be about central operating frequency ) which is typical for existing telecom multibeam systems. The selected focal distance is ap. The target relative frequency bandproximately width will be in the order of 2%, and inside this band a minimal edge of coverage directivity of 44.5 dBi should be guaranteed. Note that the edge of coverage gain/directivity is dened as the gain/directivity at the cross over between three adjacent beams: . In order to achieve the orthogonality between neighboring beams a 4 re-use scheme will be used that is very similar to the one that was described in [4]. The scheme is based on a hybrid frequency and polarization reuse. The total frequency band and cenis divided in two sub-bands of widths and , with a roll off tered at that separates the two useful bands. To sepband arate the two bands, multiple cavity band pass lters are connected to the waveguide feeds. The feeds will be polarized either vertically or horizontally, by properly shaping the irises.

RESENT and next-generation telecommunication satellite systems often require a multiple beam capability. A simple way to achieve high edge of coverage gain is to use a single aperture and a single feed per beam. However, as claried in the introduction of [1], this method may lead to several decibels (dBs) of spillover losses. In the series of works [2][4] an approach based on the use of dielectric super-layers to enhance the radiation properties of small apertures in order to excite reector antenna systems has been proposed. The same approach has also been applied to completely planar technology in [5]. The basis for the results presented in [2][5], can be traced back to the original paper of von Trentini [6]. In recent years, several other authors have also followed a similar route [7][12]. This way of proceeding [2] makes use of the shaping of the radiation of leaky waves, patterns via the excitation of a pair

Manuscript received July 29, 2010; revised January 12, 2011; accepted July 20, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. A. Neto is with the Telecom Department, the Technical University (TU) Delft, 2628 CD, Delft, The Netherlands (e-mail: a.neto@tudelft.nl). M. Ettorre is with the Groupe Antennes & Hyperfrquences, Institut dElectronique et de Tlcommunications de Rennes (IETR), UMR CNRS 6164, Universit de Rennes 1, 35042 Rennes Cedex, France. (e-mail: mauro. ettorre@univ-rennes1.fr). G. Gerini is with the TNO Defence, Security and Safety, Den Haag 2597 AK, The Netherlands (e-mail: giampiero.gerini@tno.nl). P. De Maagt is with the Electromagnetics Division, European Space Agency, 2200 AG, Noordwijk, The Netherlands (e-mail: Peter.de.Maagt@esa.int). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167909

0018-926X/$26.00 2011 IEEE

NETO et al.: LEAKY WAVE ENHANCED FEEDS FOR MULTIBEAM REFLECTORS TO BE USED FOR TELECOM SATELLITE BASED LINKS

111

Fig. 1. Hexagonal grid array with periodicity d of circular metallic waveguide D mounted on a innitely extended ground plane.

After the implementation of the re-use scheme the isolation between neighboring beams will be at least 12 dB (roughly corside lobe levels). The dimension of the responding to feed array will be minimized in order to minimize the weight of the entire front end. B. Benchmark An array based on standard waveguide horn technology cannot reach the target 44.5 dB edge of coverage directivity. However, in order to establish a benchmark to which to compare the performances of the leaky wave enhanced arrays, such a standard waveguide array is investigated rst (Fig. 1). It consists of an hexagonal grid array composed of 19 circular waveguide horns mounted on a nite ground plane of dimensions ( ). The reference system has the z axis normal to the ground plane. The periodicity of the array is set to . The circular apertures are obtained by aring to a the feeding waveguides of diameter diameter of , over a taper of length . With this arrangement the apertures cover almost the entire central part of the ground plane. The length of the taper greatly inuences the efciency of the feeds. If the length can be made arbitrarily long (innite taper) the performance of the feeds are optimal with each one of them presenting an aperture efciency in the order of 80%. In that case the apertures are illuminated with essentially uniform phase [13]. Almost independently from the taper length the and the mutual coupling coefcients, reection coefcients, of each of the waveguides can be designed to be lower than and over the entire BW, respectively. Such an array is appropriate to feed reector systems characterized by ratios. large To accurately evaluate the performance of a reector system fed by an array of waveguides, the far elds after the reector need to be calculated. Throughout this paper a dual-offset Gregorian antenna, satisfying the Mizugutchi [14] condition is assumed (to avoid the blockage of the feed array). As claried in [4], in rst approximation the design of the system is still and focal distance , espebased on equivalent diameter is relatively large. The geometry of the cially when the system is shown in Fig. 2. Which also highlights the blockage that the over-dimensioned focal plane array manufactured in this activity would imply.

Fig. 2. Schematic view of a dual-offset Gregorian reector conguration. Dimensions are for operation at 10 GHz. The possible blockage due to the oversized prototype array developed in this work is highlighted.

Fig. 3. Edge of coverage directivity as a function of the frequency associated to each element of the iris loaded waveguide array of Fig. 4. The selected antenna :  :  and focal distance F geometry has the diameter D (corresponding to = : ). The three curves refer to the directivity simulated for the free space reference case with in phase eld distribution, to the directivity simulated in double dielectric conguration and to the one from the prototype for the double dielectric conguration.

= 143 6 F D = 1 71

= 245 5

The secondary elds are obtained using a simplied Physical Optics code based on the equivalent parabola. The algorithm uses as inputs the far elds radiated by the reference array, which in turn have been calculated using Microwave Studio CST [15]. Finite array effects are negligible. Fig. 3 shows among other curves the edge of coverage directivity as a function of the frequency, calculated from the radiation patterns predicted by the reference waveguide array in which the horns present in phase distributions. The study is for a reector having , and . Only the diameter corresponding to a ratio of 1.71 is considered in Fig. 3. The edge of coverage directivity obtained from the reference array is

112

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 5. Comparisons of the directivity patterns in the four signicant planes and at frequency of 9.95 GHz for three different feeds: the iris loaded waveguide operating in free space, the iris loaded waveguide operating in the presence of a : , and in phase circular aperture of diameter single dielectric slab of  :  .

22

= 20 25

Fig. 4. (a) Real array with orthogonal polarized and dual frequency (black and grey colors) feeds to realized a 4 reuse scheme; (b) the equivalent array that is obtained when considering only the same frequency same polarization feeds.

43.5 dB and essentially constant over the frequency band considered. Using reectors characterized by larger diameters does not help as narrower beams are associated to larger roll offs and consequently the edge of coverage gain saturates. Also, even if not reported here, the isolation between neighboring beams is within specications. Although other horns than smooth walled ones could be used, [16], this simple idealized horn is considered a good starting point. II. PATTERN SYNTHESIS BASED ON LEAKY WAVE ENHANCEMENT In [3] and [4], it is explained how to use a super-layer conguration to enhance the radiation from each aperture in an imaging array conguration. In those articles it was claried that the maximum achievable directivity enhancement was obtained when the mutual coupling with the closest feed was in the order (or lower). In fact, due to the rst couple of of leaky waves, in leaky wave enhanced congurations the neighboring elements contribute out of phase with the center one. When applying the design considerations described in [3], [4] to the problem at hand, it follows that an appropriate single di. The electric super-layer would be characterized by corresponding array conguration is shown in Fig. 4(a), and it is very similar to the one that was presented in [3]. In the absence of a re-use scheme the mutual coupling between neigh, which is too boring elements would be in the order of high. Accordingly a 4 reuse scheme needs to be used. When the frequency and polarization orthogonality is introduced the low-frequency and vertically polarized array in Fig. 4(a) is electrically equivalent to the one in Fig. 4(b). The closest effective (same frequency and polarization) waveguides are at a distance which corresponds to mutual coupling in the of

Fig. 6. Dielectric stratication corresponding to the double layer design.

order of . Given that each iris-loaded waveguide is characterized by width , one can observe that eventually the area of each equivalent unit cell is about 50 times larger than the dimension of each waveguide. The primary patterns from an array conguration as shown in Fig. 4(b) turn out to be clean and symmetric in all cuts. In order to visualize the impact of the dielectric super-layers in the present context, the predicted directivity patterns for a dielectric in the four signicant planes slab characterized by are reported in Fig. 5 at the frequency of 9.95 GHz. These patterns are also compared with the ones that would be achieved by the reference (in phase) circular apertures, and by the iris loaded waveguides without the super-layer enhancement. It is apparent that the super-layer enhanced patterns are much more directive than the ones without super-layers and thus, over a small frequency range more suited to feed efciently a re. ector with large In the rest of this paper a conguration with double dielectric slabs of lower dielectric constant electrically equivalent to will be considered. This a single dielectric layer of is because the prototype demonstrator was manufactured using more readily available lower dielectric constant material. The target conguration is described schematically in Fig. 6. The , and thickness two slabs of dielectric constant , with , are located at a distance one from the other. The lowest slab is at a distance from the ground plane. All the dimensions are evaluated at the central operating frequency assumed to be 10 GHz. Fig. 3 shows the edge of coverage directivity as a function of the frequency

NETO et al.: LEAKY WAVE ENHANCED FEEDS FOR MULTIBEAM REFLECTORS TO BE USED FOR TELECOM SATELLITE BASED LINKS

113

Fig. 7. Simulated S-parameters pertinent to a single iris loaded waveguide in isolation and to a three element array where the elements are at distance 4:8 one from the other.

for this preferred twin dielectric slab conguration. Also reported in the same gure are the expected edge of coverage directivities for the reference in phase eld horn array that ts in the array cell. This conguration is considered an idealized conguration against which to compare. With respect to the free space reference case, the dielectrically enhanced structure provides an increase in the edge of coverage directivity of more than 1.5 dB over a 2.1% BW (from 9.86 to 10.1 GHz).

III. FURTHER BW CONSTRAINTS Once the edge of coverage directivity against bandwidth trade off has been investigated the actual useful BW evaluation will eventually be further reduced by the considerations on the impedance matching of the structure and the shape of the primary and the secondary patterns. A. Impedance Matching The highly resonant super-layer structure has a strong impact on the input impedance of each feed. The BW is reduced and this effect is directly related to the gain enhancement at broad side. This effect has been studied in [12] rst and then by the authors of this paper with the specic reector application [2]. The simulated parameters associated to a double dielectric structure are shown in Fig. 7. The shown parameters are pertinent to a single waveguide in isolation (loaded by a double iris) and to a three waveguide array where the elements are at distance one from the other. One can see that the of the single element and the one in the array are essentially the same and that the mutual coupling between the neighboring elements are in the frequency interval dashed in grey indibelow cating the desired operating bandwidth 9.875 10.1 GHz. Note also that in this bandwidth the reection coefcient of each aperture is below . This implies a mismatch loss in the order of 0.5 dB at the beginning of the useful band.

Fig. 8. Simulated primary beam directivities generated by each iris loaded feed in array environment are plotted as a function of the frequency calculated using CST. (a) E-plane. (b) H-plane. (c) D-plane.

B. Primary Patterns The simulated primary radiation patterns (directivities) from each waveguide feed are shown in Fig. 8(a)(c) as a function of the frequency from 9.8 to 10.1 GHz, in the -, - and -plane,

114

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I PARAMETERS CHARACTERIZING THE SECONDARY BEAMS FROM EACH FEED, F=D

= 1:71

Fig. 9. Secondary beams (directivities) generated by a reector (F=D : , F :  ), when excited by the radiation patterns of Fig. 8 are plotted as a can be estimated function of the frequency. Note that the isolation of : . by plotting the adjacent beam centered at 

= 245 5

= 1 71

013 dB = 1 12

respectively. Note that the directivities do not include the mismatch losses, which are the main cause of losses at lower frequencies. It is apparent that for higher frequencies the beams atten and that the peak directivity is reduced. This is because the overall enhanced patterns are shaped by leaky waves whose pointing directions are further away from broadside as the frequency increases. This phenomenon is more evident at frequencies as high as 10.1 GHz, where the leaky wave contribution . This incan be recognized in each -cut around duces lower directivity at broad side. The fact that the primary beam degrades has a negative impact because the edge of the reector, located at 17 , is illuminated with relatively higher eld levels with respect to axial direction. This is not a big problem in terms of spill over, because the largest portion of the power is still intercepted by the reector. In fact the edge of coverage remains high, as was apparent from Fig. 3. However, this results in higher side-lobes in the secondary patterns. Fig. 8 also highlights the good pattern symmetry of the structure with cross polarized elds remaining low for all azimuthal angles [17]. C. Secondary Patterns The simulated secondary beams generated by a reector ( , ), when excited by the radiation patterns of Fig. 8 are plotted as a function of the frequency in Fig. 9. Apart from the edge of coverage gain, one of the key parameters that quanties the quality of secondary patterns is the isolation between beams using the same polarization and same frequency. A rst order approximation can be obtained assuming that all the beams are identical to the central one. In this case it is relevant to consider the isolation between the main beam and its rst image centered at two angular periods . This exercise is also summarized in of distance Fig. 9 for the highest frequency in the band which appears to be the one that presents relatively high side lobes associated

Fig. 10. Back-view picture of the feed array manufactured and mounted. It includes the dielectric super-layers, the iris ground plane, the waveguide feeds and the band-pass lters.

with the beam degradation (splitting) that was observed in the primary beams. In Table I, the most signicant system parameters for the investigated structure are summarized, considering . IV. PROTOTYPE MANUFACTURING Even though the motivation of the work was a telecom scenario at Ka-band, a prototype at the scaled frequency of 10 GHz has been designed. A prototype array of 19 waveguide feeds was manufactured in the TNO internal workshop. A picture of the nal hardware is shown in Fig. 10 which highlights the waveguide feeds, the ground plane, the super-layers and the ltering structures. The waveguide feed connecting the lters to the radiating ground plane, is only 17 mm long. The dielectric super-layers (60 cm 60 cm) were custom made on borosilicate glass by ECN (http://www.ecn.nl/home/). The dielectric layers were over dimensioned in order for the edge effects to be completely negligible. Since the structure is resonant with overall relative bandwidth of about 2%, the separation between the ground plane and the rst sheets, nominally , should be accurate to 0.015 mm or less in order not to observe a signicant frequency shift in the measured parameters. This accuracy is comparable to the one required for the lters. The thickness of the slabs received was veried to be . The separation, , beas requested, tween the two dielectric slabs was frozen to its nominal value . at the design frequency of 10 GHz, Details of the slabs are visible in Fig. 11. The dielectric con, stant of the received material was measured to be accurate to the second signicant digit, with very small imaginary part. To allow for small variations in the dielectric constant,

NETO et al.: LEAKY WAVE ENHANCED FEEDS FOR MULTIBEAM REFLECTORS TO BE USED FOR TELECOM SATELLITE BASED LINKS

115

Fig. 13. Picture of the ensemble of the lters after the mounting. Fig. 11. Front and side view of the feed array manufactured and mounted. On the top right the micro-metric tuning mechanism is visible, while in the bottom right the two dielectric layers are explicitly visible.

Fig. 12. Picture of the pass band lter. The structure is milled from two halves of solid steel and screwed together.

micro-metric screws were used to ne tune the distance between the ground plane and the ensemble of the two dielectric slabs so that the radiation frequency band would be in correspondence with the pass band frequencies of the lters. A. Filtering Structures The lters are realized in waveguide technology. The total frequency band is divided in two sub-bands of widths and centered at and , with a roll off band that separates the two useful bands. A picture of one of the low-frequency band pass lters before mounting is shown in Fig. 12. The lters are realized by cascading ve cavities realized in WR 90 waveguides. The lters are obtained by machine milling solid steel blocks. Although this is not a manufacturing procedure that is considered for space ight hardware, it is representative for the RF prototype. Overall 19 lters were manufactured. A close view of the
Fig. 14. Measured S parameters from the two different lters: L stands for low-frequency lter while H stands for high-frequency lter. (a) Amplitudes of all four parameters and (b) phases of the input coefcients only, (S ).

mounted lters is shown in Fig. 13. The measured parameters of both lters are shown in Fig. 14. The smooth variation of the phase of the reected signals in the reection bandwidths is also presented in Fig. 14(b). It is apparent that the band pass behavior of the two lters is very well behaved, with reection coefcients in band lower and out off band rejections lower than . It than appears that with respect to the desired design bands, the present technological implementation of the lower band lter presents a

116

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 15. Element numbering of the measured feed elements. Elements 1 and 5 are active in the same polarization but in the lower and the higher frequency ranges, respectively. Fig. 16. E-plane copolar primary feed radiation patterns associated to the central element (1), operating in the lower band, for different tuning heights.

5 MHz shift toward higher frequencies. A relative shift of 0.05% appears unavoidable with the technology used at TNO and in absence of tuning screws. Overall the realization of the lters was considered satisfactory. V. MEASURED PERFORMANCES After the lters were characterized as a self standing unit, the system (lters+feeds+super-layers) was assembled so that the performance of the entire primary feed structure could be evaluated. Before freezing the actual measurement geometry the separation between the metallic ground plane and the lowest slab, . In order to decide the most was ne tuned around appropriate distance two sets of measurement of the copolarized radiation patterns at two frequencies were done. With respect to Fig. 15 the elements chosen for the ne tuning were the rst (1) and the fourth(4). The far eld patterns were obtained by post processing after the near eld was sampled with a commercial near eld scanner. Fig. 16 shows the low-frequency (9.9 GHz) E-plane copolar primary feed radiation patterns associated to the central element for different tuning heights. It is apparent that for all heights the radiation patterns have very similar shapes. The measured gain, in this graph, includes the lters and the mismatch losses. For the smallest height, the gain is almost 2 dB lower than for the higher values, highlighting the effects of the mismatch losses, and also the dependence of these losses as a function of frequency. The response of the higher frequency radiator (4) is complementary. The patterns at 10 GHz are shown in Fig. 17. It is apparent that the higher slabs give rise to much reduced gain at broadside which in turns implies higher side lobe from the secondary patterns as was discussed in Section III. Eventually it was decided that the best compromise distance to evaluate the system performance of the array was the one that was measured as 14.93 mm. This choice reduces slightly the actual gain in the lower band due to mismatch losses but guarantees that at the highest frequency the isolation of the beams is still within specication.

Fig. 17. E-plane copolar primary feed radiation patterns associated to a lateral element (4), operating in the higher band, for different tuning heights.

A.

-Parameters

Once the entire array was assembled and the -value frozen the parameters of the waveguide array, with the inclusion of parameters for two the lters were measured. The measured feeding waveguides operating at two different central frequencies are reported in Fig. 18. The feeds considered are the ones indicated as 1 through 7 with reference to Fig. 15. The results indicate that on the lower band the reection coefcients are approximately ( ) which correspond to about 15% of efciency reduction ( ). In the high band the reection coefcients are in the order of , about 3% of efciency reductions. The mutual coupling between ports (after ltering) was lower than . This means mutual coupling can be neglected for the system.

NETO et al.: LEAKY WAVE ENHANCED FEEDS FOR MULTIBEAM REFLECTORS TO BE USED FOR TELECOM SATELLITE BASED LINKS

117

Fig. 18. Measured S parameters pertinent to the array.

B. Measured Patterns Figs. 19 and 20 show the measured primary co and cross polar radiation patterns from element 1 and 5, respectively. The Ludwig 3 denition has been used for dening co and cross polarization [18]. The two gures are pertinent to the , , ( ) planes, respectively and are shown for different frequencies all within the band allowed by the lower or higher frequency lter, respectively. The gures show that the radiation patterns are very symmetrical, characterized by very low cross-polarization and that the frequency behavior is essentially as predicted. However, a difference of 1.5 dB in the gain at the lower frequencies is observed between Fig. 19 and the directivities calculated curves in Fig. 8. The present implementation in fact includes 1 dB of losses associated to the ohmic losses in the lter and the return loss. The in the latter is frequency dependent and amount to up lower frequency band. Once the mismatch and ohmic losses are included in the calculations the agreement between calculation and measurements is within 0.5 dB in the lower frequency band. For the highest frequencies, since mismatch losses are negligible, the agreement between calculations and predictions is good. It is worth noting that the patterns from all copolarized and same frequency elements are very similar. This implicitly means that the interaction among neighboring elements can be neglected and the equivalent array in Fig. 4(b) was eventually achieved. The copolar, with respect to the Ludwig 3 denition, , 30 , radiation pattern for different azimuthal angles, 60 , 90 are plotted in Fig. 21 for the frequencies: (a) 9.875 GHz representing the lowest measured and useful frequency and (b) 10.1 GHz representing the highest measured and useful frequency. The cross-polarization levels are also very low as anticipated in the calculations performed using CST [15]. The secondary , ), beams generated by a reector ( when excited by the measured primary radiation patterns are plotted as a function of the frequency in Fig. 22. Also in this case, the agreement between simulation and measurements is very good. Finally, the edge of coverage directivity for the prototype has been calculated as given in Fig. 3. A difference of about

Fig. 19. Measured copolar and cross polar primary beams generated by the element 1 of the array plotted for three different frequencies within the lower frequency band on the: (a) E-plane, (b) H-plane, and (c) D-plane, respectively. The scale in dB represents gain.

is found between simulation and measurements in the lower frequency band. However, the increase of the edge of coverage directivity by using a super-layer conguration is still preserved and in the order of 2 dB demonstrating the concept.

118

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 21. Co-polar radiation pattern for different azimuthal angles,  = 0, 30, 60, and 90 are plotted for the frequencies: (a) 9.875 GHz representing the lowest measured and useful frequency and (b) 10.1 GHz representing the highest and useful frequency. The scale in dB represents gain.

Fig. 20. Measured copolar and cross polar primary beams generated by the element 5 of the array plotted for three different frequencies within the higher frequency band on the: (a) E-plane, (b) H-plane, and (c) D-plane, respectively. The scale in dB represents gain.

Fig. 22. Secondary directivity beams generated by a reector (F=D = 1:71, F = 245:5 ), when excited by the measured primary patterns are plotted as a function of the frequency. Note that the isolation of 16 dB can be estimated by plotting the adjacent beam centered at  = 1:12 .

C. Comments on the Measurements The comparison between measured and calculated directivities as a function of the frequency demonstrates the reproducibility of the calculations. Moreover the combined use

of the calculated edge of coverage directivity in Fig. 3 and the inclusion of the ohmic and mismatch losses also allows one to obtain an accurate prediction of the edge of coverage gain. On top of these losses also the lter insertion loss will contribute for approximately a dB depending on the chosen material for

NETO et al.: LEAKY WAVE ENHANCED FEEDS FOR MULTIBEAM REFLECTORS TO BE USED FOR TELECOM SATELLITE BASED LINKS

119

fabricating the lters. Considering the complexity of the experiment, the agreement between measurements and simulation is good. The specic objective was to design a system meeting real mission specications over a bandwidth of 2%. It has to be concluded that such a bandwidth is stretching the capabilities of such a resonant structure. In fact for the low-frequency lter , which implies the overall matching is the order of of a reduction in gain. This gain reduction reduces partly the gain enhancement (in the order of 2 dB) achieved thanks to the pattern shaping. VI. CONCLUSIONS A leaky wave enhanced focal plane feed array has been designed, manufactured and measured. The array is composed of 19 very compact iris-loaded waveguides operating in X band. It is a scaled version of a system that would be suited to feed a multi beam reector system with the purpose to provide contour coverage of portions of the earth from a geostationary telecommunication satellite. The stringent manufacturing tolerances required to achieve 20 dB gains from narrow band structures, resulted in a challenging engineering exercise. Overall the measurements provided in this paper demonstrate that it is possible to use the design methodology described in (in the [2][5] for reector systems characterized by large order of 2). Thus the performances of a multi beam single-feedper-beam reector system can be signicantly enhanced to the level of becoming comparable to the ones of existing systems that rely on arrays of four reectors, on a BW in the order of 2%. ACKNOWLEDGMENT The authors would like to thank P. Sarassa of ESA/ESTEC for designing the microwave ltering structures. REFERENCES
[1] S. K. Rao, Design and analysis of multiple-beam reector antennas, IEEE Antennas Propag. Mag., vol. 41, no. 4, pp. 5359, Aug. 1999. [2] A. Neto, N. Llombart, G. Gerini, M. Bonnedal, and P. de Maagt, EBG enhanced feeds for the improvement of the aperture efciency of reector antennas, IEEE Trans. Antennas Propag., vol. 55, no. 8, pp. 21852193, Aug. 2007. [3] N. Llombart, A. Neto, G. Gerini, M. Bonnedal, and P. de Maagt, Impact of mutual coupling in leaky wave enhanced imaging arrays, IEEE Trans. Antennas Propag., vol. 56, no. 4, pp. 12011206, Apr. 2008. [4] N. Llombart, A. Neto, G. Gerini, M. Bonnedal, and P. de Maagt, Leaky wave enhanced feeds arrays for the improvement of the edge of coverage gain in multibeam reector antennas, IEEE Trans. Antennas Propag., vol. 56, no. 5, pp. 12801291, May 2008. [5] M. Ettorre, A. Neto, G. Gerini, and S. Maci, Leaky wave slot array antenna fed by a dual reector system, IEEE Trans. Antennas Propag., vol. 56, no. 10, pp. 31433149, Oct. 2008. [6] G. von Trentini, Partially reecting sheet arrays, IRE Trans. Antennas Propag., vol. 4, no. 4, pp. 666671, Oct. 1956. [7] D. R. Jackson, A. A. Oliner, and A. Ip, Leaky wave propagation and radiation for a narrow-beam multiple-layer dielectric structure, IEEE Trans. Antennas Propag., vol. 41, no. 3, pp. 344348, Mar. 1993. [8] C. Cheype, C. Serier, M. Thevenot, T. Monediere, A. Reineix, and B. Jecko, An electromagnetic bandgap resonator antenna, IEEE Trans. Antennas Propag., vol. 50, no. 9, pp. 12851290, Sep. 2002. [9] Y. J. Lee, J. Yeo, R. Mittra, and W. S. Park, Application of electromagnetic bandgap (EBG) superstrates with controllable defects for a class of patch antennas as spatial angular lters, IEEE Trans. Antennas Propag., vol. 53, no. 1, pp. 224235, Jan. 2005.

[10] N. Guerin, S. Enoch, G. Tayeb, P. Sabouroux, P. Vincent, and H. Legay, A metallic FabryPerot directive antenna, IEEE Trans. Antennas Propag., vol. 54, no. 1, pp. 220224, Jan. 2006. [11] R. Sauleau, P. Coquet, and T. Matsui, Low-prole directive quasiplanar antennas based on millimetre wave FabryPerot cavities, Inst. Elect. Electr. Proc.-Microw. Antennas Propag., vol. 150, no. 4, pp. 274278, Aug. 2003. [12] R. Gardelli, M. Albani, and F. Capolino, Array thinning by using antennas in a FabryPerot cavity for gain enhancement, IEEE Trans. Antennas Propag., vol. 54, no. 7, pp. 19791990, Jul. 2006. [13] C. A. Balanis, Antenna Theory: Analysis and Design, 2nd ed. New York: Wiley, 1997. [14] Y. Mizugutch, M. Akagawa, and H. Yokoi, Offset dual reector antenna, in Proc. IEEE AP-S Int. Symp., Oct. 1976, pp. 25. [15] CST Microwave Studio, User Manual Version 5.0. Darmstadt, Germany, CST GmbH. [16] O. Sotoudeh, P.-S. Kildal, P. Ingvarson, and S. P. Skobelev, Singleand dual-band multimode hard horn antennas with partly corrugated walls, IEEE Trans. Antennas Propag., vol. 54, no. 2, pp. 330339, Feb. 2006. [17] A. Polemi and S. Maci, On the polarization properties of a dielectric leaky wave antenna, IEEE Antennas Wireless Propag. Lett., vol. 5, pp. 306310, Dec. 2006. [18] A. C. Ludwig, The denition of cross polarization, IEEE Trans. Antennas Propag., vol. AP-21, no. 1, pp. 116119, Jan. 1973.

Andrea Neto (M00SM10) received the Laurea degree (summa cum laude) in electronic engineering from the University of Florence, Florence, Italy, in 1994 and the Ph.D. degree in electromagnetics from the University of Siena, Siena, Italy, in 2000. Part of his Ph.D. was developed at the European Space Agency Research and Technology Center, Noordwijk, The Netherlands, where he worked for the antenna section for over two years. In the years 20002001, he was a Postdoctoral Researcher at California Institute of Technology, Pasadena, working for the Sub-mm wave Advanced Technology Group. From 2002 to January 2010, he was Senior Antenna Scientist at TNO Defence, Security and Safety, The Hague, The Netherlands. In February 2010, he was appointed Full Professor of Applied Electromagnetism at the EEMCS Department, of the Technical University of Delft, the Netherlands. His research interests are in the analysis and design of antennas, with emphasis on arrays, dielectric lens antennas, wide band antennas, EBG structures and THz antennas. Dr. Neto was co-recipient of the H.A. Wheeler award for the best applications paper of the year 2008 in the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION. He was co-recipient of the best innovative paper prize at the 30th ESA Antenna Workshop in 2008. He was co-recipient of the best antenna theory paper prize at the European Conference on Antennas and Propagation (EuCAP) in 2010. He presently serves as associate editor of the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION and of IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS (AWPL).

Mauro Ettorre (M08) was born in Tricarico, Matera, Italy, in 1979. He received the laurea degree (summa cum laude) in telecommunication engineering and the Ph.D. degree in electromagnetics from the University of Siena, Siena, Italy, in 2004 and 2008, respectively. During his Masters degree studies he spent ve months at the Technical University of Denmark (DTU), Lyngby, Denmark. Part of his Ph.D. degree has been developed at the Defence, Security and Safety Institute of the Netherlands Organization for Applied Scientic Research (TNO), The Hague, The Netherlands, where afterwards he worked as an Antenna Researcher. From 2008 to 2010, he was a Postdoctoral Fellow at the Institut dElectronique et de Tlcommunications de Rennes (IETR), Universit de Rennes 1, France. Since October 2010, he has been appointed CNRS researcher at IETR. His research interests include the analysis and design of leaky-wave antennas, periodic structures and compact planar antennas. Dr. Ettorre received the Young Antenna Engineer Prize during the 30th ESA Antenna Workshop 2008 in Noordwijk, The Netherlands.

120

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Giampiero Gerini (M92SM08) received the M.Sc. degree (summa cum laude) and the Ph.D. degree in electronic engineering from the University of Ancona, Ancona, Italy, in 1988 and 1992, respectively. From 1992 to 1994, he was Assistant Professor of Electromagnetic Fields at the University of Ancona. From 1994 to 1997, he was Research Fellow at the European Space Research and Technology Centre (ESA-ESTEC), Noordwijk, The Netherlands, where he joined the Radio Frequency System Division. Since 1997, he has been with the Netherlands Organization for Applied Scientic Research (TNO), The Hague, The Netherlands. At TNO Defence Security and Safety, he is currently Chief Senior Scientist of the Antenna Unit in the Transceiver Department. In 2007, he has been appointed as part-time Professor in the Faculty of Electrical Engineering of the Eindhoven University of Technology, The Netherlands, with a chair on Novel Structures and Concepts for Advanced Antennas. His main research interests are phased arrays antennas, electromagnetic bandgap structures, frequency selective surfaces and integrated antennas at microwave, millimeter and sub-millimeter wave frequencies. The main application elds of interest are radar, imaging and telecommunication systems. Dr. Gerini was co-recipient of the 2008 H. A. Wheeler Applications Prize Paper Award of the IEEE Antennas and Propagation Society. He was co-recipient also of the Best Innovative Paper Prize of the 30th ESA Antenna Workshop in 2008 and of the best antenna theory paper prize of the European Conference on Antennas and Propagation (EuCAP) in 2010.

Peter de Maagt (S88-M88-SM02F08) was born in Pauluspolder, The Netherlands, in 1964. He received the M.Sc. and Ph.D. degrees from Eindhoven University of Technology, Eindhoven, The Netherlands, in 1988 and 1992, respectively, both in electrical engineering. He is currently with the European Space Research and Technology Centre (ESTEC), European Space Agency, Noordwijk, the Netherlands. His research interests are in the area of millimeter and submillimeter-wave reector and planar integrated antennas, quasi-optics, photonic bandgap antennas, and millimeter- and sub-millimeter-wave components. Dr. de Maagt was co-recipient of the H.A. Wheeler award of the IEEE Antennas and Propagation Society for the best applications paper of the year 2001 and in 2008. He was granted a European Space Agency award for innovation in 2002. He served as associate editor of the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION from 2005 to 2010

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

121

94 GHz Substrate Integrated Monopulse Antenna Array

Yu Jian Cheng, Member, IEEE, Wei Hong, Fellow, IEEE, and Ke Wu, Fellow, IEEE
AbstractA planar W-band monopulse antenna array is designed based on the substrate integrated waveguide (SIW) technology. The sum-difference comparator, 16-way divider and 32 32 slot array antenna are all integrated on a single dielectric substrate in the compact layout through the low-cost PCB process. Such a substrate integrated monopulse array is able to operate over 93 96 GHz with narrow-beam and high-gain. The maximal gain is measured to be 25.8 dBi, while the maximal null-depth is measured to be 43.7 dB. This SIW monopulse antenna not only has advantages of low-cost, light, easy-fabrication, etc., but also has good performance validated by measurements. It presents an excellent candidate for W-band directional-nding systems. Index TermsMonopulse antenna array, substrate integrated circuits (SICs), substrate integrated waveguide (SIW), W-band.

I. INTRODUCTION

ONOPULSE antennas are attractive for high-resolution tracking applications [1] and developed in the variety of technologies [2][8]. Conventional metallic waveguide, which provides low-loss and coupling-free features, may be the best choice for the design of W-band monopulse array. However, it is expensive, heavy and difcult to fabricate. Todays microwave and millimeterwave system design have to consider the integration problems between those planar and non-planar structures. To cope with those problems, the design and fabrication costs may become prohibitively high. In the other side, planar circuit topologies provide a better avenue for achieving the goals such as enhancing integration density and low-cost mass-production. Up to now, several planar and quasi-planar W-band monopulse systems have been demonstrated using various antenna geometries [9][11], but their performance are still not good enough. There is an

overpowering need to develop a low-cost high-quality W-band printed monopulse array. Recently, substrate integrated waveguide (SIW) have been proposed and under intensive investigation around the world. This planar scheme is able to combine the excellent features of both planar transmission lines and non-planar waveguide [12][16], and allows a high integration degree of millimeter-wave circuits at low-cost fabrication. Therefore, the SIW technique is very appropriate to construct the W-band planar monopulse array. The comparator, divider and antenna array are all integrated in a single substrate through the low-cost normal PCB process. Compared with the Ka-band version developed in [8], there are several new challenges and design considerations in the design of a higher-frequency larger-scale antenna array. Firstly, the W-band SIW has more strict design rule to avoid high fabrication tolerance and radiation leakage, thus its dimension must be determined carefully. In simulation, the delay line used in [8] has narrow phase shift bandwidth. To improve phase characteristics, another phase shifter is employed in our design. Considering the small horizontal size of W-band SIW and the fabrication limitation in normal PCB process, it is impossible to synthesize a 4-order conventional T-type divider with good performance. As such, a broadband W-band SIW divider is developed to construct this 16-way divider. In addition, the W-band SIW have to be fed by waveguide instead of the SIW-microstrip transition, which is widely used in low frequency. Finally, a prototype array is fabricated and tested to verify the correctness of our design. II. DESIGN PROCESS As shown in Fig. 1, there are several sub-components within the substrate integrated monopulse antenna array, including four 3 dB directional couplers, four 90 phase shifters, four 16-way dividers, and four 16 16 slot antenna sub-arrays. All circuits are integrated in a single Rogers 5880 substrate with the relative permittivity, , of 2.2 and the thickness of 0.508 mm. The width of the SIW, , is 1.76 mm, the diameter of the metallic via, , is 0.4 mm, and the distance between adjacent vias, , is 0.6 mm. A. W-Band SIW The W-band SIW has the small width and relative large metallic posts. As such, its characteristics are more vulnerable to fabrication tolerances compared with its low frequency counterpart. The SIW phase constant, , may be the most important factor in our design. Therefore, the W-band SIWs dimensions should be determined carefully to weaken the tolerance inuence on , which is primarily functions of , , , and .

Manuscript received March 05, 2011; revised May 12, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the National Natural Science Foundation of China (NSFC) under Grant 61001028, in part by the Research Fund for the Doctoral Program of Higher Education of China (RFDP) under Grant 20100185110014, and in part by the Fundamental Research Funds for the Central Universities under Grant ZYGX2010J019. Y. J. Cheng is with the EHF Key Laboratory of Fundamental Science, School of Electronic Engineering, University of Electronic Science and Technology of China (UESTC), Chengdu 611731, China (e-mail: yjcheng@emeld.org). W. Hong is with the State Key Laboratory of Millimeter Waves, Southeast University, Nanjing 210096, China. K. Wu is with Poly-Grames Research Center, Department of Electrical Engineering, Ecole Polytechnique, Montreal, QC H3C 3A7, Canada. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167945

0018-926X/$26.00 2011 IEEE

122

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 1. Block diagram of the proposed 94 GHz monopulse antenna.

With the change due to fabrication tolerances, the SIW actual variables will be changed and cause drifting from the designed value. Its worst case behavior, , can be calculated by use of the sensitivity method as described in [17]: (1) where (2) A complete set of graphs for the sensitivities of with respect to , , and as functions of the width-to-diameter ratio and the diameter-to-distance ratio are given in Fig. 2. All cases guarantee that the SIW only works in the dominant mode zone. The maximal errors for different variables from manufacturer are: , and . From this gure, it is noted that is more sensitive to tolerance in and less in . Larger contributes to smaller error, while smaller contributes to smaller error. From above analysis, the W-band SIW is preferred to have larger , smaller , and larger to achieve a better tolerance in . But the choices of and are also restricted by available fabrication techniques. As an example, the minimal drillable diameter of metallic via is 0.3 mm, while the minimal distance between the edges of adjacent vias is 0.2 mm in our lab. In the rst experiment, we chose . But there exists a big difference between the design and the measured values. That is because the 0.3 mm post diameter is the machinable utmost limit with large errors. Thus we change from 0.3 mm to 0.4 mm to guarantee machining precision. On the other hand, the electromagnetic eld is more difcult to be conned within the two arrays of metallic vias with the increase of frequency. The design rule for the W-band SIW should be more rigorous. Here, three different SIWs operated at different frequency, i.e., 10 GHz, 30 GHz and 94 GHz, are investigated respectively. They are simulated with perfect conductor and ideal dielectric material. As such, the only source of loss is the radiation. As shown in Fig. 3, should be less than 0.65 mm

Fig. 2. Sensitivity curves of SIW phase constant with respect to , , : , : ).

or . ( :

Fig. 3. versus . ( , and 94 GHz:

, 10 GHz: ).

, 30 GHz:

at 94 GHz to keep affordable radiation loss, i.e., [18]. In conclusion, the selection of should be considered with a view to the tolerance and radiation loss. B. SIW Comparator The SIW comparator consists of four 3 dB directional couplers and four 90 xed phase shifters. The proposed 3 dB SIW-based coupler structure is shown in Fig. 4. It has the continuous coupling aperture. Steps are used to achieve the good matching. Each input or output port is connected with a 90 corner, which has a metallic post at the corner. Full-wave simulator HFSS is used to optimize , , and to ensure excellent performance of the coupler over the whole frequency bandwidth. The nal dimensions are: , , and . Two different types of 90 SIW xed phase shifters are connected after 90 couplers to produce the in- or out-of-phase feeding mechanism. One is realized by unequal-length paths, naming delay line (I), to realize the 90 phase shift in H-plane. The other is the equal-length unequal-width phase shifter (II) to

CHENG et al.: 94 GHz SUBSTRATE INTEGRATED MONOPULSE ANTENNA ARRAY

123

Fig. 4. Conguration of the proposed 3 dB SIW-based directional coupler.

Fig. 5. Different 90 phase shifters connected after a 90 coupler. Fig. 7. Topology of the proposed SIW comparator.

Fig. 6. Comparison of different phase shifters (phase(S21S31)). Fig. 8. Simulated S-parameters of the proposed SIW comparator.

nish the 90 phase shift in E-plane. To achieve the better tolerance and reduce mismatching, one branch has unchanged width and the other branch has wider width compared with rest SIW circuits. To make a comparison, a 3 dB coupler cascading with different types of 90 phasers are modeled as shown in Fig. 5. As shown in Fig. 6, type II, which has two equal-length branches with different widths (i.e. 1.96 mm and 1.76 mm), offers phase shift over 90 98 GHz in simulation, while type I introduces phase errors within the same frequency band. The phase shifter II is better judging from the simulations. That is because the curves of propagation constants with respect to different SIW widths behave almost in parallel with each other. Now, the SIW comparator can be constructed in a compact layout. It surrounds enough area reserved for dividers and slot arrays. Four 90 round bends, which has double metallic vias to avoid the energy leakage at such high frequency, are used to connect these couplers. The topology of the SIW comparator is illustrated in Fig. 7. Excited at different input ports , the required amplitude and phase excitations can be generated at output ports . The simulated results of the whole SIW comparator are shown in Figs. 8, 9 after considering all losses. The phase characteristic in E-plane is much better.

Fig. 9. Simulated phase differences of the proposed SIW comparator.

C. SIW 16-Way Divider Four-stage 2-way dividers can be grouped together to realize the W-band 16-way divider, which is another challenge in this design. As shown in Fig. 10, the rst used type (I) is the conventional T-type divider. In our design, the SIWs width is narrow and only 1.76 mm. The T-type SIW divider requires at least an inductive via as the matching post within the corner structure [19]. Considering the dimension limitation of drillable via in the

124

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 10. Different SIW dividers.

Fig. 12. Conguration of the proposed SIW 16-way divider.

Fig. 11. Simulated reections of different SIW dividers.

normal PCB process, the matching post is usually big relative for the SIW width. Thus it is impossible to synthesize a 4-order divider with good performance in this case. The second employed divider (II) is modied from the rst version. The junction is expanded in order to locate the matching post at a proper position as shown in Fig. 10. This structure is able to offer good reection at the center frequency. However, the junction can be viewed as a resonator and thus the bandwidth of this divider is narrow. As shown in Fig. 11, the bandwidth better than 15 dB of 2-way divider is 93.79 95.74 GHz, while the bandwidth of 4-way divider grouped by two-stage 2-way dividers is 93.88 94.46 GHz. The performance of 8-way divider grouped by three-stage 2-way dividers is not good even within a narrow bandwidth. It is also difcult to design a large-scale divider. Next, we attempt to build a modied Y-divider (III) with the conguration as shown in Fig. 10. It has much better performance compared with the second type. As shown in Fig. 11, the bandwidth better than 15 dB of 2-way divider is more than 90 98 GHz, the bandwidth of 4-way divider grouped by twostage 2-way dividers is 91.39 97.46 GHz, the bandwidth of 8-way divider grouped by three-stage 2-way dividers is 91.78 96.99 GHz, and the bandwidth of 16-way divider grouped by four-stage 2-way dividers is 92.19 97.71 GHz. Besides, the topology of this 16-way Y-type divider is convenient to feed a compact antenna array. Using the original and modied Y-type dividers, the topology of the proposed SIW 16-way divider is shown in Fig. 12. The symmetric conguration guarantees all transmission coefcients equal. Inductive posts are used at the corners to help the wave travel uently and reduce the return loss. Fig. 13 depicts the simulated results of the proposed structure after considering all losses.

Fig. 13. Simulated S-parameters of the proposed SIW 16-way divider.

Fig. 14. Conguration of the proposed back-to-back SIW slot antenna array.

D. SIW Slot Array Antenna Waveguide slot antenna is an important class of millimeterwave antennas. In this design, the antenna array consists of 16 linear SIW slot arrays, each of which includes 16 slots. The conguration of the back-to-back SIW slot array is illustrated in Fig. 14. It is proposed to operate at the center frequency of 94 GHz. The back-to-back array should be designed, simulated, and optimized together. For example, if the experimental array is set to follow a certain distribution, the total number of slots will be 32 but not 16. The desired amplitude distribution for each slot can be calculated respectively, and the rst half of them will be selected to design a 16-slot array antenna. Then, the back-to-back array will be built in HFSS in its entirety as shown in Fig. 14 and be optimized together. Sidelobe suppression is not of concern in the present application because the parameters are too sensitive in W-band. The nal dimensions are listed in Table I. The simulated sum and difference radiation patterns at 94 GHz are shown in Fig. 15. In this topology, there exists a 180 phase difference between adjacent antennas. Therefore, the out-of-phase feed generates a

CHENG et al.: 94 GHz SUBSTRATE INTEGRATED MONOPULSE ANTENNA ARRAY

125

TABLE I FINAL PARAMETERS OF THE PROPOSED SLOT ARRAY ANTENNA (UNIT: mm)

Fig. 16. Simulated sidelobe level of H-plane sum beams at 94 GHz versus amplitude and phase errors.

Fig. 17. Simulated amplitude imbalances of H-plane difference beams at 94 GHz versus amplitude and phase errors.

Fig. 15. Simulation radiation patterns of the proposed back-to-back SIW slot antenna array.

sum beam while the in-phase feed generates a difference beam. There is no problem for this topology just inverting beams. The errors of two feeding signals (A1 and A2) in amplitude and phase deteriorate sidelobe level of sum beams and cause amplitude imbalances of difference beams. Figs. 16, 17 give 94 GHz H-plane simulated results of this back-to-back array considering the inuences of amplitude and phase imbalances. When , 50 phase error leads to 8.3 dB deterioration of sidelobe level for the H-plane sum beam and 4.7 dB amplitude imbalance for the H-plane difference beam. E. Waveguide-SIW Transition The SIW, which has a full-closed structure, is able to avoid the additional radiation coupling and interfere. In practical, an SIW is usually fed by planar transmission lines, such as microstrip line, CPW, etc. But in antenna applications, the power will be radiated by these open feed lines, which spoils the antenna performance especially in high frequency. As such, the SIW can be fed by waveguides to avoid the direct radiation from the feed discontinuity.

Fig. 18. Conguration of the proposed waveguide-SIW transition.

The transition between SIW and air-lled standard waveguide has the similar conguration as described in [20], [21]. Two coupled apertures are etched on the top conductor layers with two 90 corners as shown in Fig. 18. The nal dimensions are: , , , and . The simulated results are shown in Fig. 19 after considering all losses. The bandwidth for S11 less than 16 dB is from 90 GHz to 98 GHz. F. Loss Consideration There exists a very important issue for the W-band substrate integrated monopulse antenna array, namely loss. From manufacturers data sheet, the used Rogers5880 substrate has 0.0009 of at 10 GHz. Here, we should know at 94 GHz to evaluate the W-band circuit loss. Two SIWs

126

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 19. Simulation results of the proposed waveguide-SIW transition.

Fig. 21. Measured H-plane radiation patterns of the fabricated planar monopulse array at different frequency excited at port 1.

Fig. 22. Measured E-plane radiation patterns of the fabricated planar monopulse array at different frequency excited at port 1. Fig. 20. Paragraph of the fabricated substrate integrated monopulse array.

III. EXPERIMENTS with different length were fabricated and measured. Compared with different simulated curves with different , the simulation of is in good agreement with the experimental result [22]. This value may be not very accuracy due to other inuencing factors, but it can be viewed as the equivalent one. As such, all simulations presented in the paper have been carried out with this estimated loss value. Now, the total loss of the whole structure will be estimated with an excitation at the sum port. In simulation, the average loss from comparator is 4.66 dB, the average loss from divider is 1.32 dB, and the loss from transition is 0.65 dB at 94 GHz. Considering 75.14% radiation efciency of the slot array, the total radiation efciency of the proposed monopulse array is estimated to be 16.3%. The total loss of SIW monopulse array in W-band is a little high and the efciency is relative low, even it is much better than other planar printed counterparts. The multi-layer conguration is effective in shortening the propagation path of the comparator, thus improving the radiation efciency. Each center of top-layer slot array antennas can be fed in turn by a crosswise SIW using series slots to avoid the use of large divider. The most key to such a design is the reliability of multi-layer fabrication process in W-band. A. Results Fig. 20 gives the photograph of the fabricated 94 GHz substrate integrated monopulse antenna array with the size of 130 125 mm. Two standard WR-10 waveguides are xed with the measured antenna and metallic blocks by positioning pings and screws to feed the array. Excited at port 1, the monopulse antenna array generates the sum beam in both E-plane ( plane) and H-plane ( plane). Figs. 21, 22 give the measured results at different frequency. The detailed measured data are listed in Table II, including gain, sidelobe level, and 3 dB beam-width. The 3 dB beam-width are in the range of (expect 96 GHz) in H-plane and in E-plane. The sidelobe level are in the range of 5.95 19.74 dB in H-plane and 9.82 12.14 dB in E-plane. The maximum gain is measured to be 25.75 dBi at 93.6 GHz while the minimum one is 21.29 dBi at 96 GHz. At 94 GHz, the radiation efciency is measured to be 15.2%, which is in good agreement with our expected value. Excited at port 2, the monopulse array generates the difference beam in H-plane ( plane). The measured radiation patterns are shown in Fig. 23. The maximum null-depth can be low to 45.81 dB at 93.8 GHz. The best amplitude balance between the two difference peaks is 2.12 dB at 95.2 GHz. More detailed

CHENG et al.: 94 GHz SUBSTRATE INTEGRATED MONOPULSE ANTENNA ARRAY

127

TABLE II MEASUREMENTS OF THE MONOPULSE ARRAY EXCITED AT PORTS 1 VERSUS FREQUENCY

Fig. 24. Measured E-plane radiation patterns of the fabricated planar monopulse array at different frequency excited at port 3.

Excited at port 3, the monopulse array generates the difference beam in E-plane ( plane). The measured radiation patterns are shown in Fig. 24. The maximum null-depth can be low to 37.71 dB at 95.2 GHz. The best amplitude balance between two difference peaks is 0.40 dB at 93.6 GHz. Detailed measured data are also listed in Table III.

B. Discussions Compared with the simulated and measured results, the radiation performance of the antenna array is worsened, especially for sidelobe level of sum beams and null-depth and amplitude imbalance of difference beams. Due to unavoidable fabrication tolerances, the characteristics of the feeding network will depart from the designed specications. The amplitude imbalance of the feeding network mainly comes from tolerances in couplers and different transmission paths with different lengths. The phase imbalance mainly comes from tolerances in phase shifters. As an example, the worst phase errors are simulated to be 73.6 for the delay line and 48.2 for the unequal-width phase shifter considering uniform distribution tolerances: . As discussed in part D, Section II, the H-plane patterns are sensitive to the amplitude and phase imbalances. Once considering errors from the other plane and tolerances in substrate such as thickness and dielectric constant, the antenna performance would be deteriorated further. On the other hand, fabrication tolerances in slot array antennas will inuence the radiation performance as well, which is difcult to be evaluated accurately. These effects from the feeding network and the slot array antenna should be taken into consideration together in the accurate design. Besides, the input ports are terminated nothing (open) because of lacking enough W-band matching loads. In this case, the performance of antenna will be deteriorated and the sidelobe level will be increased. Such an SIW monopulse array has better performance in E-plane than that in H-plane considering the uniformity of beams versus frequency. The rst reason is that the phase of the feeding network in H-plane deviates from the designed

Fig. 23. Measured H-plane radiation patterns of the fabricated planar monopulse array at different frequency excited at port 2.

TABLE III MEASUREMENTS OF THE MONOPULSE ARRAY EXCITED AT PORTS 2 AND 3 VERSUS FREQUENCY

measured data are listed in Table III, including null-depth and amplitude imbalance.

128

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

value much more versus frequency than that in E-plane. In simulation,

This will leads to a larger variation of radiation patterns in H-plane. Furthermore, the slot array antenna is a resonant type antenna, thus its H-plane performance is more varied by frequency compared with its E-plane. IV. CONCLUSION In this paper, a high gain 94 GHz planar substrate integrated monopulse antenna array is proposed. All sub-components are integrated in a single substrate and fabricated through the lowcost PCB process. Design considerations are introduced for such a complex and large W-band application. It makes several improvements in the aspects of machining difculty, weight, cost, packaging, etc. Thus this type of substrate integrated monopulse antenna presents an excellent candidate in the development of W-band integrated radar and tracking systems. However, it has lower radiation efciency compared with the metallic waveguide-based structure since a dielectric must be used for lling the waveguide. Furthermore, there is still room for the sidelobe level improvement. REFERENCES
[1] S. M. Sherman, Monopulse Principles and Techniques. Dedham, MA: Artech House, 1984. [2] M. Sierra-Castaner, M. Sierra-Perez, M. Vera-Isasa, and J. L. Fernandez-Jambrina, Low-cost monopulse radial line slot antenna, IEEE Trans. Antennas Propag., vol. 51, no. 2, pp. 256262, Feb. 2003. [3] L. Yan, W. Hong, and K. Wu, Simulation and experiment on substrate integrated monopulse antenna, in Proc. IEEE Antennas Propag. Soc. Int. Symp., Jul. 2005, vol. 1A, pp. 528531. [4] Y. J. Cheng, W. Hong, and K. Wu, Millimetre-wave monopulse antenna incorporating substrate integrated waveguide phase shifter, IET Microw. Antennas Propag., vol. 2, no. 1, pp. 4852, Jan. 2008. [5] Y. J. Cheng, W. Hong, and K. Wu, Design of a monopulse antenna using a dual V-type linearly tapered slot antenna (DVLTSA), IEEE Trans. Antennas Propag., vol. 56, no. 9, pp. 29032906, Sep. 2008. [6] H. Wang, D. G. Fang, and M. Li, A single-channel microstrip electronic tracking antenna array with time sequence phase weighting on sub-array, IEEE Trans. Microw. Theory Tech., vol. 58, no. 2, pp. 253259, Feb. 2010. [7] H. Wang, D. G. Fang, B. Zhang, and W. Q. Chen, Dielectric loaded substrate integrated waveguide (SIW) H-plane horn antennas, IEEE Trans. Antennas Propag., vol. 58, no. 3, pp. 640648, Sep. 2010. [8] B. Liu, W. Hong, Z. Q. Kuai, X. X. Yin, G. Q. Luo, J. X. Chen, H. J. Tang, and K. Wu, Substrate integrated waveguide (SIW) monopulse slot antenna array, IEEE Trans. Antennas Propag., vol. 57, no. 1, pp. 275279, Jan. 2009. [9] C. C. Ling and G. M. Rebeiz, 94 GHz integrated horn monopulse antennas, IEEE Trans. Antennas Propag., vol. 40, no. 8, pp. 981985, Aug. 1992. [10] C. C. Ling and G. M. Rebeiz, A 94 GHz planar monopulse tracking receiver, IEEE Trans. Microw. Theory Tech., vol. 42, no. 10, pp. 18631871, Oct. 1994. [11] S. Raman, N. S. Barker, and G. M. Rebeiz, A W-Band dielectriclens-based integrated monopulse radar receiver, IEEE Trans. Microw. Theory Tech., vol. 46, no. 12, pp. 23082317, Dec. 1998. [12] H. Uchimura, T. Takenoshita, and M. Fujii, Development of a laminated waveguide, IEEE Trans. Microw. Theory Tech., vol. 46, no. 12, pp. 24382443, Dec. 1998.

[13] J. Hirokawa and M. Ando, Efciency of 76-GHz post-wall waveguide-fed parallel-plate slot arrays, IEEE Trans. Antennas Propag., vol. 48, no. 11, pp. 17421745, Nov. 2000. [14] E. Moldovan, R. G. Bosisio, and K. Wu, W-Band multiport substrateintegrated waveguide circuits, IEEE Trans. Microw. Theory Tech., vol. 54, no. 2, pp. 625632, Feb. 2006. [15] Y. J. Cheng, W. Hong, and K. Wu, Broadband self-compensating phase shifter combining delay line and equal-length unequal-width phaser, IEEE Trans. Microw. Theory Tech., vol. 58, no. 1, pp. 203210, Jan. 2010. [16] Y. J. Cheng, W. Hong, K. Wu, and Y. Fan, Millimeter-wave substrate integrated waveguide long slot leaky-wave antennas and two-dimensional multibeam applications, IEEE Trans. Antennas Propag., vol. 59, no. 1, pp. 4047, Jan. 2011. [17] K. C. Gupta, R. Garg, I. Bahl, and P. Bhartia, Microstrip Lines and Slotlines, 2nd ed. Norwood, MA: Artech house, 1996. [18] D. Deslandes and K. Wu, Accurate modeling, wave mechanisms, and design considerations of a substrate integrated waveguide, IEEE Trans. Microw. Theory Tech., vol. 54, no. 6, pp. 25162526, Jun. 2006. [19] J. Hirokawa, K. Sakurai, M. Ando, and N. Goto, An analysis of a waveguide T junction with an inductive post, IEEE Trans. Microw. Theory Tech., vol. 39, no. 3, pp. 563566, Mar. 1991. [20] T. Kai, J. Hirokawa, and M. Ando, A stepped post-wall waveguide with aperture interface to standard waveguide, in Proc. IEEE AP-S Int. Symp., Jun. 2004, vol. 2, pp. 15271530. [21] J. A. Lee, J. Hirokawa, and M. Ando, Width of the step tolerable for offset of the aperture in a transducer between post-wall and hollow standard waveguides, in Proc. IEEE AP-S Int. Symp., Jun. 2007, pp. 42644267. [22] J. F. Xu, Investigations on the substrate-integrated antennas and arrays, Ph.D. dissertation, Southeast Univ., Nanjing, China, 2010. Yu Jian Cheng (S08M11) was born in Sichuan Province, China, on April, 1983. He received the B.S. degree in electric engineering from University of Electronic Science and Technology of China, in 2005 and is working toward the Ph.D. degree (without going through the conventional Masters degree) at Southeast University, Nanjing, China. His current research interests include microwave and millimeter-wave passive circuits, antennas. Mr. Cheng has served on review boards of various technical journals, including IEEE MICROWAVE COMPONENTS LETTERS and IEEE ANTENNAS PROPAGATION

AND WIRELESS MAGAZINE.

Wei Hong (M92SM07F12) received the B.S. degree from the University of Information Engineering, Zhengzhou, China, in 1982, and the M.S. and Ph.D. degrees from Southeast University, Nanjing, China, in 1985 and 1988, respectively, all in radio engineering. Since 1988, he has been with the State Key Laboratory of Millimeter Waves and serves for the Director of the lab since 2003, and is currently a Professor and the Associate Dean of the School of Information Science and Engineering, Southeast University. In 1993, 1995, 1996, 1997, and 1998, he was a short-term Visiting Scholar with the University of California at Berkeley and Santa Cruz, respectively. He has been engaged in numerical methods for electromagnetic problems, millimeter wave theory and technology, antennas, electromagnetic scattering, RF technology for mobile communications, etc. He has authored and coauthored more than 200 technical publications, and authored the books Principle and Application of the Method of Lines (in Chinese: Southeast University Press, 1993) and Domain Decomposition Methods for Electromagnetic Problems (in Chinese, Science, 2005). Dr. Hong was thrice awarded the rst-class Science and Technology Progress Prizes issued by the Ministry of Education of China and the Jiangsu Province Government. He also received the Foundations for China Distinguished Young Investigators and for Innovation Group issued by NSF of China. He is a senior member of CIE, Vice-President of the Microwave Society and Antenna Society of CIE, and has served as a reviewer for IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, IET Proc.-H, and Electronic Letters. He also served as an

CHENG et al.: 94 GHz SUBSTRATE INTEGRATED MONOPULSE ANTENNA ARRAY

129

Associate Editor of the IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES from 2008 to 2010, and is an Editorial Board Member for IJAP and RFMiCAE.

Ke Wu (M87SM92F01) is currently a Professor of electrical engineering, and Tier-I Canada Research Chair in RF and millimeter-wave engineering with the cole Polytechnique of Montral, Montral, QC, Canada. He holds the rst Cheung Kong Endowed Chair Professorship (visiting) with Southeast University, the rst Sir Yue-Kong Pao Chair Professorship (visiting) with Ningbo University, and an Honorary Professorship with the Nanjing University of Science and Technology and the City University of Hong Kong. He has been the Director of the Poly-Grames Research Center and the founding Director of the Center for Radiofrequency Electronics Research of Quebec (Regroupement stratgique, FRQNT). He has also held guest and visiting professorship in many universities around the world. He has authored or coauthored over 730 referred papers and a number of books/book chapters. He has served on the editorial/review boards of many technical journals, transactions, and letters, as

well as scientic encyclopedias as both an editor and guest editor. He holds numerous patents. His current research interests involve substrate integrated circuits (SICs), antenna arrays, advanced computer-aided design (CAD) and modeling techniques, and development of low-cost RF and millimeter-wave transceivers and sensors for wireless systems and biomedical applications. He is also interested in the modeling and design of microwave photonic circuits and systems. Dr. Wu is a Fellow of the Canadian Academy of Engineering (CAE) and the Royal Society of Canada (The Canadian Academy of the Sciences and Humanities). He is a member of the Electromagnetics Academy, Sigma Xi, and the URSI. He has held key positions in and has served on various panels and international committees including the chair of technical program committees, international steering committees, and international conferences/symposia. He will be the General Chair of the 2012 IEEE Microwave Theory and Techniques Society (IEEE MTT-S) International Microwave Symposium (IMS). He is currently the Chair of the joint IEEE Chapters of MTTS/APS/LEOS, Montral, QC, Canada. He is an elected IEEE MTT-S Administrative Committee (AdCom) member (20062012) and is the Chair of the IEEE MTT-S Member and Geographic Activities (MGA) Committee. He is an IEEE MTT-S Distinguished Microwave Lecturer (20092011). He was the recipient of many awards and prizes including the rst IEEE MTT-S Outstanding Young Engineer Award, the 2004 Fessenden Medal of the IEEE Canada, and the 2009 Thomas W. Eadie Medal of the Royal Society of Canada.

130

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

The Planar Ultrawideband Modular Antenna (PUMA) Array

Steven S. Holland, Member, IEEE, and Marinos N. Vouvakis, Member, IEEE
The authors humbly dedicate this work in memory of Professor Benedikt A. Munk. scans [3]. Additionally, the integrated Knorr balun of the Vivaldi elements provides an unbalanced 50 feed interface. Despite excellent electrical performance, wideband Vivaldis consist of deep, vertically-integrated notches that must be electrically connected to avoid resonances [4], resulting in complicated dual-polarized arrays that are often expensive to manufacture and assemble. Modular Vivaldi array variations, such as the body of revolution (BOR) Vivaldi array [5], or the Mechanotch array [6] exist, but rely on special machining and are difcult to implement at high frequencies (above X-band). Several dipole-like elements have been developed as low-prole and modular alternatives to Vivaldi arrays, and consist of vertically integrated PCB cards backed with a ground plane. These low-prole (depth , independent of bandwidth) elements do not require electrical connection, and are often tapered vertically from the feed-lines. The Bunny-ear [7], balanced antipodal Vivaldi antenna (BAVA) [8], doubly-mirrored BAVA (Dm-BAVA) [9] and Banyan tree antenna (BTA) [10], [11], [33] arrays belong to this class. The Bunny-ear array has demonstrated a 5:1 bandwidth, using external baluns and resonance-suppression resistors between its element arms and ground [7]. The Dm-BAVA alleviates these resonances using element mirroring in E-and H- planes, but requires 180 hybrids to maintain beam collimation. The BTA array has achieved bandwidths up to 4:1without baluns or hybridsby using shorting strips between the arms and the ground plane. Vertically-integrated arrays are difcult to integrate in dual-polarized arrangements at higher frequencies, thus fully planar wideband topologies are desirable. To date, several quasi-planar wideband arrays have been published. The current sheet antenna (CSA) array [12], which cleverly realizes Wheelers current sheet, [13], uses tightly-coupled horizontal dipoles above a ground plane to achieve bandwidths of 9:1 with a printable element layer. Alternatively, array apertures formed from periodically fed continuous sheets or slots, such as the fragmented aperture array (FAA) [14], [15], and the long slot array, [16], can produce a stable impedance over wide bandwidths; however, both arrays exhibit bidirectional radiation, and when backed by a ground plane suffer from catastrophic resonances that limit their performance to approximately 4:1 [17]. Alternatively, these resonances can be suppressed using Jaumann screens [18] or ferrite loading [19], which decreases radiation efciency and power handling, and increases the arrays noise gure. More importantly, these array technologies are not fully printable and planar because they require elaborate, non-planar 3D cable organizers [20] between the ground plane and the printed element layers. These cable

AbstractA fully planar ultrawideband phased array with wide scan and low cross-polarization performance is introduced. The array is based on Munks implementation of the current sheet concept, but it employs a novel feeding scheme for the tightly coupled horizontal dipoles that enables simple PCB fabrication. This feeding eliminates the need for cable organizers and external baluns, and when combined with dual-offset dual-polarized lattice arrangements the array can be implemented in a modular, tile-based fashion. Simple physical explanations and circuit models are derived to explain the arrays operation and guide the design process. The theory and insights are subsequently used to design an exemplary dual-polarized innite array with 5:1 bandwidth and at broadside, and cross-polarization out to in the D- plane. Index TermsDipole arrays, phased arrays, planar arrays, ultra wideband antennas.

I. INTRODUCTION

LTRAWIDEBAND antenna arrays with wide-scan ability and polarization diversity are in great demand for multifunctional systems, [1]. Such systems use phased arrays that can simultaneously accommodate multiple frequency bands, beams, and polarizations in order to consolidate multiple antennas into a single ultrawideband aperture. Ultrawideband apertures are typically non-modular and expensive to build and assemble, and often require external baluns to integrate with RF front-ends. One of the most popular wideband arrays is the tapered-slot (Vivaldi) array, which has been utilized in many UWB systems [2]. Extensive work spanning three decades has resulted in Vivaldi arrays that achieve bandwidths in excess of 10:1 at wide-

Manuscript received October 09, 2010; revised January 25, 2011; accepted June 16, 2011. Date publication September 15, 2011; date of current version January 05, 2012. This work was supported by the Naval Research Laboratory under Grant PG#11320000000008. S. S. Holland was with the Center for Advanced Sensor and Communication Antennas, Electrical and Computer Engineering Department, University of Massachusetts at Amherst, Amherst, MA 01003-9292 USA. He is now with MITRE Corporation, Bedford, MA 01730-1420 USA (e-mail: steven.s.holland@gmail.com). M. N. Vouvakis is with the Center for Advanced Sensor and Communication Antennas, Electrical and Computer Engineering Department, University of Massachusetts at Amherst, Amherst, MA 01003-9292 USA. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167916

0018-926X/$26.00 2011 IEEE

HOLLAND AND VOUVAKIS: THE PLANAR ULTRAWIDEBAND MODULAR ANTENNA (PUMA) ARRAY

131

organizers are critical in preventing scan-induced resonances [21] by electrically shielding the vertical balanced feed lines. Additionally, all of these arrays require external wideband baluns/hybrids at each element. The need for wideband external baluns poses a non-trivial problem. Wideband passive baluns do exist, e.g., Marchand/ compensated baluns, [22], but are electrically large, [23], and have high insertion loss. Conversely, active baluns can readily achieve wide bandwidths, but are unidirectional, noisy, and have low power-handling that limits their use to receive-only applications. Moreover, non-zero phase and amplitude imbalances, [24], nite common-mode rejection ratios (CMRR), [25], and low active gains at high frequencies are some other practical active balun limitations. Unlike other UWB arrays, the planar ultrawideband modular array (PUMA) is fabricated with planar etched circuits and plated vias, thus it can be fabricated as a simple multilayer microwave PCB, and does not require external baluns. In addition, the array has a low prole and can be constructed modularly. The array consists of a dual-offset dual-polarized version of Munks tightly-coupled dipoles above a ground plane, fed by a novel unbalanced feed-line scheme. The PUMA [26] has shorting vias at its dipole arms, enabling direct connection to standard RF interfaces and modular construction. The placement of the plated vias controls the frequency of a catastrophic common mode that would otherwise occur near mid-band since the array is fed unbalanced. This topology modication, along with the dual-offset dual-polarized arrangement, gives rise to new phenomenology that is explained using simple physical models and equivalent circuits. Thus far, the PUMA array has demonstrated low VSWR and good scan performance out to over a bandwidth of 3:1 when fed directly from a 50 unbalanced interface [27], [28]. When the array is used in conjunction with a specially designed planar matching network printed on the opposite side of the ground plane, it achieves a 5:1 bandwidth with at broadside, out to 45 scan in the H- plane, and approximately cross-polarization at in the D- plane. The remainder of the paper is organized as follows. Section II describes the baseline structure of the PUMA array. Section III develops the theory, starting with an overview of Munks current sheet array theory, and a detailed discussion on the commonmode and loop-mode resonances supported by the unbalanced fed dipoles. A simple method for controlling surface waves on the thick grounded dielectric layers is also discussed. Section IV presents the design of a dual-polarized innite PUMA array that validates the wideband performance of this topology. The paper concludes in Section V. II. THE PUMA ARRAY Though the PUMA array can be implemented in either singleor dual-polarized versions, this paper focuses on the dual-polarized form. The radiating layer of the PUMA array is comprised of printed dipoles in a dual-offset dual-polarized lattice with strong capacitive coupling between cross-polarized elements, as shown in Fig. 1(a). A thin layer of PTFE substrate material, , is placed above and below the dipole layer, shown in Fig. 1(b), which provides

Fig. 1. The PUMA array topology. (a) Top view of dipole layer; (b) crosssectional view of a unit-cell, showing the location where a module split occurs; (c) isometric view of a 2 2 2 PUMA module with exploded dielectric cover layers.

a printable substrate and increases the power handling of the array [29] by preventing dielectric breakdown between capacitor gaps. An electrically thick where is the wavelength at mid-band) cover layer with low permittivity is used for wideband tuning and acts as a wide angle impedance matching (WAIM) layer. The bottom dielectric layer, , with thickness , is also a low permittivity PTFE material that can support plated vias that form the feed lines and shorting posts, as discussed next. Together, the dipole elements, ground plane, and dielectric layers provide wideband performance, based upon the current sheet principle, [12]. However, the feed and dipole arrangements of the PUMA array are unique, and are what chiey differentiate this topology from all other implementations of the CSA. A cross-sectional view of a PUMA unit cell is depicted in Fig. 1(b), which shows the unbalanced feed and the shorting vias. This fundamental change in topology allows unbalanced feed lines to be utilized without exciting the catastrophic common-mode resonance found in 2 D unbalanced fed arrays [11]. More importantly, this feeding method avoids cable organizers, since the unbalanced feed lines do not support the scan-induced common-modes typical of balanced fed arrays. This allows the entire array (radiating elements and feed lines) to be fabricated as a single microwave multilayer PCB, with the feed lines and shorting posts implemented as plated vias, shown in Fig. 1(c). Also, the unbalanced feed lines readily connect to standard 50 interfaces (coax, stripline, microstrip, CPW, etc.) without an external balun. An additional advantage derived from the unbalanced feed arrangement and the dual-offset, dual-polarized offset (egg-crate) lattice is modularity. As shown in Figs. 1(b) and 2, array modules can be formed by

132

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

intersecting planes passing between the feed line vias, therefore a PUMA array can be built and assembled modularly. III. THEORY The PUMA arrays wideband performance depends on careful tuning of its complex structure, and design is carried out using full-wave numerical analysis; nevertheless, key physical models are presented to develop insights and design strategies. A. The Current Sheet Concept As early as 1968, Staiman, [30], recognized that tightly spaced dipole elements placed over a ground plane exhibit wideband impedance characteristics. In 2003, Munk, [12], exploited this concept to develop the wideband CSA array, using tightly-coupled printed horizontal dipoles arranged over a ground plane. Munk developed an insightful transmission line model, [31], illustrating that the capacitive impedance of the dipole layer is compensated by the inductive reactance of the ground plane, resulting in a wideband impedance. Placing a dielectric layer in front of the array is shown to further improve the impedance match. Part of the PUMA array design relies on both the groundplane and dielectric cover compensation principles. Since the main innovation of this work is not the design principle of tightly coupled arrays, the interested reader is referred to [31]. However, in an attempt to show the challenges associated with the feeding of tightly-coupled dipole arrays above a ground plane, e.g., PUMA, a simplied analysis from [29] is used to show that the resistance is much higher than 50 . Consider an innite, single-polarized dipole array placed a distance above a ground plane, as shown in Fig. 3. The dipoles reside in vacuum , and are fed by ideal delta-gap sources at their center. The active resistance of the array in Fig. 3 is, [29],
Fig. 2. Top view of the dual-offset, dual-polarized (egg-crate) PUMA array lattice, showing two of the many possible module sizes, 1 1 2 and 2 2 2, where the module split locations are shown dashed. Circles indicate feed line vias.

Fig. 3. Sketch of innite dipole array located in the x-y plane, above a ground direction. plane, scanned in the

(1) where is the free-space wave impedance, and and are the E- and H-plane element spacings. and are the normal and parallel scalar pattern function components (referenced to the scan plane), and and are the associated transmission coefcient terms that account for the presence of the ground plane. The scalar pattern func, tion is given by is the current along the dipole length . The current where on closely-spaced, tightly-coupled dipole elements is approximately constant, as observed by Hansen, [32], thus . The transmission coefcient terms are two-element array factors consisting of the excited dipole at and its image located at , where . The resulting input resistance at broadside becomes

Fig. 4. Broadside resistance of a single-pol dipole array spaced above . The element current is asa ground plane versus periodicity, with sumed to be constant.

In Fig. 4, is plotted versus for various values and for . For most wideband arrays at mid-band, , leading to . This high resistance level poses a major challenge when matching to standard (unbalanced) 50 interfaces. B. Feeding A successful feeding mechanism for a tightly-coupled dipole array above a ground plane should be able to connect the dipole arms to a feed network or connectors at the back side of the ground plane and transform the high dipole impedance to 50 .

(2)

HOLLAND AND VOUVAKIS: THE PLANAR ULTRAWIDEBAND MODULAR ANTENNA (PUMA) ARRAY

133

Fig. 5. Feeding approaches for tightly-coupled dipole arrays (only one unit cell is shown). (a) Balanced feeding, with 3D cable organizer; (b) unbalanced feeding.

Unbalanced excitation is accomplished using the feed shown in Fig. 5(b), where one dipole arm is connected to the ground with a vertical line, and the other dipole arm is excited directly with the inner conductor of a coax line. This results in unequal magnitude currents on the two feed lines, producing a net vertically polarized current. This net current excites a problematic resonance, as shown in Fig. 6, referred to as a common-mode resonance in Section IV. 2) Common-Mode on Unbalanced Fed Tightly-Coupled Dipole Arrays: When driving the dipoles with unbalanced feeds, the net vertical current distribution couples into a resonance that results in a short circuit in the input impedance of the elements. This resonance is referred to throughout the rest of this paper as the common-mode resonance and occurs near mid-band, at the frequency (see [33] for an in depth analysis of the common-mode). This resonance is unique to unbalanced-fed arrays, and is different from the common-mode issues reported in certain types of balanced-fed arrays, [34]. Away from , there are no vertically polarized ( -polarized, normal to the ground plane) electric elds present in an array unit cell. Conversely, observing the elds at reveals a strong vertical electric eld distribution with a cosine amplitude variation along the diagonal feed-to-feed direction, with nulls at the grounded line of each feed, as shown in Fig. 7. This eld distribution is observed along both diagonal planes, and forms a distribution as shown in the overhead view of Fig. 8, which is a contour map sketch of the vertically-polarized electric eld amplitude. The common-mode resonance occurs when the diagonal plane path length between shorted lines is equal to half a wavelength, i.e., , where an are the E- and H- plane spacings respectively, leading to the common-mode frequency (3) is the speed of light, and is the bottom dielectric where layer, shown in Fig. 1(b). Equation (3) shows that will always occur below the grating lobe onset frequency, inside the operating band. To validate (3), Table I shows the analytic predictions of compared with results obtained using Ansoft/Ansys HFSS, [35], for a single-polarized (along ) unbalanced-fed tightly-coupled dipole array for various , , and , with , where is the guided wavelength in dielectric layer 3). The -polarized E- elds at are mostly conned to dielectric layer 3, thus layers 1 and 2 have negligible effect on and are set to vacuum for these simulations . Good agreement is shown, with error less than 7.60%. Equation (3) implies strategies to tune out of the band. A very large element spacing, , , can move below the lowest frequency of the band, but (the grating lobe onset frequency when scanned to the horizon, ) will also move to the low end of the band, resulting in a narrow usable bandwidth. Alternatively, very small element spacing can move above the operating band, but this results in an over-populated aperture that unnecessarily increases the

Fig. 6. Typical broadside VSWR of single-pol, tightly-coupled dipole arrays with balanced and unbalanced feed arrangements (as shown in Fig. 5). Neither design is optimized for impedance match or bandwidth.

This challenging task could be achieved with either balanced or unbalanced vertical feed lines, as shown in Fig. 5. 1) Balanced Vs. Unbalanced Excitation: Exciting the dipoles with a vertical balanced line requires an external balun at each port, as shown in Fig. 5(a). The balun forces a 180 phase difference between currents and voltages on each line, with the ground plane acting as a third reference conductor with 0V. Assuming ideal balun operation and broadside array excitation, this arrangement results in resonance-free performance over a wide bandwidth, as shown in Fig. 6 (it is noted that ideally fed (e.g., delta gap) tightly-coupled dipole arrays could achieve bandwidths up to 4:1 with broadside , [31], though this example is not optimized). However, when the array is scanned along the E- plane the push-pull currents become unbalanced (due to mutual coupling), resulting in scan-induced anomalies if the feed lines are not appropriately shielded [21]. Practically, the shielding of these lines is achieved using a vertical cable organizer [20]. Cable organizers are 3D metallic structures that require machining, assembly, and soldering, rendering the array non-planar. More importantly, this feeding approach is not easily scaled to very high frequencies.

134

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 9. Top view of a PUMA unit cell, showing new resonant length the shorting posts.

due to

TABLE II COMMON-MODE FREQUENCY CONTROL WITH SHORTING POSTS

Fig. 7. Electric eld distribution at the common-mode frequency unbalanced fed tightly-coupled dipole array.

in an

shorting posts can be used to suppress the resonance, much like the shorting posts used to suppress board resonances in PCB design, without signicantly disrupting the desired radiating elds. The shorting posts are shown in Fig. 1, and an overhead view of the array in Fig. 9 shows the placement of the shorting posts along the dipole arms. In this structure (PUMA), the commonmode resonant frequency is (4)
Fig. 8. Overhead view of the -polarized electric eld distribution at the in an unbalanced fed tightly-coupled dipole common-mode frequency array.

TABLE I COMPARISON OF COMMON-MODE RESONANCE THEORY WITH NUMERICAL SIMULATIONS

number of T/R modules, and makes connectorization difcult. Finally, increasing reduces only marginally, due to the inverse square root relationship. These limited options are thus not practical solutions for controlling the common-mode resonance. 3) Common-Mode Mitigation in PUMA Arrays: Since the resonant elds are polarized normal to the ground plane, orthogonal to either of the desired transmitting polarizations,

where is the shorting post separation from the center of the unit cell. From (4), one can shift up in frequency and out of band by increasing the shorting post separation . The results of an innite single-pol ( -polarized) PUMA array simulation using Ansoft/Ansys HFSS, [35], having , , , , and with the dipole layer located above the ground plane, are shown in Table II. The full-wave numerical results show good agreement with the analytic predictions of (4), demonstrating less than 4% error for all cases. Increasing the shorting post spacing increases , until , where is close to . Thus the novel shorting post arrangement of the PUMA array enables common-mode free, wideband performance at broadside without external baluns or 3D metalized feed structures. As discussed in [33], when unbalanced fed dipole arrays are scanned the problematic common-mode disappears along the D -and H- planes, and is weakly excited along the E-plane, where increases with ; thus broadside suppression of the common-mode implies common-mode free scan operation in the PUMA. C. Low Frequency Loop Resonance in PUMA Despite alleviating the in-band common-mode problem, and displaying minor impact on impedance over most of the

HOLLAND AND VOUVAKIS: THE PLANAR ULTRAWIDEBAND MODULAR ANTENNA (PUMA) ARRAY

135

COMPARISON

OF

TABLE III LOOP-MODE RESONANCE THEORY WITH NUMERICAL SIMULATION

Fig. 10. Impedance of an unbalanced fed Fig. 5(b) tightly-coupled single-polarized dipole array, without and with shorting posts (PUMA).

of loop antennas. A schematic of the current distribution at the rst resonance, , is shown in Fig. 11(a). Currents on the excited feed line ow in the opposite direction from those on its adjacent shorting post, and the currents on both feed lines ow in the same direction, forming a loop between elements. Since the structure is placed over a ground plane, image theory allows the full current loops to be revealed, as in Fig. 11(b). Two loops are formed by the circulating currents, where a small driving loop couples energy into a large resonant loop formed between the shorting posts of adjacent dipole arms, shown in the circuit model of Fig. 11(c). Using the circumference of the resonant loop, is found to be

(5) where represents the effect of the capacitive loading due to inter-element capacitance . The dominant terms of (5) are and , and Table III demonstrates their effect on , using the same simulation model parameters as in Section III-B.3, and with . The predicted values agree with full-wave numerical values with less than 8.35% error. (Note that the analytic values in Table III assume ; a properly chosen may further reduce the error levels). From (5), reducing (wider bandwidth) requires minimal short spacing , and maximal capacitive coupling, leading to an interesting compromise with (4). As shown in Section IV-C, removing the shorting post on the grounded arm increases the loop size sufciently to move well below the operating band, while maintaining an above the operating band. D. Surface Wave Mitigation in PUMA Tightly-coupled dipole arrays such as the CSA and PUMA exhibit a broader frequency response and avoid surface waves when the dielectric between the dipoles and ground is air . In the case of the PUMA, such a choice would make fabrication complicated and difcult to realize above X-band, since the shorting posts and feed lines require mechanical support. It is thus desirable to design the PUMA array using PTFE substrates to allow simple fabrication of the feed lines and shorting posts as plated vias. Unfortunately, this results in a thick grounded dielectric layer that supports surface

Fig. 11. Loop mode resonance model. (a) Current distribution on PUMA ele; (b) current distribution at , using ments at the loop-mode resonance image theory to remove ground plane; (c) circuit model showing a small nonresonant loop driving a large resonant loop.

operating band, the shorting posts signicantly alter the low-frequency impedance behavior of the array. To demonstrate this, the impedance of a PUMA and its no-short counterpart are plotted in Fig. 10. Without shorting posts, the impedance at the low frequency limit is capacitive, and an extrapolation to DC leads to an open-circuit, suggesting a series resonance typical of dipoles. In contrast, the PUMA array exhibits an inductive low frequency impedance, with an eventual short-circuit impedance at DC, suggesting a parallel resonance typical

136

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 12. Perforated dielectric arrangement. (a) Top view of substrate, showing , and radius, , of the holes in the dielectric; (b) top and side spacing view of a single-pol, tightly-coupled dipole array on a perforated dielectric. Fig. 13. Normalized reection coefcient, , variation versus for E-plane and various perforated dielectric hole radii. scan angles at All curves assume the array is conjugately matched to the source impedance at broadside.

waves at certain scan angles, resulting in scan blindnesses inside the desired scan volume. The location of such scan blindnesses can be approximately predicted in , space, by, [36]

TABLE IV SCAN BLINDNESS ANGLES: THEORY VS. NUMERICAL SIMULATIONS

(6) where is the grounded dielectric surface wave propagation constant, is the free space wavenumber, and and are integers. Since increases with and in a PUMA array, the scan blindness angle could potentially move inside the usable scan volume, posing a severe limitation. Since is an important design parameter in achieving wide bandwidth, [31], the only option is to reduce . Currently, the lowest for PTFE materials is (Rogers 5880LZ), but even that is not sufcient to move surface waves out of the scan cone. The effective permittivity can be further reduced by perforating the dielectric layers with a series of holes in the region between the dipoles. The effective permittivity, , is reduced by the volume averaging of the permittivities of the dielectric and air-lled holes, and is given (based on a derivation similar to [37]) by

surface wave occurs at . These full-wave results are compared with the theoretical predictions from (6) and (7) in Table IV, and the results are in excellent agreement. Experience has shown that this method can move by approximately 1020 . IV. DUAL-POLARIZED INFINITE ARRAY DESIGN A dual-polarized PUMA array design is presented to provide a representative example of how the theory and insights presented thus far can be applied to achieve a 5:1 PUMA array with good scan performance out to in all planes. A. Array Geometry

(7) where is the radius of the holes in a rectangular grid as shown in Fig. 12(a). The effectiveness of the holes is demonstrated using a simple single-polarized, tightly-coupled dipole array placed between two dielectrics, of thicknesses with , and backed by a ground plane, shown in Fig. 12. The ideal, gap-fed dipoles are of length and width , with spacing . The array was analyzed at as a doubly-periodic, innite array using Ansoft/Ansys HFSS, [35], for scanning in the E-plane. The hole radius is varied from . The results in Fig. 13 show the unperforated substrate dipole array has a scan blindness at , while the dielectrics with holes of increasing radius lead to progressively larger scan blindness angles. At , the

The overall array arrangement is shown in Fig. 1, where the elements are arranged in a dual-pol, dual-offset (egg-crate) lattice. As arranged in Fig. 1(a), the interdigited capacitors form a rhombic shaped loop. When large capacitance is needed to extend the low frequency limit of the operating band, the mean circumference of this loop becomes approximately half a wavelength, and a new resonance appears when the array is scanned in the H- plane. To concurrently maximize capacitance and avoid this scan resonance, the proposed PUMA design places each polarization on a separate dielectric layer, with parallel plate capacitors formed by the overlapping portions of orthogonally polarized dipole arms, as shown in Fig. 14(a). As elucidated in Section IV-C, the shorting post on the grounded dipole arm is removed to further enhance the low frequency performance. Dielectric layer 3 is perforated with cylindrical holes of radius and is comprised of Rogers 5880LZ , while Dielectric layer 1 is an un-perforated layer

HOLLAND AND VOUVAKIS: THE PLANAR ULTRAWIDEBAND MODULAR ANTENNA (PUMA) ARRAY

137

Fig. 15. Full-wave analysis of array and matching network for broadside scan, showing impedance loci for: the array impedance seen at the ground plane, the impedance seen after the capacitor, and nally the impedance seen at the end of . the impedance transformer. All results referenced to

Fig. 14. A 5:1 dual-pol PUMA array. (a) Top view of unit cell; (b) side view; (c) bottom view of the matching network; (d) circuit model of the matching network.

of Rogers 5880 . Dielectric layers 2 and 4 are thin dielectric bonding layers (Gore SPEEDBOARD C Prepreg bond layer), with , and serve as both a means of joining the dielectric layers, as well as a useful dielectric substrate, which is critical to a backplane matching network, described in Section IV. B. Backplane Matching Network Due to the high impedance of the dipole aperture layer , obtaining an array impedance well matched to can be accomplished in two ways. The rst is to carefully design the array and vertical feed lines to act as impedance transformer sections to provide matching to 50 at the ground plane interface; this method is typically limited to 3:1 bandwidths, [27], [28]. The second option is to use an impedance transformer below the ground plane. Instead

of tuning the array to , the array is tuned to have a specic mismatched impedance prole that, when fed through the backplane matching network, results in a wideband impedance well-matched to 50 . The matching network employed in this design is inspired by a classic wideband impedance matching concept, which uses the combination of a series open-circuit stub and an impedance transformer [38]. However, the proposed matching network avoids the open stub with a substantially smaller series LC network, formed by a parallel plate capacitor and high impedance T-line. The array backplane (PCB 2 in Fig. 14(b) and (c)) contains a printed matching network, comprised of a series capacitor, series inductor, and a transformer section. The capacitor is formed by a pair of circular plates of radius , which are separated by the thin dielectric layer 4. One plate is attached to the end of the feed line plated via, and the other plate is attached to a narrow, high impedance microstrip feed line of length and width that approximates an inductor. This is followed by a microstrip line of width , with an impedance greater than 50 , and has a length . The circuit schematic of this structure is shown in Fig. 14(d). The series capacitor, , introduces a reactance that partially cancels out the inductive reactance of the antenna; additionally, two parasitic shunt capacitors are shown dashed in Fig. 14(d), representing the parasitic capacitances between each plate and the ground plane, though these have second-order impact on performance. Next, a series inductor introduces the reactance that forms a series LC network with the capacitor. Finally, a transmission line with length and propagation constant transforms the impedance to , where , as shown in the circuit schematic of Fig. 14(d). In the schematic, is the antenna impedance at the ground plane (see Fig. 14(b)). The entire structure, array and backplane printed matching network, were analyzed full-wave (no approximate T-line models) using Ansoft HFSS, [35], and the results are shown in Fig. 15. At broadside, the antenna impedance is purposefully tuned to be inductive over most of the operating band, with a poor match to 50 ; however, the combination of

138

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 16. Comparison of the port coupling (H to V) for an optimized PUMA array with one and two shorting posts.

is shown to be centered near , though strongly capacitive. By adding the inductor reactance in series, the entire locus is moved toward the inductive part of the Smith chart (though this step is not shown in order increase the readability of Fig. 15). Finally, the impedance locus is matched to 50 by passing through a quarter wave transformer with ( for this design). The nal impedance is shown to be well matched, with a over the band. This simple, planar matching network is used on all ports, and readily ts inside a unit cell with meandering. As shown in Fig. 14(b), the array can be fabricated as one PCB, labeled PCB 1, and the backplane matching network is fabricated on a separate PCB, labeled PCB 2, these PCBs are then joined by a thick bonding layer, dielectric layer 4, without any conductive electrical connection, further reducing the fabrication complexity. C. Shorting Post Placement As noted in Section IV-A, this design uses only one shorting post, attached to the non-grounded dipole arm. The loop mode resonance, discussed in Section III-C, can be shifted to lower frequencies by maximizing the resonant loop sizeaccomplished here by removing one shorting post. The of the optimized PUMA with one and two shorting posts behaves very similarly above mid-band. However, the cross-polarized coupling (H-V pol coupling) is quite different. As shown in Fig. 16, the double-short PUMA has strong coupling near 1 GHz, due to the presence of the loop mode. Therefore, using only one shorting post increases impedance bandwidth and reduces cross-polarization coupling. D. Numerical Results The array and the backplane matching network were simulated in an innite array environment using Ansoft/Ansys HFSS, [35], assuming PEC for vias and metal layers, and standard dielectric models for layers 2 and 4 (Gore SPEEDBOARD C Prepreg bond layer) and layers 1 (Rogers 5880) and 3 (Rogers 5880LZ). All results are referenced to . 1) Impedance Vs. Scan: Fig. 17 shows the VSWR for various E-plane and H-plane scan angles out to 45 . These results are

Fig. 17. VSWR vs. frequency and scan angle of the innite dual-polarized PUMA array. (a) E-plane; and (b) H-plane. The D-plane impedance (not shown here) is approximately the average of the two.

calculated by exciting one polarization and terminating the other in 50 . As discussed in Section IV-C, the cross-polarization coupling is very low, below over the band, therefore this is a valid means of evaluating performance. The vertical lines in the plot indicate the band edges (1.06 GHz and 5.3 GHz) of this array. Fig. 17(a) shows a maximum broadside VSWR of approximately 2.1, with a VSWR less than 2 over most of the band. A maximum VSWR of 2.5 is observed at the high frequency band edge for scanning in the E-plane. The H-plane results are shown in Fig. 17(b), indicating a maximum VSWR of 2.9 at at the low end of the band, which is a typical rise for dipole arrays. The VSWR in each of these planes is seen to vary little with scan, a benet of the arrays low-prole and WAIM layer. The D-plane results are omitted since they follow an approximate average of the E- and H-planes. 2) Cross-Polarization Vs. Scan: A dual-polarized innite array can radiate arbitrarily polarized plane waves, thus the radiated power can be decomposed into two orthogonally polarized plane waves. For this analysis, the polarizations are chosen according to Ludwigs 3rd denition of cross-pol [39]. A surface is placed parallel to the ground plane above the array, and the power owing through this surface is calculated by integrating the Poynting vector of each plane wave over a unit cell. The coand cross-polarization radiated powers per unit cell are shown in Fig. 18 over the frequency band for scanning in the E- and D-planes; all power levels are plotted in dB and are normalized to the incident power at the input port, and therefore include mismatch, dielectric, and back-radiation losses.

HOLLAND AND VOUVAKIS: THE PLANAR ULTRAWIDEBAND MODULAR ANTENNA (PUMA) ARRAY

139

TABLE V 5:1 DUAL-POL PUMA ARRAY DIMENSIONS

TABLE VI BACK-RADIATION LOSS OF THE 5:1 PUMA

Fig. 18. Co- and cross-polarization (Ludwigs 3rd denition) radiated power vs. frequency and scan angle of the innite dual-polarized PUMA array. (a) E-plane; and (b) D-plane. The H-plane polarization levels (not shown here) are approximately the same as the E-plane.

The co-polarized power in the E-plane, Fig. 18(a), is nearly 0 dB throughout the band, exhibiting minor decreases only due to mismatch loss, indicating good efciency. The cross-polarized power is below over the majority of the band, and reaches a maximum of at the low-end band edge. Overall, the cross-polarization remains very low even at wide scan angles. The H-plane results are omitted, since they are very similar (or better) than the E-plane. For all angles, the D- plane co-polarized power, Fig. 18(b), is very similar to that of the E-plane at broadside. The crosspolarized power is shown to be below over the full band for angles out to , and reaches a maximum of at . The low-prole of the PUMA array helps maintain very low cross-polarization throughout the band. Finally, to demonstrate that the back-radiated power from the matching network is very small, Table VI shows the ratio of the back-radiated power, , to the input power, , for various frequencies. The array is excited at broadside. The back radiation loss is very small , and as expected, it increases slightly with frequency. This behavior is approximately the same for other scan angles as well. It is worth noting that throughout the band the radiation efciency (including mismatch, port coupling, conductor, dielectric and back-radiation losses) of this innite PUMA array was predicted to be above 90%. V. CONCLUSION The proposed PUMA array offers wideband and wide-scan performance in a simple, modular, truly planar architecture that

allows both the array aperture and feeds to be printed using standard microwave fabrication techniques, which results in lowcost and frequency scalable manufacturing. Additionally, the unbalanced feed arrangement removes the necessity for external wideband baluns and cable organizers. The enabling feature is the introduction of shorting posts, which are shown to eliminate a catastrophic resonance due to unbalanced feeding of balanced dipoles. Simple circuit models are presented that provide insight into the shorting post operation, including control of a new low frequency loop mode. The PUMA array uses a thick grounded substrate that supports surface waves, and a method to control the associated scan blindnesses was proposed. Finally, a dual-polarized PUMA array design with a 5:1 bandwidth with at broadside, and approximately cross-polarization out to 45 scan in the D-plane was shown. REFERENCES
[1] G. C. Tavik, C. L. Hilterbrick, J. B. Evins, J. J. Alter, J. G. Crnkovich, J. W. de Graaf, W. Habicht, G. P. Hrin, S. A. Lessin, D. C. Wu, and S. M. Hagewood, The advanced multifunction RF concept, IEEE Trans. Microw. Theory Tech., vol. 53, no. 3, pp. 10091020, Mar. 2005. [2] L. R. Lewis, M. Fassett, and J. Hunt, A broadband stripline array element, presented at the Antenna Applications Symp., Allerton Park, Monticello, IL, Sep. 1974. [3] M. Stasiowski and D. H. Schaubert, Broadband array antenna, presented at the Antenna Applications Symp., Allerton Park, Monticello, IL, Sep. 2008. [4] D. H. Schaubert, A gap-induced element resonance in single-polarized arrays of notch antennas, in Proc. IEEE Antennas and Propagation Society Int. Symp., Jul. 1994, vol. 2, pp. 12641267. [5] H. Holter, Dual-polarized broadband array antenna with BOR-elements, mechanical design and measurements, IEEE Trans. Antennas Propag., vol. 55, no. 2, Feb. 2007. [6] E. W. Lucas, M. A. Mongilio, K. M. Leader, C. P. Stieneke, and J. W. Cassen, Notch radiator elements, U.S. patent 5 175 560, Dec. 29, 1992. [7] J. J. Lee, S. Livingston, and R. Koenig, A low-prole wide-band (5:1) dual-pol array, IEEE Antennas Wireless Propag. Lett., vol. 2, pp. 4649, 2003. [8] J. D. S. Langley, P. S. Hall, and P. Newham, Balanced antipodal Vivaldi antenna for wide bandwidth phased arrays, Inst. Elect. Eng. Proc.-Microw. Antennas Propag., vol. 143, no. 2, pp. 97102, Apr. 1996.

140

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[9] M. Elsallal, Doubly-Mirrored balanced antipodal Vivaldi antenna (Dm-BAVA) for high performance arrays of electrically short, modular elements, Ph.D. dissertation, Dept. Elect. Comput. Engirg., Univ. Massachusett, Amherst, 2007. [10] S. S. Holland, M. N. Vouvakis, and D. H. Schaubert, The Banyan Tree Antenna Array, U.S. patent application 61/230,768, Aug. 3, 2009. [11] S. S. Holland, M. N. Vouvakis, and D. H. Schaubert, A new modular wideband array topology, presented at the Antenna Applications Symp., Allerton Park, Monticello, IL, Sep. 2009. [12] B. Munk et al., A low-prole broadband phased array antenna, in Proc. IEEE Antennas and Propagation Society Int. Symp., Jun. 2003, vol. 2, pp. 448451. [13] H. A. Wheeler, Simple relations derived from a phased-array antenna made of an innite current sheet, IEEE Trans. Antennas Propag., vol. 13, no. 4, pp. 506514, 1965. [14] J. G. Maloney, M. P. Kesler, P. H. Harms, and G. S. Smith, Fragmented aperture antennas and broadband antenna ground planes, U.S. 6,323,809, May 26, 2000. [15] P. Friederich, L. Pringle, L. Fountain, and P. Harms, A new class of broadband planar apertures, presented at the Antenna Applications Symp., Allerton Park, Monticello, IL, Sep. 2001. [16] A. Neto and J. J. Lee, Ultrawideband properties of long slot arrays, IEEE Trans. Antennas Propag., vol. 54, no. 2, pp. 534543, Feb. 2006. [17] J. J. Lee, S. Livingston, R. Koenig, D. Nagata, and L. L. Lai, Compact light weight UHF arrays using long slot apertures, IEEE Trans. Antennas Propag., vol. 54, no. 7, pp. 20092015, Jul. 2006. [18] W. Croswell, T. Durham, M. Jones, D. Schaubert, P. Friederich, and J. Maloney, Wideband antenna arrays, in Modern Antenna Handbook, C. A. Balanis, Ed. Hoboken, NJ: Wiley, 2008. [19] J. J. Lee, S. Livingston, and D. Nagata, A low prole 10:1 (2002000 MHz) wide band long slot array, in Proc. IEEE Antennas and Propagation Society Int. Symp., Jul. 511, 2008, pp. 14. [20] M. Jones and J. Rawnick, A new approach to broadband array design using tightly coupled elements, in Proc. IEEE Military Communications Conf., Oct. 2007, pp. 17. [21] J. R. Bayard, D. H. Schaubert, and M. E. Cooley, E-plane scan performance of innite arrays of dipoles printed on protruding dielectric substrates: Coplanar feed line and E-plane metallic wall effects, IEEE Trans. Antennas Propag., vol. 41, no. 6, pp. 837841, Jun. 1993. [22] G. Oltman, A compensated balun, IEEE Trans. Microw. Theory Tech., vol. 14, no. 3, pp. 112119, Mar. 1966. [23] J.-S. Lim, U.-H. Park, S. Oh, J.-J. Koo, Y.-C. Jeong, and D. Ahn, A 800-to 3200-MHz wideband CPW balun using multistage Wilkinson structure, in IEEE Microwave Symp. Digest, Jun. 2006, pp. 11411144. [24] K. Jung, W. R. Eisenstadt, R. Fox, A. Ogden, and J. Yoon, Broadband active balun using combined cascode-cascade conguration, IEEE Trans. Microw. Theory Tech., vol. 56, no. 8, pp. 17901796, Aug. 2008. [25] M. Ferndahl and H.-O. Vickes, The matrix balun-a transistor-based module for broadband applications, IEEE Trans. Microw. Theory Tech., vol. 57, no. 1, pp. 5360, Jan. 2009. [26] S. S. Holland and M. N. Vouvakis, The planar ultrawideband modular antenna (PUMA) array, U.S. Application 61/230,271, Jul. 31, 2009. [27] S. S. Holland and M. N. Vouvakis, A 721 GHz planar ultrawideband modular antenna array, in Proc. IEEE Antennas and Propagation Society Int. Symp., Toronto, ON, Jul. 1117, 2010, pp. 14. [28] S. S. Holland and M. N. Vouvakis, A fully planar ultrawideband array, in Proc. Antenna Applications Symp., Allerton Park, Monticello, IL, Sep. 2123, 2010, pp. 221232. [29] B. Munk, Frequency Selective Surfaces, Theory and Design. New York: Wiley, 2000.

[30] D. Staiman, M. Breese, and W. Patton, New technique for combining solid-state sources, IEEE J. Solid-State Circuits, vol. Sc-3, no. 3, pp. 238243, Sep. 1968. [31] B. Munk, Finite Antenna Arrays and FSS. New York: Wiley, 2003. [32] R. C. Hansen, Linear connected arrays, IEEE Antennas Wireless Propag. Lett., vol. 3, pp. 154156, 2004. [33] S. S. Holland and M. N. Vouvakis, The banyan tree antenna array, IEEE Trans. Antennas Propag., vol. 59, no. 11, pp. 40604070, Nov. 2011. [34] D. Cavallo, A. Neto, and G. Gerini, Analysis of common-mode resonances in arrays of connected dipoles and possible solutions, presented at the 6th Eur. Radar Conf., Rome, Italy, Sep./Oct. 20/2, 2009. [35] Ansoft HFSS [Online]. Available: www.ansoft.com [36] D. M. Pozar and D. H. Schaubert, Scan blindness in innite phased arrays of printed dipoles, IEEE Trans. Antennas Propag., vol. 32, no. 6, pp. 602610, Jun. 1984. [37] J. B. Muldavin, T. J. Ellis, and G. M. Rebeiz, Tapered slot antennas on thick dielectric substrates using micromachining techniques, presented at the IEEE Antennas and Propagation Society Int. Symp., Jun. 1997. [38] J. A. Nelson and G. Stavis, Impedance matching, transformers and baluns, in Very High-Frequency Techniques, H. J. Reich and L. S. McDowell, Eds. New York: McGraw-Hill, 1947. [39] A. Ludwig, The denition of cross polarization, IEEE Trans. Antennas Propag., vol. 21, pp. 116119, Jan. 1973.

Steven S. Holland (S05M11) was born in Chicago, IL, in 1984. He received the B.S. degree in electrical engineering from the Milwaukee School of Engineering (MSOE), Milwaukee, WI, in 2006, and the M.S. and Ph.D. degrees in electrical and computer engineering from the University of Massachusetts Amherst, in 2008 and 2011, respectively. From 2006 to 2011, he was a Research Assistant working in the Antennas and Propagation Laboratory, Department of Electrical and Computer Engineering, University of Massachusetts Amherst. Currently, he is a Senior Sensors Engineer with MITRE Corporation, Bedford, MA. His research interests include ultrawideband antenna arrays, electrically small antennas, Radar systems, and HF antennas and ionospheric propagation Dr. Holland is a member of Tau Beta Pi.

Marinos Vouvakis (S99M05) received the Diploma degree in electrical engineering, from Democritus University of Thrace (DUTH), Xanthi, Greece, in 1999, and the M.S. degree from Arizona State University (ASU), Tempe, in 2002 and the Ph.D. degree from The Ohio State University (OSU), Columbus, in 2005, both in electrical and computer engineering. Currently he is an Associate Professor with the Center for Advanced Sensor and Communication Antennas, Electrical and Computer Engineering Department, University of Massachusetts at Amherst. His research interests are in the area of computational electromagnetics with emphasis on domain decomposition, fast nite element and integral equation methods, hybrid methods, model order reduction and unstructured meshing for electromagnetic radiation and scattering applications. His interests extend to the design and manufacturing of ultrawideband phased array systems.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

141

A Two-Channel Time Modulated Linear Array With Adaptive Beamforming

Yizhen Tong and Alan Tennant, Member, IEEE
AbstractA Time modulated linear array (TMLA) can be congured to perform many of the functions of phased array antennas but at much lower cost as they do not require phase shifters. However, conventional time modulated linear arrays which are congured for beam steering based on a single output channel topology. Such a topology is inefcient in terms of time utilization of the array elements. In this contribution we have proposed a multiple output channels time modulated linear array which exploits the time redundancy of conventional systems. The concept is introduced by considering conventional harmonic beam steering and then extending the problem to two-channel adaptive beamforming. Index TermsAntenna linear array, phased array, time modulation.

I. INTRODUCTION

HE concept of using time as an additional parameter, or a fourth dimension, in the design and control of array antennas was rst proposed by Shanks and Bickmore in the late 1950s [1]. This work was then advanced both theoretically and experimentally by Kummer et al. in the early 1960s [2]. The fundamental principle of time modulated or time switched, linear array is to periodically energize the elements of the array by switching each element on and off using high speed RF switches in such a way that the pattern radiated by the array conforms to a prescribed function. There are two basic functions that can be realized using time modulated linear arrays: pattern shaping and harmonics beam steering. To implement pattern shaping, or pattern synthesis, in a time modulated array the elements of the array are periodically energized in such a way that the time averaged effective-amplitude distribution across the array equates to that of a conventional array weighting function such as a low sidelobe Taylor distribution. Although this technique can be successfully used to provide low-sidelobe level radiation patterns at the fundamental operating frequency of the array, the process of periodic element switching also generates radiation at harmonics of the modulation frequency. In general such harmonics are undesirable as they reduce the efciency of the array and may also interfere with other bands of the frequency spectrum [3]. Consequently, much of the recent research into

Manuscript received February 02, 2011; revised May 16, 2011; accepted July 15, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with Department of Electronic and Electrical Engineering, University of Shefeld, Shefeld, S1 3JD, U.K. (e-mail: elp08yt@shefeld.ac.uk; a.tennant@shefeld.ac.uk). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167936

time modulated arrays has addressed the problem of minimizing sideband levels and improving radiation efciency using various adaptive optimization techniques. Yang et al. addressed this issue by investigating the use of various adaptive optimization techniques, including genetic algorithms, to calculate the on-times of array elements required to reduce both the sidelobe and sideband levels [4]. The same authors also investigated the use of a differential evolution algorithm to optimize the inter-element spacing of the array element in order to suppress sidelobe levels of a time modulated linear array [5]. Other approaches to sidebands reduction have been considered by various researchers and included the application of simulated annealing, particle swarm optimization, and element thinning via differential evolution and invasive weed optimization [6][14]. Although the harmonic patterns generated by a time modulated array are often undesirable, there are applications in which such harmonic beams can be exploited and this was also investigated by Shanks who described the process of harmonic beam steering in time modulated linear arrays [15]. In this application the elements along the length of the array are sequentially energized. This effectively introduces a linear time delay in the switching sequence across the face of the array, and generates harmonic radiation patterns which have a directional response at prescribed steering angles. Therefore a properly controlled time modulated array can be congured to replicate the properties of an electronically scanned array antenna. Crucially however, the time modulated array achieves this function without the use of expensive phase shifters. Recent work in this area has also demonstrated how a time modulated linear array can be congured to provide harmonic beam steering and simultaneous control of sidelobe levels by adjusting the switching time of the array elements [16]. Other recent papers have considered how time modulated linear array can be employed for null steering applications for direction nding [17], [18], and their use in pulsed Doppler radar, direction of arrival estimation and phase switched screens [19][21]. Research into time modulated array has increased over the past decade as evidenced by the references detailed above. However, all of the previously described time modulated linear arrays have a common hardware architecture in which the signals from the switched array elements are combined to provide a single output channel at any given frequency. Although the use of bandpass ltering can provide multiple array outputs at the fundamental and harmonic frequencies [22], these systems do not exploit the potential time redundancy in time modulated linear array. In the case of harmonic beam steering using an N element array, the conventional time modulated linear array architecture is inefcient as the elements of the array are only utilized

0018-926X/$26.00 2011 IEEE

142

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

for 1/N of the array switching period. In this paper we therefore propose an alternative array architecture in which multi-throw switches are employed to provide multiple and simultaneous output channels. The approach is described and applied to a specic example of a two-channel communication system employing adaptive beamforming functions. II. THEORETICAL BACKGROUND Consider a linear array of N identical elements with uniform spacing d, positioned along the z-axis of a conventional co-ordinate system as illustrated in Fig. 1. Let each element of the array be energized by a sinusoidal signal of angular frequency for a period dened by: (1) (6) Equations (5) and (6) allow any conventional linear array weighting function to be implemented by an equivalent time sequence at the mth harmonic pattern radiated by the array and these results will be used later in the paper to realize adaptive beamforming functions. However, in order to introduce the concept of redundancy in time modulated linear array, let us simplify our analysis to consider the example of a time modulated linear array in which each element is energized sequentially with equal, normalized duration times given by 1/N. In this case the time modulated linear array factor may be expressed explicitly as:

Fig. 1. Conventional single channel time modulated linear array receiving structure.

elsewhere

and represent the switch-on and switch-off times where of element i respectively and is the overall switching period. The time-modulated linear array factor can be formulated as [3]:

(2) represent the Fourier series coefcients of where ned by: de-

(3) is the nominal operating angular frequency of the array, is the angular frequency of the element switching sequence given by , m is the number of the harmonics and represent the complex static element weights. Evaluating the Fourier coefcients for the case in which all the static weights are unity produces the following result:

(4) Hence the Fourier coefcients may be interpreted as complex weights in which the term controls the amplitude of the array elements and the term introduces a progressive linear phase shift at the harmonic frequencies dened by m. Now assume we wish to apply a time average weighting function to the array dened by and which is effective at the mth harmonic radiation pattern. To implement this weighting function using element switching we use the analysis described in [16] to formulate appropriate expressions for the switch-on and switch-off times of the array elements as:

(5)

(7) Equation (7) indicates that for the special case of sequential element switching, the harmonic beams of the array are progressively scanned in angle to locations determined by . Consider now the specic case of a 16 element array in which the array elements are sequentially energized for a normalized period of 1/16. This switching sequence is shown in Fig. 2 and the corresponding array factors at the fundamental and rst 5 positive harmonic frequencies are shown in Fig. 3. Examination of Fig. 2 demonstrates that for this particular switching sequence each element of the array is only utilized for 1/16 of the switching period and is effectively redundant for the remaining 15/16 of the period. To more fully utilize the capabilities of a time modulated linear array we propose modifying the feeding structure and switching mechanism controlling the array elements from a simple single channel system with SPST switches, to a multi-channel system with multi-throw switches. The use of SPMT switches does not incur any signicant drawbacks in terms of cost or complexity particularly for low values of M. Theoretically, an N element linear array could be congured to provide up to N independent channels (depending on the switching sequence used) but we will examine the case of a

TONG AND TENNANT: A TWO-CHANNEL TIME MODULATED LINEAR ARRAY WITH ADAPTIVE BEAMFORMING

143

Fig. 2. Conventional sequential switching sequence for a 16 element time modulated array designed to generate a rst positive harmonic beam pointing at the direction of 7.2 .

Fig. 4. Example of a two-channel time modulated linear array receiving structure.

Fig. 3. Array factors at the fundamental and rst 5 positive harmonic frequencies produced by the switching sequence in Fig. 2 (relative maximum values of the rst 5 positive harmonics beams are 0.0560 dB, 0.2260 dB, 0.5119 dB, 0.9121 dB and 1.4431 dB respectively).

Fig. 5. The corresponding time sequences for the two-channel time modulated linear array.

III. A TWO-CHANNEL TMLA FOR ADAPTIVE BEAMFORMING In this section we will examine the specic implementation of a two-channel TMLA congured for use in an adaptive beamforming system. However, to introduce the concept we will initially consider a single channel TMLA and apply adaptive beamforming functions at the rst positive harmonic frequency of the array. The following analysis relates to a conventional 16 elements linear array of half wavelength spacing isotropic elements. The array is assumed to receive a binary phase shift keying (BPSK) signal in an environment that is subject to additive white Gaussian noise (AWGN) with zero mean and unity variance. In the following examples, the amount of BPSK signal transmitted and AWGN is considered to be . Under these assumptions the general form of the output signal from the antenna array can be expressed as [23]: (8) where X(t) is signal received by the array elements, , S(t) is the transmitted signal, I(t) is the interfering signal, N(t) represents AWGN, and are the

simple two-channel system. Fig. 4 shows a generalized diagram representing an N element time modulated linear array congured to utilize the same elements to provide two independent output channels. This has been achieved by replacing the simple binary switches of a single channel array with multithrow switches. In this particular example SP3T switches have been used with the third switch state corresponding to the element off state. Fig. 4 indicates that a suitably congured time modulated linear array can be congured to provide multiple output channels which utilize the same elements. To illustrate this we return to our example of a 16 element time modulated linear array which has been congured for two channels operation. Fig. 5 shows two possible switching sequences which could be used to control the array and provide two simultaneous output signals. In this example the two switching sequences provide identical outputs at channel one and channel two (ignoring a relative phase offset) but in the following example we examine a more useful implementation of a two channels time modulated linear array system which is congured for a communication system employing adaptive beamforming functions.

144

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

phase delays of desired and interference signal respectively, W is a set of complex element weights, and is the complex conjugate transpose of W. The desired weights W could be calculated using one of the many techniques described in the literature but in our implementation we have used the adaptive least mean square (LMS) algorithm [23]. Once the required set of weights has been computed, we can use (5) and (6) for the case of to obtain the corresponding TMLA switching sequence. For completeness, we also note that the ideal, theoretical bit error rate (BER) for a BPSK signal with power S and noise power N upon reception is given by [24]:

(9) where . Armed with the above theoretical model we may now investigate some specic beamforming examples applied to a single channel TMLA. Consider rst a scenario in which we have one desired signal and one interference signal arriving at the array from the directions of 40 and respectively and with a signal to interference ratio (SIR) of 0 dB. Our initial objective is to calculate a set of conventional elements weights which will optimize the array out SINR and this may be achieved by using the previously mentioned LMS algorithm. The complex weights are then used to obtain the corresponding element switching sequence necessary to implement the beamforming response at a specic array harmonic which, in our example, is chosen to be the rst positive harmonic . The TMLA switching time sequence for this example is shown in Fig. 6 and the resultant array factor at the rst positive harmonic is shown in Fig. 7 where it is observed that the main beam response is in the direction of the desired signal and a pattern null has been formed in the direction of the interfering signal. The simulated bit error rate graph of this example is compared to the ideal, theoretical curve in Fig. 8 where it observed to be in good agreement. It should be noted that although equation (9) provides an ideal solution for the BER of a BPSK modulation scheme, there is no closed form expression for the BER performance of the TMLA considered here and we therefore used a numerical technique to estimate the BER based on a Monte Carlo approach. In the Monte Carlo simulations we used 50 random runs in order to get the BER curves. For the bit error rate calculations it is assumed that the signals at the rst harmonic frequency have been properly down-converted ltered to reject the fundamental frequency and also other harmonic frequencies (including negative harmonics). Although it may be argued that such an approach is inefcient as the power in the remaining harmonics has been discarded, we do not consider this to be an issue for two reasons: rstly the power level of the received signal is very low and hence there is no issue associated with signicant power loss; secondly, and most importantly, it is the pattern gain of the array at the desired harmonic frequency which determines the performance of the system. However, for completeness and based on the approach presented in [3], the efciencies at the fundamental

Fig. 6. One channel time sequence designed to produce a main beam at 40 and a null located at .

Fig. 7. Array factor at rst positive harmonic corresponding to time sequence in Fig. 6.

Fig. 8. BER plot of a one channel adaptive beamforming system against the ideal and theoretical BER.

and rst 5 positive harmonics of this proposed technique are approximately 9.7%, 9.4%, 8.5%, 7.2%, 5.7% and 4.13% respectively. We now examine a more complex scenario in which we have two desired signals incident on the array from two independent

TONG AND TENNANT: A TWO-CHANNEL TIME MODULATED LINEAR ARRAY WITH ADAPTIVE BEAMFORMING

145

Fig. 9. Two-channel adaptive beamforming time sequences (BlackUser 1, GrayUser 2, and Whiteoverlapping region).

Fig. 10. The two-channel adaptive beamforming radiation patterns when there . is no time offset

directions and use two channels TMLA to extract and calculate the BER from both sources simultaneously. In the scenario we wish to study we have two users (user 1 and user 2) transmitting signals to the array from directions of 40 and respectively. Initially, the signals from user 1 and user 2 are both assumed to have unit magnitude and hence the power ratio between them in this case is 0 dB. Firstly, consider channel 1 of the two channels TMLA and assume that the desired signal is from user 1. In this case the signal from user 2 becomes a source of interference and hence we may use the previously described procedure to obtain the desired switching time sequences. The result is an array radiation pattern identical to that shown in Fig. 7 with a main beam response in the direction of user 1 and a null formed towards the interference signal. Subsequently, we consider channel 2 of the array as an independent output in which user 2 provides the desired signal and user 1 becomes the source of interference. A similar procedure to that described for channel 1 is then used to calculate the time sequence of channel 2. The switching time sequences required for simultaneous performance of both channel 1 and channel 2 are presented in Fig. 9, where it is observed that for certain elements of the array there is a conict as both channel 1 and channel 2 require connection to the same element at the same time. This is shown by the overlap region in Fig. 9 and, in this example, applied to elements 1, 3, 5, 7, 9 and 11. There are many ways to address this overlap problem depending on the particular scenario being considered. For example, if the signal from channel 1 was deemed to have a higher priority than that of channel 2, then the time sequence of channel 1 would also take priority and be left unchanged. Channel 2 however would then have a non-optimum time sequence and its performance would be severely degraded. In another scenario it may be decided to share the conicting overlap region equally between channels 1 and 2. Adopting this approach degrades the response of both channels and an example of this is shown in Fig. 10, where it is observed that although the main-beam response of both channel 1 and channel 2 is as desired, the required pattern nulls in the directions of the interfering signals have not been formed. Various other time-slot sharing strategies could

Fig. 11. Channel One BER plot at various time offset values . interference ratio is 0 dB

when signal to

be examined but an alternative approach is to introduce a small time offset between the switching sequences of channels 1 and 2 such that the degree of mutual overlap is decreased or eliminated. We may therefore modify the two normalized switching sequences as follows:

(10) (11) where and are the original switching time sequences of channels 1 and 2 when there is no time offset applied, and are the modied switching time sequences of channels 1 and 2 respectively and represent the time offset. To illustrate this approach let us return to our two-channel example in which the desired signals for user 1 and user 2 are from the directions of 40 and respectively. Initially consider the case where both signals are equal in magnitude so that the corresponding SIR for both channel 1 and channel 2 is 0 dB. Figs. 11 and 12 show the simulated BER for this scenario for values of time offset in the range to . For this particular example, and because the SIR is relatively small, the

146

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 12. Channel Two BER plot at various time offset values . interference ratio is 0 dB

when signal to

Fig. 14. Channel Two BER plot at various values of time offset when signal to . interference ratio is 20 dB

Fig. 13. Channel One BER plot at various values of time offset when signal to . interference ratio is 20 dB

Fig. 15. Two-channel adaptive beamforming time sequences when the time (BlackUser 1, GrayUser 2). offset

BER plots of both channel 1 and channel 2 do not deviate significantly from the ideal result, even for the case of no time offset . Now consider a more demanding problem in which the SIR is reduced to 20 dB. The BER curves for this example are given in Figs. 13 and 14 for channel 1 and channel 2 respectively. In this example the BER performance corresponding to the case when there is no time offset between the two time sequences is signicantly degraded. However, as the time offset between the two channels is increased, the BER performance gradually improves and when there is no overlap between the two time sequences ( in Figs. 13 and 14), the BER curve approaches that of the ideal case. The time sequences of channel 1 and channel 2 which result from implementing a time offset of 0.05 are shown in Fig. 15 and the corresponding radiation pattern plots are illustrated in Fig. 16. It is now apparent that deep nulls have been formed in the directions of the interfering signals. For this particular example a normalized offset value of 0.05 is sufcient to modify the time sequences such that they do not overlap. However, any time adjustment applied to the original, ideal switching sequences will inevitably compromise the

performance of the system. In particular the addition of a time offset to the switching sequences will affect the original desired complex element weights W and hence result in reduced BER performance. To illustrate this, Fig. 13 and Fig. 14 show the simulated BER curve of the system when the normalized time offset value is increased to 0.12. In this example, although the two time sequences do not overlap, they have produced a severe degradation of the BER performance on both channels. Although the use of time offset can often be used to solve the problem of overlapping switching sequences, this is not always the case. An example where the overlap problem cannot be solved will start to emerge when signals from two users are located close to each other. In this respect the proposed two channel system is subject to the general constraints of beam forming as applied to any array system. IV. CONCLUSION A new topology for a time-modulated linear array congured for harmonic beam steering has been presented. The topology exploits the redundancy in conventional systems designs and allows the beam forming network of the array to be more fully utilized. In a conventional, N element TMLA each element of the array is only energized for 1/N of the switching period.

TONG AND TENNANT: A TWO-CHANNEL TIME MODULATED LINEAR ARRAY WITH ADAPTIVE BEAMFORMING

147

Fig. 16. Radiation patterns of channels 1 and channel 2 corresponding to the . time sequences in Fig. 15

This redundancy can easily be exploited by modifying the array switching network to incorporate multi-throw switches to provide several independent array output channels. We have illustrated the concept by considering the specic examples of a TMLA congured to perform adaptive beamforming. The system provides two, simultaneous output channels at the same harmonic frequency and each channel can perform independent beamforming functions. Such an approach doubles the capacity of a single channel system for minimal increase in system complexity. REFERENCES
[1] H. E. Shanks and R. W. Bickmore, Four-dimensional electromagnetic radiators, Canadian J. Phys., vol. 37, pp. 263275, 1959. [2] W. H. Kummer, A. T. Villeneuve, T. S. Fong, and F. G. Terrio, Ultra-low sidelobes from time-modulated arrays, IEEE Trans. Antennas Propag., vol. 11, pp. 633639, Nov. 1963. [3] J. C. Bregains, J. Fondevila, G. Franceschetti, and F. Ares, Signal radiation and power losses of time-modulated arrays, IEEE Trans. Antennas Propag., vol. 56, no. 6, pp. 17991804, Jun. 2008. [4] S. Yang, Y. B. Gan, A. Qing, and P. K. Tan, Design of uniform amplitude time modulated linear array with optimized time sequences, IEEE Trans. Antennas Propag., vol. 53, no. 7, pp. 23372339, Jul. 2005. [5] G. Li, S. Yang, M. Huang, and Z. Nie, Sidelobe suppression in time modulated linear arrays with unequal element spacing, J. Electromagn. Waves Applicat., vol. 24, no. 56, pp. 775783, 2010. [6] J. Fondevila, J. C. Bregains, F. Ares, and E. Moreno, Optimising uniformly excited linear arrays through time modulation, IEEE Antennas Wireless Propag. Lett., vol. 3, pp. 298300, 2004. [7] L. Manica, P. Rocca, L. Poli, and A. Massa, Almost time-independent performance in time-modulated linear arrays, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 843846, Aug. 2009. [8] L. Poli, P. Rocca, L. Manica, and A. Massa, Handling sideband radiations in time-modulated arrays through particle swarm optimization, IEEE Trans. Antennas Propag., vol. 58, no. 4, pp. 14081411, Apr. 2010. [9] P. Rocca, L. Poli, G. Oliveri, and A. Massa, Synthesis of time-modulated planar arrays with controlled harmonic radiations, J. Electromagn. Waves Applicat., vol. 24, no. 56, pp. 827838, 2010. [10] P. Rocca, L. Manica, L. Poli, and A. Massa, Synthesis of compromise sum-difference arrays through time-modulation, IET Radar Sonar Navig., vol. 3, pp. 630637, Nov. 2009.

[11] L. Poli, P. Rocca, L. Manica, and A. Massa, Pattern synthesis in timemodulated linear arrays through pulse shifting, IET Microw. Antennas Propag., vol. 4, no. 9, pp. 11571164, Sep. 2010. [12] L. Poli, P. Rocca, L. Manica, and A. Massa, Time modulated planar arraysAnalysis and optimisation of the sideband radiations, IET Microw. Antennas Propag., vol. 4, no. 9, pp. 11651171, Sep. 2010. [13] E. Aksoy and E. Afacan, Thinned nonuniform amplitude time-modulated linear arrays, IEEE Antennas Wireless Propag. Lett., vol. 9, pp. 514517, May 2010. [14] A. Basak, S. Pal, S. Das, A. Abraham, and V. Snasel, A modied invasive weed optimization algorithm for time-modulated linear antenna array synthesis, in Proc. IEEE Congress on Evolutionary Computation, Jul. 2010, pp. 18. [15] H. E. Shanks, A new technique for electronic scanning, IEEE Trans. Antennas Propag., vol. 9, no. 2, pp. 162166, Mar. 1961. [16] Y. Tong and A. Tennant, Simultaneous control of sidelobe level and harmonic beam steering in time-modulated linear arrays, IET Electron. Lett., vol. 46, no. 3, pp. 200202, 2010. [17] A. Tennant and B. Chambers, A two-element time-modulated array with direction-nding properties, IEEE Antennas Wireless Propag. Lett., vol. 6, pp. 6465, 2007. [18] A. Tennant, Experimental two-element time-modulated direction nding array, IEEE Trans. Antennas Propag., vol. 58, no. 3, pp. 986988, Mar. 2010. [19] G. Li, S. Yang, and Z. Nie, A study on the application of time modulated antenna arrays to airborne pulsed Doppler radar, IEEE Trans. Antennas Propag., vol. 57, no. 5, pp. 15781582, May 2009. [20] G. Li, S. Yang, and Z. Nie, Direction of arrival estimation in time modulated linear arrays with unidirectional phase center motion, IEEE Trans. Antennas Propag., vol. 58, no. 4, pp. 11051111, Apr. 2010. [21] A. Tennant and B. Chambers, Time-switched array analysis of phase switched screens, IEEE Trans. Antennas Propag., vol. 57, no. 3, pp. 808812, Mar. 2009. [22] G. Li, S. Yang, Y. Chen, and Z. Nie, A novel electronic beam steering technique in time modulated antenna array, Progr. Electromagn. Res., vol. 97, pp. 391405, 2009. [23] B. Widrow, P. E. Mantey, L. J. Grifths, and B. B. Goode, Adaptive antenna systems, Proc. IEEE, vol. 55, pp. 21432159, Dec. 1967. [24] J. G. Proakis and M. Salehi, Digital Communications, 5th ed. New York: McGraw-Hill, Nov. 2007. Yizhen Tong was born in Wuhan, China. He received the B.Eng. degree in electronic and electrical engineering in communications from the University of Shefeld, Shefeld, U.K., in 2005, where he is currently working toward the Ph.D. degree. His current research interests include antenna arrays, time modulation and array signal processing.

Alan Tennant (M92) received the B.Eng. degree in electronic engineering and the Ph.D. degree in microwave engineering from the University of Shefeld, Shefeld, U.K., in 1985 and 1992, respectively. Previously, he was with BAe Systems, Stevenage. He then joined DERA where he worked on phased array antenna systems before taking up an academic post at Hull University. He returned to Shefeld University in 2001 as a Senior Lecturer in the Communications and Radar Group where he is involved in research into techniques, materials and signal processing for adaptive radar signature management, novel three-dimensional phased array antenna topologies, acoustic array systems, and new research into time-modulated array antennas. He has published over 100 academic papers including several invited papers on adaptive stealth technology. His research has attracted substantial funding from both industry and government sources.

148

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Aperiodic Array Layout Optimization by the Constraint Relaxation Approach

Theodoros N. Kaifas, Senior Member, IEEE, Dimitrios G. Babas, Member, IEEE, George S. Miaris, Katherine Siakavara, Member, IEEE, Elias E. Vaadis, Member, IEEE, and John N. Sahalos, Life Fellow, IEEE

AbstractAn optimization procedure for the layout assessment of electrically large but nite planar arrays is presented. The synthesis takes into account the desired directivity pattern that is prescribed employing bound constraints. Moreover, the size of the radiators is taken into account, which results in a hard nonoverlapping, between the elements, constraint. The latter should not be violated if we want the attained solution not only to obey the far-eld mask, but also to be physically realizable. As stated, the optimization problem is twofold. An antenna design is associated with a packing problem. In order to take the constraints on the layout into account and solve the whole problem, we propose the constraint relaxation approach, which is equipped with a packing algorithm. Our study is applied to various initial geometries, and the resulting arrays appear to comply with the desired pattern and the nonoverlapping constraint. Several examples for different cases including symmetric arrays and a study on maximally sparse arrays are presented, which show the applicability and merit of the method. Index TermsAperiodic arrays, layout optimization, maximally sparse arrays, packing problem.

I. INTRODUCTION EOMETRY synthesis methods focus on forming the actual layout of the array by placing the radiating elements in the appropriate positions and with the appropriate orientations. In the literature, most of the array design techniques determine the unknown excitation for a given, often periodic, arrangement of the radiating elements. The excitation for a desired electric eld is found by using one of the classical procedures [1][6]. The problem is much different if the array layout has to be derived. In this case, the underlying grid usually exhibits nonperiodicity. Thus, the designed arrays are called aperiodic. They are also known as sparse arrays. Aperiodic arrays are useful for a limited eld of view systems [2]. Among other characteristics, they exhibit the following [7]: 1) half-power beamwidth (HPBW) (selectivity), which depends mainly on the occupied aperture; 2) sidelobe level (SLL), which is regulated by the total number of elements; and 3) wide operational bandwidth. The desired specications
Manuscript received February 27, 2010; revised March 03, 2011; accepted May 06, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. T. N. Kaifas, D. G. Babas, G. S. Miaris, K. Siakavara, and E. E. Vaadis are with the Department of Physics, Aristotle University of Thessaloniki, Thessaloniki 54124, Greece (e-mail: tkaif@skiathos.physics.auth.gr). J. N. Sahalos was with the Department of Physics, Aristotle University of Thessaloniki, Thessaloniki 54124, Greece. He is now with the Electrical and Computer Engineering Department, University of Nicosia, Nicosia 1700, Cyprus. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167901

that are usually taken into account range from wideband (multiband) response [8] to far-eld pattern indices (SLL, HPBW, and Directivity) [9][13]. For the design of planar arrays, either deterministic or stochastic approaches can be used. The deterministic approaches usually rely on the equivalent space tapering. They approximate an aperture current distribution with a uniformly excited and nonuniformly spaced array [14][18]. On the other hand, most of the stochastic approaches are based on the use of evolutionary optimization algorithms [19][23]. For aperiodic arrays, the common characteristic of the geometry synthesis methods is the adjustment of the spacing among the array elements according to certain criteria [24], [25]. To date, these studies do not systematically take the element characteristics (size, shape, and orientation) into account. Obviously, these extra parameters increase the difculty of obtaining a solution. Indeed, the solution procedure, besides (a) shaping the far-eld, must also (b) properly arrange the elements of the array producing a physically realizable layout. The above problems, (a) and (b), are of increased complexity on their own, and thus their coupled version, which is addressed here, makes the solution process even more difcult to cope with. Moreover, problem (b) should be classied as a packing one [26] since the goal is to place the element radiators into the host area without overlapping while optimizing the directivity pattern. To come up with acceptable layout solutions, besides the geometry methods, there is also a set of different design approaches in which some kind of spatial order still stands. Due to this order, the elements must be properly placed in adequate space. One can note approaches like [23] and [27], in which the thinning method is used. According to this method, in a fully populated array, a number of element radiators are terminated by matched loads. This is achieved by means of a probability distribution function that resembles an underlying aperture current distribution that produces an acceptable far eld. As a subset of this method, one can mention the difference sets approach [27][29]. According to this, no underlying aperture distribution is assumed, but rather the mathematic approach and the denition of the difference sets method are followed. In addition, the approach of the polyomino-based array [30] should be cited, in which techniques similar to the ones used for the puzzle arrangement are used. Finally, we can mention the tiling-based arrays [8] and [31], in which the available area of the host is divided into properly chosen tiles, usually with the aim of producing a wideband antenna array. Most of the above methods sacrice a signicant percentage of the degrees of freedom (DoF) in order not to tackle the

0018-926X/$26.00 2011 IEEE

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

149

layout arrangement problem directly. In contrast, our approach simultaneously solves both the far-eld and the layout/packing problem in order to exploit the whole number of the DoF and eventually come up with the desired results. It should be noted that in the present effort, which is an extension of the Perturbation Method, [11], [12], and [24], the element size, shape, and orientation are taken into account. The design focuses on the directivity pattern, and the optimization search is, obviously, feasibility oriented. Our problem is stated as a minimization of a nonlinear cost function that encompasses the various constraints and prescriptions. As the number of elements increases, the quest for the global minimum-optimal solution becomes more and more counterproductive since it degrades the efciency of the search irrespective of the approach followed. To bypass this problem here, the cost function utilizes bound constraints, and in this way the search is concerned only with nding a solution that complies with all the requirements. If a better solution with stricter bounds is needed, the effort should be repeated. As soon as an initial array and a far-eld target mask are given, the optimization procedure can be initiated. We perturb the element positions and orientations by taking the desired pattern into account. The perturbation procedure is applied simultaneously along the independent variables (element phase center coordinates and element orientations) for each radiator. The nal conguration results from the last iteration, in which the solution adequately fullls the constraints. Our study has been applied for various initial geometries, and the resulting arrays comply with the desired pattern without overlapping between the elements. Several examples for different cases and a study on maximally sparse arrays ([32] and references therein) are presented and show the applicability of the method. Our paper is arranged as follows. Section II provides the statement of the problem and reviews its complexity. In Section III, the optimization algorithm is presented. In fact, the perturbation method is accompanied by a layout optimization approach, which solves the underlying packing problem. In Section IV, a detailed performance evaluation is given, and the paper ends with the conclusions. II. PROBLEM STATEMENT AND COMPLEXITY REVIEW One of the key problems in array design is the shaping of the directivity pattern to form a pencil beam obeying a mask like the one given in Fig. 1. This mask could describe a common design problem of a multibeam satellite system based on a direct radiating array (DRA). An axial symmetry for the mask, around the arrays broadside, is assumed. Angles and dene the end of the passband (to use the low-pass lter notation in order to make things easier to follow) and the start of the stopband, respectively. The corresponding pass and stop directivity levels are dened by and . Thus, for , the produced directivity pattern should be higher than , while for , it should be lower than . In mathematical form, the directivity pattern of the array, , should follow (1)

Fig. 1. (a) Bound prescription of the desired directivity pattern of a planar . array. (b) Zoom on the main beam area. Solid lineupper bound: . Dashed linelower bound:

where is the upper bound and is the lower bound of the mask. As far as the array layout is concerned, the nonoverlapping between the elements should be included in the requirements. For example, if we assume the array is assembled of circular elements, each of them (the th) identied by the phase center Cartesian coordinates and and radius , then the following constraint should be valid: (2) Indeed, it is readily apparent that since the element size/area is taken into account, the array layout should obey an additional set of constraints. These constraints on the layout should not be violated if we want the attained solution not only to obey the mask constraints, but also to be physically realizable. The approach that we use to solve this problem is given in Section III.

150

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

III. OPTIMIZATION ALGORITHM As stated, the optimization problem is twofold. An antenna design is associated with a packing problem [26]. By relaxing the packing problem (the size of the radiators is not taken into account), the directivity pattern design is solved by a perturbation method, which is presented in Section III-A. In order to take the constraints on the layout into account and solve the whole problem, we propose the constraint relaxation approach. Together with the packing algorithm, they are given in Section III-B. A. Orthogonal Perturbation Method (OPM) In the orthogonal perturbation method, we perturb the position and the orientation of the elements. The total electric eld of an array composed of radiators is the summation of the element elds and can be expressed in matrix form as follows: (3) where stands for transpose. The th element, , of the column vector is the electric far eld produced by the th radiator. The th element, , of is the respective complex excitation coefcient. In the present study, we keep the excitation vector constant, while in (3) we permit the vector to change, following the array layout modication. It is assumed that the number and type of elements of the planar array are held constant and each element (the th) contributes with three design variables to the problem: the coordinates of the phase center and the orientation . Assembling those in matrix form, column vectors , and are formed respectively. To start the planar array design, we derive the cost function (4) where

and (8) In (7), is the operator transforming a row vector to a diagonal matrix. , and are the column vectors of the , variation of the design variables. in (8) are the respective eld derivatives and given in column vector form. In (6), the eld variation is expressed as a linear combination on the basis dened by . Following the Orthogonal an orthonormal basis by Method [1] we construct from using the GramSchmidt procedure (9) In the new basis, the eld variation is written as follows:

(10) where the new column vector obeys (11) Following (3) and (6), the new eld is written as

(12) Inserting (12) into (4), we have

(5) where is the desired electric eld, is the resulting one, and is the corresponding mean square error. A perturbation of the initial element positions and orientations yields a eld variation, , which reads as follows: (6) with denotes

(13)

(14)

equals the unity matrix as is an orNote that thonormal basis. Equations (10) and (13) form the basis for an iterative gradient/deterministic procedure. The outcome of the iteration is after computing via [(11)]. Vector is obtained by zeroing the rst derivative of the cost function with respect to (15)

(7)

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

151

After some straightforward manipulations, we reach the following formula: (16) Using , we have (17) Thus, after (17), (11), and (7), reads as follows:

(18) must be kept Under the perturbation assumption, relatively small. Therefore, a signicant change in the initial pattern can be obtained only by using an iterative procedure. The value of at the th iteration follows the next formula:

constraints focus on the directivity pattern; and 3) the search is feasibility oriented. It is common knowledge that, in gradient-based methods, one has to use small intermediate steps and start from an unbiased initial solution. Here, it should be made clear that our current work does not address the problem of nding a starting solution (relative studies can be found in the literature [14]). In general, the initial solution should be unbiased toward the nal one (it should not have characteristics that predetermine the optimization outcome). Usually, this can be achieved by using a uniformly spaced initial arrangement. As far as the step size of the perturbation procedure is concerned, it should be noted that it denitely has a serious effect on the speed of the search. Usually, the effort can start from a regular step size and then, as the solution is approached, it should be reduced (see the description of the second phase of the constraint relaxation approach). B. Constraint Relaxation Approach and the Packing Algorithm Following the previous approach, there are cases in which the elements get too close or even overlap. Thus, we face the problem of arranging a set of geometrical shapes (the element radiators) in the hosts available area. Viewed in this perspective, this problem is a packing one [26]. There are various ways in which the radiators size can be incorporated into the formulation. In our case, the problem to be solved is minimize subject to (24)

(19) In (19), we have

(20) From using , the design variables , and are derived (21) The nal placement of the elements is found from the last iteration, in which all the bound constraints are fullled. In order for the method to accept directivity specications in a bound constraint form, the desired electric eld should be described accordingly. Using (1), the amplitude of the produced electric eld should obey (22) is the directivity of the array. The cost function In (22), should be given by (4). Thus, the right part of (4) denes a multivalued function as shown in (23) at the bottom of the page. Thus, if the bound constraints are obeyed, the desired electric eld is set equal to the produced one in order for the cost to be zero. This concludes the presentation of the OPM when: 1) the orientation of the radiators is taken into account; 2) the far-eld

where is the cost function relative to the far eld [given, for example, by (13) or (23)], whereas is the respective cost function for the geometry-layout packing problem. Possible approaches to solve the problem are the Lagrange Multipliers Method and the Hard Barrier Method (Penalty Method) [33]. Moreover, stochastic methods could be employed, although in practice the large number of real variables precludes their use, [22]. In our study, we propose and implement the Constraint Relaxation Approach. Before we present the adopted algorithm, we need to comment on the problem feasibility. In Fig. 2, the search/solution space of the twofold optimization problem is depicted. The area dened by the curve labeled ( ) denes the feasible space of the far-eld (layout packing) problem. In Fig. 2(a), the two areas are separated, which means that although the individual problems can be solved, the problem is infeasible. Fig. 2(b) gives a pictorial representation of the respective feasible problem. For those solution approaches in which the far-eld shaping and the packing problems are simultaneously pursued, the

if otherwise.

(23)

152

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Feasibility of the problem as it is depicted into the solution/search space. (a) Although the individual problems can be solved, the total problem is infeasible. (b) The problem is feasible.

Fig. 3. Typical search path. Curve : Phase one. Curve : Phase two. (a) Start from an overlapped layout. (b) Start from a nonoverlapped layout.

Fig. 4. Flowchart of the rst phase. The Perturbation Method.

problem of the complexity of the algorithm and the respective code arises; so do the demands on computational resources and time. It is the nonlinearity of the problem that renders the simultaneous efforts highly inefcient to pursue. Alternatively, the element overlapping may be allowed to occur for the intermediate solutions until the nal one that should be physically realizable. This is analogous to the tunnel effect in quantum mechanics [34], where the initial and nal states (starting and nal array) are stable (physically realizable), but this is not necessary for the intermediate ones. This is the main concept of the constraint relaxation approach. The constraint relaxation approach consists of two phases: 1) the OPM synthesis phase; 2) the layout packing phase. During the OPM synthesis phase, we attempt to reach a solution of the far-eld synthesis without (or better, by relaxing) the overlapping check. During the layout packing phase, we reduce the step size and include the overlapping check by applying the packing algorithm. In Fig. 3, a pictorial representation of a typical search path is given, in which both optimization phases are shown. In Fig. 3(a), we start from an overlapped layout ( in the initial array: Point ), while in Fig. 3(b), from a nonoverlapped one ( in the initial array: Point ). The rst phase is depicted by the curve, whereas the second phase by the one. Following Fig. 3(a) and (b), in accordance with the constraint relaxation approach, during the rst phase, we drive the solution search into the solution space of the rst problem . Then, during the second phase, while trying not to get away from the space (keep the cost on the far eld close or, even better, equal to zero), we pursue a solution for the second problem.

The owcharts of the rst and second phases are given in Figs. 4 and 5, respectively. In Fig. 4, the Minimize phrase exactly denotes the implementation of the OPM theory. The second phase of optimization consists of two modes: the competitive mode and the settle-down one. Note that the algorithm (as given in Fig. 5) tries to nd a solution while checking . This is regarded as a competition among radiating elements (to nd a free, from overlapping, placement) when . On the other hand, if is easy to maintain, then the constraint can be relaxed (the elements settle down). In the beginning, the competitive mode among the radiating elements is followed. The settle-down mode is usually followed during the last stages of the search when overlapping is small. In the second phase (Fig. 5), the Settle-down mode implements the Minimize and Minimize algorithms sequentially. The Minimize follows. Minimize : 1. Find overlapped couples (OCs). 2. Sort OCs in descending overlapping area. 3. Produce random arrangements ( and ). 4. Compute of the current and the N arrangements. 5. Choose the one with the smallest overlap (minimum The Competitive mode implements the Minimize subject to the algorithm. Minimize subject to : 1. Find overlapped couples (OCs). 2. Sort OCs in descending overlapping area.

).

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

153

trapping, the packing algorithm is equipped with a migration procedure. Following this procedure, when the overlapped area cannot be reduced any further, the most overlapped element is moved into the available space in a position in which overlapping is low. Equivalently, this element migrates to a more suitable place. Thus, the solution also migrates to the basin of another attractor, which, hopefully, may provide both mask compliance and physical realization. IV. PERFORMANCE EVALUATION In this section, various examples are presented in order to demonstrate the merit of the proposed method. First, a detailed example is provided, demonstrating a typical evolution of the search and presenting a performance analysis (Figs. 68). Second, a comparison to previously published cases is given (Figs. 912). New results are listed next, where we focus on the design of symmetric arrays (Figs. 13 and 14). Finally, an example study for the determination of maximally sparse design curves is presented (Figs. 15 and 16). A. Detailed ExamplePerformance Analysis In the rst example, we assume an initial array of 331 elements in a circular ring arrangement. The th ring has a radius of and elements. Each element is a circular disc of radius exhibiting uniform aperture eld. In order to make the results independent of the aperture eld polarization, we assume that each radiator contributes to the arrays directivity pattern through its aperture space factor [5, Ch. 12]. In this detailed example, the weight vector is a constant vector with all the elements equal to one. The eld vector is given as follows:

(25)
Fig. 5. Flowchart of the second phase. Layout Packing Algorithm.

3. Choose the th OC. 4. Choose the rst or second element of the th OC. 5. Produce random arrangements of the chosen radiator ( and ). 6. Compute and of the current and the arrangements. 7. Keep the best 10% of the and . 8. From those: a. with probability 90%, choose the one with the best . b. with probability 10%, choose one at random. Despite all efforts and due to the nonconvexity of the problem, the search is possible to be trapped in an unacceptable solution, namely an array that obeys the directivity mask while overlapping cannot be further eliminated/reduced. To combat

are the Cartesian coordinates in wavelengths of the where th element and is the total number of antenna elements. is the Bessel function of the rst type and of order equal to one. Let us assume that we wish to design an improved array with the following mask parameters: dBi, dBi, . In order to optimize this solution, we enforce the proposed method. The results after 0 [Fig. 6(a) and (b)], 30 [Fig. 6(c) and (d)], 66 [Fig. 6(e) and (f)], and 120 [Fig. 6(g) and (h)] steps are given next. In the directivity schemes, the pattern of the produced array for various phi cuts is given, while the respective mask is denoted by the straight lines. Let us now give a general description/outline of the evolution of the search before we go into further details. Phase One of the optimization procedure spans from the rst up to the 66th iteration step. During this phase of optimization, the grating lobes are eliminated rst [Fig. 6(a) (c)]. In addition, while the level is slowly rising, the difference is also gradually rising. This is

154

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 6. Evolution of the search. The resulting patterns and layouts after (a), (b) 0; (c), (d) 30; (e), (f) 66; and (g), (h) 120 steps.

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

155

accomplished by shrinking the array aperture [Fig. 6(b) (d)]. The rst sidelobe, which was initially between passbands and stopbands [see Fig. 6(a)], starts shifting toward the stopband. As a result, the proposed algorithm lowers the lobe level. Therefore, this lobe is moved outwards and, at the same time, it is lowered in order to go under the upper bound of the mask curve [Fig. 6(c) (e) (g)]. Note also that the intermediate solutions [see Fig. 6(d) and (f)] are not viable. There are element radiators that overlap. An overlapped couple of radiators is denoted by a line segment connecting the phase centers of the two elements. Due to overlapping, after the 67th iteration step, we proceed to Phase Two of the optimization procedure. As it can be seen from Fig. 6(h), besides obeying the directivity mask [Fig. 6(g)], the nal layout is physically realizable. Let us see how the two phases of the algorithm are applied in the above detailed example. According to the constraint relaxation approach, in Phase 1 (iteration steps: 166) we drive the solution search into the solution space of the rst problem , relaxing the constraint on the element overlapping and on the physical realizability. In Phase 2 (iteration steps: 67120), while trying not to get out of the space (keep the cost on the far eld close or, even better, equal to zero), we pursue a solution with zero overlapping that shares a lot of common features with packing problem solving. Let us now study the response of our method in more detail via the evolution monitoring of various performance indexes: a) the far-eld directivity pattern cost function; b) the overlapped area; c) the peak sidelobe level (PSL); d) the arrays directivity pattern at the angle; e) the aperture radius. Evolution of the Far-Field Cost Function: In order to have a better understanding of the evolution of the far-eld cost function and since the bounds on the mask constraints consist of two sets (the bound near the main beam and the bound on the sidelobe area), we dene and study the following indexes: (26) where

(27) is the sum It is clear that the total cost on the far eld of the two cost functions . The normalized indexes, with respect to their maximum value, are shown in Fig. 7(a). During phase one, the algorithm focuses on the far-eld convergence. We can see that the far-eld indexes generally decrease, and by the end of phase one, both reach zero. From the rst up to the 25th step, the method tries to cope with the grating lobe that appears at 9 from broadside. This can be realized through the reduction of the normalized index.

Fig. 7. Evolution of the performance indexes. (a) Far-eld cost function. (b) Overlapped area. (c) Peak sidelobe level, the directivity, and the aperture radius.

Indeed, both the grating lobe and (eventually) all the sidelobes are accounted for in this index since they lie in the angle area.

156

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

At the same time (iteration steps 131), the width of the main lobe increases so that the value gets over the level. This can be clearly seen from the nearly linear decrease of the index, which accounts for the far-eld cost in the main beam area. The sidelobes become lower, as can be seen from the evolution of the index, after step 25 (iteration steps 2666). Note that for this range of iteration steps, is not monotonically decreasing. This is due to the fact that the search is now near the optimum, and a small oscillation around it, just before its attainment, should be expected. The second phase takes place after the 67th iteration step. The far-eld cost function is held nearly equal to zero, while the proposed algorithm shifts its attention to the second big problem, that of providing physical realizability. Since during Phase Two both the and indexes attain negligible values, Fig. 7(a) depicts their evolution only during Phase One. Evolution of the Overlapped Area: The evolution of the overlapped area is depicted step by step in Fig. 7(b) both for the rst and second phase of the optimization procedure. During Phase One, no check is done on the overlapping area. Since the overlapping constraint is relaxed, the overlapped area initially presents a rise reaching about at the 45th iteration step. Then, it oscillates around this value until the end of Phase One. Afterwards, it exhibits an exponential decrease from the 67th iteration step until about the 90th step. Typically, if a feasible problem is studied, the overlapping area exhibits a rise (phase one), which is followed by an exponential reduction (phase two). Less often, spikes representing sudden rises or reductions can occur, especially when rectangular elements are used. On relatively demanding problems, the decisive test about problem feasibility is the following: One should try, starting from a less strict mask to solve intermediate problems, gradually making the bounds stricter until they reach the desired form. Going back to the case study, we can see that after the 90th step, there are about 10 iteration steps where the overlapping area remains constant. As stated above, due to the nonconvexity of the problem, getting trapped in an unacceptable solution is possible, that is, one in which the designed array obeys the directivity mask but overlapping cannot be eliminated/reduced any further. This is exactly the case for those 10 iteration steps in the search. To combat trapping, according to Section III, the packing algorithm activates the migration procedure. The constant value of the overlapped area triggers the migration algorithm that releases/frees the search. After a small rise, the overlapped area decreases again till it becomes zero. To understand the migration algorithm, Fig. 8 depicts the array layouts at the 99th and 100th iteration steps. Initially, as can be seen, element 37 (the most overlapped element) is stuck between elements 19, 20, and 36. The migration algorithm throws it at a new place where it now overlaps with elements 36, 60, and 61. From this new layout, an acceptable solution can eventually be reached. Evolution of the PSL, the , and the Aperture Radius: Now, let us focus on Fig. 7(c) and the evolution of the remaining

Fig. 8. Migration algorithm example. (a) Layout at the 99th and 100th iteration steps. (b) Zoom on the migration leap of element 37.

three indexes. The aperture radius decreases from to from the rst until 50th iteration step. Then, it remains rather constant. The rises until 44 dBi, and then it remains rather constant as well. The PSL rapidly improves from the rst until the 15th step (the grating-lobe elimination phase). Then, it further reduces until step 50, where it acquires its nal value. Let us now highlight some correlation relations between the various indexes. First, indexes (a) and (c)-(d)-(e) remain rather constant during Phase Two. Thus, during this phase (in which the packing algorithm prevails), the process is ne-tuned to produce an acceptable solution. Second, indexes (d) and (e) are highly correlated. The shrinking of the arrays aperture results in the broadening of the main beam, thus increasing the . Finally, let us focus on the relation between the indexes and PSL. For the rst iteration steps, in which both normalized and are reduced, the grating lobe diminishes rapidly, which results in a rapid reduction of the PSL. After about the 15th iteration step, while normalized continues its reduction in a steady pace, the reduction of the normalized starts losing momentum. It is the rst time that the grating lobes become smaller than the rst sidelobe, and the search (while

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

157

continuing the grating lobe reduction) focuses on coping with the rst sidelobe. This reduces the pace at which the PSL reduces until about the 50th iteration step. After that, correction variations are observed. Let us conclude the study by focusing on the way in which the presented owcharts (Figs. 4 and 5) are implemented in the current example in order to easily understand the use of the algorithms. Since an acceptable solution has been found, the feasibility of the problem is depicted in Fig. 2(b). Also, in the current example, we refer to Fig. 3(b) because the initial array is physically realizable as the arrays radiators do not overlap . For steps 166, we follow the curve , while in terms of the owcharts, we follow Fig. 4. The details of the procedure minimize are given in (3)(23). We should note that if Phase one is omitted in the current example, the achievement of an acceptable solution is postponed for thousands of iteration steps. In fact, we attempted to solve the problem with a standard stochastic algorithm code, but unfortunately, it was impossible to nd a solution. This can also be advocated by the comment in [22], in which [focusing on Particle Swarm Optimization (PSO)] it is stated that The number of particles required appears to grow roughly as which indicates PSO cannot consistently return optimum solutions for large arrays. Thus, evolutionary techniques cannot manage a large number of continuous unknowns. That is why those methods (stochastic) should be accompanied by proper strategies like the one proposed here in order to nd acceptable results while using reasonable time and computational resources. For steps 67120, we follow the curve , while in terms of the owcharts, we follow Fig. 5. As stated above, the second phase of optimization consists of two modes: the competitive mode and the settle-down one. Here, the competitive mode is used. Finally, as shown in Fig. 8, during Phase Two the migration algorithm is used. B. Comparison to Previously Published Works In this section, our method is compared to some of the most powerful state-of-the-art approaches currently available: Rahmat-Samii et al.s work on PSO [23], and the Difference Set Genetic Algorithm (DSGA) of Caorsi et al. [27]. Note that, as will be shown, in the cases in which the global optimum solution is currently known, our method can reproduce it inexpensively while setting new records on nonnalized problems. In the rst two cases presented here, the algorithm uses only the rst phase. Since the average interelement distance is big, no overlapping is experienced during Phase One, and thus there is no need for the use of a packing algorithm. 1) Particle Swarm Optimization [23]: In this rst example, we compare our results to the PSO method for aperiodic antenna array designs in [23]. Let us consider a linear nonuniform array of 10 isotropic elements (point sources). The array is assumed to be symmetric about its center. The outermost elements are xed at . Hence, there are four element locations to be optimized by Real PSO (RPSO). The average element spacing is . Moreover, two adjacent elements are not allowed to get closer than . The objective of the optimization is to achieve the lowest peak SLL using uniform excitation.

Fig. 9. Evolution of the array elements placement. Edge elements are held . constant at

In order to model the nearest neighbored constraint, circular elements of radius are assumed. The element pattern of each radiator is the same as that of the point source. The mask that we use is given by the following specications: dBi, , and dB. The elements are initially placed on the -axis and are not allowed to get out of it. By implementing our method, the previously attained results are conrmed. Our method converges rapidly to the same result. Using two decimal digits as in [23], our method veries the published results and gives a 19.7 dB. Furthermore, starting from an equally spaced initial linear array, Phase One of our method is more than adequate to solve the problem. In Fig. 9, the element displacements during the convergence processes are shown. Note that the array is symmetric around although a no-symmetry constraint was enforced. The optimization procedure in [23] is executed using a large number of 20 agents (each agent represents a different array) for 200 iterations to provide better sampling of the solution space. The optimal design is observed at the 80th iteration. Here, the optimization procedure needs one agent and less than 40 iteration steps. 2) Difference Set Genetic Algorithm (Linear Array) [27]: In this second example, we work on the problem of the thinned array produced by the Cyclic Difference Set (CDS) [29] and DSGA method [27]. It is enforced on the linear implementation of the (63, 32, 16) difference set on a linear aperture. The number of the array elements (point sources) is 32. The PSL achieved is 15.39 dB. Our method improves this result by more than 2.5 dB ( 18 dB). We use the same elements as those in the previous example. The mask that we use is given by the following specications: dBi, , and dB. Our method, starting from the DSGA solution, produces this result within less than 100 iteration steps. The power pattern and the nal positions of the elements are given in Fig. 10, where

158

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 10. Power patterns and placement of the 32-element linear array. For the layouts: 0Our results; 1DSGA; 2SGA; 3CDS.

TABLE I PERFORMANCE TABLE FOR THE 32-ELEMENT LINEAR ARRAY

the respective CDS, Standard Genetic Algorithm (SGA), and DSGA results are presented for comparison. Table I presents the results from: 1) [27], in which CDS, SGA, and DSGA are discussed; and 2) the current work. 3) Difference Set Genetic Algorithm (Planar Array) [27]: In this third example, we work on the problem of the thinned array produced after the DSGA method is enforced on the planar implementation of the (63, 32, 16) difference set on a planar aperture. The number of the array elements (point sources) is 32. The record is a 14.26 dB. In implementing our method, it is assumed that the elements are point sources having a square area of . We start from a different initial solution, given in Fig. 11. This solution constrains the positions of the 32 elements in a Cartesian grid with interelement spacing of exactly as the thinning process does, but at the same time, it performs better than the CDS, SGA, and DSGA state-of-the-art methods. To produce the best currently known thinned solution for the (63, 32, 16) difference set on a planar aperture, we work as follows. It is well known from the literature that a uniform circular aperture gives a PSL of about 17 dB. Thus, sampling with our point source elements, we estimate that we could go near that index if we arrange them in such a way as to approximate a uniform circular aperture. Indeed, after placing the rst 29 elements around the origin, three elements are left and can be placed in eight positions, indicated with the cycles in Fig. 11(a), in order for the whole array to approximate a uniform circular aperture. For

Fig. 11. The (initial for the proposed algorithm) best currently known thinned solution. (a) Layout. (b) Directivity pattern for various phi-cuts.

those last three elements, their nal positions have been chosen randomly and are shown in Fig. 11(a). The record in this rst solution is 15.42 dB, already better by more than one decibel compared to the result in [27]. Continuing with the example and starting from the layout shown in Fig. 11(a), our method produces the solution shown in Fig. 12. Note that, in the current example, no rotation of the element radiators is allowed. The achieved 17 dB. The mask applied is given by the following specications: dBi, , and dB. Table II presents the results from: 1) [27], in which CDS, SGA, and DSGA are discussed; and 2) Figs. 11 and 12. Note that the increased performances, shown in Fig. 12, from the best currently known thinned solution for the (63, 32, 16) difference set on a planar aperture (Fig. 11), are due to the aperiodic nature of the nal arrangements, which also means a tradeoff in terms of realization complexity, as a regular lattice facilitates the array construction (see [35, p. 54]).

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

159

Fig. 12. Optimized 32element planar array. (a) Layout. (b) Directivity pattern for various phi-cuts. TABLE II PERFORMANCE TABLE FOR THE 32-ELEMENT PLANAR ARRAY

Fig. 13. Initial planar array with two different square element types (symmetric case). (a) Layout. (b) Directivity pattern for various phi-cuts.

C. Design of Symmetric Aperiodic Planar Arrays In the next example [Fig. 13(a)], an initial array is assumed, which employs two kinds of ideal square elements: 1) 64 with edge length and excitation 1; and 2) 64 with edge length and excitation 2. The square radiators exhibit uniform aperture eld. The produced directivity pattern of the array is given in Fig. 13(b). Now, let us assume that we need to design an improved array that obeys a mask with the following parameters:

dBi, dBi, and . Our approach is applied by working only on one of the quarters of the array. The elements of the other three quarters are placed and directed in mirror positions and orientations to the ones of the perturbed quarter. The nal array layout and respective directivity pattern are given in Fig. 14. Note that array symmetry reduces the DoF of the optimization problem, which means that the feasibility space becomes smaller. Thus, in general, asymmetric arrays perform better in the sense that they can achieve lower sidelobe levels (e.g., comply with masks of smaller ). On the other hand, due to the fewer DoF, array symmetry makes the algorithm more efcient as the problem becomes simpler because of the smaller number of elements (provided of course that the respective problem is feasible). Thus, when we deal with nondemanding problems, array symmetry is an interesting option; although it is not obligatory, it can be included in the set of constraints to facilitate the design of the underlying feeding network.

160

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 15. Maximally sparse design curve of the problem.

Fig. 14. Optimized symmetric planar array with two different square element types. (a) Layout. (b) Directivity pattern for various phi-cuts.

Finally, it should be noted that, as proved in [36], the mutual coupling for electrically large aperture radiators can be neglected even in the case of dense packing. D. Determination of Maximally Sparse Design Curves Let us now proceed to the last example. One of the key problems in aperiodic array design is, given the available types of radiators, to determine the minimum number of elements needed for the array to comply with the design prescriptions. This is the so-called maximally sparse aperiodic array problem [32], [37], [38]. Note that in [37] and [38], the minimum number of sensors is addressed using both placement coordinates and excitation coefcients in the design of linear arrays of point sources. While we have studied relative cases in the past [39], here we address a more complex problem. The radiators size is taken into account, the array is two-dimen-

sional, and the excitation is held constant in order to facilitate the array construction. We present a study using two types of rectangular elements: Type I elements, edge length normalized excitation ; Type II elements, edge length normalized excitation . The mask specications are dBi, dBi, and . We conduct the study as follows. We start from a hexagonal grid where the interring distance in the -axis is set equal to . We vary the number of elements of both types in the range : 0150 and : 040. The innermost elements of the hexagonal grid are assigned to the rst type of elements, while the rest are assigned to the second type. The initial orientation of the elements is set equal to zero, meaning that initially their edges are parallel to the - or -axis. For each combination of , and , the proposed optimization method is enforced for the composed array. The outcome of the study is given in Fig. 15, in which we can see the maximally sparse design curve of the problem. Coming from the right, the closer an array is to the curve, the more effort is needed to achieve mask compliance and physical realization. Indeed, by getting closer to the curve of Fig. 15, the convergence time rises rather exponentially. To nd a solution (assuming the problem is feasible), without consuming an unacceptable amount of time, the initial mask could be rather loose in comparison to the desired one. Then, while solving the intermediate problems, it becomes gradually stricter in order to guide the search to the nal solution. To date, no array at the left of the curve proved to be a feasible solution. Thus, the curve divides the search space between the feasible and the nonfeasible problems. In Fig. 16, the directivity pattern and layout of the single-type arrays with minimum number of elements of the rst and the second type are given. V. CONCLUSION Taking into account the size and orientation of the element radiator, the Constraint Relaxation Approach is proposed to optimize the array layout. The use of randomness in the element

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

161

most powerful ones for the solution of highly complex optimization problems [35], [40]. The method is robust enough to produce an acceptable solution that can be further improved by using possibly fewer elements and/or symmetric arrays. Indeed, usually it is important to use as few elements as possible, searching for the maximally sparse solution. Thus, after a solution is found, we can reduce the number of elements and restart checking for a less complex layout. Such layouts are also produced when symmetric arrays are pursued. Indeed, when an array layout exhibits mirror or rotational symmetries, the method leads to a simplied solution in comparison to the nonsymmetric layouts. Our study is applied to various initial geometries in order to present the relevant features of our approach. The resulting arrays are shown to comply with the corresponding desired patterns and nonoverlapping constraints. Those cases demonstrate the applicability and merit of the method. REFERENCES
[1] J. N. Sahalos, Orthogonal Methods for Array Synthesis: Theory & the ORAMA Computer Tool. West Sussex, U.K.: Wiley, 2006. [2] R. J. Mailloux, Phased Array Antenna Handbook. Boston, MA: Artech House, 2005. [3] T. Milligan, Modern Antenna Design, 2nd ed. New York: IEEE PressWiley, 2005. [4] R. E. Collin and F. J. Zucker, Antenna Theory. San Francisco, CA: McGraw-Hill, 1969. [5] C. A. Balanis, Antenna Theory: Analysis and Design. New York: Wiley, 2005. [6] R. C. Hansen, Phased Array Antennas. New York: Wiley-Interscience, 1998. [7] Y. T. Lo, A mathematical theory of antenna arrays with randomly spaced elements, IEEE Trans. Antennas Propag., vol. AP-12, no. 3, pp. 257268, May 1964. [8] T. G. Spence and D. H. Werner, Design of broadband planar arrays based on the optimization of aperiodic tilings, IEEE Trans. Antennas Propag., vol. 56, no. 1, pp. 7686, Jan. 2008. [9] J. N. Sahalos, Orthogonal methods of nonuniformly spaced arrays, Proc. IEEE, vol. 62, no. 2, pp. 281282, Feb. 1974. [10] J. N. Sahalos, A solution of general nonuniformly spaced antenna arrays, Proc. IEEE, vol. 62, no. 9, pp. 12921294, Sep. 1974. [11] J. N. Sahalos, T. N. Kaifas, and G. S. Miaris, The orthogonal method for the geometry synthesis of a 2-D antenna array, in Proc. 29th ESA Workshop Multiple Beams Recong. Antennas, Apr. 1820, 2007. [12] T. N. Kaifas and J. N. Sahalos, On the geometry synthesis of uniformly excited conformal arrays by the orthogonal method, in Proc. 2nd EuCAP, Edinburgh, U.K., Nov. 1116, 2007. [13] D. G. Babas and J. N. Sahalos, On the design of Chebyshev endre arrays subject to a performance index, Elect. Eng., vol. 90, no. 2, pp. 155160, Dec. 2007. [14] T. A. Milligan, Space tapered circular (ring) array, IEEE Antennas Propag. Mag., vol. 46, no. 3, pp. 7073, Jun. 2004. [15] Y. T. Lo and S. W. Lee, A study of space-tapered arrays, IEEE Trans. Antennas Propag., vol. AP-14, no. 1, pp. 2230, Jan. 1966. [16] A. Ishimaru, Theory of unequally-spaced arrays, IRE Trans. Antennas Propag., vol. 10, no. 6, pp. 691702, Nov. 1962. [17] R. E. Willey, Space tapering of linear and planar arrays, IEEE Trans. Antennas Propag., vol. AP-10, no. 4, pp. 369377, Jul. 1962. [18] F. B. T. Marchaud, G. D. de Villiers, and E. R. Pike, Element positioning for linear arrays using generalized Gaussian quadrature, IEEE Trans. Antennas Propag., vol. 51, no. 6, pp. 13571363, Jun. 2003. [19] R. L. Haupt, Optimized element spacing for low sidelobe concentric ring arrays, IEEE Trans. Antennas Propag., vol. 56, no. 1, pp. 266268, Jan. 2008. [20] M. I. Skolnik, , R. E. Collin and F. J. Zucker, Eds., Nonuniform Arrays, in Antenna Theory. New York: McGraw-Hill, 1969, ch. 6, pt. 2. [21] M. I. Skolnik, J. W. Sherman, and F. C. Ogg, Statistically designed density-tapered arrays, IEEE Trans. Antennas Propag., vol. AP-12, no. 4, pp. 408417, Jul. 1964.

Fig. 16. Directivity patterns and layouts of (a), (b) the and (c), (d) the maximally sparse arrays.

arrangement, during the packing algorithm phase, transforms the OPM into an evolutionary-gradient hybrid technique. Hybrid techniques of this kind are regarded in the literature as the

162

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[22] P. J. Bevelacqua and C. A. Balanis, Minimum sidelobe levels for linear arrays, IEEE Trans. Antennas Propag., vol. 55, no. 12, pp. 33423349, Dec. 2007. [23] N. Jin and Y. Rahmat-Samii, Advances in particle swarm optimization for antenna designs: Real-number, binary, single-objective and multiobjective implementations, IEEE Trans. Antennas Propag., vol. 55, no. 3, pp. 556567, Mar. 2007. [24] T. N. Kaifas, D. G. Babas, and J. N. Sahalos, Design of planar arrays with reduced nonuniform excitation subject to constraints on the resulting pattern and the directivity, IEEE Trans. Antennas Propag., vol. 57, no. 8, pp. 22702278, Aug. 2009. [25] O. M. Bucci, T. Isernia, A. F. Morabito, S. Perna, and D. Pinchera, Aperiodic arrays for space applications: An effective strategy for the overall design, in Proc. 3rd EuCAP, Berlin, Germany, Mar. 2327, 2009, pp. 20312035. [26] H. T. Croft, K. J. Falconer, and R. K. Guy, Unsolved Problems in Geometry. New York: Springer-Verlag, 1991. [27] S. Caorsi, A. Lommi, A. Massa, and M. Pastorino, Peak sidelobe level reduction with a hybrid approach based on GAs and difference sets, IEEE Trans. Antennas Propag., vol. 52, no. 4, pp. 11161121, Apr. 2004. [28] C. I. Coman, I. E. Lager, and L. P. Ligthart, The design of shared aperture antennas consisting of differently sized elements, IEEE Trans. Antennas Propag., vol. 54, no. 2, pp. 376383, Feb. 2006. [29] D. G. Leeper, Isophoric arraysMassively thinned phased arrays with well-controlled sidelobes, IEEE Trans. Antennas Propag., vol. 47, no. 12, pp. 18251835, Dec. 1999. [30] R. J. Mailloux, S. G. Santarelli, T. M. Roberts, and D. Luu, Irregular polyomino-shaped subarrays for space-based active arrays, Int. J. Antennas Propag., vol. 2009, p. 956524, 2009. [31] V. Pierro, V. Galdi, G. Castaldi, I. M. Pinto, and L. B. Felsen, Radiation properties of planar antenna arrays based on certain categories of aperiodic tilings, IEEE Trans. Antennas Propag., vol. 53, no. 2, pp. 635644, Feb. 2005. [32] R. M. Leathy and B. D. Jeffs, On the design of maximally sparse beamforming arrays, IEEE Trans. Antennas Propag., vol. 39, no. 8, pp. 11781187, Aug. 1991. [33] J. Nocedal and S. Wright, Numerical optimization, in Science Business Media LLC, ser. Operations Research and Financial Engineering, 2nd ed. New York: Springer, 2006. [34] D. J. Grifths, Introduction to Quantum Mechanics, 2nd ed. New York: Prentice-Hall, 2004. [35] R. L. Haupt and D. H. Werner, Genetic Algorithms in Electromagnetics. Hoboken, NJ: WileyIEEE Press, 2007. [36] T. S. Bird, Analysis of mutual coupling in nite arrays of differentsized rectangular waveguides, IEEE Trans. Antennas Propag., vol. 38, no. 2, pp. 166172, Feb. 1990. [37] Y. Liu, Q. H. Liu, and Z. Nie, Reducing the number of elements in the synthesis of shaped-beam patterns by the forward-backward matrix pencil method, IEEE Trans. Antennas Propag., vol. 58, no. 2, pp. 604608, Feb. 2010. [38] G. Oliveri and A. Massa, Bayesian compressive sampling for pattern synthesis with maximally sparse non-uniform linear arrays, IEEE Trans. Antennas Propag., vol. 59, no. 2, pp. 467481, Feb. 2011. [39] T. N. Kaifas and J. N. Sahalos, An iterative technique for the synthesis of active antenna oscillator arrays, Int. J. Antennas Propag., Mar. 2009. [40] R. Salomon, Evolutionary algorithms and gradient search: Similarities and differences, IEEE Trans. Evol. Comput., vol. 2, no. 2, pp. 4555, Jul. 1998. Theodoros N. Kaifas (M08SM11) received the Ph.D. degree in physics from the Aristotle University of Thessaloniki (AUTH), Thessaloniki, Greece, in 2004. From 1992 to 2004, he was a Research Assistant, and since 2004, has been a Postdoctoral Research Fellow with the Radio Communications Laboratory (RCL), Department of Physics, AUTH. During 2001 to 2008, he was a Scientic Associate with the Department of Electronics, Technological and Educational Institute of Thessaloniki, Thessaloniki, Greece, and during 2006 to 2007, with the Department of Technology Management, University of Macedonia, Thessaloniki, Greece. He has a broad knowledge on and experience with information technology and communications and an expertise in wireless communications, RF/microwave circuits,

antenna analysis and design, and computational methods in electromagnetics accompanied by extensive contribution in relative literature and funded research projects. Dr. Kaifas is a member of the Hellenic Physicist Union and the Hellenic Club of Electronic Physicists.

Dimitrios G. Babas (M93) received the B.Sc. degree in physics, M.Sc. degree in electronic physics (radioelectrology), and Ph.D. degree in physics from the Aristotle University of Thessaloniki (AUTH), Thessaloniki, Greece, in 1991, 1995, and 2001, respectively. He also received the Diploma of Postgraduate Studies in telecommunication and information systems from the University of Essex, Colchester, U.K., in 1992. Since 1992, he has been a Research Assistant with the Radio Communications Laboratory (RCL), Department of Physics, AUTH. He is also a member of the teaching staff of M.Sc. studies in electronic physics, AUTH. During 1999 to 2008, he was a Scientic Associate with the Department of Electronics, Technological and Educational Institute of Thessaloniki, Thessaloniki, Greece. His research interests include waveguide systems, antenna systems design, wireless communications, and RF measurements. Dr. Babas is a member of the Hellenic Physicist Union and Hellenic Club of Electronic Physicists.

George S. Miaris was born in Neubeckum, Germany, in 1968. He received the B.Sc. degree in physics and M.Sc. degree in electronic physics (radioelectrology) from the Aristotle University of Thessaloniki (AUTH), Thessaloniki, Greece, in 1993 and 1996, respectively. During 1995 to 1996, he was a Research Assistant with the Radio Communications Laboratory, Department of Physics, AUTH, a position he returned to after completing his military service in 1998. He is currently a Member of the Technical Staff with the same department. His research interests include antenna array design, cellular mobile systems, RF measurements, measurement automation, wave propagation, and computational electromagnetics.

Katherine Siakavara (M04) received the B.Sc. degree in physics, M.Sc. degree in electronics, and Ph.D. degree in electrical engineering from the Aristotle University of Thessaloniki (AUTH), Thessaloniki, Greece, in 1977, 1979, and 1982, respectively. She is currently an Associate Professor with the Department of Physics, AUTH. She is the author or coauthor of 36 international journal papers, 30 papers in proceedings of international conferences, two book chapters in English, and two book chapters in Greek. Her research interests are in the areas of applied electromagnetism, analysis and design of antenna systems, microwaves, and radio-wave propagation and radio communications.

Elias E. Vaadis (M86) received the B.Sc. degree in physics and M.Sc. degree in electronics from the Aristotle University of Thessaloniki (AUTH), Thessaloniki, Greece, in 1975 and 1979, respectively, and the Ph.D. degree in electrical engineering from the Democritus University of Thrace, Xanthi, Greece, in 1985. He is an Associate Professor with the Department of Physics, AUTH. From 1986 to 1993, he served as a Lecturer and an Assistant Professor with Microwaves Laboratory, Xanthi, Greece. Since 1993, he has been a member of the Radiocommunications Laboratory, Physics Department, AUTH. He has published more than 70 journal articles and conference proceedings. His research interests include the electromagnetic theory of wave guiding and radiating structures and CAD techniques of microwave circuits. Dr. Vaadis is a member of the Hellenic Physical Society.

KAIFAS et al.: APERIODIC ARRAY LAYOUT OPTIMIZATION BY THE CONSTRAINT RELAXATION APPROACH

163

John N. Sahalos (M75SM84F06LF10) received the B.Sc. and Ph.D. degrees in physics, the Diploma (BCE MCE) in civil engineering, and the Diploma of Postgraduate Studies in radio-electrology from the Aristotle University of Thessaloniki (AUTH), Thessaloniki, Greece, in 1967, 1974, 1975, and 1975, respectively. From 1971 to 1974, he was a Teaching Assistant, and from 1974 to 1976, he was an Instructor with the Department of Physics, AUTH. In 1976, he worked at the ElectroScience Laboratory, The Ohio State University, Columbus, as a Postdoctoral University Fellow. From 1977 to 1986, he was a Professor with the Electrical Engineering Department, University of Thrace, Xanthi, Greece, and Director of the Microwaves Laboratory. From 1986 to 2010, he was a Professor with the School of Science, AUTH, where he was the Director of the postgraduate studies in electronic physics and the Director of the Radio-Communications Laboratory (RCL). He is now a Professor with the Electrical and Computer Engineering Department, University of Nicosia, Nicosia, Cyprus, and Director of the Radio and Telecommunications Laboratory (RTeLab). During 1981 to 1982, he was a Visiting Professor with the Department of Electrical and Computer Engineering, University of Colorado, Boulder. During 1989 to 90, he was a Visiting Professor with the Technical University of Madrid, Madrid, Spain. He is the author of three books in Greek, seven book chapters, and more than 350 articles published in the scientic literature. He also is the author of the book The Orthogonal Methods of Array Synthesis, Theory and the ORAMA Computer Too (Wiley, 2006). He has supervised 25 Ph.D.s and

more than 100 postgraduate diploma theses. With his colleagues, he designed several well-known innovative products like the Electric Impedance Tomography (EIT), the Microwave Landing System (MLS), the ORAMA simulator, and the SMS-K monitoring system. He is the creator and leader of an EMC network with ve laboratories (three from the academy and two from the industry). His research interests are in the areas of antennas, high-frequency techniques, communications, EMC/EMI, microwaves, and biomedical engineering. Dr. Sahalos is a Professional Engineer and a consultant to industry. He also is a member of the Greek Physical Society, an Honor Fellow of Radio-electrology, and a member of the Technical Chamber of Greece. He was elected by the department representatives of AUTH as the Vice-Chairman of the Research Committee of AUTH for the period 20072010. Since 2010, he has been a member of the University of Nicosia Research Foundation (UNRF) and a member of the consulting committee of the GRNET S.A. In 2002 to 2004, he was in the Board of Directors of the OTE, the largest telecommunications company in southeast Europe. He served as a Technical Advisor in several national and international committees, as well as in several mobile communications companies. Since 1992, he has been a member of Commissions A and E of URSI. Since 1998, he has been the President of the Greek committees of URSI. He was the President of the section of Informatics, Telecommunications and Systems of the National Committee of Research and Technology. He has been honored with a special investigation fellowship of the Ministry of Education and Science, Spain. He also has been honored from several institutes and organizations. He has been in the Editorial Board of three scientic journals.

164

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Time Reversal Based Broadband Synthesis Method for Arbitrarily Structured Beam-Steering Arrays
Deshuang Zhao, Yuanwei Jin, Senior Member, IEEE, Bing-Zhong Wang, Member, IEEE, and Rui Zang
AbstractWe propose a novel broadband beam pattern synthesis method to synthesize arbitrarily structured beam-steering arrays by utilizing time reversal technique. Unlike conventional frequency-domain array synthesis methods, this method achieves array excitation by taking the Fourier Transform of the time reversed version of the transient signals received by antenna elements. Using this method, designers can determine the array excitation over a wide frequency range at once by a single run of time reversal experiment, instead of performing time-consuming multi-objective optimizations of beam pattern cost-functions at each frequency. Furthermore, it does not require explicit measurements of element mutual coupling and platform scattering effects since the time reversal signals have implicitly taken into account those effects. The proposed method is theoretically analyzed with the antenna reciprocity and numerically validated with three dipole arrays of different congurations. The results show that this new method is capable of realizing accurate beam steering over a wide frequency band if the desired steering angle is located in the feasible angular scope of beam scan. The beam patterns with accurate beam steering are successfully achieved over a bandwidth of more than 1.5 GHz for three different dipole arrays, i.e., a typical linear dipole array, a grounded circular dipole array, and a dome-shaped dipole array. Index TermsArray synthesis, beam-steering arrays, broadband synthesis, time reversal.

I. INTRODUCTION EAM-STEERING arrays have received increasing attention in recent years for applications in modern wireless communications and radar systems due to their exible beam scan ability. One of the most important approaches to beam steering realization is to control the magnitudes and phases of complex array excitation. For regular arrays, e.g., ideal linear
Manuscript received June 29, 2010; revised May 12, 2011; accepted June 21, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the Hi-Tech Research and Development Program of China under Grant 2008AA01Z206, the National Natural Science Foundation of China under Grant 61071031, and a scholarship from the China Scholarship Council. The work of Y. Jin was supported in part by the Department of Energy under award no. DE-NT-0004654, the 2010 U.S. Air Force Summer Faculty Fellowship award, and in part by the Department of Defense through the Army Research Ofce under award no. W911NF-11-1-0160. D. Zhao was with the Department of Engineering and Aviation Sciences, University of Maryland Eastern Shore, Princess Anne, MD 21853 USA. He is now with the Institute of Applied Physics, University of Electronic Science and Technology of China, Chengdu 610054, China. B.-Z. Wang and R. Zang are with the Institute of Applied Physics, University of Electronic Science and Technology of China, Chengdu 610054, China. Y. Jin is with the Department of Engineering and Aviation Sciences, University of Maryland Eastern Shore, Princess Anne, MD 21853 USA (e-mail: dszhao@uestc.edu.cn; yjin@umes.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167904

or planar arrays, the array excitation can be exactly determined by classic synthesis methods such as Taylor method, DolphChebyshev method, and Fourier synthesis [1]. However, for irregular antenna arrays, these classic synthesis methods are unsuccessful. This is primarily due to the difculties in exactly characterizing the effects of element mutual coupling and irregular-shape scattering. In order to synthesize irregular or conformal arrays, various new synthesis methods have been developed in recent years, such as projection approach [2], adaptive processing [3], least-square method [4], generic algorithm [5], [6], and particle swam optimization [7], [8]. However, most of these synthesis methods are developed in the frequency domain. For a broadband beam-steering array, these methods commonly require performing time-consuming numerical optimizations of beam pattern cost-functions at each frequency for each desired steering angle. Meanwhile, designers must know beforehand the frequency-dependent effects of element mutual coupling and platform scattering [9][11]. It is well-known that those effects are difcult to be characterized exactly because they depend on not only the operating frequency but also the element placement and the irregular structure shape of the array. One possible solution is to use full-wave numerical simulations or real-world experiments to characterize them. But the numerical simulations or real-world experiments must be carried out independently at each frequency for each antenna element. If an array has a broad bandwidth and a large number of antenna elements, the beam pattern synthesis will become rather costly. Time reversal is an adaptive transmission technique, rst proposed by Fink for acoustic detection applications [12]. Due to its unique focusing feature, time reversal has received increasing attention for wireless communications [13][15] and radar applications [16], [17]. Time reversal is typically implemented as follows: a channel sounding pulse is rst emitted from a source. Then, a time reversal mirror (TRM) is used to record, time reverse, and retransmit the received signals. If the channel reciprocity holds, the retransmitted signals automatically retrace their previously incoming propagation paths and adaptively focus the transmitted energy on the initial source antenna. By time reversal, the spatial dispersion and frequency-selective channel fading can be adaptively compensated over a wide frequency range. Research has shown that the time reversal based systems exhibit an improved performance in a highly dispersive or rich scattering environment [14][18]. Besides communication and radar applications, time reversal is also attractive to antenna design such as beam nulling [19] and beam forming [20]. In this paper, a novel broadband method is developed for beam-steering array synthesis. This method is based on the time

0018-926X/$26.00 2011 IEEE

ZHAO et al.: TIME REVERSAL BASED BROADBAND SYNTHESIS METHOD FOR ARBITRARILY STRUCTURED BEAM-STEERING ARRAYS

165

reversal technique. In this method, the array excitation is determined by taking the Fourier Transform of the time reversal signals, rather than optimizing of beam pattern cost-functions. The time reversal signals are obtained by temporally reversing the signals received by antenna elements that respond to the illumination of a pulsed uniform plane wave incident from the desired steering direction. Compared to the conventional frequency-domain array synthesis methods, the proposed method has the following advantages: 1) It does not require measuring element mutual coupling and platform scattering explicitly. As a result, the cost for array synthesis can be reduced substantially. 2) All element excitation can be determined over a wide frequency band at once by a single run of time reversal experiment, instead of performing time-consuming numerical optimizations of beam pattern cost-functions at each frequency. This advantage becomes signicant when one needs to synthesize a broadband irregularly shaped array of a large number of elements. II. PROBLEM STATEMENT AND METHOD DESCRIPTION A. Problem Statement The array synthesis problem in this paper can be mathematically stated as follows: Given a desired steering angle , the synthesis method should be able to determine a set of frequency-dependent excitation vectors (1) in a frequency range such that the beam patterns generated by them satisfy the following objectives
Fig. 1. The procedure of the time reversal based array synthesis method to , over a wide frequency band. determine array excitation

(2)

elements, where denotes the recorded signal of the -th antenna element. Step 3) Reverse each recorded signal in the order of rst-in-last-out to obtain the time reversal data . Step 4) Compute the frequency-domain excitation , for each antenna element by taking the Fourier Transform of the time reversal data, i.e., (3) After the four steps, the array excitation at any interested frequency belonging to the operating bandwidth of the array can be obtained for . If another desired steering angle is required, we repeat the above procedure by updating the incident direction of the plane wave. Note that before is fed into the corresponding elements, it is necessary to scale it by a power factor to ensure sufcient feed power for radiation. For simplication, this factor is omitted throughout the paper because it does not affect the beam pattern. Fig. 1 illustrates the procedure of the time reversal based array synthesis method. In the later sections, we will show that the array excitation obtained by (3) can produce steerable beam patterns with accurate beam pointing over a wide frequency band. III. THEORETICAL ANALYSIS In this section, we perform a theoretical analysis of the proposed array synthesis method. We rst derive the far-eld radiation pattern of the array that is excited by the time reversal excitation. Next we discuss the array gain. Finally, we analyze the beam steering direction of the synthesized beam pattern.

, denotes the excitation of the where -th element, is the total number of antenna elements, and the superscript is the vector transpose. is the beam steering angle of the realized beam pattern , and is the radiated electric eld generated by . Note that is an unknown excitation vector that needs to be determined for minimizing the beam pointing errors of and , where is the elevation angle, is the azimuth angle, and is the tolerance of the beam pointing error in angle. B. Method Description Here the time reversal based array synthesis method developed for the above problem is described. Given a desired steering angle , the proposed method determines the array excitation as follows: Step 1) Probe the studied array using a pulsed uniform plane wave with a temporal waveform of incident from the desired steering angle . The choice of the temporal waveform is somewhat arbitrary as long as it covers the operating frequency bandwidth of the array. Step 2) Record simultaneously all the transient signals , received by the antenna

166

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

A. Realized Radiation Pattern In order to perform an analysis of the proposed array synthesis method, the Plane-wave-impulse-response Element Pattern (PeP) is adopted to characterize an array. Here PeP is dened as a time-domain pattern of an antenna element that is deduced from the plane wave impulse responses in its array environment at all viewing angles. The plane wave impulse response commonly refers to the voltage or current output observed at the antenna load that responds to the illumination of an impulsive plane wave impinging from a specic viewing angle [21][23]. Notice that the plane wave impulse response usually has different denitions, typically depending on the assumed load. In [22], Baum introduced three types of plane wave impulse responses with open circuit, short circuit, and the load matched to the feed-line impedance, respectively. All of them are valid methods for describing the transient behaviors of an antenna if the antenna is considered to be a linear time invariant (LTI) system. In this paper, we dene the plane wave impulse response as the voltage output of an antenna terminated with a 50 load. This is a particularly useful denition in engineering applications because antennas are usually designed with a standard 50 terminated load. According to the time-domain antenna reciprocity [21], [23], the transmitting and receiving properties of an array are identical. Therefore, the PeP, which is dened in the receiving mode, can also be used to characterize a transmitting array. Using the PeP, an array either in the receiving mode or in the transmitting mode can be characterized by the vector

Taking the Fourier Transform of (6), we obtain (7) where the superscript denotes the complex conjugation op, and are the Fourier Transerator. , and , respectively. By comform of paring (7) with (3), we obtain (8) After the array excitation is obtained, the next step is to excite the array. In the time domain, regardless of the spatial propagation loss and actual time delay, using the radiated electric eld of an excited array can be mathematically expressed as (9) where is the transient element excitation. Here, we assume is ideally radiated by the array. By that the total energy of , (9) can be dening the vector rewritten in vector form as (10) Taking the Fourier Transform of (10) yields (11) where of Replacing arrive at , and in (11) with are the Fourier Transform , respectively. dened in (7), we nally

(4) represents the PeP of the -th antenna elewhere . In order to distinment located at guish the transmitting mode (tx) from the receiving mode (rx), in this paper two subscripts tx and rx are introduced to the . We denote and as the vector plane wave impulse response of an array operating in the transmitting mode and in the receiving mode, respectively. Notice is measured in the receiving mode, that in this paper not in the transmitting mode. For the same array it is identical , i.e., . Simultaneously, to the entries of the two vectors are correspondingly denoted as and , respectively. In our proposed method, we utilize the time reversed version of the received signals to determine array excitation. The reare the outputs ceived signals of the antenna elements in response to a pulsed uniform plane impinging from the desired angle . Using the wave LTI system theory and , the signals can be mathematically expressed as (5) where denotes the linear convolution operator. By time reversing (5), the time reversed version of the received signals, , can be obtained (6)

(12) Notice that in the above formula is the realized far-eld electric eld of the array excited by the array excitaor . Equation (12) shows that the eld pattern tion at is primarily determined by the inner product of and . This is because or is distinguished from other excitation. In or , the , intrinsic radiation information of the array, i.e., is implicitly included. By time reversal excitation, we nd that the array has some signicant radiation features. B. Array Gain The rst feature is that the array gain at the desired steering , yields the maximum magnitude over any angle, other array excitation , i.e., (13) In the above equation, is dened as [24] (14) The symbol represents the impedance of free space. is the total excitation power. Assuming that the

ZHAO et al.: TIME REVERSAL BASED BROADBAND SYNTHESIS METHOD FOR ARBITRARILY STRUCTURED BEAM-STEERING ARRAYS

167

with respect to angle, a new gain pattern can be formed. This gain pattern is eagerly desired to array designers since it can provide a direct guideline to determine the maximum gain possibly achievable at every viewing angle, the possible scan blindness, and the feasible scan coverage. C. Resulting Beam Steering Direction Another feature of the array excited by is that accurate beam pointing can be achieved over a wide frequency range. For simplicity, we rewrite (12) in its normalized form (20)
Fig. 2. The gain patterns of an array excited by (solid line) and (dashed line), respectively. The time reversal excitation yields , which is the maximum gain of the array achievable at the . angle of

by dividing the right side of (12) with where is dened as

(21) . Note that the normalized beam pattern has the same beam shape as the realized beam pattern . Therefore, the resulting beam steering direction of can be determined equivalently by searching for the maximum amplitude of , i.e., Here (22) is equal to . To Next, we discuss whether answer this question, we examine the following two cases: Case I: The norm of at any other angle is smaller than or equal to the one at , or equivalently, (23) Fig. 3(a) depicts such a case. Substituting (23) into (20), we have (24) By (21), (25) can be further written as

total excitation energy is ideally radiated, substituting (11) into (14) yields

(15) where is the Hermitian operator. Let and , (15) can be written as (16) We immediately recognize that the maximization problem of (16) over excitation vector can be solved by seeking the eigenvector with the maximum eigenvalue of the matrix [25]. The solution to this maximization problem is (17) together with the optimal array excitation (18) where is an arbitrary non-zero complex factor and represents the norm. Fig. 2 illustrates this optimal solution. Reminding and the antenna reciprocity , we nd (19) Comparing (19) with (7), and replacing with , it shows that (13) is valid. Notice that this conclusion holds at all frequencies since the reciprocal relation is independent of frequency. Notice that is the maximum gain achievable at the angle of , not the maximum gain of the array over the entire angular space. Generally, the value of varies with the viewing angle. If we plot them

(25) is the maximum The above formula shows that magnitude of the normalized pattern with the main-lobe direction equal to , i.e., (26) The result indicates that if the condition in (23) is satised, can steer its main lobe the realized beam pattern precisely to the desired steering angle for all frequencies. It also means that in this case the time reversal based synthesis method can synthesize an array to achieve precise beam pointing over a wide frequency band. It is well-known that time reversal is equivalent to phase conjugation in the frequency domain [12]. Equation (7) shows that the excitation or fed into the array is just the

168

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 4. The geometry of a linear dipole array of 10 uniformly spaced elements.

IV. NUMERICAL EXAMPLES In this section, three dipole arrays of different congurations are used as examples to demonstrate the performance of the proposed method. For each array, numerical simulations are performed with the CST Microwave Studio (MWS) [29] for array excitation determination by utilizing the procedure described in Section II. In the simulations, a Gaussian-modulated plane-wave pulse with a spectrum ranging from 1 GHz to 5 GHz is chosen to probe the array. The signals received by all antenna elements are recorded at the same sampling rate of 35 GHz with a record length of 10 ns. After array excitation is obtained, we feed them back into their corresponding elements and simulate the excited arrays with the CST MWS for far-eld radiation patterns. A. Linear Dipole Array
Fig. 3. Illustration of the beam steering angle beam pattern at different cases. (a) Case I, i.e., . (b) Case II, i.e., . of the realized

phase conjugated version of the received signals. Phase conjugation has been widely used to design retro-directive [26], [27] or phase conjugated arrays [28] for adaptive tracking applications. This kind of arrays can automatically steer its main lobe towards the direction of incident waves. We nd that the phenomenon is well explained in (25) and (26). Case II: at some other angles has the norm larger than its norm at , i.e., (27) and Fig. 3(b) depicts such a case. By examining the pattern formed by (17), we nd that the maximum gain of an array available at each viewing angle is fully characterized by . Generally, due to element mutual coupling and the non-isotropic radiation properties of the actual antenna elements, there possibly exists scan blindness in array radiation patterns where the electromagnetic energy cannot be effectively radiated by the array. From (17), it can be seen that the scan blindness is just related to dips or nulls in the pattern generated by . Under this scenario we do not consider array synthesis since the array can not effectively radiate electromagnetic energy from the dips or nulls of . But in the feasible array scan region where is relatively at and close to its global maximum magnitude, it can be proved from (20) that the realized beam pattern has at least one lobe at the desired angle . Therefore, in this region the time reversal based synthesis method is also valid to synthesize a beam-steering array in broadband.

The rst example is a linear dipole array of dipole elements, as shown in Fig. 4. The dipoles are uniformly positioned along the -axis with an inter-element spacing of 45 mm. All dipoles are designed in the same dimensions with a total arm-length of 40 mm and a center feed gap of 1 mm. The isolated dipoles operate at a center frequency of 3.05 GHz with a 10-dB bandwidth of 500 MHz ranging from 2.8 GHz to 3.3 GHz. This example aims to demonstrate the performance of the proposed method to synthesize a conventional linear array by providing one-dimensional (1D) steerable beam patterns over a wide frequency band. We chose 13 desired steering angles in the XoY plane (i.e., ) and synthesized a total of 65 beam patterns by using the proposed method. For each desired angle, we chose 5 discrete frequencies (i.e., 2 GHz, 2.5 GHz, 3 GHz, 3.5 GHz and 4 GHz) as samples to investigate the performance of the proposed method. The performance is evaluated with the averaging beam pointing error , which is dened by (28) is the total number of frequency samples in the inwhere vestigated bandwidth and is the resulting beam steering angle in azimuth obtained at the discrete sample frequency of . The numerical results show that the proposed method performs effectively in synthesizing the conventional linear dipole array in a frequency range 24 GHz. For illustration purposes, we present some samples of the synthesized beam patterns. Fig. 5 shows the beam patterns synthesized at 2 GHz, 3 GHz, and 4 GHz for the desired angles of 45 and 90 , respectively. It can be seen that the resulting beam steering directions are in agreement with their corresponding desired angles at all the

ZHAO et al.: TIME REVERSAL BASED BROADBAND SYNTHESIS METHOD FOR ARBITRARILY STRUCTURED BEAM-STEERING ARRAYS

169

Fig. 6. The average beam pointing errors calculated at different desired steering angles for the linear array.

Fig. 5. The synthesized beam patterns (XoY plane) of the linear dipole array at 2, 3, 4 GHz for the desired angles of 45 and 90 , respectively. (a) At 45 . (b) At 90 . Fig. 7. The geometry of a grounded circular dipole array consisting of 10 elements uniformly positioned around the center solid metal cylinder.

frequency points. Similar results are obtained at other desired steering angles. Fig. 6 depicts the average beam pointing errors calculated with (28) by uniformly sampling 5 discrete frequencies in 24 GHz for each desired steering angle. The results show that the proposed method provides good beam pointing accuracy in synthesizing the linear array. In this example, the maximum value of is about 3.0 , which occurs at and 165 . An interesting phenomenon is that the beam pointing errors appear symmetrically only in the angular regions between the broadside direction and the end-re direction. By investigating the gain pattern of the linear array, we observe that these errors primarily result from element mutual coupling effects. B. Circular Dipole Array The second example is a circular array, which consists of 10 dipoles uniformly positioned around a metal cylinder with a gap of 2 mm above the cylindrical surface. All dipoles have the same dimensions as the ones used in the rst example. The center solid metal cylinder is designed with height 1200 mm and radius 180 mm. Fig. 7 depicts the geometry of the grounded circular dipole array. Such type of array has wide applications in

commercial and military elds for tracking, imaging, and communications because it can produce almost identical beam patterns over a 360 scan range in azimuth. This example aims to demonstrate the performance of the proposed method to synthesize a simple conformal array with strong platform scattering. In this example, we chose 12 desired steering angles in the XoY plane (i.e., ) and 5 discrete sample frequencies (i.e., 2 GHz, 2.5 GHz, 3 GHz, 3.5 GHz, and 4 GHz) for each desired angle. We synthesized a total of 60 beam patterns excited with the array excitation determined by the proposed method. Fig. 8 depicts several samples of the realized beam patterns. Fig. 9 provides the average beam pointing errors , which are calculated with (28) by uniformly sampling 5 discrete frequencies in 24 GHz for each desired angle. The numerical results indicate that the proposed method works very well in synthesizing the circular dipole array even though the strong platform scattering effects are present. The realized beam patterns direct their main lobe accurately to their desired angles. The average beam pointing error is less than 0.35 at all steering

170

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 9. The average beam pointing errors calculated at different desired steering angles for the circular array.

Fig. 10. The geometry of a dome-shaped dipole array of 13 elements positioned above the curved surface of the dome.

Fig. 8. The synthesized beam patterns (XoY plane) of the circular array at 2, 3, 4 GHz for the desired angle of 60 and 270 , respectively. (a) At 60 . (b) At 270 .

angles of interest, as shown in Fig. 9. We also notice that the circular dipole array exhibits better beam pointing accuracy than the previously studied linear dipole array. This is mainly due to the geometrical symmetry of the circular array in the azimuth plane that effectively reduces the harmful impacts of element mutual coupling and platform scattering. C. Dome-Shaped Dipole Array The third example is a dome-shaped dipole array, which consists of 13 dipoles of the same dimensions with a total armlength of 38 mm and a center feed gap of 1 mm, as shown in Fig. 10. In this array, all dipoles are distributively placed 5 mm above the surface of a solid metal dome with a radius of 85 mm. The dome is seamlessly united together with a solid metal cylinder of height 40 mm and radius 85 mm. Among these dipoles, 8 elements are uniformly distributed around the body of the dome at the same elevation angle with an incremental azimuth interval of elements at the same elevation angle with an incremental azimuth interval of , and 1 element above the top of the dome . This example aims to demonstrate the performance of the proposed method to synthesize a relatively complex conformal

Fig. 11. The synthesized beam patterns for the dome-shaped array at 3.5 GHz and (b) for the desired steering angles at (a) , respectively.

array by providing two-dimensional (2D) steerable beam patterns over a wide frequency band. We chose 16 desired angles in the upper half-space . They are , and . For each desired angle, we selected 4 sample frequencies (i.e., 2.5 GHz, 3 GHz, 3.5 GHz, 4 GHz) for examining the characteristics of the proposed method. We synthesized a total of 64 beam patterns for this example using the proposed method. Fig. 11 shows the samples of the three-dimensional (3D) radiation patterns of the dome-shaped array that are synthesized at 3.5 GHz for the desired angles of and , respectively. Fig. 12 depicts the average

ZHAO et al.: TIME REVERSAL BASED BROADBAND SYNTHESIS METHOD FOR ARBITRARILY STRUCTURED BEAM-STEERING ARRAYS

171

Fig. 12. The average beam pointing errors calculated at different desired steering angles in the XoY and XoZ plane for the dome-shaped array, respectively.

Fig. 13. The beam pointing errors caused by signal truncation. The data are obtained using the linear dipole array.

beam pointing errors ( and ) computed at different desired steering angles. is the beam pointing error in elevation, which is calculated in the same way as by using (28). From Figs. 11 and 12, we observe that the desired beam pointing angle is reached in both elevation and azimuth. Due to the geometrical asymmetry of the array in the XoZ plane, the platform scattering and element mutual coupling cause a varying radiation efciency at different elevation angles. As a result, the beam pointing errors in the XoZ plane are slightly larger than those in the XoY plane. But the simulation results illustrate that the average beam pointing errors in the XoZ plane is less than 2.5 . V. DISCUSSION In this section, we discuss the impact of signal truncation and noise interference on beam pointing accuracy. A. Signal Truncation The tails of the transient signals received by antenna elements are sometimes truncated due to the limited physical memory of signal recorders or other reasons. In such cases, the actual array excitation becomes (29) where puted from the ideal array excitation comun-truncated signals by (3), and is the excitation deviation caused by signal truncation, which is calculated by (30) where tern is the truncation time. If the main lobe of the patgenerated by points to the desired angle , the following constraint must be satised (31) is the

or equivalently,

(32) where . Inserting (29) into (32) and discarding the negligible higher order terms, we obtain

(33) where . Equation (33) provides an equivalent constraint on the truncated signals. In order to achieve precise beam pointing, we can optimize the truncation time according to (32) or (33). In this paper we do not consider this optimization problem. Instead, we utilize the total energy ratio of the truncated received signals over their un-truncated waveforms to examine the impact of signal truncation on beam pointing accuracy. Fig. 13 shows the beam pointing errors caused by signal truncation. The results in Fig. 13 are obtained using the previously studied linear dipole array in Section IV. We observe that if the energy of the received signals after truncation is larger than 85% of the energy of their un-truncated waveforms, the beam pointing errors are small, typically less than 1 . However, if the truncation window becomes narrower, the beam pointing deviation increases. This is because more energy losses occur to the truncated signals. Similar results were observed in the other two dipole arrays. These results imply that, to achieve high beam pointing accuracy, the truncation window should be large enough to maintain the energy of the truncated signals as more as possible. B. Noise Interference For real-world implementation of the proposed method, the impact of noises must be taken into account. Commonly, an array at its early-stage is designed in anechoic chambers. In this scenario, the received signals are mainly interfered by Additive

172

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

VI. CONCLUSION In this paper we present a novel beam pattern synthesis method to synthesize arbitrarily structured beam-steering arrays for accurate beam steering over a broad frequency band. This method utilizes time reversal, rather than optimizing beam pattern cost functions at each frequency, to determine array excitation. The remarkable advantage of the proposed method is that one can synthesize a steerable array over a wide frequency band at once by a single run of time reversal experiment. The theoretical analysis is performed and the numerical validations are presented. The simulation results successfully demonstrate that the proposed method is capable of synthesizing arbitrarily-structured beam-steering arrays with accurate beam pointing over a wide frequency range for both one-dimensional (1D) and two-dimensional (2D) steerable pattern synthesis problems. REFERENCES White Gaussian Noises (AWGN). Considering the presence of the zero-mean AWGN, the actual signals received by antenna elements become (34) where is the ideal waveform received by the -th element and is its corresponding noise. By (3), the actual array excitation is (35) where is the ideal array excitation without any noise interference, and is the excitation deviation caused by AWGN. Obviously, affects the beam pattern if we feed into the array as the excitation. But we notice that the zero-mean AWGN can be effectively suppressed by averaging multiple snapshots of the received signals, i.e.,
[1] L. Josefsson and P. Persson, Conformal Array Antenna Theory and Design. New York, NY: Wiley-Interscience, 2006. [2] E. Botha and D. A. McNamara, Conformal array synthesis using alternating projections, with maximal likelihood estimation used in one of the projection operators, Electron. Lett., vol. 29, no. 20, pp. 17331734, Sep. 1993. [3] P. Y. Zhou and M. A. Ingram, Pattern synthesis for arbitrary arrays using an adaptive array method, IEEE Trans. Antennas Propag., vol. 47, no. 5, pp. 862869, May 1999. [4] L. I. Vaskelainen, Constrained least-squares optimization in conformal array antenna synthesis, IEEE Trans. Antennas Propag., vol. 55, no. 3, pp. 859867, Mar. 2007. [5] D. Marcano and F. Duran, Synthesis of antenna arrays using genetic algorithms, IEEE Antennas Propag. Mag., vol. 42, no. 3, pp. 1219, Jun. 2000. [6] J. L. Guo and J. Y. Li, Pattern synthesis of conformal array antenna in the presence of platform using differential evolution algorithm, IEEE Trans. Antennas Propag., vol. 57, no. 9, pp. 26152621, Sep. 2009. [7] S. Chamaani, S. A. Mirta, M. Teshnehlab, M. A. Shooredeli, and V. Seydi, Modied multi-objective particle swarm optimization for electromagnetic absorber design, Progr. Electromagn. Res., vol. 79, pp. 353366, 2008. [8] Z. B. Lu, A. Zhang, and X. Y. Hou, Pattern synthesis of cylindrical conformal array by the modied particle swarm optimization algorithm, Progr. Electromagn. Res., vol. 79, pp. 415426, 2008. [9] K. R. Dandekar, H. Ling, and G. Xu, Experimental study of mutual coupling compensation in smart antenna applications, IEEE Trans. Wireless Commun., vol. 1, no. 3, pp. 480487, Jul. 2002. [10] I. J. Gupta, J. R. Baxter, S. W. Ellingson, H. G. Park, H. S. Oh, and M. G. Kyeong, An experimental study of antenna array calibration, IEEE Trans. Antennas Propag., vol. 51, no. 3, pp. 664667, Mar. 2003. [11] H. M. Aumann, A. J. Fenn, and F. G. Willwerth, Phased array antenna calibration and pattern prediction using mutual coupling measurements, IEEE Trans. Antennas Propag., vol. 37, no. 7, pp. 844850, Jul. 1989. [12] M. Fink, Time reversal acoustics, Phys. Today, vol. 50, pp. 3440, May 1997. [13] G. Lerosey, J. de Rosny, A. Tourin, A. Derode, G. Montaldo, and M. Fink, Time reversal of electromagnetic waves, Phy. Rev. Lett., vol. 92, p. 194301, May 2004. [14] P. Kyritsi and G. Papanicolaou, One-bit time reversal for WLAN applications, in Proc. IEEE 16th Int. Symp. on Personal, Indoor and Mobile Radio Communications, 2005, pp. 532536. [15] R. Qiu, C. Zhou, N. Guo, and J. Zhang, Time reversal with MISO for ultrawideband communications: Experimental results, IEEE Antennas Wireless Propag. Lett., vol. 5, pp. 269273, 2006. [16] Y. Jin, J. M. F. Moura, and N. O. Donoughue, Time reversal in multiple-input multiple-output radar, IEEE J. Sel. Topics Signal Processing, vol. 4, no. 1, pp. 210225, Feb. 2010. [17] M. E. Yavuz and F. L. Teixeira, Ultrawideband microwave sensing and imaging using time-reversal techniques: A review, Remote Sensing, vol. 9, pp. 466495, 2009.

Fig. 14. The beam pointing errors caused by the zero-mean AWGN. The data in the legends is the number of are obtained using the linear dipole array. snapshots.

(36) where is the number of the received signal snapshots. Fig. 14 depicts the beam pointing errors caused by the zero-mean AWGN. The results Fig. 14 are obtained from the linear dipole array by using the averaged signals , to determine array excitation with different number of data snapshots , and 100, respectively. It can be seen clearly that the beam pointing error reduces remarkably as the number of snapshots increases. This is due to the improved signal-to-noise ratio (SNR) in the averaged signals. We applied the same averaging method to the other two dipole arrays. Similar results were observed. Ideally, if the received signals are averaged by a large number of snapshots, the impact of the noise will be negligible and the beam pattern will be the same as the one in absence of noises.

ZHAO et al.: TIME REVERSAL BASED BROADBAND SYNTHESIS METHOD FOR ARBITRARILY STRUCTURED BEAM-STEERING ARRAYS

173

[18] H. T. Nguyen, J. B. Andersen, and G. F. Pedersen, The potential use of time reversal technique in multiple elements antenna system, IEEE Commun. Lett., vol. 9, no. 1, pp. 4042, Jan. 2005. [19] A. G. Cepni and D. D. Stancil, Single antenna microwave nulling using time-reversal techniques, in Proc. IEEE IMS05, 2005, vol. 4, pp. 17231726. [20] Y. Jin, J. M. F. Moura, and N. ODonoughue, Adaptive time reversal beamforming in dense multipath communication networks, in Proc. 42nd Asilomar Conf. on Signals, Systems and Computers, Oct. 2008, vol. 4, pp. 20272031. [21] E. G. Farr and C. E. Baum, Time domain characterization of antennas with TEM feeds, Sensor Simul. Notes, vol. 426, pp. 116, Oct. 1998. [22] C. E. Baum, General properties of antennas, IEEE Trans. Electromagn. Compat., vol. 44, no. 1, pp. 1824, Feb. 2002, Sensor and Simulation Notes 330. [23] G. S. Smith, A direct derivation of a single-antenna reciprocity relation for the time domain, IEEE Trans. Antennas Propag., vol. 52, no. 6, pp. 15681577, Jun. 2004. [24] D. M. Pozar, Active element pattern, IEEE Trans. Antennas Propag., vol. 42, no. 8, pp. 11761178, Aug. 1994. [25] G. Golub and C. V. Loan, Matrix Computations, 3rd ed. Baltimore, MD: The Johns Hopkins Univ. Press, 1996. [26] R. Y. Miyamoto and T. Itoh, Retrodirective arrays for wireless communications, IEEE Microw. Mag., vol. 3, no. 1, pp. 7179, Mar. 2002. [27] Y. C. Guo, X. W. Shi, and L. Chen, Rectrodirective array technology, Progr. Electromagn. Res. B, vol. 5, pp. 153167, 2008. [28] L. D. DiDomenico and G. M. Rebeiz, Digital communications using self-phased arrays, IEEE Trans. Microw. Theory Tech., vol. 49, no. 4, pp. 677684, Apr. 2001. [29] Workow and Solver Overview, CST Microwave Studio, CST Studio Suite, 2008.

Yuanwei Jin (S99M04SM08) received the M.S. and B.S. degrees from East China Normal University, Shanghai, in 1996 and 1993, respectively, and the Ph.D. degree in electrical and computer engineering from the University of California at Davis, in 2003. From 2003 to 2004, he was a Visiting Researcher with the University of California at Santa Cruz. From 2004 to 2008, he was a Postdoctoral Research Fellow, then Project Scientist with Department of Electrical and Computer Engineering at Carnegie Mellon University, Pittsburgh, PA. Since August 2008, he has been an Assistant Professor with Department of Engineering and Aviation Sciences, University of Maryland Eastern Shore, Princess Anne. Since August 2010, he has been serving as Interim Department Chair. His research interests are in the general area of statistical signal and image processing, with applications in radar/sonar, biomedical imaging, structural health monitoring, and wireless communications. He has published over 50 technical journal and conference papers. Prof. Jin is afliated with several IEEE societies, Sigma Xi, SPIE, the American Society of Engineering Education (ASEE), and the American Association for Cancer Research (AACR). He was a recipient of the Earle C. Anthony Fellowship from the University of California at Davis in 1997 and 1998 and received the 2010 U.S. Air Force Summer Faculty Fellowship award.

Deshuang Zhao received the M.S. degree in electromagnetic eld and microwave engineering and the Ph.D. degree in optical engineering from the University of Electronic Science and Technology of China (UESTC), Chengdu, in 2001 and 2005, respectively. From March 2002 to March 2003, he was a Research Assistant with the Wireless Communications Research Centre, City University of Hong Kong. From November 2009 to October 2010, he was a Visiting Scholar with the Department of Engineering and Aviation Science, the University of Maryland Eastern Shore. Since 2006, He has been an Associate Professor with UESTC. His current research interests include antennas and propagation, ultrawideband radio, radar imaging, and microwave photonic techniques.

Bing-Zhong Wang (M06) received the Ph.D. degree in electrical engineering from the University of Electronic Science and Technology of China (UESTC), Chengdu, in 1988. He joined the UESTC in 1984 and is currently a Professor there. He has been a Visiting Scholar at the University of Wisconsin-Milwaukee, a Research Fellow at the City University of Hong Kong, and a Visiting Professor in the Electromagnetic Communication Laboratory, Pennsylvania State University, University Park. His current research interests are in the areas of computational electromagnetics, antenna theory and technique, electromagnetic compatibility analysis, and computer-aided design for passive microwave integrated circuits.

Rui Zang is currently working toward the M.S. degree at the University of Electronic Science and Technology of China, Chengdu. His main research interests include antenna theory and design, ultrawideband communications, and electromagnetic compatibility analysis.

174

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

A Generalized Hybrid Approach for the Synthesis of Uniform Amplitude Pencil Beam Ring-Arrays
Ovidio Mario Bucci, Fellow, IEEE, and Daniele Pinchera, Member, IEEE
AbstractA generalized hybrid algorithm for the synthesis of uniform amplitude ring-arrays is introduced. The method exploits the analytical properties of the eld radiated by a circular array in order to get advantages from the use of convex programming techniques. The method, aimed to the synthesis of circularly symmetric patterns, is very exible and allows the use of different kind of feeds as well as stepped excitations. The synthesis procedure is demonstrated in the case of high-directivity pencil beam patterns. Index TermsAntenna arrays, genetic algorithms, optimization, quadratic programming.

I. INTRODUCTION SE of active array antennas for satellite communications is attracting increasing interest as an alternative to reector antennas [1][3]. In order to have a competitive solution, as compared to reector antennas in terms of reliability, cost, and power efciency, it is important to reduce as much as possible the number of required ampliers and phase shifters (i.e., the control points). Furthermore, in order to achieve the maximum efciency in the DC-RF conversion chain, all the ampliers should work under the same optimal conditions, thus providing equi amplitude (isophoric) excitations. This last requirement is in contrast with the necessity to have a tapering of the array excitations, in order to satisfy the side lobe level (SLL) constraints on the synthesized beam, so that it is not possible to use uniformly spaced planar arrays. It is possible to identify at least three different architectures of non uniformly spaced arrays: thinned arrays [4][6], where the positions of the radiating elements are selected from a regular lattice, sparse arrays [7][18], where the non-regular element positions can be freely chosen within an assigned area, and clustered arrays [19], where the elements of an otherwise standard arrays are clustered into single entry sub-arrays. This work will be focused on the sparse array architecture but, as it will be shown in the following, the exibility of the proposed method allows to obtain mixed architectures with clustered elements and element-size tapering [20].

Manuscript received January 27, 2011; revised May 18, 2011; accepted July 20, 2011. Date of publication October 03, 2011; date of current version January 05, 2012. This work was supported by the European Space Agency under Contract ESA/ESTEC 21689/08/NL/ST. O. M. Bucci is with the Dipartimento di Ingegneria Biomedica, Elettronica e delle Telecomunicazioni, Universit Federico II di Napoli, I-80125 Napoli, Italy (e-mail: bucci@unina.it). D. Pinchera is with the DAEIMI, via G. Di Biasio 43, Universit di Cassino, Cassino, Italy (e-mail: pinchera@unicas.it). Digital Object Identier 10.1109/TAP.2011.2167931

Generally speaking, in the synthesis of a sparse planar isophoric array of identical elements, once the kind of radiating element has been chosen, the problem is to determine their positions in the available area, in such a way to satisfy, possibly in an optimal fashion, the design specications. From the mathematical point of view, this amounts to perform a (possibly constrained) minimization of a properly chosen functional. This minimization can be done adopting different approaches. The simplest one consists in the direct synthesis of all the element positions such that the radiated eld fullls the given constraints. Due to the presence of the elements positions in a complex exponential, the relationship between the elements positions and the array factor is strongly non-linear; for this reason, the solution of this kind of problems (minimization of a strongly non-convex functional on a non-convex search space)1 by means of local optimization techniques becomes unaffordable, since the algorithm can get easily stuck in a local minimum of the functional. On the other hand, the use of global optimization algorithms [21], [22] could be problematic, since the number of unknowns is proportional to the total number of radiating elements used, and the computational effort of global optimization algorithms increases very rapidly with the number of unknowns [23], if the attainment of the optimum must be ensured within a prescribed precision. And so, in the case of arrays with a large number of elements a brute force procedure could be ineffective. A second possibility is the use of a deterministic density taper technique: in a rst step a continuous source fullling in an optimal way the given constraint is synthesized; then the locations of the radiating elements is determined exploiting a proper density taper rule [4], [9], [10], [13], [17], [18], allowing to emulate at the best the continuous aperture distribution. Since the rst step can be dealt with convex programming techniques and the second step is deterministic, the approach is very quick, but we do not have guarantees on the optimality of the obtained solution. Moreover, in the available density taper approaches, we cannot take explicitly into account the feed pattern. In this paper we present an effective strategy able to overcome above limitations, in the relevant case of circularly symmetric pencil beams. In this case it is quite natural to put the radiating elements equispaced along circles: in this way we have just to search for the ring radii, the number of feeds per ring and the relative orientation of the feed along the rings, achieving a strong reduction of the parameters describing the array. Furthermore, as it will be shown in the next Sections, a further computational improvement is obtained by reformulating the
1The non convexity of the search space is due to the presence of the nonoverlapping constraints on the elements.

0018-926X/$26.00 2011 IEEE

BUCCI AND PINCHERA: A GENERALIZED HYBRID APPROACH FOR THE SYNTHESIS OF UNIFORM AMPLITUDE PENCIL BEAM RING-ARRAYS

175

problem in such a way to allow the use of a hybrid approach for the solution of a simplied ring-array synthesis problem, in order to exploit as much as possible its partial convexity. It has to be underlined that, in principle, we are not guaranteed that the optimal array layout should be a ring array, which, accordingly, could provide a sub-optimal solution. However, this geometry is certainly appealing, due its simplicity and to the strong reduction of the number of unknowns, which greatly improves the performance of the optimization algorithm. Once such a sub-optimal layout has been obtained, it could be rened by a further local optimization. Nevertheless, it will be shown that such a renement is usually not worthy. In the following, rst the mathematical framework for the problem will be illustrated, then the synthesis method will be presented and a number of numerical results will be given and discussed. Conclusions follow. II. STATEMENT OF THE PROBLEM As the beam can be conveniently steered to any desired direction by means of a linear phase shift, while scanning losses can be taken into account by a proper strengthening of the requirements on the far eld, we can limit ourselves to consider the case of a (circular) spot beam pointing at boresight. Let us rst consider a circular array of radius with identical equispaced isophoric elements, with common excitation (see Fig. 1). The corresponding array factor, , reads ( time convention)

Fig. 1. Geometry of the circular array.

(1) the free space propagation constant and the angular position of the element. By exploiting the well known relationship [24] being

(2) (1) can be recast as

for values of well inside , which in the case of our interest, that is, large focusing antennas, is much larger than the beam-width, which is of the order of , being the radius of the whole antenna. Accordingly, in the angular sector around the bore-sight direction wherein the eld is signicant, higher order eld harmonics are negligible and thus the array factor is practically independent. Accordingly, if we are interested in the design of a high gain circular antenna array (as in the case for satellite communications), we can simply use the term of (3) to obtain a good approximation for the array pattern in the region wherein the constraints are relevant, using the complete series (or expression (1)) for a successive validation of the results. Under this approximation, it is straightforward to obtain the expression of the pattern radiated by a set of circular arrays with radii , each one with a number of elements equal to , as

(5) where is the (copolar) pattern of the feeds on the k-th ring (considered uniform in ),2 and , so that a value of can be used to take into account the absence or presence of a feed in the center of the array. It has to be underlined that under approximation (5), the array factor is independent of , thus achieving a further reduction of the search space. From this point on we will thus search only for the set of the ring radii and the number of feeds per ring.
2Unless we deal with strongly asymmetric feeds, this assumption is likely to be satised in the angular region of interest. Anyway, if required, it can be removed by exploiting the azimuthal Fourier expansion of the feed pattern to generalize relation (5).

(3) As well known, Bessel functions of rst kind go rapidly to zero when the argument is smaller than the order. Accordingly, even the rst non-zero harmonic in (3) becomes negligible for

(4) wherein is the wavelength and the distance between the feeds. And so, all higher order eld harmonics are negligible

176

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

This is still a strongly non linear, non convex, problem. However, relation (6) shows that if the radii and the element factors are xed, the relationship between the radiated eld and the excitations is linear. Accordingly, given the feeds and the rings geometry, the corresponding optimal excitations can be found by solving a simple, convex, quadratic programming problemsee [25] for more details. This circumstance suggests to extend to our case the Hybrid Approach (HA), proposed in [26], in order to reduce the number of unknowns to be dealt with a global optimization algorithm, thus improving its effectiveness. III. THE HYBRID APPROACH AND ITS IMPLEMENTATION
Fig. 2. Pattern mask; the mask is uniform with the variable.

The synthesis problem can be thus reduced to the search of the appropriate and (and, possibly, ) with in such a way to satisfy given constraints (pattern shape, directivity, physical dimensions of the array, non overlapping of the feeds and so on) in an optimal way. We have to underline that, while neglecting the dependence is quite justied in the region wherein the pattern constraints are relevant, so that the synthesis of the array layout can be safely performed under this approximation, higher order harmonics becomes signicant far from the boresight, decreasing the directivity of the actual array with respect to that given by (5). Accordingly, this last expression, which physically corresponds to substitute the actual array with one consisting of continuous ring sources, provides an upper bound for the array performances. Obviously, such continuous rings source provides a more realistic and tight upper bound as compared to a continuous aperture antenna, often used as reference source, since it is closer to the actual array and can include some of the physical constraints of the design. Once each ring array is substituted with a continuous source, there is no reason to restrict its excitation, i.e., (see (5)) to be an integer. And so, substituting in place of in (5), we obtain

(6) In the case of satellite communications, the far eld specications are usually given in term of a minimal directivity to be achieved at the end of coverage (EOC) of the beam and of an upper bound mask for the relative (to the EOC value) power pattern outside the EOC (see Fig. 2). Accordingly, the continuous ring version of our problem can be stated as follows: Find the rings radii, the corresponding excitations (and, possibly, feed types), such that the radiated pattern is equal to one (i.e., 0 dB) at the EOC, fullls the mask constraints and maximizes the EOC directivity, i.e., minimizes the radiated power.

In its essence, the idea lying at the basis of the HA is very simple: exploit the partial convexity of the overall synthesis problem with respect to some variables (in our case the rings excitations) to eliminate these variables from the overall cost function, by substituting in it their optimal values (for given values of the other variables). In such a way the global optimization algorithm must deal only with the non convex variables (rings radii and feed type, in our case), with a benecial effect on its convergence, hence on the reliability of the obtained results. Of course, this requires the solution of a convex programming problem for each step of the global optimization: however, due to the availability of very efcient convex programming algorithms, the overall balance is favorable to the HA. The synthesis can be thus decomposed into two parts, a global optimization problem and a convex programming one; in particular the convex part will be called by the optimization algorithm within its cost function. Not withstanding its simplicity, the actual implementation of the HA is by no means trivial. Since we have divided the synthesis procedure in two parts we have to understand how the constraints on the search space inuence the behavior of each one of them. A rst constraints is that the feeds belonging to the same ring should not overlap: we can code this constraint only in the convex part of the hybrid algorithm. A second constraint is that the feeds of different rings should not overlap: this constraint could be coded in the evolutionary part of the hybrid algorithm. The same is true for the possible choice among various kind of feeds. These points inuence both the choice and implementation of the global optimization algorithm and the implementation of the convex sub-problem, as detailed in the following. A. Choice and Implementation of the Global Algorithm While a generally optimal global search algorithm does not exists (No-Free Lunch Theorems [27][29]), if we have enough information on the problem we are going to solve we can exploit this information to choose and set up an algorithm tailored to the class of problems to be solved, in order to achieve a better convergence. This led us to choose a genetic algorithm [30] (GA) as a global optimizer. This choice has been driven by the fact that GAs are very exible and tunable; furthermore, each set of radii needs to verify radial non overlapping feeds constraint, that is very easy to implement in a GA if the gene contains the coding of the radii as real numbers.

BUCCI AND PINCHERA: A GENERALIZED HYBRID APPROACH FOR THE SYNTHESIS OF UNIFORM AMPLITUDE PENCIL BEAM RING-ARRAYS

177

Taking into account the statement of the problem (see Section II), the natural cost function is obviously the total radiated power, or, equivalently, the inverse of the EOC directivity, . Actually, since we are not guaranteed that the considered set of rings are able to radiate a pattern that ts the mask, we should modify the denition of the cost function in order to take this issue into account. We would like the algorithm to search rst for a eld belonging to the mask: only when this specication has been met, the algorithm can search for the maximum directivity pattern. The mask-tting constraint can be taken into account by introducing a second cost function, f. i., the norm, , of the eld that outts the mask. This would lead to a multi-objective problem, but since the power cost has to be taken into account only when the mask-tting cost is zero (in practice, below a threshold), we can adopt the following stepped cost function: (7) is a threshold that has to be chosen according to the where numerical precision of the used computer. The cost is positive and equal to when the mask specication is not veried; it becomes negative, and equal to the opposite of the EOC directivity when the mask specication is veried. We are thus guaranteed that any layout with a pattern not fullling the mask would have a cost greater than one with pattern belonging to the mask, and so we are sure that the algorithm will try rst to nd a pattern tting the mask, and then it will optimize for the directivity. It has to be underlined that the tournament selection scheme is robust with respect to the use of a cost function that can assume positive and negative numbers; this approach, in the large performed investigation campaign, has been proven to be very robust and effective. It should be also noted that the nature of the genetic algorithm is intrinsically parallel, thus allowing a great speed-up when properly programmed to exploit the presence of multicore/multi-threaded processors. B. The Implementation of the Convex Part of the Problem As seen in Section II, the ring excitations , to be determined in the convex sub-problem, are, apart from a discretization, proportional the numbers of feed to put on the rings. Now, for each ring, we must satisfy the azimuthal non-overlapping constraint (i.e. non overlapping of the feeds belonging to the same ring). Assuming circular feeds, this implies that we cannot have for a ring of radius a number of circular feeds greater than (8) where is the feed radius of the elements on the -th ring. This constraint on the number of feeds can be translated into a constraint on the dynamic of the excitations, i.e. (9)

if a central element is present, and

(10) if there is no central element.3 Equation (9) assumes that we are putting a single element in the center of the array, while (10) assumes that we are putting the maximum possible number of feeds in the rst ring. It is important to underline that both are convex constraints, so the overall problem remains convex. Once the values of satisfying the non overlapping constraint are obtained, the values of can be found as

(11) In some cases it could be useful to x the overall number of feed of the nal array. This specication could be also coded directly in the convex problem, by adding another linear constraint

(12) that obviously does not change the convex nature of the problem. It is important now to recall one of the main results of Section II. The angle for which the effect of the higher order harmonics in (3) needs to be taken into account is function of the distance between the feeds belonging to the same ring. This means that we could control the inuence of higher harmonics by imposing a maximum inter-feed distance on each ring, in order to verify the condition (4) for a larger angular sector. This would lead to the following further convex constraints: (13) if a central element is present, and

(14) imposes a if there isnt any central element. The factor maximum element distance on each ring. From the geometrical viewpoint, (9), (13) or (10), (14) restrict the excitations, seen as function of the radii, to belong to an angular sector. A nal observation is now in order. The optimal solution and the corresponding number of feeds per circle will be, in general, slightly different, since there is a quantization error, due to the discrete nature of the feeds number; for the synthesis it is thus preferable to calculate the optimal solution and then to compute the cost
3Equations (9) and (10) can be easily generalized to the case of feeds of non circular shape.

178

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I SUMMARY OF THE SPECIFICATIONS OF [2];

IS THE FREE SPACE WAVELENGTH FOR THE CENTRAL FREQUENCY OF THE EMPLOYED X-BAND

Fig. 4. a) Layout of the 13-ring UCAs array; b) layout of the same layout after a local optimization of the positions.

Fig. 3. The 12 radii found by 4 instances of the GA after a variable number of iterations. (a) GA after 100 iterations; (b) GA after 3000 iterations.

function using the corresponding ; in this way the global search algorithm would also optimize for a set of radii that minimizes the discretization error. IV. NUMERICAL EXAMPLES In order to assess the effectiveness and the exibility of the proposed approach, in the following we present a number of numerical examples. To be denite, we will focus on a specic case of interest for the realization of an array antenna for satellite applications, able to radiate multiple pencil beams (see [2] for more details). The main requirements for the beam pattern are reported in Table I. All simulations have been performed assuming as feeds uniform circular apertures (UCA), possibly with variable radius. As a preliminary step, in order to verify the possibility to actually achieve, with good precision, the global optimum, an analysis of the performances of the implemented genetic algorithm has

been carried out. The considered problem is that of nding the optimal radii in the case of 12 continuous ring sources, using as feeds UCA of radius . Fig. 3(a) reports the values of the radii found by four different instances of the GA after 100 iterations, while Fig. 3(b) shows the radii found after 3000 iterations. While the solutions found after 100 iterations are slightly different, after 3000 iterations they are almost perfectly coincident, conrming that the absolute optimum has been reached. The continuous ring case directivity in the four cases of 100 iterations is in the interval 46.11 dBi46.17 dBi, and the directivity after 3000 iterations is 46.19 dBi. This behavior has been conrmed in other cases, indicating that we have either a high number of close local minima or a relatively at region surrounding the absolute minimum. And so if we are interested in nding the global optimum with high precision we have to use an high number of iterations (or a certain number of separate instances of the optimizer), whereas if we are interested in assessing the performance level of a certain geometry we could adopt relatively few iterations. It must also be noticed that the ring radii turn out to be linearly related to the radius index; this is probably due to the fact that we have not enforced any constraint on the overall number of feeds or on the dynamic of the auxiliary solution , so the ring array behaves similarly to a linear array. According to these results, unless differently stated, the number of iterations for the GA will be xed to 100. A. Isophoric Arrays of Identical Elements As a rst example, an isophoric array of 13 rings of identical UCAs of radius equal to has been synthesized. The obtained

BUCCI AND PINCHERA: A GENERALIZED HYBRID APPROACH FOR THE SYNTHESIS OF UNIFORM AMPLITUDE PENCIL BEAM RING-ARRAYS

179

TABLE II PERFORMANCE SUMMARY

Fig. 5. a) Pattern of the 13-ring UCAs array; b) pattern of the same layout after a local optimization of the positions.

Fig. 6. Comparison of the performances for a variable

layout, consisting of 379 feeds, is shown in Fig. 4(a), whereas the -cuts of the corresponding directivity pattern are shown in Fig. 5(a). The EOC directivity is equal to 44.10 dBi and the maximum side lobe level on Earth is 20.02 dB. It is interesting to observe that, as expected, the -cuts are overlapped up to . It has to be underlined that the values of have been set in a random way, so it could be possible to obtain slightly better results by an optimal choice of these values. However, as said before, the amount of improvement is very marginal, so it has not been considered in this work. In order to judge the quality of the obtained solution, we recall that the optimal continuous aperture sources with a radius, which has an aperture efciency of 0.8 dB and achieves an EOC directivity of 46.8 dBi [25], while that of the continuous ring array is 46.24 dBi, showing that the azimuthal discretization has a much higher impact that the radial one. This should have been expected, as the directivity decrease is just the reex of the decrease of the aperture efciency due to its discretization and sparsication. It is easy to realize that, if we neglect coupling effects, in the case of isophoric feeds with uniform aperture distribution, the aperture efciency coincides with the lling factor, i.e., the ratio between the total surface occupied by the feeds and that of the continuous aperture, which, in our case, amounts to 3.7 dB, with a decrease of 2.9 dB with respect to the optimal continuous distribution, in good agreement with the actual EOC directivity decrease. To quantify the degree of optimality of our solution we can compare it with the result of a further local optimization of the feed position, allowing the feeds to move outside the circles [31]. The obtained layout is depicted in Fig. 4(b), and the corresponding pattern is shown in Fig. 5(b); the obtained EOC direc-

tivity is 44.5 dBi, so with an increase of only 0.4 dB.4 For sake of completeness a summary of the results is given in Table II, where we have added the case of a full populated layout of 698 non isophoric UCAs located on 15 rings (whose excitation have been chose to deliver the optimal directivity) and that of a layout of 382 elements obtained by means of the deterministic density taper approach (DA) presented in [18]. Note that the EOC directivity of the HA isophoric layout is only 1.55 dB lower than the full populated non-isophoric layout, which has almost a double number of control points. On the other side, the improvement over the deterministic approach is marginal, showing that in the case of isophoric array of identical elements the DA and HA are essentially equivalent. Accordingly, the main advantage of the HA is its exibility, which allows to deal with a much larger class of architectures, as it will be shown in the following. In the example above, no constraints on the number of elements have been given, apart from the non-overlapping one, so that the total number of feeds is not xed in advance, but derives from the discretization of the continuous excitations via (10) and (11). Because, as pointed out before, the EOC directivity is essentially related to the lling factor, which depends on the number of feeds, it is interesting to see as the EOC directivity varies with . This can be done by enforcing the constraint (12) in the convex part of the synthesis. Fig. 6 shows the behavior of both the continuous and discretized ring EOC directivities as a function of .
4This behavior has been conrmed in other cases analyzed: if the GA converged to a good solution, the successive local optimization typically improves the directivity of less than 0.5 dB.

180

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Since completely removing the constraints on the excitation amplitude would lead either to a non optimal use of the feeding ampliers or to the necessity of adopting complex and inefcient beamforming networks, it could be interesting to see what happens if we group the elements into subarrays by means of 3-dB power dividers. When we use a power divider the intensity of the excitation on the two feeds is reduced of a factor , but the number of antennas doubles, thus resulting in an overall increase of the ring excitation of a factor. According to this consideration, if we have (15) when a central element is present and

Fig. 7. Performance comparison for the coefcient .

UCAs ring array by varying the

(16) when there isnt a central element, we can put on the -th circle a number of feeds grouped in couples controlled by a 3-dB power divider equal to

As can be seen, the continuous ring directivity decreases as the number of elements is increased; this is due to the fact that enforcing the constraint (12) does not allow the use of the optimal set of excitations. On the contrary the performances of the discretized case increase with the number of feeds (because of the reduction of the effect of the higher order harmonics) up to a value of , practically equal to that of the considered example, and then stays constant. This is a very interesting result, since it shows that, for a sparse isophoric array of identical elements, the EOC directivity cannot be increased beyond a certain level by simply increasing the number of feeds, because at the end this would induce a performance decrease, due to the impossibility to realize the proper tapering of the aperture distribution. The increase of the feeds number can be also obtained by enforcing constraints (13), (14). In Fig. 7 it is possible to see the effect of varying the coefcient . The gure is organized in 4 subplots: the rst reports the directivity for the continuous ring case; the second that of the discretized ring case; in the third plot we report the side lobe level up to 16 , whereas in the fourth plot the overall number of elements is reported. It is interesting to see that introducing a lower limit for the excitation dynamic decreases signicantly the performances in the continuous ring cases only for ; on the contrary the directivity in the discretized array increases with the decrease of , and values of in the range 1.53 are particularly interesting. B. Relaxing the Isophoricity Constraint The major drawback of the above ways for increasing the lling factors is the increase of the number of elements, which is in contrast with the request of a minimal number of control points. Furthermore, the analysis of the results reported in the previous sub-section indicates that the increase of the number of elements per ring forces the equivalent aperture distribution of the array to depart more and more from the optimum one, as testied by Figs. 6 and 7.

(17) This procedure, in the following referred to as stepped excitation, can be used also in case of nested 3-dB dividers; in the case of a single 3-dB divider we will speak of 1-level stepped excitation, in the case of a double 3-dB divider we will speak of 2-level stepped excitation and so on. The use of power divider allows an higher number of feeds on each ring for a given number of control points, without the need of restricting the excitation dynamic by means of the coefcient . The drawbacks of this solution are the introduction of losses by the power dividers, and the reduction of the steering performance due to the grouping of the feeds. Fig. 8 shows a layout of 12 rings of UCAs in which we have introduced the possibility to have stepped excitations; a has been used in order to have the maximum number of feeds per ring when using the stepped excitations. The obtained EOC directivity is 44.84 dBi, but we are now using 508 feeds for an overall number of 342 control points, that is still a very large number. The directivity -cuts are plotted in Fig. 9. It is interesting to observe that, because of the increased number of feeds per ring, the -cuts coincide up to about . Of course, it would be possible to use a multistepped excitation, but while the directivity improvement turns out to be marginal, possibly canceled be the increased losses, scanning performances would be signicantly worsened, due to the larger grouping of the elements. C. Using Feeds of Different Sizes According to (5) and (6), our model allow us to use a different kind of feed per ring. And so, we could try to exploit a kind of size tapering in order to improve the radiating characteristics

BUCCI AND PINCHERA: A GENERALIZED HYBRID APPROACH FOR THE SYNTHESIS OF UNIFORM AMPLITUDE PENCIL BEAM RING-ARRAYS

181

Fig. 8. Layout of the geometry employing stepped excitations; white feeds: no power divider employed; gray feeds: 3-dB power divider employed.

Fig. 10. Layout of the geometry employing UCAs of

and

radii.

Fig. 9. Directivity plot -cuts for the layout employing stepped excitations.

Fig. 11. Directivity plot -cuts for the layout employing UCAs of radii.

and

of our array [20]. As a matter of fact, the previously reported results show that, in order to achieve the proper tapering of the equivalent aperture distribution, looser feeds or stepped excitation must be employed in the more external rings. This naturally suggest to use bigger feeds in these parts of the array. To this end, a rst possibility could be to leave the user to identify which rings are to be lled with a kind of feed and which have to be lled with the other one. A more interesting possibility, which we adopt here, is that of leaving to the GA to choose both the ring radii and the type of feed for each radius. This would add a discrete variable vector to the search space, so the convergence speed would be only slightly inuenced, with only minor modications of the optimization code. Fig. 10 shows the result of the optimization of a 11 ring layout employing UCAs of radii equal to and . The EOC directivity obtained with this layout is 44.73 dBi, with a side lobe level on Earth of 20.72 dB and an overall number of feeds equal to 314. The directivity -cuts are reported in Fig. 11. As foreseen, the use of feeds of different sizes allows a significant increase of the EOC directivity, while the number of feeds is reduced.

Since both the use of a stepped excitation and of feeds of different sizes improve the array performances, with a minor impact on the complexity of the antenna, we could try to synthesize a layout employing both approaches, in order to achieve the minimum number of control points. Fig. 12 shows 10-ring layout employing two kind of UCAs, of radii and , and using stepped excitations when needed. The corresponding directivity -cuts are depicted in Fig. 13. The performances of this architecture are very good, achieving a side lobe level of 20.59 dB and an EOC directivity of 44.85 dBi, i.e. only 0.8 dB less than a fully populated array of about 700 elements (see Table II), using only 280 feeds and 230 control points. V. CONCLUSIONS A generalized hybrid approach for the synthesis of large sparse arrays has been presented. The method exploits the analytical properties of the eld radiated by a circular array in order to split the synthesis in a global optimization procedure, on a strongly reduced search space, and a convex programming problem.

182

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

possibly extended to the case of any shaped circularly symmetric pattern, with a proper modication of the convex part of the algorithm. REFERENCES
[1] G. Toso, C. Mangenot, and A. G. Roederer, Sparse and thinned arrays for multiple beam satellite applications, in Proc. 29th ESA Antenna Workshop on Multiple Beams and Recongurable Antennas, Apr. 2007, pp. 207210. [2] ESA/ESTEC Tender AO/1-5598/08/NL/ST, Innovative Architectures for Reducing the Number of Controls of Multiple Beam Telecommunications Antennas, [Online]. Available: www.esa.int [3] ESA/ESTEC Tender AO/1-6338/09/NL/JD, Active Multibeam Sparse Aray Demonstrator, [Online]. Available: www.esa.int [4] R. E. Willey, Space tapering of linear and planar arrays, IRE Trans. Antennas Propag., vol. 10, pp. 369377, 1962. [5] D. G. Leeper, Thinned aperiodic antenna arrays with improved peak sidelobe level control, U.S. Patent No. 4071848, Nov. 1976. [6] Y. Cailloce, G. Caille, I. Albert, and J. M. Lopez, A ka-band direct radiating array providing multiple beams for a satellite multimedia mission, in Proc. IEEE Int. Conf. on Phased Array System and Technology, 2006, pp. 403406. [7] A. Ishimaru, Theory of unequally-spaced arrays, IRE Trans. Antennas Propag., vol. 10, pp. 691702, 1962. [8] A. Ishimaru and Y. S. Chen, Thinning and broadbanding antenna arrays by unequal spacings, IEEE Trans. Antennas Propag., vol. 1, pp. 3442, 1965. [9] W. Doyle, On Approximating Linear Array Factors, RAND Corp. Mem RM-3530-PR, 1963. [10] M. I. Skolnik, Nonuniform arrays, in Antenna Theory, Part I. New York: McGraw-Hill, 1969, ch. 6. [11] Y. T. Lo, A study of space-tapered arrays, IEEE Trans. Antennas Propag., vol. 14, no. 2, pp. 2230, 1966. [12] B. P. Kumar and G. R. Branner, Synthesis of unequally spaced arrays using Legendre series expansion, in Proc. IEEE Antennas and Propagation Society Int. Symp., 1997, vol. 4, pp. 22362239. [13] T. M. Milligan, Space-tapered circular (ring) array, IEEE Antennas Propag. Mag., vol. 46, no. 3, pp. 7073, 2004. [14] M. Vincente-Lozano, F. Ares-Pena, and E. Moreno, Pencil-beam pattern synthesis with a uniformly excited multi-ring planar antenna, IEEE Antennas Propag. Mag., vol. 42, no. 6, pp. 7073, 2000. [15] G. Toso, M. C. Vigan, and P. Angeletti, Null-matching for the design of linear aperiodic arrays, presented at the IEEE Antennas and Propagation Society Int. Symp., Honolulu, HI, Jun. 1015, 2007. [16] O. M. Bucci, M. DUrso, T. Isernia, P. Angeletti, and G. Toso, Deterministic synthesis of uniform amplitude sparse arrays via new density taper techniques, IEEE Trans. Antennas Propag., vol. 58, no. 6, pp. 19491958, 2010. [17] O. M. Bucci, T. Isernia, A. F. Morabito, S. Perna, and D. Pinchera, Aperiodic arrays for space applications: An effective strategy for the overall design, in Proc. EuCAP, Mar. 2009. [18] O. M. Bucci and S. Perna, A deterministic two dimensional density taper approach for fast design of uniform amplitude pencil beam arrays, IEEE Trans. Antennas Propag., vol. 59, no. 8, pp. 28522861, Aug. 2011. [19] G. Caille, Y. Cailloce, C. Guiraud, D. Auroux, T. Touya, and M. Masmousdi, Large multibeam array antennas with reduced number of active chains, in Proc. 2nd Eur. Conf. on Antennas and Propagation (EUCAP 2007), Edinburgh, U.K., Nov. 1116, 2007, pp. 19. [20] O. M. Bucci, T. Isernia, A. F. Morabito, S. Perna, and D. Pinchera, Density and element-size tapering for the design of arrays with a reduced number of control points and high efciency, presented at the 4th Eur. Conf. on Antennas and Propagation, Barcelona, Spain, Apr. 2010. [21] D. G. Kurp, M. Himdi, and A. Rydberg, Synthesis of uniform amplitude unequally spaced antenna arrays using the differential evolution algorithm, IEEE Antennas Propag. Lett., vol. 51, pp. 22102217, 2003. [22] N. Jin and Y. Rahamat-Samii, Advances in particle swarm optimization for antenna design: Real-number, binary, single-objective and multiobjective implementation, IEEE Trans. Antennas Propag., vol. 55, no. 3, pp. 557567, 2007. [23] A. S. Nemirovsky and D. B. Yudin, Problem Complexity and Method Efciency in Optimization, ser. Interscience Series in Discrete Mathematics. New York: Wiley, 1983.

Fig. 12. Layout of the geometry employing stepped excitations and UCAs of and radii; white feeds: no power divider employed; gray feeds: 3-dB power divider employed.

Fig. 13. Directivity plot -cuts for the layout employing stepped excitations and radii. and UCAs of

The proposed synthesis method is able to achieve practically optimal solutions, and its exibility allows to synthesize layouts with different kind of feeds per each ring, since the radiating element pattern is directly taken into account in the synthesis algorithm, and/or the use of stepped excitation. This allows a strong reduction of the number of control points, thus making the use of direct radiating arrays interesting for a larger class of problems. Furthermore, with a proper and relatively simple modication of the mathematical framework, the technique can be also extended to the case of feeds with patterns depending on the variable, f.i., square apertures or sub-arrays. The performances achievable, even with a small number of global optimization steps, are very good, so that the obtained solutions do not need to be further ameliorated by further local optimization which would move the feeds out of the rings. It can be also foreseen that the synthesis procedure, here demonstrated for the case of pencil beam patterns, could be

BUCCI AND PINCHERA: A GENERALIZED HYBRID APPROACH FOR THE SYNTHESIS OF UNIFORM AMPLITUDE PENCIL BEAM RING-ARRAYS

183

[24] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, ser. Applied Mathematics Series. Washington, DC: National Bureau of Standards, 1964. [25] O. M. Bucci, T. Isernia, and A. F. Morabito, Optimal synthesis of directivity constrained pencil beams by means of circularly symmetric aperture elds, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 13861389, 2009. [26] T. Isernia, F. J. A. Pena, O. M. Bucci, M. DUrso, J. F. Gmez, and J. A. Rodrguez, A hybrid approach for the optimal synthesis of pencil beams through array antennas, IEEE Trans. Antennas Propag., vol. 52, pp. 29122918, 2004. [27] D. H. Wolpert and W. G. Macready, No free lunch theorems for optimization, IEEE Trans. Evol. Comput., vol. 1, no. 1, pp. 6782, Apr. 1997. [28] D. Corne and J. Knowles, Some multiobjective optimizers are better than others, in Proc. Congress on Evolutionary Computation, Dec. 812, 2003, vol. 4, no. 1, pp. 25062512. [29] B. Weinberg and E. G. Talbi, NFL theorem is unusable on structured classes of problems, in Proc Congress on Evolutionary Computation, Jun. 1923, 2004, vol. 1, pp. 220226. [30] A. Fraser and D. Burnell, Computer Models in Genetics. New York: McGraw-Hill, 1970. [31] ESA/ESTEC Tender AO/1-5598/08/NL/ST, Innovative architectures for reducing the number of controls of multiple beam telecommunications antennas, J.N. Sahalos WP 4000 Report: Assessment of the Array Synthesis Technique. Ovidio Mario Bucci (F93) was born in Civitaquana, Italy, on November 18, 1943. He graduated summa cum laude in Electronic Engineering from the University of Naples, Italy, in 1966 and received the Libera Docenza degree in theory and technique of electromagnetic elds (allowing one to become a tenured university professor) in 1971. He was an Assistant Professor at the Istituto Universitario Navale of Naples, 19671975, then Full Professor of electromagnetic elds at the University of Naples. He was Director of the Department of

Electronic Engineering, 198486 and 198990, Vice Rector of the University of Naples, 19942000, Director of the CNR Institute of Electromagnetic Environmental Sensing (IREA) 20012010. He is the author or coauthor of more than 370 scientic papers, mainly published on international scientic journals or proceedings of international conferences. His scientic interests include scattering from loaded surfaces, reector and array antennas, efcient representations of electromagnetic elds, near-eld far-eld measurement techniques, inverse problems and noninvasive diagnostics, biological applications of nanoparticles and electromagnetic elds. Prof. Bucci is a Fellow of the IEEE since 1993 and a Member of the Academia Pontaniana. He was President of the National Research Group of Electromagnetism, of the MTT-AP Chapter of the Centre-South Italy Section of IEEE, and Director of the Interuniversity Research Centre on Microwaves and Antennas (CIRMA). He was a recipient of the International Award GUIDO DORSO for Scientic Research, 1996, and of the Presidential Gold Medal for Science and Culture, 1998, among others.

Daniele Pinchera (S05M08) received the Dr. Eng. degree (summa cum laude) in telecommunication engineering and the Ph.D. in information and electronic engineering from the University of Cassino, in 2004 and 2008, respectively. He is currently working as a Postdoctoral Researcher for the Faculty of Engineering, University of Cassino. His current research interests are in the elds of smart antennas and MIMO systems, large array synthesis, compressed sensing, sensor networks and industrial and medical applications of microwaves.

184

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Polarimetry With Phased Array Antennas: Theoretical Framework and Denitions

Karl F. Warnick, Senior Member, IEEE, Marianna V. Ivashina, Member, IEEE, Stefan J. Wijnholds, Member, IEEE, and Rob Maaskant, Member, IEEE
AbstractFor phased array receivers, the accuracy with which the polarization state of a received signal can be measured depends on the antenna conguration, array calibration process, and beamforming algorithms. A signal and noise model for a dual-polarized array is developed and related to standard polarimetric antenna gures of merit, and the ideal polarimetrically calibrated, maximum-sensitivity beamforming solution for a dual-polarized phased array feed is derived. A practical polarimetric beamformer solution that does not require exact knowledge of the array polarimetric response is shown to be equivalent to the optimal solution in the sense that when the practical beamformers are calibrated, the optimal solution is obtained. To provide a rough initial polarimetric calibration for the practical beamformer solution, an approximate single-source polarimetric calibration method is developed. The modeled instrumental polarization error for a dipole phased array feed with the practical beamformer solution and single-source polarimetric calibration was dB or lower over the array eld of view for elements with alignments perturbed by random rotations with 5 degree standard deviation. Index TermsArray signal processing, phased array antennas, polarimetry.

I. INTRODUCTION

HE radio astronomy community is currently developing polarimetric aperture arrays and phased array feeds (PAFs) for large reectors [1][4]. Accurate polarization state measurements for observed sources is critical to the science goals for current and planned phased array instruments. With a traditional waveguide feed, the polarization properties of the receiver are xed at the time of manufacture and unwanted instrumental polarization can be calibrated by observing sources with known polarization parameters. For a phased array receiver, the polarimetric properties of each formed beam or image pixel can be adjusted on the y by changing beamformer coefcients. If array output correlations are computed

Manuscript received January 12, 2011; revised June 20, 2011; accepted June 21, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. K. F. Warnick is with the Department of Electrical and Computer Engineering, Brigham Young University, Provo, UT 84602 USA (e-mail: warnick@ee.byu.edu). M. V. Ivashina is with the Department of Earth and Space Sciences, Chalmers University of Technology, S-41296 Gothenburg, Sweden (e-mail: marianna. ivashina@chalmers.se). S. J. Wijnholds is with the Netherlands Institute for Radio Astronomy (ASTRON), NL-7991 PD, Dwingeloo, The Netherlands (e-mail: wijnholds@astron.nl). R. Maaskant is with the Department of Signals and Systems, Chalmers University of Technology, S-41296 Gothenburg, Sweden (e-mail: rob. maaskant@chalmers.se). Digital Object Identier 10.1109/TAP.2011.2167926

and stored, one set of observation data can be processed with multiple sets of beamformer coefcients tuned to optimize sensitivity, sidelobe level, or polarimetric accuracy. Exploiting this exibility and achieving best possible system performance requires the development of a theory for polarimetric phased arrays, including gures of merit, optimal beamformer solutions, and practical calibration strategies. Key questions that must be answered by this theory include the following: How do astronomical performance criteria relate to the standard IEEE denitions for polarimetric antennas? What beamforming algorithm will simultaneously optimize for high SNR and polarimetric accuracy? Which requirements should be set on the antenna array and beamformer design to achieve optimal performance? How can a polarimetric array be accurately and efciently calibrated? This paper will consider the rst two questions in detail and addresses the third empirically through a numerical study. An approximate single-source calibration scheme is presented to address the fourth issue. A full treatment of polarimetric calibration is beyond the scope of this paper and will be addressed in future work. The rst question arises because antenna engineers assess the polarimetric performance of antenna systems in terms of the axial ratio, cross-polarization discrimination (XPD), and crosspolarization isolation (XPI), while astronomers judge instrument performance and express system requirements in terms of Stokes parameters [5], [6]. Another challenge is that the standard IEEE denitions of the axial ratio, XPD and XPI have been established for single port systems, and these gures of merit must be extended to phased array systems that are capable of forming multiple dual-polarized beams simultaneously. Astronomical antenna applications also have unique constraints because radiation in terrestrial communication systems is typically highly polarized, whereas astronomical sources have a small but important polarized component of a few percent or less relative to the total power signal ux density. The starting point for answering these questions is the development of a signal and noise model for a polarimetric phased array receiver. This is accomplished in Section II. The treatment leads to the general problem of beamforming for rank-two signals and the concept of a polarimetric array beam pair. In Section III, the standard IEEE denitions for polarimetric gures of merit are related to the beam pair Jones matrix. In Section IV, the optimal beam pair solution is derived for a perfectly known instrument, and a practical beamforming method based on signal correlation matrix eigen-

0018-926X/$26.00 2011 IEEE

WARNICK et al.: POLARIMETRY WITH PHASED ARRAY ANTENNAS: THEORETICAL FRAMEWORK AND DEFINITIONS

185

vectors is considered for partially characterized polarimeters. An important result is that if sample estimation error (the error incurred by computing the array output correlation matrix from a nite number of voltage samples) is neglected, the practical eigenvector max-SNR method is equivalent to the optimal solution when polarimetrically calibrated. The performance of practical beamformers are compared to the ideal solution using a numerical model in Section V. Voltage and eld quantities are phasors with the convention. An overbar is used to denote three-dimensional eld and position vectors, whereas vectors of voltages are typeset in boldface . The superscript designates the complex conjugate and the conjugate transpose. denotes expectation over time.

II. POLARIMETRIC PHASED ARRAY MODEL The purpose of a radio polarimeter is to measure the polarization properties of an incident electromagnetic wave as a function of the angle of arrival. Fig. 1 illustrates an -element polarimetric beamforming array comprised of two groups of antenna elements with nominally orthogonal polarization. The antenna system is assumed to be illuminated by a point source radiating partially-polarized elds. The electric eld intensity vector at the point radiated by such a source can be approximated in the neighborhood of the receiver by the incident plane wave (1) where and are orthogonal unit vectors according to one of Ludwigs polarization denitions relative to the coordinate system of the array [7], and is the wave vector corresponding to the angle of arrival. Since and are dened with respect to the coordinate system of the array, we are neglecting in this treatment a rotation of the polarization state from the astronomical coordinate system on the sky to the coordinate system of the array. In the following theoretical development, we will consider the polarimetric calibration problem for one beam-steering direction. The calibration process must be applied to every formed phased array beam for each desired beam direction, as the numerical results make clear. Array calibration is accomplished with point sources, rather than extended sources, so all signal response quantities are assumed to arise from a point source. For an imaging array, each pixel represents a different beam steering direction and set of array beamforming coefcients. The gures of merit and beamforming procedures developed in this paper apply independently to each image pixel. The source of interest is assumed to be a point source, and all results for gures of merit are calculated at the beam center. Since extended astronomical sources are common, beam polarization patterns are important, but this aspect of phased array polarimetry will be considered in future work. The antenna output signals are amplied to form the -element output voltage vector , which is subsequently combined into the output voltages and using the beamformer weight vectors and respectively, each of which is a column
Fig. 1. A radio polarimeter comprised of a dual-polarized actively beamformed receiving antenna array.

vector of size . Together, and constitute a polarimetric beam pair for a given sky pointing direction1. The electric eld components and are complex random processes in the phasor or complex baseband representation. The polarization state of the plane wave is determined by the covariance matrix of the two eld components, which is the Hermitian matrix (2) The covariance matrix has two real and one complex degrees of freedom, or four real degrees of freedom. The time-average power ux density of the incident wave is (3) is the characteristic impedance of free space and tr is where the matrix trace operation. Some authors rearrange the covariance matrix to form the coherency vector [6] (4)

where is the vector obtained by stacking the columns of . The relative magnitudes and phases of these quantities determine the degree of polarization and the polarization state of the polarized part of the wave. A. Array Signal Response With reference to Fig. 1, we will model the antenna array output signals in terms of the voltage responses of
the less general bi-scalar method, and are weight vectors of size and are only associated to the - and -oriented antenna elements, respectively.
1In

186

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

the array to the unit intensity, time-harmonic, linearly polarized waves (5a) (5b) where and are vectors of the respective voltages induced by the two waves at the array receiver outputs before beamforming. Even though and may not be orthogonal due to possible non-ideal polarimetric response of the receiver. The voltage responses and are not exactly known in practice, but are useful in developing a system model for a beamforming phased array antenna. We will show in Section IV.A that if and were known, the beam pair can be exactly calibrated polarimetrically. Experimental procedures for calibrating the array in the case that and are not known are discussed in Section IV.B. Since and are orthogonal, it follows by linearity that for an arbitrarily polarized wave, the array signal voltage response vector can be written using and as (6) The array output signal voltage covariance matrix is

The covariance matrix

of the two beam outputs is (12)

In terms of the array signal and noise covariance matrices, the beam pair output covariance matrix is (13) Introducing the vector (14) leads to the more compact expression (15) where (16a) (16b) This result provides a signal and noise model for the polari. metric beam pair III. POLARIMETRIC DEFINITIONS

(7) which is of rank one for a fully polarized wave and of rank two for a partially polarized or unpolarized incident wave. Upon introducing the matrix (8) we can write (7) in the more compact form (9) Assuming that the phased array system noise can be characterized by the noise covariance matrix , the complete array output voltage covariance matrix can be described as (10) This expression provides a signal and noise model for a polarimetric array. The precise form of the noise response for an array receiver is not important here, but is considered in detail in [8][10]. B. Beam Pair Signal and Noise Response After beamforming, the two output voltages obtained with the beamformer pair are (cf. Fig. 1) (11a) (11b)

In this section, we consider the relationship between IEEE denitions for polarimetric gures of merit and the Jones and Mueller matrix formulations that are common in radio astronomy and remote sensing. We also consider the problem of quantifying the joint sensitivity of a polarimetric beam pair. A. IEEE Denitions for Single Port Antennas The IEEE denitions for polarization terms for single port antennas are as follows [5]: Cross-polarization discrimination (XPD) is the ratio of the power level at the output of a receiving antenna, nominally co-polarized with the transmitting antenna, to the output of a receiving antenna of the same gain but nominally orthogonally polarized to the transmitting antenna. Cross-polarization isolation (XPI) is the ratio of the wanted power to the unwanted power in the same receiver channel when the transmitting antenna is radiating nominally orthogonally polarized signals at the same frequency and power level. Other gures of merit such as axial ratio and intrinsic crosspolarization ratio [11] can be applied to polarimetric phased arrays but will not be considered further in this paper. B. Denitions for Phased Array Antennas With reference to the polarimetric phased array model developed in Sec. II, the signal response of an array antenna to coand cross-polarized incident elds and is described by the two beamformer output voltages (17)

WARNICK et al.: POLARIMETRY WITH PHASED ARRAY ANTENNAS: THEORETICAL FRAMEWORK AND DEFINITIONS

187

(18) and are the output voltage vectors of the receiving where elements in response to pure and signals. The weight vectors and of the two beamformers are assumed to be dened such that they optimally receive or at least approximate the optimal reception of and , respectively, according to some specied criterion to be discussed in Section IV. This implies that the polarizations of the two formed beams are nominally aligned with the and directions. While and are dened as real vectors in (1), the denitions in this section hold for any pair of waves and with orthogonal polarizations, whether linear, circular, or elliptical. Ideally, one expects the two output voltage vectors to be directly proportional to and , but this may not hold due to polarization leakage caused by imperfections in the element geometry or mutual coupling between the array elements, in particular when the antennas are placed in a nite array exhibiting strong truncation effects. In the latter case, the polarization characteristics of the embedded element patterns differ signicantly from each other. The degree of polarization leakage is dependent not only on the array geometry and mechanical construction, but also on the values of the beamformer coefcients. To quantify the beam-dependent polarization leakage, we extend the standard denitions in Section III.A as follows: Cross-polarization discrimination is dened as the ratio of powers received at the beamformer outputs 1 and 2 due to the same incident eld, i.e., or : (19) (20) and are the signal covariwhere ance matrices in response to pure and signals of unit intensity. Cross-polarization isolation is dened as the ratio of powers received at the same beamformer output, 1 or 2, due to orthogonally-polarized incident elds and : (21) (22)

modeled or measured in order to infer the incident wave polarization state. For a phased array, each polarimetric beam pair has an associated Jones matrix. The general relationship between the beam outputs and the incident electric eld intensity vector is (23) where the two by two matrix on the right side of this equation is the beam pair Jones matrix, which we will denote in the following as . Using the notation developed above, for an array this relationship becomes

(24) The Jones matrix can be identied as (25) where (26) Using this result, the beam pair output signal voltage covariance matrix (16a) can be written as (27) The goal of polarimetric calibration is to transform a given beam to a new beam pair for which the radiation patterns pair are steered to the same angle of arrival as the original beam pair but having Jones matrix as close as possible to the identity matrix. If , the cross-polarization gures of merit are large in value. If the initial Jones matrix is known, it can be readily seen that the beam pair (28) is ideally polarimetrically calibrated and the realized Jones matrix is . If the beam pair output signal coveriance matrix is measured for one signal with a known polarization state , the relationship (27) does not uniquely x the beam pair Jones matrix. Given known input and output polarization parameters, by substitution it can be shown that the general solution to (27) has the form (29) where is an arbitrary unitary matrix. This shows that with only a single calibrator source, there remains a 2 2 unitary degree of freedom in the Jones matrix. The physical interpretation of the unitary degree of freedom is discussed in [6], [13]. The ambiguity can be removed using knowledge of the nominal element polarizations (see Section IV.C), measurements of

C. Jones Matrix Formulation Unlike radio communications, for which antennas are fabricated to meet a xed polarization purity requirement, applications such as radio astronomy and remote sensing that rely on accurate measurement of wave polarization states also require operational polarimetric calibration. When modeling system effects that contribute to the measured wave polarization state, it is convenient to use the Jones matrix formulation [12]. The Jones formulation allows matrix terms for various system components to be chained together into an overall Jones matrix that must be

188

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

additional calibrator sources, or correlation with a second calibrated polarimetric antenna. In terms of the elements of the Jones matrix, the cross-polarization gures of merit (19)(22) are [11] (30a) (30b) These expressions show that the standard antenna polarization gures of merit are measures of the magnitudes of the off-diagonal elements and relative to the diagonal elements and . D. Mueller Matrix Formulation While Jones matrices operate in the voltage phasor domain, 4 4 Mueller matrices represent transformations on wave polarization states in the correlation or Stokes parameter domain. Since the Mueller matrix formulation is commonly used in the astronomical literature, the treatment will be rehearsed here and placed into the mathematical framework developed in earlier sections. For a signal characterized by the eld coveriance matrix (2), the Stokes vector containing the four Stokes parameters is dened as [14, pp. 9798]

parameters, (35), are identical, apart from a constant gain factor. In the following, we will include the gain calibration with polarimetric calibration, and consider the system to be polarimetrically calibrated when . The Mueller and Jones matrix formulations are connected by known formulas, which we will review here. Stokes parameters are related to the coherency vector (4) by (36) where (37)

The inverse of this transformation matrix is (38)

Using this relationship for the incident wave and the beam output voltages (39) Using the properties of the Kronecker product, (27) can be rearranged in the form (40)

(31)

Some authors dene the Stokes parameters with a factor of , where is the intrinsic impedance of space, so that the Stokes parameters have units of power density (W/m . It can be shown that [15, p. 29] (32) The degree of polarization is (33) For an unpolarized wave, . A Stokes polarimeter can be represented by the relationship (34) is a vector of the Stokes parameters of the incident where wave given by (31) and is the Mueller matrix of the system. The vector of Stokes parameters referred to the beam output voltages is

In view of (4), this becomes (41) Combining these results shows that the system Mueller matrix is related to the Jones matrix by (see also [16]) (42) In the present context, the Jones matrix is weight dependent, so that, by using (25) along with the properties of the Kronecker product

(43) If the polarimeter is ideally calibrated, then , and . For a realistic system congured as a Stokes polarimeter, polarimetric calibration means measuring the beam pair Mueller matrix and using that information to transform the weight pair , or in general adapt the beam pair covariance matrix , such that the effective Mueller matrix becomes proportional to the identity matrix. Polarimetric accuracy can be quantied by the error in the measured Stokes parameters for a given source or by the deviation of the Mueller matrix from the identity. The relative RMS Stokes error is

(35)

The beam pair is said to be polarimetrically calibrated if the Stokes parameters of the source, (31), and the measured Stokes

WARNICK et al.: POLARIMETRY WITH PHASED ARRAY ANTENNAS: THEORETICAL FRAMEWORK AND DEFINITIONS

189

(44) This metric can be applied to computed Stokes parameters before the instrument is polarimetrically calibrated, to determine the raw or uncalibrated Stokes error, or it can be applied after calibration, to assess the performance of both the calibration procedure and the native instrumental polarization properties. As with the Jones formulation, source-independent measures of polarimetric accuracy are also desirable. By the denition of the induced operator norm, we have the bound (45) For weakly polarized sources, imately , and we have approx-

where is the system noise equivalent bandwidth, is Boltzmanns constant, and is the power ux density of the incident wave in W/m . As required by the denition of effective area, the incident wave is polarization matched with the polarization of the beam. The signal correlation matrix for a time-harmonic incident wave can be expressed as (49) . The polarization of the incident wave where is matched to the beam when is maximized, which occurs for the incident eld state (50) where the scale factor follows from (3). Using these results in (48) leads to (51) This same result can be obtained from (48) for an unpolarized with in the leading scale factor rewave placed by half the power ux density of the unpolarized wave. To dene the beam pair sensitivity, we seek a gure of merit that is independent of the beam pair polarimetric calibration. This means nding bounds on the beam sensitivity (48) with an arbitrary linear combination of and . Using (51), the sensitivity is (52) where , where (53) (52) becomes (54) This shows that the ratio of quadratic forms in (52) lies within the eld of values of the matrix [17]. Since is Hermitian, the eigenvalues are real and the eld of values is an interval in the on the real line. It follows that for any beamformer subspace spanned by the beam pair the sensitivity is bounded by (55) and are the eigenvalues of . By substituting , where is an arbitrary invertible 2 2 matrix, it can be seen that the matrix is independent of linear transformation of the beam subspace and hence of the beam pair polarimetric calibration. Since the response to orthogonal polarized waves is not directly available for a phased array in practice, it is of interest to express the sensitivity bound in terms of measurable array where is an arbitrary vector. With

(46) The quantity (47) therefore is an approximate upper bound on the relative RMS might be referred to as the Stokes instruStokes error. mental polarization bound. This bound is derived for the case of a noise-free system. Since estimation error due to noise can be reduced to arbitrarily low levels by integration, (47) is adequate for many purposes, including antenna design optimization, but when analyzing astronomical observation data in practice, the additional effect of noise on polarimetric accuracy should be considered. E. Beam Pair Sensitivity The above treatment has considered traditional polarimetric formulations for the case of a phased array antenna. Another important aspect of beamforming for polarimetric arrays is that the beamformer coefcients control the beam output SNR and antenna sensitivity. For one beam, it is straightforward to determine the beamformer coefcients that maximize sensitivity given a knowledge of the array signal and noise responses. For a beam pair, the sensitivity of both beams should ideally be as high as possible. In particular, the beam sensitivities of an uncalibrated beam pair and the polarimetrically calibrated beam pair are in general different, and there are many possible polarimetrically calibrated beam pairs that have lower sensitivity than the classical maximum-SNR beamformer solution. Therefore, it is of interest to characterize the sensitivity of both outputs of a beam pair with a metric that is independent of the beam pair polarimetric calibration. For one beam , the sensitivity is (48)

190

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

output quantities. In the Appendix it is shown that the sensitivity bound can be expressed in the form

(56) The calibration-independence of this form of the bound follows from the properties of the matrix trace. We will refer to two beam pairs and as sensitivity equivalent if they lead to the same upper and lower sensitivity bounds in (55) or (56). By the above derivation, if the beam pairs are related by a linear transformation, they are sensitivity equivalent, and a beam pair is sensitivity equivalent to its polarimetrically calibrated counterpart . Finally, we observe that a polarimetrically calibrated beam pair does not necessarily achieve the upper bound in (55). The degree to which polarimetric calibration reduces beam sensitivity will be studied empirically in Section V. IV. POLARIMETRIC BEAMFORMING For a polarimetric phased array, calibration has two aspects: 1) determining the beamformer pair so that sensitivity and other beam gures of merit are optimal, and; 2) determining the beam pair Jones matrix or the Mueller matrix so that the incident wave polarization state can be inferred from the beam pair outputs. Combining 1) and 2) enables high sensitivity polarimetry. Beamformer coefcients applied to a phased array can be updated dynamically to accomplish various goals, such as sensitivity maximization as the noise environment changes, interference mitigation, or beam pattern sidelobe control. Consequently, the polarization properties of the array beam outputs are not necessarily xed. This introduces the possibility of combining the beamformer optimization and polarimetric calibration steps, so that the beam pair satises a specied set optimality criteria and also is polarimetrically calibrated with both the Jones and Mueller matrices approximately equal to identity matrices. We will consider the ideal case of a perfectly known system, in order to understand the ultimate performance limits of a given beam pair, as well as the practical case for which polarization properties of a beam pair must be calibrated empirically using observations of sources with known polarization parameters. To avoid sacricing observation time for polarimetric calibration, one can observe with non-polarimetrically calibrated beam pairs having imperfect polarization discrimination. Therefore, it is important to nd an approach for obtaining best possible polarimetric beam pairs using the single-source calibration data required for non-polarimetric array beamforming, and to characterize the polarization purity of beam pairs that are obtained without additional effort at polarimetric calibration. A. Optimal Beamforming for a Perfectly Known System Any beam pair that responds to two non-colinear incident polarizations can be polarimetrically calibrated, but the resulting beams may have poor sensitivity, which implies a low SNR and

inaccurate measured Stokes parameters for a given integration time. We will show that there is a unique polarimetrically calibrated beam pair that minimizes estimation error in measured Stokes parameters for a point source at a given angle of arrival. Measured Stokes parameters are a linear combination of the elements of a sample estimate of the 2 2 covariance matrix given by (16a). We will denote the sample estimated ma. With the standard trix for a given integration length as is the difference between radiometric detection technique, an on-source measurement and off-source measurement of the beam pair output coveriance matrix: (57a) (57b) signies a sample-estimated coveriance matrix. The where difference is

(58) After polarimetrically calibrating the beam pair outputs, the output covariance matrix becomes

(59) This is the same measurement equation that would be obtained with (58) subject to the calibrated beam pair (60) Therefore, if Stokes parameters are computed by applying the beam pair inverse Jones matrix to the polarization parameters obtained from an uncalibrated beam pair , the measurement error due to system noise can be analyzed as if the beam pair were replaced by a polarimetrically calibrated beam pair. The goal is to nd the polarimetrically calibrated beam pair that minimizes the estimation error in (59), which we will write with (60) as (61) The matrix is described stochastically by the Wishart distribution [18], which means that is the difference of two identically distributed Wishart random matrices. The mean of is the zero matrix. In the low SNR limit, the variance of the entries of is (62) where is the number of voltage samples that are averaged to produce and . This result shows that in order to minimize estimation error in the Stokes parameters, the diagonal elements of the 2 2 beam pair output noise correlation matrix must be minimized.

WARNICK et al.: POLARIMETRY WITH PHASED ARRAY ANTENNAS: THEORETICAL FRAMEWORK AND DEFINITIONS

191

This consideration reduces the polarimetric beamformer calibration problem to the joint constrained minimization problem (63) (64) Since is positive denite, we can add the two objective functions and recast the constrained minimization problem in the form (65) The constrained optimization problem can be solved by setting to zero the matrix derivative (66) where is a matrix of Lagrange multipliers, which will be chosen to satisfy the constraint in (64). Evaluating the derivative leads to (67) (68) Using the constraint , and solving for , we nd that (69) from which it follows that (70) The solution to the optimization problem is found by substituting (70) in (68) to yield (71) This pair of beamformer weight vectors minimizes the output system noise power while constrained to be polarimetrically calibrated with respect to a source with a given angle of arrival. The optimal solution is related to a simple rank-two generalization of the classical maximum-SNR beamformer [19]: (72) This beam pair realizes two maximum sensitivity beamformer weight vectors, with the Jones matrix . The optimal beamformer solution (71) can be written as (73) which shows that the beam pair is not in general polarimetrically calibrated but is sensitivity equivalent to the optimal beam pair.

B. Practical Polarimetric Array Calibration and the Eigenvector Method In practice, the voltage response matrix is difcult to measure for a PAF on a reector antenna, since ideal, orthogonally polarized astronomical sources are unavailable. A practical polarimetric array calibration procedure using observations of unpolarized and partially polarized sources is needed. For a dual-polarized array, neglecting estimation error, the signal covariance matrix is of rank two for an unpolarized source. The latter is readily concluded from (9) by taking , so that (74) which is a sum of two rank-one matrices. The eigenvalues and eigenvectors of are dened by (75) Neglecting estimation error, two of the eigenvalues are nonzero. Veidt proposed to use the two principal eigenvectors and of as conjugate eld match (CFM) beamformer weights [3]. We will refer to the beam pair (76) as the eigenvector CFM beamformer. The eigenvector CFM method can be modied to form the maximum-SNR eigenvector beamformer weight vectors (77a) (77b) This is the eigenvector max-SNR beamformer algorithm. To compare this beamformer with the optimal solution presented in the previous section, we note that the voltage response vectors and must span the same subspace as the eigenvectors and . This implies that (78a) (78b) which is equivalent to (79) Substitution in (71) gives

(80) which has a similar form as (71). Upon introducing and , (80) can be written as (81)

192

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

This result shows that the eigenvector max-SNR beam pair is sensitivity equivalent to the optimal beam pair. A modication of the above approach can be obtained using the generalized eigenvalue form of the max-SNR beamformer: (82) is rank two, the generalized eigenvalue problem has Since two nonzero eigenvalues. The corresponding principle eigenvalues provide a maximum-SNR beam pair. It can be easily shown that this beam pair is also sensitivity equivalent to the optimal beam pair (71). The same beam pair can also be obtained by using the noise correlation matrix to pre-whiten the array output signals and using the eigenvector CFM method with the pre-whitened signal correlation matrix. To summarize, we have so far ve polarimetric beamforming algorithms: (83a) (83b) (83c) (83d) (83e) The rst and second beam pairs require knowledge of the array response to orthogonally polarized incident waves, which is not directly available in practice. The third, fourth, and fth are practical beamformers in the sense that they can be computed from measurable signal output correlation matrices for unpolarized or partially polarized sources. The rst beam pair is exactly polarimetrically calibrated, but in general the others are not. In the absence of estimation error, all of these beam pairs except for CFM are sensitivity equivalent. Sample estimation error in the array output correlation matrices may affect each beamformer algorithm differently, but the effect of estimation error on the beamformer weights can be driven to low levels by integrating the array output during the calibration phase for a long period of time, so this effect is not considered here. C. Approximate Calibration With a Single Unpolarized Source The practical beamforming algorithms, (83b)(83e), are not polarimetrically calibrated. The eigenvectors computed from (75) are arbitrary up to a scale factor, and the resulting beam pair has a unitary degree of freedom according to (29). In order to give meaning to cross-polarization gures of merit, these degrees of freedom must be xed by some concrete, repeatable algorithm. Since each beam pair can be viewed as the output of a dualpolarized single pixel feed, any polarimetric calibration technique that is used for conventional feeds could be applied to calibrate the beam pair. Standard polarimetric calibration methods require observations of multiple sources with known Stokes parameters or tracking a polarized source over time as it rotates relative to the feed. Since these observations would have to be repeated with the telescope steered so that the source was at the center of each formed phased array beam, a straightforward implementation of standard polarimetric calibration could re-

quire many hours or even days. It is possible that efcient multiple-source calibration methods for phased arrays can be found, but a full treatment of efcient polarimetric calibration procedures for phased arrays is beyond the scope of this paper. Since determining beamformer weights for a phased array requires observations of a bright (and typically unpolarized) source over a grid of telescope pointings at the center of each beam [20], the goal here is to nd the best possible polarimetric calibration procedure given this already available observation data. We will show that a single unpolarized source observation per beam together with the known nominal element polarization can be used to calibrate each beam pair approximately. Given an uncalibrated beam pair , the rst step in the calibration procedure is to use the nominal polarization of array elements to nd the approximate Jones matrix. The signal correlation matrix has the block form (84) where and represent outputs from -polarized elements and -polarized elements, respectively. Neglecting estimation error, the matrices (85) are rank one and have principal eigenvectors and , respectively. These vectors are orthogonally rotated as needed to obtain and , where the hat indicates that these are only approximate responses to orthogonally polarized waves, whereas and in (5) are exact. These vectors can be used directly as beamformer weights (the biscalar method), but we will employ them here to obtain the approximate Jones matrix (86) where . This approximate Jones matrix can be used to calibrate the beam pair to obtain (87) and are only determined up to Since the eigenvectors a scale factor, it remains to specify the complex scaling of the beamformer weight vectors and in the beam pair . Assuming that the calibrator source is unpolarized, the magnitudes of the weight vectors can be xed by equalizing the responses to the calibrator source. The overall phases of the weight vectors are set by dividing by the phase of the largest weight vector element, so that the largest weights are real. This leads to the approximately calibrated beam pair (88a)

(88b) In practice, the phase lengths of the receiver signal paths will be different, but this phase can be measured using injected cali-

WARNICK et al.: POLARIMETRY WITH PHASED ARRAY ANTENNAS: THEORETICAL FRAMEWORK AND DEFINITIONS

193

bration tones or other means and incorporated in the phase constraint. The magnitude scaling in (88) ensures that the diagonal elements of are unity, which through (42) implies that the element of the Mueller matrix is equal to unity. This approximate single-source polarimetric calibration procedure is essentially equivalent to a polarimeter that uses the nominally orthogonally polarized outputs of a standard dual-polarized antenna (including a rotation from the astronomical coordinate system to the coordinate system of the antenna). As with a standard antenna, coupling between phased array elements and mechanical imperfections mean that the beam pair will not be exactly calibrated, and additional observations of polarized sources may be required to remove residual instrumental polarization effects. Numerical results will be given in the following section to assess the performance of this single-source polarimetric calibration method. V. NUMERICAL RESULTS The polarimetric gures of merit and beamforming algorithms will be illustrated for a 19 2 element hexagonal array of thin, lossless, - and -polarized crossed dipoles spaced apart and backed by a ground plane. The phased array feeds a 20-meter reector with . The open-circuited element response is approximated using the analytical expression for radiation by a dipole. Physical optics is used to compute secondary elds scattering from the reector. The array mutual impedance matrix is approximated by conservation of energy from the element pattern overlap integrals. The PAF noise model includes sky noise, spillover, and receiver noise due to low noise ampliers with parameters , and K. Ludwigs rst convention [7] is used to dene the and directions, which means that and are aligned with the array - and -coordinate frame. The array is calibrated using the approach of [20] using observations of an unpolarized calibrator source for each desired beam steering direction. Beamformer weights are computed using the eigenvector max-SNR and CFM algorithms described above, and polarimetrically calibrated using the approximate single-source method of Section IV.C. Results are given for two array congurations: (1) perfectly aligned elements and (2) perturbed element orientations with small random rotations away from the nominal - and -polarizations. Since only a small amount of power is received by elements more than one or two wavelengths away from the focal spot associated with a given formed beam, simulations of polarimetric accuracy for the 19 element case are expected to be representative of results for larger arrays. A. Perfectly Aligned Elements The rst study is a comparison of the sensitivity of the eigenvector max-SNR and CFM beam pairs to the optimal solution (Fig. 2). The single-source calibration procedure of Section IV.C is used with both beamformers. The sensitivity of the max-SNR beam pair calibrated with the single-source procedure is nearly indistinguishable from that of the optimal solution. For the polarimetrically uncalibrated case, the max-SNR beam pair sensitivity is different, but still lies within the upper and lower bounds over arbitrary linear transformations of the beam pair. The beam

Fig. 2. Beam pair sensitivity for dipole array with perfectly aligned elements for beam steering angles from the feed boresight to 2 half power beamwidths (HPBW) from boresight. The cut is at an azimuth angle of 30 from the -direction in the feed coordinate system. The sensitivity bound in (55) is indicated with gray shading.

Fig. 3. Beam center cross-polarization gures of merit as a function of steering angle for dipole array with perfectly aligned elements.

pair sensitivity bounds are given by (55) and indicated in the gure with gray shading. For some beam steering angles, the sensitivity of one of the uncalibrated beams is larger than that of the optimal beam pair. This illustrates that the calibration constraint in (65) leads to a beam pair with sensitivity slightly smaller than is achieved by other uncalibrated beam pairs in the subspace spanned by the optimal beam pair. The single-source calibrated CFM beam pair sensitivity also lies within its corresponding bound (55), but the overall sensitivity is lower than that of max-SNR. Cross-polarization gures of merit are shown in Fig. 3. Results are given for each beam center as a function of beam steering angle. For the beam steered in the boresight direction, the polarimetric calibration is essentially perfect, but for beams steered off-boresight, the gures of merit decrease. Results are shown for the polarization. Curves for the polarized beams are nearly identical. Error in measured Stokes parameters relative to the total source intensity for the single-source calibrated max-SNR and CFM beam pairs is shown in Fig. 4. The source Stokes parameters are tabulated values from the NRAO C Band VLA/VLBA

194

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 4. Relative Stokes parameter error (44) and instrumental polarization bound (47) as a function of steering angle for dipole array with perfectly aligned elements.

Fig. 6. Beam center cross-polarization gures of merit for dipole array with perturbed element orientations.

Fig. 5. Beam center sensitivity for dipole array with perturbed element orientations. The cut is at an azimuth angle of 30 from the -direction in the feed coordinate system. The sensitivity bound (55) is indicated with gray shading.

Fig. 7. Relative Stokes parameter error and instrumental polarization bound for dipole array with perturbed element orientations.

Polarization Calibration Database for Jy, . The error closely follows the instrumental polarization bound (47) for both beam pairs. It may be surprising that the perfectly aligned array, with no receiver gain imbalances, no mechanical defects, and exact analytical formulas for the element response, is only exactly calibrated for the boresight beam. For the boresight beam, the response vectors and in (7) are orthogonal, and the eigenvectors obtained from (75) are proportional to and . Thus, at boresight the eigenvector max-SNR and CFM beam pairs can be exactly calibrated without making use of the biscalar transformation (87). For off-boresight steered beams, due to the depolarizing effect of the reector, and are not orthogonal, and the eigenvector pair and for the perfectly aligned array require polarimetric calibration. The Square Kilometer Array (SKA) design target for relative instrumental polarization is 25 dB for general polarimetric imaging and 40 dB for specialized applications (see [21] and other available SKA documents). For the perfectly aligned array, the instrumental polarization is better than 40 dB only

for beam steering angles near boresight. For off-boresight beams, it is apparent that an additional polarimetric calibration beyond the approximate single-source method of Section IV.C is required. B. Imperfectly Aligned Elements We now study the effect of mechanical imperfections on array feed polarimetric performance. A rotation in the - plane was applied to the orientation of each element in the array, with rotation angle chosen from a zero-mean normal distribution with 5 standard deviation. Fig. 5 shows that the beam sensitivity is not reduced by the perturbation, but the upper and lower limits of the sensitivity bound (55) are no longer equal for the boresight beam. The cross-polarization gures of merit (Fig. 6) and Stokes instrumental polarization (Fig. 7) are signicantly degraded. The instrumental polarization is poorer than 25 dB for many beam steering angles. For observations that require better polarimetric accuracy, the nominal calibration approach presented in this paper is not adequate, and a further calibration step for each beam pair similar to the methods used for traditional single-pixel feeds would be required.

WARNICK et al.: POLARIMETRY WITH PHASED ARRAY ANTENNAS: THEORETICAL FRAMEWORK AND DEFINITIONS

195

VI. CONCLUSION A signal and noise model for a dual-polarized beamforming array receiver has been used to develop polarization gures of merit and a theory of ideal polarimetric beamforming. An optimal polarimetrically calibrated maximum-SNR beamforming algorithm was given. For each angle of arrival, the algorithm provides a beam pair with maximum sensitivity subject to a polarimetric calibration constraint. The optimal polarimetric beamforming algorithm requires exact knowledge of the responses of the array to orthogonally polarized waves from a given angle of arrival. In practice, the array responses to orthogonally polarized waves are not readily available. A practical eigenvector-based maximum-SNR beamformer solution has been shown to optimize sensitivity in the sense that when the beams are polarimetrically calibrated, in the absence of estimation error the beams become equivalent to the optimal solution. In general, the eigenvector maximum-SNR beam pair is not polarimetrically calibrated, and additional correction through estimation of the beam pair Jones matrix is required. An approximate single-source calibration procedure based on the approximate orthogonality of the array element polarizations was developed. The instrumental polarization of the approximately calibrated beams for a perturbed dipole phased array feed was better than 10 dB over the eld of view. The approximate single-source method therefore provides a rough calibration method for routine observations that do not require high polarimetric accuracy. In future work, more accurate polarimetric calibration methods for phased arrays should be studied. To achieve lower instrumental polarization than the single-source method used in this paper, standard polarimetric calibration techniques could be applied on a beam-by-beam basis to achieve lower instrumental polarization. To avoid time-consuming observations of multiple calibrator sources or long observations of a single polarized source for each formed beam, methods for reducing the number of required telescope pointings required to calibrate all formed beams are needed. Beam cross-polarization response patterns and the temporal stability of formed beam polarization responses are also of interest. APPENDIX Here we derive (56). By diagonalizing shown that is bounded by in (53), it can be (89) In view of (3) and (55), the sensitivity of any beam in the beam pair subspace is therefore bounded by the largest and smallest values of (90) Using the invariance of the trace with respect to the ordering of a product of two matrices, this can be expressed as (91)

Using (16),

becomes (92)

This leads directly to the bound (56). Because of the properties with the calibrated beam pair of the trace, replacing does not change the value of the trace:

(93) is invariant with respect which shows that the trace of to polarimetric calibration of the beam pair, and the bound in (55) is a polarimetric calibration-independent measure of the intrinsic beam pair sensitivity. REFERENCES
[1] S. G. Hay et al., Focal plane array development for ASKAP (Australian SKA Pathnder), presented at the 2nd Eur. Conf, on Antennas and Propagation, Edinburgh, U.K., Nov. 1116, 2007. [2] O. A. Iupikov, M. V. Ivashina, and O. M. Smirnov, Reducing the complexity of the beam calibration models of phased-array radio telescopes, presented at the Eur. Conf. on Antennas and Propagation (EuCAP), Rome, Italy, Apr. 1115, 2011. [3] B. Veidt, G. Hovey, T. Burgess, R. Smegal, R. Messing, A. Willis, A. Gray, and P. Dewdney, Demonstration of a dual-polarized phased-array feed, IEEE Trans. Antennas Propag., vol. 59, no. 6, pp. 20472057, Jun. 2011. [4] K. F. Warnick, D. Carter, T. Webb, J. Landon, M. Elmer, and B. D. Jeffs, Design and characterization of an active impedance matched low noise phased array feed, IEEE Trans. Antennas Propag., vol. 59, no. 6, pp. 18761885, Jun. 2011. [5] IEEE Standard Denitions of Terms for Radio Wave Propagation, IEEE Std. 211-1997, 1997. [6] J. P. Hamaker, J. D. Bregman, and R. J. Sault, Understanding radio polarimetryI. Mathematical foundations, Astron. Astrophys. Supplement Series, no. 117, pp. 137147, May 1996. [7] A. Ludwig, The denition of cross polarization, IEEE Trans. Antennas Propag., vol. 21, no. 1, pp. 116119, Jan. 1973. [8] K. F. Warnick and B. D. Jeffs, Efciencies and system temperature for a beamforming array, IEEE Antennas Wireless Propag. Lett., vol. 7, pp. 565568, 2008. [9] K. F. Warnick, B. Woestenburg, L. Belostotski, and P. Russer, Minimizing the noise penalty due to mutual coupling for a receiving array, IEEE Trans. Antennas Propag., vol. 57, no. 6, pp. 16341644, Jun. 2009. [10] M. V. Ivashina, R. Maaskant, and B. Woestenburg, Equivalent system representation to model the beam sensitivity of receiving antenna arrays, IEEE Antennas Wireless Propag. Lett., vol. 7, pp. 733737, 2008. [11] T. Carozzi, A fundamental gure of merit for radio polarimeters, IEEE Trans. Antennas Propag., vol. 59, no. 6, pp. 20582065, Jun. 2011. [12] O. Smirnov, Revisiting the radio interferometer measurement equation. I. A full-sky Jones formalism, Astron. Astrophys., vol. 527, 2011. [13] R. J. Sault, J. P. Hamaker, and J. D. Bregman, Understanding radio polarimetryII. Instrumental calibration of an interferometer array, Astron. Astrophys. Supplement Series, no. 117, pp. 149159, May 1996. [14] A. R. Thompson, J. M. Moran, and G. W. Swenson, Jr., Interferometry and Synthesis in Radio Astronomy, 2nd ed. Weinheim, Germany: Wiley-VCH Verlag GmbH & Co. KGaA, 2004. [15] H. C. v. d. Hulst, Light Scattering by Small Particles. New York: Dover Publications, 1981. [16] R. J. Sault and T. J. Cornwell, The Hamaker-Bregman-Sault measurement equation, in Proc. Synthesis Imaging in Radio Astronomy IIASP Conf. Series, 1999, vol. 180, pp. 657669. [17] R. Horn and C. Johnson, Topics in Matrix Analysis. Cambridge, U.K.: Cambridge Univ. Press, 1994. [18] J. Wishart, The generalised product moment distribution in samples from a normal multivariate population, Biometrika, vol. 20A, no. 12, pp. 3252, 1928.

196

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[19] H. L. Van Trees, Optimum Array Processing. New York: Wiley, 2002. [20] B. D. Jeffs, K. F. Warnick, J. Landon, J. Waldron, D. Jones, J. R. Fisher, and R. Norrod, Signal processing for phased array feeds in radio astronomical telescopes, IEEE J. Sel. Topics Signal Processing, vol. 2, no. 5, pp. 635646, Jul. 2008. [21] J. Lazzio, SKA Science Working Group, The square kilometre array design reference mission: SKA-mid and SKA-lo, [Online]. Available: www.skatelescope.org 2010

Karl F. Warnick (SM04) received the B.S. degree (magna cum laude) with University Honors and the Ph.D. degree from Brigham Young University (BYU), Provo, UT, in 1994 and 1997, respectively. From 1998 to 2000, he was a Postdoctoral Research Associate and Visiting Assistant Professor in the Center for Computational Electromagnetics at the University of Illinois at Urbana-Champaign. Since 2000, he has been a faculty member in the Department of Electrical and Computer Engineering at BYU, where he is currently a Professor. In 2005 and 2007, he was a Visiting Professor at the Technische Universitt Mnchen, Germany. He has published many scientic articles and conference papers on electromagnetic theory, numerical methods, remote sensing, antenna applications, phased arrays, biomedical devices, and inverse scattering, and is the coauthor of the book Problem Solving in Electromagnetics, Microwave Circuits, and Antenna Design for Communications Engineering (Artech House, 2006), and the author of Numerical Analysis for Electromagnetic Integral Equations (Artech House, 2008) and Numerical Methods for Engineering: An Introduction Using MATLAB and Computational Electromagnetics Examples (Scitech, 2010). Dr. Warnick was a recipient of the National Science Foundation Graduate Research Fellowship, Outstanding Faculty Member award for Electrical and Computer Engineering (2005), and the BYU Young Scholar Award (2007). He has served the Antennas and Propagation Society as a member of the Education Committee and as a session chair and special session organizer for the International Symposium on Antennas and Propagation and other meetings afliated with the Society. He is a frequent reviewer for the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION and Antennas and Wireless Propagation Letters. He has been a member of the Technical Program Committee for the International Symposium on Antennas and Propagation for several years and served as Technical Program Co-Chair for the Symposium in 2007.

(SKA). The results of these early PAF projects have led to the denition of APERTIFa PAF system that is being developed at ASTRON to replace the current horn feeds in the Westerbork Synthesis Radio Telescope (WSRT). She was involved in the development of APERTIF during 20082010 and acted as an external reviewer at the Preliminary Design Review of the Australian SKA Pathnder (ASKAP) in 2009. In 2002, she also stayed as a Visiting Scientist with the European Space Agency (ESA), ESTEC, in the Netherlands, where she studied multiple-beam array feeds for the satellite telecommunication system Large Deployable Antenna (LDA). She is currently a Senior Scientist at the Department of Earth and Space Sciences (Chalmers University of Technology). Her interests are wideband receiving arrays, antenna system modeling techniques, receiver noise characterization, signal processing for phased arrays, and radio astronomy. Dr. Ivashina received the URSI Young Scientists Award for GA URSI, Toronto, Canada (1999), APS/IEEE Travel Grant, Davos, Switzerland (2000), the 2nd Best Paper Award (Best team contribution) at the ESA Antenna Workshop (2008) and the International Qualication Fellowship of the VINNOVAMarie Curie Actions Program (2009) and The VR project grant of the Swedish Research Center (2010).

Stefan J. Wijnholds (S06M10) was born in The Netherlands in 1978. He received the M.Sc. degree in astronomy and the M.Eng. degree applied physics (both cum laude) from the University of Groningen, in 2003, and the Ph.D. degree (cum laude) from the Delft University of Technology, Delft, The Netherlands, in 2010. After graduation, he joined the R&D Department, ASTRON, The Netherlands Institute for Radio Astronomy, Dwingeloo, where he works with the system design and integration group on the development of the next generation of radio telescopes. His research interests lie in the area of array signal processing, specically calibration and imaging.

Marianna V. Ivashina (M00) received the Ph.D. degree in electrical engineering from the Sevastopol National Technical University (SNTU), Ukraine, in 2000. From 2001 to 2004, she was a Postdoctoral Researcher and, from 2004 to 2010, an Antenna System Scientist at The Netherlands Institute for Radio Astronomy (ASTRON). During this period, she carried out research on an innovative phased array feed (PAF) technology for a new-generation radio telescope, known as the square kilometer array

Rob Maaskant (M10) was born in the Netherlands on April, 14th, 1978. He received the M.Sc. (cum laude) and Ph.D. (cum laude, best dissertation of the Electrical Engineering Department) degrees in electrical engineering from the Eindhoven University of Technology, in 2003 and 2010, respectively. From 20032010 he was employed as an Antenna Research Scientist at the Netherlands Institute of Radio Astronomy (ASTRON). He is currently a Postdoctoral Researcher at the Chalmers University of Technology, Sweden. His current research interest is in the eld of receiving antennas for low-noise applications, meta-material based waveguides, and computational electromagnetics to solve these types of problems. Dr. Maaskant received a Rubicon Postdoctoral Fellowship from the Netherlands Organization for Scientic Research (NWO), in 2010. He won the 2nd Best Paper Prize (Best Team Contribution) at the 2008 ESA/ESTEC workshop, Noordwijk, and is the primary author of the CAESAR software; an advanced integral-equation based solver for the analysis of large antenna array systems.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

197

An Amplifying Recongurable Reectarray Antenna

Krishna Kumar Kishor, Student Member, IEEE, and Sean Victor Hum, Senior Member, IEEE
AbstractElectronically tunable reectarrays and amplifying reectarrays have attracted considerable research in recent years, however, little work has been done to combine these two features simultaneously in a single design. This paper focuses on the design of such a reectarray, which can be used as a high gain, recongurable transmitting antenna for communication links. The reectarray element is an aperture-coupled microstrip patch that accepts a linearly polarized wave, phase shifts and amplies the guided-waves in the transmission lines, and then re-radiates an orthogonally-polarized wave. First, the element design, modelling, stability analysis and experimental results are presented. Then a 48 element reectarray prototype operating at 5.7 GHz is described and its two dimensional beam steering capability and amplifying nature are successfully demonstrated and veried. Index TermsActive antennas, antenna arrays, beam steering, microstrip arrays, power combiners, reectarrays, varactors.

I. INTRODUCTION IGH gain, recongurable antennas are candidates for a variety of wireless applications. The use of high gain antennas in long distance communication links improves link gain and helps to suppress the interference from other devices or users in the vicinity of the terminal. The ability to recongure the radiation pattern of the antenna is also very desirable, especially if the antenna is to be deployed in a dynamic environment. Recongurable antennas nd applications in RADAR and satellite systems which often require scanning or recongurable shaped-beam radiation patterns. Reectarray antennas [1], with their large aperture size and spatial feeding mechanism, are strong candidates for some of these applications. Their advantages over the parabolic reectors and phased arrays are also well known [2]. They can be classied into the following four categories: (i) passive, xed (ii) passive, recongurable (iii) amplifying, xed and (iv) amplifying, recongurable. The fourth category is the focus of this paper. Passive xed reectarray antennas are the most pervasive in the literature, and a wide variety of design techniques have been proposed. These include those employing variable-sized patches [3], [4], stub-loaded patches [5], variable-sized slots in the ground plane [6], variable-length transmission lines in aperture-coupled designs [7], and so on. The desire to adaptively shape the beam patterns from passive xed reectarrays has
Manuscript received February 15, 2011; revised June 09, 2011; accepted July 25, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the Edward S. Rogers Sr. Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON M5S 3G4 Canada (e-mail: kishor@waves.utoronto.ca; svhum@ieee.org). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167939

shifted the focus to passive recongurable versions. These include reectarrays utilizing technologies like PIN diodes [8], varactor diodes [9], [10], MEMS switches [11][13] and liquid crystal materials [14] to recongure the phase of the scattered eld from each element. Amplifying xed reectarray antennas have also been demonstrated in literature [15], [16], where ampliers are integrated with individual reectarray elements to achieve power amplication from the reectarray surface. An interesting application of these antennas is in spatial power combining [17], where spatial feeds are used to distribute power to and collect power from an array of antennas connected to ampliers. However, most amplifying implementations of reectarrays do not have recongurable beam patterns. The combination of recongurability and amplication in reectarray antennas could make them suitable candidates for long range communication systems operating at millimeter-wave frequencies, where device power handling capabilities are limited. Such reectarray implementations have not been widely explored. Recently, such an antenna has been demonstrated using aperture-coupled patches at 5.8 GHz [18] using IQ modulators. However, a major drawback of this reectarray is its complexity in terms of its layout, the components used and the associated bias circuitry. This paper presents a less complicated reectarray prototype that is simultaneously capable of amplifying the incident elds and steering its far-eld beam in the azimuth and elevation directions. The organization of this paper is as follows. Section II presents a brief introduction to recongurable reectarray antennas, but focuses on the notion of stability in an amplifying reectarray element. Section III then covers the design of the reectarray element and presents its experimental results. Section IV describes the fabricated array and Section V discusses its measured performance. Lastly, the conclusions are stated in Section VI. II. BACKGROUND The principle of operation of a reectarray antenna is well documented [1]. The difference between xed and recongurable reectarrays is that in xed designs, the phase shift provided by each element is decided during the design of the array, and once the element is fabricated, it cannot be changed. However, in recongurable designs, the phase shift can be changed even after the fabrication of the element, and it is varied depending on the desired location of the main beam. There are two kinds of phase shifting mechanisms used in tunable reectarray elements: (i) integrated phase shifting, where the properties of the radiating structure (e.g. electrical length of a microstrip patch) are directly modied to obtain the desired phase shifts (for example [9]), and (ii) guided-wave phase shifting, where a space-wave transitions to a guided-wave along a transmission line and then the desired phase shifts are

0018-926X/$26.00 2011 IEEE

198

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 1. (a) A generic cross-polarized reectarray element. (b) Visualization of a cross-polarized reectarray element in signal ow form. Fig. 2. Complete signal ow graph in a cross-polarized, amplifying reectarray element.

obtained using a phase shifting network (for example [10]). In this work, the second approach was chosen, since it lends itself most readily for integration with microwave ampliers. Introducing an amplier in a reectarray element poses a new challenge in reectarray design: the need to prevent network instability and consequent oscillations. In an amplifying co-polarized element, where the incident and scattered elds are of the same polarization, the use of a reection-mode amplier is required. This design is difcult to make stable at the out-of-band frequencies due to the narrow impedance bandwidth of the reectarray elements. In-band instability can also occur if the phase shifter is not designed carefully to maintain a good impedance match while the phase is tuned. Given the challenging requirements in achieving a stable design, co-polarized elements were considered unsuitable for this work. On the other hand, in an amplifying cross-polarized reectarray element, where the incident and scattered elds are orthogonally-polarized, stability can be easily attained because of the natural isolation between the two polarization ports on the radiating structure. A schematic representation of a generic reectarray element is shown in Fig. 1(a), and it is represented in signal ow form in Fig. 1(b). As seen from these gures, the reectarray element consists of a radiating structure (for example, a dual-polarized aperture-coupled microstrip patch) and an amplifying and phase shifting (APS) circuit, which is connected in a path between the two ports of the radiating structure. A small-signal analysis of the signal ow in Fig. 1(b) requires the denition of four reection coefcients: , , and , similar to linear amplier design [19]. The necessary and sufcient criteria for stability can therefore be dened as follows [16]: (1) (2) Expressions for the above four reection coefcients can be computed using the corresponding signal ow graph shown in Fig. 2 [16]. In Fig. 2, the and terms represent the incident and reected voltage travelling waves, while and represent the two-port S-parameters of the radiating structure and the APS circuit, respectively. The signal ow graph shown in Fig. 2 shows all possible paths of signal ow in the reectarray element, including those that can lead to instability. Note that mutual coupling between the

reectarray elements has not been included in this analysis, but several antenna elements could be represented using a multiport network which includes the coupling if desired. However, for the design shown in this work, mutual coupling was low enough between elements such that it did not affect the stability analysis. If we assume that the input to the APS circuit and the second port of the radiating structure are impedance-matched (that is, ), the signal ow graph in Fig. 2 can be simplied to contain a single dominant feedback path. This path is shown in Fig. 2 using thick solid lines. The stability criteria in (1) and (2) reduces to (3) which states that the small-signal gain provided by the amplier is limited by the amount of isolation between the two polarization ports. Thus in the proposed reectarray element, attempts were made to impedance match the input of the APS circuit and the second port of the radiating structure, as well as maximize the isolation between the two ports. This ensured the use of (3) as a valid stability criteria. III. PROPOSED REFLECTARRAY ELEMENT Fig. 3(a) depicts a three dimensional view of the proposed multilayered reectarray element. A top view of the element is also shown in Fig. 3(b). The principle of operation of the proposed element can be summarized as follows: a -polarized wave incident on the microstrip patch couples into the input terminals of the phase shifter through slot 1. An electronicallycontrolled phase shift is then applied to the signal by a tunable reection-type phase shifter, followed by its amplication using a two-port amplier operating in the small-signal regime. The phase-shifted and amplied signal is coupled to the microstrip patch through slot 2 and radiates out as an -polarized wave. Fig. 3(b) shows that the two orthogonal slots in the ground plane are offset from the center of the patch only in the -direction ( and ) and not in the -direction. Placing the slots in such a fashion maintains the symmetry of the currents along the -direction (perpendicular to the polarization of the scattered wave), thereby increasing the isolation between the two slots [20]. In this design about 30 dB of isolation was observed in simulations. The - and -dimensions of the patch ( and ), the

KISHOR AND HUM: AN AMPLIFYING RECONFIGURABLE REFLECTARRAY ANTENNA

199

Fig. 4. Schematic of the phase shifter along with the tunable loads. Fig. 3. Proposed reectarray element. (a) 3D view (b) Top view.

TABLE I SPECIFICATIONS OF THE PROPOSED REFLECTARRAY ELEMENT

Fig. 5. Four-port model used in studying the ARRE. (a) 3D view and (b) top view.

lengths of the two slots ( and ), the -offset of the slots ( and ) and the lengths of the stubs ( and ) were optimized such that there was optimum coupling between the 50 transmission lines at the bottom of the circuit and the patch-slot-stub networks. Table I lists the optimized values of the various dimensions of the reectarray element indicated in Fig. 3. A. Phase Shifter Design Fig. 4 shows a schematic of the quadrature hybrid-based, electronically controlled reection-type phase shifter used in the reectarray element. An integrated circuit phase shifter, though smaller, was not chosen since the performance of the phase shifter was not a priority. The goal of the investigation was to demonstrate the amplifying recongurable reectarray concept, and hence it was desirable to have to phase shifter that could be easily and cheaply integrated with the reectarray element. The coupled and through ports of the quadrature hybrid were terminated in two identical reective tunable loads, each of which to any signal coming out of presents a reection coefcient these ports [21]. The tunable load consisted of two arms, where the top arm was connected to a grounded varactor diode, while the bottom arm consisted of a quarter wavelength transmission terminated with an identical line of characteristic impedance grounded varactor diode. This multi-pole load network provides over 360 of phase agility [22], [23]. The phase shifter was fabricated on a Rogers 3006 substrate and height of 1.28 mm. MGV100-20 hyperwith abrupt GaAs varactor diodes from Aeroex were used, and was chosen to be 50 . 4.7 nH inductors from Johanson Technology were used as RF chokes to prevent the leakage of RF

currents to the bias lines (seen in Fig. 7). Note that the bias lines were located at the back of the reectarray cell, and therefore did not disrupt the radiation pattern of the reectarray element. Measurement results at 5.7 GHz, the design frequency of the reectarray, indicated an insertion loss variation between 2.4 dB and 8 dB, a variation between 8.6 dB and 20 dB, and 360 of phase shifting performance [24]. The high insertion loss variation of 5.6 dB is attributed to 5.7 GHz not being the optimal frequency of operation of the phase shifter (which is 5.4 GHz [24]). This non-optimal operation of the phase shifter, as will be seen in Section V, leads to slightly larger losses in the array operating at 5.7 GHz, but the losses were deemed acceptable for exploring the behavior of the full array discussed in Section IV. B. Network Modelling In general, predicting the performance of a recongurable reectarray element integrated with an amplier requires non-linear circuit simulations, thereby making it challenging to simulate the proposed amplifying recongurable reectarray element (ARRE) in a full-wave simulator. To overcome this problem, it is treated as a combination of a passive, linear four-port network and a two-port, potentially non-linear, network representing the amplier. 3-D and top views of the simulated four-port network enclosed in a periodic waveguide are shown in Fig. 5. Ports 1 and 4 are spatial ports, whereas ports 2 and 3 are located at the ends of the transmission lines. The complete four-port S-parameters can be computed from the normalized incident and reected voltage travelling waves calculated at each port. For example, if port 1 is excited with a plane wave, then , and can be obtained. To calculate and , the total voltage and current are measured at the two ports and

200

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 6. Schematic of the ARRE to understand the effect of changing the transmission line length on the magnitude of the outgoing signal. Fig. 7. Bottom view of the fabricated reectarray element.

(4) or 3 is utilized. and refer to the total voltage where and current measured at port . is the port impedance at ports 2 and 3, and for the spatial ports 1 and 4. In a similar fashion, the remaining S-parameters in the fourport model are computed. As the four-port network includes the phase shifter, the four-port S-parameter matrix has to be computed for different values of the varactor diode control voltage, in order to characterize the behavior of the amplifying cell as a function of the control voltage (we note that a six-port linear circuit can also be characterized so that a more intricate circuit model of the diode can be substituted for simple lumped elements used in the full-wave four-port simulation). In this work, the four-port S-parameters were obtained using SEMCAD X [25]. The four-port network, along with the two-port network of the amplier placed between ports 2 and 3, were simulated in a circuit simulator (Agilent ADS) to yield the overall two-port S-parameters of the reectarray element. For this rst study, a small-signal LEE-39+ linear amplier from Mini-Circuits, with gain approximately equal to 14 dB at 5.7 GHz was used, allowing the small-signal S-parameters to be used in the analysis. Of course, non-linear device models can also be used and simulated. It is important to minimize the amplitude variation of the scattered eld as the phase is tuned to different values. To achieve this, the signal ow in the ARRE was studied. Fig. 6 shows a schematic representation of the components of the reectarray element. Of particular interest is the transmission line of length connecting the output of the amplier to the second slot of the patch, since it directly inuences the amplitude variation of the cell. Note that the locations and orientations of the slots in Fig. 6 are shown for descriptive purposes only, and do not reect their actual location. In addition to the main signal ow (described previously in Section III), the cross-talk between the . two slots is also represented as a transmission coefcient, and (equal to ) are the transmisSimilarly, sion gains through the phase shifter and the amplier-transmission line network respectively. Thus in the chosen element conguration where the amplier and phase shifter are placed in a closed signal loop, the latter controls the phase of the loop gain and hence the closed-loop gain produced by the reectarray element. This variation is minimized by an appropriate choice of the length , which provides another degree of

Fig. 8. Fabricated ARRE behavior. (a) and , indicating the levels , indicating phase of the co- and cross-polarized signals, respectively, (b) agility of the design.

freedom to control the phase of the loop gain. The following analysis shows this more clearly. (5) where

(6) Since ideally , (7) Equation (7) therefore indicates that , and therefore , uctuates as the bias voltage is varied, and it can be minimized with a proper choice of , or equivalently, . Using the known four-port S-parameters of the ARRE (see Fig. 5) and the S-parameters of the amplier, the optimum length of the transmission line, which minimized the variation to 3.1 dB, was found to be 12.4 mm. A wrong choice of the line length, for example 30 mm, could lead to a worst case variation of 4.8 dB. C. Measurement Results Fig. 7 shows a bottom view of the fabricated reectarray element. From here, the phase shifter used in the reectarray element is seen along with the amplifying part of the circuit. The ARRE was measured inside a C-band ortho-mode transducer (OMT) to simulate an innite array scenario at an off-broadside illumination angle [26]. The cross-sectional waveguide dimensions of the OMT were 33.56 33.56 mm. The use of the and OMT allowed the measurement of the co-polarized

KISHOR AND HUM: AN AMPLIFYING RECONFIGURABLE REFLECTARRAY ANTENNA

201

TABLE II SPECIFICATIONS OF THE FABRICATED REFLECTARRAY ANTENNA

Fig. 10. Reectarray fed in an offset-fed conguration.

Fig. 9. Fabricated reectarray antenna. (a) Top view. (b) Bottom view.

cross-polarized waves reected from the ARRE. The results obtained at 5.7 GHz are plotted in Fig. 8. Note that these results include the mutual coupling between the reectarray elements. A phase agility of 362 was observed and the gain was found to vary between 5.6 dB and 11.3 dB. The 5.7 dB gain variation is quite signicant and as explained in [24], is due to a shift in frequency of operation from simulations. Further attempts to address the gain variation were not made, since the goal of the project was to demonstrate the amplifying recongurable reectarray concept. The magnitude of the co-polarized is quite high, but is well below the signal reected back desired amplied cross-polarized signal . IV. REFLECTARRAY DESIGN AND EXPERIMENTAL SETUP The fabricated 8 6 array is shown in Fig. 9, and its specications are outlined in Table II. The array of microstrip patches and the amplifying and phase shifting circuits were fabricated on Rogers 3203 and Rogers 3006 substrates respectively, and then held together using nylon screws. The fabricated array is intended to represent a small segment of a larger reectarray and its size was chosen to minimize cost, while producing sufcient gain to be measured using a near-eld antenna scanner. The 48 elements in the array can be visualized as two 4 6 subarrays. Each subarray has 24 control lines for the varactor diodes and two main bias lines for the four rows of ampliers.

These are indicated in Fig. 9(b). Each of the 24 control lines is connected to a 40 pin ribbon cable connector, which is then linked to a voltage controller through a routing board. The voltage controller was designed in-house, and consists of a microcontroller and an array of digital-to-analog converters, leading to 24 independent outputs. In this way, the varactor diode bias voltage for every phase shifter of the array is controlled by a unique output pin of the controller. The layout of each element on the array is very similar to the tested unit cell, except that a common connection pad is used to connect the anode of the varactor diodes to ground. In the array, the MGV100-21 hyperabrupt GaAs varactors diodes from Aeroex and the LEE-39+ ampliers from Mini-Circuits were used. The reectarray was fed in an offset conguration as shown in Fig. 10, using a pyramidal feed horn having an aperture size of 140 130 mm. The offset-fed conguration was required to minimize the feed horn blockage of the broadside radiation from the relatively small-sized reector. The angle of incidence on the array was 15 and the distance between the aperture of the horn and the array was 41 cm. The incident elds from the feed horn on the array were measured using a near-eld scanner and an average edge taper of 9.2 dB and 5.7 dB were obtained at the reector edges in the E-plane and H-plane of the feed, respectively. The resulting spillover , taper and illumination efciencies were calculated to be 56.61%, 87.6% and 49.6% respectively. For optimal beamforming, the phase versus voltage curve for each element was experimentally measured using a near-eld sensing technique. This provides an accurate picture of the cell in an array layout and also accounts for inter-cell variations due to manufacturing. After removing the effect of the cables through calibration, the measurement was done by placing the OMT, which served as the transmitter and the receiver, about 3 mm from the radiating microstrip patch of the element-under-

202

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 11. Measurement setup for characterizing elements on the array.

Fig. 13. Horizontal cut of co-pol (solid line) and cross-pol (dashed line) data and various azimuth angles. for beams at

Fig. 12. Coordinate systems used during measurement of the reectarray antenna.

test, as shown in Fig. 11. For de-embedding calculations, the OMT was assumed to be lossless. Characterization of the elements on the array revealed that the average phase agility of the reectarray elements had reduced to 248.52 , with a standard deviation of 23.05 . Potential causes for this include the use of the slightly different MGV100-21 varactor diodes on the array (which causes a shift in the phasing characteristic of the element), errors introduced due to alignment offset between the two boards and possible air gaps introduced between the two boards. Nevertheless, sufcient phase agility was achieved to enable beamforming experiments which are discussed next. V. ARRAY RESULTS This section demonstrates the working of the reectarray antenna, by rst showcasing its beam steering capability, and then calculating the amount of gain provided by the array to verify its amplifying nature. The azimuth-elevation and spherical coordinate systems used for measurements and beamforming are shown in Fig. 12. Fig. 13 shows the horizontal cut of the beam (in solid lines) when it was steered in the azimuth direction at an elevation angle of 0 . The beam-pointing angles were , ,0 , 10 and 20 . Similarly, in Fig. 14, the vertical cut of the beam when it was steered in the elevation direction at an azimuth angle of 0 is shown in solid lines. The beam-pointing angles in this case were 0 , 5 , 10 and 20 . The small reector size along with the offset feed conguration produces signicant

Fig. 14. Vertical cut of co-pol (solid line) and cross-pol (dashed line) data for and various elevation angles. beams at

feed blockage at negative elevation angles, therefore measurements were not conducted in this region. Figs. 13 and 14 show that the array is capable of beamforming in both the azimuth and elevation planes. Lower antenna gains due to feed blockage are observed when comparing the beam at an elevation angle of 0 with that at 20 . Figs. 13 and 14 also indicate that there is greater than 13 dB difference between the co-polarized (solid lines) and cross-polarized (dashed lines) signal levels, which is reasonable considering the small size of the array. Figs. 15 and 16 plot the measured beams, in the azimuth and elevation directions respectively, with those calculated using array theory (cases A, B, C). The amplitude and phase distributions used in the three cases are summarized in Table III. In case A, a tapered amplitude distribution (measured) due to the feed horn was utilized and the reectarray elements were assumed to capable of providing the exact phase shifts required by array theory. This amplitude distribution therefore assumes that no insertion loss variation due to the reectarray elements is present in the array. In case B, the amplitude distribution on the array was obtained by holographically back-projecting the measured elds onto the aperture surface. The same phase distribution from case A was employed. In case C, the same amplitude distribution from case B was used. However, a more realistic phase distribution on the array surface was employed, based

KISHOR AND HUM: AN AMPLIFYING RECONFIGURABLE REFLECTARRAY ANTENNA

203

Fig. 15. Comparison of measured and calculated far-eld patterns at for the various cases listed in Table III. (a) , (b) , (d) .

, (c)

prominent nulls and errors in the location of the nulls in the measured pattern. Mutual coupling between the reectarray elements, which can lead to amplitude and phase variations in the array, and feed blockage effects have not been accounted for in the three cases plotted in Figs. 15 and 16. A comparison of cases A and B shows the dominant inuence of the insertion loss variation from the reectarray elements on the radiation pattern, the other factors being the element pattern and mutual coupling between the reectarray elements. Next, the net gain added by the amplier was determined to verify the amplifying nature of the reectarray. Ideally, to calculate the added gain provided by the ampliers, the gain from a passive reectarray of the same size would need to be measured. As building and testing the equivalent passive antenna was beyond the scope of this project, the gain from the passive reectarray was computed theoretically. Though this method does not provide an accurate measure of the gain provided by the ampliers, the purpose was to show that the contribution of the amplier to the measured gain of the amplifying reectarray was signicantly greater than unity (0 dB). For the calculations presented next, the beam pointing at and was chosen as the reference. This was done with the assumption that the measured results at this beam angle were devoid of any feed blockage effects. Mathematically, we can state the following in dB: (8) where , the gain of the fabricated antenna when compared to a standard gain horn was known to be 24.6 dBi, , the gain from the equivalent passive reectarray, can be determined as explained next, and is the average closed-loop amplier gain (that is, the average gain from all the ampliers on the array when they are placed in the feedback conguration). is the quantity of interest. can be computed as

Fig. 16. Comparison of measured and calculated far-eld patterns at for the various cases listed in Table III. (a) , (b) , (c) , (d) .

(9) where is the directivity of the passive reectarray and is the total efciency of the passive reectarray expressed on a dB scale. The directivity of the passive reectarray was computed from array theory using the realistic phase distribution (used previously to obtain the case C curves in Figs. 15 and 16) and a uniform amplitude distribution. This was found to be 23.6 dBi. The total efciency of the passive reectarray has several contributing factors, and on a dB scale, it can be computed as the sum of these factors: (10) The various efciency factors are described in Table IV. The efciency factors , and were quite difcult to determine without rigorous analysis and they were conservatively estimated to contribute about 1 dB of loss to the reectarray. As mentioned earlier, for the beam under consideration, the feed did not produce any signicant blockage. The illumination efciency for the particular conguration of the feed horn

TABLE III AMPLITUDE AND PHASE DISTRIBUTIONS USED FOR CASES A, B, C IN Figs. 15 AND 16

on the phase lookup tables obtained from the element measurements discussed in Section IV. Figs. 15 and 16 show that the location of the main beam of the measured radiation pattern is in close agreement with case C. The lack of phase agility from the reectarray elements and phase errors on the reectarray surface explain the lack of

204

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE IV GAIN CALCULATION OF THE AMPLIFYING REFLECTARRAY

Fig. 17. Gain of the fabricated reectarray versus frequency.

and the reectarray, was presented in Section IV, and was equal to 3.1 dB. indicates the losses due to reection from the elements in the array, since not all the power from the feed horn was captured by the elements for amplication. The amount of power captured by an element is given by , and it depends on the reverse bias voltage applied on the phase shifter of the element. The bias voltage applied to every element of the array was known and the variation of was also known from the simulated passive cell in SEMCAD. Therefore the amount of power captured by all the elements on the array was computed by taking a spatial average given by: (11) is the number of elements. where The contribution from the losses in the cell, given by , was computed by taking a spatial average of the insertion gain through the phase shifter and transmission lines. This was also a function of the bias voltage applied to the reectarray element, and was determined using (12) Table IV summarizes all the efciency factors obtained. Also shown is the gain the passive array calculated using (9). Knowing and , (8) gives the average closed-loop gain provided by the ampliers, which was calculated to be 10.0 dB. For verifying this average closed-loop gain from the ampliers, the expected closed-loop gain was calculated from the spatial average of the from the reectarray element presented in Section III-C. This was found to be 10.9 dB, which was within a dB of the calculated value. It is worthwhile to remember that the closed-loop gain calculated here is not expected to match the 14 dB open-loop gain from the amplier. It is also important to note that the average closed-loop gain calculated here is an estimate. Errors could be introduced in the calculation of

Fig. 18. Gain of the fabricated reectarray versus input power.

, and , due to the use of the reectarray element results presented in Section III and the theoretical directivity of the passive array. However, taking the spatial average should reduce these errors. Thus it can be concluded that the fabricated reectarray provides an amplication, in addition to recongurability of the beam pattern. The gain of the reectarray is plotted as a function of frequency in Fig. 17, which indicates a 1-dB gain bandwidth of 3.9% is achieved. The small bandwidth is expected, since it is limited by the bandwidth of the reectarray elements. Lastly, if the proposed reectarray is to be used as a transmitter, its large-signal performance is of interest. In Fig. 18 the gain of the reectarray at 5.7 GHz is plotted as a function of the power input into the feed horn. This indicates that the 1 dB compression point occurs at approximately 27.5 dBm of input power. It should be noted that the proposed reectarray was not optimized for power efciency as it used commercial small-signal ampliers in its design. However, future implementations could be designed around large signal device models and optimized for output power and power-added efciency. VI. CONCLUSION In this paper an amplifying recongurable reectarray antenna and the associated element has been presented. The design, modelling and experimental results of the reectarray element that is capable of amplication and providing 362 of phase agility were discussed. Two dimensional beam steering from 0 to 20 degrees in the elevation and 20 to 20 degrees

KISHOR AND HUM: AN AMPLIFYING RECONFIGURABLE REFLECTARRAY ANTENNA

205

in azimuth directions was demonstrated at 5.7 GHz. The average closed-loop gain from the array was then calculated to be 10.0 dB, thereby verifying its amplifying nature. This work has shown that adding an amplier requires a careful design of the reectarray element. Simulations of such a structure require the use of both full-wave and circuit simulators. Moreover, care must be taken during the assembly of the array, as oscillations can be easily induced otherwise. Some future work includes the identifying the cause of the loss in phase agility from the reectarray element in the array layout, minimizing the amount of co-polarized signal reected from the reectarray element and also characterizing the non-linear response of the element for high-power transmitter-type applications. The capabilities of the reectarray antenna discussed in this paper has not been presented previously in literature. The proposed reectarray antenna has demonstrated that amplication and 2-D beam steering can be achieved using a simple, low-prole, single reector-based antenna. The array also holds signicant promise as a high gain antenna and a spatial power combiner at millimeter-wave frequencies, where power handling limitations constrain the amount of power from a single amplier. REFERENCES [1] J. Huang and J. A. Encinar, Reectarray Antennas. Hoboken, NJ: Wiley-IEEE Press, 2007. [2] M. Bialkowski and J. Encinar, Reectarrays: Potentials and challenges, in Proc. Int. Conf. on Electromagnetics in Advanced Applications (ICEAA), Sep. 2007, pp. 10501053. [3] D. Pozar, S. Targonski, and H. Syrigos, Design of millimeter wave microstrip reectarrays, IEEE Trans. Antennas Propag., vol. 45, no. 2, pp. 287296, Feb. 1997. [4] J. Encinar, Design of two-layer printed reectarrays using patches of variable size, IEEE Trans. Antennas Propag., vol. 49, no. 10, pp. 14031410, Oct. 2001. [5] D.-C. Chang and M.-C. Huang, Multiple-polarization microstrip reectarray antenna with high efciency and low cross-polarization, IEEE Trans. Antennas Propag., vol. 43, no. 8, pp. 829834, Aug. 1995. [6] M. R. Chaharmir, J. Shaker, M. Cuhaci, and A. Sebak, Reectarray with variable slots on ground plane, IET Microw. Antennas Propag., vol. 150, no. 6, pp. 436439, Dec. 2003. [7] A. Bhattacharyya, Slot-Coupled Patch Reect Array Element for Enhanced Gain-Band Width Performance, U.S. Patent 6388620, May 2002. [8] M. Barba, E. Carrasco, J. E. Page, and J. A. Encinar, Electronic controllable reectarray elements in X band, Frequenz, vol. 61, no. 910, pp. 203206, 2007. [9] S. V. Hum, M. Okoniewski, and R. J. Davies, Modeling and design of electronically tunable reectarrays, IEEE Trans. Antennas Propag., vol. 55, no. 8, pp. 22002210, Aug. 2007. [10] M. Riel and J.-J. Laurin, Design of an electronically beam scanning reectarray using aperture-coupled elements, IEEE Trans. Antennas Propag., vol. 55, no. 5, pp. 12601266, May 2007. [11] S. V. Hum, G. McFeetors, and M. Okoniewski, Integrated MEMS reectarray elements, in Proc. 1st Eur. Conf. on Antennas and Propagation, EuCAP, Nov. 2006, pp. 16. [12] H. Rajagopalan, Y. Rahmat-Samii, and W. Imbriale, RF MEMS actuated recongurable reectarray patch-slot element, IEEE Trans. Antennas Propag., vol. 56, no. 12, pp. 36893699, Dec. 2008. [13] S. Montori, L. Marcaccioli, R. Gatti, and R. Sorrentino, Constantphase dual polarization MEMS-based elementary cell for electronic steerable reectarrays, in Proc. Eur.Microwave Conf., EuMC, Sep. 2009, pp. 033036. [14] A. Gaebler, A. Moessinger, and F. Goelden et al., Liquid crystal-recongurable antenna concepts for space applications at microwave and millimeter waves, Int. J. Antennas Propag., vol. 2009, 2009, Article ID 876989, 7 pages.

[15] M. E. Bialkowski, A. W. Robinson, and H. J. Song, Design, development, and testing of X-band amplifying reectarrays, IEEE Trans. Antennas Propag., vol. 50, no. 8, pp. 10651076, Aug. 2002. [16] R. W. Clark, G. H. Huff, and J. T. Bernhard, An integrated active microstrip reectarray element with an internal amplier, IEEE Trans. Antennas Propag., vol. 51, no. 5, pp. 993999, May 2003. [17] J. A. Navarro and K. Chang, Integrated Active Antennas and Spatial Power Combining. New York: Wiley, 1996. [18] L. Cabria, J. . Garca, J. Gutirrez-Ros, A. Tazn, and J. Vassallo, Active reectors possible solutions based on reectarrays and Fresnel reectors, Int. J. Antennas Propag., vol. 2009, 2009, Article ID 653952, 13 pages. [19] G. Gonzalez, Microwave Transistor Ampliers, Analysis and Design, 2nd ed. Englewood Cliffs, NJ: Prentice Hall, 1997. [20] S.-C. Gao, L.-W. Li, M.-S. Leong, and T.-S. Yeo, Dual-polarized slotcoupled planar antenna with wide bandwidth, IEEE Trans. Antennas Propag., vol. 51, no. 3, pp. 441448, Mar. 2003. [21] R. H. Hardin, E. J. Downey, and J. Munushian, Electronically variable phase shifters utilizing variable capacitance diodes, Proc. IRE, vol. 48, pp. 944945, May 1960. [22] R. Garver, 360 varactor linear phase modulator, IEEE Trans. Microw. Theory Tech., vol. 17, no. 3, pp. 137147, Mar. 1969. [23] J. Upshur and B. Geller, Low-loss 360 X-band analog phase shifter, in IEEE MTT-S Int. Microwave Symp. Digest, May 1990, vol. 1, pp. 487490. [24] K. Kishor and S. Hum, An amplifying recongurable reectarray element, in Proc. IEEE Antennas and Propagation Society Int. Symp., Jul. 1117, 2010, p. 14. [25] SEMCAD X Reference Manual Schmid & Partner Engineering AG, Zurich, Switzerland, May 2009 [Online]. Available: http://www.speag.com [26] P. Hannan and M. Balfour, Simulation of a phased-array antenna in waveguide, IEEE Trans. Antennas Propag., vol. 13, no. 3, pp. 342353, May 1965.

Krishna Kumar Kishor (S10) was born in Thrissur, Kerala, India. He received the B.A.Sc. and M.A.Sc. degrees in electrical engineering from the University of Toronto, Toronto, ON, Canada, in 2007 and 2010, respectively, where he is working toward the Ph.D. degree under Prof. Sean Victor Hum. He is interested in antenna design, particularly of recongurable antennas. He has worked on reectarray antennas, and is currently developing recongurable antennas for modern wireless communication systems.

Sean Victor Hum (S95M03SM11) was born in Calgary, AB, Canada. He received the B.Sc., M.Sc., and Ph.D. degrees from the University of Calgary, in 1999, 2001, and 2006 respectively. In 2006, he joined the Edward S. Rogers Sr. Department of Electrical and Computer Engineering, University of Calgary, where he currently serves as an Assistant Professor. His present research interests lie in the area of recongurable RF antennas and systems, antenna arrays, and ultrawideband communications. Prof. Hum received the Governor Generals Gold Medal for his work on radio-on-ber systems in 2001. In 2004, he received an IEEE Antennas and Propagation Society Student Paper award for his work on electronically tunable reectarrays. In 2006, he received an ASTech Leaders of Tomorrow award for his work in this area. On the teaching side, he has received three University of Toronto Departmental Teaching Awards since 2007, and most recently, an Early Career Teaching Award in 2011. He served on the steering committee and technical program committee for the 2010 IEEE AP-S International Symposium on Antennas and Propagation. In August 2010, he was appointed as an Associate Editor for the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION.

206

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Design of Retrodirective Antenna Arrays for Short-Range Wireless Power Transmission

Ying Li, Member, IEEE, and Vikram Jandhyala, Senior Member, IEEE
AbstractApplication of retrodirective antenna arrays in wireless power transmission is proposed in this paper. The time reversal or phase conjugate theory and design considerations of the array factors are examined in GHz regime and meter range. The wireless power transmission efciency is analyzed in the 2-D case with fullwave electromagnetic (EM) solution and validated by analytical array theory. The design factors of array size, array spacing, element size, and specic choice of the industrial, scientic and medical (ISM) band frequency are investigated to optimize the transmission focus. Design guidelines are provided for those factors and the impact of manufacturing variability is discussed. The proposed approach can also be extended to wireless power transfer with active retrodirective rectenna array and passive RFID systems. Index TermsRetrodirective array, time reversal, variability design, wire antenna array, wireless power transmission.

I. INTRODUCTION

ISTORICALLY, Nikola Tesla rst proposed theories of wireless power transmission in the late 1800s and early 1900s and remotely powered lights in the ground at his Colorado Springs experiment station [1]. Wireless power systems are slowly emerging in a variety of applications such as passive radio frequency identication (RFID) systems, inductively coupled methods for recharging appliances, and resonance or propagating wave structures. Inductive coupling relies on magnetic eld and current induction between two coils, and has been applied commercially in, for example, electric toothbrushes, and appliance charging mats [2]. As a drawback, inductive coupling requires the coils to be very close to the object under charge for successful induction. Soljacic and Joannopoulos proposed resonance coils of the same resonant frequency to transfer wireless power over a distance of meters [3]. Macroscopic resonant coupling needs large coils operating at MHz range and proper placement. In this work, we propose a method of efcient wireless power transfer using retrodirective arrays. This method is effective in the range of meters, but does not require large resonators and sensitive alignments. A retrodirective array transmits a signal
Manuscript received December 01, 2009; manuscript revised September 13, 2010; accepted June 20, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the Department of Electrical Engineering, University of Washington, Seattle, WA 98195 USA (e-mail: liying@uw.edu; vj@uw.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167897

back to the interrogators position without any a priori knowledge of the incoming angle, and without relying on sophisticated digital signal processing algorithms. Its reective property is realized by manipulating the phases of the array elements, which is equivalent to reversal of incoming waves in the time domain, or phase conjugation in the frequency domain. The theory of time reversal methods is originally derived from acoustics [4], and has then been applied to electromagnetics (EM) [5] and antennas. Retrodirective arrays have been used commonly in applications such as collision avoidance systems, microwave imaging or detection, and RFID systems. Well-known retrodirective antenna structures include the corner reector, the Van Atta array [6][8] and the heterodyne retrodirective array that uses an RF circuit for phase conjugation [9]. The efciency of the wireless power system in the meter range is greatly increased because the reradiated eld focuses in the incoming eld direction due to the phase conjugate relationship, regardless of the positions of the devices to be charged. The focus of the retrodirective array power transmission can be further improved by appropriate design optimization of the array size, operating frequency, and element size under given constraints. In this paper, we investigate the design choices to achieve desired focus and examine the constraints, and theoretical and computational limits of the power transmission efciency. While there has been signicant work on the design of reectarray radiation elements, from simple dipole wire antenna [10], [11] to more complicated rectennas [12], [13], the emphasis of this focused work is on array design. In Section II, we describe the structure of the proposed wireless power system, and summarize time reversal theory and retrodirective arrays. Then, we illustrate the transmission efciency dependence on the design factors using a 2-D dipole antenna retrodirective array. The results are computed with full-wave methods of moment (MOM), and also validated with analytical methods of dipole antenna and array theory. The retrodirective behavior is validated by radar cross section (RCS). The transmission efciency results are shown with different industrial, scientic and medical (ISM) band frequencies, geometric parameters, and the variability of the geometric parameters. II. THEORY We aim to charge electronic devices such as cell phones, laptops, personal digital assistants (PDAs), etc., wirelessly in a typical ofce or living room. The wireless power source is placed at a xed location of the room, for example, on the ceiling or side walls. The devices to be charged are placed randomly at different places by the user. The devices can also be moved as

0018-926X/$26.00 2011 IEEE

LI AND JANDHYALA: DESIGN OF RETRODIRECTIVE ANTENNA ARRAYS FOR SHORT-RANGE WIRELESS POWER TRANSMISSION

207

Fig. 1. Illustration of the wireless charging scheme in a room with retrodirective array.

long as they stay within the room. The arrangement is illustrated in Fig. 1. The power source is an appropriately designed retrodirective antenna array. In this gure, an active Van Atta array with a bi-directional amplier is used for reecting and amplifying the transmitted radiation signal from a device which is incident onto the array. Firstly in stage 1, this transmitted radiation signal is sent by an electronic device, and then in stage 2 the retrodirective array receives the radiation signal and reradiates a directed wireless power beam back in the reverse direction to the position of the electronic device. Several devices can be charged simultaneously. All of the rectennas on different electronic devices operate at the same frequency as the wireless power beam. The operating frequency determined by the array design is used for power transmission only, and can be different from the frequencies used for communications. Because the array elements are distributed at different distances from the electronic device, the phases of the radiation signals are different when reaching different elements. Retrodirective techniques make the phase of the wireless power beam at each element reradiate with the phase that is conjugate to the radiation signal at the same element. In this way, the relative phase differences among the array elements can be reversed and the wave front of reradiated beam can also be reversed to the original direction. Specically, in the Van Atta array, the phase conjugate technique is realized using properly designed transmission lines connecting the array elements [7]. A heterodyne retrodirective array uses RF mixers and lters for each array element [14]. The effective isotropic radiated power (EIRP) of the reradiated beam of the active Van Atta array with bi-directional amplier in Fig. 1 can be expressed as [15] (1) is a constant directly proportional to the number of where is the received antenna elements and the antenna gain,

is the radiation eld pattern of power from each antenna, a single antenna, and is the eld gain of the unilateral amplier. Equation (1) explains the situation of stage 1 in Fig. 1. In stage 1, the reradiated power generated by the retrodirective array depends on the received power from the radiation signal of the device to be charged and the characteristics of the array itself. In order to increase the transmission efciency of stage 1, we need to maximize the EIRP of the reradiated eld. In our appliis the power from the radiation signal, which cation is small from the device to be charged. can be tuned by the amplier which can be designed separately. is determined by the type of antenna element, which should be more omnidirectional so as to cover larger charging range and space. The number of array elements and antenna gain are directly related to the efciency and are our design targets. However, array elis constrained by the dimension of power ement number source implanted in the room. In addition, a large ultimately impacts power transfer efciency because of the increased mutual coupling, as will be shown later. Ideally, the power beam in stage 2 will all be reradiated back in the direction of the incident angle for the retrodirective array. Incident angle is determined by the angle between the horizontal plane and the vector from the position of the electronic device to , the center of the retrodirective array. In Fig. 1, denotes the incident angles of different devices. However, in the real situation, some portion of the power beam will be reradiated in other directions. The distribution of the power beam is measured at all possible angles of reception. is dened as the angle between the horObservation angle izontal plane and the vector from the position of observation point to the center of the retrodirective array. In a 2-D case, where the devices to be charged are placed in a line under the retrodirective array, the observation angle can be varied from 0 to 180 degrees, while the incident angle is a specic value in the same range. The observation angle from the array to each device equals the incident angle but in reversed direction. The focus of the array Focus is the quantity to represent the distribution of the power beam. For a specic device with a , Focus is dened as the percentage certain incident angle proportion of the power reradiated in a one degree span in the reversed direction of the incoming radiation signal to the total power in all directions in the lower half space under the array. The reversed direction of the incoming radiation signal can be . , the total exactly represented by the incident angle of is calculated by the received power by the device placed at . integration of the reradiated power in a narrow span around The power received by the device is highly directional reradiated from the retrodirective array; thus, one degree span receiving angle is used in this 2-D case. The total transmitted power is the summation of the reradiated power at all angles in the receiving plane where the devices are placed. Focus is dened as follows:

(2)

208

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

is used to represent the transmission efciency for . While a certain device can be placed the device placed at on any spot on the receiving plane under the array with an angle , the transmission efciency is a function of device placeof ment. Focus and transmission efciency are used interchangeably in this work. A wire dipole element based array is utilized to illustrate our theory and design. The antenna element is a standard wire dipole formed by two equal length conductors placed back to back with central feeding. It is intended for single-frequency centered narrowband operation, and its eld expression and antenna pattern are depicted in [16]. The array consists of multiple wire dipoles lined up in a row. The initial design is a half-wavelength wire-dipole array of -element with uniform spacing. The analytical result from the antenna array theory [16] is used as a reference for our full-wave analysis later. III. ARRAY DESIGN FOR WIRELESS POWER TRANSFER This section discusses design constrained associated with the retroactive array. The whole system is mainly analyzed with full wave methods of moment (MOM), besides analytical array theory. In the range of room and frequency of GHz, the electronic devices are in the far eld region of each element of the power source array. The space of the room in Fig. 1 is assumed in air and all the ceiling, bottom, or other sides are made of total absorbing material. We demonstrate that the reradiated power beam is reected in the incoming direction by the phase conjugate technique. We also analyze how the design factors affect the performance of the focusing of the power delivered to the device. The effect of manufacturing variability on the system behavior is also discussed. The amplier section of the retrodirective array system is not examined in this work, as the focus of the reradiated power beam is a ratio and not dependent on the actual power level. The actual power level is determined by the safety requirement of the wireless power regulation [17], [18]. The design objective for the power source array is to maximize its efciency in delivering the power to the electronic device to be charged. The wireless power beam needs to be as focused as possible in the reversed direction. We then examine the design factors, array element number , operating frequency freq, antenna element size , and variability of the elements spacing for their impact on the objective function Focus. A. Phase Conjugate Retrodirective Property The array consists of dipole elements of size of with element spacing at GHz. The incoming radiation signal from the electronic device is assumed to be a plane wave incident at 20 degrees. From the denition of the incident angle, 90 degrees means that the device is placed just under the center of the array in 2-D case, which is easy to receive the power beam. In contrast, 0 or 180 degrees means that the device is placed at the leftmost or rightmost of the innite large room, which is hard to receive the power beam from all array elements intuitively. It is also validated from Fig. 3 the monostatic RCS of reradiated power beam of the retrodirective array, with plane waves at different incident angles later. In a normal room of nite length, 20 degrees might be close to the
Fig. 2. Bistatic RCS of reradiated power beam of the retrodirective array, with incident plane wave at 20 degrees.

corner of the room, i.e., worst case. Thus, 20 degrees is chosen to illustrate a worst or worse than average case of transmission in Section IV. Fig. 2 shows the reradiation of the beam via phase conjugation at the array. When incident angle is 20 degrees, the largest portion of the reradiated power is also at about 20 degrees, while in other observation angles, the power level is very low. Focus in the bistatic case is the percentage of the power reradiated in a one degree span in this particular observation angle to the total power in all directions. The bistatic RCS plot validates the retrodirective property of the array at a certain incident angle. This experiment can be repeated for other values of incident angles, and the retrodirective property preserves. However, the percentage of the largest portion out of the total power, i.e., Focus, varies with different case of incident angles. Theoretically, the retrodirective property is signicant for any angle of incidence. In actual fact, the reradiated power level is lower with more oblique incidence, as shown in the monostatic RCS of the array in Fig. 3. There are obvious drops of focus with incident angles around 90 degrees, and few blind points with very low focus. If needed, a second shifted charging array can be added so that the chances of availability of blind charging spots for both arrays can be greatly reduced or eliminated. B. Number of Array Elements The focus of the array at the reective direction can be further sharpened by adding more array elements, according to (1). The phase of each array element is carefully set based on time reversal or phase conjugate theory [9]. With xed frequency GHz, and incident angle at 20 , the focus with different number of elements is shown in Fig. 4. From Fig. 3, focus will be larger for incident angles at 20 to 160 . The result of the full-wave MOM simulation result is validated by the analytical calculation by array theory. The MOM without coupling underestimates the focus, as it ignores the effect that the coupling between elements reinforces the eld in the target direction. smaller than 150, the focus of the array In Fig. 4, with increases approximately linearly with while the slope of

LI AND JANDHYALA: DESIGN OF RETRODIRECTIVE ANTENNA ARRAYS FOR SHORT-RANGE WIRELESS POWER TRANSMISSION

209

Fig. 3. Monostatic RCS of reradiated power beam of the retrodirective array, with plane waves at different incident angles.

Fig. 5. MOM simulated focus of reradiated power beam of the retrodirective array with sweep of ISM band frequencies, and with constant electronic size.

will be larger, and vice versa. All the antennas of different electronic devices work at the same frequency in each case, which is denoted by one particular result point in Fig. 5. From Fig. 5, the focuses are found to be independent of operating frequencies. Inside each band, focus is also insensitive to the frequency variation (less than 2% focus uctuation). Note that the physical antenna size is smaller, i.e., system size is reduced, at higher frequency, thus at least 2.45 GHz is suggested. D. Size of Element Antenna We then investigate the effects of the antenna element size at xed frequency of 2.45 GHz and xed . As suggested in Fig. 4, with larger the Focus value will be increased from 29% to about 55% and 89% when and 150 respectively. However, it means the size of the array will also be doubled and tripled respectively. With large allowance of the is used space and cost, large could be chosen. Here, for illustrating a space and cost-constrained case. The element spacing equals . Referring to Fig. 3 monostatic RCS, reradiated focus depends on incident angle of the radiation signal picked up by the array, so statistical analysis is undertaken under the assumption of equi-probable incident angles, to nd the optimal size of antenna. The average, maximum and minimum focus of the reradiated power beam is shown in Fig. 6 with a sweep of from to . Comprising the 3 curves, of to are reasonable choices. The maximum efciency is obtained by comparing the simulated results from all the possible incident angles at each array element size . Maximum efciency drops . The reason might be that the maximum abruptly around points obtained around the drop might come from different incident angle. Meanwhile, the average efciency comes from collective cases of simulations of all incident angles, and its curve shows a relatively smoother variation than the minimum and maximum cases. E. Manufacturing Variability We nally analyze the effect of variability of the geometry parameters. With the nominal design of operating frequency at 2.45 GHz, incident angle at 20 , and antenna element size

Fig. 4. Analytical and MOM simulated focus of reradiated power beam of the retrodirective array as a function of number of array elements.

the linear increment decreases slightly, i.e., the improvement larger than 150, of focus becomes less with larger . With the focus begins to uctuate around the limit at about 90%. So can be chosen between 50 and 150 to achieve in this case, certain focus and high efciency in terms of system physical size. The specic size chosen in this range will also depend on space and cost constraints. C. ISM Band Frequency In practice, the ISM band is used for commercial design. Communication band frequencies are also used in wireless power transmission [12]. If we keep the electronic size of the antenna elements constant at the central frequency of the band, , the focus at frequency bands centered at i.e., 915 MHz, 2.45 GHz, 5.8 GHz, and 24.125 GHz are plotted in Fig. 5. Dipole element number and incident angle are xed at 50 and 20 , respectively. Dipoles of different physical lengths are used for each ISM band frequency individually. Antenna varies depending on the specic size of the ISM band being considered. That central frequency means for lower frequencies, the physical length of the dipole

210

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I AVERAGE FOCUSES AT THE SAMPLE DESIGN WITH DIFFERENT PLACEMENT, SPACING VARIABILITY AND ARRAY SIZE

Fig. 6. Average, maximum and minimum focuses of reradiated power beam of the retrodirective array, from all incident angles, with sweep of the element antenna size at xed frequency and .

in Table I. Only one device to be charged is considered. Multiple devices can be placed at different spots under the array with different incident angles, while each one can be analyzed independently as a single device. Higher focus is obtained with larger , i.e., larger space allocation. One good sample design GHz, , , from Table I is variability tolerance of 10%, which provides and spacing an average focus of 47%. Other combination of the choices on the multiple parameters can be investigated using an extended lookup table, as in Table I, considering the limitations of the space allocated and the tradeoff between the performance and cost. IV. CONCLUSION Retrodirective wire dipole antenna array design for wireless power transmission in the range of typical room sizes is examined. Good efciency is achieved with this retrodirective property. The efciency of the passive array is analyzed with respective to multiple design factors, including number of array elements, antenna size, and spacing variability tolerance. All the simulations are carried using full wave EM solver and validated with analytical antenna array theory. Design suggestions are made for each design factor, and for the overall design. These suggestions include the use of intermediate antenna size and appropriate limits on manufacturing tolerances. In continuing work, dual-band, dual-polarized multilayered microstrip structure will be examined to replace the wire dipole antennas. Active array with amplier and rectier will be combined to the array design. The safe power level, array dimensions, and cost can then be included to determine the nal choices of the design factors. The delivered power level to the electronic devices depends on both the efciency Focus and supply power level. Under the current FCC regulation, the transmitted power density in uncontrolled environment should be less than 1 mW cm in the frequency range of 1.5100 GHz [17]. Our method can easily achieve this limit efciently. The limit of power density also implies that our method is best suitable for trickle or continuous charging as an add-on to the existing power sources. REFERENCES
[1] N. Tesla, The Transmission of Electric Energy Without Wires, The 13th Anniversary Number of the Electrical World and Engineer, 1904. [2] R. Tseng, Method and Apparatus for Wireless Power Transmission, U.S. Patent App. 11/901,158, 2007. [3] A. Kurs, A. Karalis, R. Moatt, J. D. Joannopoulos, P. Fisher, and M. Soljacic, Wireless power transfer via strongly coupled magnetic resonance, Science, vol. 317, no. 83, pp. 8386, Jul. 2007.

Fig. 7. Effect of manufacturing variability on the focus of reradiated power beam of the retrodirective array of the nominal design.

of , we examine the manufacturing variability of the array elements spacing with respective to . The nominal . We mimic manufacturing prospacing is cesses by introducing random variation in the spacing between elements. Fig. 7 shows the focus of reradiated power beam of the nominal design, compared with the case that spacing between the array elements randomly shifted by 1%, 10%, and 50%. The impact of manufacturing variability is negative on the focus. The impact is quite small when is less than 50, and when becomes larger, the impact increases to 10%20% depending on random shift amplitude. F. Design Guideline With Lookup Table In summary, after considering all the design factors above, we can create a lookup table of all the frequencies, elements number , element size , spacing , and their respective focus efGHz, , for the ciency. Suppose cases that the receiver is placed at 20 with respect to the array and at equi-probable random angles: the focus values with different spacing variability tolerance and array size are listed

LI AND JANDHYALA: DESIGN OF RETRODIRECTIVE ANTENNA ARRAYS FOR SHORT-RANGE WIRELESS POWER TRANSMISSION

211

[4] M. Fink, Time-reversal of ultrasonic eldsPart I: Basic principles, IEEE Trans. Ultrason., Ferroelectr., Freq. Control, vol. 39, no. 5, pp. 555566, May 1992. [5] G. Lerosey, J. de Rosny, A. Tourin, A. Derode, G. Montaldo, and M. Fink, Time reversal of electromagnetic waves, Phys. Rev. Lett., vol. 92, pp. 193904-1193904-3, May 2004. [6] L. C. Van Atta, Electromagnetic Reector, U.S. Patent 908,002, Oct. 6, 1959. [7] S.-J. Chung and K. Chang, A retrodirective microstrip antenna array, IEEE Trans. Antennas Propag., vol. 46, no. 12, pp. 18021809, Dec. 1998. [8] S.-J. Chung, S.-M. Chen, and Y.-C. Lee, A novel bi-directional amplier with applications in active Van Atta retrodirective arrays, IEEE Trans. Microw. Theory Tech., vol. 51, no. 2, pp. 542547, Feb. 2003. [9] C. Y. Pon, Retrodirective array using the heterodyne technique, IEEE Trans. Antennas Propag., vol. AP-12, no. 2, pp. 176180, Mar. 1964. [10] D. Pozar and T. A. Metzer, Analysis of a reectarray antenna using microstrip patches of variable size, IEEE. Electron Lett., vol. 29, no. 8, pp. 657658, Aug. 1993. [11] B. Munk, Finite Antenna Arrays and FSS. New York: Wiley, 2003. [12] Y.-H. Suh and K. Chang, A high-efciency dual-frequency rectenna for 2.45- and 5.8-GHz wireless power transmission, IEEE Trans. Microw. Theory Tech., vol. 50, no. 7, pp. 17841789, Jul. 2002. [13] Z. Popovic, R. Zane, A. Dolgov, J. Branna, J. Moroni, T. Paing, and J. Shin, Efcient low-power energy harvesting and power management, NanoPower Forum, 2007. [14] C. T. Rodenbeck and K. Chang, A limitation on the small-scale demonstration of retrodirective microwave power transmission from the solar power satellite, IEEE Antennas Propag. Mag., vol. 47, no. 4, pp. 6772, Aug. 2005. [15] L. Chiu, Investigation on retrodirective arrays: Novel component developments and system constructions, Ph.D. dissertation, City University of Hong Kong, Hong Kong, 2008. [16] C. A. Balanis, Antenna Theory, 2nd ed. New York: Wiley, 1997, pp. 133164. [17] OET Bulletin 65 Supplement C, [Online]. Available: http:// www.fcc.gov/Bureaus/Engineering_Technology/Documents/bulletins/oet65/oet65.pdf [18] FCC Regulations and Rules, Section XV.249, Part 15.

Ying Li (M06) received the B.Eng. degree in electrical engineering from Huazhong University of Science and Technology, Wuhan, China, in 2003, and the M.S. degree from McMaster University, Hamilton, ON, Canada, in 2006. She is currently working toward the Ph.D. degree at the University of Washington, Seattle. Her research interests include computational electromagnetics, signal integrity, high-speed packaging design, sensitivity analysis, engineering optimization, and wireless power transmission systems.

Vikram Jandhyala (M98SM03) received the B.Tech. degree in electrical engineering from Indian Institute of Technology (IIT), Delhi, in 1993, and the M.S. and Ph.D. in electrical engineering from the University of Illinois at Urbana-Champaign in 1995 and 1998, respectively. He is a Professor of electrical engineering at the University of Washington (UW), Seattle, and Director of the Applied Computational Engineering Lab. He is chair of the UW electrical engineering professional masters program, and is an inaugural Entrepreneurial Faculty Fellow in UWs Center for Commercialization. He is founder and chairman of Nimbic (formerly Physware), Inc., a venture-funded startup in the area of scalable electronic design automation and cloud-based scalable electromagnetic simulation with many of the worlds top semiconductor companies as enterprise customers. His research interests include several aspects of computational electromagnetics, large-scale and cloud simulation, physics-based computing algorithms, and electronic design automation. He has published approximately 200 papers in peer-reviewed journals and conferences and has directed research funded by DARPA, NSF, NASA, SRC, INTEL, IBM, the SBIR program, AFRL, Navy, and other sponsors. He developed accelerated boundary element solvers at Ansoft Corporation, Pittsburgh, from 1998-2000 prior to joining UW in 2000. Honors include an NSF CAREER award, Chairs award and Outstanding Research Advisor award at UWEE, a NASA inventor award, and graduate research awards from IEEE and UIUC. He is a Senior Member of the IEEE and an Elected Full Member of URSI Commission B.

212

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Anisotropic Impedance Surfaces for Linear to Circular Polarization Conversion

Efstratios Doumanis, Member, IEEE, George Goussetis, Member, IEEE, Jos Luis Gmez-Tornero, Member, IEEE, Robert Cahill, Senior Member, IEEE, and Vincent Fusco, Fellow, IEEE

AbstractAnisotropic impedance surfaces are employed as low-prole and broadband reectors that convert orthogonal linear to right- and left-handed circular polarization respectively. By virtue of anisotropy, it is possible to independently control the reection characteristics of two orthogonal linearly polarized incident plane waves and therefore achieve linear to circular polarization conversion. Equivalent circuits for anisotropic impedance surfaces with arbitrarily shaped elements are employed to demonstrate the operating principle and a design procedure is proposed. The proposed design procedure is demonstrated by means of an example involving a dipole array. A prototype is designed and its performance characteristics are evaluated. The 3-dB relative axial ratio bandwidth exceeds 60%, while low loss and angular stability are also reported. Numerical and experimental results on a fabricated prototype are presented to validate the synthesis and the performance. Index TermsCircular polarization, frequency-selective surfaces, high-impedance surfaces, polarization conversion.

I. INTRODUCTION

OLARIZATION converters are key elements in microwave sensors and mm-wave systems. They nd applications in millimeter-wave and sub-millimeter wave imaging applications [1]. In satellite systems, polarization converters are used to minimize the effect of Faraday rotation caused by the ionosphere [2]. They have been used in the design of circulators [3] and isolators [1], [4] as well as for remote environmental monitoring applications [5]. Polarization transformers are also important in antenna applications where polarization diversity is highly desired [6]. Various polarization converter prototypes have been demonstrated to date [1], [3][15]. A three-layer design based on a combination of wires and rectangles is presented in [7]. In order to improve tolerances, a design based on multilayer meander-line structures was proposed in [8] and was
Manuscript received November 04, 2010; revised April 24, 2011; accepted June 06, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the Royal Academy of Engineering, U.K., and was sponsored under the Department for Education and Learning Strengthening the All-Island Research Base program, Mobile Wireless Futures, and the UK Engineering and Physical Science Research Council under grant EP/E01707X/1. E. Doumanis, G. Goussetis, R. Cahill, and V. Fusco are with the School of Electronics, Electrical Engineering and Computer Science, Queens University Belfast, Belfast BT7 1NN, U.K. (e-mail: e.doumanis@qub.ac.uk, g.goussetis@ieee.org, r.cahill@ee.qub.ac.uk, v.fusco@ecit.qub.ac.uk). J. L. Gmez-Tornero is with the Department of Communication and Information Technologies, Technical University of Cartagena, Cartagena 30202, Spain (e-mail: josel.gomez@upct.es). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167920

further studied in [9] and [10]. A variant structure consisting of a cascade of meanders was proposed in [11] for arbitrary rotation of linear polarization (LP). A cascade of four identical grids of capacitively loaded dipoles was reported in [12]. The operating principle of a meander line polarizer for linear to circular polarization (CP) conversion as well as for arbitrary rotation of linear polarization is demonstrated in [13]. A variety of all-metal structures suitable for sub-millimeter wave frequencies polarization conversion were presented in [14]. They are based on double layer aperture frequency selective surfaces (FSS). An all-metal double layer array of split slot rings employed in a quasi-circulator for RCS characterization was reported in [3] while an all-metal design based on three dimensional helices was studied in [16]. All of the polarization converter surfaces reported above consist of multilayer planar arrays (or three-dimensional structures in the case of [16]). This increases bulkiness, due to the need for multilayer structures with layers commonly placed quarter wavelength apart, as well as the fabrication complexity and associated costs. A single layer split slot ring LP to circular polarization (CP) converter design was reported in [5], which however reects approximately 3 dB of the incoming power leading to high insertion loss. Additionally all the above designs operate in the transmit mode. In [15] the use of polarizers in mm-wave imaging systems is described. In this system [1], [15], schematically represented in Fig. 1, it would be benecial to combine the reector and the polarization converter in a single component that could perform both operations. This would signicantly reduce system complexity by replacing the linear to circular polarization transformer and the scanning mirror (block P in Fig. 1) in Fig. 1). by a linear to circular polarizing reector (block A linear to circular polarization reector has been proposed in [17]. This design involves two grids, one for each of the two orapart. thogonal polarizations of the incident wave, placed Although simple in concept, this solution is of severely limited usage due to its inherent narrow-band operation and poor angular stability. Doubly periodic planar metallo-dielectric arrays have over the past decade been extensively studied in the literature as engineered impedance surfaces [18][28]. When supported by a ground plane, and neglecting thermal losses or grating lobes, these structures fully reect incident plane waves in a specular direction with a tailored phase shift. Among those surfaces, anisotropic designs impose a differential phase shift to the two polarizations of the incoming plane wave [29][34]. A reected circularly polarized wave can hence be achieved by means of the differential reection phase provided by an

0018-926X/$26.00 2011 IEEE

DOUMANIS et al.: ANISOTROPIC IMPEDANCE SURFACES FOR LINEAR TO CIRCULAR POLARIZATION CONVERSION

213

Fig. 1. Simplied block diagram of an imaging mm-wave system. Proposed replacement of the scanning mirror and LP to CP converter by a reection polarization transformer.

anisotropic impedance surface [34]. In this paper, an equivalent transmission line network approach is initially employed to demonstrate the operating principle. This network is subsequently employed to propose an approximate design technique followed by an example involving a dipole array. Advantages of the proposed reection polarizer include low-prole, mass and size, wide-band operation, low-loss and angular stability. The proposed structure is also compatible with conventional single layer PCB technology, thus minimizing the associated costs and allowing scalability to mm-waves. Numerical and experimental results are presented to demonstrate the performance characteristics. II. PRINCIPLE OF OPERATION The polarization of a plane wave refers to the orientation of the electric eld vector, which may be in a xed direction or may change with time. Circular polarization is characterized by electric eld where the two orthogonal components are of the same amplitude and 90 (or odd multiples of) out of phase [35]. A linearly polarized wave may be converted to a circularly polarized wave by means of an engineered reector, which provides this difference in phase between two crossed linear components, as discussed in the following. Without loss of generality we assume a linearly polarized half space incident on the surface plane wave from the which lies on the xy-plane. The incidence plane is assumed to be normal to the y-axis (xz-plane) and the direction of propagation (wavenumber) of the incoming wave is at an angle with the z-axis (Fig. 2). Two orthogonal linearly polarized plane waves suitable for the expansion of the incoming and outgoing waves are dened by electric and magnetic elds transverse to , these are commonly rethe xz-plane respectively. For ferred to as TE and TM polarizations and are schematically depicted in Fig. 2. Next assume that the incoming wave is a superposition of a TE and a TM wave with equal magnitude and phase. Such a wave will be linearly polarized with the electric

Fig. 2. (a) TM and (b) TE incidence on an anisotropic impedance surface consisting of double periodic arrangement of arbitrary elements printed on a grounded dielectric slab. Geometrical conguration (left) and equivalent transmission line network (right).

eld lying at an angle from the positive y-axis, which can be either 45 or 135 . If the surface is lossless and no grating lobes exist, both the TE and TM components will be fully reected in the specular direction. The condition for the outgoing wave to have circular polarization is therefore that the impedance surface imposes a differential reection phase of 90 (or odd multiples of) to , if the TE compothe TE and TM component. For phase advance with respect nent is reected with 90 to the TM component, the reected wave will be characterized by left-handed circular polarization (LHCP). Similarly, if the phase advance TM component is reected with 90 with respect to the TE component, the reected wave will be right-handed circular polarized (RHCP). The opposite is true for . Similar conditions hold for the reection phases of the x- and y-polarized components for normally incident plane waves , where TE and TM polarizations are not formally dened. A. Equivalent Circuit Model for Reection Polarizer In order to illustrate the above and provide design guidelines, the reection properties of the structure under normal and oblique angles of incidence are investigated employing equivalent transmission line networks [29]. The equivalent transmission line circuits for the TM-polarized and the TE-polarized

214

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

waves are shown in Fig. 2(a) and (b) (right side), respectively. Similar circuits are valid for the two orthogonal polarizations in the case of normal TEM incidence. For TM-waves, the shunt admittance at represents the periodic array. The rest of the network can be described by sections of TM transmission lines associated with the semi-innite free-space and the grounded dielectric slab of thickness . The propagation conand stants for the two transmission lines are respectively. Geometrical considerations everyand boundary conditions impose that where. The TM characteristic admittance are and , respectively. An analogous network is given for the TE case where in this case the characteristic impedances for the semi-innite free space and the dielectric and respectively. For are TE-waves the periodic array is represented by a shunt admitplaced at . tance Assuming no losses and grating lobes, the metallic ground plane ensures that incident waves are fully reected in the specular direction for all frequencies and both polarizations. The reexperienced by TM,E waves can be ection coefcient readily derived as (1) where is given by (2) In (2), at [33] (3) Without loss of generality we assume that the incident wave . The condition for is linearly polarized at an angle converting the linear polarization of the incident wave to circular polarization for the reected beam is (4) where as discussed above, the sign refers conversion to RCHP and LHCP, respectively. Combining (1)(4), the required equivalent admittance of the periodic array for TE-polarized waves, , is obtained as a function of the equivalent susceptance for TM waves, is the admittance of the ground plane as seen

Fig. 3. Layout of a dipole array and unit cell dimensions.

metallo-dielectric dipole arrays have been extensively studied as frequency selective surfaces FSSs [36], [37] and their equivalent surface impedance can be obtained using a variety of available techniques (e.g., [21] and [29]). For TM waves the electric eld is perpendicular to the dipoles and therefore an array of thin can to a rst approximation be modeled by dipoles . Equation (5), therefore, rean open circuit, so that duces to (6) In (6), the frequency dependence is only in the argument of the cotangent function of (3). Therefore, (6) is a periodic function of frequency with period (7) Moreover, (6) reveals that tends to innity when the is equal in TM susceptance of the ground plane seen at magnitude to the TM wave impedance of free space (8) When (8) is satised, an incident TM wave is reected at . At these points RHCP or LHCP for the outgoing wave is occurs when the dipole array for TE incidence at resonance, i.e., when for TE waves the array behaves as a short circuit and reects incoming waves with a phase reversal. Away from these frequencies, (6) is decreasing with frequency. In practice, a surface with equivalent impedance decreasing with frequency violates Fosters reactance theorem [38] and therefore can not be realized. For bands where is decreasing with frequency, the equivalent admittance of a practical array can match only at a frequency point and quickly diverge away from there, so that the axial ratio bandwidth is narrow. increases with In the vicinity where (8) is satised, frequency and is therefore possible to approximate it with a practical array that follows the required impedance trend over a certain frequency band [22] yielding a nite axial ratio bandwidth. Some further algebra reveals that (8) is met when (9)

(5) An initial design procedure therefore involves determining the physical array dimensions that satisfy (5). B. Dipole Array Polarizer Next, the approximate design procedure is demonstrated by means of an example. For simplicity, an array of dipoles is considered, with unit cell as shown in Fig. 3. Doubly periodic planar Equation (9) provides with some initial guidelines for the choice of dielectric substrate. In addition, since the left-hand side of (9) is monotonically decreasing with frequency, this equation dictates that the frequency band where a practical design can approximate the required admittance for LHCP is always at lower frequencies than for RHCP.

DOUMANIS et al.: ANISOTROPIC IMPEDANCE SURFACES FOR LINEAR TO CIRCULAR POLARIZATION CONVERSION

215

In order to illustrate the above, in the following a dielectric and thickness of substrate with permittivity mm is considered for a linear to circular polarization converter within the 1015 GHz band. The angle of incidence is as exemplied in the schematic of Fig. 1. Using (6), the required TE susceptance for RHCP and LHCP is obtained and plotted in Fig. 4. As can be observed for the case of RHCP the required TE susceptance (dashed line) is decreasing monotonically with frequency in the range 5 to 18 GHz. This suggests that a practical array can convert incoming linear polarization to RHCP at a single frequency point and with a narrow bandwidth. For the case of LHCP (solid line) the required susceptance is decreasing in the frequency range 5 to almost 10.5 GHz and increasing with frequency thereafter. Beyond 10.5 GHz it is, therefore, plausible to seek the dimensions of periodic dipole arrays that will produce linear to LHCP conversion over a nite bandwidth. In practice, the surface impedance of periodic dipole arrays can not exactly match any given increasing function with frequency, such as the solid line curve in Fig. 4, over an arbitrary bandwidth. Recognizing that a deviation from the required impedance value results in an increase of the axial ratio, in practical designs a tradeoff between the axial ratio and bandwidth is often necessary. Based on the theory outlined above, it is possible to extract the range of equivalent impedance values that will produce axial ratio within xed limits. The lighter and darker grey shaded area in Fig. 4 correspond to surface admittance values that will produce a LHCP with an axial ratio of less than 1 dB and 2 dB, respectively. An initial design of the geometrical dimensions can, therefore, be based on graphs such as those depicted in Fig. 4, taking also into consideration the tradeoff between bandwidth and axial ratio levels. Analytical [21] or semi-analytical [29], [39] expressions are available for the equivalent surface impedance of periodic dipole arrays and a method for the design of a dipole array that produces a given equivalent surface impedance dispersion has been outlined in [22]. These techniques can be employed for the design of the initial dimensions of the dipole reector. For other array geometries, the more generic (5) should be used rather than the approximate (6). III. NUMERICAL AND EXPERIMENTAL RESULTS Once an estimation of the physical dimensions has been obtained, some optimization based on full-wave simulations is necessary to account predominantly for the fact that the TM is not exactly zero but instead array impedance has a nonzero susceptance, which depends on the array geometry. Using the same substrate of the previous section, a dipole array has been designed. CST Microwave Studio was employed for the full-wave simulations. Metallic and dielectric losses are accounted for in the simulations. In particular, the loss tangent of and the conductivity of copper the substrate is is used for the metal dipoles. The thickness of the dipoles and of the ground plane is assumed to be 35 m. An axial ratio requirement of less than 1.5 dB as well as fabrication tolerance constraints have been considered during the optimization. Remm, ferring to Fig. 3, the nal dimensions are mm, mm and mm.

Fig. 4. Required TE surface admittance (Y ) of a ctitious array with TM admittance equal to zero that when printed on a substrate of thickness t = 1:524 mm and permittivity " = 3:5 converts plane waves incident at angle  = 45 and linearly polarized at an angle  = 45 with respect to the positive y-axis to LHCP (solid line) and RHCP (dashed line). Light/dark grey area: TE admittance region where LHCP axial ratio is below 1 dB/2 dB. Equivalent ) of Fig. 5 (circle surface admittance for the practical impedance surface (Y ).

The equivalent sheet admittance for this array was derived using the pole zero technique described in [29] and is plotted in Fig. 4. Based on this gure, the surface is expected GHz and LHCP at to produce RHCP at GHz and GHz. Also, Fig. 4 suggests that a variation of about 2 dB in the LHCP axial ratio is expected in this freas obtained from quency range. The reection coefcient (1) assuming is plotted in Fig. 5 (grey solid line). The reection phase in the presence of the array as obtained using full-wave simulation is superimposed in Fig. 5 for comparison (solid line). The discrepancy between the two curves is attributed to the approximation of an open circuit for the dipole array upon TM illumination. As it can be seen in Fig. 5, the open circuit approximation is increasingly accurate for lower frequencies. The required reection phases, , that will convert incident linear polarization to RHCP and LHCP can be obtained by subtracting 90 and 270 respectively from the full-wave reection . The relevant curves are plotted in Fig. 5 (dotted line and dash-dotted lines respectively). In view of Fig. 5, at frequency GHz, the difference between the reection phase experienced by the TE and TM components is 90 which results in a right-hand circularly polarized wave (RHCP). At frequencies GHz and GHz the phase difference is 270 and the reected wave is left-hand circularly polarized (LHCP). Due to the smooth variation of the reection phases for both polarizations between these two frequencies, a small variation of the axial ratio is anticipated within this range. These observations are in good agreement with those derived from Fig. 4. Some discrepancy is attributed to the approximation made in (6), which is increasingly inaccurate

216

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 5. Full wave reection phase for plane waves incident at  and linearly polarized at an angle  with respect to the positive y-axis onto : ;W : ;D : and a dipole array with dimensions (in mm): L D : printed on a substrate with thickness t : mm and relative per: for TM (solid line) and TE (dashed line) polarizations. Remittivity " quired reection phase of the TE component for RHCP (dotted line) and LHCP (dash-dotted line). Reection phase of the TM polarized incident wave for the case of an un-patterned grounded dielectric (grey solid line).

= 45

=10

=35

=70 =05 = 1 524

= 45 =80

Fig. 6. Simulated axial ratio (dB) of the reected wave from the array of Fig. 5 at incidence angles  for incident plane wave linearly polarized at  (solid line),  (dashed line),  (dotted line), and  (dash-dotted line).

= 30

= 45 = 45

= = 60

towards higher frequencies. As a result, the frequencies and shown in Fig. 4 are slightly different than the ones obtained from Fig. 5 ( and ). The simulated reection loss is small and comparable for the two polarizations. The maximum reection loss is 0.2 dB and is observed for the TE polarization at 8.29 GHz, where reection phase is 0 . This frequency, associated with Articial Magnetic Conductor operation, is known to exhibit stronger resonance phenomena [20] and therefore thermal losses peak around that frequency [40]. Signicantly, the frequency range of interest lies outside strong resonance phenomena and therefore the losses for both polarizations are small. For frequencies between 10.5 to 20 GHz, the thermal loss for both components results to less than 0.04 dB reduction in the reection coefcient. The grating lobe region is well above the operational frequency range of the polarization converter for all angles of incidence considered. The above suggest that the assumption of full specular reection for both polarizations is valid to a good extend and therefore to a good approximation the design can be based on the reection phases. We note that in case this assumption does not hold, a higher absorption of either polarization can be compensated by tilting the incoming wave polarization angle with the y-axis, , to values different than 45 , thus increasing the relative strength of the component that experiences higher losses. The axial ratio as obtained from the full-wave simulations for this array for incidence angle , is shown in Fig. 6 (dotted line). The 3 dB axial ratio bandwidth is more than 63%, while the 1.5 dB axial ratio bandwidth is over 52%. The minimum axial ratio for RHCP is 0.16 dB at 7.95 GHz. For LHCP, two minimum points are observed at 10.27 GHz and 14.86 GHz where the simulated axial ratio is 0.03 dB and 0.006 dB respec-

tively. These frequencies exactly coincide with the frequencies and of Fig. 5. Fig. 6 also shows the axial ratio for various angles of incidence between 0 and 60 . The 3 dB axial ratio bandwidth for (solid line), 30 (dashed line), and 60 (dash-dotted line) are 60.8%, 63.1%, and 56.6%, respectively. Within the 9.85 GHz to 16.5 GHz band the axial ratio is below 1.5 dB for all angles of incidence with exception grazing incidence at 60 . The low prole of the structure together with the reported levels of polarization purity over a large bandwidth and angular eld-of-view is a signicant improvement compared to the state of the art [1], [10]. In order to experimentally validate the above, a prototype array has been fabricated and measured. The array has been photo-etched on a Taconic RF-35 laminate with a relative di, loss tangent and electric permittivity thickness mm. The thickness of the copper dipole elements and ground plane is 35 m. The prototype array consists of 272 34 elements with overall dimensions 30 30 cm. A photograph of the prototype is shown in Fig. 7(a). Standard-gain X-band horn antennas are used as the receiver and transmitter. A linearly polarized horn antenna (Tx) is fed from a Vector Network Analyzer (VNA) and positioned at angle to the screen at normal incidence. The array is positioned at a distance of 60 cm away from the two antennas. The reection phase in the far eld is taken using the horn antennas and is then normalized with respect to an identical measurement where the array is substituted by a fully metallic surface. TE and TM incidence is achieved by relative rotation of the horn antennas by 90 . A photograph of the measurement setup is shown in Fig. 7(b) for TE incidence. The measured axial ratio up to 13 GHz is plotted in Fig. 8 along with the simulated one as obtained with CST for comparison. The measured response is in good agreement with the simulation. Some discrepancies are attributed to experimental tolerances.

DOUMANIS et al.: ANISOTROPIC IMPEDANCE SURFACES FOR LINEAR TO CIRCULAR POLARIZATION CONVERSION

217

has been demonstrated. An approximate design technique based on Transverse Equivalent Networks for both TE and TM polarizations has been proposed. By means of an example, the operation principle and the design technique are outlined and the performance is assessed. The 3 dB axial ratio for the given example was in excess of 60% over a wide angular bandwidth. Full-wave numerical and experimental results have been presented that demonstrate the LP to CP conversion performance of the proposed design. As a closing remark we note that the frequency point where linear polarization is converted to RHCP ( in Fig. 5) can be designed to be increasingly closer to the point where reected waves are converted to LHCP ( in . Such a design could Fig. 5) by increasing the periodicity, potentially be employed for the dynamic conversion of linear to LHCP or RHCP upon inclusion of a substrate with tunable permittivity (such as nematic liquid crystals [41]), or by using tunable high-impedance surfaces [42]. ACKNOWLEDGMENT The authors would like to thank Dr. D. Robertson for fruitful discussions and Taconic Advanced Dielectric Division for providing the substrate. REFERENCES
[1] C. Dietlein, A. Luukanen, Z. B. Popovic, and E. Grossman, A W-band polarization converter and isolator, IEEE Trans. Antennas Propag., vol. 55, no. 6, pp. 18041809, Jun. 2007. [2] G. Maral and M. Bousquet, Satellite Communications Systems, Systems, Techniques and Technology. Sussex, U.K.: Wiley, 2009, ch. 5, p. 208. [3] M. Euler and V. Fusco, Split slot ring spatial quasi-circulator for RCS characterization, Electron. Lett., vol. 46, no. 6, pp. 394395, Mar. 2010. [4] S. Hollung, W. A. Shiroma, M. Markovic, and Z. B. Popovic, A quasioptical isolator, IEEE Trans. Microw. Guided Wave Lett., vol. 6, no. 5, pp. 205206, May 1996. [5] M. Euler, V. Fusco, R. Cahill, and R. Dickie, 325 GHz single layer sub-millimeter wave FSS based split ring linear to circular polarization convertor, IEEE Trans. Antennas Propag., vol. 58, no. 7, pp. 24572459, Jul. 2010. [6] K. S. Min, J. Hirokawa, K. Sakurai, M. Ando, and N. Goto, Single-layer dipole array for linear-to-circular polarization conversion of slotted waveguide array, Inst. Elect. Electr. Proc. Microw. Antennas Propag., vol. 143, no. 3, June 1996. [7] D. S. Lerner, A wave polarization converter for circular polarization, IEEE Trans. Antennas Propag., vol. 13, no. 1, pp. 37, Jan. 1965. [8] L. Young, L. A. Robinson, and C. A. Hacking, Meander-line polarizer, IEEE Trans. Antennas Propag., vol. 21, no. 3, pp. 376378, May 1973. [9] C. Terret, J. R. Levrel, and K. Mahdjoubi, Susceptance computation of a Meander-line polarizer layer, IEEE Trans. Antennas Propag., vol. 32, no. 9, pp. 10071011, Sep. 1984. [10] R.-S. Chu and K. M. Lee, Analytical model of a multilayered Meander-line polarizer plate with normal and oblique plane-wave incidence, IEEE Trans. Antennas Propag., vol. 35, no. 6, pp. 652661, Jun. 1987. [11] T. K. Wu, Meander-line polarizer for arbitrary rotation of linear polarization, IEEE Trans. Microw. Guided Wave Lett., vol. 4, no. 6, pp. 199201, Jun. 1994. [12] W. A. Shiroma, S. C. Bundy, S. H. Brenda, B. D. Bauernfeind, and Z. B. Popovic, Cascaded active and passive quasi-optical grids, IEEE Trans. Microw. Theory Tech., vol. 43, no. 12, pp. 29042909, Dec. 1995. [13] J. C. Zhang, Y. Z. Yin, and J. P. Ma, Multifunctional Meander line polarizer, Prog. Electromagn. Res. Lett., vol. 6, pp. 5560, 2009.

Fig. 7. (a) Photograph of the fabricated prototype (part of the array zoomed as an inset) and (b) the measurement setup.

Fig. 8. Measured axial ratio of the fabricated design for plane wave angle of (dotted line). The simulation for  is repeated from incidence  Fig. 6 for comparison.

= 45

= 45

IV. CONCLUSION A single-layer anisotropic impedance surface that reects orthogonal linear plane waves to RHCP and LHCP, respectively, has been presented. The basic principle of operation

218

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[14] M. Euler, V. Fusco, R. Cahill, and R. Dickie, Sub-millimetre wave linear to circular polarization converter study, in Proc. Eur. Microw. Conf., Rome, Italy, Sep. 2009, pp. 28922895. [15] K. M. K. H. Leong and W. A. Shiroma, Wafe-grid polarizer, Electron. Lett., vol. 38, no. 22, pp. 13601361, Oct. 2002. [16] M. G. Silveirinha, Design of linear-to-circular polarization transformers made of long densely packed metallic helices, IEEE Trans. Antennas Propag., vol. 56, no. 2, pp. 390401, Feb. 2008. [17] H. Meikle, Modern Radar Systems. Norwood, MA: Artech House, 2008, ch. 5, pp. 147152. [18] D. Sievenpiper, Z. Lijun, R. F. Broas, N. G. Alexopoulos, and E. Yablonovitch, High-impedance electromagnetic surfaces with a forbidden frequency band, IEEE Trans. Microw. Theory Tech., vol. 47, no. 11, pp. 20592074, Nov. 1999. [19] S. Clavijo, R. E. Diaz, and W. E. McKinzie, Design methodology for Sievenpiper high-impedance surfaces for positive gain electrically small antennas, IEEE Trans. Antennas Prop., vol. 51, no. 10, pp. 26782690, Oct. 2003. [20] G. Goussetis, A. P. Feresidis, and J. C. Vardaxoglou, Tailoring the AMC and EBG characteristics of periodic metallic arrays printed on grounded on grounded dielectric substrate, IEEE Trans. Antennas Propag., vol. 54, no. 1, pp. 8289, Jan. 2006. [21] O. Luukkonen, C. R. Simovski, G. Granet, G. Goussetis, D. Lioubtchenko, A. Raisanen, and S. A. Tretyakov, Simple and accurate analytical model of planar grids and high-impedance surfaces comprising of metal strips or patches, IEEE Trans. Antennas Propag., vol. 56, no. 6, pp. 16241632, Jun. 2008. [22] G. Goussetis, J. L. Gomez-Tornero, A. P. Feresidis, and N. Uzunoglu, Articial impedance surfaces for reduced dispersion in antenna feeding systems, IEEE Trans. Antennas Propag., vol. 54, no. 1, pp. 8289, Jan. 2006. [23] A. Monorchio, G. Manara, and L. Lanuzza, Synthesis of articial magnetic conductors by using multilayered frequency selective surfaces, IEEE Trans. Antennas Wireless Propag. Lett., vol. 1, 2002. [24] A. Feresidis, G. Goussetis, S. Wang, and J. C. Vardaxoglou, Articial magnetic conductor surfaces and their application to low prole highgain planar antennas, IEEE Trans. Antennas Propag., vol. 53, no. 1, pp. 209215, Jan. 2005. [25] P. Kildal, A. A. Kishk, and S. Maci, Special issue on articial magnetic conductors, soft/hard surfaces, and other complex surfaces, IEEE Trans. Antennas Propag., vol. 53, no. 1, Jan. 2005. [26] S. Wang, A. P. Feresidis, G. Goussetis, and J. C. Vardaxoglou, Highgain subwavelength resonant cavity antennas based on metamaterial ground planes, Inst. Electr. Elect. Proc. Microw. Antennas Propag., vol. 153, no. 1, pp. 16, Feb. 2006. [27] R. F. Jimenez Broas, D. F. Sievenpiper, and E. Yablonovitch, A highimpedance ground plane applied to a cellphone handset geometry, IEEE Trans. Microw. Theory Tech., vol. 49, no. 7, pp. 12621265, Jul. 2001. [28] G. Goussetis, A. P. Feresidis, A. Yakovlev, and C. Simovski, Chapter 31 in Metamaterial Handbook, in High Impedance Surfaces. London, U.K.: Taylor and Francis, vol. I, ISBN: 97-814-2005-3623. [29] S. Maci, M. Caiazzo, A. Cucini, and M. Casaletti, A pole-zero matching method for EBG surfaces composed of a dipole FSS printed on a grounded dielectric slab, IEEE Trans. Antennas Propag., vol. 53, no. 1, pp. 7081, Jan. 2005. [30] M. A. Hiranandani, A. B. Yakovlev, and A. A. Kishk, Articial magnetic conductors realized by frequency-selective surfaces on a ground dielectric slab for antenna applications, IEE Proc. Microw. Antennas Propag., vol. 153, no. 5, pp. 487493, Oct. 2006. [31] M. Hosseinipanah and Q. Wu, Polarization-dependent articial magnetic conductor structures using asymmetrical frequency selective surface, in IEEE Int. Symp. Microw. Antenna, Propag. Tech. for Wireless Comm., 2009, pp. 707710. [32] G. Goussetis and A. P. Feresidis, Perturbed frequency selective surfaces for multiband high impedance surfaces, IET Microw. Antennas Propag., vol. 4, no. 8, pp. 11051110, 2010. [33] J. Huang and J. A. Encinar, Reectarray Antennas. Hoboken, NJ: Wiley, 2008. [34] K. Karkkainen and M. Stuchly, Frequency selective surface as a polarization transformer, Inst. Elect. Electr. Proc. Microw. Antennas Propag., vol. 149, no. 516, pp. 248252, 2002.

[35] B. Toh, R. Cahill, and V. Fusco, Understanding and measuring circular polarization, IEEE Trans. Education, vol. 46, no. 3, pp. 313318, Aug. 2003. [36] B. Munk, Frequency Selective Surfaces: Theory and Design. Hoboken, NJ: Wiley, 2000. [37] R. Mittra, C. Chan, and T. Cwik, Techniques for analyzing frequency selective surfacesA review, IEEE Proc., vol. 76, no. 12, pp. 15931615, Aug. 2002. [38] R. E. Collin, Foundations for Microwave Engineering, 2nd ed. Hoboken, NJ: Wiley/IEEE, 2000. [39] M. Garcia-Vigueras, J. L. Gomez-Tornero, G. Goussetis, J. S. GomezDiaz, and A. Alvarez-Melcon, A modied pole-zero technique for the synthesis of leaky-wave antennas loaded with dipole-based FSS, IEEE Trans. Antennas Propag., vol. 58, no. 6, pp. 19711979, Jun. 2010. [40] F. Costa, A. Monorchio, and G. Manara, Analysis and design of ultra thin electromagnetic absorbers comprising resistively loaded high impedance surfaces, IEEE Trans. Antennas Propag., vol. 58, no. 5, May 2010. [41] H. Wenfei, R. Cahill, J. A. Encinar, R. Dickie, H. Gamble, V. Fusco, and N. Grant, Design and measurement of recongurable millimeter wave reectarray cells with nematic liquid crystal, IEEE Trans. Antennas and Propagation, vol. 56, no. 10, pp. 31123117, Oct. 2008. [42] C. Mias and J. H. Yap, A varactor-tunable high impedance surface with a resistive-lumped-element biasing grid, IEEE Trans. Antennas Propag., vol. 55, no. 7, pp. 19551962, Jul. 2007.

Efstratios Doumanis (M11) graduated from the Electrical and Computer Engineering School in the Democritus University of Thrace, Greece, in April 2005, and received the Ph.D. degree from Queens University of Belfast, Belfast, U.K., in 2010. In 1998, he joined KFKI Research Institute for Particle and Nuclear Physics, Department of Space Technology, Budapest, Hungary, as student Engineer. Between 2008 and 2009 he was with Space Engineering, Rome, Italy, as an RF Engineer. In 2010, he joined the High Frequency Electronic Circuits (HFEC) Group, Queens University of Belfast, UK, as a Research Assistant. His research interests include microwave lters, frequency-selective surfaces and high impedance surfaces. In 2011, Dr. Doumanis was awarded the Best Student Paper in the European Conference on Antennas and Propagation (EuCAP), Rome, Italy.

George Goussetis (S99M02) graduated from the Electrical and Computer Engineering School in the National Technical University of Athens, Greece, in 1998, and received the Ph.D. degree from the University of Westminster, London, U.K., In 2002, he also received the B.Sc. degree in physics (rst class) from University College London (UCL), London, UK. In 1998, he joined the Space Engineering, Rome, Italy, as RF Engineer and in 1999 the Wireless Communications Research Group, University of Westminster, Westminster, UK, as a Research Assistant. Between 2002 and 2006, he was a Senior Research Fellow at Loughborough University, Loughborough, UK. Between 2006 and 2009 he was a Lecturer (Assistant Professor) with the School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, UK. He joined the Institute of Electronics Communications and Information Technology at Queens University Belfast, UK, in September 2009 as a Reader (Associate Professor). He has authored or co-authored over 150 peer-reviewed papers three book chapters and two patents. His research interests include the modelling and design of microwave lters, frequency-selective surfaces and periodic structures, leaky wave antennas, microwave heating as well numerical techniques for electromagnetics. Dr. Goussetis received the Onassis foundation scholarship in 2001. In October 2006 he was awarded a ve-year research fellowship by the Royal Academy of Engineering, UK. In 2010, he was visiting Professor at UPCT, Spain.

DOUMANIS et al.: ANISOTROPIC IMPEDANCE SURFACES FOR LINEAR TO CIRCULAR POLARIZATION CONVERSION

219

Jos Luis Gmez Tornero (M06) was born in Murcia, Spain, in 1977. He received the Telecommunications Engineer degree from the Polytechnic University of Valencia (UPV), Valencia, Spain, in 2001, and the laurea cum laude Ph.D. degree in telecommunication engineering from the Technical University of Cartagena (UPCT), Cartagena, Spain, in 2005. In 1999 he joined the Radio Communications Department, UPV, as a research student, where he was involved in the development of analytical and numerical tools for the automated design of microwave lters in waveguide technology for space applications. In 2000, he joined the Radio Frequency Division, Industry Alcatel Espacio, Madrid, Spain, where he was involved with the development of microwave active circuits for telemetry, tracking and control (TTC) transponders for space applications. In 2001, he joined the Technical University of Cartagena (UPCT), Spain, as an Assistant Professor. From October 2005 to February 2009, he held de position of Vice Dean for Students and Lectures affairs in the Telecommunication Engineering Faculty at the UPCT. Since 2008, he has been an Associate Professor at the Department of Communication and Information Technologies, UPCT. His current research interests include analysis and design of leaky-wave antennas and the development of numerical methods for the analysis of novel passive radiating structures in planar and waveguide technologies. Dr. Gmez Tornero received the national award from the foundation EPSONIbrica for the best Ph.D. project in the eld of Technology of Information and Communications (TIC) in July 2004. In June 2006, he received the Vodafone foundation-COIT/AEIT (Colegio Ocial de Ingenieros de Telecomunicacin) award to the best Spanish Ph.D. thesis in the area of Advanced Mobile Communications Technologies. This thesis was also awarded in December 2006 as the best thesis in the area of Electrical Engineering, by the Technical University of Cartagena. In February 2010, he was appointed CSIRO Distinguished Visiting Scientist by the CSIRO ICT Centre, Sydney.

Robert Cahill (M10-SM11) received the B.Sc. (rst class honors) degree in physics from the University of Aston, Birmingham, U.K., in 1979 and the Ph.D. degree in microwave electronics from the University of Kent, Canterbury, U.K., in 1982. He joined Queens University Belfast (QUB) in 1999 after a 17 year career working in the UK space and defense industry, where he worked on antenna and passive microwave device technology projects. During this time he pioneered methods for predicting the performance of antennas on complex scattering surfaces such as satellites and has developed techniques for analyzing and fabricating mm and sub-mm wave quasi-optical dichroic lters. Recently he has established a 100700 GHz quasi-optical S parameter measurement facility at QUB. He has exploited the results of numerous research projects, sponsored by the European Space Agency, EADS Astrium Space Ltd., the British National Space Agency, the Centre for Earth Observation Instrumentation (CEOI) and the UK Meteorological Ofce, to develop quasi-optical demultiplexers for atmospheric sounding radiometers in the range 89500 GHz. These include AMSU-B, AMAS, MARSCHALS and the ESA 500 GHz demonstrator. His recent interests also include the characterization of liquid crystal materials at microwave and mm wavelengths, and strategies for broad banding and creating active reectarray antennas. He has (co)-authored over 130 publications and holds four international patents.

Vincent Fusco (S82-M82-SM96-F04) received the Bachelors degree (1st class honors) in electrical and electronic engineering, the Ph.D. degree in microwave electronics, and the D.Sc. degree for his work on advanced front end architectures with enhanced functionality, from The Queens University of Belfast (QUB), Belfast, Northern Ireland, in 1979, 1982, and 2000, respectively. He is the Technical Director of the High Frequency Laboratories at The Queens University of Belfast, U.K., and is also Director of the International Centre for Research for System on Chip and Advanced MicroWireless Integration, SoCaM. His research interests include nonlinear microwave circuit design, and active and passive antenna techniques. He has published over 420 scientic papers in major journals and international conferences, and is the author of two text books. He holds several patents on active and retrodirective antennas and has contributed invited chapters to books in the elds of active antenna design and EM eld computation. Prof. Fusco is a Fellow of the Royal Academy of Engineering, and a Fellow of the Institution of Electrical Engineers (U.K.). In 1986, he was awarded a British Telecommunications Fellowship, and in 1997 he was awarded the NI Engineering Federation Trophy for outstanding industrially relevant research.

220

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Transmitting-Mode Time Reversal Imaging Using MUSIC Algorithm for Surveillance in Wireless Sensor Network
Xiao-Fei Liu, Bing-Zhong Wang, Member, IEEE, and Joshua Le-Wei Li, Fellow, IEEE
AbstractThe electromagnetic time reversal (TR) detection and imaging using multiple signal classication (MUSIC) is extended to the Maxwell equations problems with the transmitting-mode congurations, where the transmitter and receiver arrays are bistatic and the arrays elements are arbitrarily distributed. Based on this type of setup, the multistatic response matrix is derived in detail for target-scattering by relatively small anisotropic spheres located in the imaging domain. Then the electromagnetic transmitting-mode MUSIC algorithm is proposed by dening the pseudo-spectrum deduced from the Maxwell equations. By such extension, following advantages are achieved: a) leading to more accurate location and higher imaging resolution because the noncoincident arrays actually increase the effective aperture of the array; b) making the algorithm more exible and practical; c) enlarging the coverage area of the detection system due to the reciprocal mapping between the transmitter and receiver spaces; and d) overcoming one special problem that the echo-mode encounters when one target is atop another one in the upright direction of the transceiver antenna array even they are far-separated. Based on these advantages, we have numerically discussed typical applications of the proposed method for surveillance in wireless sensor network. Through examples, the validity, the improvement, and the extended potential of the electromagnetic transmitting-mode time reversal MUSIC method are veried. Index TermsElectromagnetic scattering, green function, image analysis, inverse problems, multisensor systems.

I. INTRODUCTION HE TIME reversal method, as a novel detection and imaging technique exploiting the multipath effect in the forward scattering and inverse scattering of electromagnetic wave, has attracted a lot of attention in recent years. The major adopted methods in literature include, although not limited only to, the time reversal mirror method [1][4](TRM), the time reversal operator decomposition (DORT) [5][8], and the time reversal multiple signal classication (TR-MUSIC) [9][11].

Manuscript received May 19, 2010; revised February 15, 2011; accepted July 20, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the High-Tech Research and Development Program of China (2006AA01Z275, 2008AA01Z206) and in part by the State Scholarship Fund from the China Scholarship Council (No. 2008607035). J. L.-W. Li was supported by the University of Electronic Science and Technology of China, Chengdu, under the Chinese Governments 1000Talent Plan. X.-F. Liu and B.-Z. Wang are with the Institute of Applied Physics, University of Electronic Science and Technology of China, Chengdu, 610054, China (e-mail: lxf_uestc@hotmail.com; bzwang@uestc.edu.cn). J. L.-W. Li is with the Institute of Electromagnetics and School of Electronic Engineering, University of Electronic Science and Technology of China, Chengdu, 611731, China (e-mail: lwli@ieee.org). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167903

These methods were originally developed and improved in acoustic wave research. The TR-MUSIC algorithm was, however, recently extended to characterize electromagnetic elds [12][17] and employed to applications to subsurface detection, medical imaging, distributed sensing, radar systems and so on. It was reported that the TR-MUSIC algorithm evades the iterative back-propagation process required in TRM and overcomes the difculty in dealing with non-well-separated targets for DORT. In addition, the TR-MUSIC is valid for detections and imaging of multiple targets and extended targets with one step processing. However, the rst transmitting-mode investigation, extension from the mostly discussed echo-mode, of the TR-MUSIC algorithm is carried out under the frame of Helmholtz equation with non-coincident and arbitrary array elements distributions by E. A. Marengo and F. K. Gruber [18]. Their excellent works discussed the comprehensive theory of a nonlinear inverse scattering problem combined with the TR-MUSIC or an alternative signal subspace method in detail for estimating the locations and scattering strengths or reectivities of the targets. By considering the applications in wireless sensor network (WSN), we further extend the studies in this paper to Maxwell equations for the electromagnetic transmitting-mode TR-MUSIC with array conguration of irregular elements distribution because the node locations in WSN are often arbitrary and the transmitter and receiver are generally bistatic. Therefore, the vector elds, the dyadic Greens functions and the tensors of EM parameters are introduced for the analysis. Similar to the DORT method, the transmitting-mode TR-MUSIC method also starts from the multistatic response matrix or the transfer matrix. The difference is that the latter utilizes the so-called noise space of the multistatic response matrix while the former uses the signal space. Therefore, the location and the imaging of the targets can be achieved by taking the orthogonality between the eigenvectors in the noise space and the Greens function vectors, which are extractable from the background media. The following four improvements are expected by proposing the transmitting-mode algorithm. Firstly, the new conguration of arrays accurately increases the effective aperture so that the resolution could be further enhanced as compared to that of the echo-mode algorithm. Secondly, the degree of freedom in the choices (among the transmitter pseudo-spectrum, the receiver pseudo-spectrum or both) is apparently increased. Because of this and also the reciprocal mapping, the detection and surveillance coverage is enlarged.

0018-926X/$26.00 2011 IEEE

LIU et al.: TRANSMITTING-MODE TIME REVERSAL IMAGING USING MUSIC ALGORITHM

221

The vector wave equations satised by the electric eld and magnetic eld in a homogenous and isotropic medium have their following solutions: (1a) (1b)
Fig. 1. Physical conguration of simulation model.

In addition, the proposed method could overcome the problem that the echo-mode encounters where one target is atop another one in the upright direction of the transceiver antenna array even they are well- or far-separated. At last, the algorithm becomes now more exible and practical, especially for the highlighted applications to wireless sensor network. The detection targets are anisotropic in our analytical derivation, without any loss of generality. However, the isotropic targets are compatible in applications by alteratively dening the tensors of the electromagnetic parameters and . The remaining part of the paper is so organized subsequently. In Section II, the construction of the multistatic response matrix is discussed in detail from the electromagnetic scattering point of view. It is then followed by the investigation of the eigenvalue structure, whereas the MUSIC pseudo-spectrum is dened for the transmitting-mode TR-MUSIC algorithm. Then, the imaging simulation of targets in wireless sensor network is taken as a typical application and carried out in Section III. The results depict a good agreement with the theoretical expectation and it thus veries the improvement stated above. Section IV concludes the paper. II. THEORETICAL FORMULATIONS

where denotes the electric-electric dyadic Greens function between the eld position and the source position . The formulae lead to the incident electromagnetic elds expressed in terms of the driven currents on the transmitter dipoles as follows: (2a) (2b) where stands for the permeability of background medium, . Then, they could while be converted into a matrix form as follows: (3) and , and two matrices and , on where the two vectors, the left- and right-hand sides are explicitly dened respectively as

(4a) (4b) and

A. Multistatic Response and Time Reversal Matrix The analysis conguration of the physical model for the forward propagation, as shown in Fig. 1, is composed by a transmitter array, some desired targets, and a receiver array in homogenous background media, say, free space for most of the practical cases. The transmitter array and the receiver array, which were both assumed to be linear or planar in literature, are antenna elnow assumed to be general in distribution, where ements are located respectively at , while antenna elements are located at , , respectively. Each of the elements consists of three dipoles oriented in (with the -, - and -directions with lengths of , , and , ). Assume that the source at the th element can be , written as , , and stand for the driving currents at the where three-direction oriented dipoles of the th antenna element. Meanwhile, the scattering targets (which are assumed to be , anisotropic spheres with radii of , permittivity tensors of , respectively) are located at and permeability tensors of . the positions , where

(5a) (5b) At the same time, the sub-matrices in (5b), and

, are dened in (6), where their elements are determined by the dyadic Greens function at different locations. See (6), shown at the bottom of the next page. The derivation is carried out for anisotropic scatterers without any loss of generality. For completeness, the permittivity and permeability tensors are necessarily provided as in [19], [20] (7a)

(7b)

222

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

where (or ) represents the permittivity (or permeability) element aligned to the th electric (or magnetic) principal axis of the th scatterer, while (where , ) denotes the rotation transforming matrix and it is a function of the rotation , and . Euler angles Explicitly, it is dened that whose three respective angular contributions are described in terms of the , and by Euler angles

where

(10a)

(8a) (10b) (8b) Rewriting (9) in matrix form, one obtains (11) (8c) where the matrices are dened as

By considering the multiple scattering among the targets and using the Foldy-Lax equation, the relationship between the incident elds and the total elds nearby the targets could be obtained as

(12a) (12b) (12c) In (12c), the sub-matrices, , , and , are explicitly dened in (13), shown at the bottom of the next page. Therefore, the scattered electric elds at the receiver antenna (where ) are expressed elements located at in the vectorial summation form as

(9a)

(14) or in the matrix form as (9b) (15)

. . .

. . .

..

. . .

(6a)

. . .

. . .

..

. . .

(6b)

LIU et al.: TRANSMITTING-MODE TIME REVERSAL IMAGING USING MUSIC ALGORITHM

223

where

(16a) (16b) Then, the voltages induced on the antenna dipoles can be obtained in a vector form dened as follows: (17) where

(18a) (18b) At last, as a consequence of combining (3), (11), (15), and (17), one arrives at (19) Hence, the operator matrix (20) is referred to as the multistatic response matrix. Furthermore, the time reversal matrix is written as (21) where the operation denotes the conjugate transpose.

B. Eigenvalue Systems and Pseudospectrum Denition It was shown in literature that the DORT imaging can be performed using the back-propagation of the singular vectors associated to nonzero singular values of the and matrices. But it is not an effective-enough solution because this method requires that the targets are sufciently far separated from each other such that they can be resolved in imaging. In addition, the relationship between the individual target and the nonzero singular value is in an one-to-one manner in acoustics area. In electromagnetic applications, the relationship becomes more complicated due to polarizations, array congurations, shapes, electromagnetic characteristics of targets, and so on. Therefore, using the DORT method is inadequate for the high quality electromagnetic imaging. The TR-MUSIC algorithm is an enhanced solution for the electromagnetic detection and imaging. It is developed also based on the singular value decomposition of the matrix, which has the so-called signal subspace spanned by the background Greens function vectors evaluated at the target , locations. by considering the singular system it can be written in the following form: (22a) (22b) denotes the th singular value, while and stand where for the th column vectors of the orthonormal matrices and , respectively. Consequently, the so-called transmitter and receiver time reand , have the versal matrices, following eigenvalue system: (23a) (23b) and , The transmitter and receiver spaces are spanned by respectively. They can be both subdivided into signal space

(13a) . . . . . . . . . .. . . . .

(13b) . . . . . . . . . .. . . . .

. . .

. . .

. . .

..

. . .

(13c)

224

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

and noise space, respectively spanned by singular vectors corresponding to nonzero singular values and zero singular values. It is formulated as follows:

mode conguration to investigate the performance of the separated pseudo-spectrums corresponding to transmitter space and the receiver space, respectively. They are denoted by and , and expressed below

(24a) (24b) (24c) Meanwhile, the background medium Greens function vectors are the orthonormal bases of the signal space. Concretely, the Greens function vectors between locations of the transmitter array and the targets are bases of transmitter space, and the ones between locations of the targets and receiver array are bases of the receiver space. Therefore, the inner product of the background Greens function vectors of the target locations and singular vectors in noise space theoretically equals to zero, that is

(28a)

(28b) and It is notable that to locate all the targets, the values of must be larger than six times of the number of the targets since one target can be associated to at most six nonzero singular values [21][23], so that there exist zero singular values in the multistatic response matrix to perform the TR-MUSIC searching. In practice, the system requires more antenna elements, however, to achieve good performance due to the loss and noise. In addition, since the DORT method has the analysis processes similar to those of the TR-MUSIC method, it can be also extended to Maxwell equations. Therefore, in the subsequent numerical studies, the electromagnetic transmitting-mode DORT is also considered as a reference method.

(25a) (25b) where

III. NUMERICAL VALIDATION AND EVALUATIONS . . . (26a) In this section, we rst carry out a simulation of electromagnetic imaging in the wireless sensor network so as to verify the feasibility and applicability of the proposed electromagnetic transmitting-mode TR-MUSIC algorithm. Then, the performance comparison between the traditional DORT method and the proposed algorithm is made by performing the second numerical simulation to show the improvement in terms of resolution. The last simulation example deals with the far-eld detection where an obstacle comes up due to the multiple scatterers overlapped in the vertical direction of the transmitter aperture. The physical conguration of simulation model is shown in Fig. 1 where a 2-D imaging plane and a 3-D imaging domain are respectively set up for the 2-D and 3-D numerical examples in Section III-A. In addition, the additive white Gaussian noise is considered in all the simulations. This is done by contaminating the multistatic response matrix with the noise generated at certain level of signal to noise ratio (SNR), which was dened in [13], so will be not repeated herewith. A. Electromagnetic Transmitting-Mode TR-MUSIC Imaging , the location and imaging of By checking the peaks of the targets can be obtained numerically without any back propagation of the signals which is required in the time reversal mirror imaging. Furthermore, we dene the transmitter pseudospectrum and the receiver pseudo-spectrum in this transmittingThe 16 transmitter array elements and 32 receiver array elements are employed the analysis. The coordinates of the elements are denoted by where , with or 32 for the transmitter array and the receiver array,

. . .

(26b)

and and (where , 2, 3) denote the th column vectors of and (where ), respectively. of the transNow, one denes the pseudo-spectrum mitting-mode TR-MUSIC algorithm as follows:

(27)

LIU et al.: TRANSMITTING-MODE TIME REVERSAL IMAGING USING MUSIC ALGORITHM

225

Fig. 2. Two-dimensional detection and imaging results when SNR dB. (a) The geometrical distribution of the two scatterers on the imaging plane. (b) The transmitter pseudospectrum in (28a) of the transmitting-mode TR-MUSIC detection is utilized. (c) The receiver pseudospectrum shown in (28b) of the transmitting-mode TR-MUSIC detection is considered. (d) The transmitting-mode pseudospectrum shown in (27) for the transmitting-mode TR-MUSIC algorithm is implemented.

= 40

respectively. In this simulation, we have assumed: for transmitters

[13] for ease of validation and comparison. The Euler angles are set to be

and for receivers Meanwhile, the operation frequency is chosen to be at 100 MHz and the SNR is set to be 40 dB. Fig. 2 presents the detection and imaging results from simulations in 2-D view. Fig. 2(b)(d) employ the detection pseudospectrum dened in (28a) and (28b) and (27), respectively. It is shown that all of them can be used to accurately locate the scatterers even when the noise levels are high. The receiver pseudospectrum is found achieving relatively higher resolution since its focusing spot is much smaller than the transmitter pseudospectrum. This is due to the larger number of the antenna elements of the receiver array. If the scatterers are not on the as indicated in Fig. 1, one needs to imaging plane at perform the 3-D detection at the imaging domain to locate the

The position is randomly chosen while is also assumed. The three-direction oriented dipoles all have the same length of . Two anisotropic scatterers are and . They both have placed, respectively, at . The principle elements of the permittivity and a radius of permeability tensors are assumed to follow the same setup as in

226

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 3. Three-dimensional detection and imaging results when SNR dB. (a) The geometrical distribution of the two scatterers in the imaging domain. (b) The transmitter pseudospectrum shown in (28a) of the transmitting-mode TR-MUSIC detection. (c) The receiver pseudospectrum shown in (28b) of the transmittingmode TR-MUSIC detection. (d) The transmitting-mode pseudospectrum shown in (27) for the transmitting-mode TR-MUSIC algorithm.

= 40

scatterers. Fig. 3 presents the scatterers distribution and the detection results in 3-D view. In the 3-D simulations, three different denitions made in (28a) and (b) and (27), respectively are also considered. It is clearly seen that the scatterers are located with a good accuracy. Thus, it proves a good feasibility of the proposed transmitting-mode TR-MUSIC algorithm. Meanwhile, according to the different occasions and requirements of various applications, one can choose different algorithms to perform the detection and imaging. This actually demonstrates that the TR-MUSIC method is more exible and practical.

array are assumed. Similar to the previous setup, the coordinates of the elements are

and

B. Investigation on the Enhanced Resolution In our second simulation, the performance comparison of the existing time reversal imaging methods and the proposed method is made. The emphasis is made on the resolution of the imaging and the ability of locating the closely positioned targets. As mentioned in previous sections, one of the popular time reversal imaging methods is called DORT. Here, we also discuss its two modes, i.e., echo mode and transmitting-mode. Actually, the transmitting-mode DORT has not been reported elsewhere so far. It can be easily derived from the discussion in Section II due to the resemblance between DORT and TR-MUSIC. So the four algorithm in solving the Maxwell equations problems: echo-mode DORT, transmitting-mode DORT, echo-mode TR-MUSIC, and transmitting-mode TR-MUSIC, will be investigated in the following simulation. In the analysis using the transmitting-mode methods, 16 elements of the transmitter array and 16 elements of the receiver

coordinates are randomly chosen so that for transmitter elements and for receiver elements. All the three-direction oriented dipoles . For comparison, the above also have the same length of transmitter elements with duplex function are assumed in the echo-mode methods. In addition, the frequency is also chosen at 100 MHz but the SNR is changed to a lower value dB. Fig. 4 presents the imaging results of of SNR one anisotropic sphere, which is located at with a . It is shown that all the four methods correctly radius of locate the target sphere. The proposed transmitting-mode TR-MUSIC algorithm provides, however, the best resolution of the imaging, which can be judged by checking the cross-range of the focusing spots. Meanwhile, it can be seen that the transmitting-mode methods work is better than the echo-mode methods for both the DORT and TR-MUSIC. When the number of targets increases, these methods are all effective only in the cases where the scatterers are adequately well-separated. But if the distance between two adjacent targets are not large enough, the DORT method may rstly become invalid. This can be seen in Fig. 5 where three targets are included , and , and the locations are at respectively. Both the echo-mode TR-MUSIC method and transmitting-mode TR-MUSIC method are still working, but

while the

LIU et al.: TRANSMITTING-MODE TIME REVERSAL IMAGING USING MUSIC ALGORITHM

227

Fig. 4. Detection and imaging results for a scatterer located at (0; 3; 0) when SNR = 20 dB and at f = 100 MHz. (a) The echo-mode DORT algorithm. (b) The transmitting-mode DORT algorithm. (c) The echo-mode TR-MUSIC algorithm. (d) The transmitting-mode TR-MUSIC algorithm shown in (27).

the transmitting-mode has advantage of the higher resolution. Meanwhile, the resolution is also related to the number of the array elements. The imaging performance can be improved if the number of elements increases. It is, therefore, proved in this subsection that the enhanced resolution has been achieved for applications, and it is especially true for wireless sensor network using the proposed transmitting-mode TR-MUSIC method. C. Scenario of Limited Effective Array Aperture In the third numerical example, we perform a far eld EM detection and imaging to verify the advantage of the electromagnetic transmitting-mode TR-MUSIC over the echo-mode method in dealing with the special problem where one target is atop another one in the upright direction of the transceiver antenna array aperture even they are well-separated. The 60 elements of transmitter array and 60 elements of receiver array are placed in the far-eld zone in a similar fashion to the above examples except that the coordinates are randomly chosen for the transmitter elements; and within for the receiver elements. To increase difculty of the problem, the imaging plane shown in Fig. 1 is meanwhile rotated clockwise with respect to the -axis by 90 to the -plane. Four targets are considered in this numerical and example. Two of them are centered at , respectively. The other two scatters are overlapped in the direction and in the perpendicular direction of the approximate array aperture. Concretely, their locations are and , respectively. Their electric

and magnetic parameters are all set to be the same as those of the rst scatterer in Section II-A. This further increases the resemblance of the overlapped scatterers. The operation frequency MHz and the signal noise ratio is set to be is chosen as 50 dB. Fig. 6 depicts the imaging results of the echo-mode TR-MUSIC method, where the four scatterers are not readily detected due to the overlapping (in -direction) by two of these four scatterers, the decreased array aperture, and the smaller distance between the scatterers. Oppositely, the transmitting-mode method leads to precise true locations of all the four scatterers as shown in the Fig. 7. The problem that the echo-mode method encounters is effectively solved because of the increase in the equivalent array aperture by using the bistatic transmitter and receiver congurations. This is a common problem and it frequently appears in the wireless sensor network and also other applications. Therefore, this is seen as a remarkable improvement made by the proposed transmitting-mode TR-MUSIC algorithm. IV. CONCLUSIONS In this paper, the transmitting-mode TR-MUSIC algorithm is applied to Maxwell equations for electromagnetic detection and imaging problems. Taking the advantage of the released requirement of the sensor distributions, the proposed method is found to be very suitable for applications to wireless sensor network, where the antenna elements (or sensors) are randomly located and the arrays are nonlinear and/or planar. Besides this, the electromagnetic transmitting-mode TR-MUSIC is veried to have higher resolution imaging and to lead to more accurate location.

228

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 5. Detection and imaging results for three scatterers located at (0; 4; 0), (; 2; 0) and ( ; 2; 0) respectively when SNR = 20 dB and at f = 100 MHz. (a) The echo-mode DORT algorithm. (b) The transmitting-mode DORT algorithm. (c) The echo-mode TR-MUSIC algorithm. (d) The transmitting-mode TR-MUSIC algorithm shown in (27).

Fig. 6. Echo-mode TR-MUSIC algorithm for far eld detection and imaging of four scatterers located at ( 0:45; 0; 0:5), (0:45; 0; 0:5), (0; 0; 0:95) and (0; 0; 0:05) respectively when SNR = 50 dB, and at f = 100 MHz.

Fig. 7. Transmitting-mode TR-MUSIC algorithm for far eld detection and imaging of four scatterers located at ( 0:45; 0; 0:5), (0:45; 0; 0:5), (0; 0; 0:95) and (0; 0; 0:05) respectively when SNR = 50 dB, and at f = 100 MHz.

At meanwhile, more exible and practical conguration is believed to be very potential for many other EM applications. In addition, it effectively overcomes the problem the echo-mode encountered when one target is atop another one in the upright direction of the transceiver antenna array. The theoretical discussion is presented in detail from the view point of electromagnetic scattering and inverse scattering, by considering the multiple scattering among the targets. Then, the pseudo-spectrum for imaging using the proposed method is built up. The numerical studies convincingly justify our statements one by one in Section III. In practice, the multistatic response matrix is

obtained from the time-domain or frequency-domain measurement. The signal measurement could be carried out at one frequency point within a wide band or within an even ultra-wide band. If it is implemented in the latter case, the algorithm can be performed using multistatic response matrices of multiple frequency points. This will improve the performance of the algorithm in a complex environment. Because different targets with different electromagnetic characteristics show various responses, weaker or stronger, at different frequency points, this wide or ultra-wide band process can collect the responses at all frequencies and compensate them for all the targets to provide a better imaging. After the location of the targets is known, one

LIU et al.: TRANSMITTING-MODE TIME REVERSAL IMAGING USING MUSIC ALGORITHM

229

can employ a noniterative method based on the least squares method reported in [24] to retrieve the polarization strength tensors of the targets, where an efcient and stable estimation is shown. To sum up, the proposed electromagnetic transmitting-mode TR-MUSIC method is highly recommended to the superresolution imaging due to the exible conguration and the enhanced performance in many potential EM applications, especially in wireless sensors network, which is the primary objective of this paper.

ACKNOWLEDGMENT

X. F. Liu would like to thank the Department of Electrical and Computer Engineering at National University of Singapore (NUS), Singapore 117576 for hosting his visiting research engineer position at the NUS for one year.

[14] X. Chen and K. Agarwal, MUSIC algorithm for two-dimensional inverse problems with special characteristics of cylinders, IEEE Trans. Antennas Propagat., vol. 56, no. 6, pp. 18081812, Jun. 2008. [15] E. Iakovleva, S. Gdoura, D. Lesselier, and G. Perrusson, Multistatic response matrix of a 3-D inclusion in half space and music imaging, IEEE Trans. Antennas Propagat., vol. 55, no. 9, pp. 25982609, Sep. 2007. [16] X. F. Liu, B. Z. Wang, and L. W. Li, Trade off of transmitted power in time reversed impulse radio ultra-wideband communications, IEEE Antennas Wireless Propagat. Lett., vol. 8, pp. 14261429, 2009. [17] X. F. Liu, B. Z. Wang, S. Xiao, and S. Lai, Post-time-reversed MIMO ultrawideband transmission scheme, IEEE Trans. Antennas Propagat., vol. 58, no. 5, pp. 17311738, May 2010. [18] E. A. Marengo and F. K. Gruber, Subspace-based localization and inverse scattering of multiply scattering point targets, EURASIP J. Adv. Signal Process., vol. 2007, no. Article ID 17342, pp. 1616, 2007. [19] C. F. Bohren and D. R. Huffman, Absorption and Scattering of Light by Small Particles. Hoboken, NJ: Wiley, 1998. [20] G. B. Arfken and H. J. Weber, Mathematical Methods for Physicists, 5th ed. San Diego, CA: Harcourt/Academic, 2001. [21] D. H. Chambers and J. G. Berryman, Analysis of the time-reversal operator for a small spherical scatterer in an electromagnetic eld, IEEE Trans. Antennas Propagat., vol. 52, no. 7, pp. 17291738, Jul. 2004. [22] X. Chen, Time-reversal operator for a small sphere in electromagnetic elds, J. Electromagn. Waves Appl., vol. 21, pp. 12191230, 2007. [23] D. H. Chambers and J. G. Berryman, Time-reversal analysis for scatterer characterization, Phys. Rev. Lett., vol. 92, no. 2,023902, 2004. [24] X. Chen and Y. Zhong, A robust noniterative method for obtaining scattering strengths of multiply scattering point targets, J. Acoust. Soc. Amer., vol. 122, pp. 13251327, 2007.

REFERENCES
[1] M. Fink and C. Prada, Acoustic time-reversal mirrors, Inv. Probl., vol. 17, pp. R1R38, 2001. [2] H. C. Song, W. A. Kuperman, and W. S. Hodgkiss, Iterative time reversal in the ocean, J. Acoust. Soc. Amer., vol. 105, pp. 31763184, 1999. [3] D. Liu, G. Kang, L. Li, Y. Chen, S. Vasudevan, W. Joines, Q. H. Liu, J. Krolik, and L. Carin, Electromagnetic time-reversal imaging of a target in a cluttered environment, IEEE Trans. Antennas Propagat., vol. 53, no. 9, pp. 30583066, Sep. 2005. [4] C. Prada, J. L. Thomas, and M. Fink, The iterative time reversal process: Analysis of the convergence, J. Acoust. Soc. Amer., vol. 97, pp. 6271, 1995. [5] C. Prada, S. Manneville, D. Spoliansky, and M. Fink, Decomposition of the time reversal operator: Detection and selective focusing on two scatterers, J. Acoust. Soc. Amer., vol. 99, pp. 20672076, 1996. [6] J.-L. Robert and M. Fink, Greens function estimation in speckle using the decomposition of the time reversal operator: Application to aberration correction in medical imaging, J. Acoust. Soc. Amer., vol. 123, pp. 866877, 2008. [7] G. Micolau, M. Saillard, and P. Borderies, DORT method as applied to ultrawideband signals for detection of buried objects, IEEE Trans. Geosci. Remote Sens., vol. 41, no. 8, pp. 18131820, Aug. 2003. [8] M. Davy, J.-G. Minonzio, J. de Rosny, C. Prada, and M. Fink, Experimental study of the invariants of the time-reversal operator for a dielectric cylinder using separate transmit and receive arrays, IEEE Trans. Antennas Propagat., vol. 58, no. 4, pp. 13491356, Apr. 2010. [9] E. A. Marengo, F. K. Gruber, and F. Simonetti, Time-reversal MUSIC imaging of extended targets, IEEE Trans. Image Process., vol. 16, pp. 19671983, 2007. [10] A. J. Devaney, Time reversal imaging of obscured targets from multistatic data, IEEE Trans. Antennas Propagat., vol. 53, no. 5, pp. 16001610, May 2005. [11] A. J. Devaney, E. A. Marengo, and F. K. Gruber, Time-reversal-based imaging and inverse scattering of multiply scattering point targets, J. Acoust. Soc. Amer., vol. 115, pp. 31293138, 2005. [12] H. Ammari, E. Iakovleva, D. Lesselier, and G. Perrusson, MUSIC type electromagnetic imaging of a collection of small three-dimensional inclusions, SIAM J. Sci. Comput., vol. 29, pp. 674709, 2007. [13] Y. Zhong and X. Chen, MUSIC imaging and electromagnetic inverse scattering of multiply scattering small anisotropic spheres, IEEE Trans. Antennas Propagat., vol. 55, no. 12, pp. 35423549, Dec. 2007.

Xiao-Fei Liu received the B.S. degree in electrical engineering and the Ph.D. degree in radio physics from the University of Electronic Science and Technology of China (UESTC), Chengdu, in 2004 and 2010, respectively. From 2008 to 2009, he was a visiting research engineer in the Department of Electrical and Computer Engineering, National University of Singapore. Currently, he serves as a research engineer in the Nanjing Research Institute of Electronics Technology (NRIET), Nanjing, China. His research interests include antenna and active array technique, electromagnetic scattering, and ultrawide band communications.

Bing-Zhong Wang (M06) received the Ph.D. degree in electrical engineering from University of Electronic Science and Technology of China (UESTC), Chengdu, in 1988. In 1984, he joined the UESTC, where he is currently a Professor. He was a Visiting Scholar at the University of Wisconsin, Milwaukee; a Research Fellow at the City University of Hong Kong; and a Visiting Professor in the Electromagnetic Communication Laboratory, Pennsylvania State University, University Park. His current research interests include the areas of computational electromagnetics, antenna theory and techniques, electromagnetic compatibility analysis, and computer-aided design for passive microwave integrated circuits.

230

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Joshua Le-Wei Li (S91M92SM96F05) received the Ph.D. degree in electrical engineering from Monash University, Melbourne, Australia, in 1992. In 1992, he was a Research Fellow with Department of Electrical and Computer Systems Engineering at Monash University, sponsored by Department of Physics at La Trobe University, Melbourne. From 19922010, he was with the Department of Electrical and Computer Engineering, National University of Singapore, where he was a Professor and Director of NUS Centre for Microwave and Radio Frequency. In 19992004, he was seconded with the High Performance Computations on Engineered Systems (HPCES) Programme of Singapore-MIT Alliance (SMA) as a SMA Faculty Fellow. In May-July 2002, he was a Visiting Scientist in the Research Laboratory of Electronics at Massachusetts Institute of Technology (MIT), Cambridge; in October 2007, an Invited Professor with University of Paris VI, France; and in January and June 2008, an Invited Visiting Professor with Institute for Transmission, Waves, and Photonics at the Swiss Federal Institute of Technology, Lausanne (EPFL). He was appointed in 2009 as a QRJH Chair Professor at the University of Electronic Science and Technology of China under the Chinese Governments 1000-Talent Plan. His current research interests include electromagnetic theory, computational electromagnetics, radio wave propagation and scattering in various media, microwave propagation and scattering in tropical environment, and analysis and design of various antennas. In these areas, he has (co-)authored two books, namely, Spheroidal Wave Functions in Electromagnetic Theory (Wiley, 2001); Device Modeling in CMOS Integrated Circuits: Interconnects, Inductors and Transformers (Lambert Academic Publishing), 48 book chapters, over 320 international refereed journal papers, 49 regional refereed journal papers, and over 370 international conference papers. Dr. Li was a recipient of a few awards, including two best paper awards from the Chinese Institute of Communications and the Chinese Institute of Electronics, the 1996 National Award of Science and Technology of China, the 2003 IEEE AP-S Best Chapter Award when he was the IEEE Singapore MTT/AP Joint Chapter Chairman, and the 2004 University Excellent Teacher Award of National University of Singapore. He has been a Fellow of The Electromagnetics Academy since 2007 (member since 1998) and was IEICE Singapore Section Chairman between 20022007. As a regular reviewer of many archival journals, he was a Guest Editor of the IEICE Transactions on Electronics for ISAP2006 Special Section and the IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES for APMC2009 Special Section. He is an Associate Editor of Radio Science, the International Journal of Numerical Modeling, and the International Journal of Antennas and Propagation; an Editorial Board Member of the Journal of Electromagnetic Waves and Applications (JEWA), the book series Progress In Electromagnetics Research (PIER) by EMW Publishing, the International Journal of Microwave and Optical Technology, and the Electromagnetics Journal; and an (Overseas) Editorial Board Member of the Chinese Journal of Radio Science, Frontiers of Electrical and Electronic Engineering in China, and China Science: Information Sciences. He is an Advisory Professor at the State Key Laboratory of Electromagnetic Environments, Beijing (2002-present). He is a Guest Professor at both the Harbin Institute of Technology, Harbin (20032011), Shanhai Jiao-Tong University (20082011), and Southeast University, Nanjing (2004-), and an Adjunct Professor at both Zhejiang University, Hanzhou (20042006) and University of Electronic Science and Technology of China, Chengdu (20062010). He also serves as a member of various International Advisory Committee and/or Technical Program Committee of many international conferences or workshops, in addition to serving as a General Chairman of ISAP2006, MRS09-Meta09, and iWEM series (since 2011), Vice Chairman of PIERS2010 in Marrakech, and TPC Chairman of PIERS2003, iWAT2006, PMC2009, ISAPE2010, and ISAPE2012. He is appointed as an IEEE MTT-S Commission-15 Member in 2008, IEEE AP-S Region Representative (Region 10: Asia-Pacic) in 2010, and an IEEE AP-S Distinguished Lecturer in 2011.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

231

UWB Microwave Imaging of Objects With Canonical Shape

Navid Ghavami, Gianluigi Tiberi, David J. Edwards, and Agostino Monorchio, Fellow, IEEE
AbstractA novel technique for imaging based on ultrawideband (UWB) microwave signals is introduced. Specically, the procedure deals with the Huygens Principle (HP); using HP to forward-propagate the waves removes the need to solve inverse problems and, consequently, no matrix generation/inversion is required. Together with its simplicity, the methodology permits the capture of contrastthe extent to which different media can be discriminated in the nal image. Moreover, UWB allows all the information in the frequency domain to be utilized by combining the information from the individual frequencies to construct a consistent image. It follows that the methodology can identify the presence and location of signicant scatterers inside a volume. Validation of the technique through both simulations and measurements on cylinder and spheres with inclusions has been performed. Application of the proposed technique to medical imaging is envisaged. Index TermsBreast cancer detection, Huygens Principle (HP), microwave imaging.

I. INTRODUCTION

HERE is a considerable interest in the development of imaging methods for medical applications. Current medical imaging technologies are able to produce well-dened tomographic reconstructions of living tissues using a range of media. Most common are ultrasound scanners, X-ray-based computed tomography (CT), and the RF technique of nuclear magnetic resonance imaging (known as MRI). Each one of these techniques has its own drawbacks. For example, ultrasound is a very cost-effective technique and is extremely useful for selected tissues, but suffers from contrast problems and an inability to image objects with large acoustic impedance differences like air spaces and bones [1]. CT, on the other hand, is very effective for imaging of high-contrast subjects, but does result in a signicant dose of ionizing radiation, and thus has limited application [2]. Finally, a drawback of MRI is the scan time, which for a whole body can be as long as 2 h, during

Manuscript received July 09, 2010; revised March 29, 2011; accepted June 03, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by a Marie Curie Intra European Fellowship within the 7th European Community Framework Programme and by the U.K. Engineering and Physical Sciences Research Council. N. Ghavami and D. J. Edwards are with the Department of Engineering Science, University of Oxford, Oxford OX1 3PJ, U.K. (e-mail: navid.ghavami@sthughs.ox.ac.uk; david.edwards@eng.ox.ac.uk). G. Tiberi and A. Monorchio are with the Department of Information Engineering, University of Pisa, 56122 Pisa, Italy (e-mail: g.tiberi@iet.unipi.it; a.monorchio@iet.unipi.it). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167905

which time the patient is required to remain quite still within the connes of a rather cramped and claustrophobic environment. Moreover, concerning breast cancer detection, the limitations of conventional X-ray mammograms are well recognized. For example, this technique, even in high-resolution images, with relatively low radiation doses, misses approximately 15% of all cancers, while 75% of the identied breast lesions are in fact benign [3]. In response to all these limitations, microwave imaging has been recognized as one promising nonionizing and noninvasive alternative screening technology for these application areas. Microwave imaging has attracted growing attention in the last decade, especially for its applicability to breast cancer detection, motivated by the signicant contrast in the dielectric properties at microwave frequencies of normal and malignant tissues. Addressing an imaging problem involves estimating the distribution of the eld within a volume based upon observations of the eld on the surface of , and this involves estimating the distribution of dielectric properties within the target volume. Current research in microwave breast imaging can be divided mainly into microwave tomography and ultrawideband (UWB) radar techniques. In tomographic image reconstruction, a nonlinear inverse scattering problem is solved to recover an image of the dielectric properties in the breast [4], [5]. More recently, a new analytical approach (partly based on previous works [6], [7]) has been introduced by reformulating the inverse scattering problem as an inverse source one. The approach requires the calculation of both the radiating and the nonradiating parts of the induced currents and permits an efcient solution of 2-D problems [8], [9]. In contrast to tomography, the UWB radar approach solves a simpler computational problem by seeking only to identify the signicant scatterers [10][14]. In order to reconstruct the image, beamforming techniques of varying complexity are required. Current beamforming techniques use time-domain processing algorithms; one method is confocal imaging, which employs simple delay-and-sum beamforming [10], [11]. However, while confocal processing is computationally inexpensive, it presents major problems in imaging spaces that possess volumes of varying dielectric constant. In fact, internal refraction cannot be easily removed and would render the image inaccurate. An alternative technique, microwave imaging via space-time (MIST) beamforming [12], [13], uses lters that compensate for dispersion and fractional time delays. The lters solve a penalized least-squares problem such that signals originating from a candidate tumor location are passed with approximately unit amplitude and linear phase while the white noise gain is constrained. Filter designs improve clutter

0018-926X/$26.00 2011 IEEE

232

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

suppression and spatial discrimination, but they lead to an increase of complexity and computational burden. In [14], a data-adaptive beamforming is proposed, which accounts for some uncertainties during preprocessing steps, e.g., nonideal time delays, or errors in amplitude compensation. Time-reversal (TR)-based approaches have also been proposed in [15] and [16]. TR is based on the following assumption: If the scattered elds at the receiver are time-reversed and computationally reradiated into the medium, focusing is achieved at the source point. However, reradiation into the medium is possible if an approximate knowledge of the channel transfer function associated with the numerical back propagation is known. Further research and novel approaches need to be exploited to open new horizons to microwave imaging. In this context, the aim of this paper is to develop a novel, fast, and accurate UWB microwave imaging method. In particular, a method based on the principles of physical optics is introduced. The method deals with the Huygens Principle (HP): Using HP to forward-propagate the waves removes the need to solve inverse problems and, consequently, no matrix generation/inversion is required. Together with its simplicity, the methodology permits the capturing of the extent to which different tissues, or differing conditions of tissues, can be discriminated, and hence render contrast in the nal image. Moreover, UWB allows all the information in the frequency domain to be utilized by combining the information from the individual frequencies to construct a consistent image. It follows that the methodology can identify the presence and location of signicant scatterers inside a volume. Validation of the technique through both simulations and measurements on cylinder and spheres with inclusions has been performed, and results are presented, illustrating the effectiveness of the methods. Potential application would be breast cancer detection, internal organ imaging, and whole body imaging. This paper is organized as follows. In Section II, the HP-based procedure is introduced and validated through simulation, while a detailed characterization is presented in Section III. Validation through measurements is given in Section IV, considering a cylinder with single inclusion, a cylinder with multiple inclusions, and a sphere. Section V concludes the paper. II. HUYGENS-PRINCIPLE-BASED PROCEDURE A. Formulation Consider a cylinder in free space, having a radius , illuminated by a transmitting line source , and operating at a frequency . We assume that the dielectric properties of the cylinder, i.e., the dielectric constant and the conductivity , are known. The cylinder is envisaged to contain an inclusion (Fig. 1), here assumed to be cylindrically shaped and with a higher dielectric constant than . The problem consists of identifying the presence and location of the inclusion by using only the eld measured outside the cylinder. Suppose the eld at the surface points is known, i.e., with (1)

Fig. 1. Pictorial view of the problem, with the light and dark dots representing the receiving and transmitting positions, respectively.

Restating now the HP, we note that: Each locus of a wave excites the local matter which reradiates a secondary wavelets, and all wavelets superpose to a new, resulting wave (the envelope of those wavelets), and so on [17]. Therefore, we consider what happens if we apply the HP using as locus of a wave. With the aim in mind of exploiting such a strategy, we calculate the eld inside the cylinder as superposition of the elds radiated by the observation points of (1)

(2) where is the observation point, represents the wavenumber for the media constituting the cylinder, and is the spatial sampling. In (2), the string rcstr is used to indicate the reconstructed internal eld, while the string HP indicates that an HP-based procedure will be employed. Moreover, we note that the reconstructed eld will depend on the illuminating source and the frequency (since depends on these parameters). In (2), the Bessel functions for homogeneous media are used to forward-propagate the eld. However, when dealing with 3-D problems, Greens function can be used. Thus, we have

(3) where is the observation point. It follows that (2) can be used for 2-D problems (such as cylinders), while (3) represents the generalization to 3-D problems. It has to be emphasized that the HP procedure does not give the correct internal eld, even when dealing with homogeneous problems. This derives from the essence of HP, which is formulated for far-eld phenomena, i.e., it refers to a propagating wavefront. In this case, we are not interested in evaluating the internal eld, but we want to see if HP can capture the contrast (mismatch boundaries) and locate an inclusion within a volume (as in mode matching [18]). The capability of capturing the contrast is due to the difference in properties between the two media, which is not taken into account by (2) or (3). It follows that the reconstructed eld will exhibit a mismatch in the region of transition of the two media. This mismatch opens the way for a new strategy for detection and localization. Considering the problem in more detail, suppose the external cylinder

GHAVAMI et al.: UWB MICROWAVE IMAGING OF OBJECTS WITH CANONICAL SHAPE

233

Fig. 2. (a) Cast P.M.M.A. cylinder with a PEC rod. (b) Intensity (linear) obtained through (4a). - and -axes are in meters.

Fig. 3. (a) Cast P.M.M.A. cylinder without the PEC rod (homogeneous cast P.M.M.A. cylinder). (b) Intensity (linear) obtained through (4a). - and -axes are in meters.

is illuminated using a range of different frequencies and from different illuminating points; all the reconstructed elds will exhibit the mismatch, which will be always located in the region of transition of the two media. Thus, by summing incoherently all the solutions, i.e., the reconstructed elds, the inclusion will be detected and localized. Mathematically speaking, assuming we use transmitting sources with , and NF frequencies , it follows that the intensity of the nal image can be obtained through the following equations: (4a) (4b) and are the frequency sampling and the bandwidth, where respectively. The resolution is expected to achieve the optical resolution limit of , where represents the wavelength in the cylinder calculated at the highest frequency; this gives a rule of thumb for determining the highest frequency to be used. Finally, it is important to point out that the use of the Greens function implies a singularity for ; this singularity can be canceled by performing a multiplication between the same Greens function and . Moreover, when dealing with a medium with high losses, (2) and (3) can be used substituting with so as to compensate eld attenuation (which could mask the target feature). B. Validation Through Simulation Let us now consider a cylinder with a rod. In particular, we assume that the external cylinder has radius of 4.25 cm and is constituted by cast P.M.M.A. ( , S/m). The inclusion is a perfectly electrically conducting (PEC) cylinder having a radius of 3 mm, a radial offset of 1 cm, and an angular offset of 45 [see Fig. 2(a)]. One single transmitting line source illumination is considered, while the band spans from 8 to 10 GHz with frequency sample spacing of 10 MHz. For each frequency sample, the eld at points lying on the external surface was calculated analytically by resorting to [19] and [20]. To highlight that the eld refers to a cylinder with a rod illuminated by , we use the notation . Next,

Fig. 4. HP procedure applied to the difference : normalized intensity (linear) obtained through (4a). and -axes are in meters.

the HP procedure is applied to , and the image is obtained through (4a); a peak can be clearly detected [Fig. 2(b)]. This peak is assumed to be the image of the transmitter, and it masks the image of the rod, thus detection cannot be achieved. Next, we consider the same cast P.M.M.A. cylinder, but without an inclusion [see Fig. 3(a)], illuminated by the same single transmitting line source . HP is applied to , and the image is obtained through (4a). Again, the peak representing the image of the transmitter appears, supporting the assumption [see Fig. 3(b)]. The assumption is further supported by the observation that this feature moves as the position of the transmitter is varied. As the last step, we apply the HP to the difference: . Fig. 4 shows the normalized linear intensity after a proper image adjusting (consisting in enforcing to zero the values below 0.25 and expanding from 0 to 1 the values above 0.25) obtained through (4a). From the gure, it is possible to note that image of the transmitter has been canceled, and the peak representing the rod appears. The other peak probably corresponds to a mirror image, but its intensity is lower than 0.5.

234

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

In a practical situation, it is not possible to determine . However, it can be shown that the image of the transmitter can be removed also if applying the HP as follows:

TABLE I WITHIN-BREAST S/C AND RESOLUTION FOR BANDWIDTH OF 1, 2, 4, 6 GHz

(5a)

(5b) where represents the average of signals obtained illuminating the object using different transmitter positions. This in effect smears out the transmitter image; it can be shown that the minimum number of sources is required. The proposed HP-based procedure differs considerably from Kirchhoff migration methods [21], whose algorithms usually perform time reversal and back-propagation to nd the phase, i.e., time, traces. Conversely, here we are not interested in nding the phase traces. In fact, we use HP with the aim of reconstructing the eld. In the reconstruction process, a mismatch appears when going from a media to another, and the capability of capturing the mismatch boundaries represents the key feature of the proposed method allowing detection. Finally, it has to be pointed out that time-reversal and delay-and-sum algorithms, which do not require any inverse problem solution, perform a noncoherent summation in the time domain; here, instead, the noncoherent summation is performed in the frequency domain for each frequency component. This means that the approach allows all the information in the frequency domain to be utilized by combining the information from the individual frequencies to construct a consistent image. In Section III, it will be shown that this feature provides better performances if compared to conventional time-domain approaches. III. HUYGENS-PRINCIPLE-BASED PROCEDURE: METHODOLOGY CHARACTERIZATION In this section, a detailed characterization of the methodology is performed. For this purpose, we consider again a cylinder with a rod; in particular, we assume that the external cylinder has radius of 4.25 cm and consists of cast P.M.M.A. ( , S/m). The inclusion is a cylinder having a radius of 3 mm, dielectric constant , conductivity , a radial offset of 1 cm, and an angular offset of 45 . Note that the contrast between dielectric properties of external cylinder and inclusion is 1:5. The cylinder is illuminated by transmitting line sources with , located at a distance of 8 cm from the center and equally spaced along . The band 410 GHz is considered, and a frequency sample spacing of 5 MHz is used. For each illuminating source and for each frequency sample, the eld at points lying on the external surface is calculated analytically by resorting to [19] and [20]; note that the number of points leads to a spatial sampling of approximately , where represents the wavelength in the cylinder calculated at the highest frequency. The following points need to be addressed:

1) analysis of signal to clutter ratio and resolution with respect to the band; 2) investigating the optimum number of employed transmitting and receiving positions; 3) investigating the frequency sampling; 4) analysis of the performance with respect to uncertain knowledge of the dielectric properties of the target volume. Concerning the signal-to-clutter ratio (S/C), the following denition can be found in microwave breast cancer imaging literature [11]: The within-breast S/C ratio compares the maximum tumor response with the maximum clutter response in the same image. The maximum clutter is determined by locating the maximum pixel value outside of the volume containing the imaged tumor (dened by twice the extent of the full-width at half-maximum-FWHM response of the tumor). Thus, let us refer to the aforementioned denition. To evaluate the S/C with respect to the band, we apply (4) for different bandwidths. Specically, we perform image reconstruction assuming a bandwidth of 1, 2, 4, and 6 GHz, respectively. In this example, the lower limit of the band coincides with the lower frequency. Next, for each image the S/C is calculated. Table I shows the S/C for the bandwidths given above. It is important to highlight that, in HP procedure, the within-breast S/C approaches 8 dB for a bandwidth of 6 GHz. This value has been obtained with transmitter positions. Note that in [11], for a similar problem of 6-GHz bandwidth, a within-breast S/C of 4.1 has been obtained by employing the delay-and-sum beamforming algorithm. Next, the resolution with respect to the band is addressed. Specically, the resolution, here dened as the dimension of the region whose normalized intensity is above 0.5, is calculated assuming a bandwidth of 1, 2, 4, and 6 GHz, respectively. It can be shown (see Table I) that the resolution achieves the optical resolution limit of m, where represents the wavelength in the cylinder calculated at the highest frequency. From Table I, one could suggest to further increase the maximum frequency so to obtain a better resolution. However, it has to be pointed out that medical applications involve lossy materials, i.e., the penetration depth decreases if the frequency is increased. Turning now to the investigation of the optimum number of employed transmitting positions, it can be shown that the minimum number of illuminating sources is required to smear out the transmitter image [see (5)]. A lower number of transmitting positions leads to a lower S/C. It can be shown that a degradation of 4 dB occurs if using rather than

GHAVAMI et al.: UWB MICROWAVE IMAGING OF OBJECTS WITH CANONICAL SHAPE

235

Fig. 6. (a) Cast P.M.M.A. cylinder having a radius of 4.25 cm with PEC eccentric inclusion. (b) Measurement equipment.

IV. HUYGENS-PRINCIPLE-BASED PROCEDURE: MEASUREMENTS

Fig. 5. Normalized intensity along cut, i.e., the cut containing the inclusion, when assuming the dielectric constant in HP procedure equal to (dotted line), (continuous line), and (dashed line), respectively. The location of the peak intensity changes signicantly in the three cases here investigated. The inclusion is indicated with the stem.

A. Cylinder With Single Inclusion The capability of HP procedure to detect and locate an inclusion has been veried through measurements. In this context, a cast P.M.M.A. cylinder ( , S/m) having a radius of 4.25 cm and height of 30 cm was constructed. The cylinder has an eccentric inclusion consisting of a PEC cylinder with a radius of 3 mm; the distance between the centers of the two cylinders is approximately 2.1 cm [see Fig. 6(a)]. Frequency-domain UWB measurements in an anechoic chamber were performed using a vector network analyzer (VNA) arrangement to obtain the transfer function at 1601 discrete frequencies in the band of 610 GHz MHz ; this measurement only used the nominal output power level of the VNA (2 mW). Wideband bowtie antennas, vertically polarized and omnidirectional in the azimuth plane, were used [see Fig. 6(b)] after calibration. A careful adjustment was done for both antennas to be in vertical position to avoid any polarization mismatch. For each set of measurements, the location of the transmitting antenna was xed at approximately 20 cm away from the center of the cylinder, while the receiver antenna was positioned close to the external surface of the cylinder and placed on a computer-controlled rotating stage with 3 degrees of angular resolution. Thus, the eld at equally -spaced points lying on the external surface was measured; note that the number of points leads to a spatial sampling of approximately , where represents the wavelength in the P.M.M.A. cylinder calculated at the highest frequency (10 GHz). A total number of sets of data was recorded (with ), changing the position of the transmitting antenna along with a step of 30 [see Fig. 7(a)]. Next, the average of the 12 set is computed, i.e., , and HP is applied as illustrated in (5). Fig. 7(b) shows the normalized intensity (linear) obtained through (4a) after a proper image adjusting. A peak can be clearly detected in the region of the inclusion. Thus, both detection and localization are achieved. From the gure, it can be stated that the resolution is approximately 5 mm; this value is in agreement with the optical resolution limit of . The time required to perform all 12 sets of measurements is approximately 120 min, while the CPU time required by the

. Conversely, the S/C does not vary signicantly from that given in Table I if a lower number of receiving positions and a lower number of frequency samples are used. Specically, a spatial sampling of approximately , i.e., , and a frequency sampling of 20 MHz can be used without any degradation in terms of S/C. Finally, the characterization of the HP procedure with respect to uncertain knowledge of the dielectric properties of the target volume is addressed. We begin our analysis by evaluating the S/C when assuming the dielectric properties in the HP procedure equal to the following: 1) ; 2) ; while the dielectric properties in the analytical simulation remained xed to . Due to its very low value, is set to zero. In both cases, the S/C does not vary signicantly from that given in Table I, which refers to the matching case. However, if the dielectric properties in the HP procedure do not match the dielectric properties of analytical simulation, the inclusion is not accurately localized. Fig. 5 shows the normalized intensity along the cut, i.e., the cut containing the inclusion, when assuming the dielectric constant in the HP procedure equal to (dotted line), (continuous line), and (dashed line), respectively. The location of the peak intensity changes signicantly in the three cases here investigated. As expected, the gure shows that the inclusion (which is indicated with the stem) is most accurately localized when the dielectric properties used in analytical simulation are equal to those assumed in the HP procedure. Thus, the primary effect of mismatch between actual and assumed dielectric properties is a location bias. The same effect has been highlighted in [12]. It has to be pointed out that Fig. 5 has been obtained using 6 GHz of bandwidth and using an image adjusting (consisting in enforcing to zero the values below 0.25 and expanding from 0 to 1 the values above 0.25).

236

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 8. (a) Cast P.M.M.A. cylinder having a radius of 4.25 cm with PEC and agar-agar solution inclusions. (b) Pictorial view of the problem, with the dark and light circles representing the PEC and agar-agar solution inclusions, respectively.

Fig. 7. (a) Pictorial view of the problem, with the light and dark dots representing the receiving and transmitting positions, respectively. (b) Normalized intensity (linear) obtained through (4a). A peak can be clearly detected in the region of the inclusion. - and -axes are in meters.

HP procedure is approximately 10 min on an Intel Pentium M 2 GHz with 1 GB RAM. The high measurement time depends on the number of transmitting and receiving positions, which has been set for this example to and , respectively. In Section III, it has been shown that the S/C does not vary signicantly if a spatial sampling of is used in the receiving grid. It follows that a reduction of the measurement time of a factor 5 can be achieved. A further reduction can be achieved if employing an array conguration rather than a full mechanical scanning system, yielding to a measurement time comparable to that given in [14]. The CPU time required by the HP procedure depends on the number of transmitting and receiving positions and on the number of frequency samples. Referring again to Section III, it has been shown that the S/C does not vary signicantly if a spatial sampling of and a frequency sampling of 20 MHz is used. It follows that a reduction of the computational time of a factor 40 can be achieved. B. Cylinder With Multiple Inclusions To assess the applicability of Huygens principle in solving more complex problems, a medium with two inclusions was considered. A P.M.M.A. cylinder ( ,

S/m), with identical geometry to the one in the previous section but now with an extra inclusion, was constructed. The cylinder has an eccentric inclusion consisting of a PEC cylinder with a radius of 3 mm, and in addition, a second inclusion was placed by drilling a 3-mm-radius hole at 90 (relative to center) from the PEC and lling it with an agar-agar solution [Fig. 8(a)]. The distance between the centers of each inclusion and that of the outer cylinders is approximately 2.1 cm [Fig. 8(b)]. The agar-agar was dissolved in hot water at approximately 95 and then cooled to room temperature to form a semitransparent jell. The dielectric constant and the conductivity are functions of the concentration of the agar-agar [22]; for this experiment, we used a concentration of approximately 6%. Agar-agar gels are often used for approximating high-water-content human tissues. Frequency-domain UWB measurements in an anechoic chamber were performed, using a VNA arrangement to obtain the transfer function at 1601 discrete frequencies in the band of 68 GHz. Wideband bowtie antennas, vertically polarized and omnidirectional in the azimuth plane, were used. For each set of measurements, the location of the transmitting antenna was xed at approximately 20 cm away from the center of the cylinder, while the receiver antenna was positioned close to the external surface of the cylinder and placed on a computer-controlled rotating stage with 3 degrees of angular resolution. Thus, the eld at equally phi-spaced points lying on the external surface was measured. A total number of sets of data was recorded (with ), changing the position of the transmitting antenna along phi with a step of 60 . Next, the average of the six sets was computed, i.e., , and HP was applied as illustrated in (5). Fig. 9 shows the normalized intensity (linear) obtained through (4a) after a proper image adjusting. Two peaks can be clearly detected in the region of the inclusions, with different intensities. Specically, the peak with the higher intensity corresponds to the PEC cylinder (due to the fact that a greater contrast, i.e., greater mismatch boundaries, occurs when going from P.M.M.A. to PEC cylinder), while the less intense peak

GHAVAMI et al.: UWB MICROWAVE IMAGING OF OBJECTS WITH CANONICAL SHAPE

237

Fig. 9. Normalized intensity (linear) obtained through (4a). Two peaks can be clearly detected in the region of the inclusions. - and -axes are in meters.

Fig. 10. Cast P.M.M.A. sphere having a radius of 6 cm with PEC eccentric inclusion.

represents the location of the agar-agar solution cylinder. Hence, both detection and localization are achieved, showing the ability of the methodology in resolving multiple scatterers and paving the way for addressing problems involving tissues inhomogeneities as well. C. Extension to 3-D Problems From the previous examples, it follows that the proposed HP-based procedure can capture the contrast so to detect and locate an inclusion. Using HP to forward-propagate the wave removes the need to solve any inverse problem, thus no matrix generation/inversion is required. Moreover, extension to more realistic 3-D geometries is straightforward. In fact, for a given 3-D object, the eld can be determined in the external surface, and (5b) can be used to forward-propagate the eld so as to construct the corresponding image. The 3-D extension has been validated through measurements. In particular, a cast P.M.M.A. sphere having a radius of 6 cm, , S/m has been considered. An inclusion consisting of a PEC cylinder having radius of 3 mm and height of 6 mm has been positioned inside the sphere by drilling the sphere, putting the PEC cylinder into the hole, and relling the hole with liquid P.M.M.A. having dielectric properties similar but not equal to those of the sphere. Fig. 10 gives a pictorial view of the sphere used in the measurements. Frequency-domain UWB measurements in an anechoic chamber were performed, using a VNA arrangement to obtain the transfer function at 1601 discrete frequencies in the band of 610 GHz (frequency sampling of MHz). Wideband discone antennas, vertically polarized and omnidirectional in the azimuth plane, were used, after calibration. A careful adjustment was done for both antennas to be in vertical position to avoid any polarization mismatch. For each set of measurements, the location of the transmitting antenna was xed at approximately 20 cm away from the center of the sphere, while the receiver antenna was positioned close to the external surface of the sphere and placed on a computer-controlled rotating stage with measurements taken at intervals of 3 (Fig. 11). A total number of sets of data, , was recorded

Fig. 11. Pictorial view of the problem, with the circumference lying on the -plane representing the equatorial plane where the measures are recorded. To reproduce measurements along eight different equally spaced equatorial planes, the sphere is rotated with respect to the -axis, while the measures are recorded on the same -plane.

(with ). Each set of measurements differs from the previous ones as we used eight equally spaced equatorial planes to record the measurements, keeping both antennas in vertical position with respect to such planes. For simplicity, in order to produce measurements along eight different equally spaced equatorial planes, the sphere is rotated with respect to the -axis, while the data are recorded in the same -plane (see Fig. 11). Next, the average of the eight sets is computed, i.e., , and the HP procedure is applied as indicated in (5b). Fig. 12 shows a view of the normalized intensity (linear) obtained through (4b). A peak can be clearly detected, and its position corresponds to the position of the inclusion. The area surrounding the peak can be attributed to the difference in dielectric properties between the P.M.M.A. of the sphere and the liquid P.M.M.A. used to rell the hole. V. CONCLUSION A novel microwave imaging procedure based on the Huygens Principle has been introduced and analyzed. Using the HP procedure to forward-propagate the wave removes the need to solve inverse problems and, consequently, no matrix generation/inversion is required. Together with its simplicity, another benet

238

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 12. Normalized intensity (linear) obtained through (4b). A peak can be clearly detected in the region of the inclusion. -, -, and -axes are in meters. The intensity is plotted on three planes cutting the sphere, whose contour is given in the same gure.

of the methodology is the capability of capturing the contrast (dielectric variation). The method allows all the information in the frequency domain to be utilized by combining the information from the individual frequencies to construct a consistent image. This feature provides better performance when compared to conventional time-domain approaches; specically, the S/C reaches 8 dB. The procedure is robust to uncertainties in the medium, and it is capable of resolving multiple scatterers as well, with a resolution of approximately one quarter of the shortest wavelength in the dielectric medium. The power levels used are well within safety limits, while the bands satisfy the UWB rule of being at least 25% of the center frequencies. It follows that the methodology permits the detection and location of signicant scatterers inside a volume. Validation of the techniques through both simulations and measurements have been performed and presented, illustrating the effectiveness of the methods. It is worthwhile pointing out that the S/C given in Table I has been calculated using simulated data, i.e., noiseless data. A slight decrease in S/C can be observed when using measurement data containing noise. However, this decrease does not affect the detection and localization, as gathered from Section IV. Research is in progress for quantifying the S/C with respect to the contrast. An important issue that needs to be addressed in the future is the extension of the analysis to three-dimensional phantoms with realistic shape and realistic tissue architecture. This will permit the exploitation of the potential of the new methodology when applied to realistic medical imaging problems. REFERENCES
[1] C. R. Hill and G. R. Ter Harr, Review article: High intensity focused ultrasound-potential for cancer treatment, Brit. J. Radiol., vol. 68, no. 816, pp. 12961303, Dec. 1995. [2] Principle of Computerized Tomography, R. F. Cotellessa, Ed. et al. : IEEE Free Press, 1987. [3] Mammography and Beyond: Developing Techniques for the Early Detection of Breast Cancer. Washington, DC: Institute of Medicine, National Academy Press, 2000. [4] P. M. Meaney and K. D. Paulsen, Nonactive antenna compensation for xed-array microwave imaging: Part IIImaging results, IEEE Trans. Med. Imag., vol. 18, no. 6, pp. 508518, Jun. 1999.

[5] D. W. Winters, B. D. Van Veen, and S. C. Hagness, A sparsity regularization approach to the electromagnetic inverse scattering problem, IEEE Trans. Antennas Propag., vol. 58, no. 1, pp. 145154, Jan. 2010. [6] S. Kusiak and J. Sylvester, The scattering support, Commun. Pure Appl. Math., vol. 56, no. 11, pp. 15251548, 2003. [7] S. Kusiak and J. Sylvester, The convex scattering support in a background medium, SIAM J. Math. Anal., vol. 36, no. 4, pp. 11421148, 2005. [8] M. Chiappe and G. L. Gragnani, An analytical approach to the reconstruction of the radiating currents in inverse electromagnetic scattering, Microw. Opt. Technol. Lett., vol. 49, no. 2, pp. 354360, Feb. 2007. [9] G. L. Gragnani, Two-dimensional non-radiating currents for imaging systems: Theoretical development and preliminary assessment, Microw., Antennas Propag., vol. 3, no. 8, pp. 11641171, December 2009. [10] X. Li and S. C. Hagness, A confocal microwave imaging algorithm for breast cancer detection, IEEE Microw. Wireless Compon. Lett., vol. 11, no. 3, pp. 130132, Mar. 2001. [11] E. C. Fear, X. Li, S. C. Hagness, and M. A. Stuchly, Confocal microwave imaging for breast cancer detection: Localization of tumors in three dimensions, IEEE Trans. Biomed. Eng., vol. 4, no. 8, pp. 812822, Aug. 2002. [12] E. J. Bond, X. Li, S. C. Hagness, and B. D. Van Veen, Microwave imaging via space-time beamforming for early detection of breast cancer, IEEE Trans. Antennas Propag., vol. 51, no. 8, pp. 16901705, Aug. 2003. [13] X. Li, S. K. Davis, S. C. Hagness, D. W. van der Weide, and B. D. Van Veen, Microwave imaging via space-time beamforming: Experimental investigation of tumor detection in multilayer breast phantoms, IEEE Trans. Microw. Theory Tech., vol. 52, no. 8, pp. 18561865, Aug. 2004. [14] M. Klemm, I. J. Craddock, J. A. Leendertz, A. Preece, and R. Benjamin, Radar-based breast cancer detection using a hemispherical antenna arrayExperimental results, IEEE Trans. Antennas Propag., vol. 57, no. 6, pp. 16921704, Jun. 2009. [15] P. Kosmas and C. M. Rappaport, Time reversal with the FDTD method for microwave breast cancer detection, IEEE Trans. Microw. Theory Tech., vol. 53, no. 7, pp. 23172323, Jul. 2005. [16] Y. Chen, E. Gunawan, K. S. Low, S. Wang, Y. Kim, and C. B. Soh, Pulse design for time reversal method as applied to ultrawideband microwave breast cancer detection: A two-dimensional analysis, IEEE Trans. Antennas Propag., vol. 55, no. 1, pp. 194204, Jan. 2007. [17] P. Enders, Huygens principle as universal model of propagation, Latin Amer. J. Phys. Educ., vol. 3, no. 1, pp. 1932, Jan. 2009. [18] G. Tiberi, N. Ghavami, D. J. Edwards, and A. Monorchio, A mode matchingBessel functions based approach for UWB microwave imaging, in Proc. AP-S/URSI Symp., 2010, pp. 14. [19] R. P. Parrikar, A. A. Kishk, and A. Z. Elsherbeni, Scattering from an impedance cylinder embedded in a nonconcentric dielectric cylinder, Inst. Elect. Eng. Proc. Microw., Antennas Propag., vol. 138, no. 2, pp. 169175, 1991. [20] A. A. Kishk, R. P. Parrikar, and A. Z. Elsherbeni, Electromagnetic scattering from an eccentric multilayered circular cylinder, IEEE Trans. Antennas Propag., vol. 40, no. 3, pp. 295303, Mar. 1992. [21] X. Zhuge, T. G. Savelyev, A. G. Yarovoy, L. P. Ligthart, and B. Levitas, Comparison of different migration techniques for UWB short-range imaging, in Proc. EuRAD, 2009, pp. 184187. [22] K. Iizuka, An agar-agar chamber for study of electromagnetic waves in an inhomogeneous medium, IEEE Trans. Antennas Propag., vol. AP-19, no. 3, pp. 365377, May 1971.

Navid Ghavami received the B.Eng. degree (rst class honours) in electrical and electronic engineering from Kings College London, London, U.K., in 2006, and is currently pursuing the Ph.D. degree in communications at Oxford University, Oxford, U.K. His research interests include medical imaging and ultrawideband communications. Mr. Ghavami received seven university-wide prizes, including an IEEE prize for best nal-year project in the eld of telecommunications during his studies at Kings College London.

GHAVAMI et al.: UWB MICROWAVE IMAGING OF OBJECTS WITH CANONICAL SHAPE

239

Gianluigi Tiberi received the Laurea degree in telecommunication engineering and Ph.D. degree in information engineering from the University of Pisa, Pisa, Italy, in 2000 and 2004, respectively. In 2000, he was a Visiting Researcher with the Centre for Telecommunication Research, Kings College London, London, U.K. In October 2000, he joined the Microwave and Radiation Laboratory, Department of Information Engineering, University of Pisa. From September to December 2002, he was a Visiting Ph.D. Researcher with the Electromagnetic Communication Laboratory, Pennsylvania State University, University Park. From March 2009 to March 2010, he spent one year with the Department of Engineering Science, University of Oxford, Oxford, U.K., after receiving a Marie Curie Intra European Fellowship. He is currently conducting his postdoctoral research with the University of Pisa. His research interests include high-frequency derived approaches for solving electromagnetic scattering problems, electromagnetic propagation in complex environment, wideband/ultrawideband channel modeling, and microwave and magnetic resonance imaging for medical application. Dr. Tiberi, jointly with the Department of Engineering Science, University of Oxford, received a Marie Curie Intra European Fellowship for Career Development, Seventh Research Framework Programme (id. FP7-PEOPLE-2007-2-1IEF).

Prof. Edwards is a Fellow of the Institution of Engineering and Technology and the Royal Astronomical Society. He has received a number of awards and prizes for his research and has been extremely well supported by funding from grant-awarding bodies, industry, and government agencies. He has served on a number of national and international committees relating to the antennas and propagation elds.

David J. Edwards received the B.Sc. degree in physics, M.Sc. degree in physics of materials, and Ph.D. degree in engineering from Bristol University, Bristol, U.K., in 1973, 1974, and 1987, respectively. He is a Professor of engineering science with the University of Oxford, Oxford, U.K., and a Fellow of Wadham College, Oxford, U.K. He has been an academic for 26 years after having spent 12 years with British Telecom, London, U.K. He has a strong record of innovation in communications systems, techniques, and technologies. He is the author or coauthor of more than 300 publications. He holds a number of patents, several of which have appeared as licensed commercial products, and he serves as Founder Director of a startup company. He has also acted as a consultant to a number of industrial organizations. His current research interests include electromagnetics, magneto-inductive waveguides, ultrawideband communications, ad hoc networks, multiple-inputmultiple-output (MIMO) systems, radio astronomy technology, and medical imaging technology.

Agostino Monorchio (S89M96SM04F12) received the Laurea degree in electronics engineering and the Ph.D. degree in methods and technologies for environmental monitoring from the University of Pisa, Pisa, Italy, in 1991 and 1994, respectively. During 1995, he joined the Radio Astronomy Group, Arcetri Astrophysical Observatory, Florence, Italy, as a Postdoctoral Research Fellow in the area of antennas and microwave systems. He has been collaborating with the Electromagnetic Communication Laboratory, Pennsylvania State University (Penn State), University Park, and he is an Afliate of the Computational Electromagnetics and Antennas Research Laboratory. He has been a Visiting Scientist with the University of Granada, Granada, Spain, and with the Communication University of China, Beijing, China. In 2010, he was afliated with the Pisa Section of INFN, the National Institute of Nuclear Physics. He is currently an Associate Professor with the School of Engineering, University of Pisa, and Adjunct Professor with the Italian Naval Academy of Livorno, Leghorn, Italy. He is also an Adjunct Professor with the Department of Electrical Engineering, Penn State. He is on the Teaching Board of the Ph.D. course in Remote Sensing and on the council of the Ph.D. School of Engineering Leonardo da Vinci at the University of Pisa. He has been a reviewer for many scientic journals and has been supervising various research projects related to applied electromagnetics, commissioned and supported by national companies and public institutions. His research interests include the development of novel numerical and asymptotic methods in applied electromagnetics, both in frequency and time domains, with applications to the design of antennas, microwave systems and RCS calculation, the analysis and design of frequency-selective surfaces and novel materials, and the denition of electromagnetic scattering models from complex objects and random surfaces for remote sensing applications. Dr. Monorchio has served as Associate Editor of the IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS. He received a Summa Foundation Fellowship and a NATO Senior Fellowship.

240

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Experimental Characterization of an UWB Propagation Channel in Underground Mines

Youssef Rissa, Larbi Talbi, Senior Member, IEEE, and Mohamad Ghaddar
AbstractAn experimental characterization of the ultrawideband (UWB) propagation channel in an underground mine environment over the frequency range from 3 GHz to 10 GHz is reported in this paper. Two kinds of antennas, directional and omnidirectional, were used to investigate the effect of the antenna directivity on the path loss propagation and on the time dispersion parameters in both line-of-sight (LOS) and no-line-of-sight (NLOS) underground galleries. The measurement and simulation results show that the path loss exponents in an underground environment are larger than their counterparts in an indoor environment. In NLOS, the directional-directional (Direct-Direct) antenna combination showed better radiation efciency for reducing the time dispersion parameters while the omnidirectional-omnidirectional (Omni-Omni) case resulted better performance in term of path loss. After extracting the channel parameters, a statistical modeling of the UWB underground channel based on data measurements was conducted. Index TermsMultipath channel, ultrawideband (UWB) propagation measurement, underground mine.

I. INTRODUCTION HE ultrawideband (UWB) technology has shown to be attractive for high speed data transmissions, and may coexist with narrow band systems. In 2002, the Federal Communication Commission (FCC) proposed to dene a UWB device as any device where the (below the maximum power) fractional bandwidth is greater than 20% or occupies 500 MHz or more of spectrum [1]. The UWB frequency mask depends on the application and the environment where the devices are used. For indoor and outdoor communications, the power spectral density of is allowed in the frequency range of 3.110.6 GHz. Numerous narrowband and wideband channel measurements and models have been proposed in the literature for indoor and outdoor [2][5], and for underground mine environments [6][8] in the Canadian Center for Minerals and Energy Technology (CANMET) experimental mine in Val dOr, Canada.
Manuscript received January 13, 2011; revised May 24, 2011; accepted June 16, 2011. Date of publication October 03, 2011; date of current version January 05, 2012. This work was supported in part by the National Sciences and Engineering Research Council of Canada. Y. Rissa was with the University of Quebec in Outaouais, Gatineau, QC J9A 1L8, Canada. He is now with the Information Technology Branch, Canada Revenue Agency, Ottawa, ON K1A 1A2, Canada (e-mail: Youssef.rissa@craarc.gc.ca). L. Talbi and M. Ghaddar are with the Department of Computer Science and Engineering, University of Quebec in Outaouais, Gatineau, QC J9A 1L8, Canada (e-mail: larbi.talbi@uqo.ca; mohamad.ghaddar@uqo.ca). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167927

Furthermore, many studies on UWB communication measurements in indoor and outdoor environments have been published in the literature [9][11]. The UWB propagation in underground environments is still a relatively new research challenge. To date, only a limited amount of work has been done in conned environments with rough surfaces. In order to achieve a reliable UWB communications in conned environments, particularly, in underground mines, the investigation of data transmission at extremely high throughputs must be based on actual measurements of the channel propagation characteristics. The purpose of this paper is to report on the results of a UWB measurement campaign in underground mines with rough surfaces, in which different antenna directivities were considered to extract the UWB propagation characteristics. This paper is organized as follows. Section II provides a brief description of the underground mining environment and the measurement setup. Section III presents experimental data where the path losses along with the rms delay spreads have been evaluated in terms of the propagation distance and the antenna directivity. Finally, Section IV draws out the conclusion of this study. II. MEASUREMENT CAMPAIGN A. Description of the Underground Mining Environment The UWB measurements were performed in two typical galleries of a former gold mine at 70 m below the ground level. The mine mainly consists of very rough sidewalls surface that exhibits a difference of 25 cm between the maximum and minimum surface variations. The ground was not at and was sparse with puddles of water. The dimensions of the mine galleries vary between 2.5 m and 3 m in width and approximately 3 m in height. Fig. 1 illustrates LOS and NLOS underground galleries including transmit and receive antenna location arrangements. B. Experimental Procedure During the measurements, the transmitter antenna was kept at a xed position, while the receiver antenna was moved throughout the gallery on a 49 point grid. As shown in Fig. 2, the grid was arranged as 7 7 points with spacing of 5 cm between each adjacent point. The distance of 5 cm corresponds to one half of the wavelength of the lowest frequency component for uncorrelated small scale fading. However, to investigate the large scale path loss in the LOS gallery, the measurements were made by moving the receiver antenna from 1 m to 10 m away from that of the transmitter, with intervals of 1 m. In the NLOS gallery, the receiver antenna displacement was in the range 4 m to 10 m. The transmitting and receiving antennas

0018-926X/$26.00 2011 IEEE

RISSAFI et al.: EXPERIMENTAL CHARACTERIZATION OF AN UWB PROPAGATION CHANNEL IN UNDERGROUND MINES

241

Fig. 3. Overview of the measurement setup.

Fig. 4. Channel transfer function measured at

under LOS condition.

Fig. 1. Photos of the underground galleries and of the antenna locations (a) Gallery 1 (LOS condition) and (b) Gallery 2 (NLOS condition).

Fig. 2. The underground gallery plan.

were maintained at the same height of approximately 1.7 m, and the channel was kept stationary by ensuring that there was no movement in the surrounding environment. The UWB measurements were performed in the frequency domain using the frequency channel sounding technique based on the scattering parameter . The measurement setup consisted of an Agilent E8363B network analyzer (PNA) and two

different types of antennas, both are vertically polarized: a double bridged directional horn antenna DRG-118/A covering the frequency range from 1 to 18 GHz and an omnidirectional antenna, EM-6116, covering the frequency range from 2 to 10 GHz. Fig. 3 shows the experimental setup for UWB measurements. The transmitting port of the PNA swept 7000 discrete frequencies ranging from 3 GHz to 10 GHz which are uniformly distributed over the 7 GHz bandwidth. The receiving port measured then the magnitude and the phase of each frequency component. Thus, at each distance between the transmitter and the receiver, the complex channel transfer function (CTF) was measured 30 times to reduce the effects of the random noise on the data measurements, and then stored on a computer hard drive via a GPIB interface. The sweeping time of the network analyzer was decreased to validate the quasi-static assumption of the channel. Fig. 4 shows a typical complex CTF measured with the network analyzer. The theoretical delay resolution for the UWB signal is 143 ps, the inverse of the bandwidth, i.e., 1/BW and due to the use of windowing, the time resolution is estimated to be 2/BW. Consequently, the time axis was quantized into bins of 286 ps width which correspond to a bin size slightly greater than the pulse width and the frequency resolution of 1 MHz has provided a maximum delay range of , which was largely enough when compared to the highest maximum excess delay of all measurements. Before the measurements, a calibration of the equipment was performed in order to reduce the inuence of unwanted cables effects. Table I summarizes the systems setup parameters. Since the measurements were performed in the frequency domain, the inverse Fourier transform (IFT) was applied to the measured complex transfer function using a Kaiser window in order to obtain the channel impulse response or, in logarithmic scale, the power delay prole (PDP). The Kaiser

242

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I MEASUREMENT SYSTEM PARAMETERS

window with parameter was designed as a nite impulse response (FIR) lter to reduce the side-lobes of the PDP and to improve the dynamic range. For all the measurements, a predened threshold of was applied during the data processing. III. MEASUREMENT RESULTS AND ANALYSIS A. Path Loss Gradient Large scale measurements were performed to determine the propagation distance-power law in the underground mine. The average path loss in for arbitrary transmitter-receiver separation distance can be represented as

Fig. 5. Path loss vs. Tx-Rx separation distance (a) LOS and (b) NLOS.

(1) is the measured complex frequency response, where, represents the number of data points measured during a sweep of 7000 discrete frequencies ranging from 3 GHz to 10 GHz and is the number of sweeps that have been averaged. During this phase of analysis, the measured CTF data were tted using the linear least squares regression. Using different combinations of transmitting and receiving antennas, the computations have shown that the path loss at any location in the gallery can be described by a random lognormal (normal in ) distribution [12]:

TABLE II SUMMARY OF PATH LOSS EXPONENTS AND STANDARD DEVIATION (a) LOS and (b) NLOS

(2) is the path loss at the reference distance where, set to 1 m, is the path loss exponent and is a zero-mean Gaussian distributed random variable in dB with the standard deviation . At a distance of 1 m of antenna separation, is measured to be 48.65 dB, 23.82 dB and 40.28 dB using Omni-Omni, Direct-Direct, and Omni-Direct combinations

respectively. Hence, for each antenna selection, the path loss is evaluated as a function of distance in both LOS and NLOS galleries as shown in Fig. 5. It is important to underline the fact that the gain variation on 210 GHz range of each antenna has been measured and then removed from the experimental data. As listed in Table II, the path loss exponent in the LOS gallery is equal to 2.11, 2.01, and 1.99 for each antennas combination, respectively. It can be noted that the different path loss exponents for all these combinations are close to the free space path loss , where the smallest value is recorded in the Direct-Omni case. However, in the NLOS gallery, the path

RISSAFI et al.: EXPERIMENTAL CHARACTERIZATION OF AN UWB PROPAGATION CHANNEL IN UNDERGROUND MINES

243

TABLE III PARAMETER VALUES OF BEST FITTED DISTRIBUTION FUNCTIONS

Fig. 6. Statistical fading distribution of experimental data (a) LOS and (b) NLOS.

loss exponent with directive antenna is twice as large as that of the omnidirectional antenna due to the limited multipath signal reaching the directional antenna. The standard deviation of Gaussian random variable is lower for directional antenna in the LOS gallery. The path loss exponent values observed for indoor UWB propagation [13][17] are lower than the values obtained for underground UWB propagation. In the indoor environment, such as in a corridor or a hallway clear of obstacles, the results show a lower path loss exponent due to multipath signal contributions. In the mine, however, the walls irregularities and their roughness are very considerable, and hence, resulting multiple scattered rays that are severely attenuated before reaching the receiver. These rays superimpose the LOS, resulting in a closer arrangement with the free space path loss exponent. To this end, it is essential to describe the probability distribution that the signal will exceed a specied reference level. It is expected that Rice fading distribution is a suitable model when a direct LOS path is present between the transmitter and the receiver. While, Rayleigh statistics, which assumes the absence of any dominant multipath component will be the worst-case distribution. However, from the latest results generated by [18], the lognormal distribution is recommended to best t the measured amplitude fading in a UWB multipath channel. The cumulative distribution functions (CDFs) for both the LOS and NLOS cases

Fig. 7. Power delay prole for Omni-Omni scenario (a) LOS and (b) NLOS.

are depicted in Fig. 6. Also Table III shows the parameter values of experimental data best tted curves. B. Time Dispersion Characteristics The PDP shows the received signal power as a function of time delay, giving an intuition inspection of the multipath channel. Its statistical characterization is a useful process for describing the rapid uctuations of the amplitude, phase, and multipath propagation delays of UWB signals. The presence of multipath phenomenon in an underground mine is due to the numerous reections and scatterings from the ground and surrounding rough surfaces. Fig. 7 shows two typical power delay proles (PDP) of the channel, they are measured using

244

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE IV , MEAN, STANDARD DEVIATION AND MAXIMUM OF FOR DIFFERENT ANTENNA ARRANGEMENTS

AND

Fig. 8. Cumulative probability of the rms delay spread tted to Lognormal delay (a) LOS and (b) NLOS.

Omni-Omni combination under LOS and NLOS conditions respectively. In the LOS scenario, Fig. 7(a), the strongest power peak of the PDF corresponds to the rst observable path, followed by gradual smearing and exponential decrease in power multipath components with various propagation delays. However, in the NLOS scenario, Fig. 7(b), the absence of the direct path between transmit and receive antennas affected signicantly the PDP. One can observe three important sharp peaks each followed by a multipath cluster, the power decays exponentially with cluster decay as well as excess delay within a cluster. These clusters are resulted from multiple propagation mechanisms such as reection, diffraction and scattering from walls irregularities, metal pipes and other objects nearby the transmitter and the receiver. In the rst cluster of Fig. 7(b), the strongest path occurs 5 ns after the rst detected signal, and exhibits an increase in power of 6 dB compared to the latter. The main propagation mechanism of this cluster may be the specularly reected wave from metal pipes which are the only objects presented in

the channel. The second cluster appears at an excess delay of about 44 ns with a dominant component of , and last for about 26 ns. It may be caused by the interaction between the metal pipes and the opposite irregular wall. The third cluster, 3 dB weaker than the previous one, could be the result of diffraction from the corner edge shape appeared in Fig. 2 and surface irregularities close to the receiver. To compare different multipath channels of different antennas combinations, the channels dispersive parameters such as the mean excess delay , maximum excess delays and the square root of the second central moment of the PDP were evaluated [19]. The various channels delays are spanned between the rst crossing above the predened threshold level and the last crossing below it. The signal-threshold level denes the removal of the noise components and the calibration errors. The choice of threshold was considered to be relevant after several tests of thresholds by considering all experimental PDPs. In order to investigate the effect of antenna radiation pattern on the time dispersion properties of the channel, the CDFs of the rms delay spread were computed for the three antennas combinations and plotted in Fig. 8. According to the results shown in Fig. 8, one can observe that for 50% of all locations in the LOS gallery, the rms delay spread is less than or equal to 2.10, 7.75 and 9.55 ns for the Direct-Direct, Direct-Omni and Omni-Omni antennas combinations, respectively. It can be clearly seen that the radiation pattern of the directional antenna has a major advantage to reduce

RISSAFI et al.: EXPERIMENTAL CHARACTERIZATION OF AN UWB PROPAGATION CHANNEL IN UNDERGROUND MINES

245

TABLE V ESTIMATED COHERENCE BANDWIDTH AT 0.5 CORRELATION LEVEL DIFFERENT ANTENNA ARRANGEMENTS

FOR

This is apparent in Fig. 9, which shows the experimental and simulated CDFs in both LOS and NLOS galleries respectively. It is also worthy to evaluate the channel coherence bandwidth which is dened as the range of frequency components that have a strong potential for amplitude correlation. The coherence bandwidth is inversely proportional to the multipath rms delay spread. Based on the channel mean values of given in Table IV, and frequency correlation function above 0.5, the channel coherence bandwidth is approximately [12]:

(3) Table V shows the different values of the channel coherence bandwidth for different antennas combination. Obviously, LOS coherence bandwidths are signicantly greater than those of NLOS for both antennas arrangement, Omni-Omni and DirectDirect. Moreover, the latter arrangement, improves the coherence bandwidth, consequently, a maximum transmitted symbol rate in UWB underground channel is expected.
Fig. 9. Experimental and simulated CDF of rms delay spread (a) LOS and (b) NLOS.

IV. CONCLUSION Wireless communication systems in underground mines are becoming an increasingly important part of the mining industry, allowing connection between various electronic devices. UWB is a potential candidate to such wireless communication system. Thus, a characterization and modeling of the UWB propagation channel are presented in this paper. The basic characteristics of this channel are outlined based on an extensive measurement campaign. For LOS and NLOS underground galleries, the path loss exponents for different antennas combination were found higher than their corresponding in indoor environment. Moreover, the omnidirectional antennas offer a better signal coverage in NLOS gallery, while the directional antennas have an efciency effect to reduce the time dispersion parameters that lead to a better channel capacity. ACKNOWLEDGMENT The authors would like to thank the anonymous reviewers for their constructive and valuable comments and suggestions. The comments clearly helped to improve the manuscript. REFERENCES
[1] FCC, First Report and Order, Revision of Part 15 of the Commissions Rules Regarding Ultra-Wideband (UWB) Transmission Systems Report no. ET Docket 98-153, FCC, 2002 [Online]. Available: http://www.fcc.gov/Bureaus/Engineering_Technology/Orders/2002/fcc02048.pdf

the number of multipath components in the LOS underground gallery, and consequently maximizes the data transfer rate. In the NLOS gallery, for 50% of all locations the use of directional antenna can reduce the rms delay by 13.5 ns as compared to the omnidirectional antenna. A similar directivity effect can be observed on the mean excess delay and the maximum excess delay . The mean, standard deviation and maximum values of , and in both LOS and NLOS galleries have been computed from the time domain responses and are summarized in Table IV. The effects of antenna directivity in the underground LOS environment conform to the results reported in indoor channel [20], [21].

C. Time Dispersion Statistical Analysis Several distribution models have been used to model the statistical rms delay spread including lognormal, Nakagami, Rice and Weibull. First, the CDF of each distribution is obtained after estimating the maximum likelihood of the distribution parameters. Then the Kolmogorov-Smirnov test was used to identify the distribution which best represented the experimental results. The analysis showed that the best-t distribution of rms delay in LOS gallery can be the lognormal distribution however in a NLOS gallery Weibull appears to be the best tting distribution.

246

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[2] A. A. Saleh and R. A. Valenzuela, A statistical model for indoor multipath propagation, IEEE J. Select. Areas Commun., vol. 5, pp. 128137, Feb. 1987. [3] T. S. Rappaport, S. Y. Seidel, and K. Takamizawa, Statistical channel impulse response models for factory and open plan building radio communication system design, IEEE Trans. Commun., vol. 39, no. 5, pp. 794807, May 1991. [4] H. Hashemi, The indoor radio propagation channel, Proc. IEEE, vol. 81, no. 7, pp. 943968, Jul. 1993. [5] J. Kunisch and J. Pamp, Measurement results and modeling aspects for the UWB radio channel, in Proc. IEEE Conf. UWB Systems and Technol., May 2002, pp. 1923. [6] Y. P. Zhang, Y. Hwang, and R. G. Kouyoumjian, Ray-optical prediction of radio-wave propagation characteristics in tunnel environmentsPart2: Analysis and measurements, IEEE Trans. Antennas Propag., vol. 46, no. 9, pp. 13371345, Sep. 1998. [7] C. Nerguizian, C. Despins, S. Affes, and M. Djadel, Radio channel characterization of an underground mine at 2.4 GHz, IEEE Tran. Wireless Commun., vol. 4, no. 5, pp. 24412453, Sep. 2005. [8] M. Boutin, S. Affes, C. Despins, and T. Denidni, Statistical modelling of a radio propagation channel in an underground mine at 2.4 and 5.8 GHz, in Proc. IEEE 61st Vehicular Technology Conf. VTC 2005-Spring, Jun. 30, 2005, vol. 1, pp. 7881. [9] S. Ghassemzadeh, L. Greenstein, T. Sveinsson, A. Kavcic, and V. Tarokh, UWB indoor path loss model for residential and commercial environments, in Proc. IEEE Vehicular Technol. Conf (VTC 2003Fall), Orlando, FL, Sep. 2003, pp. 629633. [10] C. Chong and S. Yong, A generic statistical-based UWB channel model for high-rise apartments, IEEE Trans. Antennas Propag., vol. 53, no. 8, pp. 23892399, Aug. 2005. [11] A. Molisch, D. Cassioli, C. Chong, S. Emami, A. Fort, B. Kannan, J. Karedal, J. Kunisch, H. Schantz, K. Siwiak, and M. Win, A comprehensive standardized model for ultrawideband propagation channels, IEEE Trans. Antennas Propag., vol. 54, pp. 31513166, Nov. 2006. [12] T. S. Rappaport, Wireless Communications: Principles and Practice, 2nd ed. Englewood Cliffs, NJ: Prentice Hall, 2002. [13] R. C. Bernhardt, Macroscopic diversity in frequency reuse systems, IEEE J. Sel. Areas Commun., pp. 13561360, 1989. [14] A. Muqaibel, A. Safaai-Jazi, A. Attiya, B. Woerner, and S. Riad, Pathloss and time dispersion parameters for indoor UWB propagation, IEEE Trans. Wireless Commun., vol. 5, pp. 550559, Mar. 2006. [15] D. Cassioli, M. Z. Win, F. Vatalaro, and A. F. Molisch, Low complexity rake receivers in ultrawideband channels, IEEE Trans. Wireless Commun., vol. 6, pp. 12651275, Apr. 2007. [16] J. A. Dabin, A. M. Haimovich, and H. Grebel, A statistical ultra-wideband indoor channel model and the effects of antenna directivity on path loss and multipath propagation, IEEE J. Sel. Areas Commun., vol. 24, no. 4, pp. 752758, Apr. 2006. [17] L. J. Greenstein, S. S. Ghassemzadeh, S. Hong, and V. Tarokh, Comparison study of UWB indoor channel models, IEEE Trans. Wireless Commun., vol. 6, no. 1, pp. 128135, Jan. 2007. [18] S. S. Ghassemzadeh and V. Tarokh, UWB path loss characterization in residential environments, in Proc. IEEE Radio Frequency Integrated Circuits (RFIC) Symp., Jun. 810, 2003, pp. 501504. [19] J. Forester, Channel Modeling Sub-Committee Report Final, IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs), IEEE P802.15-02/490-SG3a 2003 [Online]. Available: http://grouper.ieee.org/groups/802/15/pub/2003/Mar03/02490r1P80215_SG3a-Channel-Modeling-Subcommittee-Report-nal.zip [20] H. Jeffrey, An Introduction to Ultra Wideband Communication Systems. Englewood Cliffs, NJ: Prentice Hall, 2005. [21] J. Dabin, A. M. Haimovich, and H. Grebel, A statistical ultra-wideband indoor channel model and the effects of antenna directivity on path loss and multipath propagation, IEEE J. Sel. Areas Commun., vol. 24, no. 4, pp. 752758, Apr. 2006.

Youssef Rissa received B.Sc. degree in physic electronic from the University of Rabat, Morocco, in 1996, the M.Sc.E. degree in electrical engineering from the University of Transport of Beijing, China, in 2002, and the M.Sc. degree in computer science and IT from the University of Quebec in Outaouais (UQO), Gatineau, QC, Canada, 2007. Previously, he joined the Department of Computer Science and Engineering, UQO. He was involved in a project supported by Telebec Inc. for the experimental characterization of the underground mine channel at the ultralarge band. Since 2008, he has been an IT Analyst and Developer in the Information Technology Branch, Canada Revenue Agency (CRA-ACR). Ottawa, ON, Canada.

Larbi Talbi (S95M97SM05) received the M.S. and Ph.D. degrees in electrical engineering from Laval University, Quebec City, QC, Canada, in 1989 and 1994, respectively. He completed a Postdoctoral Fellowship within the Personal Communications Research Group, INRS-Telecommunications, Montreal, QC, Canada (19941995), where he led projects supported by Bell-Canada. From 1995 to 1998, he was an Assistant Professor at the Electronics Engineering Department, Riyadh College of Technology, Saudi Arabia. During 19981999, he was an Invited Professor at the Electrical and Computer Engineering Department, University Laval, Quebec, Canada. Since 1999, he is a Professor with the Department of Computer Science and Engineering, University of Quebec in Outaouais (UQO), Gatineau, QC, Canada, where he is the Ph.D. Program Chair in Sciences and Information Technologies. In 2006, he was a Visiting Invited Professor with the Electrical and Electronics Engineering Department, Dumlupinar University, Turkey. His research activities include experimental characterization and modeling of UHF/EHF indoor radio propagation channels and design of antennas and microwave circuits for wireless communication systems. Currently he is actively involved in major projects related to the deployment of wireless technologies in Underground mines, mainly, experimental characterization of the underground mine channels using MIMO antennas at 60 GHz, design of microwave and RF components using SIW technique and metamaterials, antenna array for wireless applications. He has authored and coauthored over 200 journal and conference technical papers. Prof. Talbi is a member of the Ordre des ingnieurs du Qubec. He frequently serves as a technical program committee member of international and national conferences. He regularly acts as a Reviewer for many international scientic journals and conferences and also for research funding organizations.

Mohamad Ghaddar was born in Beirut, Lebanon, in 1967. He received B.Sc. degree in electrical engineering from Beirut Arab University (BAU), Lebanon, in 1991 and the M.Sc.E. degree in telecommunications from the Institut National de la Recherche Scientique (INRS), Montreal, Canada, in 2004. He is currently working toward the Ph.D. degree at the University of Quebec in Outaouais (UQO), Gatineau, QC, Canada. His main research interests concern wireless propagation measurements, modeling and signal processing, especially in the emerging area of 60 GHz wireless communications and UWB systems.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

247

Statistical Prediction of Site Diversity Gain on Earth-Space Paths Based on Radar Measurements in the UK
Chethana Nagaraja and Iok E. Otung
AbstractThis paper presents an empirical model for the statistical prediction of site diversity gain on earth space propagation paths. A modelling approach is described which incorporates 46 months of Nimrod rain radar data recorded in South England. The dual site diversity (SD) gain is developed as a function of several parameters, including time percentage of average year, site separation, baseline orientation, link frequency and path elevation angle. The model performs with an overall root mean square error of 1.87 dB computed over a wide variety of link congurations with elevation angles ranging from 10 to 50 and transmission frequencies from 16 to 50 GHz. The performance of the new model is checked against satellite beacon measurements as well as the empirical and physical models recommended by ITU-R. Index TermsFade mitigation, rain attenuation, satellite communication, site diversity, site diversity gain.

I. INTRODUCTION EVERE fading due to rain-induced impairments has led to serious consideration being given to the incorporation of site diversity in the design of high-availability satellite communication systems at Ka-band frequencies and above [1]. Site diversity takes advantage of the usually localized nature of intense rainfall by receiving the satellite downlink signal at two or more earth stations congured to minimize the likelihood of diversity stations being simultaneously subject to signicant rain attenuation. The rain impact on the communication system is reduced by drawing on the signal received via the less attenuated path. This reduction is often quantied using a parameter referred to as site diversity gain, which is the difference between the equi-probable attenuations exceeded in a single link system and a system employing site diversity. Various empirical and physical models for the prediction of site diversity gain have been reported in the literature. Determination of the single site and joint site attenuation distributions estimated from the knowledge of spatial rainfall occurrence forms the basis for physical models [2][5], [7]. ITU-R P.618-10 [7] incorporates physical model based on log-normal

distribution of rain rate and rain attenuation to predict joint probability of dual site attenuation. Empirical models on the other hand involve mathematical regression analysis performed on long term measurements. Hodge [6] proposed the rst empirical prediction of site diversity gain as a product function of individual gains contributed by single site attenuation, site separation, baseline orientation, link frequency and path elevation angle. This model has been adopted by ITU-R [7] for the estimation of gain up to a site separation of 20 km, while a different model by Alnutt et al. [8] provides gain prediction for separations greater than 20 km. Goldhirsh [9] dened SD gain as an aggregate function of single site attenuation and site separation. A recent model proposed by Panagopoulos et al. [10] predicts the gain using a slightly different regression t obtained along the lines of [6]. The aforementioned models have been validated for link frequencies operating up to 30 GHz. Moreover, majority of the dataset used for regression analysis amongst the empirical models are limited to measurements of up to 20 GHz. We propose a novel regression analysis procedure for the prediction of dual SD gain as an ensemble of multi-parametric function incorporating outage probability and extended EHF frequencies covering from 16 GHz to 50 GHz.

II. DATASET The required spatial rain measurements were obtained from the UK Met ofce Nimrod system which processes data from a network of 15 C-band rainfall radars covering the UK mainland region. The radar located at Chenies was chosen, which performs ground based scans every 5 min, processed to yield rain rate record on a 100 100 Cartesian grid of 2- km resolution. The data period used in our study covered 46 months from April 2004. A reference site for the main earth station was set at OS), from Grid location (45 900, 24 100) m or (52.064 , 1.14 which the elevation and azimuth to a geostationary satellite at 13.2 are 30 and 162 , respectively. The reference site is located at a ground height of 89 m above the mean sea level and the associated rain height was found to be 2.4 km [11]. The corresponding diversity stations were chosen to achieve site separations covering from 4 km to 38 km aligned along xed baseline orientations spanning from 0 to 90 . This resulted in ve

Manuscript received November 30, 2009; revised April 26, 2011; accepted May 16, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the Department of Electronics, University of Glamorgan, Pontypridd CF37 1DL, U.K. (e-mail: nagaraja_chethana@hotmail.com; ieotung@glam.ac.uk). Digital Object Identier 10.1109/TAP.2011.2167896

0018-926X/$26.00 2011 IEEE

248

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 1. Cumulative distribution of the data compared with ITU-R predictions, (a) Rain rate distribution and Rain attenuation distributions at (b) 16 GHz, (c) 20 GHz, and (d) 50 GHz.

groups of diversity stations corresponding to 5 baseline orientations, namely 0 , 26 , 45 , 71 and 90 . Rain rate interpolation was performed where necessary to maintain a constant spacing of 2 km between the diversity earth stations for the benet of uniform regression analysis. The nal set-up thus provided 18 earth stations in each of the ve groups giving a total of 90 diversity sites. III. RAIN RATE, RAIN ATTENUATION AND SITE DIVERSITY GAIN DISTRIBUTIONS The rain rate distribution at the reference site is compared with the ITU-R [12] predictions in Fig. 1(a). Overall, there is a satisfactory agreement between the measurement and the predictions with root mean square error percentage of about 4%. The ITU-R appears to overestimate the rain rate exceeded at 0.01% by about 22.6%. This was also inferred from radar rain measurements for a similar rain zone [13], where the prediction error was observed to be about 21.5% at 0.01%. The rain attenuation at the reference and the diversity earth stations was computed in line with ITU-R [7] procedure for 45 hypothetical slant paths with frequencies in the range 16 GHz50 GHz and elevation angles from 10 50 as shown in Table I. The slant path rain attenuation using ground based radar data has been deduced in line with the procedure outlined in [14], (mm/hr) recorded at time t for each [15]. The rain rate pixel location indicating the reference and diversity earth stations were used to compute the specic attenuation ( , dB/km)

TABLE I FREQUENCY AND ELEVATION ANGLE OF 45 SLANT PATHS

Fig. 2. Cumulative distributions of diversity gain for dual sites at 10 km separation and 0 baseline orientation.

given by (1), where and are the link frequency dependent coefcients dened in ITU-R [16]. The total slant path rain attenuation at time t ( , dB) was determined using (2), where

NAGARAJA AND OTUNG: STATISTICAL PREDICTION OF SITE DIVERSITY GAIN ON EARTH-SPACE PATHS

249

Fig. 3. Regression analysis of g(p, s,

).

is the effective slant path length at time t. was evaluated using the calculations of the horizontal and vertical reduction factors outlined in ITU-R [7] (1) (2) distributions with the ITU-R atThe comparison of tenuation predictions are illustrated for 6 hypothetical paths in Fig. 1(b)(d). In all cases, the ITU-R model appears to over estimate the attenuation for time percentages down from 0.005%. It can also be seen that the distributions at 10 path elevation angles are under estimated by ITU-R for higher outage probabilities spanning from 0.1% to 2% in Fig. 1(c) and 0.05% to 5% in Fig. 1(d). The CDF of SD gain (G, dB) was determined using the statistical distribution of the single site (A, dB) and joint site (J, dB) attenuation as shown in (3) (3) in (3) was obtained using the joint site attenuation time series ( , dB) dened as (4) and are the instantaneous rain attenuation where values at the reference and the diversity station, respectively. The diversity gain distributions thus obtained are shown in

TABLE II TIME PERCENTAGES

Fig. 2 at a site separation of 10 km and 0 baseline orientation for varying slant path congurations indicated within Fig. 2. IV. MODELLING PROCEDURE A. Stage 1 Using the approximate median values from Table I, the reference slant path frequency and elevation angle were chosen to be 20 GHz and 30 , respectively, for initial regression analyses. The rst stage involved the formulation of diversity gain g(p, s, ) as a function of outage probability (p), site separation (s) at xed link parameters of 20 GHz and baseline orientation and 30 . Table II shows the 28 values. The regression analyses involved in modelling g(p, s, ) are shown in Fig. 3. The regression t of diversity gain as a function of along with standard error is demonstrated in Figs. 4 and 5. The standard error was evaluated as rms percentage error [17] using (5)

(5)

250

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 4. Non-linear regression t of diversity gain as a function of time percentage at xed , and (d) . (c)

and

and (a)

, (b)

Fig. 5. Non-linear regression t of diversity gain as a function of time percentage at xed , and (d) . (c)

and

and (a)

, (b)

By comparing the data plots in Figs. 4(a) and 5(a), it can be found that improved diversity gain can be obtained at 90 baseline orientation. This is in line with the prevailing NorthWestern wind direction which is inclined to 90 baseline ori-

entation and thereby minimizes the probability of simultaneous rain fading at the dual sites. It is noted from Figs. 4 and 5 that regression t (5) (see Fig. 3) for time percentages below 0.003% do not follow the

NAGARAJA AND OTUNG: STATISTICAL PREDICTION OF SITE DIVERSITY GAIN ON EARTH-SPACE PATHS

251

ii) For iii) For iv) For ;

; . , . , . , , : ;

Fig. 6. Non-linear regression t of diversity gain as a function of link param, . and . eters at

. and in (iv) were The coefcients of link parameters modied through iterative regression of nal expression dB to achieve minimum rms error. V. PERFORMANCE EVALUATION

data pattern where diversity gain values continue to increase as the time percentages decrease. Hence it was decided to x 0.003% as the lower limit of time percentage for model validation purposes. B. Stage II The second stage of the modelling procedure consisted of incorporating the link parameters in the statistical gain prediction. The link frequency f was regressed using the diversity gain values at 30 elevation angle, 0.005%, 10 km separation and 45 baseline orientation for the 9 frequency values mentioned in Table I. Equation (6) tted the data plot with a standard error 6.16% as shown in Fig. 6(a) (6) was The regression analysis for the elevation angle obtained using the diversity gain data at 20 GHz frequency, 0.005%, 10 km separation and 45 baseline orientation. This resulted in (7) which tted the data with a standard error of 3.25%. The data and the t are shown in Fig. 6(b) (7) C. Final Expression of Dual SD Gain (G) The regression ts obtained in stages I and II were then merged to establish the nal composite function of the form (8) Depending on the values of the parameters in (8), the new model involves the following steps in predicting the statistical dual SD gain: , : i) For ; ;

The rms error was determined as the difference between predicted and data-derived SD gain for the 26 values from 0.003% to 1%. These rms errors averaged using (9) over the 45 hypothetical slant paths at each of the 90 diversity stations was found to be 1.87 dB

(9) Since the modelling procedure included interpolated data, the predictions were tested against real measurements. The new model performance are compared with direct SD gain predictions provided by ITU-R and Panagopoulos [10] models as well as SD gain derived from joint probability of dual site attenuation predicted by ITU-R under section 2.2.4.1 in [7]. These tests along with comparisons to the ITU-R and Panagopoulos [10] models for applicable frequency range of up to 30 GHz are shown in Figs. 710. The rms error (dB) values are summarized in Tables III, IV, and V at 10 , 30 and 50 elevation angles, respectively. The comparison of model predictions at link frequencies greater than 30 GHz with the SD gain derived from ITU-R predictions of joint probability of dual site attenuations are shown in Figs. 1113 and the associated rms error (dB) values are indicated in Table VI. It should be noted that the rms errors of the new models predictions spanning between 1 dB and 2 dB at link frequencies above 20 GHz are acceptable in line with the large magnitude of the diversity gain achieved at higher frequency bands between 20 GHz and 50 GHz. It can be observed that the new models performance is comparable with the recommended ITU-R model to estimate SD performance via joint probability of dual-site attenuation. Moreover the new model provides improved predictions at 10 elevation angle. Dual site diversity experiments performed using GBS satellite beacon at 20.7 GHz [18] were used to compare the model predictions. Measurements were recorded using two ) beacon receivers located at Chilbolton (51.1445 , 1.437 and Sparsholt (51.0838 , 1.3908 ). The receiver dynamic

252

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 7. Comparison of the statistical diversity gain predicted by new model, ITU-R and Panagopoulos et. al., [10] with data at 16 GHz and (a) , and (b) , , .

Fig. 8. Comparison of the statistical diversity gain predicted by new model, ITU-R and Panagopoulos et. al., [10] with data for (a) , and (b) , , , .

Fig. 9. Comparison of statistical diversity gain predicted by the proposed model, ITU-R and Panagopoulos et. al., [10] with data at 20 GHz and (b) , , and (c) , , .

range was 13 dB which limited the measurements distributions for the corresponding exceedance time percentages below 0.02% [18]. The diversity gain is compared for the ref-

erence site at Chilbolton whose elevation angle and azimuth to the satellite beacon at 23 are 28 and 207 , respectively.

NAGARAJA AND OTUNG: STATISTICAL PREDICTION OF SITE DIVERSITY GAIN ON EARTH-SPACE PATHS

253

Fig. 10. Comparison of statistical diversity gain predicted by the proposed model, ITU-R and Panagopoulos et. al., [10] with data at 30 GHz and (a) , , (b) , , , and (c) , , .

TABLE III SD GAIN PREDICTION RMS ERRORS BY NEW MODEL, ITU-R AND PANAGOPOULOS [10] MODELS AT 10 ELEVATION

TABLE IV SD GAIN PREDICTION RMS ERRORS BY NEW MODEL, ITU-R AND PANAGOPOULOS [10] MODELS AT 30 ELEVATION

TABLE V SD GAIN PREDICTION RMS ERRORS BY NEW MODEL, ITU-R AND PANAGOPOULOS [10] MODELS AT 50 ELEVATION

The site separation and baseline orientation between Chilbolton and Sparsholt are 7.5 km and 53 , respectively. The statistical diversity gain extracted from one year data from 01/2004 is compared with the new model in Fig. 14. It can be seen that the new model gives an excellent t for time percentages between 0.03% and 1%. The observed divergence

below 0.03% was due to the receivers limited dynamic range of 13 dB. VI. CONCLUSION A new empirical model for the statistical prediction of dual site diversity gain has been developed using 46 months of exten-

254

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 11. Comparison of statistical diversity gain predicted by model with data at 10 elevation and (a) , , (c) , , , and (d) ,

, , .

, (b)

Fig. 12. Comparison of statistical diversity gain predicted by model with data at 30 elevation and (a) , , (c) , , , and (d) ,

, , .

, (b)

sive rain radar data from Nimrod. The novelty is contributed by incorporating exceedance time percentage and very high EHF bands comprising link frequencies from 16 GHz to 50 GHz. The performance validation of the new prediction method was carried out over a broad arrangement of diversity stations covering

438 km site separation range and baseline orientations from 0 to 90 . Comparisons with ITU-R and Panagopoulos. et al., empirical predictions showed that the proposed model provided a better performance for slant paths operating up to 30 GHz. Further more the rms error of the new model estimations at very

NAGARAJA AND OTUNG: STATISTICAL PREDICTION OF SITE DIVERSITY GAIN ON EARTH-SPACE PATHS

255

Fig. 13. Comparison of statistical diversity gain predicted by model with data at 50 elevation and (a) , , (c) , , , and (d) ,

, , .

, (b)

TABLE VI NEW SD MODEL PREDICTION ERROR RMS VALUES AT FREQUENCIES ABOVE 30 GHz

Fig. 14. Comparison of SD gain predicted by the new model with GBS beacon , , and . data at

ITU-R recommended model [7] to estimate SD performance. Overall the proposed model was found to perform satisfactorily with averaged rms error of 1.87 dB for signicantly varying congurations of slant paths comprising link frequencies from 16 to 50 GHz and elevation angles from 10 to 50 at each of 90 diversity sites. The model evaluation using the satellite beacon data at 20.7 GHz and 28 elevation angle showed very good agreement for time percentages down to 0.03% beyond which the measurements underestimated the diversity gain due to signicantly limited receiver dynamic range of 13 dB. Since the new model has been derived and validated using rain data from temperate climatic region, further comparisons would be required to test the model performance for different climatic regions. REFERENCES
[1] T. Wrede, ASTRA broadband interactive system: A Ka-band return path for IP connectivity, presented at the IEE Coll. Multimedia Services and Digital Television by Satellite, London, UK, 1999.

high link frequencies of 40 GHz and 50 GHz ranged up to only about 2 dB. Moreover SD performance prediction errors of the new model were found to be close to rms errors provided by the

256

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[2] J. Mass, A simulation study of rain attenuation and diversity effects on satellite links, COMSAT Tech. Rev., vol. 17, no. 1, pp. 159186, 1987. [3] R. K. Crane and H. C. Shieh, A two component rain model for the prediction of site diversity performance, Radio Sci., vol. 24, no. 6, pp. 641665, 1989. [4] E. Matricciani, Prediction of site diversity performance in satellite communications systems affected by rain attenuation: Extension of the two layer rain model, Eur. Trans. Telecommun., vol. 5, no. 3, pp. 327336, 1994. [5] A. V. Bosisio and C. Riva, A novel method for the statistical prediction of rain attenuation in site diversity systems: Theory and comparative testing against experimental data, Int. J. Satellite Commun., vol. 16, pp. 4752, 1998. [6] D. B. Hodge, An improved model for diversity gain on earth-space propagation paths, Radio Science, vol. 17, no. 6, pp. 13931399, 1982. [7] Propagation Data and Prediction Methods Required for the Design of Earth-Space Telecommunication Systems Recommendations of the ITU-R, Rec. P.618-10, 2009. [8] J. E. Allnut and D. V. Rodgers, Novel method for predicting site diversity gain on satellite-to-ground radio paths, Electron. Lett., vol. 18, no. 5, pp. 233235, 1982. [9] J. Goldhirsh, Space diversity performance for earth-satellite paths using radar modelling techniques, Radio Sci., vol. 17, no. 6, pp. 14001410, 1982. [10] A. D. Panagopoulos, P. D. M. Arapoglou, J. D. Kanellpoulos, and P. G. Cottis, Long term rain attenuation probability and site diversity gain prediction formulas, IEEE Trans. Antennas Propag., vol. 53, no. 7, pp. 23072313, Jul. 2002. [11] Rain Height Model for Prediction Methods Recommendations of the ITU-R, Rec. P. 839-3, 2001. [12] Characteristics of Precipitation for Propagation Modelling Recommendations of the ITU-R, Rec. P.837-5, 2007. [13] K. S. Paulson, R. J. Watson, and I. S. Usman, Diversity improvement estimation from rain rate radar databases using maximum likelihood extimation, IEEE Trans. Antennas Propag., vol. 54, no. 1, pp. 168174, Jan. 2006. [14] C. Nagaraja and I. E. Otung, Spatial rain correlation analysis for application in site diversity design, Radio Sci., vol. 44, 2009, 10.1029/ 2008RS003841. [15] I. E. Otung and C. Nagaraja, Seasonal variation of site diversity performance, Electron. Lett., vol. 45, no. 14, 2009. [16] Specic Attenuation Model for Rain for Use in Prediction Methods Recommendations of the ITU-R, Rec. P.838, 2005. [17] I. E. Otung and A. Savvaris, Estimating tropospheric scintillation intensity on earth-space propagation paths, Electron. Lett., vol. 42, no. 7, pp. 381381, 2005.

[18] S. A. Callaghan, B. Boyes, A. Couchman, J. Waight, C. J. Walden, and S. Ventouras, An investigation of site diversity and comparison with ITU-R recommendations, Radio Sci., vol. 43, pp. 18, 2008.

Chethana Nagaraja received the B.E degree (with Distinction) from Visveswaraiah Technological University, Karnataka, India, and the M.Sc. (with Distinction) and Ph.D. degrees in electrical engineering from University of Glamorgan, Wales, U.K, in 2009. She is an Assistant Professor teaching courses in wireless network communications at Herguan University, Sunnyvale, CA. Her current research interests include satellite and wireless communications signal propagations and physical channel modelling.

Iok E. Otung received the B.Sc. (First Class Honors) and M.Sc. degrees in electronic and electrical engineering from the University of Ife Nigeria and the Ph.D. degree in satellite communications from the University of Surrey, Surrey, U.K. He is currently with the University of Glamorgan in Wales, U.K., where he teaches satellite and digital communications courses, supervises Ph.D. degree work, conducts satellite link measurements on the Glamorgan campus, and manages an M.Sc. degree program in Mobile and Satellite Communications that attracts around 80 international students each year. His research interests include characterization and prediction of propagation impairments on earth-space paths, design of fade mitigation techniques for EHF satellite communication systems, development of bandwidth-efcient modulation and coding schemes, data analysis and signal processing in communication systems, capacity optimization for wireless communication networks, and performance of smart antennas in realistic propagation scenarios. He maintains an active relationship with industry through knowledge transfer projects (KTP), consultancy work, short courses for practicing engineers, and free 16-week projects carried out under supervision by M.Sc. degree students on a problem of interest to the company. He is the author of 110 publications, including several textbooks and in particular the 664-page Communication Engineering Principles (Palgrave, 2001), which is used in leading engineering departments around the world. Dr. Otung is a Chartered Engineer in the U.K.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

257

Wideband Characterization of Backscatter Channels: Derivations and Theoretical Background

Daniel Arnitz, Member, IEEE, Ulrich Muehlmann, Member, IEEE, and Klaus Witrisal, Member, IEEE
AbstractThe wireless channel of backscatter radio systems is a two-way pinhole channel, created by the concatenation of two standard wireless channels. We present a method to calculate wideband channel parameters of backscatter channels based on the parameters of the constituent one-way channels. The focus is on characteristics that are vital for narrowband and wideband ranging, such as the K-factor w.r.t. the direct (line-of-sight) path and the RMS delay spread. The presented analyses include uncorrelated as well as correlated channel pairs and are thus valid for bistatic and monostatic antenna setups. We also show that the uncorrelated scattering (US) assumption holds for the backscatter channel provided that the constituent channels are US. Index TermsAverage excess delay, backscatter channels, K-factor, multipath channels, parameter approximation, RMS delay spread, wideband parameters.

I. INTRODUCTION
Fig. 1. Combination of individual channels (uplink) to the backscatter channel (downlink) and .

INHOLE channels are created by the concatenation of the channel from the transmitter to the pinhole and the channel from the pinhole to the receiver. In (semi-)passive ultra-high frequency radio-frequency identication (UHF RFID), a prime example for backscatter radio systems, the pinhole is formed by the RFID tag, as illustrated in Fig. 1. The overall backscatter channel between TX and RX is composed of the two individual channels to and from the tag, which are both multipath radio channels. Pinhole (a.k.a. keyhole, dyadic, or double fading) channels are quite well-understood in terms of fading statistics for narrowband signals [2][5] and channel capacity for multiple-input multiple-output (MIMO) systems [6][8]. Even though channel capacity and fading statistics are fully sufcient to analyze the performance of communication over a channel, they are insufcient for the analysis of narrowband and wideband indoor ranging systems (with the exception of receive-power-based methods). The performance of these systems depends on the shape of the power-delay-prole (PDP), and thus on shape parameters like the K-factor w.r.t.
Manuscript received December 20, 2010; revised April 07, 2011; accepted June 20, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. Part of this work was presented at the European Wireless Conference, Vienna, Austria, Apr. 27-29, 2011 [1]. This work was supported by NXP Semiconductors, Gratkorn, Austria and the Austrian Research Promotion Agency (FFG) under Grant 818072. D. Arnitz and K. Witrisal are with the Signal Processing and Speech Communication Laboratory, Graz University of Technology, 8010 Graz, Austria (e-mail: daniel.arnitz@tugraz.at; witrisal@tugraz.at). U. Muehlmann is with NXP Semiconductors, 8101 Gratkorn, Austria (e-mail: ulrich.muehlmann@nxp.com). Digital Object Identier 10.1109/TAP.2011.2167923

the line-of-sight (LOS) and the root-mean-square (RMS) delay spread. We present a generic expression for the average power-delayprole (APDP) of pinhole channels based on the APDPs of the constituent channel pair, both for correlated and for uncorrelated individual channels, and a closed-form expression for exponential APDPs. A similar expression for the generic APDP is derived in [9], however this result is limited to orthogonal channels. We also present approximations for common shape parameters of the pinhole APDP, such as K-factor w.r.t. the , RMS delay spread , and avLOS component , based on the parameters of the indierage excess delay vidual channels, cf. Fig. 1. These formulas can be used to calculate backscatter channel parameters from point-to-point measurements of the constituent channel(s), e.g, [10][14]. This paper is organized as follows: A generic expression for the APDPs of backscatter channels is derived in Section II, while a closed-form solution for the typically exponential APDP of short-range indoor channels is introduced in Section III. Section IV deals with the K-factor w.r.t. the LOS path, while expressions for the RMS delay spread and the average excess delay of the backscatter channel are derived in Sections V and VI, respectively. Validation for exponential APDPs and a discussion of robustness w.r.t. nonexponential decays can be found in Section VII, and a measurement-based validation is provided in Section VIII. The Appendix contains an investigation of the uncorrelated scattering (US) assumption for the combined backscatter channel.

0018-926X/$26.00 2011 IEEE

258

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

II. STATISTICAL DERIVATION OF THE BACKSCATTER AVERAGE POWER-DELAY-PROFILE We dene the time-invariant complex baseband-equivalent channel impulse responses of the band-limited channels to and from the pinhole by uniformly sampled tapped delay lines [15] (1) (2) where is the tap interval, is the LOS component of , is the -th reected (non-line-of-sight, NLOS) component at delay , and is the Kronecker delta. All channels are causal, thus and , and have been normalized to have their LOS component at . This normalization has no effect on the results, as the analyzed parameters are shift-invariant.1 Time-invariance is assumed for notational simplicity, although only short-time wide-sense-stationarity (WSS) is required. Also the sampled representation is chosen purely for notational simplicity; we will use continuous-time representations in other parts of this paper. Both representations are equivalent due to the band limitation of the channels. In the tapped delay line model, each tap represents a large number of physical paths. As a consequence of the central limit theorem [16] each NLOS component in the complex baseband representation can be modeled as zero-mean circulant-symmetric complex Gaussian (ZMCSCG) random variable. Also the sampled LOS component will be random to some degree even in direct line of sight because of indirect physical paths close to the LOS path that are mapped to the LOS component in the sampled model. In the ideal case, which we will assume for now, the LOS component has a constant amplitude and uniform phase. We will discuss the inuence of random LOS amplitudes at the end of this section (3) (4) (5) The individual multipath components are assumed to be uncorrelated (US), cf. [15], [17], [18], i.e., (6) As a rst step, the combined channel is obtained by convolution of the two constituent channels

The rst term in (7) is the LOS component of the combined pinhole channel. All the other components form the NLOS part. We will focus our derivation on the NLOS part and rewrite it in vector notation in order to remove the delta pulses and further simplify the notation. By dening (8) (9) (10) for the random NLOS part of the channel impulse responses (CIRs), and the convolution matrix

. . .

. . .

. . .

..

(11) (12)

we can write the NLOS part of the backscatter CIR as

The APDP for WSSUS channels is the second moment (autocorrelation) of the CIR, cf. [15], [17], [18]. A proof for the validity of the US assumption for the backscatter channel can be found in the Appendix. We will denote the APDP by in the vector notation and in the continuous-delay notation, and start by writing down the NLOS part of the backscatter APDP

(13) where denotes the element-wise multiplication of vectors. Using the zero-mean property of and , we can dene the NLOS APDPs of the individual channels

(14) (15) and calculate the expected values of the rst two summands in (13) (16) The next term

(7)
1The K-factor is a power ratio and does not depend on the delay axis at all. The RMS delay spread is a central moment of the APDP, hence the average time shift is irrelevant. The average excess delay is dened relative to the LOS delay and thus shift-invariant.

(17) . . .

ARNITZ et al.: WIDEBAND CHARACTERIZATION OF BACKSCATTER CHANNELS

259

contains fourth-order cross-terms between all channel taps. It can be shown that [19]

Finally, it is straightforward to show that the last two remaining terms of (13) are zero due to the zero-mean property of , , and (25) (26) Hence, the NLOS part of the backscatter APDP is

(18) . Using for zero-mean complex Gaussian variables (18), we can split the fourth-order terms into auto- and cross-correlation terms of the CIRs

(19) and , can easily be The autocorrelation terms, obtained from the denitions in (4)(6). For the two cross-correlation terms we know from the Cauchy-Schwarz-inequality that (20) and that equality holds if and are fully correlated [20]. We will limit the derivation to zero correlation (ideal bistatic setup) and full correlation (monostatic setup), and thus provide bounds for full/no correlation. Note that full correlation implies that the channels are linearly dependent, and thus in every practical sense . Hence

uncorr. identical

(27)

From (27), we can deduce two major conclusions: 1) For uncorrelated constituent channels, the backscatter APDP is the convolution of the individual APDPs. Proof: By separating the APDPs of the channels into LOS and NLOS parts (28) where APDPs of is the Dirac delta, the convolution of the and (29)

and and

uncorr. identical

(21)

results in (30) and

By applying this approach, it is slightly tedious but straightforward to show that uncorr. identical with

(22) (31) This decomposition is also illustrated in Fig. 2. Switching to vector notation and inserting (14), (15), and (23) into (31), we obtain for the NLOS part of the combined channel, which matches (27) for uncorrelated channels. 2) For fully correlated constituent channels, the NLOS part of the backscatter APDP is multiplied by a factor of two compared to two independent but identically parameterized constituent channels. Proof: With identical constituent channels, , (27) reduces to (32) while for two uncorrelated channels with we get and (33)

(23) . . . . . . . . . .. .

All the remaining terms in (13) contain as least one of the non-Gaussian RVs and , hence (18) does not immediately hold for these terms. However, by extracting the random phase of and we can exclude the remaining constant amplitudes from the expectation operators. The phase shifts are absorbed by the NLOS channel taps without any effect due to their circular symmetry. It follows that:

uncorr. identical

(24)

An example is presented in Fig. 3.

260

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

channels, the power of the Gaussian LOS part will be doubled in the backscatter APDP, just like the NLOS part. As a theoretical limiting case, the entire APDP is doubled if the sampled LOS component is composed entirely of reected paths. Again, this is not limited to specic types of APDPs. III. BACKSCATTER AVERAGE POWER-DELAY-PROFILE FOR EXPONENTIALLY DECAYING CHANNELS Indoor average power-delay-proles are typically exponential due to rich scattering (e.g., [21], [22]). We will now calculate a closed-form indoor backscatter APDP, separated into LOS and NLOS parts

(34) For the channels to and from the pinhole, assume short-range indoor APDPs [23] and , we (35)
Fig. 2. Decomposition of the constituent channels into LOS and NLOS parts (for uncorrelated channels). This compares to the structure of (30) and (31), as well as (37) and (38), and also illustrates the steps in calculating the pinhole . channels LOS K-factor,

(36) where is the LOS power, is the NLOS power density, and is the NLOS decay constant. Using these denitions, we obtain (37) and

(38) This is the result for uncorrelated channels. Using the conclusions from Section II, we know that coherent combining of the NLOS part will double the power (density) of (38). Expressions that link , , and the overall power with , , and are provided by [23] along with the APDP (39) (40) To sum up, the APDP of the backscatter channel can be calculated via convolution of the APDPs of the individual channels if these channels are uncorrelated. This matches the results in [9] and is also similar to the Kronecker product for the overall correlation matrix in [7]. For fully correlated channels, the NLOS part of the backscatter APDP is merely multiplied by a factor of two; the general shape stays the same. This is not limited to a specic type of APDP and thus applies to all pinhole channels. The same reasoning can be applied to the LOS component if the assumption of a constant amplitude does not hold. Due to the linearity of the models the LOS component can be split into a component with constant amplitude and uniform phase and an additive ZMCSCG random variable. Both random variables are uncorrelated due to the US assumption. For fully correlated (41)

Fig. 3. Simulation-based example showing the doubling of NLOS power due to coherent combining. Both constituent channels are exponentially decaying, and , and are with either identical (monostatic) or independent (bistatic). The shown APDPs are instantaneous PDPs. averaged over

IV. K-FACTOR W.R.T. THE LINE-OF-SIGHT COMPONENT The K-factor w.r.t. LOS between LOS and NLOS power only depends on the ratio (42) and is thus independent of the actual shape of the APDP and the . overall power The simplest way to calculate of the backscatter channel, , is based on the decomposition of each

ARNITZ et al.: WIDEBAND CHARACTERIZATION OF BACKSCATTER CHANNELS

261

individual channel into its LOS and NLOS components, as illustrated in Fig. 2. Using

V. RMS DELAY SPREAD Unlike for the K-factor, the RMS delay spread of the backscatter channel cannot be calculated completely independent of the shapes of the APDPs of up- and downlink channel. More precisely, calculating solely based on and works for uncorrelated channels to and from the pinhole, and for correlated channels with low K-factors. For correlated channels with high K-factors though, the individual shapes and K-factors have to be taken into account, as will be shown below. We will start by formally dening the RMS delay spread, followed by a calculation of for uncorrelated channels. We will then extend this result to correlated channels and combine both cases in one equation. Finally, we will specify this equation for exponential APDPs. A discussion of the robustness of these results for nonexponential APDPs and instantaneous PDP can be found in Section VII. A. General Denitions

(43) the power of the direct (LOS) path after the rst channel is (44) At this point, the signal passes the pinhole and is sent over the second channel. It is important to note that both impulse responses, and , are direct representations of physical channels and thus causal per necessity. Hence

(45) Using (43) again, the overall K-factor w.r.t. the LOS path is (46) So far, weve treated each channel as an individual, thus implicitly assuming independent channels. According to the results in Section II, full correlation between the channels will add a factor of two to the NLOS power. The LOS component, on the other hand, ideally has constant amplitude and is thus not changed by the coherent combining (see Fig. 3). As a consequence, is reduced by the same factor of two according to (42). Combining these observations, the K-factor of the combined backscatter channel is (47) where stituent channels represents correlation between the conand and uncorr. identical

The RMS delay spread is the square root of the second central moment of the normalized APDP [15]

(49) , its only inuence on As the LOS component is at this formula is via the normalization term . We can thus again split the APDP into LOS and NLOS parts and skip the LOS term in the numerators (50) with the average excess delay (51)

B. From Uncorrelated to Correlated Channels The RMS delay spread for uncorrelated constituent channels can easily be obtained using a similarity to statistics: If two independent random variables are added, their distribution functions are convolved and the variances, i.e., second central moments, add up. It is shown in Section II and [9] that the APDPs of two uncorrelated channels are convolved as well. Hence (52) Extending this analogy, the variance of the sum of two arbitrary random variables and is [20] (53) where (54)

(48)

even Note that a stochastic LOS amplitude will cause in monostatic setups, see Section II, last paragraph. From (47) it can be seen that the combined backscatter channel always has a lower K-factor than each individual channel, and that the difference is considerably more severe for low K-factors. Assuming and uncorrelated channels, for example, asymptotically approaches for , and for . As a consequence, the LOS component is only dominant on the backscatter link if both constituent channels have extremely high K-factors. Channel correlation additionally reduces the backscatter K-factor.

262

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

and uncorr. identical. (55) We can thus assume that the RMS delay spread for fully correlated channels will be similar to (56) based on (53) and the analogy of squared RMS delay spreads and variances. Unfortunately, this is as far as the analogy goes. Unlike for uncorrelated channels, the backscatter APDP for correlated channels is not identical to the convolution of the singlechannel APDPs, even though the CIRs are still convolved. As demonstrated in Section II, correlation will lead to increased NLOS power, thus modifying the distribution of power within the backscatter APDP. In the limiting case of identical channels with deterministic LOS amplitudes, the NLOS power is doubled while the LOS component remains untouched, i.e., (57) and (58) Adapting the denitions of RMS delay spread and average excess delay accordingly, they will contain terms like (59) The effect of correlation on the RMS delay spread thus depends on the ratio between LOS and NLOS part. In case of a negligible LOS component, the correlation has no effect on at all. Conversely, the maximum effect (56) is reached for an extremely dominant LOS part, i.e., very high . This variable inuence of channel correlation on is taken into account by adding a parameter to (56), similar (but not identical) to a correlation coefcient (60) The parameter can be obtained by comparing the results for correlated and uncorrelated channels. Looking at (60) and (52), obviously has to be zero for uncorrelated channels. In case of full correlation, has to model the dependence of on the K-factor. Assuming uncorrelated channels with identical RMS delay spreads , the squared RMS delay spread of the pinhole channel is . For identical channels, which necessarily have identical RMS delay spreads, the result is . This translates to (61) on the shape of the Unfortunately, the dependence of APDP prevents any further specication without the limitation to a specic type of decay. From a generic point of view, some facts are apparent from (60) and (61): Obviously, is always larger than and , with an additional factor (65) For correlated channels with high K-factors and deterministic LOS amplitudes though, the average excess delay of the backscatter channel depends on the shapes of the individualchannels APDPs. This conclusion can be reached following (62) A comparison between the full formula and the approximation for is shown in Fig. 4. Following this result, correlation between the constituent channels can be ignored as long as of the individual channels is below 0 dB. For higher K-factors, the correlation term has to be considered. The approximate version should be sufciently accurate in these cases (roughly 5%). Hence and and uncorr. identical. (63)
Fig. 4. Correlation factor in the RMS delay spread approximation for exponentially decaying constituent channels.

for fully correlated constituent channels. Moreover, is dominated by the channel with the higher RMS delay spread due to the geometric sum. C. Parameter for Exponential APDPs

For exponential APDPs, the factor is derived by a rather lengthy but straightforward derivation from (37) and (38) using the above denitions. For full correlation and thus , we obtain

VI. AVERAGE EXCESS DELAY The average excess delay is the rst moment of the normalized APDP, cf. (51). As such, and following the same analogy to statistics as for the RMS delay spread, the backscatter parameter for uncorrelated constituent channels can be calculated via (64) independent of the type of decay. Naturally, (64) can also be derived for specic types of APDPs: Calculating (51) for exponential APDPs (35), (36) and substituting (39)(41) for , , and leads to same result

ARNITZ et al.: WIDEBAND CHARACTERIZATION OF BACKSCATTER CHANNELS

263

Fig. 5. Comparison between approximation and true backscatter parameter for , RMS delay spread simulated exponential PDPs. Shown are K-factor , and average excess delay .

issue [24], but also nonexponential decays, especially in industrial environments with large metal reectors [10], [13]. Moreover, even if the average PDP is exponential, each instantaneous PDP has an arbitrary shape. It is thus important that the approximations are robust to nonexponential APDPs and that they are unbiased when used for instantaneous PDPs. This robustness is discussed below. The derivations in Section II do not assume a specic type of decay. All results in this section are thus valid for all types of APDPs. This includes the doubling for NLOS power for fully correlated constituent channels, see (27). Also the derivations leading to the nal formula for the backscatter K-factor, (47), are independent of the type of multipath decay. The approximation for is thus valid for all types of wireless channels. The same is true for the backscatter RMS delay spread for uncorrelated channels, (52), and also for the corresponding average excess delay, (64). In addition, the equations for and with correlated constituent channels, (60), (63), and (66) also work for nonexponential APDPs with relatively low K-factors. This is demonstrated in Section VIII, where the formulas are used for measured channels with nonexponential decays. VIII. MEASUREMENT RESULTS In order to fully validate the above derivations, we compare the single-channel-based approximations to the true backscatter parameters for measured channels. The used measurements were taken in a UHF RFID warehouse portal with metal backplanes in an industrial environment, see [10], [11]. Portal setup and antenna gain patterns resembled a typical UHF RFID portal, and the channels were recorded in a frequency range of 0.5 through 1.5 GHz [11]. The tag-antennas were mounted on a pallet containing liquids and metal. This environment is characterized by intense and very variable multipath propagation with strong deterministic reections [10]. The measurements cover exponential and nonexponential APDPs, LOS and NLOS scenarios, channels with heavily correlated components, as well as channels where the LOS component is inuenced by spatially close reections. See [10] for a full analysis of the measurements w.r.t. ranging and localization. Comparisons between the constituent-channel-based approximations and the corresponding parameters directly estimated from the backscatter channel are shown in Figs. 68. The parameters here are estimated from instantaneous PDPs. As a consequence, each individual estimate is a random variable, which explains the noise in the plots. For an unbiased approximation, the results are clustered around the 45 line, where the constituent-channel-based estimates and the true backscatter parameter are identical. This has been marked as dashed line in all plots. Class averages2 are shown as reference for the reader. The constituent-channel-based estimates generally follow the true backscatter parameters, i.e., the backscatter parameters can be approximated quite well from the single-channel estimates. The approximations use (47) and (48) for , (60) and (63) for , and (64) and (66) for . Note that (63) and (66)
2Averages

the same reasoning as for the RMS delay spread above. Again solving (51) for exponential APDPs, but this time doubling the NLOS power (see Section V-B), we obtain

with decays.

(66) for identical channels with exponential

VII. VALIDATION FOR EXPONENTIAL APDPS AND DISCUSSION OF ROBUSTNESS W.R.T. NONEXPONENTIAL MULTIPATH DECAYS Fig. 5 shows a comparison between the backscatter parameters and their respective single-channel approximations for simulated ensembles of CIRs following exponential APDPs. The simulations cover a range of 20 through 20 dB and 3 to 50 ns for the single-channel K-factor and RMS delay spread, respectively. The results in Fig. 5 were generated using ensemble averages over 100 instantaneous PDPs for each combination of , , , and . Approximations are based on (47) and (48) for , (60) and the exact version of (62) for , and (64) and (66) for the average excess delay, . The single-channel approximation can be expected to match the true backscatter parameters perfectly for exponential APDPs. Fig. 5 reects and conrms this expectation: All markers are on the dashed line, which marks a perfect match between approximation and true parameter. This validates the approximation formulas for exponentially decaying channels. A more thorough analysis can be found in [1]. The assumption of exponential APDPs for indoor environments is of course a simplication. In reality, clusters will be an

for ranges of the abscissae, similar to a histogram calculation.

264

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 6. Approximation of the backscatter K-factor w.r.t. LOS for measured power-delay-proles. Estimates are based on instantaneous PDPs and assume for the monostatic case.

Fig. 8. Approximation of the backscatter average excess delay for measured power-delay-proles. Estimates are based on instantaneous PDPs and use the formulas for exponentially decaying channels, (66), for the monostatic setup.

IX. CONCLUSION A method to compute wideband parameters of wireless backscatter radio channels from the shape parameters of the constituent channel pair was presented. Expressions for the K-factor w.r.t. the LOS path, the RMS delay spread, and the average excess delay of the backscatter channel based on the individual channels parameters are given. The presented approximations are valid independent of the actual shape of the average power-delay-prole (APDP), with exceptions for fully correlated channels with dominant LOS components. These exceptions are discussed in detail in the paper. It was also shown that the uncorrelated scattering (US) assumption holds for the backscatter channel if both individual channels are US, independent of channel correlation and for all types of APDPs. As anticipated, the backscatter channel always has a considerably lower K-factor and considerably higher delay spreads than both individual channels. For the K-factor, this is especially true for constituent channels with weak LOS . All parameters are dominated by the most dispersive constituent channel, i.e., the channel with the lowest K-factor and the highest delay spread. It was shown that full correlation between up- and downlink channel doubles the NLOS energy compared to uncorrelated constituent channels. Correlation thus additionally decreases the K-factor and increases the delay spread of the backscatter channel. The presented results have been veried based on simulated and measured indoor channel impulse responses. Although the shape parameters are dened for the average PDP, the presented methods can be readily applied to nonaveraged, i.e., instantaneous PDPs, as shown in Figs. 68. APPENDIX A. Uncorrelated Scattering of the Backscatter Channel It stands to reason that the uncorrelated scattering (US) assumption does not hold for the combined backscatter channel

Fig. 7. Approximation of the backscatter RMS delay spread for measured power-delay-proles. Estimates are based on instantaneous PDPs and use the with exponentially decaying channels, (63). simplied formula for

work well for these largely nonexponential APDPs (cf. [10]), even though they have been calculated specically for exponential decays. The slight bias for the monostatic case in Fig. 6, especially for medium to low K-factors, is caused by indirect physical paths that have been mapped to the LOS component. Because the approximation assumes , the estimate of from the constituent channels is slightly too low in these cases. Finally, the outliers in Fig. 6 for very low K-factors are created by false-positives in the LOS detection. If the LOS is too weak to be detected, which is increasingly the case for very low , the detection locks onto the next higher component, thus leading to an almost arbitrary positive error for the backscatter K-factor. RMS delay spread and average delay, on the other hand, are more robust against such errors, hence no such outliers are present in Figs. 7 and 8.

ARNITZ et al.: WIDEBAND CHARACTERIZATION OF BACKSCATTER CHANNELS

265

because of the convolution operation, even if the individual channels are US. Uncorrelated scattering implies uncorrelatedness between the individual paths of a channel impulse response, i.e., the covariance between different taps has to be zero (67) The NLOS part of the backscatter CIR was derived in Section II, (12). For convenience, we will repeat this result here (68) Clearly, the rst two terms in (68) do not create any correlation between the taps of , because and have uncorrelated elements. The last term, on the other hand, is the convolution of and , which might lead to correlation. The structure of this term is

Looking at the structure of (69), it is clear that neither combination of indices is present in any cross-covariance. Hence (75) . for all combinations between different elements of The cross-covariances between the rst two terms in (68) and contain the LOS components of the constituent channels, and . These are non-Gaussian random variables,3 hence splitting the terms using (18) is not possible. We will instead extract the deterministic amplitude from the expected value. The uniform phase is then absorbed by the remaining ZMCSCG variables, cf. (24). This results in (76) and (77) Clearly, the rst-order moments here are zero, and it is straightforward to show that also the third-order moments are zero due to the zero-mean properties of the variables. We can thus nally conclude that (78)

. . .

. . .

. . .

..

. . .

. . .

(69)

It can easily be shown that (70) and are arbitrary random variables. Inserting the where elements of into (67) and using (70), we end up with covariances between individual terms (71) for example and, following the same reasoning

This leaves possible correlation between the LOS part, (cf. (7)), and NLOS-components. Such correlations can only be present for terms that either contain or , and thus only for the rst two terms of (68). Applying (67) and extracting the constant amplitudes of the LOS parts, we obtain for all (79) (80) Combining (78), (79), and (80) then leads to (81)

(72) The second-order terms in (71) are zero for all circulant symmetric complex random variables (73) and the fourth-order term can be split into second-order terms using (18)

which proves the uncorrelated scattering assumption for pinhole channels with correlated or uncorrelated constituent channels. REFERENCES
[1] D. Arnitz, U. Muehlmann, and K. Witrisal, Wideband characterization of backscatter channels, in Proc. Eur. Wireless Conf., Vienna, Austria, Apr. 2011, pp. 205211. [2] J. D. Grifn and G. D. Durgin, Gains for RF tags using multiple antennas, IEEE Trans. Antennas Propag., vol. 56, no. 2, pp. 563570, Feb. 2008. [3] T. Taniguchi, Y. Karasawa, and M. Tsuruta, An analysis method of double fading MIMO channels including LOS environments, in Proc. IEEE 19th Int. Symp. Personal, Indoor Mobile Radio Commun., 2008, pp. 15. [4] J. D. Grifn and G. D. Durgin, Link envelope correlation in the backscatter channel, IEEE Commun. Lett., vol. 11, no. 9, pp. 735737, Sep. 2007.
3Any Gaussian fraction in an LOS component is independent from the constant-amplitude part and can thus be modeled as an additive variable (see discussions at the end of Section II). Consequently, (75) applies.

(74) The rst summand here is zero due to uncorrelated scattering of and , except for the combination and . The second summand is always zero for uncorrelated channels, and only nonzero for correlated channels if and , cf. (6).

266

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[5] D. Kim, M. A. Ingram, J. Smith, and W. W. , Measurements of smallscale fading and path loss for long range RF tags, IEEE Trans. Antennas Propag., vol. 51, no. 8, pp. 17401749, Aug. 2003. [6] P. L. Kae, A. Intarapanich, A. B. Sesay, J. McRory, and R. J. Davies, Spatial correlation and capacity measurements for wideband MIMO channels in indoor ofce environment, IEEE Trans. Wireless Commun., vol. 7, no. 5, pp. 15601571, May 2008. [7] P. Almers, F. Tufvesson, and A. F. Molisch, Keyhole effect in MIMO wireless channels: Measurements and theory, IEEE Trans. Wireless Commun., vol. 5, no. 12, pp. 35963604, Dec. 2006. [8] A. Paulraj, R. Nabar, and D. Gore, Introduction to Space-Time Wireless Communications. Cambridge, U.K.: Cambridge Univ. Press, 2003. [9] X. Yin, Y. Zhou, and F. Liu, A generic wideband channel model for keyhole propagation scenarios and experimental evaluation, in Proc. 4th Int. Conf. Commun. Networ. China, 2009, pp. 15. [10] D. Arnitz, U. Muehlmann, and K. Witrisal, Wideband Characterization and Modeling of UHF RFID Channels for Ranging and Localization, IEEE Trans. Antennas Propag., to be published. [11] D. Arnitz, G. Adamiuk, U. Muehlmann, and K. Witrisal, UWB channel sounding for ranging and positioning in passive UHF RFID, in Proc. 11th COST2100 MCM, Aalborg, Denmark, Jun. 2010 [Online]. Available: http://www.spsc.tugraz.at/publications [12] M. K. Awad, K. T. Wong, and Z. bin Li, An integrated overview of the open literatures empirical data on the indoor radiowave channels delay properties, IEEE Trans. Antennas Propag., vol. 56, no. 5, pp. 14511468, May 2008. [13] J. Karedal, S. Wyne, P. Almers, F. Tufvesson, and A. F. Molisch, A measurement-based statistical model for industrial ultra-wideband channels, IEEE Trans. Wireless Commun., vol. 6, no. 8, pp. 30283037, Aug. 2007. [14] M. S. Varela and M. G. Sanchez, RMS delay and coherence bandwidth measurements in indoor radio channels in the UHF band, IEEE Trans. Veh. Technol., vol. 50, no. 2, pp. 515525, 2001. [15] A. Molisch, Wireless Communications. Hoboken, NJ: Wiley, 2005. [16] H. Cramr, Mathematical Methods of Statistics. Princeton, NJ: Princeton Univ. Press, 1946. [17] R. Kattenbach, Statistical modeling of small-scale fading in directional radio channels, IEEE J. Sel. Areas Commun., vol. 20, no. 3, pp. 584592, Mar. 2002. [18] P. Bello, Characterization of randomly time-variant linear channels, IEEE Trans. Commun. Syst., vol. 11, no. 4, pp. 360393, 1963. [19] K. S. Miller, Complex Stochastic Processes. Reading, MA: AddisonWesley, 1974. [20] A. Papoulis and S. U. Pillai, Probability, Random Variables, and Stochastic Processes, 4th ed. New York: McGraw-Hill, 2002. [21] L. J. Greenstein, S. S. Ghassemzadeh, S.-C. Hong, and V. Tarokh, Comparison study of UWB indoor channel models, IEEE Trans. Wireless Commun., vol. 6, no. 1, pp. 128135, Jan. 2007. [22] D. Cassioli, M. Z. Win, and A. F. Molisch, The ultra-wide bandwidth indoor channel: From statistical model to simulations, IEEE J. Sel. Areas Commun., vol. 20, no. 6, pp. 12471257, 2002. [23] K. Witrisal, Y.-H. Kim, and R. Prasad, A new method to measure parameters of frequency-selective radio channels using power measurements, IEEE Trans. Commun., vol. 49, no. 10, pp. 17881800, Oct. 2001.

[24] A. Saleh and R. Valenzuela, A statistical model for indoor multipath propagation, IEEE J. Sel. Areas Commun., vol. 5, no. 2, pp. 128137, 1987.

Daniel Arnitz (M11) received the Dipl.-Ing. degree in electrical engineering from the University of Applied Sciences FH Joanneum Kapfenberg, Austria, in 2005 and the Ph.D. degree from Graz University of Technology, Austria, in 2011 (both degrees with honors). His Ph.D. thesis covered the eld of tag localization in passive UHF RFID and his diploma thesis focused on a feasibility study of (burst) error correcting codes for long-range RFID systems. He is currently with the Signal Processing and Speech Communication Laboratory of Graz University of Technology, Austria. His research interests include positioning and radar systems, channel and system modeling, ultrawideband technologies, and nonlinear systems.

Klaus Witrisal (M09) received the Dipl.-Ing.degree in electrical engineering from Graz University of Technology, Austria, in 1997, the Ph.D. degree (cum laude) from Delft University of Technology, The Netherlands, in 2002, and the Venia Docendi (Habilitation) from Graz University of Technology, in 2009. From 1997 to 2001, he was a Ph.D. student and Research Engineer at Delft University of Technology. Currently, he is an Associate Professor at the Signal Processing and Speech Communication Laboratory (SPSC), Graz University of Technology. His research interests are in signal processing for broadband and UWB wireless communications, propagation channel modeling, and positioning. Dr. Witrisal is Co-Chair of the MTT/COM Society of the IEEE Austria Section.

Ulrich Muehlmann (M05) received the Dipl.-Ing. degree in telematics and the Ph.D. degree from Graz University of Technology, Austria, in 2000 and 2005, respectively. Since 2005, he has been with NXP Semiconductors (former Philips Semiconductors) working in the eld of UHF RFID technology focused on systems and analog innovation. He is author and/or coauthor of several publications in the eld of optical tracking, machine vision, and RFID.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

267

Calibration of Electric Field Sensors Onboard the Resonance Satellite

Manfred Sampl, Member, IEEE, Wolfgang Macher, Christian Gruber, Thomas Oswald, Member, IEEE, Helmut O. Rucker, and Mikhail Mogilevsky
AbstractStrategies and results for calibrating electric eld sensors (antennas), as used in radio astronomy, onboard the spacecraft Resonance are presented. Calibration is performed for four boom antennas and four cylindrical sensors at the boom tips. These antennas are devised for the measurement of electric elds and plasma parameters. It is shown that the electrical representations of the antennas, the effective length vectors, differ from their mechanical originals and are shortened and tilted by several degrees of angle. The knowledge of the acquired parameters is of great benet to the Resonance mission. In particular, goniopolarimetry techniques like polarization analysis and direction nding depend crucially on the effective axes. For the rst time, this kind of analysis is performed for a space-borne antenna system consisting of boom monopoles and cylindrical tip antennas. Index TermsAntenna measurements, HF antennas, numerical analysis, radio astronomy, satellite antennas, space vehicle antennas.

anechoic chamber measurements, might be applied at a later stage of the project. II. MISSION RESONANCE The Russian Space Research Institute (IKI, Russian Academy of Sciences, Moscow, Russia) is leading the scientic effort for a space-borne mission, called Resonance, to investigate the inner magnetosphere and auroral region [1]. The mission is consisting of four Earth orbiting satellites, which are to be investigating the wave-particle interaction and dynamic processes in near Earths plasma in detail. Details of the project can be found in [2]. Among these processes are the following: ducted propagation of electromagnetic waves of whistlermode and ion cyclotron frequency ranges, which play a signicant role in the dynamics of electron and ion radiation belts; motion of energetic particles from the regions of auroral acceleration and magnetic reconnection; investigation of the magnetospheric cyclotron maser instability; source mechanism of the Auroral Kilometric Radiation (AKR). Targeted observation area of the proposed experiments are magnetospheric ux tubes. To achieve a maximized observation time along a bundle of geomagnetic eld lines (ux tube) the spacecraft has to y a magneto-synchronous orbit. Since geomagnetic eld lines corotate with Earth, the spacecraft has to move along a trajectory ensuring the presence in the ux tube. Fig. 1 outlines such a trajectory where the spacecraft moves along a selected ux tube. In Fig. 1, denotes the initial position of spacecraft inside the magnetic ux, where is the position at the moment of time . The spacecraft Resonance is planned to be launched in 2014 as an assembly of the new Russian standardized scientic satellite micro platform called Karat. This platform is intended to serve as a bus for various kinds of space science missions and is developed and built by NPO S.A. Lavochkina. The platform will have a mass of 250 kg and is going to offer a science payload capacity of up to 80 kg. It consists of a service module with standardized interfaces, telemetry and control, telecom, a power supply system, a propulsion module, and the payload module. The Resonance spacecraft will be spin stabilized along the -axis (Fig. 2), pointing the solar panels (located at the bottom of the spacecraft) at a maximum of the possible time towards the sun. The knowledge of the true antenna properties is of vital importance to the scientic results of the Resonance mission by greatly enhancing the accuracy of the acquired radio data.

I. INTRODUCTION ITH the launch of the rst space-borne radio astronomy observatories in the last century, it soon became clear that the true antenna parameters of the launched instruments differ from their mechanical representation. For reliable and correct measurements, nding the true antenna characteristics is vital. In this paper, we present two methods applied to the spacecraft Resonance to nd the true antenna and instrument parameter for the quasi-static case. The applied methods are electrolytic tank measurements (rheometry), which is a method to determine the effective length vectors of electrically short antennas, and numerical computer simulations. The accuracy of the applied methods is about 1 for directions of effective axes and some percent for effective lengths and antenna capacitances. Two other calibration methods, in-ight calibration and
Manuscript received October 29, 2010; revised March 11, 2011; accepted July 20, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was part of the science project RESONANCE electric eld sensors: Determination of the optimum conguration, which was supported by the Austrian Research Promotion Agency (FFG) in the framework of ASAP 4, Project 816159. M. Sampl, W. Macher, and H. O. Rucker are with the Space Research Institute, Austrian Academy of Sciences, Graz 8042, Austria (e-mail: manfred. sampl@ieee.org; wolfgang.macher@oeaw.ac.at; rucker@oeaw.ac.at). C. Gruber is with the University of Graz, Graz @oeaw.ac.at, Austria (e-mail: gruber@alumni.tugraz.at). T. Oswald is with the Thomas Oswald Aerospace Software, Weinitzen 8044, Austria (e-mail: thomas.oswald@aeroware.at). M. Mogilevsky is with the Space Research Institute of the Russian Academy of Sciences, Moscow 117997, Russia (e-mail: mogilevsky@romance.iki.rssi. ru). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167918

0018-926X/$26.00 2011 IEEE

268

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 3. Picture of the gold-plated model of antenna A4 and the boom tip sensor B4 (stub on the right-hand side), kept apart by an insulating black spacer. The threaded part on the left of the boom is used to screw the antenna into the socket on the model.

Fig. 1. Magnetic ux tubes rotate together with Earth as the spacecraft moves along the orbit. The thick lines mark the distance the spacecraft passes along the magnetic ux tube.

and eases the solution of the problem. In an antenna system, where the wavelength is much larger than the geometrical dimensions , the exponential term in (1) gets close to 1. Accountable for this is the fact that for points on the antenna, applies [6], thus (2) Comparing (1) and (2), it is obvious that since the term containing the wave vector vanishes, becomes independent of the wave incident direction. However, this simplication does not conclude on the frequency independence of the present antenna system because the current distribution on the antenna is still frequency-dependent. Further clarifying, Sinclairs effective length vector has no radial component, which is in contrast to (3)

Fig. 2. Antenna numbering: M1M4 are the mechanical booms, A1A4 the effective axes of the boom antenna, and B1B4 the effective axes of the cylindrical tip sensors.

Here, denotes the wave propagation vector and wave constant with

the

III. EFFECTIVE LENGTH VECTORS The investigated monopole antennas and sensors are subject to parasitic inuence from the conducting spacecraft body. The elaborate shape of the spacecraft body acts as ground plane for the antennas, signicantly altering their properties. In the present radio astronomy context, the denition of the complex vector quantity effective length vector is slightly different from the classical approach found in [4] and [5]. Sinclair [4] denes the direction of effective length vector perpendicular to the wave vector , and therefore the effective length vector is depending on the wave incident direction. This dependence makes the use of complicated because for every wave incident, direction is different. In the present context, it appears to be more useful to have a denition for the effective length vector that reects the antenna reception property, regardless of the wave incident direction, at least for the quasi-static case. A denition fullling this prerequisites is extensively discussed in [6], but can also be found in [5] and is given by (1) is related to the vector potential and the current at the antenna terminals . In (1), the integral of the current density has to be taken over the antenna volume. For electrically short antennas, it is possible to state further simplications that reduce the complexity of the above relations The electric far eld radiated when driving the antenna with current is (4) where denotes the intrinsic impedance of space, and the wavelength. Subsequently, we refer to the effective length vector as in denition (1) since this is the established quantity in the radio astronomy community. IV. RHEOMETRY Rheometry is an experimental method to nd the quasi-static properties of an antenna system. It is essentially an electrolytic tank measurement using a scale model as receiver of waves to acquire the effective length vector. Rheometry was investigated initially in the Projet Pyramides [7], [8] in the beginning of the 1970s. Since then, the method has been developed and applied to a number of spacecraft, such as CassiniHuygens [3] and Stereo [9], [10]. The frequency range of the proposed radio waves experiment is from 0.1 Hz to 2 MHz, so rheometry is well suited. In the present experiment, a gold-plated model (scale 1:30), shown in Fig. 3, was built according to the specication of the satellite manufacturer NPO S.A. Lavochkina. A detailed description of the model, as well as pictures and rst results, can

SAMPL et al.: CALIBRATION OF ELECTRIC FIELD SENSORS ONBOARD THE RESONANCE SATELLITE

269

TABLE I MECHANICAL ANTENNA VECTORS

OF THE RESONANCE ANTENNAS A1A4 AND B1B4, FOR OPEN PORTS, AS OBTAINED BY RHEOMETRY AND

COMPARED TO THEIR EFFECTIVE LENGTH VECTORS NUMERICAL SIMULATIONS

Fig. 5. Denition of spherical coordinates (colatitude) and (azimuth) in the spacecraft-xed reference frame as used for the representation of antenna axes.

Fig. 4. Scheme of the rheometry setup, a tank holding an electrolyte, two metal plates creating an electric eld , the spacecraft scale model, and measurement equipment (adapted from [3]).

be found in [11]. The corner properties of the model are the following: approximate height and diameter of the spacecraft body is 40 mm, length of the boom antennas (A1A3) perpendicular to the spacecraft spin axis is 340 mm, and length of the boom antenna (A4) in line to the spin axis is 99 mm. Every boom antenna holds a cylindrical tip sensor (B1B4), each of which is 5 mm long and has the same diameter as the boom antennas, namely 2.2 mm. All antennas are treated as monopoles. During the measurements, the model is immersed in an electrolytic tank, with tap water as a sufcient electrolyte. Metal plates are attached at two opposite sides of the tank to form a large capacitor (Fig. 4). A signal generator is connected to the plates to establish a homogeneous electric eld in the tank, and the plates are fed with approximately 0.4 V RMS, 1.03 kHz. The distinctive frequency of 1.03 kHz is chosen to be apart of multiples of the mains frequency of 50 Hz. This experimental setup satises the quasi-static conditions, which prevail in the real receiving situation as outlined in [3]. The measured open port voltages [3] of the antennas are (5) measured open port voltage of th antenna. effective length vector of th antenna. electric eld in tank.

Voltages are measured, using a high-precision voltmeter, at the model antennas as a function of the model orientation with regard to the electric eld. Collected data is written in a voltage vector , where is the number of antennas, in the case of the Resonance spacecraft . For that purpose, the model can be rotated around a vertical axis. Several different suspensions of the model at this axis are used to nd enough independent measurements. A detailed description of the measurement process can be found in [3] and [12]. The effective length vector for all antennas as obtained by rheometry can be found in found [11] and is repeated in Table I for the sake of completeness. The spherical spacecraft coordinate system denition is shown in Fig. 5. Beside the acquisition of the effective length and axes, rheometry is an established method to nd the impedance matrix of the instruments antenna system. The impedance matrix contains the self- and mutual impedances of the antenna system and is used to represent the antenna system in a multiport network description [9]. In the quasi-static case, the impedance matrix can be expressed in terms of a pure capacitance matrix [6], [9], with the capacitances . These values are of special interest for the so-called mutual impedance experiment planned in the Resonance mission, which measures plasma parameters via the mutual impedances. Relevant for this experiment are the values between the boom antennas and the cylindrical tip sensor. The open-port values reect the characteristics of the insulated antenna, i.e., open feeds, without any additional hardware connected (cables, preamplier, etc.). The input impedance of common preampliers and the cable from the antennas to the actual preamplier-plugs are also dominantly capacitive [13]. The method for determining is similar as for acquiring the effective axes and lengths. In the course of the measurements, the hardware model is immersed into the electrolytic tank. A

270

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE II SPACECRAFT CAPACITANCE MATRIX [pF]; OBTAINED (NO METAL PLATES IN TANK)

BY

RHEOMETRY

current is driven through one of the antennas , all other antennas being left open. The occurring voltages on all antennas are recorded, yielding the voltage vector for calculating the th column, , in the impedance matrix . This has to be repeated for every antenna . The impedance matrices in vacuum and in the electrolyte are then related by [9] (6) In order to apply (6), one has to calculate the conductivity (7) of the electrolyte using the tank geometry, which is dened as width m , height m , and length m (see Fig. 4). The resistance of the electrolyte volume between the tank plates mounted on the smaller sides of the tank can be easily measured (7) , the metal plates used for For the actual measurements of creating the capacitor in the tank were removed. In order to investigate the boundary condition, two measurement cycles were performed, one with the metal plates in the tank and one without the plates. The difference is marginal and obviously holds the assumption that the ideal boundary conditions, vanishing of potential at innity, are not signicantly violated. Table II shows the measured values of the capacitance matrix without the metal plates in the tank. V. NUMERICAL SIMULATIONS Due to the obviously complex structure of spacecraft and antenna system assembly, it is not possible to nd the antenna reception characteristics by analytical means. The numerical simulations were based on the well-known numerical solvers Antenna Scatterers Analysis Program (ASAP)1 and COde for the Numerical Computation of Electromagnetic Processes for thin wire and thin shell structures (CONCEPT-II).2 Both programs are based on the method of
1ASAP was originally written by J. W. McCormack on the code basis of J. H. Richmond, The Ohio State University, Columbus. 2CONCEPT-II is developed by the Institute of Electromagnetic Theory at the Technische Universitt HamburgHarburg, Hamburg, Germany.

Fig. 6. Simple wire-grid model, oblique view.

moments [14] to solve numerically the underlying integral equations for the sought current distribution (ASAP: reaction integral equation; CONCEPT-II: electric eld integral equation). The desired open-port effective length vectors (8) and the capacitance matrix can be calculated as soon as the current distributions on all conducting bodies are given. Thus, for a wire-grid, (2) becomes (8) ( counting The integrals are taken over all wires the wires), is the innitesimal line element along the th wire, and is the current through the th wire (positive in -direction) when the th port is driven with current , and the other ports are left open. The software model was gradually improved during the course of the project. Starting with a simple wire-grid model (Fig. 6) and using both solvers, ASAP and CONCEPT-II, rst results were obtained [11]. In a second stage, the satellite body was patch modeled (Fig. 2), still using wires as antennas. The nal results, shown in Tables I and III, were calculated using a full patch-grid model (antennas and body, Fig. 7) and the solver CONCEPT-II. Previous numerical calculations for scientic spacecraft were performed with the spacecraft body modeled as patches [9], [10] or wire-grid [6], antennas always realized as wires. This is the rst time that antenna cylinders are modeled as patches. Notable is the fact that models with wires as antennas were resulting in low precision for the cylindrical tip sensor B1B4 . It veries that a wire is not an appropriate representation for a short cylindrical structure, and the solvers fail in obtaining a good result due to the implemented thin-wire approximation [14][16]. The patch-grid antenna modeling was performed with two different patch sizes in order to check the necessary neness. The reference model is ner, representing the antennas as cylinders with 12-edge polygons. For comparison, a hexagonal version (cross section with six edges) was evaluated as well. The change of most results is below the inherent errors of the method. Especially the inuence on effective length vectors

SAMPL et al.: CALIBRATION OF ELECTRIC FIELD SENSORS ONBOARD THE RESONANCE SATELLITE

271

Fig. 7. Final detailed patch-grid model, oblique view.

TABLE III SPACECRAFT CAPACITANCE MATRIX [pF]; OBTAINED BY NUMERICAL CALCULATIONS (SOLAR PANELS INSULATED FROM THE S/C HULL)

Fig. 8. (color online) Spacecraft body and the attached antennas (black lines) shown in - and -planes. The blue rod depicts the effective length vector of the cylindrical tip sensor B3, and the red rod depicts the effective length vector of the boom antenna A3. In the -plane, the effective axes are tilted toward the -axis, but in the -plane (azimuthal), the effective axes are practically not effected, less than 1 angular offset, due to the high symmetry of the spacecraft.

is negligible. Only some mutual capacitances change by almost 9%, which is the maximum error that can be expected by numerical and modeling inaccuracies. This proves that the six-edge modeling introducesdue to its coarse antenna surface representationa perceptible additional bias to the mutual capacitances, but the 12-edge modeling is certainly ne enough. Detailed results can be found in [17]. Another aspect that has been investigated numerically is the inuence of the connection between the solar panels and the spacecraft body on the antenna properties. To check this inuence, numerical calculations had been made with a dened impedance of 50 M connection between the spacecraft body and the solar panels, and with a short circuit. The difference amounts in a 1.4 higher colatitude for the boom antennas A1A3 and 0.9 for the cylindrical tip sensor B1B3 in case of the high impedance connection, i.e., the effective axes are further tilted away from their originals in the short-circuit case. Neither the effective length of the antennas nor the azimuthal angle are affected, accounting for the high symmetry of the spacecraft. The cylindrical tip sensors A4 and B4 are also not affected to a relevant degree. VI. SUMMARY The performed rheometry measurement campaign during November 2008 has proven the expected tendency of the effective length vector offsets. The results presented in Table I and Figs. 2 and 8 show clearly that the effective length vectors

(red and blue rods) are shortened and tilted away from the mechanical antennas (black rods). All antennas perpendicular to the spin axis of the spacecraft are tilted toward the -axis. Most of the parasitic spacecraft body is below the antenna mount and therefore repels the effective axes upwards. Fig. 8 outlines the difference between the effective lengths of the antennas in detail. Whereas the boom antenna A3 shows the expected reduction in magnitude [5], of being shortened to less than half the mechanical length, the cylindrical tip antenna shows an entirely different behavior. In the present measuring setup, the cylindrical tip sensor measures the potential difference between the respective boom tip and the spacecraft body. Under ideal conditions, the effective length of B3 would be equal to the mechanical length of the span created by the length of the boom antenna, but the boom antenna as a parasitic body draws potential from the spacecraft body toward the cylindrical tip sensor and therefore shortens the effective length of the tip sensor B3. This behavior is comparable to relations holding for spherical probes as described in [18]. Similar considerations hold for the other antennas. In the -plane, the axes of effective length vectors are virtually not affected (Fig. 8, bottom panel) due to the high symmetry of the satellite design. The tilt is less than 1 , and this is considered to be within the measurement accuracy. The axes of the antennas A4 and B4, parallel to the -axis, are also not considerably affected. Comparing the effective length vectors and capacitances as obtained by rheometry (Tables I and II) and CONCEPT-II (Tables I and III), with solar panels insulated, shows that the effective axes practically coincide, being far better than expected.

272

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

The deviation is less than 0.2 , whereas the characteristic error of the rheometry method is about 1 [3]. The differences in effective lengths are less than 4% for A1A4 and B1B3, and 5% for B4. This is in line for the typical error to be expected from measurement imprecision and modeling inaccuracy. The good agreement of rheometry and numerical computations is based on patch modeling and cannot be achieved on basis of antenna wire grid modeling. Pure wire-grid codes, especially ASAP, have to cope with high systematic errors, in particular, for the cylindrical tip sensors (B1B4). The modeling of these antenna stubs (very thick cylinder, wire diameter mm and length mm) is very coarse and seems to violate the common thin-wire approximation applied in this type of numerical computer program. The results of the pure wire calculations and additional investigations on the effect of wire radii can be found in [17]. Similarly, the capacitances as obtained by the two methods are in good concurrence, except for the self-capacitances of B1/B4 and the mutual capacitance of A4/B4. The capacitances between the cylindrical tip sensors and their mounting booms are very sensitive to the modeling of the related feed zones.

ACKNOWLEDGMENT The authors want to thank Dr. M. Yanovsky of the Russian Space Research Institute for the invaluable information about the Resonance spacecraft design, and Dr. J.-L. Rauch from the Laboratoire de Physique et Chimie de lEnvironnement (CNRS) for information on the cylindrical tip antennas.

[10] T. Oswald, W. Macher, H. Rucker, G. Fischer, U. Taubenschuss, J. Bougeret, A. Lecacheux, M. Kaiser, and K. Goetz, Various methods of calibration of the STEREO/WAVES antennas, Adv. Space Res., vol. 43, no. 3, pp. 355364, 2009. [11] M. Sampl, W. Macher, C. Gruber, T. Oswald, H. Rucker, and M. Mogilevsky, Resonance spacecraft antenna calibration: Rheometry and numerical simulations, in Proc. Loughborough Antennas Propag. Conf., Loughborough, U.K., 2009, pp. 709712. [12] M. Sampl, Experimental and numerical antenna calibration in the resonance mission, Masters thesis, University of Graz, Graz, Austria, 2010. [13] D. A. Gurnett, W. S. Kurth, D. L. Kirchner, G. B. Hospodarsky, T. F. Averkamp, P. Zarka, A. Lecacheux, R. Manning, A. Roux, P. Canu, N. Cornilleau-Wehrlin, P. Galopeau, A. Meyer, R. Bostrm, G. Gustafsson, J.-E. Wahlund, L. hlen, H. O. Rucker, H. P. Ladreiter, W. Macher, L. J. C. Woolliscroft, H. Alleyne, M. L. Kaiser, M. D. Desch, W. M. Farrell, C. C. Harvey, P. Louarn, P. J. Kellogg, K. Goetz, and A. Pedersen, The Cassini radio and plasma wave investigation, Space Sci. Rev., vol. 114, no. 1, pp. 395463, Sep. 2004. [14] R. F. Harrington, Field Computation by Moment Methods. Malabar, FL: Krieger, 1968. [15] A. Schroth, Moderne Numerische Verfahren zur Lsung von Antennenund Streuproblemen. Munich, Germany: Oldenbourg, 1985. [16] Moment Methods in Electromagnetics. Techniques and Applications, J. Moore and R. Pizer, Eds. Baldock, U.K.: Research Studies, 1984. [17] W. Macher, M. Sampl, C. Gruber, T. Oswald, and H. O. Rucker, Resonance electric eld sensors. Report on the ASAP4-Project, Space Research Institute, Austrian Academy of Sciences, Graz, Austria, Tech. Rep. IWF-183, 2009. [18] D. A. Gurnett, Principles of Space Plasma Wave Instrument Design Measurement Techniques in Space PlasmasFields, ser. Geophysical Monograph, R. F. Pfaff, J. E. Borovsky, and D. T. Young, Eds. Washington, DC: AGU, 1998.

REFERENCES
[1] A. G. Demekhov, V. Y. Trakhtengerts, M. M. Mogilevsky, and L. M. Zelenyi, Current problems in studies of magnetospheric cyclotron masers and new space project resonance, Adv. Space Res., vol. 32, no. 3, pp. 355374, 2003. [2] M. Mogilevsky et al., Resonance project. Study of wave-particle interaction and plasma dynamics in the inner magnetosphere. REPORT ON THE PHASE A, Space Research Institute, Moscow, Russia, Tech. Rep. 54-R/262-131/R-E, 2002. [3] H. O. Rucker, W. Macher, R. Manning, and H. P. Ladreiter, Cassini model rheometry, Radio Sci., vol. 31, pp. 12991311, 1996. [4] G. Sinclair, The transmission and reception of elliptically polarized waves, Proc. IRE, vol. 38, no. 2, pp. 148151, Feb. 1950. [5] C. A. Balanis, Antenna Theory: Analysis and Design, 3rd ed. Hoboken: Wiley, 2005. [6] W. Macher, Transfer matrix description of multi-port antennas and its application to the Mars express/Marsis radar, Ph.D. dissertation, Graz University of Technology, Graz, Austria, 2005. [7] G. Dumas and S. Hoang, Mesure de rayonnement des antenneslectriquement courtes implantes sur une sonde spatiale pour le projet pyramide, Dp. de Rech. Spaciale, Obs. de Paris-Meudon, Paris, France, Tech. Rep. 104/PYR/10, 1973. [8] R. Hulin and G. Epstein, Projet pyramides simulation dntennes en cuve rheographique, Dp. de Rech. Spaciale, Obs. de Paris-Meudon, Paris, France, Tech. Rep. 103/PYR/9, 1973. [9] W. Macher, T. Oswald, G. Fischer, and H. O. Rucker, Rheometry of multi-port spaceborne antennas including mutual antenna capacitances and application to STEREO/WAVES, Meas. Sci. Technol., vol. 18, no. 12, pp. 37313742, 2007.

Manfred Sampl (S06M10) was born in Austria in 1976. He received the Dipl.-Ing. degree in electrical engineering from the Graz University of Technology, Graz, Austria, in 2007, and the M.Sc. degree in space sciences from the University of Graz, Graz, Austria, in 2010. He was a Student Trainee with Siemens Building Technologies, Salzburg, Austria, from 2002 to 2004 and with NXP Semiconductors, Gratkorn, Austria, from 2004 to 2007. Since 2007, he has been a Research Assistant with the Department of Extraterrestrial Physics, Austrian Space Research Institute, Graz, Austria, where he is concerned with antenna properties of radio and plasma wave experiments on scientic spacecraft. His research interests include antenna properties, RF engineering and wave propagation, numerical codes, and signal processing. Mr. Sampl is member of the European Geosciences Union (EGU) and the Austrian Electrotechnical Association (OVE). Wolfgang Macher was born in Voitsberg, Austria, in 1965. He received the Dipl.-Ing. degree and the Ph.D. degree in physics from Graz University of Technology, Graz, Austria, in 1997 and 2005, respectively. In 1990, he joined the Space Research Institute, Austrian Academy of Sciences, Graz, Austria, where he is currently a Staff Scientist. He worked on the calibration of antennas, accelerometers, and other sensors of scientic instruments onboard spacecraft. He developed methods and computer programs to analyze the properties of space-borne multimonopole antenna systems for the observation of radio and plasma waves. His current research activities focus on techniques for the investigation of thermal and dielectric properties of planetary and cometary surfaces. His research interests include electromagnetic theory, numeric computation, and signal analysis.

SAMPL et al.: CALIBRATION OF ELECTRIC FIELD SENSORS ONBOARD THE RESONANCE SATELLITE

273

Christian Gruber was born in 1981. He received the Dipl.-Ing. degree in technical physics from the Graz University of Technology, Graz, Austria, in 2010, and is currently pursuing the Ph.D. degree within the nano-optics group at the University of Graz, Graz, Austria, investigating the interaction of light emitters with metal nanostructures. In 2008 and 2009, he was a Research Assistant with the Radio Science team, the Space Research Institute, Austrian Academy of Sciences, Graz, Austria, and was involved in numerical simulations.

He has been the Head of the Department of Extraterrestrial Physics since 1999 and Deputy Director at the SRI since 2010. In the course of his work, he was a Visiting Researcher with various institutes including the Goddard Space Flight Center, NASA, Greenbelt, MD; Observatoire de Paris, Paris, France; the Department of Astronomy, University of Iowa, Iowa City; and the Institute for Radioastronomy, Kharkov, Ukraine. He is author and coauthor of more than 240 publications in international refereed journals and proceedings. Dr. Rucker is a Corresponding Member of the Austrian Academy of Sciences and a member of the European Geosciences Union (EGU), American Geophysical Union (AGU), and American Astronomical Society (AAS).

Thomas Oswald (M06) was born in Vienna in 1970. He received the B.Sc. degree in physics from the Open University, Milton Keynes, U.K., in 2003, and the M.Sc. degree in space physics and Ph.D. degree in physics from the University of Graz, Graz, Austria, in 2005 and 2010, respectively. His work with the Space Research Institute, Austrian Academy of Sciences, Graz, Austria, from 2003 to 2008 includes callibration of the STEREO and INTERBALL spacecraft antennas. Currently, he is a professional Pilot and Flight Instructor, as well as entrepreneur and independent scientist. His research areas include numerical electromagnetics and plasma-electrodynamics. Dr. Oswald is member of the American Geophysical Union (AGU).

Mikhail Mogilevsky was born in 1949. He received the magisterial diploma from the Gorky State University, Kharkiv, Ukraine, in 1972, and the Ph.D. degree from the Russian Academy of Sciences, Moscow, Russia, in 1988, both in physics. From 1972 to 1990, he was a Researcher with the Institute of Terrestrial Magnetism, Ionosphere and Radiowave Propagation, Russian Academy of Sciences. Since 1990, he has been with the Space Research Institute, Russian Academy of Sciences, and from 2003 he is Head of Laboratory. He is author and coauthor of more than 110 refereed journal publications and technical reports. His main research interests are space plasma physics and radiophysics, electromagnetic wave generation and propagation, magnetospheric physics, and nonlinear wave-particle interaction. Dr. Mogilevsky is a member of the European Geosciences Union (EGU) and the American Geophysical Union (AGU).

Helmut O. Rucker was born in Austria in 1948. He received the Ph.D. degree in geophysics and theoretical physics from the University of Graz, Graz, Austria, in 1974. He was awarded the Habilitation (Venia Docendi) in geophysics (physical space research) from the University of Graz in 1988. Since 1974, he has been a Scientist with the Space Research Institute (SRI), Austrian Academy of Sciences, Graz, Austria. His teaching at the University of Graz includes space plasma physics, planetary radio emission, and magnetospheres and waves in plasmas.

274

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

A Second-Order Asymptotic Approximation for the Sommerfeld Half-Space Problem

Won-seok Lihh
AbstractInvestigation is made on the Sommerfeld half-space problem that concerns the spherical-wave elds on the sparser side of a planar interface involving both permittivity and permeability contrasts. For a point current source placed at the interface, the elds are described by a second-order approximation that properly evaluates the asymptotic component having the inverse-square dependence on the distance from the source. The validity of this approximation is demonstrated in the time domain by comparisons with the waveforms of the simpler approximations and the wavenumber-synthetic waveforms. Index TermsApproximation methods, electromagnetic radiation, electromagnetic transient propagation, half space.

I. INTRODUCTION

classical topic in the literature is the Sommerfeld halfspace problem concerning the wave elds over a at earth radiated by a Hertz-dipole source [1, Ch. 6]. Historical remarks can be found in [2, Pref.] and [3]. For a point current source placed at the interface of permittivity and permeability contrasts, this paper seeks to better describe the elds of the spherical wave in the sparser medium (with smaller permittivity and permeability) above the interface. The spherical-wave elds are associated with the stationary-phase (SP) point in the complex plane of a wavenumber variable. A simple asymptotic evaluation of the wavenumber-synthetic integrals around the SP point gives the rst-order elds proportional to the simple inverse of the distance from the source. This approximation is valid for observations well off the interface. In the absence of the permeability contrast, the tangentialelectric (TE) wave involves the SP point only, while the tangential-magnetic (TM) wave involves also a pole known as the Sommerfeld pole. Although this pole makes no isolated (guidedwave) contribution, there needs to be a consideration for the coupling effect of the SP point and the pole [3], [4, Sec. 15-515-7], for observations made near the interface. This consideration modies the rst-order approximation by a correction term containing the (Sommerfeld) attenuation function. The higher the contrast (in permittivity), the more signicant the correction is. The TM spherical-wave elds thus obtained do not match with the head-wave elds in [5]. The mismatch can be resolved by a second-order approximation that properly evaluates the asymptotic component proportional to the inverse square of the dis-

Fig. 1. Sommerfeld half-space problem for a point current source placed tangentially or normally at the interface. Abbreviations are HW, head wave; IW, inhomogeneous (surface) wave; and SW, spherical wave. The angle  denes the critical cones   and   . Of interest are the spherical-wave elds above the interface in the region  >  .

= 180 0

tance. This approximation is shown to nicely describe the elds for low (as well as high) contrast, particularly near the interface. Although this paper is primarily intended for the permittivity contrast, the formulation is extended to cover the permeability contrast for which the TE wave may also involve a Sommerfeld pole. The analysis begins with the wavenumber-integral representations of the electric and magnetic elds in Section II. Section III describes 1) the rst-order approximation for the SP point, and 2) its modication by including the coupling effect of the SP point and the Sommerfeld pole. Section IV makes the second-order approximation. The validity of the three approximations is examined in Section V by comparisons in the time domain against the wavenumber-synthetic waveforms. The author adds a comparison with the results of a formula derived by King [6]. Finally, the effect of the permeability contrast is illustrated. II. INTEGRAL REPRESENTATION OF FIELDS The conguration of the Sommerfeld problem is shown in Fig. 1. A point current source is placed at the origin of the plane that coincides with the interface between lossless half-

Manuscript received August 26, 2010; revised February 10, 2011; accepted April 04, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The author is with GCT Research, Inc., Seoul 156-714, Korea (e-mail: lws@gctsemi.com). Digital Object Identier 10.1109/TAP.2011.2167908

0018-926X/$26.00 2011 IEEE

LIHH: SECOND-ORDER ASYMPTOTIC APPROXIMATION FOR THE SOMMERFELD HALF-SPACE PROBLEM

275

space media. The upper medium has the permittivity and the permeability , both positive. The lower medium has and , and the impedance mismatch is assumed: , or, . The angles and are such that , , , and , where and . The eld detector makes ideal observations of the wave elds above , the interface in the region where is the critical angle that denes the critical cones. Assuming that the source is excited by the Dirac impulse in the time domain, the analysis is performed under the . With the letter Fourier transformation standing for , , or , the notation denotes the -directional electric eld generated by the tangential ( -directional) source of density C m . The notation denotes the eld generated by the normal ( -directional) source of . The same for the magnetic density C m and . The normal source at generates elds times the elds of that at . syntheses and be dened by Let the wavenumber (1) and are the second-kind Hankel functions of where order 0 and 1, respectively. The integration is taken along a path (in the complex plane) that runs from the left extreme just below the real axis to the right extreme just above, through the with for , origin. Let 2. The square root of a complex quantity, say , is such that . By the decomposition of the TE and TM waves [7] (transverse-electric and -magnetic to the axis, respectively), the TE elds are expressed as

closing parenthesis nor by a multiplication. The partial elds, indicated by TE and TM, are such that (7) (8) The remaining three elds , , and are zero. The can nonphysical transients of the partial elds at as . Apbe avoided by treating proximations are made for large , after replacing the Hankel functions by their asymptotic representations. III. TWO SIMPLE ASYMPTOTIC APPROXIMATIONS Using the asymptotic representation [8, Sec. 9.2] (9) for large , the elds in (2)(6) can be put in the form (10) . The values of and the exwhere pressions of are listed in Table I ( and are for later use). The integral in (10) is along , a segment around the SP point that satises of , the prime denoting differentiation. The elds with do not exhibit the dominant simple-inverse dependence on the distance. Two simple approximations for (10) are described in this section. The rst involves the isolated contribution of the SP point. The second adds a correction term arising from the coupling effect of the SP point and the Sommerfeld pole. The integral in (10) becomes (11)

(2)

(3)

and where integral is along a path to which The relation between and is

, and where the is mapped (and deformed). (12)

and the TM elds are expressed as corresponding (piecewise) to either , where (4) A. First-Order SP Approximation Placing (5) in (11) allows the evaluation or .

(13) (6) for large , where expression of is obtained as . The (14)

where the factor C m has been omitted on the right-hand sides, and is not interrupted by a and where the operation of

276

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

, (k ), ^ , AND

FOR THE

TABLE I FIELDS IN (2)(6) ( ^ FOR  = < " =" )

B. Modied Approximation Supplementing the approximation above demands a con. sideration for the Riemann surface (hence, the sign) of Suppose that , and let be the residue of at the real-valued TM Sommerfeld pole (16) . The expressions of are listed satisfying in Table I. The weak inuence of the TE imaginary poles (17) is neglected. Note that satisfying for the TE elds, and for the TM elds with . Adapting the procedure in [4, Sec. 15-6 and 15-7], the product in (11) is approximated as

(18) where (13) is replaced by . The asymptotic evaluation (19) where (20)

in terms of the complewith is known as the atmentary error function. The factor tenuation function [4, Sec. 15-7]. The correction term accounting for the coupling effect of the SP point and the pole , the modied rst-order approximation [for (10)] reads

(21)
^

k = (" k  = k

= Res

0" k 0k .

)=("

0 ;  ; k), where 0 " ).

. Then, (10) is approximated as (15)

with

having mixed dependence on the distance. This approximation agrees with [3] which treats a lossy lower medium as [4, Ch. 15], for . The pole has no guided-wave effect. can also be expressed as The term (22) in terms of the inwith complete gamma function [8, Sec. 6.5], [9, Sec. 6.4], [10, Sec. vanishes while becomes 2.1]. At the interface,

which gives the dominant rst-order components proportional to the simple inverse of the distance, for . They vanish since , failing to give at a proper description near the interface. Those components of , , and have another factor arising from in the numerators of the expressions.

(23)

LIHH: SECOND-ORDER ASYMPTOTIC APPROXIMATION FOR THE SOMMERFELD HALF-SPACE PROBLEM

277

TABLE II
FOR

 = > " ="

implies ). with the values of listed in Table I ( in this paper is for the pole The assumption not to approach the singularity at in (27). The integral in (27) is approximated by (28) for which the approximation in (18) is supplemented by [4, App. 15C], [11, Sec. 4.4a] (29) the rst term making little contribution due to its oddness in . Then, the rst integral in (28) is approximated by (30) where is that appearing in (19), and where [9, Th. 1.2], [12, Sec. 27.1], [13, Sec. 4.7d]

where . For elds of the normal source as

, for example, (23) gives the

(24) (31) where with (25) and . The second integral in (28) is approxi:

mated by . Thus, the following term is added to

. Apart from the discontinuity factor for with , the expression agrees with the head-wave correction part [5, Eq. (42)] arising from the coupling effect of and the pole . However, the rst term the branch point [in (24)] does not match with the canonical head-wave part [5, Eq. (14)] contributed by the branch point in an isolated way. This means that (21) does not correctly describe the secondorder elds proportional to the inverse square of the distance. The mismatch is resolved in the next section. , the TE elds are affected by the real pole If . The entries in Table II replace those in Table I. The TM elds have imaginary poles whose inuence is neglected (as ). implied by IV. SECOND-ORDER ASYMPTOTIC APPROXIMATION To correctly handle the second-order elds, the Hankel functions are represented as [8, Sec. 9.2] (26) The elds in (2)(6) are put in the form (27)

(32) whose component cancels the rst of (22). A little arrangement establishes the second-order approximation [for (27)]

(33) which is regarded as a modied approximation in the sense that, without the component (modication term), it involves the SP is simply . point only. The parenthesis factor after For large such that [8, Eq. 6.5.32] (34) vanishes as , and the component the factor has an inverse-cube dependence on the distance. In case this

278

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

component can be neglected, the TE and TM elds are subjected to a common formula (involving the SP point only) regardless . of whether At the interface, the rst term of (33) vanishes since . The second term becomes (35) by the fact that (36) (37) . The rst term in the brackets with designating , (35) gives the of (35) differs from that of (23). For elds of the normal source as

(38)

agreeing with the head-wave elds in [5] (apart from the factor for ). This agreement contrasts with the mismatch exhibited by (24). The respective elds in (24) and (38) become equal for small (high contrast in permittivity). V. NUMERICAL RESULTS The validity of the three approximations (15), (21), and (33) at m is can be tested in the time domain. The eld taken for illustration. The upper medium is assumed to be free space, having the permittivity and the permeability . The lower medium also has unless otherwise stated. The source is excited by a Gaussian pulse, and the (time-domain) waveforms are constructed using the spectrum (39)

Fig. 2. Comparison of the (dashed) and (dotted) waveforms against and (b) the respective k -synthetic waveforms (solid) for E at (a)   (" " ," " ,   ,r m). The waveform is shown in (c) (dotted). at 

= 88 = = 88

=2

=5

= 70 7

with . In the following, the waveforms of the rst, modied rst-order , and (modorder elds are compared ied) second-order against the waveforms of the -synthetic eld evaluated numerically [using the expression as in (6)]. and . Fig. 2(a) makes the comparison for and The dashed and dotted lines are the waveforms of the elds, respectively. They show only small discrepancies with the -synthetic waveform in solid line, and may serve as waveform exhibits slightly acceptable approximations. The better agreement. If the observation is made near the interface, and components become signithe corrections by the waveform (dashed) cant. Fig. 2(b) shows the failure of the . to approximate the -synthetic waveform (solid) at waveform (dotted) shows an improvement, but there The

Fig. 3. Magnitudes of = plotted versus  for E at ! rad/s (medium), and rad/s (thick) (" (thin), :   ,r  m).

3 5210 = =

7 7 =5

4210

= 3210 rad/s = " , " = 2" ,

comporemains discrepancy, which is accounted for by the waveform in Fig. 2(c) shows good agreement. nent. The The transient around ns belongs to the inhomogeneous wave (shown in Fig. 1), which is not of interest with respect to in this paper. The amounts of correction by can be roughly guessed from Fig. 3 that plots for a few values of . The approximation for the permittivity contrast can also be applied to the case of a lossy lower medium, if every is formally replaced by , where the conductivity charac, terizes the loss. Fig. 4 shows the waveforms for as compared with the lossless case in Fig. 2(b) and (c). The good agreement in Fig. 4(b), compared with Fig. 4(a), supports the

LIHH: SECOND-ORDER ASYMPTOTIC APPROXIMATION FOR THE SOMMERFELD HALF-SPACE PROBLEM

279

Fig. 4. Comparison of the (a) (dashed) and (dotted), and (b) (dotted) waveforms against the k -synthetic waveform (solid) for E (" " " ,   , : = ,r m,  ).

=2

7 7 = 0 05 S m = 5

= 88

7 ="

(dashed) and (dotted), and (b) Fig. 6. Comparison of the (a) (dotted) waveforms against the k -synthetic waveform (solid) for E [" " " ,   ,r (and  m,  )].

= 10

7 =5

= 88

=0

7 ="

Fig. 5. Magnitudes of = plotted versus  for E at ! rad/s (medium), and rad/s (thick) (" (thin), :    ,r m,  ).

3 5210 = =

7 7 4210 = 5 = 88

7. = 3210 rad/s Fig. rad/sMagnitudes of 7 =7 plotted versus " =" 10 = " , " = 2" ,  =  (thin),,3r:5210 rad/s (medium),and 4210 =  = 5 m,  = 88 (and = 0)].

for E at ! rad/s (thick) ["

= 32 =" ,

applicability of the approximation to the lossy case. The correction by the component becomes less signicant for larger , as indicated in Fig. 5. Fig. 6 compares the waveforms for a larger permittivity and no loss . The approximation in Fig. 6(a) (dotted line) shows better agreement than in Fig. 2(b) of . The waveform in Fig. 6(b) is nearly exact. Fig. 7 reduces as increases. For shows that the correction by approximation may sufhigh contrasts (in permittivity), the ce. A comparison is in order with the result of the Kings formula [6, Eq. (37)]

shown in Fig. 8(b), where the singular behavior implies that the Kings formula does not assess the eld in an asymptotically apstrict manner. Fig. 8(c) is the error waveform for the proximation in Fig. 6(b), having no singular but a step-like behavior. The absence of a Dirac-impulsive error (in this lossless component. Using the case) conrms the correctness of the expression of in Table I, one may evaluate the amount , due to the excitation, by of the step at

(41) afforded by a third-order analysis that employs [8, Sec. 9.2] (40) . For where and , the waveform of (40) is shown in Fig. 8(a) by dotted line, the good agreement rivaling that of Fig. 6(b). There emerge, however, different behaviors of the error waveforms, upon evaluating the errors of the dotted waveforms with respect to the solid-line waveform. The error waveform for Fig. 8(a) is (42) To be further mentioned regarding the Kings formula is the case of low contrast in permittivity. Because this formula is subject , or , it does not to the condition (when ), while serving work well for small values of , the disas a good approximation for large values. For crepancy in Fig. 8(d) exhibited by the Kings formula contrasts

280

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

waveforms (dotted) of (a) " " Fig. 9. Comparison of the and (b) " " against the respective k -synthetic waveforms (solid) for E [" " ,   ,r m,  (and  )].

= 10

(7 ) = =5

=2

= 88

=0

Fig. 10. Magnitudes of = plotted versus  = for the normal-source elds E (solid), E (dashed), and H (dotted) at ! rad/s (thin), rad/s (medium), and rad/s (thick) [" " ," " , :  ,r (and   m,  )]. The solid lines almost overlap.

7 7

3 5 2 10 = =5

= 88

4 2 10 =0

= 3210 =

=2

Fig. 8. (a) Waveform of the Kings formula (40) (dotted) compared against the k -synthetic waveform (solid) in Fig. 6 (for " " ). (b) The error waveform approximation in Fig. 6(b). (d) The for (a). (c) The error waveform for the discrepancy exhibited by the Kings formula for " " , contrasting with the good agreement in Fig. 2(c).

= 10

=2

with the good agreement in Fig. 2(c) afforded by the approximation. The discrepancy, being the price for the simplicity of (40), is comparable to that of Fig. 2(b) exhibited by the waveform (dotted line). The Kings formula has worked better approximation for [Figs. 6(a) and 8(a)]. than the The price for the wide-range validity of the approximation is having to evaluate the rst- and second-order derivatives [in (33)]. of Consider now the second-order approximation involving the and approxiSP point only, for comparison with the mations. The effect of the (real) Sommerfeld pole can be evaded . This leads to the approximation by by formally placing , illustrated in Fig. 9 for and . The waveform for in Fig. 9(a) is waveform in Fig. 2(b), though a little rather better than the waveform in Fig. 2(c). That waveform for worse than the in Fig. 9(b) is worse than both the and waveforms in Fig. 6.

Finally, consider the case in which the permeability of the . The lower medium, , is different (larger) than can be guessed from the dashed amounts of correction by in Fig. 10, which plots versus lines for the elds of the normal source. If particularly, the factor of is zero and the approximation is employed. See Fig. 11(a) for , where makes no correction since . The correction by in Fig. 11(b) may be acceptable, though there remains an error whose step-like component can be corrected by (41) [for the Gaussian pulse (spectrum) in (39)], as shown in Fig. 11(c). It is is less favorable noted in passing that the case , to the replacement of by than for the lossy lower medium. When , in the former case, the remoteness between the SP point and each of the imaginary is such as to allow igpoles noring these poles. As increases after that replacement, however, the negative imaginary pole happens to approach the SP point. Their coupling effect becomes appreciable, and needs to . be taken into account as in the case Similarly as Fig. 10, the plots for the tangential-source elds are shown in Fig. 12(a) and (b) for the TE and TM elds, respec. In these plots, the transition region around the tively

LIHH: SECOND-ORDER ASYMPTOTIC APPROXIMATION FOR THE SOMMERFELD HALF-SPACE PROBLEM

281

Fig. 12. Magnitudes of = plotted versus  = for the tangentialsource elds with , similarly as Fig. 10. (a) For the TE elds E (solid), H (dashed), and H (dotted) at  . (b) For the TM elds (solid), E (dashed), and H (dotted) at  . E

7 7 =1

= 90 =0

Fig. 11. Comparison of the (a) (dashed) and (dotted), (b) (dotted), and (c) step-corrected (dotted) waveforms against the k -synthetic " ," " ,  ,  ,r m, waveform (solid) for E ["  (and  )]. In (a), the waveform is the same as the waveform, since .

= 88

=0 ^=0

=2 7

=3

7 =5 7

This paper has treated the Sommerfeld poles in such a manner as for the permittivity contrast only and the permeability contrast only. The applicability of the second-order approximation is limited to the case in which the two media show a mismatch in impedance. More elaborate analysis may have to be made for a slight (or no) mismatch. REFERENCES
[1] A. Sommerfeld, Partial Differential Equations in Physics. New York: Academic, 1949. [2] A. Baos, Jr., Dipole Radiation in the Presence of a Conducting HalfSpace. London, U.K.: Pergamon, 1966. [3] R. E. Collin, Hertzian dipole radiating over a lossy earth or sea: Some early and late 20th-century controversies, IEEE Antennas Propagat. Mag., vol. 46, no. 4, pp. 6479, Apr. 2004. [4] A. Ishimaru, Electromagnetic Wave Propagation, Radiation, and Scattering. Upper Saddle River, NJ: Prentice-Hall, 1991. [5] W. Lihh, A correction to head-wave elds for a simple planar contrast of permittivity, IEEE Trans. Antennas Propagat., vol. 58, no. 6, pp. 20782087, Jun. 2010. [6] R. W. P. King, Electromagnetic eld of a vertical dipole over an imperfectly conducting half-space, Radio Sci., vol. 25, pp. 149160, Mar./Apr. 1990. [7] W. Lihh, Asymptotic elds in frequency and time domains generated by a point source at the horizontal interface between vertically uniaxial media, IEEE Trans. Antennas Propagat., vol. 55, no. 10, pp. 27332745, Oct. 2007. [8] Handbook of Mathematical Functions, M. Abramowitz and I. A. Stegun, Eds. Mineola, NY: Dover, 1972. [9] V. A. Borovikov, Uniform Stationary Phase Method. London, U.K.: Inst. Elect. Eng. Press, 1994. [10] M. A. Chaudhry and S. M. Zubair, On a Class of Incomplete Gamma Functions With Applications. Boca Raton, FL: Chapman & Hall/CRC, 2002. [11] L. B. Felsen and N. Marcuvitz, Radiation and Scattering of Waves. New York: IEEE, 1994. [12] L. M. Brekhovskikh, Waves in Layered Media, 2nd ed. New York: Academic, 1980. [13] J. A. Kong, Electromagnetic Wave Theory. New York: Wiley, 1990. Won-seok Lihh received the B.S., M.S., and Ph.D. degrees in electronics/electrical engineering from Seoul National University, Korea. He is currently with GCT Research, Inc., Seoul, Korea. His research interests include asymptotic and time-domain analysis for wave elds, electromagnetic modeling for passive integrated components and packages, and the design of CMOS circuits and antennas for wireless front-ends.

vacant zone corresponds to which is forbidden by assumption in this paper. The curves of in[of in Table I] suffer less abrupt volving the factor , , , and involving , transition than those of and than those of , , , and involving . The elds , , and have regions of , at due to the factor resulting from 1) , and 2) due to . VI. CONCLUSION This paper has investigated a second-order asymptotic approximation for the spherical-wave elds of the Sommerfeld half-space problem, radiated by the point current sources placed tangentially and normally at the interface. The approximation correctly handles the second-order component proportional to the inverse square of the distance from the source, thus matching the spherical-wave elds with the head-wave elds. Improvement in accuracy upon the rst-order approximations has been presented by comparisons in the time domain. The analysis has treated the permeability contrast as well as the permittivity contrast. The approximation for the lossless media can be applied to the case of a lossy lower medium by a simple formal introduction of conductivity, if the permeability contrast is nil or low. The analysis for the interfacial source can be extended without difculty to the case of an elevated source.

282

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Direct Rational Function Fitting Method for Accurate Evaluation of Sommerfeld Integrals in Stratied Media
Theodoros N. Kaifas, Member, IEEE
AbstractThe current work proposes a direct rational function tting method, employing cylindrical waves alone, for the accurate evaluation of Sommerfeld integrals for planar multilayered structures. Three are the key points of the effort. 1) Until now, relative works require the extraction of the quasi-static/asymptotic terms, and branch cut/continuous wave contribution explicitly. In the current one, the explicit treatment of those terms is avoided. The proposed methodology is based on the direct tting of the spectrum of the Greens function by rational functions. Thus, it provides the spatial Greens function solely in terms of cylindrical waves. 2) The effectiveness, robustness and accuracy improvement of the rational function t rely upon the proper sampling of the spectrum. This accurate tting is possible because instead of avoiding large variations of the spectral kernel, we introduce proper paths to include more variation, and thus more spectrum information, before we apply the modied VECTFIT algorithm. 3) Furthermore, proper weighting of the VECTFIT is proposed in order to guide the algorithm in providing increased accuracy in specic desired areas of the horizontal distance between the source and observation points. Armed with the above the direct rational function tting method provides accurate results both for the near and far-eld regions. Various examples, among them the correct treatment of a two branch case, are given that prove the excellent performance and robustness of the proposed approach. Index TermsClosed form Greens functions (GFs), multilayered media, rational function t, Sommerfeld integral (SI).

I. INTRODUCTION HE Greens functions, (GFs), for stratied multilayer media are obtained as closed-form expressions in the spectral domain. The route back to spatial domain passes via the inverse Hankel transform through the computation of Sommerfeld integrals, (SIs). Those integrals do not lend themselves to an easy computation. Their integrands exhibit singularities near the integration path and an oscillatory, (due to the asymptotic form of the Hankel function), slow-decaying, [due to the Spectral Domain Greens Function, (SDGF)], nature. This complex structure, of continuous and discrete spectrum, in the spectral domain, results in the existence, into the actual eld, of spherical, cylindrical and also lateral waves, [1][3].
Manuscript received September 25, 2010; manuscript revised February 20, 2011; accepted May 06, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. T. N. Kaifas is with the Radiocommunications Lab, Department of Physics, Aristotle University of Thessaloniki, Thessaloniki 54124, Greece (e-mail: tkaif@skiathos.physics.auth.gr). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167915

Various methods have been proposed to address the problem. Here we focus on those that aim rst at the accurate representation of the SDGF. The use of exponential series, [following the discrete complex image method, (DCIM)], [4] and [5], or rational functions, [following various forms of the rational function tting method, (RFFM)], [6][9], or combinations of those, [10], lead, mainly but not solely, to spherical and/or cylindrical waves. It is the use of the various SI identities that render the inverse Hankel transform an easy task leading to closed form representation of the actual produced eld. Although the cited methods are highly efcient and fully developed, nevertheless there are still problems that remain or reenter the play when more complex structures, increased range of validity and/or decreased computational effort, are needed. The list of problematic structures includes cases without poles, stratied media with lossy Right or left-handed material, (LHM), and in general those that the branch cut contribution can be dominant. The proposed approach is based on the rational function representation of the SDGF where we evolve the works of [6] and [7]. Note that unlike those studies we use the rational function approximation not only for the surface waves and the dynamic part of the spectrum but also for representing the quasi-static and the continuous ones. Also, for accurate tting, we introduce proper paths to include more variation of the spectral kernel before we apply the modied VECTFIT algorithm, [6]. Furthermore, proper weighting of the VECTFIT is proposed in order to guide the algorithm in providing increased accuracy in specic desired areas of the horizontal distance between the source and observation points. The rest of the paper is organized as follows. In Section II, the problem statement and a review of the various problems impairing the effort are given. In Section III the proposed Direct RFFM is presented and in Section IV various implementation and performance issues are addressed. In Section V, proof of the effectiveness of the proposed method on benchmark structures is provided. Conclusions are drawn in Section VI. II. PROBLEM STATEMENT AND COMPLEXITY REVIEW Consider the planar stratied medium shown in Fig. 1. It is assembled by layers that each one, (the th), is characterized , the relative permittivity, , and the relative by the height, . permeability, Various structures can t in this model. Their full wave analysis can be provided by the MoM, which in turn should make use of the mixed potential integral equation, (MPIE), [11].

0018-926X/$26.00 2011 IEEE

KAIFAS: DIRECT RATIONAL FUNCTION FITTING METHOD FOR ACCURATE EVALUATION

283

or in pole-residue form

(3)

The rst identity is extensively used in DCIM approaches, [4] and [5], [16] and [17]. Here we focus on the second one. III. DIRECT RFFM In the current section the details of the evaluation of the SI by the use only of equivalent line sources are given. A. Direct Rational Fit
Fig. 1. Generic multilayered structure.

In both [6] and [7] the SDGF is written as (4) being the th pole and residue and being the with and total number of poles, used. term according to [6] represents quasiThe static terms that refer to the high spatial frequency part of the spectrum which is responsible for the near-eld behavior of the space domain GF. The remaining, dynamic, part of the spectrum is written in a pole-residue form. The quasi static terms employed by the procedure are not unique and their computation can become problematic. Their extraction scheme relies on the fact that for high spatial frequencies the waves emitted by the source are mainly determined by the material interfaces near to the source. The accurate representation of the dynamic part of the SDGF in pole-residue form is provided by a modied version of the VECTFIT algorithm. represents the asymptotic According to [7], terms dominant in the near eld and terms that account for the continuous spectrum, (due to the branch cut), in the far eld. Indeed, previous implementations of the rational t method provided inaccurate far-eld results in the vicinity of interfaces for some congurations. This was due to the lack of lateral wave contributions in the closed-form cylindrical wave representations in the far-eld region. Thus, a new correcting term was properly added in [7]. In this work, the accurate representation of the dynamic part of the SDGF in pole-residue form is provided by the total least squares algorithm. Unfortunately, even though these approaches provide accurate results, they may require problem specic preprocessing. In contrast to the previous works we propose the direct use of the RFFM (5) Equation (5) means that we avoid the use of the term which hinders the effort to fully automate the SI computation procedure. In this way, after (1), (3),

There, the spatial domain GFs are given in the form of the vector and scalar potentials of elementary sources. It is the computation of those potentials, given in the form of the SIs that makes the whole effort rather demanding. Indeed, in general, using formulation C of [12] for the MPIE, the spectral and spatial GF vector or scalar potentials are associated by the SIs given by the following equation:

(1)

with being the horizontal and the vertical distance between the source and the observation points. is the radial wavenumber. , (associated with and ), is the is the vertical wavenumber of the th layer. wavenumber and and are the spatial and spectral domain GFs, (see for example [11] and [12] for the SDGF closed form computation). is the Hankel function of the second kind and of order . The integration in (1) is along the Sommerfeld integral path, (SIP). For the structure given in Fig. 1, there is only one branch point and associated branch cut singularities, related with the free space layer. If the PEC is removed and and/or go to innity, two branch cut singularities are supported. There are also surface waves between the minimum and maximum wavenumbers of the layers. If LHM and/or metals, in optical frequencies, are used then the surface waves can extend to a greater range. The evaluation of (1) can be done, apart from numerical integration, [13] and [14], expressing the SDGF in complex exponential series, [making use of the Sommerfeld Identity, (see also [15])]

(2)

284

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

and (5), the space domain form of the VECTFITted GF is stated as

C. Weighted Modied VECTFIT The accurate, robust and efcient representation of the whole SDGF is provided by the VECTFIT algorithm. The objective of the VECTFIT is to identify the unknown co, of the transfer function such that the efcients, least-squares distance between the rational function and the , is minimized over a predened data samples, . The modied radial wavenumber sample range of interest, VECTFIT algorithm minimizes the following nonlinear cost function:

(6)

B. Sampling Path(s) Efforts involved with the traditional RFFM tried to avoid the large variations of the spectral kernel. In the application of the method of total least squares, [7][9], it was convenient to sample along a path in the complex -plane that detours around the singularities. In [6], a path very similar to the one used in the two levels DCIM was adopted. Note that those paths dynamic are employed for the part of the spectrum only and they pass away from the areas of rapid variations of the spectrum. In this paper, we employ the reverse tactics: we try to include more variations, and thus more information, of the spectral kernel before we employ the modied VECTFIT. The idea is implemented by moving the path closer to the singularities. In this way, the line sources that were supposed to approximate the spatial domain GFs only in the medium distance region, can be also accurate both in the near and in the far eld regions. Let us consider the three sampling paths dened in the following:

(8) in an iterative fashion. In (8), is the total number of samples. One can further modify the previous equation introducing weights:

(9) Here we propose the use of (10) because the respective cost function, given in the next equation, when the range of the sampling path and the number of sampling points go to innity

(7) Note that is a small number or even zero. is assembled by logarithmically equally spaced points that range to large values of the radial wavenumber . This is made in order to capture mainly the near eld behavior of the GF. , which is a path of uniformly sampled points, is used to sample the SDGF values that reside on relative small values of the radial spectral variable, . This is made in order to capture mainly the intermediate eld behavior of the GF. is focused on the branch point, , On the other hand, to capture the branch cut contribution. Those paths can be used in a relay race fashion or the VECTFIT can be employed in their concatenation/union in a single effort run. The previous paths are not optimum. Better results can be attained by fewer points scattered, adaptively, along the sampling route. Nevertheless, for the scope, focus and orientation of the present work the paths given in (7) together with their concatenation are adequate.

(11) provides an estimate of the error introduced in a specic point, , into the spatial domain due to the square of the whole error in the spectral domain. Thus VECTFIT enforced in (9) tries to radial distance minimize the error estimate mainly on the between the source and observation points. D. Direct RFFM Procedure In detail the procedure of the proposed Direct RFFM is as follows. Surface wave poles are accounted for by implementing the VECTFIT algorithm employing a sampling path between the minimum and the maximum radial wavenumbers of the stratied media. From the provided poles those are selected that have negative imaginary part and a real part between and . Note that this step is optional and can be readily integrated into the next one. Nevertheless its use usually results in reducing both the number of the sampling points and the number of the used poles.

KAIFAS: DIRECT RATIONAL FUNCTION FITTING METHOD FOR ACCURATE EVALUATION

285

The asymptotic/quasi static terms are accounted for by implementing the VECTFIT algorithm employing a sampling path on or close to the real axis. The range of this path should include large values of the radial wavenumber in order to properly characterize the near eld spatial domain behavior of the GF. The proposed weights, (see Section III-C), may be necessary, depending on the desired accuracy range in the near eld. Furthermore, relative dense sampling for the rst few periods of the spectrum may be necessary for the accurate representation of the intermediate eld. The branch cut/continuous wave contribution is accounted for by implementing the VECTFIT algorithm employing a dense sampling path near the branch point on the proper Riemann sheet. Armed with the listed steps, the proposed procedure is proved to be accurate and efcient in all the cases examined. It is highly appropriate in this point to note that although the presented method is an evolution of [6] and [7], it bears a strong resemblance to the Direct DCIM proposed in [16]. Both methods are direct. In [16] and here the whole spectrum is tted by the same type of tting functions, exponentials and rational function respectively. Both methods propose that sampling paths should pass near the singularities. Nevertheless, in paths has to be assembled of constant sample [16], the density strait lines. This constraint is highly restrictive. That is the reason why latest DCIM approaches, [17], in order to achieve excellent results using reasonable recourses, have given up using a single type of functions to t the whole spectrum (the surface waves contribution is extracted rst). Here, the paths can by arbitrary, can even have loops, [6], and the sampling density can vary. Also proper weighting can be introduced, (see Section III-C). Furthermore, the paths can be extracted adaptively to save recourses. IV. IMPLEMENTATION AND PERFORMANCE ISSUES In the current section, implementation and performance issues for the evaluation of the SI by the Direct RFFM are addressed. We will study our method on the computation of known identities and specically the tests are on the Sommerfeld identity, (2), and its second derivative with respect to the vertical distance
Fig. 2. Performance of the proposed Direct RFFM for the integral in (2) with z :  (a) Relative error versus the electrical horizontal distance between source and observation points. (b) Respective result.

=00

A. Sampling Path(s)-Relay Race First we will determine the relation between the sampling regions, (paths), in the spectral domain and respective areas in the spatial domain. Indeed, sampling on specic regions of the spectral variable results in characterizing respective regions of the radial distance between the source and the observation points. Let us start from the rst identity. The dotted curve of Fig. 2(a) represents the relative error achieved in the spatial do, , main GF when the linear path is used, ( ). Although it provides an accuracy number of poles window of nearly 3.5 decades of radial distance for respective error of 0.1% other paths can exhibit better performance. The continuous curve of Fig. 2(a) is provided by the logarithmic , , number of poles ), while path, ( the dash-dotted curve by the branch focused path, (concatena, ), and path ( , tion of path ( ), number of poles ). Respective curves in Fig. 2(b) give the Direct RFFM result on the evaluation of (2), . for

(12) The presented results for the Sommerfeld identity are for . The results for (12) are for . The performance of the method is assessed through the provision of the , as a funcrelative error in log format, , between source tion of the horizontal distance, and observation points.

286

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2 proves that the presented method can provide an accurate relay race representation of the spatial domain multilayer GF in a broad range of distances between the source and the observation points. Even more, the achieved range can be extended both in the near and in the far eld. For the near eld we can use again the previously used logarithmic path but now we employ nonuniform weights. Specically we use the weights of (10) with equal to . Note that the use of weights has transformed the continuous curve into the dashed curve and the 0.1% accurange (-2.8, 1.8) racy window has moved from the to (-6.5,-1.7). Let us now examine the performance of the method for the second identity. The dash-dotted curve of Fig. 3(a) represents the relative error achieved in the spatial domain GF when the , , number of poles logarithmic path, ( ), is used. The continuous curve of Fig. 3(a) is provided by the branch focused path, (concatenation of path , and path , number of poles ). Respective curves on Fig. 3(b) give the Direct RFFM result on . the evaluation of (12), Note that the near eld is represented with adequate accuracy. , More than 6 signicant digits, for are shown for the dash-dotted curve. Thus, while the Hankel functions have a logarithmic singularity, their linear combination, as it is given by the modied VECTFIT algorithm, does not exhibit this singularity at least for the distances studied here. This should be expected, for the case of nonzero vertical distance between source and observation points, as it was shown in [6]. The achieved accuracy range can be extended into the far eld region by intensifying the focus on the branch point. Indeed using the following paths , , path concatenation of path , and path , , number of poles , , number of poles the dashed and dotted accuracy curves are produced, respectively. From the presented results one can make various useful comments. Increasing the number of sampling points and the number of the poles increases the accuracy of the representation. There is a correspondence between spectral path used and spatial distance area accurately characterized. This fact can lead to accurate relay race implementation of the Direct RFFM. The weighed modied VECTFIT can be useful especially when studying the response of the system for extremely small horizontal sourceobservation distances. The Hankel case is proved functions logarithmic singularity for the not to be a problem at least for the distances studied here. Last, intensifying the focus of the path on the branch point provides accurate far eld results when the structure presents losses. B. Single Effort-Path Concatenation While in the previous subsection we examined the performance of the method in a relay race fashion, here we focus on the single effort case. We use path concatenation in order

Fig. 3. Performance of the proposed Direct RFFM for the integral in (12) with z :  (a) Relative error versus the electrical horizontal distance between source and observation points. (b) Respective result.

= 0 001

to examine the methods ability to characterize large horizontal distance ranges using a single set of pole-residue couples. Using 30 pole-residue terms, the performance of the proposed Direct RFFM is depicted in Fig. 4. Fig. 4(a) is given for the integral in (2) while Fig. 4(b) is for the integral in (12). Specically, the paths used are the logarithmic and branch focusing ones. In , , and the examples we used or 4 and . The curves with or symbols in while the curves Fig. 4 are for . Also without symbols are for the continuous and curves are when is included in the process while the dotted and ones are in the case where is not used and thus the branch cut contribution is not properly accounted for. Note in Fig. 4(a) that extending the integration range, (even while at the same time keeping the number of the ), improves the accusampling points unaltered, term racy in the near eld region. Also the inclusion of the improves the accuracy in the far eld region. In Fig. 4(b) note does not that extending the integration range affect the accuracy in the near eld. This is because in the range to the integrand is practically zero due to the from

KAIFAS: DIRECT RATIONAL FUNCTION FITTING METHOD FOR ACCURATE EVALUATION

287

Fig. 4. Performance of the proposed Direct RFFM in terms of relative error versus the electrical horizontal distance between source and observation points. (a) For the integral in (2) with z :  , (b) For the integral in (12) with z :  .

= 0 001

= 00

damping exponential, . Indeed, since we sample near the real axis the vertical wavenumber moves on the negative axis. For the examples given, an adequate, ( imaginary signicant digits ), accuracy window is clearly depicted in Fig. 4. C. Representation of the Spherical and Lateral Waves Through the Cylindrical Ones While in Sections IV-A and IV-B proper rational function sampling paths were introduced and performance issues were addressed, in the current subsection details focusing on the mechanism of representation of the SDGF through the cylindrical waves are given. It is interesting to study the behavior of the partial contribution of the VECTFIT produced poles. Both the residues and the respective poles are properly scattered into their domains in order to represent the respective function. As it is shown in Figs. 24 the pole-residue expansion is able to represent the approximating SDGF in the near, intermediate and far eld
Fig. 5. Representation of spherical waves by the proposed Direct RFFM. (a) Partial contribution of each of the pole-residue couples versus the electrical horizontal distance between source and observation points. Turning points are indicated by dots. (b) Residue magnitude versus normalized by k imaginary part of the pole. (c) Normalized imaginary part of the pole versus the normalized real part of the pole. The results are for the integral in (2) with z :  . The path ;N used is the branch focusing one with concatenation of path t , and path t ;N and number of poles equal to 20. These results correspond to the red colored curve of Fig. 2.

10 )

=5

= 20)

=00 ( =1

regions. Specically, the pole-residue expansion can represent both spherical [Fig. 5(a)] and lateral [Fig. 6(a)] waves.

288

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

area the contributing term becomes negligible. Indeed, for small distances the partial contribution of a pole has a logarithmic distance between the source and the obserform. As the vation points increases, an exponential decay, associated with the imaginary part of the pole, is exhibited. In between, the behavior is observed, especially for poles with small imaginary part, [see Fig. 6(a)]. The pole-residue couples exhibit a specic ordering necessary to perform proper rational function approximation for spherical or lateral waves. The higher is the logarithmic part, the faster is the exponential decay. This ordering results from the fact that big residues are associated with big imaginary part of the respective pole. Indeed, in Figs. 5(b) and 6(b) the magnitude of the residue with respect to the normalized by imaginary part of the pole is given for the case of both the spherical and the lateral waves, respectively. An estimate of the adequate representation region/area, for the case of spherical and lateral waves, is between the smaller and bigger exponential decay turning points. The abscissas of those points are the solution to the equation

(13) The exponential decay points are shown both in Figs. 5(a) and 6(a) for each one of the pole-residue partial contribution curves. Let us now examine the placement of the poles in the -plane. In Figs. 5(c) and 6(c), the normalized by imaginary part with respect to the normalized real part of the poles, are given. Note that for the lateral wave representation the real part of the poles is practically equal to . Before we close the current section we make a last note about the determination of the proper number of the poles, , needed. If this number is exceeded then each new pole added is assigned a residue of negligible magnitude and practically exhibits no contribution. From the stated results it is proved that if the Direct RFFM is given enough information to approximate the spectral kernel then it can easily approximate the spatial domain GF in the fareld region. This holds irrespective of the fact that the GFs have or while the surface-wave the decay rate of poles have a different one, , or exponential). From a theoretical point-of-view, a continuous function, (hence and the GF), can be analytically expressed by a set of functions, (expan), through the minimization of a sion basis, here the norm representing the distance between the function and its approximating expansion. Following this path, accurate results are provided not only for the far but also for the intermediate and near eld regions. V. BENCHMARK PROBLEMS Reviewing the relative literature, there are problems hampering the accuracy of most of the proposed methods. Indeed, the following cases appear to be quite demanding: 1) no poles; 2) pole near branch cut; 3) lossy materials;

Fig. 6. Representation of lateral waves by the proposed Direct RFFM. (a) Partial contribution of each of the pole-residue couples versus the electrical horizontal distance between source and observation points. Turning points are indicated by dots. (b) Residue magnitude versus normalized by k imaginary part of the pole. (c) Normalized imaginary part of the pole versus the normalized real part of the pole. The results are for the integral in (12) with z :  . The ,N and number path used is the branch focusing one with t of poles equal to 20. These results correspond to the brown colored curve of Fig. 3.

=9

= 0 001 = 10

Note that each term is active, (contributing), in a specic distance range, which is determined by the magnitude of the residue and the location of the respective pole. Outside of this

KAIFAS: DIRECT RATIONAL FUNCTION FITTING METHOD FOR ACCURATE EVALUATION

289

Fig. 7. Log -Magnitude of the scalar potential GF for the PEC backed slab with parameters h = 10 mm, relative permittivity " = 4:4, and relative permeability  = 1. The frequency of operation is f = 4:075 GHz.

Fig. 8. Log -Magnitude of the scalar potential GF for the PEC backed slab with parameters h = 10 mm, relative permittivity " = 4:4 1 (1 0 j 1 0:02), and relative permeability  = 1. The frequency of operation is f = 10 GHz.

4) left-handed materials; 5) media with two branch points. The excellent performance of the Direct RFFM for the cases where the poles are absent has already been given in the previous subsection since (2) provides the eld of a (Horizontal) Electric Dipole, ((H)ED), immersed into free space. The presented method exhibits the same performance in the case of the with no excited surface waves. vector potential For cases 2 and 3, we will apply our method on benchmark problems both given in [14] and [8]. The structure under study is a PEC backed slab with a HED at the interface between the dielectric and the free space. For the second case, (pole near mm, relative branch cut), the dielectric layer has height , and relative permeability . The permittivity GHz, [14, Fig. 6(a)], [8, frequency of operation is Fig. 4b]. For the third case, (losses are present), the dielectric mm, relative permittivity layer has height , and relative permeability . The frequency GHz, [14, Fig. 4(b)], [8, Fig. 8(b)]. of operation is The sampling paths are those that used and tested in and included, the Section IV-B with . In both Figs. 7 and 8 the magnitude of the scalar potential GF is given with respect to the horizontal distance between the HED and the observation point that also resides at the interface. The dotted curves present the surface wave contribution, the dashed the remaining while the continuous ones the total eld. The inner boxes present the respective ranges of the gures of [14] and [8]. For the fourth case we will apply our method on a LHM benchmark problem given in [17]. The structure under study is a PEC backed LHM slab with a HED at the interface with cm, relathe free space. The LHM layer has height , and relative permeability tive permittivity . The frequency of operation is GHz. In Fig. 9, the magnitude of the scalar potential GF is given with respect to the horizontal distance between the HED and the observation point that also resides at the interface.

Fig. 9. Log -Magnitude of the scalar potential GF for the PEC backed lossy LHM slab with parameters h = 15:5 cm, relative permittivity " = 020j 10:01, and relative permeability  = 01:5 0 j 1 0:01. The frequency of operation is f = 1 GHz.

The sampling path is given as a concatenation of two linear and path b: paths, (path a: ) and a branch focusing one, . Four pole-residue terms for the surface waves plus 16, (same number with the ones used in [17]), for the rest of the SDGF are used. The dotted colored curves present the surface wave contribution, the dashed colored the remaining while the continuous one the total eld. Note that, the current example is included in order to prove that the proposed scheme is at least as efcient as the [7][9] and [17]. Indeed, with a combination of 16 plus 4, (to account for the surface waves), Hankel functions the proposed approach is able to characterize the structure for 9 orders of magnitude with respect to the horizontal distance of the source and the observation points. Last, we will apply our method on a two branch cut case studied in [18]. The structure under study is a dielectric slab

290

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

GHz. The semi-innite dielectric materials have relis and and relative permeability ative permittivity . The locations of the observation and the source, both (HED), points are at the interface between the air and the upper boundary of the dielectric layer. The initial implementation of the VECTFIT between gives a pole with respective residue . The sampling path is given as a concatenation and two branch of a linear path, , with focusing ones. The one on , and the other on , , again with . 16 poles were used in order to characterize the behavior of the structure. is given In Fig. 10 the magnitude of the vector potential with respect to the horizontal distance between the HED and the observation point. The dotted curves present the pole contribution, the dashed the remaining while the continuous one the total eld. In Fig. 10(a), only the linear path is used and that is why there is no branch point contribution. In Fig. 10(b), the branch point are used linear and the path focusing on the , is inand thus the contribution of the rst branch point, reduction of the magnitude cluded. See that there is a . The second branch cut contribution is not for included and this is readily understood due to the lack of intermagnitude ference pattern. There are no oscillations of the and this indicates that the contribution of for the rst branch cut dominates. In Fig. 10(c), after the inclusion branch focusing path the interference pattern is apof the parently there and thus the second branch cut is included into the analysis. To the best of our knowledge, currently there is no other method apart from the one presented here that can cope with structures that support two branch points. The lack of approximating the second lateral wave has been readily alleviated by introducing an additional sampling path on the second branch point. The current example again proves that a combination of cylindrical waves can accurately represent lateral waves. VI. CONCLUSION A Direct RFFM has been introduced in this paper. Compared with the traditional RFFM, this method is characterized by three key points. 1) No quasi-static or asymptotic or branch cut contribution terms are extracted before the rational tting. 2) The method includes more variations of the spectral kernel introducing proper paths. 3) Furthermore, proper weighting of the VECTFIT is proposed in order to guide the algorithm in providing increased accuracy in specic desired areas of the horizontal distance between the source and observation points. The above result in a spatial-domain GF, (expressed as a sum of cylindrical waves), that is accurate both in the near, the intermediate and the far-eld regions. REFERENCES between two different semi-innite dielectric materials. The dimm, relative permittivity , electric slab has height and relative permeability . The frequency of operation
[1] L. B. Felsen and N. Marcuvitz, Radiation and Scattering of Waves. New York: IEEE Press, 1994. [2] W. C. Chew, Waves and Fields in Inhomogeneous Media, ser. Electromagnetic Waves. New York: IEEE Press, 1995.

Fig. 10. Log -Magnitude of the vector potential GF for a two branch cut case. The structure under study is a dielectric slab between two different semi-innite dielectric materials. The dielectric slab has height h = 10 mm, relative permittivity " = 2, and relative permeability  = 1. The frequency of operation is f = 1 GHz. The semi-innite dielectric materials have relative permittivity " = 1, (free space), and " = 4 respectively and relative permeability both  = 1. (a) No branch point contribution, (b) Inclusion of the rst branch point contribution, and (c) Inclusion of the second branch point contribution.

KAIFAS: DIRECT RATIONAL FUNCTION FITTING METHOD FOR ACCURATE EVALUATION

291

[3] J. R. Wait, Electromagnetic Waves in Stratied Media. New York: IEEE Press, 1996. [4] D. C. Fang, J. J. Yang, and G. Y. Delisle, Discrete image theory for horizontal electric dipoles in a multilayered medium, Proc. Inst. Elect. Eng., pt. H, vol. 135, pp. 297303, Oct. 1988. [5] Y. L. Chow, J. J. Yang, D. G. Fang, and G. E. Howard, A closed-Form spatial Greens function for the thick microstrip substrate, IEEE Trans. Microw. Theory Tech., vol. 39, no. 3, pp. 588592, Mar. 1991. [6] V. N. Kourkoulos and A. C. Cangellaris, Accurate approximation of Greens functions in planar stratied media in terms of a nite sum of spherical and cylindrical waves, IEEE Trans. Antennas Propag., vol. 54, no. 5, pp. 15681576, May 2006. [7] R. R. Boix, F. Mesa, and F. Medina, Application of total least squares to the derivation of closed-Form Greens functions for planar layered media, IEEE Trans. Microw. Theory Tech., vol. 55, no. 2, pp. 268280, Feb. 2007. [8] F. Mesa, R. R. Boix, and F. Medina, Closed-Form expressions of multilayered planar Greens functions that account for the continuous spectrum in the far eld, IEEE Trans. Microw. Theory Tech., vol. 56, no. 7, pp. 16011614, Jul. 2008. [9] R. R. Boix, A. L. Fructos, and F. Mesa, Closed-Form uniform asymptotic expansions of Greens functions in layered media, IEEE Trans. Antennas Propag., vol. 58, no. 9, pp. 29342945, Sep. 2010. [10] A. G. Polimeridis, T. V. Yioultsis, and T. D. Tsiboukis, A robust method for the computation of Greens functions in stratied media, IEEE Trans. Antennas Propag., vol. 55, no. 7, pp. 19631969, Jul. 2007. [11] K. A. Michalski and J. R. Mosig, Multilayered media Greens functions in integral equation formulations, IEEE Trans. Antennas Propag., vol. 45, no. 3, pp. 508519, Mar. 1997. [12] K. A. Michalski and D. Zheng, Electromagnetic scattering and radiation by surfaces of arbitrary shape in layered mediaPart I: Theory, IEEE Trans. Antennas Propag., vol. 38, no. 3, pp. 335344, Mar. 1990. [13] J. R. Mosig, Integral equation technique, in Numerical Techniques for Microwave and Millimeter-Wave Passive Structures, T. Itoh, Ed. New York: Wiley, 1989, pp. 133213. [14] J. R. Mosig and A. A. Melcn, Greens functions in lossy layered media: Integration along the imaginary axis and asymptotic behavior, IEEE Trans. Antennas Propag., vol. 51, no. 12, pp. 32003208, Dec. 2003.

[15] Y. Ge and K. P. Esselle, New closed-Form Greens functions for microstrip structures theory and results, IEEE Trans. Microw. Theory Tech., vol. 50, no. 6, pp. 15561560, Jun. 2002. [16] M. Yuan, T. K. Sarkar, and M. S. Palma, A direct discrete complex image method from the closed-Form Greens functions in multilayered media, IEEE Trans. Microw. Theory Tech., vol. 54, no. 3, pp. 10251032, Mar. 2006. [17] A. Alparslan, M. I. Aksun, and K. A. Michalski, Closed-Form Greens functions in planar layered media for all ranges and materials, IEEE Trans. Microw. Theory Tech., vol. 58, no. 3, pp. 602613, Mar. 2010. [18] A. Alparslan, Study of Greens functions of potentials and elds in layered media composed of left-handed and right-handed materials, M.S. thesis, Koc Univ., , 2008.

Theodoros N. Kaifas (M08) received the Ph.D. degree from the Department of Physics, Aristotle University of Thessaloniki (AUTH), Thessaloniki, Greece, in 2004. From 1992 to 2004, he was a Research Assistant, and since 2004, has been a Postdoctoral Research Fellow with the Radio Communications Laboratory (RCL), Department of Physics, AUTH. From 2001 to 2008, he was a Scientic Associate with the Department of Electronics, Technological and Educational Institute of Thessaloniki, Greece, and from 2006 to 2007, with the Department of Technology Management, University of Macedonia, Greece. He has a broad knowledge and experience on Information Technology and Communications and an expertise on Wireless Communications, RF/Microwave Circuits, Antenna Analysis and Design, and Computational Methods in Electromagnetics accompanied by extensive contribution in relative literature and funded research projects. Dr. Kaifas is a member of the Hellenic Physicist Union and the Hellenic Club of Electronic Physicists.

292

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Integral Equation Modeling of Doubly Periodic Structures With an Efcient PMCHWT Formulation
Samuel Nosal, Paul Soudais, and Jean-Jacques Greffet
AbstractA surface integral equation modeling is described for complex doubly periodic structures. To avoid long computations of the slowly convergent pseudoperiodic Greens function, ctitious surfaces between translated unit cells are set in order to bound regions of the structure within the symmetry cell and use the freespace Greens function. The integral operators on top and bottom surfaces are computed with an algorithm originally used for planar frequency-selective surfaces. This approach uses a unique periodic PMCHWT formulation in all the regions, with two different kinds of Greens function. The method and its advantages are illustrated by two cases in the near-IR domain and in the radar domain. A frequency selective structure is studied, that shows a large at-top bandwidth under oblique incidence and TM polarization. Index TermsBoundary integral equations (BIE), frequency selective surfaces (FSS), hybrid method, metamaterials, method of moments (MoM), periodic Greens function.

I. INTRODUCTION OUBLY-PERIODIC (or biperiodic) structures can be used to design frequency selective radomes or antenna substrates. At IR or optical wavelengths, they can be used as antireective coatings or lters. These structures are examples of metamaterials and photonic crystals. As devices built on these structures are getting more and more complex, there is a need for versatile numerical methods for the scattering from complex biperiodic structures with short computation times. Purposefully, we present here a method based on surface integral equations (SIE), which is able to handle biperiodic structures with a nite thickness made up of an arbitrary conguration of perfectly conducting (PEC) and homogeneous isotropic dielectric regions. Numerical methods have been developed for decades to model the diffraction by periodic structures, but they strike a

Manuscript received December 21, 2009; revised May 23, 2011; accepted June 16, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The work of S. Nosal was supported by the DGA (National Armament Directorate, French Ministry of Defense), through a Ph.D. scholarship. S. Nosal was with EM2C Laboratory, cole Centrale Paris. He is now with the Laboratory for Electromagnetic Fields and Microwave Electronics, Swiss Federal Institute of Technology, ETH-Zentrum, ETZ K97, CH-8092 Zurich, Switzerland (e-mail: samuel.nosal@ensta.org). P. Soudais is with Dassault-Aviation, 92210 Saint-Cloud, France (e-mail: paul.soudais@dassault-aviation.com). J.-J. Greffet was with EM2C Laboratory, cole Centrale Paris. He is now with the Charles-Fabry Laboratory, Institut dOptique, Universit Paris Sud, CNRS, F-91127 Palaiseau cedex, France (e-mail: jean-jacques.greffet@institutoptique. fr). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167898

compromise between applications to specic geometries (e.g. layered planar structures), accuracy and computation time. Among them, the SIE-based methods take a large place: the method of moments for planar FSS [1][3] has been proven a fast and accurate method. It is limited to FSS made of innitely thin metallic or resistive patches, whose shapes can be described on a regular grid. Two extensions for stacked FSS separated by dielectric layers can be used. The scattering problem can be handled by a S-matrix cascading technique [1] or layered Greens functions [4]. If the symmetry cell is made of regions of 3D arbitrary shape, one can use volume methods with periodic boundary conditions (PBC) based on nite difference time domain (FDTD) method [5] or nite element (FE) methods (e.g. [6] and [7]). But the number of unknowns for volume formulations will usually be much higher than for surface formulations. Another group of methods is based on pseudoperiodic SIE. They involve the pseudoperiodic Greens function, which is a very slowly convergent double series. Several techniques to accelerate the convergence of this series have been studied, some of them have been recently compared in [8]. A pseudoperiodic SIE method was proposed in [9]. A pseudoperiodic PMCHWT (from Poggio, Miller, Chang, Harrington, Wu and Tsai) formulation was used in all the regions. The large computational cost of the Greens function for doubly periodic problems was alleviated by a tabulation of both the pseudoperiodic Greens function and its gradient. In order to reduce the number of evaluations of the pseudoperiodic Greens function, hybrid methods were suggested: by hybrid method, we designate a method that solves differently the outer region (often vacuum) and the inner region (the doubly periodic structure). In [10], the outer innite regions are treated with SIE using the pseudoperiodic Greens function, while the inner bounded regions in the biperiodic structure are treated with a FE method. These methods are often referred to as hybrid nite element-boundary integral (FE-BI) methods. Their main problem is that the (volume) FE method requires the creation of many degrees of freedom. A hybrid method using two different SIE formulations was developed in [11]. The internal regions of the structure were bounded with ctitious surfaces, and then the free-space Greens function was used inside the structure. This resulted in a tremendous reduction in the assembly time, usually offsetting by far the extra-solution time due to the extra-unknowns on the ctitious surfaces. The outside problem was formulated with a MoM derived from the MoM for planar FSS of [1][3]. In this paper, we present a hybrid method that uses only one integral formulation, but both the pseudoperiodic Greens func-

0018-926X/$26.00 2011 IEEE

NOSAL et al.: INTEGRAL EQUATION MODELING OF DOUBLY PERIODIC STRUCTURES WITH AN EFFICIENT PMCHWT

293

tion and the free-space Greens function by using ctitious surfaces to bound the inner regions of a unit cell. The key point of the method, as presented in Section II, is the use of an adapted PMCHWT formulation [12], which is often used for general purpose codes dealing with several dielectric or conducting regions and impedance conditions: for 2D and 3D scatterers [13], for bodies of revolution [14], for large scatterers [15]. This formulation requires the computation of contributions from the entire boundary of regions neighboring the symmetry cell and we will show how to recover these terms in an easy way. To the best of our knowledge, it is the rst time that a pseudoperiodic formulation is presented, that uses a unique PMCHWT formulation for all the regions, including regions bounded by ctitious surfaces with PBC. Another important point is that the PMCHWT contribution from the top and bottom unbounded regions are efciently computed thanks to a method derived from a Method of Moments (MoM) specic to planar frequency selective surfaces (FSS). This is achieved by using triangular RWG functions on part of the mesh and rooftop basis functions on top and bottom surfaces, as presented in Section III. Note that in [11], the top planar surface was meshed with isosceles rectangle triangles and a linear combination of the triangular basis functions had to be made to obtain rooftop basis functions on the top surface. The use of these techniques makes it possible to obtain a numerical method, whose computational time is similar to that of a case where the symmetry cell would be computed as an isolated object (i.e. a non-periodic 3D case with similar number of unknowns), as illustrated by the validation cases in Section IV. II. PSEUDOPERIODIC PMCHWT FORMULATION A. Description of the Pseudoperiodic Problem 1) Biperiodic Scattering Structure: A biperiodic diffractive structure consists of a unit cell which is repeated over two diand rections of periodicity, given by the vectors , as schematized on Fig. 1. Note than in this work, and could be made non orthogonal without any modication except in Section III-B. where the algorithm for planar FSS should be used in its skewed lattice version [1]. The unit cell is composed of an arbitrary number of homogeneous material regions with complex permittivity and permittivity or PEC regions. The array is of nite thickness and is placed between two semi-innite regions, generally air or PEC. 2) Scattering Problem: An incident plane wave impinges on the biperiodic scattering structure. The time-harmonic dependence is taken as . At a given frequency, the incident eld is completely dened by the components of its normalized wave vector along the x and y axes ( and , from which is deduced), its polarization (TE or TM), the region where its source is located (top or bottom) and its amplitude, where , being the speed of light in the region where the source is located. Due to the pseudoperiodicity of the incident eld, the biperiodicity of the scattering structure and the linearity of Maxwells equations, the scattered eld is pseudoperiodic: it is only necessary to nd its values on the symmetry cell. The values of the

Fig. 1. Denition of the reference unit cell, which is repeated according to the and . Note that only the left and bottom boundaries translation vectors with the translated cells belong to the reference unit cell (0,0).

eld in any other cell can be calculated using the following relation (1) where is the phase coefcient, being the incident wave vector. Pseudoperiodic boundary conditions on opposite faces of the symmetry cell are a consequence of (1). B. PMCHWT Formulation for 3D Problems 1) Electric and Magnetic Field Integral equations (EFIE and MFIE): For the sake of simplicity, we give here the outline of the formulation in the case where there are only two homogeneous dielectric regions, separated by a closed surface S. The extension to an arbitrary number of regions can be easily achieved. Following J.-M. Jins notation [14], we can write in each region both the electric eld integral equation (EFIE)

(2) and the magnetic eld integral equation (MFIE)

(3) where , ,

(4) (5)

and electric region .

being the relative impedance of the di-

294

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

In these equations, is the free-space Greens function in the region (=1 or 2) and the incident elds only exist in the region where their source is located. They are uniformly zero in the other region. 2) PMCHWT Formulation: The PMCHWT formulation is used for the scattering by composite dielectric objects. On the boundary S separating two dielectric regions (e.g. and ), the EFIE (2) and MFIE (3) from each region are summed to give the PMCHWT, using the fact that tangential elds are continuous across S: and . Hence, the equations that are to be solved are

Fig. 2. Top view of a junction between several dielectric regions and composite function on a junction edge.

(6)

posite function is made of two RWGs or two rooftops functions with opposite signs. The PMCHWT formulation writes for [13], [17]:

(7) The associated variational formulation of (6) and (7) is discretized by a boundary element method, with RWG and rooftop basis and test functions, both denoted by (Galerkin method). Equation (6) is tested with test functions and (7) is tested with test functions. The discretized operators read (e.g. [13]): (11) The PMCHWT formulation writes for [13], [17]:

(12) Please recall that the incident eld is non-zero only in the region where the source is located. On PEC boundaries, , and there are no test functions and we write only (11) for test functions . Note that (11) only includes contributions from dielectric regions. C. Hybrid SIE Formulation: Fictitious Surfaces (9) This leads to the following linear system, which can usually be solved directly: A way to adapt the PMCHWT formulation to doubly periodic scatterers is to replace, in each unbounded region, the Greens function by the pseudoperiodic Greens function, i.e. the Greens function for a double array of point sources in this region. Each of these Greens functions satises the PBC (1). We investigate here another way, following [11], to address the computational time issue: reducing as much as possible the number of evaluations of the pseudoperiodic Greens function. We add ctitious surfaces between the reference unit cell and its translated neighboring cells, as shown on Fig. 3. Thus the inner regions of the periodic structure are bounded, allowing the use of the free-space Greens function, whose computation cost is negligible. D. Writing a Pseudoperiodic PMCHWT Formulation With Fictitious Surfaces For the sake of simplicity, we rst consider the case where two homogeneous regions bounded by ctitious surfaces ll the

(8)

(10) Note that we use unknowns instead of unknowns in order to obtain a symmetrical discretized matrix [16]. This formulation can be generalized to an arbitrary number of dielectric regions. Due to the continuity of tangent elds, for an edge belonging to two or more dielectric regions numbered to k, there is one composite basis and test function for J and one composite basis and test function for M [17]. The value for a composite function in region d is (cf. Fig. 2). Note that it is necessary to use composite functions on every edge: on edges belonging to two dielectric regions, the com-

NOSAL et al.: INTEGRAL EQUATION MODELING OF DOUBLY PERIODIC STRUCTURES WITH AN EFFICIENT PMCHWT

295

Fig. 3. Example of a pyramid array on a silicon slab. Left: mesh used for a 3D pseudoperiodic SIE. There are three innite regions: the two outer regions are lled with air; the inner region is lled with silicon. Right: mesh used for the proposed hybrid method. Here only the outer regions are innite and the inner region is bounded by adding four vertical ctitious surfaces.

Fig. 5. Left: The PMCHWT formulation involves terms from four different unit cells. Right: Those terms can be obtained through phase relations with the terms that are directly available in the reference unit cell.

volving the image of a test function cell can be computed with:

in the (14)

Fig. 4. Pseudoperiodic electric equivalent currents on two opposite ctitious surfaces with pseudoperiodic boundary conditions.

symmetry cell (Fig. 4). The mesh is assumed to have been created so that the surface meshes on opposite ctitious surfaces are identical. Only unknowns belonging to the (0,0) symmetry cell are considered. The values of the currents in other cells can be computed thanks to the PBC (1). The pseudoperiodicity of the solution is enforced strongly by eliminating the unknowns of the boundary of cell (0,0) with cells (1,0), (0,1) and (1,1) using the PBC (1). Therefore, contributions to a matrix entry involving the image of unknown in the cell should be replaced as follows (13) We want to use a PMCHWT formulation in all the homogeneous regions for consistency and accuracy and for an easier treatment of junction cases. The PMCHWT formulation consists in summing the contribution of the integral operators on both sides of a surface. The question for the pseudoperiodic PMCHWT formulation with ctitious surfaces is how to get the contributions coming from the neighboring cells. For a test function between the (0,0) and ( 1,0) cells, the PMCHWT involves terms from the entire boundary of the and regions. We remark that the contribution from region to a test function can be obtained from the contribution from region to the corresponding test function and the PBC (1). Contributions to a matrix entry in-

, Some edges belong to four regions of Fig. 4 (e.g. , and ); on such edges, there are contributions from the entire boundary of these four regions. Each of the last three contributions is obtained from a contribution and (14). All the contributions will be computed from those of the (0,0) cell (Fig. 5), by looping over the elements bounding the regions of the reference cell and applying the PBC relations (13) and (14), accordingly to the source and reception elements. This scheme also applies to the top and bottom surface and to complex junctions (e.g. when there are several arbitrary internal PEC or dielectric regions, thin sheets, impedance boundary conditions ). III. EFFICIENT COMPUTATION OF THE INTEGRAL OPERATORS ON THE TOP AND BOTTOM BOUNDARIES A. Structured Mesh on Top and Bottom Boundaries We use RWG and rooftop elements on composite meshes made of triangles and quadrangles. Only the top and bottom regions contributions will involve computations of the pseudoperiodic Greens function. If the boundaries of the top and bottom regions are planar, it is possible to use a structured mesh made of identical quadrangles. If the top (resp. bottom) boundary is not, we suggest to add a ctitious slab on the top (resp. bottom) of the biperiodic structure, in order to use a structured mesh on its top (resp. bottom) boundary, as illustrated on Fig. 6. B. Use of the Algorithm for Planar FSS Screens The algorithm for planar FSS of [1], [3] is based on the PEC or resistive version of the EFIE (2). In fact, this algorithm leads to a highly efcient discretization of the operator with the pseudoperiodic Greens function for a planar surface discretized with identical rooftop elements. This is precisely what we need to compute the contributions of the upper and lower regions to the PMCHWT formulation, since and on the planar surfaces bounding these regions.

296

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 6. Adding a ctitious slab above the biperiodic scattering structures (a) makes it possible to have a structured mesh (c) on the top and bottom surfaces of the entire mesh (b).

Fig. 8. (a) Mesh of a single free stranding dielectric sphere. (b) Mesh with ctitious surfaces, with structured mesh on top and bottom surfaces.

Fig. 7. Array of dielectric spheres in a dielectric outer region. The radii of the or 0.23 . spheres is taken to be 0.2

The computation time of the contributions obtained with the planar FSS algorithm is much shorter than what is needed to compute the integral operators in the 3D case. The overall assembly time of this pseudoperiodic SIE method is then similar to that of a 3D SIE with the same number of degrees of freedom. IV. VALIDATION The validation of the method has been carried out extensively in [18]. Here three cases are presented, that correspond to two different frequency ranges: infrared (IR) and radar frequencies. For all validation cases, we have used a direct solution method to solve the discretized linear system. A. Array of Dielectric Spheres at Near IR Wavelengths 1) Description: The rst test case consists in a square array of dielectric spheres embedded in a dielectric innite region. This test case was suggested by T. Teperik [19] and is illustrated on Fig. 7. The lattice constants are 1 . The relative permittivity of the outer region is 2.1 and the relative permittivity of the spheres is 3.05. We compute the transmission efciency under normal incidence for wavelengths from 1.448 to 1.460 in vacuum for two spheres radii and .

Fig. 9. Comparison between analytical approach and numerical modeling (with and without ctitious surfaces) for two arrays of dielectric spheres (radii 200 nm and 230 nm).

2) Modeling: To assess the effect on accuracy and computation time of ctitious surfaces, the case has been computed with and without ctitious surfaces. For the rst model (cf. Fig. 8(a)), the pseudoperiodic Greens function is used for the external region and the free-space Greens function for the internal region. A single model with 2160 unknowns is used for both radii and corresponding to an average edge length of , where is the wavelength in the dielectric sphere. Since the number of unknowns is small, we didnt try to use a mesh with a minimal number of unknowns. In the second model (Fig. 8(b)) we add ctitious surfaces meshed with rectangles with edge length. The total number of unknowns is now 7808. 3) Results and Computation Time: We present the transmission efciency under normal incidence computed with our method and ctitious surfaces, with an asymptotic model [19] and with our method without ctitious surfaces on Fig. 9. There

NOSAL et al.: INTEGRAL EQUATION MODELING OF DOUBLY PERIODIC STRUCTURES WITH AN EFFICIENT PMCHWT

297

Fig. 10. Scheme of the FSS embedded in a dielectric slab: (a) in the xy-plane (b) in the yz-plane.

is very good accuracy between all the results with a slight frequency shift, which is negligible compared to the wavelength. The computation is carried out with an Intel Itanium II node with 8 processors at 1.4 GHz. Without ctitious surfaces, the computation time is 0.44 h for one wavelength; 99.9% of which is matrix assembly time. With ctitious surfaces, the computation time is only 0.028 h, 82% of which being spent during the matrix assembly. In this case, the ctitious surfaces add a lot of extra degrees of freedom (3.2 times more than without the ctitious surfaces) but even in this case the total computation time is improved by a factor 15.7.

Fig. 11. Three different modeling of a FSS made of an array of perfectly conducting patches embedded in a dielectric slab: (a) BEM modeling where the unit cell consists of a patch embedded in a box of dielectric (b) BEM modeling where each patch is placed on two translated unit cells (c) Stack calculation with FSS algorithm.

B. FSS at Radar Frequencies 1) Description: We consider now the case of a square array of perfectly conducting rectangular patches embedded in a dielectric slab. The patches are 5 mm long and 2.5 mm wide with lattice constants of 10 mm along the x and y directions. The FSS is placed in the median plane of a 2 mm thick dielectric slab with and . The geometrical and material characteristics of this case are illustrated on Fig. 10. 2) Modeling: In a rst model (Fig. 11(a)), the patch is in the middle of the symmetry cell. There are 3464 unknowns and the maximal edge length at 24 GHz is , where is the wavelength in the dielectric region. With the second model, we test junctions through ctitious surfaces. We set the patch across the ctitious side walls (Fig. 11(b)) also adding a horizontal ctitious surface. The surface of the PEC patch is considered as two separate surfaces with one set of test functions on each side of the patch. On junction edges between the patch and the horizontal ctitious surface, there is only one test function. There are 7050 unknowns. The maximal edge length at 24 GHz is . On this simple geometry, reference results are obtained with a planar FSS code based on [1], as shown on Fig. 11(c). 3) Results: We compare the reection efciencies for TE (H) and TM (V) polarizations, at normal and 30 incidences on Figs. 12 and Fig. 13. The comparison between the three models is very good, indicating that there are no signicant errors due to ctitious surfaces and to junctions with the ctitious surfaces.

Fig. 12. Reection efciency of the array of metallic patches embedded in a dielectric layer under normal incidence.

Fig. 13. Reection efciency of the array of metallic patches embedded in a dielectric layer under oblique incidence (30 ) for TE and TM polarizations.

298

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 14. Geometrical parameters and mesh of the unit cell of the array of evolutive coaxial cavities in a perfectly conducting slab.

Fig. 16. Transmission efciency of the array of evolutive coaxial cavities under oblique incidence (64 ) and TE (H) polarization: comparison with arrays of straight coaxial cavities.

Fig. 15. The cavity is created by rotating two arcs of circle around the z axis. The circles are passing by three points given by the inner and outer radii in and .

Fig. 17. Transmission efciency of the array of evolutive coaxial cavities under oblique incidence (64 ) and TM (V) polarization: comparison with arrays of straight coaxial cavities.

C. Array of Evolutive Coaxial Cavities in a Perfectly Conducting Metallic Slab An array of straight coaxial cavities in a silver slab in the visible optical wavelengths has been presented in [20]. This structure has band-pass properties, which are independent on the polarization and on the angle of incidence. It was shown that these properties are due to a resonance of the TE11 mode of the waveguides. Total transmission is not achieved though, due to losses in the silver slab. We present here a structure based on evolutive coaxial cavities in X band which is a broadband lter under oblique incidence in the TM polarization. 1) Description: The coaxial cavities are arranged in a square lattice 8 mm 8 mm. The perfectly conducting slab is 10 mm thick. The cavities have a horn-like shape whose vertical section is dened on Fig. 14 and Fig. 15. The cavities are lled with PTFE, with a refractive index of in order to reduce the frequency for the rst TE11 resonance.

2) Modeling: Two ctitious slabs of air are added in order to use locally structured meshes on the top and bottom boundaries. There are 16488 triangles and 512 quadrangles on the mesh we used and 31552 unknowns. All the edges are shorter than at the highest frequency, even in the PTFE. The mesh is shown on Fig. 14. 3) Results and Computation Time: The computation is carried out with an Intel Itanium II node with 8 processors at 1.4 GHz. Both the matrix assembly and the direct resolution are parallelized. For one frequency and one incidence, the computation time is 1.3 h. The transmission efciencies under oblique incidence at 30 in both TE (H) and TM (V) polarizations are presented on Fig. 16 and Fig. 17 and compared to arrays of straight cavities (narrow, wide or median). The bandwidth is narrow in TE polarization. But the bandwidth in TM polarization is larger with a at-top of 0.6 GHz. This phenomenon is due to a coupling between the TE11 and TEM coaxial modes in TM polarization (cf. [21], [22]).

NOSAL et al.: INTEGRAL EQUATION MODELING OF DOUBLY PERIODIC STRUCTURES WITH AN EFFICIENT PMCHWT

299

V. CONCLUSION We have introduced a general method to model the scattering by complex doubly periodic structures, which is based on a single PMCHWT formulation for all regions. The inner regions are bounded by using ctitious surfaces between translated unit cells. In these regions, the free-space Greens function is used with PBC. We have shown how to derive the PMCHWT contributions from regions neighboring the symmetry cell from terms internal to the symmetry cell. This integral formulation ensures the best accuracy of the nal results. The assembly time of the part of the matrix corresponding to the inner region is similar to what we would have had with a 3D case. The pseudoperiodic Greens function must be used on the top and bottom boundaries. These contributions can be efciently computed by using an adaptation of the algorithm for planar FSS. The assembly time for these contributions is negligible, due to the local Toeplitz matrix. Examples of calculation and computation times have been given. They highlight the reduction of the computation time due to the use of ctitious surfaces: for the array of dielectric spheres, it was reduced by a factor 15.7. The method is a general formulation for arbitrary geometries combining PEC and dielectric regions and all possible junctions used in 3D models. Localized or surface impedance conditions and thin wires can be taken into account. Furthermore, the method can be extended to include a local coupling with the nite element method to handle anisotropic materials.

REFERENCES
[1] R. Mittra, C. H. Chan, and T. Cwik, Techniques for analysing frequency selective surfacesA review, Proc. IEEE, vol. 76, no. 12, pp. 15931615, Dec. 1988. [2] T. A. Cwik and R. Mittra, Scattering from a periodic array of freestanding arbitrarily shaped perfectly conducting or resistive patches, IEEE Trans. Antennas Propag., vol. 35, pp. 12261234, Nov. 1987. [3] C. H. Chan and R. Mittra, On the analysis of frequency-selective surfaces using subdomain basis functions, IEEE Trans. Antennas Propag., vol. 38, pp. 4050, Jan. 1990. [4] R. A. Kipp and C. H. Chan, A numerically efcient technique for the method of moments solution for planar periodic structures in layered media, IEEE Trans. Microwave Theory Tech., vol. 42, pp. 635643, Apr. 1994. [5] P. Harms, R. Mittra, and W. Ko, Implementation of the periodic boundary-condition in the nite-difference time-domain algorithm for FSS structures, IEEE Trans. Antennas Propag., vol. 42, pp. 13171324, Sept. 1994. [6] J. M. Jin, The Finite Element Method in Electromagnetics. New York: Wiley, 1993. [7] J. M. Jin, Finite Element Analysis of Antennas and Arrays. New York: Wiley, 2009, ch. 9. [8] G. Valerio, P. Baccarelli, P. Burghignoli, and A. Galli, Comparative analysis of acceleration techniques for 2-D and 3-D Greens functions in periodic structures along one and two directions, IEEE Trans. Antennas Propag., vol. 55, pp. 16301643, June 2007. [9] X. Dardenne and C. Craeye, Method of moments simulation of innitely periodic structures combining metal with connected dielectric objects, IEEE Trans. Antennas Propag., vol. 56, pp. 23732380, Aug. 2008. [10] D. T. McGrath and V. P. Pyati, Phased array antenna analysis with the hybrid nite element method, IEEE Trans. Antennas Propag., vol. 42, pp. 16251630, Dec. 1994.

[11] L. C. Trintinalia and H. Ling, Integral equation modeling of multilayered doubly-periodic lossy structures using periodic boundary condition and a connection scheme, IEEE Trans. Antennas Propag., vol. 52, no. 9, pp. 22532261, Sep. 2004. [12] A. J. Poggio and E. K. Miller, Integral equation solutions of three-dimensional scattering problems, in Computer Techniques for Electromagnetics, R. Mittra, Ed., 2nd ed. New York: Pergamon Press, 1973, pp. 159264. [13] L. N. Medgyesi-Mitschang, J. M. Putnam, and M. B. Gedera, Generalized method of moments for three-dimensional penetrable scatterers, J. Opt. Soc. Am. A, vol. 11, no. 4, pp. 13831398, Apr. 1994. [14] J.-M. Jin, A highly robust and versatile nite element-boundary integral hybrid code for scattering by BOR objects, IEEE Trans. Antennas Propag., vol. 53, pp. 22742281, Jul. 2005. [15] W. C. Chew, J.-M. Jin, E. Michielssen, and J. Song, Fast and Efcient Algorithms in Computational Electromagnetics. Norwood, MA: Artech House, 2001. [16] J.-J. Anglini, C. Soize, and P. Soudais, Hybrid numerical method for harmonic 3-D Maxwell equations: Scattering by a mixed conducting and in homogeneous anisotropic dielectric medium, IEEE Trans. Antennas Propag., vol. 41, Jan. 1993. [17] B. M. Kolundija, Electromagnetic Modeling of Composite Metallic and Dielectric Structures, IEEE Trans. Microwave Theory Tech., vol. 47, pp. 10211032, Jul. 1999. [18] S. Nosal, Modlisation lectromagntique de structures priodiques et matriaux articiels (in French) Ph.D. dissertation, Ecole Centrale Paris, Chtenay-Malabry, France, 2009 [Online]. Available: http://theses.abes.fr/2009ECAP0030 [19] T. Teperik, R. Sainidou, and F. J. Garcia de Abajo, Electromagnetic mode trapping for giant optical eld enhancement, presented at the META08, NATO Advanced Research workshop on Metamaterials for Secure Information and Communication Technologies Marrakesh, Morocco, 2008. [20] D. Van Labeke, D. Grard, B. Guizal, F. I. Baida, and L. Li, An angleindependent Frequency Selective Surface in the optical range, Optics Expr., vol. 14, no. 25, 2006. [21] F. I. Baida, Enhanced transmission through subwavelength metallic coaxial apertures by excitation of the TEM mode, Appl. Phys. B, vol. 78, 2008. [22] S. Nosal, P. Soudais, and J.-J. Greffet, Enhancing the bandwidth of coaxial aperture arrays in the radar frequencies, presented at the META10, 2nd Int. Conf. on Metamaterials, Photonic Crystals and Plasmonics, Cairo, Egypt, 2010.

Samuel Nosal was born in Strasbourg, France, in 1983. He received the M.S. degree in engineering from Ecole Nationale Suprieure de Techniques Avances (ENSTA), Paris, France and the M.S. degree in applied mathematics (modeling and simulation) from Institut National des Sciences et Techniques Nuclaires (INSTN), Saclay, France, both in 2006, and the Ph.D. degree in engineering sciences from Ecole Centrale Paris, Chtenay-Malabry, France, in 2009. From 2006 to 2009, he worked as a Ph.D. student at EM2C Laboratory, Ecole Centrale Paris, Chtenay-Malabry, France, and at Dassault Aviation, Saint-Cloud, France, under supervision of Prof. Jean-Jacques Greffet and Dr. Paul Soudais.

Paul Soudais received the Dipl. Ing. and Ph.D. degrees from cole Centrale de Paris, in 1988 and 1995, respectively. From 1989 to 2001, he has been a Researcher at ONERA, the French Aerospace Research Ofce, in the Structures Department, then in the Electromagnetism and Radar Department. In 1994, he was a Visiting Researcher at the Department of Electrical Engineering and Computer Science, University of Illinois at Chicago. In 2001, he joined Dassault Aviation to work on numerical codes and modeling for RCS and antenna applications.

300

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Jean-Jacques Greffet is an alumnus of Ecole Normale Suprieure de Cachan. He received the agrgation de physique degree, the M.S. degree in solid state physics, the Ph.D. degree and the Habilitation degree from Universit Paris-Sud Orsay, France, in 1982, 1983, 1988 and 1992, respectively. He was a Professor at Ecole Centrale Paris from 1994 to 2008. He is currently a Professor at the Institut Universitaire de France and at the Institut dOptique and Head of the Nanophotonics and Quantum Information Division of the Nanoscience Center in Paris (cnano). His current research interests include nanophotonics (nanoantennas, plasmonics) and radiative heat transfer on the nanoscale.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

301

Reduced-Order Models of Finite Element Approximations of Electromagnetic Devices Exhibiting Statistical Variability
Prasad Sumant, Hong Wu, Andreas Cangellaris, Fellow, IEEE, and Narayana Aluru, Member, IEEE
AbstractA methodology is proposed for the development of reduced-order models of nite element approximations of electromagnetic devices exhibiting uncertainty or statistical variability in their input parameters. In this approach, the reduced order system matrices are represented in terms of their orthogonal polynomial chaos expansions on the probability space dened by the input random variables. The coefcients of these polynomials, which are matrices, are obtained through the repeated, deterministic model order reduction of nite element models generated for specic values of the input random parameters. These values are chosen efciently in a multi-dimensional grid using a Smolyak algorithm. The generated stochastic reduced order model is represented in the form of an augmented system that lends itself to the direct generation of the desired statistics of the device response. The accuracy and efciency of the proposed method is demonstrated through its application to the reduced-order nite element modeling of a terminated coaxial cable and a circular wire loop antenna. Index TermsFinite element, Krylov methods, Model order reduction, polynomial chaos, random input, stochastic.

I. INTRODUCTION

OLLOWING a growing trend in a variety of engineering elds, model order reduction (MOR) has become one of the methods of choice for tackling the complexity of numerical electromagnetic eld modeling and simulation inside domains of large geometric and material complexity. The basic idea of MOR is to replace the original discrete system resulting from the discretization of Maxwells equations over the domain of interest, the dimension of which is too large for an expedient numerical solution to be possible, in terms of one of much smaller dimension, yet capable of approximating the electromagnetic behavior of the original one with sufcient accuracy [1][8].

Manuscript received January 31, 2011; revised June 12, 2011; accepted June 21, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the Defense Advanced Research Projects Agency (DARPA), under the N/MEMS Science & Technology Fundamentals Research Program. P. Sumant was with the Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, IL 61820 USA. He is now with ExxonMobil Upstream Research Company, Houston, TX 77098 USA (e-mail: ps.sumant@gmail.com). H. Wu is with Extreme DA Corporation, Santa Clara, CA 95054 USA. A. Cangellaris and N. Aluru are with University of Illinois at Urbana-Champaign, IL 61820 USA. Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167935

In addition to their application for expediting the calculation of the broadband response of passive electromagnetic systems (a process often referred to as fast frequency sweep), MOR has been demonstrated as an efcient technique for use in conjunction with domain decomposition modeling of large-scale electromagnetic structures (e.g., [9]) and for the purposes of sub-domain macro-modeling for the purpose of alleviating the complexity of electromagnetic analysis of multi-scale, passive electromagnetic structures (e.g., [10], [11]). In their majority, MOR applications for electromagnetic eld modeling reported in the literature consider deterministic structures of known material properties and geometry. For those cases where some of the input parameters are allowed to vary in a deterministic fashion, schemes for parameterized model order reduction have been proposed and demonstrated (e.g., [12], [13]). However, it is often the case that geometric and/or material variability of either random or difcult to predict nature may be present in a structure under investigation. For example, surface roughness is a source of modeling uncertainty pertinent to a variety of electromagnetic applications, with notable examples the prediction of attenuation in metallic electromagnetic waveguides and the analysis of electromagnetic scattering by rough surfaces. Another source of modeling uncertainty is the micro/nanomanufacturing process-induced geometric and material variability in the very-ne-scale features encountered in integrated microwave/millimeter-wave and optical devices and circuits, as well as in the context of synthesized metamaterials. The predictive electromagnetic analysis of such structures, accounting for the impact of material and geometric uncertainty, is becoming a growing need in the computer-aided exploration, design, and prototyping of such devices and synthesized materials. Of particular interest to this work is the case where one or several portions (or domains) of a complex structure exhibit such statistical variability or uncertainty in the parameters that dene their geometric and material composition. In the following, such domains will be referred to as stochastic domains. For such structures, the development of stochastic, reduced-order models of the stochastic domains offers an interesting option for facilitating and expediting system-level electromagnetic analysis in the presence of uncertainty. The demonstration of a methodology for the development of stochastic reduced order models of nite element approximations of passive electromagnetic structures is the objective of this paper. More specically, our objective is to put forward a method that treats MOR as a black-box deterministic solver; thus,

0018-926X/$26.00 2011 IEEE

302

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

the method is independent of the specic MOR algorithm employed. Our methodology takes advantage of recent advances in the ideas of generalized polynomial chaos and stochastic collocation for the polynomial interpolation of the random variables involved and the establishment of a framework for the mathematical representation of the stochastic reduced-order model. Among the numerous contributions to the rapidly growing literature on the generalized polynomial chaos methodology for stochastic computation, the review article by Xiu [14] is worth mentioning because of its comprehensive bibliography on the subject. The rest of the paper is organized as follows. The mathematical framework is developed in Section II. Following a brief review of the deterministic MOR process and the use of polynomial chaos as a means of representing uncertainty, the proposed methodology for stochastic MOR is presented for the specic case of a nite element approximation of Maxwells equations in a linear, isotropic domain. The validity and accuracy of the proposed methodology are discussed in Section III in the context of its application to the reduced-order modeling of terminated coaxial cable and a circular loop antenna. The paper concludes with a summary of the contribution and a discussion of topics of on-going research. II. MATHEMATICAL FORMULATION A. Deterministic Model Order Reduction In order to keep the mathematical framework simple, we restrict ourselves to the nite element approximation of a linear, electromagnetic boundary value problem inside a domain with isotropic, non-dispersive media. However, it should be noted that the proposed methodology is also applicable when the media is anisotropic. As far as the assumption of non-dispersive media, this is driven by our intention to demonstrate in a simple manner the fact that the proposed methodology is applicable to the reduced-order modeling in both the frequency and the time domain. For the case of dispersive media, the model order reduction framework presented in [9], [15] must be employed in place of the simpler, reduced-order modeling algorithm presented here. Following [2], we develop a nite element approximation of the vector Helmholtz equation for the electric eld using edge elements. It is assumed that the structure under modeling interacts with its exterior through ports. Hence, the Laplace-domain representation of the nite element model of the -port system is described by the following system of equations [2]

matrix is used to sample the calculated electric eld to generate the desirable output quantity. The generalized multi-port impedance matrix (GIM) is given by [2], (2) Calculation of the GIM at a given frequency requires the inversion of a nite element matrix that is of large dimension, typically in the order of several tens or even hundreds of thousands. This is where MOR offers substantial reductions in cost. A reduced-order model attempts to approximate the original system in terms of a subset of the vectors in the eigenspace of the full nite element system. More specically, the eigenvectors of interest are the ones that are the most inuential in the calculation of the response of the structure as seen by the ports involved in the denition of . As mentioned in the introduction and as it will be demonstrated later in this section, the proposed methodology for stochastic MOR is such that it is compatible with anyone of the popular approaches used in practice for deterministic MOR. For our purposes, the Krylov subspace algorithm described in [2] will be used. In this approach, a transformation matrix in is constructed, where . The reduced order system, which is of dimension , can then be computed as, (3) where,

(4) Clearly, the matrices are in .

B. Polynomial Chaos Expansion Next, we consider the way statistical variability is introduced in the nite element model. Toward this, let denote the physical domain of interest with denoting the position vector in . To keep the presentation simple, and without loss of generality, we assume that the uncertainty in the denition of the electromagnetic boundary problem of interest can be described in terms of two continuous, orthogonal random variables, , where represents an event in our probability space. For random variables that are correlated, they can always be transformed into orthogonal random variables using procedures denote the such as Gram-Schmidt orthogonalization. Let support of . There are three kinds of support, , namely, the bounded support in , the half-space support , and the whole space support . Let , be the probability density function (PDF) of the random variable . Then, the joint probability density of is , with support . In this context, the electromagnetic eld response of interest for a given excitation is the random process in . Before we proceed with the discussion of polynomial interpolation in the two-dimensional random space , it is worth noting that the description of the uncertainty or statistical variability in terms of a nite set of random variables is one of the most important steps in the pursuit of a meaningful solution to the boundary

(1) is the number of degrees of freedom in the nite elwhere ement approximation, the vector contains the coefcients in the nite element approximation of the electric eld, the maare in , the matrices are in , trices the vectors are in and is the complex angular frequency. The vector indicates unit excitation at each port. The matrix is dependent on the port characteristics and is used to map the unit excitation to the state space variables. Similarly, the

SUMANT et al.: REDUCED-ORDER MODELS OF FINITE ELEMENT APPROXIMATIONS OF ELECTROMAGNETIC DEVICES

303

value problem of interest. While statistical variability of constitutive material parameters is, in most cases, easily quantied in terms of random variable, variability or uncertainty in geometrical features and boundary conditions often pose challenges in their modeling in terms of random variables. An overview of approaches in use today for tackling such challenges is provided in [14]. The basic idea of polynomial chaos is to approximate a random function in in terms of orthogonal polynomial expansions. Let , be a set of orthogonal polynomials of maximum degree , for interpolation in . The pertinent orthogonality condition is (5) denotes the inner product given by, (6) It is noted that the PDF serves as the integration weight in the orthogonality relation and denes the class of orthogonal polynomials. For example, for a uniformly distributed random variable in , the PDF is a constant and the orthogonality relation denes Legendre polynomials. For the case of a Gaussian distribution, the orthogonality relation denes Hermite polynomials. The examples presented in the next section make use of uniform and Gaussian random variables; thus, Legendre and Hermite polynomials are the two classes of polynomials that will be used in our polynomial chaos expansions. The extension to polynomial interpolation in the two-dimensional space is straightforward. Thus, we dene the polynomial space , with elements products of polynomials in the two random variables (7) where, , with being the number of expansion functions in the polynomial space thus dened. is given by (8) where, denotes the dimension of the polynomial space . In view of the orthogonality of the polynomials, it is (9) denotes the inner product given by, (10) Equipped with the polynomial expansion functions in the random space , the electric eld unknown at a position is approximated as follows (11) Similar expansions are used for the approximation of any quantity inuenced by the uncertainty in the computational domain.

For example, for and for the case of Legendre polynomials, linear expansions are used to approximate the stochastic matrices in (1). For example, the matrix , is approximated by (12) It is noted that the coefcient matrices in the above expression are deterministic. Once they are found, the stochastic matrix is completely dened. The process we follow for the calculation of these coefcient matrices is discussed later in this section. C. Stochastic Model Order Reduction Let the stochastic reduced order system be represented by,

(13) with

(14) Clearly, the development of the stochastic reduced-order model requires the construction of the stochastic projection matrix . However, before we delve into the process used for its construction, let us assume that the matrix has been constructed and is available in terms of a polynomial chaos expansion over the probability space . In a similar manner, assume that polynomial chaos expansions are available for the stochastic matrices of the original, full nite element system. For example, for the case of a linear approximation using Legendre polynomials, these expansions are of the form

(15) For now, we assume that all the coefcient matrices in the above expansions are known. Using these in (14), the matrices of the stochastic reduced-order model are readily obtained. For example, the matrix is obtained as

with similar expansions for the remaining matrices. The R.H.S. in the above expression can be expanded and terms arranged such that they can be cast in the form of orthogonal polynomial expansion. The resulting expressions may be truncated to the desirable degree of polynomial chaos. For instance,

By utilizing a similar linear polynomial chaos for the expansion of the reduced-order model matrices and vectors on the two-

304

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

dimensional probability space as that of the full system, the reduced stochastic matrix is approximated as follows: (16) where the coefcient matrices are given by the following expressions:

ample, the coefcient matrix in the polynomial chaos expansion of is readily computed as follows

(21) In a similar manner, the remaining coefcients, obtained through the expression , are (22) The efcient calculation of these integrals is discussed next. D. Smolyak Algorithm for Numerical Integration in Multiple Dimensions denote one of the integrals in (21), (22). ConsidLet ering the case of one-dimensional integration rst, let us assume, without loss of generality, that the support of each one of the two random variables is the interval . Then may be computed in terms of an appropriate quadrature rule through a summation of the form (23) This integration rule involves the calculation of the integrand at points, , with corresponding weights . Among the various possible choices for the case of a polynomial integration rule, the extrema of the Chebyshev polynomials is a commonly used option. In this case, the sampling points are given by (24) In extending the above integration rule to higher dimensions, the key issue is to ensure the accuracy of the numerical integration while at the same time maintain the efciency of the numerical calculation by tackling the challenge of the exponential growth of the number of sampling points. For the development that follows we adopt the terminology used in [16]. Assume an -dimensional random space and let denote the one-dimensional integration rule in the dimension, , consisting of points on the support interval . Thus, (25) based on nodal sets (26) A straightforward approach for extending the above one-dimensional numerical integration to multiple dimensions is through the tensor product

(17)

(18)

(19) Similar expansions can be written for the rest of the matrices in the reduced-order model. Using a linear polynomial chaos expansion for the output , we have

Thus, combining this with (13) and (16) the stochastic reducedorder system assumes the form of (18). By rearranging the terms and making use of the orthogonality of the polynomials, (18) may be cast in terms of the deterministic system in (19). Clearly, the dimension of this system is . In a compact form, we write

(20) where the correspondence between the names of all matrices and vectors between the two sets of equations is apparent, while the subscript is used to remind us that these matrices and vectors are dened in terms of the coefcients in the polynomial chaos expansions of the stochastic reduced order model. Clearly, the specic form of these matrices depends on both the choice of the polynomials used in the polynomial chaos expansions and the highest polynomial degree kept in the expansions. The ones derived above are for the case of orthogonal Legendre polynomials. At this point we return to the issue of computing the coefcients in the polynomial chaos expansions of the stochastic projection matrix and full nite element matrices (16). Clearly, the orthogonality relation (9) can be used for this purpose. For ex-

SUMANT et al.: REDUCED-ORDER MODELS OF FINITE ELEMENT APPROXIMATIONS OF ELECTROMAGNETIC DEVICES

305

(27) Clearly, the number of terms in the above summation equals the number of nodal points nodal points. If we choose the same number of points in each dimension, i.e., , the total number of points is . The rapid growth of the number of nodes and, consequently, of the computational cost of the integration, when using tensor products is evident. A computationally efcient alternative makes use of sparse grids for multi-dimensional numerical integration. This is discussed next. Sparse Grids: Sparse grids have been used for solving random differential equations by stochastic collocation approach in multi-dimensional random spaces [17]. For our purposes, we will make use of the Smolyak algorithm ([18][20]). Starting with the one-dimensional integration formula (25), the Smolyak algorithm is given by (28)

(29) . To compute where need to evaluate function on the sparse grid we only (30) In this paper, we use the Smolyak formulas based on one-dimensional polynomial integration at the extrema of the Chebyshev polynomials (24). In addition, we dene if and choose and for . This denition makes the one-dimensional nodal sets nested, so . Let be the space of -variate orthonormal polynomials of total degree at most . It can be shown that, if we set , then is exact for integration of polynomials in a space larger than , and that the total number of nodes, Q, for sufciently large is given, approximately, by (31) is much weaker than Clearly, the dependence on dimension tensor product rule. Henceforth in this paper, we will refer to in as the level of the sparse grid integration. As a specic example, Fig. 1 depicts a comparison of two dimensional grids constructed using the tensor product and the Smolyak algorithm for level 4. Clearly, use of the Smolyak algorithm reduces signicantly the number of nodes in the numerical calculation of the integral. With the Smolyak grid dened, the integrals in (21) and (22) for the case of the polynomial chaos expansion of matrix of the original system are computed as follows (32)

Fig. 1. Grids for two-dimensional integration using the extrema of Chebyshev polynomials. (a) Tensor product grid (total number of points 81). (b) Smolyak grid (total number of points 29).

where is the deterministic system matrix computed for the specic values of the input random variables at the th Smolyak grid point, and . The coefcients in the polynomial chaos of all original system matrices are obtained in a similar fashion. The calculation of the polynomial chaos expansion of the stochastic projection matrices proceeds as follows. For each point on the Smolyak grid a deterministic full system is obtained. For this system, using a Krylov subspace method (or any other method of choice) a projection matrix, is obtained. Once these matrices have been constructed, the coefcients in the polynomial chaos expansion of the projection matrix are computed as follows:

(33) This completes our presentation of the mathematical framework for stochastic model order reduction. The proposed methodology may be summarized in terms of the series of steps described in Algorithm 1.

306

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Algorithm 1 Algorithm for Stochastic Model Order Reduction 1: Identify statistically varying material and geometric parameters and represent them in terms of independent random variables . 2: Choose an appropriate order and type of polynomial chaos expansion. 3: Generate a Smolyak grid and corresponding weights . Each point represents a combination of values for different input random variables. 4: For each point on the Smolyak grid, calculate the full nite element system matrices. Perform deterministic MOR for each system to generate the corresponding projection matrix. 5: Using above information, calculate the coefcients in the polynomial chaos expansion of the stochastic full nite element system matrices and the stochastic projection matrix ((32), (33)). 6: Compute the coefcients in the polynomial chaos expansion of the stochastic, reduced matrix (17). 7: Compute the augmented system (20). 8: The augmented system can be used to calculate calculating the mean, standard deviation and other statistics of the system response. III. NUMERICAL STUDIES A. Input Impedance of Terminated Coaxial Cable The rst example considers a structure for which an analytic solution can be obtained, namely, a terminated air-lled coaxial cable. The length of the coaxial cable is m, and its two electrodes are assumed to be perfectly conducting. The radius of the inner, circular cylindrical electrode is mm. The inner radius of the outer circular cylindrical electrode is mm. The cable is terminated at its far end by a resistor. First, we would like to investigate the effectiveness of Smolyak sparse grid for performing numerical integration. We make use of the transmission-line theory formula for the input impedance of the terminated cable, (34) where, , is the cable characteristic impedance, (35) is the propagation constant. We consider two and random inputs. One is the permittivity of the insulating medium in the cable, which is modeled as a uniform random variable in the interval [3.64.4]. The second is the load resistance, which is modeled as a uniform random variable in [2535]. With the frequency set at MHz, using a standard Monte Carlo approach with 102400 random samples, the mean value of the real part of the input impedance is obtained as 16.7386 and its standard deviation as 2.8995 . For sparse grid integration, we

TABLE I PERFORMANCE OF STOCHASTIC COLLOCATION

Fig. 2. Real part of input impedance of a coaxial cable. (a) Mean (real part of impedance). (b) Standard deviation (real part of impedance).

consider different levels starting from level 3. Table I summarizes the results. It is clear that the Sparse grid integration converges to the linear expansion solution level 4 onwards. Hereafter in this paper, we make use of level 4 for sparse grid integration. A full nite element model was developed for the coaxial cable. The nite element system is of dimension 36840. The generated reduced order model is of order 20. 10201 Monte Carlo simulations are performed to generate the reference solution. For the proposed method, a level 4 Smolyak algorithm is used that involves 29 points in the two-dimensional grid. Hence, only 29 deterministic problem solutions are involved in the sparse grid instead of the 10201 used for the standard Monte Carlo. Shown in Fig. 2 is a comparison of the two approaches in the calculation of the real part of the input impedance of the terminated coaxial cable. Very good agreement is observed. In practice it is customary to rely on corner analysis for assessing the impact of variability of the input parameters on the response. To contrast this approach with that of a stochastic modeling, results for the input impedance, obtained using extreme values of the two stochastic input parameters, are generated. More specically, three sets of corner values are considered. A minimum values set, ,a

SUMANT et al.: REDUCED-ORDER MODELS OF FINITE ELEMENT APPROXIMATIONS OF ELECTROMAGNETIC DEVICES

307

Fig. 3. Comparison of real part of input impedance using extreme values for the random parameters with the statistics obtained using the proposed stochastic MOR approach.

Fig. 4. Comparison of calculated mean value of the real part of the input impedance of the wire loop antenna obtained using the proposed stochastic MOR (SC) and standard Monte Carlo (MC).

mean values set, , and a maximum values set, . The calculated real part of the input impedance versus frequency for these three sets is compared in Fig. 3 with the mean response obtained using the proposed algorithm. The error bars indicate one standard deviation. This comparison clearly shows that the statistics of the response generated by the proposed stochastic MOR approach provides a tighter estimate of the system response. It is also worth noting that there is a discrepancy between the response obtained from using mean parameter values from the calculated mean response, a result that is expected for the case where multiple random variables are involved. B. Circular Wire Loop Antenna The next numerical study considers the electromagnetic modeling of a circular wire loop antenna in air. The radius of the circular loop is 1 m. The computational domain used for the nite element model is a cube of side 5 m. A rst-order absorbing boundary condition was imposed on all six walls of the domain. Further details of the nite element model used can be found in [9]. A reduced-order model of order 30 was generated. We are interested in assessing the impact on the transient response of the antenna of statistical variability in geometric parameters such as the wire radius and the loop radius. The wire radius is modeled as a Gaussian random variable with mean of 0.05 m and standard deviation of 0.01 m. The loop radius is modeled as a Gaussian random variable with mean of 1 m and standard deviation of 0.03 m. We rst establish the accuracy of the proposed method by comparing the mean input impedance with that obtained using Monte Carlo. 9801 Monte Carlo simulations are performed to generate the reference solution. A level 4 in the Smolyak algorithm results in sparse grid sampling with 29 points on the two-dimensional probability space. Shown in Fig. 4 is a comparison between the results produced by our approach and standard Monte Carlo for the real part of the mean input impedance. Of particular interest in this problem is the mean fundamental frequency obtained using the two methods. The fundamental frequency obtained using Monte Carlo is 21.3 MHz as against 21.32 MHz obtained using our proposed algorithm.

Next, we consider the transient response of the loop antenna to a pulse-current source excitation. More specically, a rectangular current pulse of rise and fall times of 10 ns, pulse width of 60 ns and amplitude of 1 A, is used to excite the wire loop antenna. Plotted in Fig. 5(a) is the mean input voltage obtained using the generated stochastic reduced-order model. Also plotted in the same gure is the input voltage obtained using a deterministic reduced order model employing the mean values for the two input parameters (wire radius and loop radius). We observe that, as time progresses, the deviation between the two responses is appreciable. This highlights the necessity of using a stochastic reduced order model for computing the mean system response when more than one input variables exhibit statistical variability. Finally, we compare the stochastic system response, plotted in terms of mean voltage with error-bars corresponding to one standard deviation , with responses from corner solutions, obtained using deterministic reduced order models employing minimum and maximum values of the input parameters (Fig. 5(b)). It is clear from Fig. 5(b) that corner simulations can be unreliable and that care should be exercised in their use for the development of error bounds in the system response. IV. CONCLUSION In this paper, we have proposed a method for performing model order reduction of passive electromagnetic devices in the presence of uncertainty or statistical variability in geometric or material parameters. The method makes use of generalized polynomial chaos expansions on the probability space dened by the stochastic variables for the nite element system matrices and the projection matrix used in the development of the reduced-order model. The coefcient matrices in these expansions are obtained from a numerical integration using sparse grids of sampling points on the probability space, generated by means of the Smolyak algorithm. At each sparse grid point a deterministic model order reduction operation is carried out to calculate the corresponding transformation matrix. Thus, the computation complexity of the proposed process is dictated by the number of points on the sparse grid. However, thanks to the mutual independence of the model order reduction operations at the sparse grid points, the parallelism of these calculations can be

308

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

parameters can be unreliable. Furthermore, it was shown that use of corner solutions, utilizing extreme values of the uncertain variables, may lead to gross inaccuracy in the systems response. The second study involved a circular wire loop antenna and it demonstrated that the stochastic reduced-order model can be used to assess the impact of uncertainty of input parameters on the transient response of the antenna. As already indicated in the introduction, stochastic reducedorder modeling offers a computationally expedient means for system-level electromagnetic analysis in the presence of statistical variability in portions of the overall system. This is a topic of current investigation and will be reported in a forthcoming paper. REFERENCES
[1] A. Cangellaris and L. Zhao, Model order reduction techniques for electromagnetic macromodelling based on nite methods, Int. J. Numer. Model.: Electron. Netw., Devices Fields, vol. 13, pp. 181197, 2000. [2] Y. Zhu and A. C. Cangellaris, Finite element-based model order reduction of electromagnetic devices, Int. J. Numer. Model.: Electron. Netw., Devices Fields, vol. 15, pp. 7392, 2002. [3] A. C. Cangellaris, M. Celik, S. Pasha, and L. Zhao, Electromagnetic model order reduction for system-level modeling, IEEE Trans. Microw. Theory Tech., vol. 47, pp. 840850, 1999. [4] H. Wu and A. C. Cangellaris, Krylov model order reduction of nite element models of electromagnetic structures with frequency-dependent material properties, in IEEE MTT-Int, Microwave Symp., 2006, pp. 5255. [5] H. Wu and A. C. Cangellaris, Model-order reduction of nite-element approximations of passive electromagnetic devices including lumped electrical-circuit models, IEEE Trans. Microw. Theory Tech., vol. 52, pp. 23052312, 2004. [6] M. Celik and A. C. Cangellaris, Simulation of multiconductor transmission lines using Krylov subspace order-reduction techniques, IEEE Trans. Comput.-Aided Design Integr. Circuits Syst., vol. 16, pp. 485496, 1997. [7] A. Odabasioglu, M. Celik, and L. Pileggi, PRIMA: Passive reducedorder interconnect macromodeling algorithm, in Proc. IEEE/ACM Int, Conf. on Computer-Aided Design, 1997, pp. 5865. [8] R. W. Freund, Krylov-subspace methods for reduced-order modeling in circuit simulation, J. Comput. Appl.Math., vol. 123, pp. 395421, 2000. [9] H. Wu, Methodologies and algorithms for nite element analysis of electromagnetic devices and systems, Ph.D. dissertation, UIUC, Urbana, 2007. [10] Y. Zhu and A. C. Cangellaris, Multigrid Finite Element Methods for Electromagnetic Field Modeling, ser. IEEE Press Series on Electromagnetic Wave Theory. Hoboken, NJ: Wiley, 2006. [11] L. Kulas and M. Mrozowski, Multilevel model order reduction, IEEE Microw. Wireless Compon. Lett., vol. 14, pp. 165167, Apr. 2004. [12] P. Li, F. Liu, X. Li, L. T. Pileggi, and S. R. Nassif, Modeling interconnect variability using efcient parametric model order reduction, in Proc. Conf. on Design, Automation and Test in Europe (DATE05), 2005, vol. 2. [13] Y.-T. Li, Z. Bai, Y. Su, and X. Zeng, Parameterized model order reduction via a two-directional Arnoldi process, in Proc. IEEE/ACM Int. Conf. on Computer-Aided Design, ICCAD, San Jose, CA, Nov. 2007, pp. 868873. [14] D. Xiu, Fast numerical methods for stochastic computations: A review, Commun. Comput. Phys., vol. 5, no. 24, pp. 242272, Feb. 2009. [15] H. Wu and A. C. Cangellaris, Krylov model order reduction of nite element models of electromagnetic devices with frequency-dependent material parameters, Int. J. Numer. Model., vol. 20, no. 5, pp. 217235, 2007. [16] D. Xiu, Efcient collocational approach for parametric uncertainty analysis, Commun. Comput. Phys., vol. 2, pp. 293309, 2007. [17] D. Xiu and J. S. Hesthaven, Higher order collocation methods for differential equations with random inputs, SIAM J. Sci. Comput., vol. 27, pp. 11181139, 2005.

Fig. 5. Loop antenna transient response to a pulse current source excitation. (a) Comparison of mean stochastic response with the response obtained using mean values of the input parameters. (b) Comparison of stochastic response with corner responses obtained using extreme values of input parameters.

exploited for containing the computational cost. The stochastic, reduced-order model that is generated is in terms of a polynomial expansion in terms of the uncertain parameters in the model. Thanks to the small dimension of these matrices, the reduced-order system lends itself to the expedient computer-aided assessment of the impact of uncertainty on the electromagnetic attributes of the system under study. For example, for the numerical studies presented in this paper, compared to a standard Monte Carlo approach, use of the Smolyak grid and the generated reduced-order model provided for at least two orders of magnitude reduction in the computation of the statistics of the system response. An important feature of the proposed method is that it is independent of the choice of the algorithm used for model order reduction. Thus, even though for our purposes a Krylov subspace projection scheme was utilized, the proposed methodology can be used in conjunction with all other popular reduced-order modeling algorithms. The proposed method was used to calculate the impact of statistical variability in the permittivity of the insulating material and the value of the load resistance on the frequency-dependence of the input impedance of a terminated lossless coaxial cable. Through this example it was shown that, when multiple uncertain parameters are involved, prediction of the mean value response of the system using the mean values of the uncertain

SUMANT et al.: REDUCED-ORDER MODELS OF FINITE ELEMENT APPROXIMATIONS OF ELECTROMAGNETIC DEVICES

309

[18] V. Barthelmann, E. Novak, and K. Ritter, High dimensional polynomial interpolation on sparse grid, Adv. Comput. Math., vol. 12, pp. 273288, 1999. [19] E. Novak and K. Ritter, High dimensional integration of smooth functions over cubes, Numer. Math., vol. 75, pp. 7997, 1996. [20] E. Novak and K. Ritter, Simple cubature formulas with high polynomial exactness, Construct. Approx., vol. 15, pp. 499522, 1999. Prasad S. Sumant received dual B.Tech. and M.Tech. degrees from the Indian Institute of Technology Bombay (IITB), Mumbai, India, in 2004, and both the M.S. degree in electrical and computer engineering and the Ph.D. degree in mechanical engineering from the University of Illinois at Urbana-Champaign, in 2010. He is currently working in the Computational Sciences Function at ExxonMobil Upstream Research Company, Houston, TX. His research interests include scientic computing, computational electromagnetics and numerical methods in geosciences and engineering. Dr. Sumant was elected to the honor society of Phi Kappa Phi in 2007. He is a recipient of the Outstanding Scholar Fellowship from the Department of Mechanical Science and Engineering at UIUC.

Van Valkenburg Professor and Department Head of ECE. His expertise and research interests are in applied and computational electromagnetics and its applications to the advancement of modeling methodologies and CAD tools in support of electrical performance analysis and noise-aware design of integrated electronic systems. His research group has conducted pioneering work in the advancement of modeling methodologies and computer tools for predictive electromagnetic response performance of the signal and the power distribution networks of integrated electronics at the board, package and chip levels. Several of the prototype tools developed by his group have been transferred to the microelectronics industry and have contributed to the development of several EDA signal integrity tools. He has authored and coauthored more than 250 papers in peer-reviewed journals and conference proceedings on the topics of nite methods for electromagnetic eld modeling; multi-conductor transmission line modeling; and signal and power integrity of integrated electronic systems. Prof. Cangellaris is a Fellow of IEEE. In 2000, he received the University of Illinois, ECE Department Outstanding Faculty Teaching Award. In 2005, he received the Alexander von Humboldt Research Award for outstanding contributions to electromagnetic theory. He was the co-founder of the IEEE Topical Meeting on Electrical Performance of Electronic Packaging in 1991. He is currently serving as Editor of the IEEE Press Series on Electromagnetic Wave Theory. He is an active member of the IEEE Microwave Theory & Techniques Society and the IEEE Components Packaging and Manufacturing Technology Society, contributing to numerous professional activities, conferences and symposia organized and sponsored by them. During 20102012 he served as Distinguished Lecturer for the IEEE Microwave Theory & Techniques Society.

Hong Wu received the B.S. degree in computational mathematics from Nanjing University, China, in 2002 and the Ph.D. degree in electrical and computer engineering from the University of Illinois at Urbana-Champaign, in 2007. He is currently working in Extreme DA Corporation, Santa Clara, CA. His research interests include FEMs for electromagnetic modeling of integrated RF and high-speed digital circuits, IC capacitance extraction, timing analysis and multi-corner, multi-mode timing optimization.

Andreas C. Cangellaris (F00) received his Diploma in Electrical Engineering from the Aristotle University of Thessaloniki, Greece in 1981, and the M.S. and Ph.D. degrees in electrical engineering from the University of California, Berkeley, in 1983 and 1985, respectively. After a two-year tenure with the Electronics Department, General Motors Research Laboratories, he joined the Electrical and Computer Engineering (ECE) Department, University of Arizona, rst as Assistant Professor (19871992) and then Associate Professor (19921997). He joined the ECE Department, University of Illinois, Urbana-Champaign, as Full Professor in 1997, where he is currently the M.E.

Narayana Aluru (M95) received the B.E. degree (with honors and distinction) from the Birla Institute of Technology and Science (BITS), Pilani, India, in 1989, the M.S. degree from Rensselaer Polytechnic Institute, Troy, NY, in 1991, and the Ph.D. degree from Stanford University, Stanford, CA, in 1995. He was a Postdoctoral Associate at the Massachusetts Institute of Technology (MIT), Cambridge, from 1995 to 1997. In 1998, he joined the University of Illinois at Urbana-Champaign (UIUC) as an Assistant Professor. He is currently a Professor in the Department of Mechanical Science and Engineering, University of Illinois at UrbanaChampaign. He is also afliated with the Beckman Institute for Advanced Science and Technology, the Department of Electrical and Computer Engineering and the Bioengineering Department at UIUC. Prof. Aluru received the NSF CAREER award in 1999, the NCSA faculty fellowship in 1999 and 2006, the 2001 CMES Distinguished Young Author Award, the Xerox Award for Faculty Research in 2002, the ASME Gustus L. Larson Memorial Award in 2006, the USACM Gallagher Young Investigator Award in 2007 and was named a Willett Faculty Scholar by the College of Engineering at UIUC for the period 20022008. He is a Subject Editor of the IEEE/ASME Journal of Microelectromechanical Systems, served as the Associate Editor of IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS PART II: EXPRESS BRIEFS for 20042005, and currently serves on the Editorial Board of a number of other journals.

310

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Spherical ADI FDTD Method With Application to Propagation in the Earth Ionosphere Cavity
Dominique L. Paul, Member, IEEE, and Chris J. Railton
AbstractThe alternating direction implicit (ADI) nite-difference time-domain (FDTD) technique is implemented using the spherical coordinate system and applied to wave propagation in the lossy Earth ionosphere cavity. The paper shows, for the rst time, that the ADI technique can be developed in conjunction with spherical FDTD in order to overcome the typically long computing times involved in Earth ionosphere cavity problems. Numerical accuracy of the full wave three-dimensional (3D) ADI implementation is validated by comparison with spherical FDTD results. While not free from late time instability, transient responses are shown to be obtained accurately for time steps as large as 15 times the time step derived from cells at the Equator. Hence the spherical ADI is demonstrated to be a promising technique to efciently characterize Schumann resonances and other complex propagation phenomena around the Earth, while overcoming the space cell eccentricity typically generated by spherical grids. Index TermsAlternating direction implicit (ADI) nite-difference time-domain (FDTD), Earth ionosphere cavity, extremely low frequencies (ELF) propagation, Schumann resonances, spherical FDTD.

Fig. 1. Illustration of cell size reduction near the Poles for typical spherical grids. (a) coarse (b) ne.

I. INTRODUCTION HERE has been much research on the characterization of the fundamental resonances generated by intense lightning discharges in the Earth cavity waveguide [1][8], [10][16]. This research was initially prompted by the discovery that shifts in these so called Schumann resonances were sensitive to climate variations [1], exhibiting diurnal and seasonal periodicities [1][4]. It has been shown that Schumann resonances could be used among others as indicators of weather anomalies including in the prediction of extreme weather events [5], Earthquakes [6], global climate modication [1][4] as well as for the remote sensing of the ionosphere D-layer [7] or perturbations in space weather [8]. Although the nite-difference time-domain (FDTD) technique [9] has been developed extensively for Cartesian coordinates, the spherical coordinate system has been employed only in very rare occasions. Full-wave prediction of the rst ve Schumann resonances of the Earth has been achieved using the FDTD technique based on the spherical coordinate system [10][16] but only by very few groups worldwide. Assumption
Manuscript received February 17, 2011; revised May 18, 2011; accepted June 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the Centre for Communications Research, Department of Electrical & Electronic Engineering, University of Bristol, Bristol BS8 1UB, U.K. (e-mail: d.l.paul@bristol.ac.uk). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167940

of rotational symmetry of the Earth was made in [10] and [11] to reduce computing requirements, whereby only one cell was considered in azimuth with periodic boundaries applied at the limits of the domain. More general 3D spherical FDTD models were developed in [12][15]. The most advanced model was elaborated by Jamesina Simpson (one example being [15]) and includes a full representation of continents and oceans. Although not featuring such capability, all other models still enable local perturbations to be included due to their full-wave 3D nature [12][14]. An excellent review of all 2D and 3D models of the Earth can be found in [16]. As pointed out in [15], a drastic reduction of the CourantFriedrichs-Lewy (CFL) time step computed from cells that are almost square at the Equator is generally necessary, due to the shape of the spherical grid which inherently reduces near the , being the Poles, as depicted in Fig. 1. A guideline of smallest cell size at the Equator and the Earth radius, is given in [15] for the reduction of the time step. This can be as low as 0.015 for cells sizes of 100 km only and will be far lower for even ner grids. To overcome this difculty, a progressive merging to larger cells is recommended in [15] so as to maintain the time step dened from cells at the Poles in the whole FDTD domain. Although this proves to be a very effective technique, it requires careful interpolation of eld components at cell edges. As far as the authors are aware, this is the only technique proposed up to now to get over this challenging restriction of the spherical FDTD method. An alternative way to overcome the space cell geometrical eccentricity is to employ the ADI technique, rst presented by Namiki [17] and Zheng-Chen-Zhang [18] in 1999. The ADI technique is unconditionally stable, the time step being limited by accuracy rather than stability. The standard ADI technique has been widely applied to cases which include ne geometrical details in the Cartesian coordinate system [17][20]. It has also been translated to problems with body of revolution symmetry

0018-926X/$26.00 2011 IEEE

PAUL AND RAILTON: SPHERICAL ADI FDTD METHOD WITH APPLICATION TO PROPAGATION IN THE EARTH IONOSPHERE CAVITY

311

using the cylindrical ADI FDTD [21], [22] but to the authors knowledge, has never been attempted on spherical geometries. This paper describes the implementation of the ADI FDTD in the spherical coordinates system with application to lossy Earth system. Although the relations for the iteration mechanism appear to be substantially more involved than for the standard spherical FDTD technique, full description is provided in this contribution to facilitate its future adoption. It can be noted that while the implementation of the ADI FDTD is described with specic reference to Earth resonances in this paper, the method described is equally applicable for other scenarios which require a spherical coordinate system including spherical antenna arrays [23]. Section II introduces the basis of the technique with the derivation of ADI formulae along with the treatment of boundary conditions. Section III presents some numerical results to validate this approach. II. FORMULATION OF SPHERICAL ADI FDTD A. Derivation of ADI Formulae The 3D full wave spherical FDTD constitutes an extension to three dimensions of the formulation in [11] and is equivalent to the full 3D model presented in [12][14]. It will be used as a reference in this contribution. Starting from similar relations, the basic equations for the spherical ADI FDTD can be formulated, each iteration in time being split in two steps as follows. 1) Sub Time-Step1: In this rst half time step, , and are synchronous with , , and respectively

(5)

(6) 2) Sub Time-Step2: In this second half time-step, alternating directions, , and are synchronous with , , and respectively

(7)

(8)

(9)

(1)

(10)

(2)

(11)

(3)

(12) Where the cell edge lengths and cell face areas are dened as in [23], the indices relating here to the center points of either the edges or the faces, hence taken at different positions for E and H eld components (13) (14) (15)

(4)

312

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

(16) (17) (18) The indices for permittivity and conductivity have been omitted for simplicity in relations (1) to (12) and are taken at the center of cell edges for each E eld component. Also, the quantity stands for half a time step. Note that following [11] and [14], the terms to account for the conductivity losses in the Earth atmosphere are introduced in the E eld equations only, whereas the H eld equations, which are derived from Faraday law, are not affected. Also note that in (1) to (12), the FDTD relations are written in a scaled fashion, with the quantities for E and H elds standing in fact for the eld value multiplied by the curved length of its supporting edge. This has the effect of substantially simplifying the ADI formulae. As in the standard ADI method, the relations for the H eld components are substituted into the E eld components relations to generate an implicit system. Hence, after substituting (5) into (1) and rearranging, (1) can be written as (1) with (2')

(23) (24) (25) By substituting (4) into (3) and after rearrangement, simultaneous equations are obtained for the component

(3') with (1') With the quantities in (1) dened as: (27) (19) (20) (21) (22) The simultaneous equations of (1) to solve for all eld can be written in the form of a tridiagonal components matrix which can be solved using the procedure described in [24]. Similarly, by substituting (6) into (2) and rearranging, simultaneous equations are obtained for the component After solving for the three tridiagonal systems, the E eld components are obtained at sub time-step 1. The H eld components are then readily calculated using the explicit equations (4) to (6). Operating in a similar manner to obtain the elds at sub timestep 2 leads to the following ADI relations: After substituting (12) into (7), the relation for writes (28)

(26)

(7')

PAUL AND RAILTON: SPHERICAL ADI FDTD METHOD WITH APPLICATION TO PROPAGATION IN THE EARTH IONOSPHERE CAVITY

313

with

(29) (30) After substitution of (10) into (8) and simplication, the relation for writes

as perfect conductors. As recommended in [25] for stability purposes, these boundary conditions are introduced directly into the tridiagonal matrix at sub time-step 1 by zeroing the coefcients and , and and as well as the right hand sides in (2) and reand (3) for the tangential components spectively. For sub time-step 2, , and and are zeroed for the components , and respectively along with the corresponding right hand sides of (8) and (9). 2) Azimuthal Direction: One extra cell is added to close the domain in azimuth and circular boundary conditions are applied explicitly in the azimuthal direction by matching the eld components at both sub time-steps as (34) (35) Where represents the number of nodes (with extra cell) azimuthally, stands for H eld and tangential E eld components and the indices for and refer to the relevant eld value. 3) Singularities at the Poles: As for the ADI technique in cylindrical coordinates [21], [22], the boundary conditions at the Poles are applied explicitly. We follow the procedure outlined in [11], [14], [15] and [23], whereby the radial E eld component is updated by integrating the H eld component on a contour surrounding the discontinuity at the Pole. For completeness, the relations at the North Pole are reproduced here and write as

(8') with

(31) Similarly, after substitution of (11) into (9), the relation for is derived as

(36)

(9') with with dened as

(37)

(32) (33) After solving the implicit systems for the E elds at sub timestep 2 using (7), (8) and (9), the H eld quantities are derived using explicit relations (10)(12). B. Treatment of Boundary Conditions 1) Radial Direction: Perfect electric wall conditions are applied to limit the FDTD domain in the radial direction as both the ground and the upper edge of the ionosphere are considered

(38) Similar relations are established at the South Pole, with a minus sign in the second part of (36) and (37). III. NUMERICAL VALIDATION A spherical FDTD model of the Earth was generated using a uniform mesh in each direction. Cell sizes of 5 km were considered in the radial direction, and an angular resolution of 3 degrees was considered both in latitude and longitude, with cell sizes of about 330 km at the Equator to ensure a mesh based on 20 cells per wavelength at 50 Hz. A diameter of 6,370 km was assumed for the Earth [12]. The ground was considered as

314

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Comparison between spherical FDTD and spherical ADI FDTD for time step of 0.5 dt with probe at the Equator. (a) whole response and (b) zoom.

Fig. 3. Comparison between spherical FDTD and spherical ADI FDTD for time step of 0.5 dt with probe at 45 degrees in the Northern hemisphere. (a) whole response and (b) zoom.

a perfect conductor and a height of 100 km was taken for the ionosphere, delimiting the FDTD domain in the radial direction between two perfectly conducting walls [11][14]. A representative conductivity prole described as quiet atmosphere in [11] was considered for the atmosphere with conductivity values ranging from 5e-14 S/m at ground level to 7.5e-3 S/m near the upper edge of the ionosphere. Similarly to [15], a hypothetical lightning source was launched 2.5 km above the ground at the Equator, whereby a radial electric eld was excited in the form of a current source together with a raised cosine excitation waveform of width 1.5 ms. Probes were located at an altitude of 7.5 km and at various positions in the Northern hemisphere to record propagation away from the source and enable comparison between standard and ADI spherical FDTD schemes. As the two schemes have been incorporated in the same program for convenience, the relevant update mechanism can be activated by simply setting a ag to 0 or 1 in the input le. 1) Results Obtained Using a Small Time Step: Initially, a time step of 0.5 dt was used, dt being the value of the time step calculated from cells at the Equator and equal to 1.6674 . Fig. 2 shows the radial electric eld recorded three cells away from the excitation source at the Equator. It can be observed from Fig. 2(a) that the transient responses

produced by both schemes are in excellent agreement. Fig. 2(b) shows that even when the scale on the y axis is reduced by a factor of 10 approximately, the results are still in almost perfect superposition. The agreement is also excellent further away from the source as conrmed by Figs. 3 and 4 which show the electric eld sampled at latitudes of 45 degrees and 75 degrees (Arctic Circle) respectively. Although the time domain response changes substantially from one position to another, it is equally well tracked down by both schemes for which results almost superimpose, even when the scale is made smaller by a factor of around 4 as in Figs. 3(b) and 4(b). 2) Results Obtained Using Larger Time Steps: The time step was subsequently made larger than the CFL value for a conductivity prole corresponding to the quiet atmosphere. Fig. 5 shows results obtained at the Equator using the spherical ADI FDTD technique with the time step xed at 1, 5, 10 and 15 times the CFL value, for which the standard spherical FDTD technique is unstable. It can be observed that despite some minor uctuations due to the slightly different amplitudes generated by a current source excitation, the agreement is extremely good as all curves almost superimpose to the one obtained with half the CFL value. Although good accuracy is achieved, it is however to be mentioned that late time instability occurs for these cases, therefore

PAUL AND RAILTON: SPHERICAL ADI FDTD METHOD WITH APPLICATION TO PROPAGATION IN THE EARTH IONOSPHERE CAVITY

315

TABLE I RESONANT FREQUENCIES (IN HERTZ) FOR QUIET ATMOSPHERE CONDUCTIVITY PROFILE

ILLUSTRATION

OF

TABLE II TIME STEP REDUCTION REFINEMENT

AT THE

POLE WITH MESH

Fig. 4. Comparison between spherical FDTD and spherical ADI FDTD for time step of 0.5 dt with probe at 75 degrees in the Artic Circle of Northern hemisphere. (a) whole response and (b) zoom.

Fig. 5. Spherical ADI FDTD transient responses for time steps varying from 0.5 dt to 15 dt with probe at the Equator.

preventing the full transient response of 0.7 s shown in Fig. 2(a) to be obtained. The reason for this is still under investigation. Despite this, the results for the rst four resonant frequencies when the CFL condition is maintained throughout the grid were compared to the results presented in [11] and are listed in Table I. It is to be noted that whereas symmetry in azimuth is exploited in [11], the standard and ADI spherical FDTD schemes

in this study are fully three-dimensional. Very good agreement can be observed among all three techniques as similar values are obtained for all modes, and these compare well with the experimental values listed in [26]. The spherical ADI scheme with CFL value in particular gives very satisfactory results. Table II illustrates the drastic reduction in the time step required in order to ensure stability in standard FDTD as the mesh is rened. In this example, the radial cell size was xed at 5 km and resolutions of 3, 1.5 and 0.75 degrees in latitude and longitude were considered. This corresponds to square cells of length 333.55 km, 166.78 km and 83.39 km at the Equator. In this table, is the CFL time step determined from the cell size at the equator, is the CFL time step determined from the cell size at the Poles and is the time step used in the ADI method and is set at . It can be seen that whereas the time step determined from cells at the Equator, , only suffer minor reductions due to the small cell size along the radius (which is predominant for CFL determination), the time step determined from the smallest cell at the Poles, , is decreased drastically. This results in typical reductions in time step of around 90% in order to avoid early instability by conventional FDTD when using the nest mesh with cell lengths of 83.39 km. The spherical ADI on the other hand, maintains for all runs and this actually corresponds to a time step as large as 17 to 138 times the one derived from cells at the Pole. The potential advantage of the spherical ADI FDTD over the conventional spherical FDTD in terms of computing requirements can be appreciated in Table III for the simulation of a transient response of 0.05 s with for ADI and the largest possible time step for FDTD. All runs were performed on a 3.2 GHz Pentium 4 personal computer equipped with 2 GB RAM. Whereas the ADI scheme typically employs some 25% extra memory for all meshes, the reduction in CPU time is particularly noticeable for rened meshes with a factor of almost 4 for the nest mesh. These gures are consistent with those reported in [21], [22] for the cylindrical ADI FDTD. IV. CONCLUSION This contribution sets out the basis for the development of the ADI FDTD in the spherical coordinate system and full-wave

316

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

COMPARISON

OF

TABLE III CPU REQUIREMENTS FOR CONVENTIONAL SPHERICAL FDTD SCHEMES

AND

ADI

three-dimensional ADI formulae were established for structures with high electrical losses such as in the Earth atmosphere. While not free from late time instability, the method developed herein produced very accurate results for time steps varying from 0.5 to 15 times the CFL value. The memory required by spherical ADI FDTD was typically increased by some 25% over the spherical FDTD due to the more complex iteration mechanism but reductions of up to 75% were observed in CPU time. The rst four Schumann resonances were obtained with good accuracy while maintaining the time step at the equatorial CFL value throughout the spherical grid. Future work will consist in addressing the cause of the late time instability and in establishing a provably unconditionally stable implementation of the technique. Beside the aforementioned advantage for structures where high mesh renement is required, potential future applications for the spherical ADI method include the development of spherical antenna arrays as well as the characterization of highly resonant nanospheres for biosensing devices.

ACKNOWLEDGMENT The authors would like to thank their colleagues in the Communication Systems & Networks Group for helpful discussions and the Centre for Communications Research for provision of facilities.

REFERENCES
[1] E. R. Williams, The Schumann resonanceA global tropical thermometer, Science, vol. 256, pp. 11841184, 1992. [2] E. R. Williams, The global electrical circuit: A review, Atmos. Res., vol. 91, pp. 140152, 2009. [3] C. Price and A. Melnikov, Diurnal, seasonal, and inter-annual variations of the Schumann resonance parameters, J. Atmos. Sol. Terr. Phys., vol. 66, no. 1314, pp. 11791185, 2004. [4] M. Fullekrug and A. C. Fraser-Smith, Global lightning and climate variability inferred from ELF magnetic eld variations, Geophys. Res. Lett., vol. 24, no. 19, pp. 24112414, 1997. [5] C. Price, Lightning sensors for observing, tracking and now casting severe weather, SENSORS, vol. 8, pp. 157170, 2008. [6] M. Hayakawa, K. Ohta, A. P. Nickolaenko, and Y. Ando, Anomalous effect in Schumann resonance phenomena observed in Japan, possibly associated with the Chi-Chi earthquake in Taiwan, Ann. Geophys., vol. 23, pp. 13351346, 2005. [7] J. J. Simpson and A. Taove, ELF radar system proposed for localized D-region ionospheric anomalies, IEEE Geosci. Remote Sensing Lett., vol. 3, no. 4, pp. 400403, 2006. [8] K. Schlegel and M. Fullekrug, Schumann resonance parameter changes during high-energy particle precipitation, J. Geophys. Res., vol. 104, no. A5, pp. 10,11110,118, May 1999.

[9] A. Taove and S. C. Hagness, Computational Electrodynamics: The Finite-Difference Time-Domain Method, 3rd ed. Norwood, MA: Artech House, 2005. [10] T. Otsuyama, D. Sakuma, and M. Hayakawa, FDTD analysis of ELF wave propagation and Schumann resonances for a subionospheric waveguide model, Radio Sci., vol. 38, no. 6, pp. 11031103, 2003. [11] A. Soriano, E. A. Navarro, D. L. Paul, J. A. Porti, J. A. Morente, and I. J. Craddock, Finite difference time domain simulation of the Earth-ionosphere resonant cavity: Schumann resonances, IEEE Trans. Antennas Propag., vol. 53, no. 4, pp. 15351541, Apr. 2005. [12] H. Yang and V. P. Pasko, Three-dimensional nite difference time domain modeling of the Earth-ionosphere cavity resonances, Geophys. Res. Lett., vol. 32, no. L03114, 2005. [13] H. Yang and V. P. Pasko, Three-dimensional nite difference time domain modeling of the diurnal and seasonal variations in Schumann resonance parameters, Radio Sci., vol. 41, no. RS2S14, 2006. [14] E. A. Navarro, A. Soriano, J. A. Morente, and J. A. Porti, Numerical analysis of ionosphere disturbances and Schumann mode splitting in the Earth-ionosphere cavity, J. Geophys. Res., vol. 113, no. A09301, pp. 114, 2008. [15] J. J. Simpson and A. Taove, Three-Dimensional FDTD modelling of impulsive ELF propagation about the earth-sphere, IEEE Trans. Antennas Propag., vol. 52, no. 2, pp. 443451, Feb. 2004. [16] J. J. Simpson, Current and future applications of 3-D global Earthionosphere models based on the full-vector Maxwells equations FDTD method, Surveys Geophys., vol. 30, no. 105130, pp. 105130, 2009. [17] T. Namiki, A new FDTD algorithm based on alternating-direction implicit method, IEEE Trans. Microw. Theory Tech., vol. 47, no. 10, pp. 20032007, Oct. 1999. [18] F. Zheng, Z. Chen, and J. Zhang, A nite-difference time-domain method without the Courant stability conditions, IEEE Microw. Guided Wave Lett., vol. 9, no. 11, pp. 441443, Nov. 1999. [19] C. C.-P. Chen, T.-W. Lee, N. Murugesan, and S. C. Hagness, Generalized FDTD-ADI: An unconditionally stable full-wave Maxwells equations solver for VLSI, in Proc. IEEE/ACM Int. Conf. on Computer Aided Design (ICCAD-2000), 2000, pp. 156163. [20] Y. Yang, R. S. Chen, Z. B. Ye, and Z. B. Wang, Analysis of planar antennas using unconditionally stable three-dimensional ADI-FDTD method, in Proc. IEEE Int. Symp. Antennas and Propag., 2005, pp. 170173. [21] C. Yuan and Z. Chen, A three-dimensional unconditionally stable ADI-FDTD method in the cylindrical coordinate system, IEEE Trans. Microw. Theory Tech., vol. 50, no. 10, pp. 24012405, Oct. 2002. [22] H.-L. Chen, B. Chen, Y. Yi, and D.-G. Fang, Unconditionally stable ADI-BOR-FDTD algorithm for the analysis of rotationally symmetric geometries, IEEE Microw. Wireless Compon. Lett., vol. 17, no. 4, pp. 304306, Apr. 2007. [23] O. Franek, G. F. Pedersen, and J. B. Andersen, Numerical modeling of a spherical array of monopoles using FDTD method, IEEE Trans. Antennas Propag., vol. 54, no. 7, pp. 19521963, Jul. 2006. [24] G. D. Smith, Numerical Solution of Partial Differential Equations: Finite Difference Methods, 3rd ed. Oxford, U.K.: Oxford Univ. Press, 1985, pp. 2428. [25] J. Chen and J. Wang, PEC condition implementation for the ADIFDTD method, Microw. Opt. Techn. Lett., vol. 49, no. 3, pp. 526530, Mar. 2007. [26] M. Balser and C. A. Wagner, Observations of Earth-ionosphere cavity resonances, Nature, vol. 188, pp. 638641, Nov. 1960.

Dominique L. Paul (M90) received the D.E.A. degree in electronics from Brest University, Brest, France, in June 1986 and the Ph.D. degree from Ecole Nationale Superieure des Telecommunications de Bretagne (LEST-ENSTBr), Brest, in January 1990. From 1990 to 1994, she was a Research Associate at the Centre for Communications Research, University of Bristol, Bristol, U.K. During 19951996, she worked as a Research Associate at the Escuela Tecnica Superior de Ingenieros de Telecomunicacion of Madrid, Spain under a grant from the Spanish Government. Since 1997, she has been a Research Fellow in the Centre for Communications Research, University of Bristol, with a permanent position from 2003. Her research interests include

PAUL AND RAILTON: SPHERICAL ADI FDTD METHOD WITH APPLICATION TO PROPAGATION IN THE EARTH IONOSPHERE CAVITY

317

the electromagnetic modelling of passive devices such as microwave heating systems, dielectric structures at millimeter wavelengths, MIMO systems, low prole antennas, conformal antenna arrays and textile antennas for wearable applications. Dr. Paul is a member of the IET, IEEE Microwave Theory and Techniques Society and IEEE Antennas and Propagation Society.

circuits. He currently works in the Centre for Communications Research, University of Bristol, Bristol, U.K. where he leads the Computational Electromagnetics group which is engaged in the development of new algorithms for electromagnetic analysis and their application to a wide variety of situations including planar and conformal antennas, microwave and RF heating systems, radar and microwave imaging, EMC, high speed interconnects and the design of photonic components.

Chris J. Railton received the B.Sc. degree in physics with electronics from the University of London, London, U.K., in 1974 and the Ph.D. degree in electronic engineering from the University of Bath, Bath, U.K., in 1988. During the period 19741984, he worked in the scientic civil service on a number or research and development projects in the areas of communications, signal processing and EMC. Between 1984 and 1987, he worked at the University of Bath on the mathematical modelling of boxed microstrip

318

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Analysis of Directional Logging Tools in Anisotropic and Multieccentric Cylindrically-Layered Earth Formations
Guo-Sheng Liu, Fernando L. Teixeira, and Guo-Ji Zhang
AbstractWe develop a pseudoanalytical method for the analysis of directional resistivity well-logging tools consisting of multiple tilted-coil antennas in cylindrically-layered Earth formations which have anisotropic conductivities and where each layer may be eccentric to the others (multieccentric formations). The cylindrically-layered, anisotropic, and multieccentric scenarios considered here are often encountered in deviated/horizontal drilling for hydrocarbon exploration, where the logging tool may be eccentric to the borehole due to mechanical vibrations and/or gravitational pull, and where eccentric invaded zones (of mud ltrate into the formation) may also arise because of gravitational pull on the mud ltrate. Index TermsAnisotropic media, geophysical exploration, logging-while-drilling (LWD).

I. INTRODUCTION

IRECTIONAL well-logging resistivity tools are routinely used in hydrocarbon exploration to determine, for example, horizontal and vertical resistivities in anisotropic Earth formations and relative dip angle during deviated drilling [1][5]. For logging-while-drilling (LWD) tools, tilted-coil antennas provide directional data for real-time geosteering. Resistivity tools typically consist of multiple tilted-coil antennas wrapped around a cylindrical metallic mandrel. A transmitter coil excites eddy currents in the surrouding Earth formation [6]. This current is proportional to the formation conductivity and produces a secondary electric eld and induced voltage in the receivers. Various techniques have been proposed to model this problem. These include brute-force numerical methods such as nite-differences (FD) [7][10] and nite elements (FE) [11][13], and pseudoanalytical approaches [4], [5], [14][19]. The former can handle arbitrarily complex Earth formations

Manuscript received July 21, 2010; revised February 20, 2011; accepted May 30, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the OSC under Grants PAS0061 and PAS-0110 and in part by the Chinese Scholarship Council under Grant 2008615058. G.-S. Liu is with the Department of Information Management Engineering, Guangdong University of Technology, Guangzhou 510520, China (e-mail: armstrongliu@yahoo.cn). F. L. Teixeira is with the ElectroScience Laboratory and Department of Electrical and Computer Engineering, The Ohio State University, Columbus, OH 43212 USA (e-mail: teixeira@ece.osu.edu). G.-J. Zhang is with the Department of Mathematics, South China University of Technology (SCUT), Guangzhou 510641, China (e-mail: magjzh@scut.edu. cn). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167907

but suffer from high computational costs. The latter are less exible but less computationally costly. Many pseudoanalytical approaches approximate coil antennas as magnetic dipoles, which ignores mandrel effects and nite-size antenna effects. Pseudoanalytical approaches that do include these two effects [4], [5] have attracted much interest in recent years as they provide better accuracy. Lovell and Chew [14] studied the response of a centered dipole source in a cylindrically-layered formation, and later extended it to horizontal eccentric coil antenna sources [15]. Hagiwara et al. [4] extended this approach to arbitrarily-tilted coils, and Hue et al. [5] to eccentric and arbitrarily-tilted coils. All these references have assumed concentric isotropic cylindrical layers in the Earth formation, i.e., the only eccentricity, if any, is that of the logging tool itself. In a number of practical scenarios, it is of interest to study multieccentric formations where each successive cylindrical layer may have a different axis of symmetry, as illustrated in Fig. 5. This geometry may arise due to (1) mechanical vibrations and/or gravitational pull causing the logging tool to become eccentric with respect to the borehole and (2) the presence of eccentric invasion zones of mud ltrate (borehole uid) in the formation. The eccentricity of the latter may be a consequence of gravitational pull during deviated or horizontal drilling.1 The analysis of multieccentric formations is also useful to model, as a limit case, the response of logging tools near lateral beds, see Fig. 8. Finally, consideration of anisotropy is important because Earth formations often have different conductivities along horizontal and vertical directions [6], [9]. Indeed, this consideration is one of the original motivations for employing tilted-coil antennas [2], [20], [21]. In this work, we extend the pseudoanalytical approach of [5] to model anisotropic and multieccentric Earth formations. The formalism employs cylindrical eigenfunctions [22], [23] in anisotropic media to expand the eld in each cylindrical layer. Generalized reection and transmission matrices for multiple cylindrical layers are derived in a recursive fashion from singleinterface ones. The addition theorem [22] for cylindrical waves is applied in a nested fashion to model multieccentric layers. II. CYLINDRICAL EIGENFUNCTIONS IN UNIAXIAL MEDIA In this Section, we write down the eld solutions for Maxwells equations in terms of cylindrical eigenfunctions
1In practice, ltrate invasions are expected to deviate from a circular shape, but an eccentric circular layer serves as a good rst approximation of the slumping effect of gravity.

0018-926X/$26.00 2011 IEEE

LIU et al.: ANALYSIS OF DIRECTIONAL LOGGING TOOLS

319

by extending the basic approach outlined in [5] to uniaxial anisotropic media.2 Assuming a time-harmonic dependence of , we write Maxwells equations as [22] the form

(1) where and are the electric and magnetic elds respectively, is the causative electric current density, is the permeability, and is the complex permittivity tensor. We consider a conductivity tensor expressed in cylindrical coordinates with , and where as is the conductivity in transverse plane (denoted by ), so that . We next employ a Fourier transformation (2) and decompose the problem into axial and transverse components as

or being the Bessel function of rst with kind and the Hankel function of rst kind, respectively. From physical considerations, the square root for is chosen such [24]. The amplitude coefcients and that are generic 2 1 column vectors to be determined from the boundary conditions. Since only Hankel functions of the rst kind are used in what follows, we will drop the superscript in for notational simplicity. The transverse eld components can be obtained by substituting (7) into (4). In the case of logging tools with coil antennas, the -component of the eld does not contribute to the voltage at receiver coil since the tangent vector to the coil(s) does not have a radial component. The -components can be written as (8) where

and similarly for the other elds, so that Maxwells equations are rewritten as where the prime denotes derivative with respect to the argument. III. TILTED-COIL SOURCES IN ANISOTROPIC, MULTIECCENTRIC LAYERED MEDIA Fig. 1 depicts a tilted coil in a multilayered cylindrical media. In the case of directional logging-tools, this geometry includes a , the borehole , and the Earth metallic mandrel layers). To conform to the practical formation ( scenarios of interest, we assume the rst two layers isotropic, may and the remaining possibly anisotropic. The layer refer to an invaded zone of mud ltrate (borehole uid) into the adjacent Earth formation. The boundary between the th layer and th layer is denoted by . The coordinates of the transmitter coil antenna are given by , where is the radial coordinate of is the vertical coordinate of the center of the coil, the coil, is the elevation tilt angle, and is the azimuthal direction of the tilt [5]. The coordinates of receiver antennas are dened similarly. A prototypical directional logging tool would have one transmitter coil antenna and two receiver coil antennas. The current density on the transmitter antenna is given by3 (9) where is the current source amplitude and is the tangent vector to the tilted-coil antenna plane given by . During drilling, the transmitter coil is excited by an sinusoidal waveform, and phase difference and amplitude ratio between the voltages excited at the receiver coils are measured. The voltage
3Note that ~ ~ a is not a unit vector, and that . Even though the magnitude of the current density ~ varies with , the ux of it over any (innitesimal) cross section (and hence the total current owing along the coil) is constant and equal to I .

(3) Using and , and after a few manipulations, we have in a source-free region (4) and a dual equation for , where , and [24]. Substituting (4) into (3) and using the expansion (5) and similarly for tial equations , we obtain the following ordinary differen-

(6) Equation (6) can be recognized as different versions of Bessel equation, which admit general solutions of the form (7) where

r1J = 0

2Note

that the notation adopted here differs from that in [5] in some places.

320

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

with similar expressions for and , where and are generalized reection and transmission matrices5, as deand can be rived in Appendix A, and the coefcients obtained from the eld produced by the (normalized) current source (9) in a homogeneous medium (incident eld), as detailed in Appendix B (see also [5]). In multilayered media, and in expressions such as (12) vary from layer to layer. To simplify the notation, we drop layer indices in and except when necessary to distinguish them. A. Multieccentric Anisotropic Layers Consider two eccentric cylindrical layers as shown by the solid line circles in Fig. 2. The centers of the inner and outer and nonlayers coincide with the origins of primed coordinate systems, respectively. The nonprimed primed on the coordinates can be treated as a displacement of primed coordinate. Consider the elds at a certain point between the two circles. Using (2) and (5), the spectral components and can be expressed as

Fig. 1. Geometry of a tilted-coil antenna in a cylindrically multilayered formation, with tilt angle ( ;  ). The coil antenna is located within the i = 0 layer (borehole). The rst layer i = 1 represents to the metallic mandrel. The n vector ~ represents the normal unit vector to the plane of the coil antenna. Additional layers i = 1; 2; . . . can be included to model complex Earth formations including invaded zones.

(13) in the primed coordinate, or

at a receiver coil for a unit current source at the transmitter (transimpedance) can be written as [5]

(14) in the nonprimed coordinate. Applying Grafs addition theorem for cylindrical waves [22], we have

(10) where (15) whereby the elds in nonprimed coordinate are expressed in terms of the eigenfunctions in the primed coordinate. This leads to the following transformation rules between eld coefcients in the two coordinate systems

(11) , and . Details on the derivation of the above expression are found in [5], [25]4. The analysis in [5] was limited to isotropic media and one single eccentric layer. In the following, we lift these two restrictions. , for We can express the solutions in the borehole layer computing the transimpedance in (10), as and

(16)

We next introduce a ctitious boundary, represented by the dashed circle in Fig. 2 and concentric to the inner layer. Note that the ctitious boundary is chosen in a different way than in (12)
4The 5The terminology reection and transmission is used in a loose sense here. More precisely, these coefcient matrices refer to the ratio between standing wave and outgoing wave amplitudes in each of the layers i, and vice-versa.

integral in [5] is a folded (less general) version of the above.

LIU et al.: ANALYSIS OF DIRECTIONAL LOGGING TOOLS

321

where

is the Kronecker delta, a convenient choice is so that

(21) In this fashion, a geometry of two eccentric layers has been transformed into a concentric geometry by introducing a ctitious boundary. In the case of multiple eccentric layers, the procedure above is simply iterated, starting from the outermost layer and moving towards the inner layers. Suppose outer layer and inner layer are eccentric to each other, with nonprimed and primed coordinates centered on and , respectively. In this case, the equivalent relations to (17) and (18) are (22)
Fig. 2. Cross-section view of an eccentric borehole geometry in cylindrical coordinates. The nonprimed coordinate system (; ) is concentric to the outer circular boundary and has a displacement of (d, ) with respect to the primed coordinate system ( , ), which is concentric to the inner circular boundary. The ctitious circular boundary shown by a dashed circle is also concentric to the primed coordinate system.

(23)

[5], [15]. This choice facilitates the treatment of multiple eccentric layers to be considered later in this Section and the derivation of an iterative solution for the eld coefcients as detailed in Appendix B. Some further numerical considerations are discussed in Appendix C. Different azimuthal modes do couple when an outgoing wave from the inner layer is reected by the ctitious boundary (due to the effect of the outer eccentric boundary). The equivalent relating and is dened by reection matrix (17) at the outer(most) On the other hand, the reection matrix boundary can be written independently for each mode in the nonprimed representation, as before, through (18) Substituting (17) and (18) into (16) leads to

where is the known reection matrix from layer , simin (18). Note that, as opposed to (18), the matrix in ilarly to (23) in general couples azimuthal modes because of the possible presence of eccentric layers external to layer . Following the addition theorem procedure delineated above, the reection layer from a ctitious boundary, within layer , that is external and concentric to layer can now as be obtained from

(24) In this equation, and refer to the radial and angular relative displacement between the centers of layers and , and is is computed from (24), the wavenumber of layer . Once can be determined using a relation equivalent to (31) of Appendix A, i.e.,

(25) to layer The same procedure can then repeated from layer and so on. The procedure ends at the borehole layer where the receiver antennas are located, to obtain . Note that there is no need here to effect the analogous pro(which in this case would proceed cedure for obtaining outwardly from the innermost layer to the outer layers) because in practice we are only interested in computing the elds in the borehole layer, and there is only one layer internal to it (the ), see Fig. 1. metallic mandrel, with

(19)

Equation (19) can be solved for [26]

since

is arbitrary. Since

(20)

322

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

The nal expression for the elds in the borehole layer that generalizes (12) in the presence of modal coupling caused by the eccentric layers is of the form

TABLE I TILTED MAGNETIC DIPOLES IN ECCENTRIC BOREHOLE

(26) where is the reection matrix at the mandrel given by (27)

IV. NUMERICAL RESULTS Unless indicated otherwise, the logging tool geometry considered here consists of three tilted-coil antennas with 11.43 cm radius each, wrapped around a metallic mandrel with 10.16 cm in radius. One coil antenna functions as transmitter and the other two as receivers. The midpoints of the two receiver coils are located at 60.96 cm and 76.2 cm away from the midpoint transmiter coil along the -direction shown in Fig. 1. The phase difference (PD) and amplitude ratio (AR) between the voltages measured at two receivers coils are the parameters of interest to estimate the formation conductivities. We assume the relative permeability and permittivity of all media to be unity. A. Magnetic Dipole Model We rst validate the proposed method against a magnetic dipole model as considered in [17], [22]. In this case the mandrel is absent. In our model, we set the radius of coil antennas to 0.5 cm so that the elds produced by the tilted coils are well approximated by magnetic dipoles. The radius of the borehole is 10.16 cm. All the coils are eccentric, with a displacement of 7.62 cm from the center of borehole along the same direcand tion. The transmitter and receiver coils are tilted by , respectively. The two receiver coils are positioned 30.46 cm and 43.18 cm away from transmitter coil, respectively. The borehole is lled by an isotropic uid (mud ltrate) with 100 m resistivity. The surrounding Earth formation consists of an anisotropic medium with horizontal and vertical resistivities equal to 0.1 m and 4 m, respectively. The transmitter coil is excited by a 12 kHz alternating current with 1 A. Table I shows the voltages at the two receivers computed by three different approaches: an analytical method, as described in [17], [21], a numerical result based on the nite element method (FEM), and and denote the the proposed pseudoanalytical method. real and imaginary part of the voltages, respectively. The relative errors are below 0.25%. Assuming the analytical method as benchmark, the pseudoanalytical results are in a slight better agreement.

Fig. 3. Directional tool response (phase difference between receiver voltages) in an anisotropic formation versus anisotropy ratio: Higher conductivity case.

B. Tilted-Coil Antennas With Anisotropic Formation In this subsection, the effect of the anisotropy on the response of directional tools is investigated. The radius of the mandrel, coil antennas, and borehole are 10.16 cm, 11.43 cm, and 12.7 cm respectively, all concentric. The vertical distances of the receiver coils to the transmitter coil are 60.96 cm and 76.2 cm. The boreS/m. The tilt hole is lled with an oil-based mud having angle of the 2 MHz transmitter coil is xed at , while four different tilt angles are considered for the two receivers: . Fig. 3 shows the PD as a function of the ratio of vertical and horizontal components of the formation conductivity (the horizontal conductivity is xed at S/m, and the vertical conductivity is varied). In practice, the vertical conductivity is typically lower than the horizontal one. It , the tool has maximum sencan be seen that when sitivity to the anisotropy. Fig. 4 shows a similar result now for S/m. Note the a formation with lower conductivity, different scales used in Figs. 3 and 4, indicating that the tool is more sensitive to the anisotropy ratio for higher conductivities. C. Eccentric Tool in a Formation With Eccentric Invaded Zone We next consider a multieccentric problem with cross-section as depicted in Fig. 5. In this case, the mandrel is eccentric to the borehole, and the rst formation layer is eccentric to both mandrel and borehole. This rst layer may constitute an invaded

LIU et al.: ANALYSIS OF DIRECTIONAL LOGGING TOOLS

323

Fig. 4. Directional tool response (phase difference between receiver voltages) in an anisotropic formation versus anisotropy ratio: Lower conductivity case.

Fig. 6. Logging tool response (phase difference and amplitude ration between the two receiver voltages) in a multieccentric environment with invaded zone as depicted in Fig. 5. The eccentric offset of the tool with respect to the borehole is j cm. The parameter represent the eccentric xed and equal to j offset of the invaded zone with respect to the borehole, j j, which is varied between 0 and 10 cm. The parameter represents the conductivity 2 S/m (borehole of the invaded zone, which is varied between conductivity) and S/m (formation conductivity).

OO

= 5:08

 = 10

d= OO   = 5 10

Fig. 5. Multieccentric environment with eccentric tool and eccentric invaded zone. , and refer to the centers of tool mandrel, borehole, and invaded zone respectively.

O ;O

zone caused by diffusion of the borehole uid into formation, , and having mixed electrical properties. The points and indicate the centers of mandrel, borehole, and invaded zones, respectively. The corresponding radii are cm, cm, and cm. The radii of coil antennas are equal to 11.43 cm. The tilted angles of transmitter and 45 respectively, all of them and receiver coils are tilted along the -direction. The operation frequency is 2 MHz. The borehole is lled with oil-based mud having

S/m. The conductivity of formation is 10.0 S/m. We x the mandrel eccentricity to 5.08 cm in the -direction in Fig. 5, i.e., in and . The magnitude of the eccentric offset of invaded zone varies, but is also in the -direction. Fig. 6 plots the results of PD and AR for different conductivities in the invaded zone, S/m, and for different offsets cm. It can be seen from Fig. 6 that the phase difference is less sensitive to the eccentric offset with an invaded zone having conductivity very close to either the borehole mud or earth formation conductivities. Fig. 6 (top) shows that the PD sensitivity to offset S/m. In addition, Fig. 6 (bottom) shows peaks at about that the AR varies more strongly with the invaded zone conductivity as the eccentric offset is increased. Fig. 7 is an inversion plot showing (PD, AR) traces as the eccentric offset and invaded zone conductivity are varied. Note that the traces converge at the S/m, as the conductivity of the invasion becomes point equal to that of the external Earth formation.

324

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 7. Plot of eccentric offset d between invaded zone and borehole in cm and invaded zone conductivity  in S/m as function of the phase difference and amplitude ratio between the receiver voltages.

D. Directional Logging Example Directional LWD tools can provide azimuthal information about the surrounding formation by rotating the tilt azimuth along the drilling axis. As is varied to cover a angle 360 angular span, the measured well-data can be used to determine the next drill direction for real-time geosteering purposes. Directional tools provide capability of detecting nearby beds given sufcient resistivity contrast [21]. In this subsection, we evaluate the azimuthal sensitivity of a directional LWD tool response when approaching a conductive bed boundary. We use an example similar to the one considered in [21]. We assume a conductive bed located in the proximity of the logging tool along the 135 direction, as shown in Fig. 8. The bed is assumed to have an approximately at boundary, which can be approximated by a circle of sufciently large radius as illustrated in Fig. 8. While drilling occurs, the driller can be made to rotate 360 on its axis, thus providing variation on . We assume the elevation tilt anof the transmitter and receivers to be xed, as specied gles below. We consider a one-transmitter, two-receiver LWD tool, with the receiver coils arranged 60.96 cm and 76.2 cm away from the transmitter coil. The bed boundary is located 68.6 cm cm in Fig. 8. away from the borehole boundary, i.e., (with denoting the We use a series of eccentric circles eccentric offset in cm) to approximate the bed boundary. The other parameters of this problem are as follows. The conductivities of borehole, intervening formation, and conductive bed are S/m, 0.5 S/m, and S/m, respectively. The radii of mandrel, coil antennas, and borehole are 10.16 cm, 11.43 cm, and 12.7 cm, respectively, and the tool is concentric to the borehole. The elevation tilt angles of transmitter, bottom receiver, 0 , and 45 , respectively, and the opand top receiver are erating frequency is 2 MHz. Fig. 9 shows the PD response for and . The results clearly converge as the eccendifferent tric offset of the approximating circle is increased. The PD , precisely when the response reaches an extrema at

Fig. 8. Solid circles represent the mandrel (gray area), coil antennas (in blue color), and borehole boundary respectively, with radii of 10.16 cm, 11.43 cm, and 12.7 cm, respectively. The mandrel and antennas rotate anticlockwise while drilling. A conductive bed with at boundary is located at a distance of 68.6 cm to the borehole boundary along the 135 direction. The dashed eccentric circles are used to approximate the bed boundary (for sufciently large radius).

Fig. 9. Azimuthal response (phase difference) of directional LWD tool, where d denotes the eccentric offset in cm. The tool rotates throughout a period of which is 360 . The response converges for large d, and peaks at  angle corresponding to (the nearest point of) the bed boundary.

= 135

azimuth tilt angle points in the direction of (the nearest point of) the bed boundary. V. CONCLUSION A pseudoanalytical formulation to analyze resistivity logging tools developed in [25] was extended to anisotropic and multieccentric cylindrically-layered Earth formations, where sucessive cylindrical layers may have different axes of symmetry. A recursive approach based on a successive application of Grafs addition theorem from the outermost layer to the borehole layer is developed to account for multiple eccentric layers external to the

LIU et al.: ANALYSIS OF DIRECTIONAL LOGGING TOOLS

325

borehole and to yield generalized reection matrices associated to a ctitious boundary concentric to the logging tool. This allows for an efcient iterative approach to be used for computing the elds due to a tilted-coil source within the borehole layer. Various numerical examples are shown to validate the method and to illustrate its applicability for hydrocarbon exploration, in particular in the modeling of formation resistivity log responses and in the detection of lateral beds. APPENDIX A GENERALIZED REFLECTION AND TRANSMISSION MATRICES Here, we extend the methodology in [22] to compute local reection and transmission coefcients between anisotropic cylindrical layers. Assuming that a standing wave is incident from onto the boundary , continuity of transverse medium eld components gives

Fig. 10. Boundary conditions in terms of reection and transmission coefcients for (a) standing and (b) outgoing waves, respectively. (a) A standing wave onto boundary R , an outgoing wave is reected is incident from layer i and a standing wave is transmitted into layer i. (b) An outgoing wave is incident from layer i onto boundary R , a standing wave is reected and an outgoing wave is transmitted into layer i .

+1

+1

with , and Similarly, we rewrite the azimuthal components as .

(28) (33) see Fig. 10(a). Similarly, assuming that an outgoing wave is incident from medium onto the boundary , the continuity of transverse eld components gives where

(34) The normalized matrices are better conditioned than the original ones. The elds in the borehole are expressed as (29)

see Fig. 10(b). In (28) and (29), and are the 2 2 local reection matrices at the boundary , while and are the 2 2 local transmission matrices at . Following [5], [14] recursive relations to obtain the matrices and in (12) are given by (30) (31) and are computed from the Note that the matrices innermost to the outermost layer and vice-versa, respectively. In conductive layers, the coefcient matrices of (28) and (29) include exponentially large as well as exponentially small terms. This can cause numerical instabilities for large arguments. It is benecial to normalize the solutions as

(35)

where

(36)

APPENDIX B ITERATIVE SOLUTION Generalized reection matrices in multieccentric media can be obtained through a recursive application of Grafs addition theorem. After the innermost ctitious boundary is introduced, the nal geometry becomes equivalent to that of a problem consisting of mandrel, coil antennas, and one ctitious outer

(32)

326

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

boundary (all concentric). Similar to [5], we obtain in the borehole layer

(37) where drel [4], [5], and is the reection matrix of the steel manare 2 1 column vectors given by [5]

(38) where stands for with superscript or with superscript. in (37) involves a matrix inversion, which Direct solve for can become ill-conditioned for large , and/or . We can instead write an iterative scheme (39) for . The rst term in the right-hand side corresponds the initial wave excited by the transmitter coil (hois the (inmogeneous solution). The second term coming) reected wave produced by (reection of the initial outgoing wave from) the external boundary. Iteration (39) starts ; then is solution of the wave traveling across with borehole once (without mandrel effect); is solution the wave traveling across borehole twice (reected once by the mandrel); is the solution of the wave traveling in the borehole times. An adaptive stop criterion can be set up for this iterative process. The above xed-point iterative procedure is convergent because and correspond to reection cothe elements of efcients with magnitude less than one. A detalied discussion on convergence criteria of iterative procedures such as above is presented [28] from a more general standpoint.

for to the asymptotic estimate denotes the truncating point for in the large . Here, numerical integration and we assume a PEC mandrel [5]. The above estimate agrees with prior numerical investigations [25]. For truncating Grafs series, we note that the argument of the , whose largest value occurs for large , Bessel functions is . Since Bessel functions decrease rapidly for when increasing order and xed argument when the absolute value of the order becomes larger than the absolute value of the argu. Using ment, the series can be truncated at order the above estimate and some typical values for desired accuracy cm, cm, and geometrical parameters, say cm, we have . This number of terms was found to be sufcient in most of the examples considered here. ACKNOWLEDGMENT The authors would like to thank Dr. B. Donderici from Halliburton Energy Services for providing validation data. REFERENCES
[1] M. Sato, J. Fuziwara, M. Miyairi, K. Kashihara, and H. Nitsuma, Directional induction logging methods, in Proc. SPWLA 35th Annu. Logging Symp., Jun. 1922, 1994, pp. 116. [2] M. Bittar, Electromagnetic Wave Resistivity Tool Having a Tilted Antenna for Determining the Horizontal and Vertical Resistivities and Relative Dip Angle in Anisotropic Earth Formations, U.S. Patent 6,163,155, Dec. 19, 2000. [3] T. Hagiwara and H. Song, Directional Resistivity Measurements for Azimuthal Proximity Detection of Bed Boundaries, U.S. Patent 6,181,138, Jan. 30, 2001. [4] T. Hagiwara, E. J. Banning, R. M. Ostermeier, and S. M. Haugland, Effects of mandrel, borehole, and invasion for tilt-coil antennas, presented at the SPE 78th Annu. Tech. Conf. Exhibition, Denver, CO, Oct. 58, 2003, paper SPE 84254. [5] Y. K. Hue and F. L. Teixeira, Analysis of tilted-coil eccentric borehole antennas in cylindrical multilayered formations for well-logging applications, IEEE Trans. Antennas Propagat., vol. 54, no. 4, pp. 10581064, Apr. 2006. [6] J. R. Wait, Geo-Electromagnetism. New York: Academic, 1982. [7] T. Wang and S. Fang, 3-D electromagnetic anisotropy modeling using nite differences, Geophys., vol. 66, no. 5, pp. 13861398, 2001. [8] Y. K. Hue, F. L. Teixeira, L. S. Martin, and M. S. Bittar, Three-dimensional simulation of eccentric LWD tool response in boreholes through dipping formations, IEEE Trans. Geosci. Remote Sens., vol. 43, no. 2, pp. 257268, Feb. 2005. [9] H. O. Lee and F. L. Teixeira, Cylindrical FDTD analysis of LWD tools through anisotropic dipping-layered earth media, IEEE Trans. Geosci. Remote Sens., vol. 45, no. 2, pp. 383388, Feb. 2007. [10] D. Wu, J. Chen, and C. R. Liu, An efcient FDTD method for axially symmetric LWD environments, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 6, pp. 652656, Jun. 2008. [11] D. Pardo, L. Demkowicz, C. Torres-Verdn, and M. Paszynski, Two-dimensional high-accuracy simulation of logging-while-drilling (LWD) measurements using a self-adaptive goal-oriented hp nite element method, SIAM J. Appl. Math., vol. 66, no. 6, pp. 20852106, 2006. [12] D. Pardo, L. Demkowicz, C. Torres-Verdn, and C. Michler, PML enhanced with a self-adaptive goal-oriented hp nite-element method: Simulation of through-casing borehole resistivity measurements, SIAM J. Sci. Comput., vol. 30, no. 6, pp. 29482964, 2008. [13] M. S. Novo, L. C. Silva, and F. L. Teixeira, Finite volume modeling of borehole electromagnetic logging in 3-D anisotropic formations using coupled scalar-vector potentials, IEEE Antennas Wireless Propagat. Lett., vol. 6, pp. 549552, 2007. [14] J. R. Lovell and W. C. Chew, Response of a point source in a multicylindrically layered medium, IEEE Trans. Geosci. Remote Sens., vol. GE-25, no. 6, pp. 850858, Nov. 1987. [15] J. R. Lovell and W. C. Chew, Effect of tool eccentricity on some electrical well-logging tools, IEEE Trans. Geosci. Remote Sens., vol. 28, no. 1, pp. 127136, Jan. 1990.

APPENDIX C FURTHER NUMERICAL CONSIDERATIONS The integral in (10) can be accelerated by subtracting the homogeneous solution [5]. A possible way to evaluate the is to estimate the relative error introduced by truncating ratio where refers to the homogeneous solution. Combining the second term for in (26) with the rst term in (39), we have = . For the nonhomogeneous contribution to be sufciently small, the last two terms in (39) and the rst term ) should approach zero. For typical tool in (26) (take geometries, the outgoing wave term in (26) dominates those . Hence, the error can be estimated from (39) for large by the ratio of outgoing and incoming wave contributions and using (27), we have in (26). Assuming , which leads

LIU et al.: ANALYSIS OF DIRECTIONAL LOGGING TOOLS

327

[16] H. Wang, P. So, S. Yang, W. J. R. Hoefer, and H. Du, Numerical modeling of multicomponent induction well-logging tools in the cylindrically stratied anisotropic media, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 4, pp. 11341147, Apr. 2008. [17] L. Zhong, J. Li, A. Bhardwaj, L. C. Shen, and R. C. Liu, Computation of triaxial induction logging tools in layered anisotropic dipping formations, IEEE Trans. Geosci. Remote Sens., vol. 46, no. 4, pp. 11481163, Apr. 2008. [18] S. Ebihara and Y. Inoue, Analysis of eccentered dipole antenna for borehole radar, IEEE Trans. Geosci. Remote Sens., vol. 47, no. 4, pp. 10731088, Apr. 2009. [19] G. L. Wang, C. Torres-Verdin, and S. Gianzero, Fast simulation of triaxial borehole induction measurements acquired in axially symmetrical and transversely isotropic media, Geophysics, vol. 74, no. 6, pp. e233e249, Nov.Dec. 2009. [20] M. S. Bittar, Electromagnetic Wave Resistivity Tool Having a Tilted Antenna for Determining the Horizontal and Vertical Resistivities and Relative Dip Angle in Anisotropic Earth Formations, U.S. Patent 7,557,579 B2, Jul. 7, 2009. [21] L. Chou, Q. Li, A. Darquin, J.-M. Denichou, R. Grifths, N. Hart, A. McInally, G. Templeton, D. Omeragic, I. Tribe, K. Watson, and M. Wiig, Steering toward enhanced production, Oileld Rev., vol. 17, no. 3, pp. 5463, Oct. 2005. [22] W. C. Chew, Waves and Fields in Inhomogeneous Media. Piscataway, NJ: IEEE, 1995. [23] L. W. Pearson, A construction of the elds radiated by z-directed point sources of current in the presence of a cylindrically layered obstacle, Radio Sci., vol. 21, no. 4, pp. 559569, Jul.Aug. 1986. [24] W. C. Chew, The singularities of a Fourier-type integral in a multicylindrical layer problem, IEEE Trans. Antennas Propagat., vol. AP-31, no. 4, pp. 653655, Jul. 1983. [25] Y. K. Hue, Analysis of Electromagnetic Well-Logging Tools, Ph.D. dissertation, The Ohio State Univ., Columbus, 2006. [26] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, ser. Applied Mathematics Series55. Gaithersburg, MD: NIST, 1972, p. 363, 10th printing. [27] N. J. Vilenkin, Special Functions and the Theory of Group Representations. Providence, RI: AMS, 1968. [28] M. DUrso, I. Catapano, L. Crocco, and T. Isernia, Effective solution of 3-D scattering problems via series expansions: Applicability and a new hybrid scheme, IEEE Trans. Geosci. Remote Sens., vol. 45, no. 3, pp. 639648, Apr. 2007.

Guo-Sheng Liu was born in Hubei Province, China, in 1983. He received the B.S. degree in science and the Ph.D. degree in engineering, both from the South China University of Technology (SCUT), Guangzhou, China, in 2005 and 2010, respectively. From 2008 to 2010, he was a visiting scholar with the ElectroScience Laboratory, The Ohio State University, Columbus. Since 2011, he has been as an Assistant Professor at the Department of Information Management Engineering, the Guangdong University of Technology, Guangzhou, China. His current research interests include nite-difference time-domain (FDTD) methods and numerical modeling of well-logging tools.

Fernando L. Teixeira received the Ph.D. degree in electrical engineering from the University of Illinois at Urbana-Champaign, Urbana, in 1999. From 1999 to 2000, he was a Postdoctoral Associate with the Massachusetts Institute of Technology. In 2000, he joined the The Ohio State University, where he is now a Professor with the Department of Electrical and Computer Engineering and also afliated with the ElectroScience Laboratory. His current research interests include modeling of electromagnetic sensors for hydrocarbon exploration, timedomain computational electromagnetics, metamaterials, and ultrawideband inverse scattering. Dr. Teixeira received a number of awards for his research, including the CAREER Award from the National Science Foundation, the triennial Booker Fellowship from the International Union of Radio Science (USNC/URSI), and the Outstanding Young Engineer Award from the IEEE Microwave Society (MTT-S). He has served as chair of the Joint IEEE AP/MTT-S Columbus Chapter and currently serves as an Associate Editor for the IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS.

Guo-Ji Zhang received the B.S. degree in computer science and applications and the Ph.D. degree in circuit and systems from the South China University of Technology (SCUT), Guangzhou, China, in 1977 and 1999, respectively. Since 1977, he has done research work at SCUT, where he is currently a Professor with the Department of Mathematics. His research interests include computational electromagnetics and computational intelligence.

328

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Analytic Transient Analysis of Radiation From Ellipsoidal Reector Antennas for Impulse-Radiating Antennas Applications
Shih-Chung Tuan, Member, IEEE, Hsi-Tseng Chou, Senior Member, IEEE, Kung-Yu Lu, and Hsi-Hsir Chou
AbstractA time-domain (TD) analysis of ellipsoidal reector antennas is presented to predict the transient scattered elds when it is illuminated by a cosine-tapered and transient-step feed radiation whose phase center is located at the nearby focus. A transient-step function of the feeds radiation is considered because it will produce an impulsive response near the second focus of the reector as interested in the applications of impulse-radiating antennas. This TD solution is in a closed form and remains valid both near and far from the reector, and can be used via the convolution theorem to efciently obtain the early time transient elds generated by the same ellipsoidal reector antenna when it is illuminated by a realistic nite-energy pulse which emanates as a spherical wave from the focus. Numerical results are presented for the transient elds both near and far from the reector. Index TermsAperture integration, ellipsoidal reector antennas, near-eld focus and impulse-radiating antenna, time domain, transient analysis.

I. INTRODUCTION N this paper we present an early-time transient analysis of electromagnetic (EM) elds scattered from a perfectly conducting ellipsoidal reector when it is illuminated by a spherical wave of a feeds radiation with a cosine-tapered pattern and a transient time-step response. This work is motivated by the needs of transient EM analysis in the booming applications of ultra-wide-band or short-pulse antenna systems for target identication and remote sensing such as in the design of impulse-radiating antennas (IRA) [1][6]. An ellipsoidal reector antenna is considered here because it has been widely used as the sub-reector of Gregorian dual-reector antenna system. Recently it has also attracted great attention in the applications of near-eld focused radiations for RFID [7], the detections of vital signal and underground objects [8][10], or noncontact microwave inspection system [9], [10] because of its nature to focus the scattering eld in the near zone of antenna. Furthermore, an ellip-

Manuscript received November 01, 2010; revised May 25, 2011; accepted July 21, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. S.-C. Tuan is with the Oriental Institute of Technology, Department of Communication Engineering, Taipei, Taiwan (e-mail: t1200475@ms32.hinet.net). H.-T. Chou is with the Yuan Ze University, Department of Communications Engineering, Chung-Li, Taiwan (e-mail: hchou@saturn.yzu.edu.tw). K.-Y. Lu is with the National Taiwan University, Graduate Institute of Communication Engineering, Taipei, Taiwan (e-mail: fo012@mail.oit.edu.tw). H.-H. Chou is with the Yuan Ze University, Communication Research Center, Chung-Li, Taiwan, 320 (e-mail: hsi-hsir.chou@trinity.cantab.net). Digital Object Identier 10.1109/TAP.2011.2167919

soidal reector may reduce to a either spherical or parabolic reector with a proper parameter assignment. Thus, the developed solution has an advantage of exhibiting eld phenomena when the surface changes from a spherical to a parabolic reector. Past works on the transient analysis of reector antennas have been focused on the parabolic types. In addition to conventional time-domain (TD) numerical analysis techniques such as nite-difference time-domain (FDTD) and TD integral approaches, which are computationally inefcient for the analysis of large antennas, most efforts have been focused on developing quasi-analytical TD solutions [1][25] that appear in closed forms and thus provide physical interpretations of wave behaviors. Typical solutions using TD uniform geometrical theory of diffraction (TD-UTD) [11][13] and physical theory of diffraction (TD-PTD) [14] appear shortcomings in the analysis of reector antennas because of the existence of singularity in the vicinity of ray caustics. TD physical optics (TD-PO) approximation [15][17] and TD aperture integration (TD-AI) technique [18] that were obtained by either a direct inverse Laplace transform [15], [16], [18] or an analytical time transform (ATT) [17] of corresponding frequency-domain (FD) formulations [26] were mainly used to develop many analytic and closed form solutions for the transient analysis of parabolic reectors. In particular, Baum et al. [1], [2] have presented a closed-form TD analysis for predicting the IRA elds on the boresite but far from the reector. They had also extended their analysis in [3] to provide a closed-form TD solution for the IRA elds in the two principal planes; also for observation points that are far from the reector. Recently, analytic TD solutions valid for arbitrary observation points off the reector surface had been developed in [18] for a Huygens source illuminations using TD-AI. In [5], a closed form solution of boresite radiation was developed for a defocused parabolic reector antenna while [4] computed the radiation from a substitution of the parabolic reector by an exciting V-antenna imaged from the reector. The analytic transient solution of scattering from a reector illuminated by a plane wave and Gaussian beams were developed in [16], [17] using TD-PO technique. TD-AI is employed in this paper to develop the transient analytic solution for an ellipsoidal reector antenna. The development rst denes an aperture in the front of reector that has a uniform phase distribution (or a constant time delay) of equivalent currents determined by using an approximation of Geometrical Optics (GO) ray tracing techniques [26]. This uniform phase aperture occurs on the surface of a sphere centered at one of the two focuses of ellipsoidal reector. The TD elds

0018-926X/$26.00 2011 IEEE

TUAN et al.: ANALYTIC TRANSIENT ANALYSIS OF RADIATION FROM ELLIPSOIDAL REFLECTOR ANTENNAS

329

radiated from these electric and magnetic currents, respectively are found by using the radiation integral, which are then inverted into TD to achieve a closed form solution in the TD-AI approach. The solution developed in this paper appears to be more general and useful than previously developed solutions because the ellipsoidal reector may resemble a class of reectors shapes, and, for example, it may reduce to either a spherical or a parabolic reector by properly adjusting the parameters of an ellipse. Mostly importantly, the developed solution explicitly and theoretically exhibits a transient time-impulsive response [1][3] at the second focus of reector when it is illuminated by a transient time-step wavefront as interested in IRA applications. Furthermore, the wave phenomena interpretations in terms of diffraction mechanism have been achieved, where the total scattering eld is decomposed into contributions from reection and edge diffractions. This paper is organized as follows. Section II describes the characteristics of ellipsoidal reector antenna, its feed modeling and the FD radiation integral of equivalent currents. Development of transient analytic solution, as well as its characteristics, is presented in Section III. Numerical investigation on the TD characteristics of antenna is shown in Section IV. Concluding remarks are discussed in Section V. Detailed derivations of TD formulations are shown in the Appendix for a reference. II. ELLIPSOIDAL REFLECTOR AND ITS FD FIELDS Fig. 1 shows the structure of a symmetric ellipsoidal reector that is illuminated by the radiation of a -polarized feed at . The ellipsoidal reector is formed by taking a part of a rotationally symmetric ellipse described by (1) where and are the radii of its two principal axes with as illustrated in Fig. 1 (i.e., it is truncated at ). The ellipse has two focuses (referred as and , respectively, hereafter) located on the z-axis at and 0, where was selected as the origin of a global coordinate system for a convenience of computation. In the following analysis, is located at for a focus feeding. It is worthy to note that when , and consequently , then the ellipse reduces to a sphere, where the two focuses coincide at the center of sphere. On the other hand, when , this ellipsoidal reector will become a parabolic reector. According to the denition of an ellipse, the distances between and via the lines connecting to an arbitrary point, , on the ellipse surface are a given constant, , i.e. , which assures the scattered elds to have an in-phase eld superposition at when the reector is illuminated by the radiation of a feed located at . A. Modeling of Feeds Radiation The radiation of feed is assumed to exhibit a transient step function spherical wave in a time domain, and can be expressed with a cosine-taper as (2a)

Fig. 1. A rotationally symmetric ellipse and a near-eld focused reector which is taken from a part of ellipse.

Fig. 2. Equivalent aperture and denitions of related coordinate systems. and are spherical surfaces centered at O and , respectively. is formed (the solid line part of the circular by the normal projection of reector on . and . (a) Denition of related cycle) is formed by the intersection of coordinate system; (b) illustration of integration path.

(2b) is the Heaviside step function, is the speed of light where in free space with being the wave impedance, and [18] (3) denotes the initial value of feeds excitation, and is dened in the feeds spherical coordinate system with and as illustrated in Figs. 1 and 2(a). Note that for an arbitrary position vector, , on the reector surface, as shown in Fig. 1. In (3), B. Equivalent Sources in Aperture Integration (AI) Techniques The elds scattered from the ellipsoidal reector when it is illuminated by in (2) can be approximated in terms of the elds radiated from a set of equivalent currents, ,

330

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

that are dened by the EM elds, , over an aperture, , selected in the vicinity of reector surface as illustrated in Fig. 2(a). In this case, a good selection of considers the projection of reector surface onto the surface, , of a sphere centered at with a radius as shown on Fig. 2(b). The aperture eld at may be approximated adequately via GO [18], [26] by

(4) where are the spherical coordinates dened over the aperture sphere as illustrated in Fig. 2(a). It is noted that the selection of a spherical aperture gives an advantage of resulting in aperture elds in (4) with a constant delay time. In this case, the unit vector normal to the aperture surface is , which is further used to dene the equivalent currents by (5) C. Scattering Field in FD via AI Using (5), the scattered electric eld can be expressed as [26]
Fig. 3. Interpretation of and at which starts and ceases to touch . The dashed lines illustrate the existence of , which is a sphere centered at O. is a part of by the projection of reector on . (a) Reection point is off . (b) Reection point is on .

(6) where with being the location of observation measured from . In (6), is the Dirac delta function. Substituting (4) and (5) into (6) gives

(7a) where (7b) III. ANALYTIC DEVELOPMENT FOR SCATTERED FIELDS In this section we investigate the TD characteristics of scattering elds. A. Characteristics of Constant Time-Delay (or Equal-Phase in the Frequency Domain) Contours If (i.e., not at ), in (7) may be used to reduce the surface integration into a line one at a condition of zero value in its parameter, which occurs at (8) This equation describes a geodesic contoured path on , from which the elds radiated from a feed and scattered from the reector to arrive the eld point, , will have an equal propaga-

tion distance and cause a constant time delay. Based on the illustration in Fig. 2(b), (8) describes a circular contour, , formed by the intersection of and , where is a spherical surface centered at the eld point, Q, with a radius, . It is straightforward to show that the center, , of exists on the straight line going through and Q. This straight line is actually a part of the propagation path of feeds radiation to arrive Q via a reection from the ellipsoidal surface. The intersection of this line and is referred to a reection point, , on , which is uniquely determined by every reection propagation path. It is also noted that the location of , as well as , changes with respect to , and moves along this line as increases. Since exists only on , (7a) exists a non-zero value only when has an overlap with . Let and representing the time period that starts and ceases to have an overlap with . It is understandable that (7a) vanishes at or . Fig. 3(a) and (b) show two situations where is off and on , respectively. In Fig. 3(a), and are the time that starts and ceases to tough the edge of . In Fig. 3(b), is the time that toughs at while is the time that ceases to tough the edge of . The mathematical expression for can be simplied in a coordinate system which is dened by where are the spherical coordinates of . In this new coordinate system, the position vector at Q is expressed as (0, 0, ) with . In this case, can be described by (9)

TUAN et al.: ANALYTIC TRANSIENT ANALYSIS OF RADIATION FROM ELLIPSOIDAL REFLECTOR ANTENNAS

331

It is noted that this contour exists only when

, i.e., (10a)

or (10b) B. Development of Analytic Solution Equation (7) can be simplied if one addresses its integration variables, , in coordinate system by , and considers the following variable transformation: (11) It can be shown that , and where both and in all equations afterward are now functions of time because both and given in (9) are function of time, and are given by and . In (15) have been evaluated in closed-form solutions, and are given in (A23)(A28). Equation (16) is a transition function dened in [16]. It is noted that (A23)(A28) appear singular upon an observation over their forms when observer is located on axis. The cancellation of singularities is not apparent in these formulations. Their solutions for eld points on axis have been evaluated in Appendix A.4 and B. C. Transient Characteristics of Scattering Fields (13a) with (13b) represents s variation on as illustrated in Fig. 4 where with and representing the starting and ending angles that overlaps with . Based on the derivations in Appendix A and B, (7a) can be solved in a closed-form as Based on the observations on (14)(16), the characteristics of scattering elds are further investigated. 1) Reected Field Component, : The reected eld component is characterized using GO concepts which result in a polarization unit vector by (17) The selection of sign in (17) is same as the sign of larized component can be found by (14) where . The co-po-

Fig. 4. Illustration of variation.

that is formed by

on

in terms of angular

(12) Thus if gives (Q is away from ), substituting (12) into (7a)

(18) The cross-polarized component, which is caused by the depolarization effect of reecting surface, can be found by (15) and (16) (19) It is noted that the reected eld due to the contribution from the reection point, , can be obtained by setting (or

332

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

COEFFICIENTS,

AND

TABLE I , IN (A18). COEFFICIENTS NOT LISTED ARE EQUAL TO 0

in Appendix A), which is found by using (A31a)(A33a) and given by (20) This is identical to that obtained by GO. The effects of surface curvatures can be found by (21) 2) Field Distribution as the Reection Point Located Right , on the Edge: In this case, (14) remains valid except at where , and and are undetermined since it is only a point on . In this case, one may assume a straight edge with , which makes . 3) Field Distribution at Ray Caustic (i.e., ): At , and will not overlap at any time except at . In this case they are identical. Thus and (7) becomes and

(25d) (25e)

(25f) (22) which is identical to (23) A closed form solution of (23) is found in (A36). This solution explicitly shows a transient impulsive eld behavior at when the reector is illuminated by a transient time-step incident eld. 4) A Special Case of Spherical Reector: In this case, and , and (A16) becomes (24) . The solution can be reduced from One may choose (14) directly. Without going through the details, the following functions can be obtained (25a) and (26b) One may simply replace the parameters in Table I by and zeros for other parameters. In this case, . Also it can be shown that (27a) (25b) (25c) and (27b) where the parameters, and are dened in Tables I and II. Also in (25), , and are dened in (A12b), (A12b) and (A15b), respectively. The functions, and are dened in (A21a) and (A21c), respectively. 5) A Special Case of Parabolic Reector [18]: For a parabolic reector, one considers where is the focal length of reector and . In this case, . Thus in (A9) and makes in (A8). In this case, the coordinates of observer are measured from and denoted by . As illustrated in Fig. 5, it can be shown that (26a)

TUAN et al.: ANALYTIC TRANSIENT ANALYSIS OF RADIATION FROM ELLIPSOIDAL REFLECTOR ANTENNAS

333

Fig. 5. Geometrical interpretation of parameters for a parabolic reector.

where (27c) In (A20), and vanish, where and the rest of

Fig. 6. Transient scattering elds along z-axis, where is measured from . The time delays were shifted to have a common time reference at the center of each pulse.

(28) The solution in (14) reduces to

(29) where , and

Fig. 7. Transient scattering eld with respect to varying observation angles.

(30a)

(30b) and (30c) In the above equations, (A17c), respectively. and are given in (A17b) and

IV. NUMERICAL EXAMINATIONS Numerical examples are presented to demonstrate the characteristics of scattering elds. One rst examines the behavior of in (13a) which governs the transient behavior of scattering elds. In this case, the parameters of ellipsoidal reector are , , and , which makes the diameter of reector by 3.48 m. Fig. 6 shows the variation of

transient scattering elds, which is normalized by the spreading factor, , as a function of measured from along axis. In Fig. 6, the time delay of each pulse is shifted in order to compare the pulse width. The pulse width is equal to where and are the rising and ending time of the nite pulse. By the denition of an impulse, as , tends to become impulsive. As shown in Fig. 6, as , i.e., approaching to . It explicitly demonstrates that the scattering eld has a time-impulse response at . It is noted that is corresponding to a ray caustic in the FD asymptotic evaluation. Secondly one considers the transient behavior of the scattering elds with respect to the angles of observation as shown in Fig. 7. In this case, the reector has a diameter of 2 m (i.e., ), and . It is observed that along the z-axis, where the reection dominates, the scattering eld exhibits a narrower pulse. The scattering elds at the observation angles away from z-axis are contributed from the edge diffractions as previously demonstrated in [16], [18]. One next considers the transient behaviors of scattering elds for the special cases of spherical and parabolic reectors. In these cases, all reectors have a same size with a diameter of 2 m for the front planar and circular aperture. The observers are on z-axis with 8.9 m away from the reector aperture. Also the

334

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 8. Total transient scattering elds of spherical and parabolic reectors at an observer of 8.9 m away for their aperture. All reectors have a same size of 2 m in the diameter of reectors edge ring for the planar and circular aperture.

spherical reector has a radius of 5 m while the focal length of parabolic reector is 3 m. The results are compared with that of an ellipsoidal reector shown in Fig. 7. Fig. 8 shows the total elds scattered from spherical and parabolic reectors. It is observed that the magnitudes of both cases are similar. The case of spherical reector exhibits a narrower pulse width (0.16 ns) than that of parabolic reector (0.166 ns). In comparison with the results shown in Fig. 7, the ellipsoidal reector exhibits a largest value in magnitude and narrowest pulse width. The reason is caused by a fact that the focal points of spherical and parabolic reectors are at the center of sphere and innite, respectively. In this current example, the distances to the focal points are 4, 1994 and 1.06 m, respectively for spherical, parabolic and ellipsoidal reectors (we assume for a parabolic reector in Section III-C.5). One next examines the characteristics of reected elds as discussed in Section III-C.1 in terms of cross-polarization and curvature effects as shown in Fig. 9(a)(c), where the GO elds, cross-polarization and curvature effects are shown, respectively. The cross-polarization and curvature effects are produced by the de-polarization effects of surface curvature. Fig. 9(a) shows the total transient scattering elds and the GO components. In this case, only the eld distributions within the pulse duration were shown for the GO components since GO elds exhibit a transient-step eld response according to TD-UTD in [11]. In this case, the GO components will exhibit an uniform and rectangular pulse response. On the other hand, the total elds exhibit tapered eld response as time changes, which is caused by the existence of surface curvature. Fig. 9(b) shows the cross-polarized components of the scattering elds, which are also caused by the surface curvature. In comparison with the total eld, the cross-polarization is sufciently small even though the current solution exhibits this component. Also Fig. 9(c) shows the curvature effects, which were obtained from the subtraction of co-polarized components from the GO elds. It is observed that the ellipsoidal reector exhibits a higher level of curvature effects while the spherical reector exhibits least curvature effects. In comparison with the level of cross-polarization, the curvature effects are apparently more signicant.

Fig. 9. Characteristics of Reected elds in terms of de-polarization and curvature effects. (a) Total & GO elds; (b) cross-polarization; (c) curvature effect.

V. CONCLUSIONS In this paper we presented a quasi-analytic TD solution to perform a transient analysis of scattering from an ellipsoidal reector antenna when it is illuminated by the radiation of a feed with a cosine-taper pattern. The characteristics of ellipsoidal reector with two focuses have made this solution very useful because this type of reector can be used for the applications of either sub-reector in a dual reector antenna system or near-eld focus antennas. It may also reduce to either a spherical or a parabolic reector, and make the developed solutions useful for a class of reector antenna analysis. In particular, we have explicitly shown a transient time-impulsive response at the

TUAN et al.: ANALYTIC TRANSIENT ANALYSIS OF RADIATION FROM ELLIPSOIDAL REFLECTOR ANTENNAS

335

second focus of the reector when it is illuminated by a transient time-step wavefront, which was not mathematically derived. When the reector is reduced to be a parabolic reector, the solution explicitly and rigorously validates the concept of impulsive radiating antennas as shown in [18]. Future works include the transient analysis of dual-reector antennas using dual PO or GTD-PO techniques. APPENDIX DERIVATIONS OF CLOSED-FORM ANALYTIC SOLUTION FOR (18)

1) Evaluation of form by substituting

in (A5) can be found in a closed-

(A7) into (7b), and readdressing (7b) in the following (A8) where

A. Field Point is Not at in Consider a position vector on , which can be expressed and , respectively, as (A9) Using the relations shown in Figs. 1 and 2(a), the following equations can be obtained: (A10) where the upper and lower signs are used for respectively. Using a fact that term in (A5) can be expressed as and , , the common

(A1) The vector cross-products in (13) can be expressed as

It is noted that is denoted by purpose of brevity. It is noted that value independent of , and

(A2) in the following for the in (A2) is a real

(A3) (A11) Thus (13) becomes Also using (A1) results in the following relationships: (A12a) with (A12b) (A12c) (A4) and (A13) The integrals in (A4) involve with the following integrals: with (A5) (A14a)

With these denitions, (A4) becomes

(A14b)

(A15a) and (A6) (A15b)

336

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

It is noted that both and in (A13b) are larger than zero. Substituting (A12) into (A11) gives the following expression:

each term in (A19) can be integrated in a closed-form solution by using the following formulation:

(A16) (A21a) where (A17a) (A17b) (A17c) and (A17d) Using (A12)(A13), (A9) can also be expressed as a power seby ries of (A21d) because in (A15b) and (A17d). It is noted that the denition domain in (A21d) is and . In this case, in (A5) can be expressed as (A22) (A18) or , and the coefcients, and , are where shown in Table I. Combining (A16) and (A18) and applying factorization give the integrands of (A5) by where with (A23) and where (A21c) (A21b)

(A24) : Following the derivations in Section 2) Evaluation of A.2, can be expressed in a form similar to (A22) by replacing the subscript 1 to c where (A25) and

(A19) where the coefcients, are given by

(A20) and can be found from (A20) by replacing with . In this case, because in Table I. It is noted that

(A26)

TUAN et al.: ANALYTIC TRANSIENT ANALYSIS OF RADIATION FROM ELLIPSOIDAL REFLECTOR ANTENNAS

337

COEFFICIENTS,

AND

TABLE II , IN (A30). COEFFICIENTS NOT LISTED ARE EQUAL TO 0

3) Evaluation of : Following the derivations in Section A.2, can be expressed in a form similar to (A22) by replacing the subscript 1 to s

(A32a)

(A27) and (A32b) (A33a)

(A33b)

(A28) where since in Table I. 4) Special Cases at : A special case occurs as where in (A7a). In this case, (A11) becomes

B. Field Point at The integration in (23) can be found in a closed-form solution by using and . The integration over can be found by following an identical procedure in section A.4 of this appendix since is located at z-axis. Thus,

(A29) and (A14) becomes

(A34) is the maximum span angle of the aperture meawhere sured from . One may simply use the results in (A31) with and , and reduce (A34) to

(A30) where and , are shown in Table II. Substituting (A21) and (A22) into (A3a) gives the following solution: (A35) where and are given in (A29) and Table II, respectively. The integration in (A35) is found in a closed form given by

(A31a)

(A31b)

(A36)

338

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

where

(A37a) (A37b) (A37c) (A37d) and

(A37e)

REFERENCES
[1] C. E. Baum, E. G. Farr, and D. V. Giri, Review of impulse-radiating antennas, in Review Radio Sciece. Oxford, U.K.: Oxford Univ. Press, 1999, pp. 403439. [2] S. P. Skulkin and V. I. Turchin, Transient elds of parabolic reector antennas, in Ultra-Wideband Short Pulse Electromagnetics, C. E. Baum, L. Carin, and A. P. Stone, Eds. New York: Plenum, 1997, vol. 3, pp. 8187. [3] E. G. Farr, C. E. Baum, W. D. Prather, and L. H. Bowen, Multifunction impulse radiating antennas: Theory and experiment, in UltaWideband, Short-Pulse Electromagnetics 4. New York: Kluwer Academic/Plenum Publishers, 1999. [4] O. V. Mikheev et al., New method for calculating pulse radiation from an antenna with a reector, IEEE Trans. Electromag. Compat., vol. 37, pp. 4854, Feb. 1997. [5] R. de Oliveira and M. Helier, Closed-form expressions of the axial prompt responses of a defocused parabolic reector antenna, IEEE Trans. Antennas Propag., vol. 57, no. 11, p. 3685, Nov. 2009. [6] J. S. Tyo, M. Dogan, J. H. Boddeker, and C. J. Buchenauer, Increasing the focused response of impulse radiating antennas through polarization control of aperture elds, IEEE Trans. Antennas Propag., vol. 54, no. 2, pp. 586594, Feb. 2009. [7] A. Buf, A. A. Serra, P. Nepa, H.-T. Chou, and G. Manara, A focused planar microstrip array for 2.4 GHz RFID readers, IEEE Trans. Antennas Propag., vol. 58, no. 5, pp. 15361544, May 2010. [8] M. Bogosanovic and A. G. Williamson, Antenna array with beam focused in near-eld zone, Electron. Lett., vol. 39, no. 9, pp. 704705, May 2003. [9] M. Bogosanovic and A. G. Williamson, Microstrip antenna array with a beam focused in the near-eld zone for application in noncontact microwave industrial inspection, IEEE Trans. Instrum. Meas., vol. 56, no. 6, pp. 21862195, Dec. 2007. [10] K. D. Stephan, J. B. Mead, D. M. Pozar, L. Wang, and J. A. Pearce, A near eld focused microstrip array for a radiometric temperature sensor, IEEE Trans. Antennas Propag., vol. 55, no. 4, pp. 11991203, Apr. 2007. [11] P. R. Rousseau and P. H. Pathak, Time-domain uniform geometrical theory of diffraction for a curved wedge, IEEE Trans. Antennas Propag., vol. 43, pp. 13751382, Dec. 1995. [12] P. R. Rousseau, P. H. Pathak, and H.-T. Chou, A time domain formulation of the uniform geometrical theory of diffraction for scattering from a smooth convex surface, IEEE Trans. Antennas Propag., vol. 55, no. 6, pp. 15221534, Jun. 2007. [13] H.-T. Chou, P. R. Rousseau, and P. H. Pathak, TD-UTD solutions for the transient radiation and surface elds of pulsed antennas placed on PEC smooth convex surfaces, IEEE Trans. Antennas Propag., to be published. [14] P. M. Johansen, Time-domain version of the physical theory of diffraction, IEEE Trans. Antennas Propag., vol. 47, pp. 261270, Feb. 1999.

[15] E. Y. Sun, Transient analysis of large paraboloidal reector antennas, IEEE Trans. Antennas Propag., vol. 43, pp. 14911496, Dec. 1995. [16] H.-T. Chou and S.-C. Tuan, Analytic analysis of transient scattering from a nite second-order surface illuminated by an incident plane wave, IEEE Trans. Antennas Propag., vol. 54, no. 9, pp. 24632471, September 2006. [17] H.-T. Chou, S.-C. Tuan, and H.-H. Chou, Transient analysis of scattering from a perfectly conducting parabolic reector illuminated by a Gaussian beam electromagnetic eld, IEEE Trans. Antennas Propag., vol. 58, no. 5, pp. 17111719, May 2010. [18] H.-T. Chou, P. H. Pathak, and P. R. Rousseau, Analytic solution for early-time transient radiation from pulse-excited parabolic reector antennas, IEEE Trans. Antennas Propag., vol. 45, pp. 829836, May 1997. [19] R. de Oliveira and M. Hlier, Closed-form expressions of the axial step and impulse responses of a parabolic reector antenna, IEEE Trans. Antennas Propag., vol. 55, no. 4, pp. 10301037, Apr. 2007. [20] J. S. Tyo, E. G. Farr, and D. I. Lawry, Effect of defocus on the prompt response of a reector IRA, IEEE Trans. Antennas Propag., vol. 53, no. 10, pp. 32473254, Oct. 2005. [21] R. C. Hansen, Short-pulse excitation of reector antennas, Proc. Inst. Elect. Eng., vol. 134, pt. H, pp. 557559, Dec. 1987. [22] R. C. Hansen, Correction to short-pulse excitation of reector antennas, Proc. Inst. Elect. Eng., vol. 139, pt. H, pp. 119120, Feb. 1992. [23] E. Y. Sun and W. V. T. Rusch, Time-domain physical-optics, IEEE Trans. Antennas Propag., vol. 42, pp. 915, Jan. 1994. [24] Y. M. Chen, The transient behavior of diffraction of plane pulse by a circular cylinder, Int. J. Engrg. Sci., vol. 2, pp. 417429, 1964. [25] J. R. Wait, Transient response of the penumbral currents for plane wave diffraction by a cylinder, Can. J. Phys., vol. 47, pp. 13071312, 1969. [26] P. H. Pathak, High-frequency techniques for antenna analysis, Proc. IEEE, vol. 80, pp. 4465, Jan. 1992.

Shih-Chung Tuan (M05) was born in Kaohsiung, Taiwan, R.O.C., on November 19, 1962. He received the B.S. degree from National Taiwan Ocean University, Keelung, in 1990, the M.S. degree from Tatung University, Taipei, Taiwan, in 1997, and the Ph.D. degree from Yuan Ze University, Taiwan, in 2006, all in electrical engineering. He is with the Oriental Institute of Technology. His interests are in the development and use of asymptotic high-frequency methods as well as numerical methods for analyzing electromagnetic antenna and scattering problems.

Hsi-Tseng Chou (S96M97SM01) was born in Taiwan, in 1966. He received the B.S. degree in electrical engineering from National Taiwan University in 1988, and the M.S. and Ph.D. degrees in electrical engineering from Ohio State University (OSU), in 1993 and 1996, respectively. He joined Yuan-Ze University (YZU), Taiwan, in August 1998, and is currently a Professor in the Department of Communications Engineering. His research interests include wireless communication network, antenna design, antenna measurement, electromagnetic scattering, asymptotic high frequency techniques such as uniform geometrical theory of diffraction (UTD), novel Gaussian beam techniques, and UTD type solution for periodic structures. He has published more than 250 journal and conference papers. Dr. Chou is an elected member of URSI International Radio Science US commission B. He has received two awards from Taiwanese Ministry of Education and Ministry of Economic Affairs, in 2003 and 2008, respectively, to recognize his distinguished contributions in promoting academic researches for industrial applications, which were the highest honors these two ministries have given to university professors to recognize their industrial contributions.

TUAN et al.: ANALYTIC TRANSIENT ANALYSIS OF RADIATION FROM ELLIPSOIDAL REFLECTOR ANTENNAS

339

Kung-Yu Lu was born in Tainan, Taiwan in 1967. He received the B.S. degree in applied physics from Chung Cheng Institute of Technology, Taoyuan, Taiwan, in 1989, the M.S. degree in electrical engineering from Yuan Ze University, Taoyuan, Taiwan, in 1996. He is currently working toward the Ph.D. degree at National Taiwan University, Taipei, Taiwan. He has been involved in the development of radar technologies at Chung-Shan Institute of Science and Technology. His current research interests include computation parallelization of EM numerical simulation and radar detection in both near- and far-zone applications.

Hsi-Hsir Chou was born in ChangHua Taiwan, in 1975. He received the Ph.D. degree in engineering from Cambridge University, U.K. in 2008. From 2004 to 2008, he was involved in collaborating with ALPS UK Co. Ltd. and Dow Corning Co. Ltd. in the development of patented free-space optical interconnection technologies, ferroelectric liquid crystal devices and carbon nanotube dielectric devices during his Ph.D. program at Cambridge University. He joined the Department of Engineering Science, Oxford University, U.K., in July 2008 as a Postdoctoral Researcher in the development of high-speed visible light communication technologies sponsored by Samsung Electronics Co. Ltd., Korea, before he returned Taiwan to join the Communication Research Center, Yuan Ze University, Taiwan, as a Researcher in May, 2009. His current research interests include free-space optical interconnection technologies, ferroelectric liquid crystal devices, carbon nanotube dielectric devices and antenna design. Dr. Chou is a lifetime member of Trinity College, Cambridge and a Fellow of Cambridge Overseas Society since 2005.

340

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

An Analytic Solution of Transient Scattering From Perfectly Conducting Ellipsoidal Surfaces Illuminated by an Electromagnetic Plane Wave
Hsi-Tseng Chou, Shih-Chung Tuan, Associate Member, IEEE, Kung-Yu Lu, and Hsi-Hsir Chou
AbstractThis paper presents an analytical and closed-form solution, using a time domain (TD) physical optics (PO), for the fast analysis of transient scattering from a nite and perfectly conducting ellipsoidal surface when it is illuminated by a transientstep plane wave. The advantage of ellipsoidal shapes to resemble a variety of realistic surfaces such as spherical, parabolic or planar surfaces allows the developed solution applicable to model a realistic scattering object such as an aircraft in an effective fashion. Physical appealing interpretation of wave phenomena in terms of reection and diffraction mechanisms is also provided in the solution. Numerical examples are presented to demonstrate its physical phenomena of scattering mechanisms. Index TermsElectromagnetic scattering, ellipsoidal surfaces, time-domain physical optics, transient analysis.

I. INTRODUCTION physically appealing, direct in time domain (TD) analysis of electromagnetic (EM) elds scattered from a perfectly conducting ellipsoidal surface truncated by nite edges, when it is illuminated by a transient-step plane wave, is presented. The analysis achieved a solution in a closed form and is applicable to the increasing applications in the ultrawideband or short pulse target identication and remote sensing applications [1][10]. In this aspect, ellipsoidal shapes appear to be more exible in modeling a variety of realistic surfaces such as spherical, parabolic and planar surfaces, where these various nite ellipsoidal-like surfaces may be used to resemble a realistic scattering object such as an aircraft. This type of scattering analysis from electrically large objects remains challenging even with todays computational power. Among the potential analysis techniques, quasi-analytical TD solutions are attractive because not only the conventional TD numerical techniques, such as the nite difference time domain (FDTD) and TD integral techniques, are still suffered from computational inefciency in practical applications, but also

Manuscript received January 21, 2011; revised June 07, 2011; accepted July 25, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. H.-T. Chou is with the Department of Communications Engineering, Yuan-Ze University, Chung-Li, Taiwan (e-mail: hchou@saturn.yzu.edu.tw). S.-C. Tuan is with the Department of Communication Engineering, Oriental Institute of Technology, Taipei, Taiwan (e-mail: t1200475@ms32.hinet.net). K.-Y. Lu is with the Graduate Institute of Communication Engineering, National Taiwan University, Taipei, Taiwan (e-mail: lky5719@gmail.com). H.-H. Chou is with the Communication Research Center, Yuan Ze University, Chung-Li, Taiwan (e-mail: hsi-hsir.chou@trinity.cantab.net). Digital Object Identier 10.1109/TAP.2011.2167930

the quasi-analytical solutions may provide interpretations of wave behaviors. This advantage has driven the efforts of many past works as referenced in [1][32], among which TD uniform geometrical theory of diffraction (TD-UTD) [20][24] and physical theory of diffraction (TD-PTD) [25], [26] have been successfully obtained by using either a direct inverse Laplace transform [20], [21], [25][29], [31], [32] or an analytical time transform (ATT) [22][24], [30] of the corresponding FD formulations [33], [34]. It has been shown that these ray-based closed form solutions provide accurate analysis for the early-time or narrow pulse responses. However, the suffering of becoming singular near the optical ray caustics or multiple-overlapped regions of ray shadow boundaries has limited the applications of those solutions. As used in FD, the TD physical optics (TD-PO) [27][30] and TD aperture integration (TD-AI) [31], [32] techniques may be used to overcome these difculties by using a numerical integration over the radiation integrals of a set of well dened induced currents. These surface-integration based solutions appear to lack the capability of wave phenomenon interpretation as interested in TD-UTD and TD-PTD. We have attempted to resolve this restriction, as motivated in this paper, by developing solutions for a relatively general surface. One of the successful works is the development of TD analytic solution of scattering from a parabolic surface in [29] by using TD-PO approximation. The current work further extends it to treat a more sophisticated ellipsoidal surface. However, the current solution appears to be more superior and complete than that in [29] because an ellipsoidal surface may be used to more accurately model realistic and general scatters which can not be achieved by simply using parabolic ones, such as a sphere. Furthermore, the current solution can be reduced to the previous one in [29] because an ellipsoidal surface will reduce to a parabolic one when the radius of the ellipse approaches to an innite. Fig. 1 illustrates several realistic scatterers modeled by truncating ellipsoidal surfaces with nite edge contours including a sphere, a parabolic surface, an ellipsoidal body, a nite cylinder and a planar surface, which appear in many practical objects. This paper is organized in the following format. Section II describes the TD-PO formulations for a transient-step incident wave. Section III shows the reductions of the formulation for special cases illustrated in Fig. 1(b). Transient behaviors of the scattered eld are discussed in Section IV. Examples that demonstrate the physical phenomena of the scattering mechanisms are presented in Section V. Finally some conclusive remarks are discussed in Section VI.

0018-926X/$26.00 2011 IEEE

CHOU et al.: AN ANALYTIC SOLUTION OF TRANSIENT SCATTERING FROM PERFECTLY CONDUCTING ELLIPSOIDAL SURFACES

341

Fig. 1. Typical structures of scatters to assemble a realistic structure of scatter such as aircraft: (a) shows a realistic aircraft model and (b) shows a variety of shaped surfaces that can be used to approximate the model in (a).

Fig. 2. The relation of integration contour to the ellipsoidal surfaces with nite is formed by the edges in the lit region and the shadow boundary of edges. is the surface in the lit region to the incident eld, and is scatterer, and the plane described by (4). (a) Illustration of PO edge the intersection of contour; (b) illustration of integration contour.

II. FORMULATIONS OF THE TRANSIENT STEP RESPONSE Consider an ellipsoidal surface described by (1) where is a position vector on the surface with , and being its radii along z-, x- and y-axes, respectively. The actual surface is formed by taking a part from the ellipsoidal surface described in (1) with an edge contour, which is then illuminated by a plane wave incident from the direction, . Since PO considers only the induced surface currents in the lit region of incident eld, an effective edge contour, , as illustrated in Fig. 2(a), is formed by the original edge contour and the shadow boundary of incident eld such that the surface under examination is also in the lit region of incident eld. It is noted that (1) may approximately model a variety of realistic surfaces such as these shown in Fig. 1. For example, when , it reduces to a sphere. The electrical eld of the incident plane wave in TD is described by (2) where is a Heaviside step function, and the time reference, , is selected at , which allows for a

. In (2), is a real vector denoting negative value of the amplitude and polarization of the plane wave with and being the wave speed in free space. The elds scattered from the ellipsoidal surface can be obtained by TD-PO. At in the far zone of the scatter, it can be approximated by [29]

(3) with being the outward surface normal unit vector pointing to the opposite direction of the incident eld, and is the portion of the ellipsoidal surface truncated by in the lit region of the incident eld. The superscript indicates that the eld is the response to a transient-step incident wave. Also in (3), is the Dirac delta function, which may reduce (3) to a line integral at the condition of zero parameter for the Dirac delta function at by (4) This equation describes a line contour on as illustrated in Fig. 2(b). Let , is an intersection between the ellipsoidal surface and a plane with being its normal vector. This plane shifts in parallel along the direction of at different time values. As shown in

342

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Appendix A.1, the projection of part of an ellipse expressed by

on a

plane is a

(5) where

(6)

(7) and (8) In (6)(8) (9) where , and (10)

where . Note that (15) has a direction pointing to , and the negative sign in (14) should be used when points to . This sign selection can be justied by . In (13), is the angular span of formed by the intersections of and . The step functions in (13) are used to indicate whether and have intersections, where and are beginning and ending time that the intersection exists. These behaviors have been discussed in [29], and are omitted here for brevity. It is assumed that there are only two intersection points, and in the following study. The case with multiple pairs of intersection points can be treated by using a superposition of each pairs contribution. It was shown in Appendix A.2 that (16) Thus substituting (15) and (16) into (13) makes the integrand in (13) become

(17)

and results in a closed-form expression by coordinates can be related to the This by the following transformation: coordinates (18) (11) The detailed derivations of (5)(11) are shown in Appendix A.1 for reference. Equation (3) can be expressed in a closed-form by following the procedure in [29]. Thus using the following representation (12) and employing (15) in [29] for the delta functions identity result in a form identical to (16) in [29] by where are dened in (A13) by (20a) where each term was derived in Appendix A.3, and given by (19a)

(19b)

(13) where (14a) (14b) (14c) For an ellipsoidal scatterer, it can be shown that

(20b) Substituting (19) into (18), and re-organizing the terms give (21) where and are referred to reected and edge diffracted components, respectively. In particular,

(15)

(22a)

CHOU et al.: AN ANALYTIC SOLUTION OF TRANSIENT SCATTERING FROM PERFECTLY CONDUCTING ELLIPSOIDAL SURFACES

343

(22b) ( is used to denotes either or ) and (22c) with (22d) The reection component in (22a) can be re-addressed as (23) because it is evident that the vectors in (22a) and (23) point to the same direction normal to the surface at . Thus one may dene the surface normal unit vector at by (24) Substituting (7), (20) and (24) into (22a) gives (25) Substituting (25) into (13) gives

Fig. 3. Unit vectors to illustrate the relationship between incident and reected elds.

existence of each term. Especially the reected eld component is accounted by , and the edge diffracted terms will be accounted by their difference of values at and . III. REDUCED FORMULATIONS FOR SPECIAL CASES The formulation of the TD transient response in Section II can be reduced to several special cases as shown below. A. A Rotationally Symmetric Ellipsoidal Surface In this case, one considers . The formulations in Section II remain identical except the following reduction of parameters. First the coordinate rotation angle in (10) become (30) and thus (31) which makes (7) and (9) become

(32a) (32b) The parameters, and in (6) reduce to

(26) where the unit vectors are illustrated in Fig. 3. It is noted that at the point of reection, where in (8) (27) which make (26) become The formulations in (20) becomes

(33)

(34) Substituting these formulation reductions in (30)(34) into (26), (28) and (29) results in the eld solutions by

(28) The diffraction component can be obtained by substituting (22b) into (13): (29) where the step functions, , in (26), (28) and (29) were dropped without losing the validity of the solutions as pointed in [29]. The main reason is because this solution is based on the search of and , which automatically accounts the

(35a) at any , and (35b)

344

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

where

In this case, (6) becomes

(36a) with (41) (36b) and (36c) The eld solutions are identical to (26), (28) and (29).

D. Finite Planar Surface In this case, one makes . The coordinate rotation relations are identical to (30) and (31). Also The parameters, and in (33) reduce to (42) remains same in (32b), and in (32a) become

B. Spherical Surface In this case, , and the formulations can be further reduced from part A of Section III. Thus the coordinate rotation angle is identical to (30) and (31), and (33) becomes (37) and (38) The eld solutions now become (39a) and (39b) where Also in this case, in (21) and (22) become makes It is noted that the projection of scribed by

(43) is approximately de-

(44) . The formulations

(45a)

(45b) It is noted that since where eld can be described by (39d) (46) C. Finite Cylindrical Surface In this case, one assumes , and makes being the axis of the cylinder. Thus the coordinate rotation angle in (10) becomes (40) it can be approximated by is illustrated in Fig. 4. The

(39c) and

IV. CHARACTERISTIC DISCUSSIONS Based on the derivations, several observations on the characteristics are discussed in the following:

CHOU et al.: AN ANALYTIC SOLUTION OF TRANSIENT SCATTERING FROM PERFECTLY CONDUCTING ELLIPSOIDAL SURFACES

345

where

(50a) (50b) and

Fig. 4. Illustration of transient scattering phenomena for a nite planar surface. At , the plane of (4) will intersect with the nite planar surface at a line with a length of .

(50c) It is noted that can be found from (36a) as

A. Forward Scattering, or In this case, , and (4) becomes . In this case, (3) becomes where (51a) (47) which exhibits a transient impulse function in time. B. Back Scattering, , for a Closed Surface and contour on function of plane be . Thus . Let the projection of edge which is also a which gives (51b) with at .

(50d)

, and makes the transient function In this case, equal to an unity and edge diffraction component vanishing in (22b) and (22c), respectively. The reected eld component in (26) becomes

E. Response to a General Astigmatic Finite TD Pulse (48) which exhibits a transient response of ramp-function in time. C. Impulse-Scattering Field In addition to the discussion in part A, an impulse scattering eld occurs when (4) vanishes. This condition will be true when the surface is planar with pointing to its normal direction. In another word, is in the reection direction of the incident wave. D. Response to an Impulse Incident Plane Wave The response to an impulse incident plane wave can be obtained by differentiating the total eld as or alternatively (52b) The response of a general astigmatic nite-energy pulse of a , of the incident wave can be obtained by a time function, direct time convolution with (49) by (52a)

V. NUMERICAL EXAMPLES In this section we present some numerical examples to demonstrate the characteristics of transient scattering from a variety of scatters. A. Transient Scattering From a Planar Surface The rst example considers a circular PEC disk located on the x-y plane, which has a radius of 1 meter. The plane wave is incident from . The observation points are located at . The transient scattering elds are shown

(49)

346

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 5. Transient responses of a PEC circular disk of 1 m in radius, which is , . (a) illustrated by a plane wave incident from plane: shows the responses in the specular scattering directions, while (b) plane: shows the responses in the backscattering directions.

Fig. 6. Transient responses of a PEC rectangular disk with in dimen, . sions, which is illustrated by a plane wave incident from : shows the responses in the specular scattering directions, while (a) : shows the responses in the backscattering directions. Rectangular (b) pulses of transient responses are observed.

in Fig. 5(a) and (b) for the observations located on and planes, respectively, which are in the specular and backward scattering directions, respectively. As demonstrated in Section III-D that the amplitude of the transient response is proportional to the length of the line formed by the intersection between this disk and the plane of (4) at . Note that the plane of (4) shifts linearly in the direction of when changes. Due to the symmetry of the disk, it can be seen that the initial and ending times, and , for the transient responses to have non-zero values occur at the condition that the plane of (4) is tangent to the edge circle of the disk, where two tangent points at and , respectively can be found. In this case, the transient responses are both zero at and . Fig. 5(a) shows the transient responses in the forward directions at , 30 , 60 and 90 . It is observed that the TD responses exhibit narrower pulses with higher amplitudes as the observation direction approaches to the reection direction of the incident plane eld. Away from this specular direction, the transient scattering eld drops rapidly to low values, which can also be observed in Fig. 5(b) for the cases in the backward scattering directions. In all cases, the amplitudes increase gradually from a null eld at to a maximum, which occurs as the intersection of these two planes penetrates through the center of the disk, and then continue to decrease to become a null eld again at .

The next example considers a rectangular disk of oriented on the x-y plane between and . The illuminating plane wave is incident from . The observation points are same as those in Fig. 5. In this case, the initial and ending times, and , occur at where the plane of (4) intersects with this disk. Thus after , the amplitude jumps to a value, and remains constant between and , where the responses are rectangular pulses. In this case, the length of the intersected line interface is a constant and equal to 4 m for all cases. The differences in amplitudes of the transient responses shown in Fig. 6 are resulted from the outer unit vector products in (46). Also it is observed that except the behaviors of rectangular pulses, the phenomena shown in Fig. 6 are identical to those shown in Fig. 5 including the characteristics of pulse widths which are smaller when the observation point moves toward the specular direction of reection. It is noted that as demonstrated in Section IV-C the transient response becomes impulsive as in the direction of reected eld as dened by geometrical optics (GO). B. Transient Scattering From a Half Sphere The examination considers a half sphere (the radius is 1 m) in the space, which is illuminated by a plane wave incident from z-axis. This case also illustrates the situation of a full

CHOU et al.: AN ANALYTIC SOLUTION OF TRANSIENT SCATTERING FROM PERFECTLY CONDUCTING ELLIPSOIDAL SURFACES

347

Fig. 7. Illustration of transient scattering from a half sphere, and its transient responses for various observation angles. The plane wave is incident from z-axis. (a) Illustration of and on the x-z plane; (b) transient responses.

sphere illuminated by the same plane wave because the illuminating lit region of the sphere is its portion in space. The shadow boundary is now the circle of 1 m in radius at plane, which forms an equivalent edge contour, . One rst considers an observation point on the z-axis at and (vertical observation in Fig. 7(a)). In this case, the integration contour, , for between and is a circle, where and occur at the time when the plane of (4) intersects with the sphere at and the edge contour. Thus the transient function in (22c), and the diffraction term in (29) vanishes. The phenomena of reection term in (26) are shown in Fig. 7(b). It can be observed that at the transient response jumps to a large value, which may be viewed as the reected eld because the surface at is the point of reection in the point of GO view. After , the transient response decreases linearly to a value of zero at . The reason is due to a fact that the induced currents estimated by PO will decrease linearly from a maximum value at the reection point to zero at the shadow boundary, which appears at edge contour at . This phenomenon of curvature effects is distinguished from that observed in the case of a parabolic surface shown in [29], where a constant value appears for the reection term. One next considers the cases of observation in the oblique angles on the x-z plane, where , 60 , and 90 are considered. As illustrated in Fig. 7(a), three times of , and need to be considered, which are the time that the plane of (4) touches the surface (reection point), the edge point at and the edge point at , respectively. At , only the reection term in (26) exists, which exhibits the phenomena of transient response identical to these exhibited in the case of as demonstrated in Fig. 7(b). At , the transient function and diffraction term come into effects. In this case, both the reection term and transient function continue to decrease and become zero at

Fig. 8. Transient response of scattering from a half- and a full-sphere illuminated by a normally incident plane wave. TD-PO gives a same solution for the scattering from these two structures. The late time responses in CST results (indicated by sphere and Half-sphere in the gures) are caused by the edge or surface diffractions, which were not predicted by TD-PO because of null currents at the shadow boundary. (a) At 0.5 m; (b) at 1.25 m.

, while the diffraction term starts to increase its effect. It is noted that the diffraction term increases initially, reaches a maximum, and then decreases to zero at . To further validate the feasibility of the presented solution, a comparisons with numerically exact solutions using CST [35], a commercial code based on a TD nite integration method, were made as shown in Fig. 8(a) and (b). In this case, the sphere has 0.25 m in radius, and is illuminated by a normally incident plane wave with a Gaussian distribution in TD. Due to a fact that TD-PO considers only the induced currents on the surface of sphere in the lit to the incident eld, both cases of half- and full-spheres result in the same responses by TD-PO, which are however different by using CST as indicated by sphere and half-sphere in the gures. Furthermore, TD-PO only accounts the contributions of reection and rst-order diffractions. In this case, TD-PO results in vanishing diffractions from the shadow boundary because of null induced currents. Fig. 8(a) and (b) show the transient responses at and 1.25 m, respectively. It is noted that CST assumes a nite aperture to illustrate a plane wave and incurs a truncation effect. Thus our simulation uses the TD waveform computed by CST near the sphere as the input of the plane wave as shown in Fig. 8(a). It is observed that this TD-PO predicts accurate transient responses for the early time. The late time responses shown in CST simulations are caused by edge or surface diffractions, which were not predicted by TD-PO.

348

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

complex scatter such as an aircraft. The developed solutions will be very effective in analyzing the transient behaviors of such scattering problems. APPENDIX A. Derivation of the Elliptic Contour in (5) Substituting (1) into (4) gives (A1) Moving the linear terms on the left-hand side to the right-hand side and taking a square of the resulted equation give

(A2) One thus introduces a coordinate transformation by (A3) which is substituted into (A2) and makes the coefcient of vanished. Thus it can be shown that (A4) and on the plane, (5)(9) will be resulted.

Fig. 9. Illustration of transient scattering from a nite half-cylinder with end caps, and its transient responses for various observation angles. (a) Illustration of the scattering problem; (b) transient responses.

C. Transient Scattering From a Finite Half Cylinder Finally one considers the transient scattering from a nite half-cylinder as illustrated in Fig. 9(a). The cylinder has a ra). dius of 1 m, and a length of 4 m (truncated at The plane wave is incident from and . In this case, the illuminating lit region includes the surface of end cap on the left-hand side and the cylinder surface (the end cap on the right-hand side is in the shadow region). Similar to the phenomena shown in Section V-B, four times of , , and need to be considered as also shown in Fig. 9(a). For , the scatterings from the end cap on the left-hand side and a part of the cylinder surface contribute to the total eld responses. For , only the scattering from the cylinder surface contributes to the total eld. In this case, the and integration contour is a half circle, and thus . Thus the transient responses in this time period the integraremain constant. On the other hand, for tion contour is only a portion of circle on the cylinder surface, . The transient rewhich gradually decreases to zero at , 30 , sponses are shown in Fig. 9(b) for the cases of plane. It is observed that a narrower 60 , and 90 on pulse will give a higher amplitude of transient response. VI. CONCLUSIONS This paper presents a quasi-analytic TD solution to perform a transient analysis of scattering from an ellipsoidal surface with a nite edge contour when it is illuminated by a transient-step plane wave. The nite ellipsoidal-like surfaces may resemble a variety of realistic surfaces that are parts of a realistic and large

B. Derivation of (16) The derivation of (16) starts from re-addressing (4) by dening (A5) , . Following a procedure similar Apparently at to (A1), an elliptic equation similar to (5) can be obtained by (A6) where and are dened in (6), and (A7) and (A8) with and dened in (9). One then substitutes (12) into (A6) and takes the derivative of (A6) with respect to . Afterward making at , gives (A9) Thus (A10)

CHOU et al.: AN ANALYTIC SOLUTION OF TRANSIENT SCATTERING FROM PERFECTLY CONDUCTING ELLIPSOIDAL SURFACES

349

C. Obtaining the Closed-Form Solution of (17) Substituting (14), (15) and (16) into (17) gives (18) with

It is noted that at the following:

, one may use (4), (11) and (12) to nd

(A12) where (A13) Apparently using (A12) will make the integration in (A11) in closed-forms, which are shown in (19). REFERENCES
[1] C. E. Baum, E. G. Farr, and D. V. Giri, Review of impulse-radiating antennas, in in Review of Radio Science. Oxford, U.K.: Oxford Univ. Press, 1999, pp. 403439. [2] S. P. Skulkin and V. I. Turchin, Transient elds of parabolic reector antennas, in Ultra-Wideband Short Pulse Electromagnetics, C. E. Baum, L. Carin, and A. P. Stone, Eds. New York: Plenum, 1997, vol. 3, pp. 8187. [3] E. G. Farr, C. E. Baum, W. D. Prather, and L. H. Bowen, Multifunction impulse radiating antennas: Theory and experiment, in UltraWideband, Short-Pulse Electromagnetics 4. New York: Kluwer Academic/Plenum Publishers, 1999. [4] J. S. Tyo, E. G. Farr, and D. I. Lawry, Effect of defocus on the prompt response of a reector IRA, IEEE Trans. Antennas Propag., vol. 53, no. 10, pp. 32473254, Oct. 2005. [5] J. B. Keller and A. Blank, Diffraction and reection of pulses by wedges and corners, Comm. Pure Appl. Math., vol. 4, pp. 7594, Jun. 1951. [6] F. G. Friedlander, Sound Pulses. Cambridge: Cambridge Univ. Press, 1958. [7] R. de Oliveira and M. Hlier, Closed-form expressions of the axial step and impulse responses of a parabolic reector antenna, IEEE Trans. Antennas Propag., vol. 55, no. 4, pp. 10301037, Apr. 2007. [8] R. C. Hansen, Short-pulse excitation of reector antennas, in Proc. Inst. Elect. Eng., Dec. 1987, vol. 134, pp. 557559, pt. H. [9] Y. M. Chen, The transient behavior of diffraction of plane pulse by a circular cylinder, Int. J. Engrg. Sci., vol. 2, pp. 417429, 1964. [10] J. R. Wait, Transient response of the penumbral currents for plane wave diffraction by a cylinder, Canadian J. Phys., vol. 47, pp. 13071312, 1969. [11] L. B. Felsen, Propagation and diffraction of transient elds in nondispersive and dispersive media, in Transient Electromagnetic Fields, L. B. Felsen, Ed. New York: Springer-Verlag, 1976. [12] L. B. Felsen, Diffraction of the pulsed eld from an arbitrarily oriented electric or magnetic dipole by a perfectly conducting wedge, SIAM J. Appl. Math, vol. 26, pp. 306312, Mar. 1974. [13] K. I. Nikoskinen, M. E. Ermltlu, and I. V. Lindell, Transient image theory for 2-D and 3-D conducting wedge problems, IEEE Trans. Antennas Propag., vol. 42, pp. 1515l520, Nov. 1994. [14] R. Ianconescu and E. Heyman, Pulsed eld diffraction by a perfectly conducting wedge: A spectral theory of transients analysis, IEEE Trans. Antennas Propag., vol. 42, pp. 78l789, Jun. 1994. [15] R. Ianconescu and E. Heyman, Pulsed eld diffraction by a perfectly conducting wedge: Exact solution, IEEE Trans. Antennas Propag., vol. 42, pp. 13771385, Oct. 1994. [16] E. Heyman and R. Ianconescu, Pulsed eld diffraction by a perfectly conducting wedge: Local scattering models, IEEE Trans. Antennas Propag., vol. 43, pp. 519528, May 1995. [17] E. Heyman and L. B. Felsen, Weakly dispersive spectral theory of transients (STT). Part I: Formulation and interpretation, IEEE Trans. Antennas Propag., vol. AP-35, pp. 8086, Jan. 1987.

[18] E. Heyman and L. B. Felsen, Weakly dispersive spectral theory of transients (STT), Part : Evaluation of the spectral integral, IEEE Trans. Antennas Propag., vol. AP-35, pp. 574580, May 1987. [19] E. Heyman, Weakly dispersive spectral theory of transients (STT), Part III: Applications, IEEE Trans. Antennas Propag., vol. AP-35, pp. l2581987, Nov. 1987. [20] T. Jirapunth and R. G. Kouyoumjian, Early-time responses of currents and charges on wedges and strips, in IEEE Antennas Propagation Symp. Digest, 1979, vol. 2, pp. 590593. [21] T. W. Veruttipong, Time domain version of the uniform GTD, IEEE Trans. Antennas Propag., vol. 38, pp. 17571764, Nov. 1990. [22] P. R. Rousseau and P. H. Pathak, Time-domain uniform geometrical theory of diffraction for a curved wedge, IEEE Trans. Antennas Propag., vol. 43, no. 12, pp. 13751382, Dec. 1995. [23] P. R. Rousseau, P. H. Pathak, and H.-T. Chou, A time domain formulation of the uniform geometrical theory of diffraction for scattering from a smooth convex surface, IEEE Trans. Antennas Propag., vol. 55, no. 6, pp. 15221534, Jun. 2007. [24] H.-T. Chou, P. R. Rousseau, and P. H. Pathak, TD-UTD solutions for the transient radiation and surface elds of pulsed antennas placed on PEC smooth convex surfaces, IEEE Trans. Antennas Propag., to appear on. [25] P. M. Johansen, Time-domain version of the physical theory of diffraction, IEEE Trans. Antennas Propag., vol. 47, no. 2, pp. 261270, Feb. 1999. [26] A. Altintas and P. Russer, Time-domain equivalent edge currents for transient scattering, IEEE Trans. Antennas Propag., vol. 49, no. 4, pp. 602606, Apr. 2001. [27] E.-Y. Sun and W. V. T. Rusch, Time-domain physical optics, IEEE Trans. Antennas Propag., vol. 42, no. 1, pp. 915, Jan. 1994. [28] E. Y. Sun, Transient analysis of large paraboloidal reector antennas, IEEE Trans. Antennas Propag., vol. 43, no. 12, pp. 14911496, Dec. 1995. [29] H.-T. Chou and S.-C. Tuan, Analytic analysis of transient scattering from a nite second-order surface illuminated by an incident plane wave, IEEE Trans. Antennas Propag., vol. 54, no. 9, pp. 24632471, Sep. 2006. [30] H.-T. Chou, S.-C. Tuan, and H.-H. Chou, Transient analysis of scattering from a perfectly conducting parabolic reector illuminated by a Gaussian beam electromagnetic eld, IEEE Trans. Antennas Propag., vol. 58, no. 5, pp. 17111719, May 2010. [31] H.-T. Chou and P. H. Pathak, Analytic solution for early-time transient radiation from pulse-excited parabolic reector antennas, IEEE Trans. Antennas Propag., vol. 45, no. 5, pp. 829836, May 1997. [32] H.-T. Chou, S.-C. Tuan, K.-Y. Lu, and H.-H. Chou, Analytic transient analysis of radiation from ellipsoidal reector antennas for impulseradiating antennas applications, IEEE Trans. Antennas Propag., vol. 60, no. 1, pp. 328339, Jan. 2012. [33] P. H. Pathak, High-frequency techniques for antenna analysis, Proc. IEEE, vol. 80, pp. 4465, Jan. 1992. [34] P. H. Pathak, Techniques for high frequency problems, in Antenna Handbook: Theory, Application and Design, Y. T. Lo and S. W. Lee, Eds. New York: Van Nostrand Reinhold, 1988, ch. 4. [35] CST Studio Suite 2009 [Online]. Available: www.cst.com

Hsi-Tseng Chou (S96M97SM01) was born in Taiwan, in 1966. He received the B.S. degree in electrical engineering from National Taiwan University, in 1988, and the M.S. and Ph.D. degrees from Ohio State University (OSU), Columbus, in 1993 and 1996, respectively, all in electrical engineering. He joined Yuan-Ze University (YZU), Taiwan, in August 1998, and is currently a Professor in the Department of Communications Engineering. His research interests include wireless communication network, antenna design, antenna measurement, electromagnetic scattering, asymptotic high frequency techniques such as uniform geometrical theory of diffraction (UTD), novel Gaussian beam techniques, and UTD type solution for periodic structures. He has published more than 250 journal and conference papers. Dr. Chou is an elected member of URSI International Radio Science US commission B. He has received two awards from the Taiwanese Ministry of Education and Ministry of Economic Affairs, in 2003 and 2008, respectively, to recognize his distinguished contributions in promoting academic researches for

350

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

industrial applications, which were the highest honors these two ministries have given to university professors to recognize their industrial contributions.

Shih-Chung Tuan (A06) was born in Kaohsiung, Taiwan, R.O.C., on November 19, 1962. He received the B.S. degree from National Taiwan Ocean University, Keelung, Taiwan, in 1990, the M.S. degree from Tatung University, Taipei, Taiwan, in 1997, and the Ph.D. degree from Yuan Ze University, Taiwan, in 2006, all in electrical engineering. He is with the Oriental Institute of Technology, Tapei, Taiwan. His interests are in the development and use of asymptotic high-frequency methods as well as numerical methods for analyzing electromagnetic antenna and scattering problems.

Hsi-Hsir Chou was born in ChangHua Taiwan, in 1975. He received the Ph.D. degree in engineering from Cambridge University, London, U.K., in 2008. From 2004 to 2008, he was involved in collaborating with ALPS UK Co. Ltd. and Dow Corning Co. Ltd. in the development of patented free-space optical interconnection technologies, ferroelectric liquid crystal devices and carbon nanotube dielectric devices during his Ph.D. program at Cambridge University. He joined the Department of Engineering Science, Oxford University, U.K. in July 2008 as a Postdoctoral Researcher in the development of high-speed visible light communication technologies sponsored by Samsung Electronics Co. Ltd., Korea, before he returned Taiwan to join the Communication Research Center, Yuan Ze University, Taiwan as a Researcher in May, 2009. His current research interests include free-space optical interconnection technologies, ferroelectric liquid crystal devices, carbon nanotube dielectric devices and antenna design. Dr. Chou is a lifetime member of Trinity College, Cambridge and a Fellow of the Cambridge Overseas Society since 2005.

Kung-Yu Lu was born in Tainan, Taiwan, in 1967. He received the B.S. degree in applied physics from Chung Cheng Institute of Technology, Taoyuan, Taiwan, in 1989, the M.S. degree in electrical engineering from Yuan Ze University, Taoyuan, Taiwan, in 1996. He is currently pursuing the Ph.D. degree at National Taiwan University, Taipei, Taiwan. He has been involved in the development of radar technologies in the Chung-Shan Institute of Science and Technology. His current research interests include computation parallelization of EM numerical simulation and radar detection in both near- and far-zone applications.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

351

Greens Function Extraction for Interfaces With Impedance Boundary Conditions

Evert Slob and Kees Wapenaar
AbstractTheory and experiments to obtain the response between two receivers from cross correlations of wave elds recorded at these receivers are well established. The principle relies on mutually spatial and temporal uncorrelated contributions from sources on a boundary enclosing the two receivers, which upon cross correlation interfere only constructively for signals traveling between the two receivers. It has, therefore, become generally known as interferometry. The theory includes situations with ow, mechanical, and electromagnetic eld uctuations, and their mutual coupling. Here, we present an electromagnetic theory for Greens function retrieval from cross correlations that incorporates general bianisotropic media in which interfaces are present where bianisotropic impedance boundary conditions apply. The derived Greens function representation shows that in lossless media and for interfaces with lossless impedance boundary conditions the Greens function between two receivers is obtained by cross correlating the recordings of these receivers from sources on a boundary enclosing them. We show numerical examples in 2-D where proper solutions are numerically tractable and practical approximations illustrate that numerically less intensive algorithms lead to acceptable results and accurate extraction of reections from impedance boundary surfaces. Because the method is data driven, it is suitable for experimental Greens function extraction from measured data. Index TermsBianisotropic media, impedance boundary conditions, interferometry.

I. INTRODUCTION

NTERFEROMETRIC Greens function extraction involves cross correlation and integration of recorded elds. Here it is used to obtain the Greens function between two receivers from cross correlation of their recordings. The method of extracting the Greens function from cross correlating observations simultaneously made at two different locations can be envisaged as coherent interferometric radiometry. Presently known electromagnetic formulations rely on continuity conditions of the transverse electric and magnetic eld components across nite jump discontinuities in medium parameter values [1][5]. The interaction of electromagnetic wave elds with interfaces across which the wave eld components satisfy jump-average

conditions have been studied for a long time. In the earth sciences this was discussed for elds in the diffusive approximation, for a strongly conductive interface [6]. In the case of a conductive interface charges cannot be built up and the current that is generated at the interface gives rise to a jump in the tangential magnetic eld components, linearly proportional to the tangential electric eld components. The ratio of the electric eld and the magnetic eld jump is the specic boundary resistivity. Interface conditions of this type are known as impedance boundary conditions. Impedance boundary conditions are used for modern scattering computations of large and curved objects, and corrugated surfaces [7][13], of half planes [14], in biomedical engineering [15] and for bianisotropic impedance boundaries with jump conditions [16], [17]. In this work general linear electromagnetic media are incorporated, which are represented by a full constitutive matrix [18]. An early example is the dual polarized ring laser [19] that can be used in downhole formation testing in oil exploration. For bianisotropic media with continuity boundary conditions, for the transverse components of the electric and magnetic elds, the interferometric relations are derived in [20]. Bianisotropic media are of growing importance in the fabrication of metamaterials [21] that can be used for creating impedance boundary conditions [22], [23]. In this paper, the Greens function retrieval is formulated for interfaces with general linear impedance boundary conditions. General symmetry properties of the interface material parameter matrix is given for reciprocal and nonreciprocal lossless impedance boundary conditions that lead to Greens function extraction from contributions of sources located on a boundary. We show numerical examples in 2-D to illustrate the possibilities and restrictions. II. RECIPROCITY The theory is developed in six-vector notation [24] and three , are introduced as unitary six-matrices, (1) while the matrix is obtained as and the unit matrix is used for the 3 3 and 6 6 unit matrices, but no confusion occurs. Note that , and , which relations are used many times. The macroscopic space-time electromagnetic eld is determined by , the magnetic eld , the electhe electric eld tric and magnetic ux densities , and the external source volume densities of electric and magnetic currents, , respectively. The time-Fourier transform of a space-time dependent quantity is dened as

Manuscript received March 02, 2011; manuscript revised July 21, 2011; accepted July 30, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the Department of Geotechnology, Delft University of Technology (TU Delft), Stevinweg 1, 2628 CN Delft, The Netherlands (e-mail: e.c.slob@tudelft.nl; c.p.a.wapenaar@tudelft.nl). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167942

0018-926X/$26.00 2011 IEEE

352

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

, where j is the imaginary unit and denotes angular frequency. The frequency domain constitutive reand lations are given by where electric permittivity and magnetic permeability tensors denote the magneto-electric are given by and , while tensors. The effects of moving media and all possible timerelaxation mechanisms are incorporated in the frequency dependent complex valued material tensors. Maxwells equations , where the eld vector is given by read and the superscript denotes transposi) is the source vector, while tion, is the matrix of spatial differential operators given by (2) The material matrix is dened as (3) We have the following symmetry property for the derivative ma. trix, Macroscopic impedance and admittance boundary conditions dene linear relations between the electric and magnetic eld components tangential to an interface [17]. General jump-average conditions are adopted in the frequency domain [25] (4) with

Fig. 1. Conguration for the reciprocity theorem internal interfaces with impedance boundary conditions.

such that

and then (4) can be written as , with . A reciprocity theorem relates two states, labeled and , that can be nonidentical everywhere. Reciprocity of the timeconvolution type is applied to a bounded spatial domain , and outer boundary with outward pointing unit normal vector , and internal interfaces , where the boundary conditions of (4) apply, see Fig. 1. With the above denitions, the theorem reads [25]

(8) (5) with dened as but with the components of the spatial derivative, , replaced by the components of the unit vector , 2, 3. This formulation normal to the interface, , allows for approximations of thin high contrast layers [6], but they are exact for interfaces with perfect conductors, either electric, magnetic or electromagnetic [23] and generalized soft-and-hard (SHS) layers [22] or other conceivable interfaces with impedance boundary conditions (IBC) [26]. The jump , and average across the interface are represented by , respectively; hence (6) (7) where is chosen at the interface. The interface permittivity and and , while permeability are the magneto-electric interface parameter tensors are given by and . First, the elds at both sides of the interface are separated to avoid sign problems with the matrices containing unit vectors normal to the interface. Let the two sides of the interface be denoted 1 and 2 and we use to represent the unit vector, normal to the interface, see Fig. 1, where is dened similar to , but with replaced by , and , and has been used. Note that for the integral over the nonperfect and , and the matrix are all lointerfaces the elds cated on the 1-side. Equation (8) is the general representation for two independent electromagnetic states in bianisotropic media. The sources and source locations as well as the media in the two states can be completely different. The rst integral in the right-hand side of (8) represents the boundary integral over the outer boundary, where continuity conditions apply. The third integral in the right-hand side of (8) represents the boundary integral over all internal interfaces, where the impedance boundary conditions of (4) apply. This integral and interfaces satisfying these vanishes when conditions are each others adjoint. The adjoint of an interface and it is related to the interface impedance is denoted as impedance through . This implies that is given by the adjoint of the interface parameter matrix , and

(9)

SLOB AND WAPENAAR: GREENS FUNCTION EXTRACTION FOR INTERFACES WITH IMPEDANCE BOUNDARY CONDITIONS

353

When these conditions hold for one and the same interface, it is called a self-adjoint, or reciprocal, interface and we have and . The volume integral in the right-hand side of (8) vanishes , which implies that , , when and . This situation is discussed in [20] and here only the necessary results are used. The adjoint and its relation to the material matrix medium is denoted . If these conditions hold in one is given by and the same medium, the medium is called self-adjoint or re. ciprocal and we have A. Source-Receiver Reciprocity The Greens function expression of source-receiver reciprocity is obtained by taking state B as the adjoint of state A, and hence the second and third integrals in the right-hand side of (8) vanish. The material matrices are given by and , while the interface impedance matrices are and . The 6 1 source vectors are replaced by the 6 6 unit strength point source matrices , where is the identity matrix. The eld vector is correspondingly replaced by the 6 6 Greens matrix , while the eld vector is replaced by . In the Greens the adjoint Greens matrix matrix each column represents the Greens functions for all the electric and magnetic eld components for a single source type and direction, while each row represents a single eld type and component for all source types and directions. If we take and inside and assume that outside some sphere with nite radius the medium is isotropic and homogeneous, then the boundary integral also vanishes, leaving the source-receiver reciprocity relation as (10) which expresses the equality of a measurement in a certain medium to an other measurement in an adjoint medium, with interchanged source and receiver type, vector component and accounts for possible sign changes location. The matrix upon interchanging source and receiver. III. POWER BALANCE The here derived interferometric relation originates in the correlation-type reciprocity theorem, which reads [27][29]

has been used and the superscript dewhere notes matrix transposition and complex conjugation. If in (11) the interface is lossless, which can occur either when the interfaces in both states are lossless or when the interface in one state dissipates energy from the eld and in the other state it delivers the same amount of energy to the eld in . Then it is found that , which case and , . The parameter matrix of the adjoint interface is related to that of the interface as (12) (13) where the superscript denotes complex conjugation. , the interface is lossless, but not necesWhen . When sarily reciprocal. The medium is lossless when and are the same, and both the interface and the states medium are lossless the last two integrals of (11) vanish. A. Correlation Type Greens Matrix Representation Equation (11) is used to derive a representation of the Greens matrix in terms of cross correlations. Point source matrices and Greens matrices replace the source and eld vectors. The points and are chosen in and both states have the same inand terface and medium parameters . With these choices the correlation type Greens matrix representation is given by

(14)

where the contrast functions are both Hermitian and given by (15) (16) It is noted that now both states exist in one and the same medium. Furthermore, these states can occur simultaneously, but that is not mandatory. No assumptions have been made about the internal interfaces and material matrices, and . Obviously, at each interface where , the interface is lossless and the third integral in the right-hand side of (14) vanishes. Only the anti-Hermitian part of remains in the representation, which is the part accounting for energy dissipation. Equation (14) is a general representation of the electromagnetic Greens matrix for arbitrary bianisotropic media, including arbitrary IBC interfaces. Equation (14) represents the Greens

(11)

354

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

functions between and obtained from integral contribu, at the tions of received Greens functions at the boundary and in the volume , in a heterogeneous internal interfaces bianisotropic medium with piecewise continuous interfaces. IV. GREENS FUNCTION RETRIEVAL Equation (11) is used to derive an interferometric representation of the Greens matrix in terms of cross correlations. The and are chosen in , but the adjoint states are points taken for both and , hence again both states occur in one and and the same medium but now with . Point source matrices are used and, as a consequence of the medium and interface impedance parameters, adjoint Greens matrices replace the source and eld vectors. Equation (10) is used together with the symmetry relations , and , and for and . Transposing both sides for of the resulting equation yields

boundary of . In applications of remote sensing without a source, the most interesting situation is the lossless variant with , where the interfaces can be dispersive, but dissi. pate no energy, and with similar medium properties A. Lossless Interfaces For lossless interfaces and media the matrices satisfy and , leading to and . Note , and that this occurs when for the interface parameter matrices and the same conditions apply for the material parameter matrices. These choices allow for nonreciprocal media and impedance interfaces. Then (17) directly reduces to

(19) For lossless IBC interfaces in lossless media the Greens maand is obtained from cross correlations of trix between recordings from responses to independent impulsive sources on only. When the data, generated by one of the sources at the boundary and recorded by two receivers, are cross correlated with each other and summed over all the source contributions at the boundary, the result is a pulse echo experiment with one of the receivers acting as a source and in absence of the original sources at the boundary, as sketched in Fig. 2. To make must be diag(19) suited for uncorrelated noise sources, onalized. This involves the separation of contributions from the sources for inward and outward traveling waves and the procedure is outlined in [28]. A suitable diagonalization procedure is available when the noise sources lie on a boundary in a lossless isotropic medium [30]. Then, e.g., the electric subset of (19) is in the time domain given by

(17)

is the same as in (16), while where the contrast function the impedance boundary contrast matrix is now given by (18) No assumptions have been made about the internal interfaces and material matrices, and , other than that their adjoint exist. Equation (17) is a general representation of the and located electromagnetic Greens matrix, between in the same medium, obtained from integral contributions , at the internal generalized from sources at the boundary, impedance interfaces, , and inside the volume, , in an arbitrary heterogeneous bianisotropic medium, with IBC interfaces. It is valid for reciprocal and nonreciprocal IBC interfaces and media. When the media and IBC interfaces dissipate no energy, the full Greens matrix can be obtained from sources on the boundary only. Hence, even for an arbitrary heterogeneous bianisotropic medium with IBC interfaces, absence of energy dissipation is a sufcient condition for obtaining the Greens function from the cross correlation of two recordings from sources on a closed boundary only. In the time domain for the Greens matrix is causal; hence, , and the time reversed Greens function is time reversed causal; hence, for . For this reason or can be easily retrieved from the left-hand side of (17), , by taking the causal or time-reversed causal part, respectively. In general, the application of (17) requires independent measurements of sources at all points in the domain and at the

(20) where denotes the autocorrelation of the noise sources denotes a spatial ensemble average. This expression is and similar to the results obtained for anisotropic and bianisotropic media [3], [20]. In (20) time convolution is indicated by , the Greens function corresponding to the electric eld generated by , while dean electric current source is represented by notes the observed electric eld vector due to uncorrelated noise and the sources. For random white noise sources, Greens function is retrieved. V. ELECTRIC FIELD GREENS FUNCTION RETRIEVAL We now reduce (17) to retrieve the Greens function for the electric eld generated by an electric current source. We rst write the interface impedance matrix as (21)

SLOB AND WAPENAAR: GREENS FUNCTION EXTRACTION FOR INTERFACES WITH IMPEDANCE BOUNDARY CONDITIONS

355

Fig. 2. Right graph depicts the situation for one source; when the data recorded at x and x is cross correlated for each source at the boundary and summed over all sources as expressed in the right-hand side of equation (19), the result is an experiment in absence of sources at the boundary, but where one of the original receivers acts as a source as depicted in the left graph as an illustration of the rst Greens function in the left-hand side of equation (19).

where the subscripts denote the Cartesian tensor components for and . The of each sub-matrix matrix for the impedance condition is (22) with the sub-matrices given by (23) (24) (25) (26) We also introduce two new directional tensors , is equal to in subscript notation, and where , being equal to in subscript notation, and . These newly introduced tensors are used in (11) to write it out in Cartesian tensor components for the electric eld. We keep electric current sources in both states, and but take the magnetic current sources to be zero . For a lossless medium with interfaces satisfying impedance boundary conditions, we nd

, where is the inverse . of the magnetic permeability tensor, given by In case the interfaces have impedance boundary conditions that do not dissipate energy all boundary condition tensors vanish, , and only the integral over the outer boundary remains. In case the interfaces are perfect electric (PEC), or perfect magnetic (PMC) conductors, the integrals vanish because the tangential electric and/or magnetic elds vanish. In those situations the perfect conducting boundary can be taken , because such interface beas part of the outer boundary haves as a perfect reector with possible polarization rotation. When small losses occur, e.g., when the conducting boundaries are not perfect, but still good conductors [31], a small error will be introduced when the boundary integrals are ignored, similar to ignoring the volume integral when the background medium is slightly dissipative [30]. Equation (27) is the starting point to formulate the numerical examples. Even when the impedance conditions involve energy dissipation at the interface and sources on the interface are required to retrieve correct Greens functions, no numerical singularities occur, unless a receiver is located at the interface with an impedance condition. In such situations it is not useful to retrieve Greens functions for a source and/or receiver at the impedance boundary, because then the easiest way is direct computation. We are interested in remote sensing without having a source close to the impedance boundary and we will show numerical examples for two difference impedance conditions. VI. EXAMPLES The most interesting situations occur when the propagating medium is lossless. We will investigate two different types of IBC interfaces in two dimensions. The TE-mode Maxwells equations for a 2-D bianisotropic medium is written as

(28) (27) Correlation reciprocity is used in a bounded 2-D domain , with , in a medium that is adoutward unit normal joint to the medium for which the electric eld scalar Greens function is retrieved.

Because the magnetic current sources are taken zero, the magnetic eld that is still present in the integrals can be replaced by

356

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

The reciprocity theorem of the time correlation type of (27) reduces to

(29) in this 2-D example. Equawhere it is understood that tion (29) is the 2-D equivalent of (27) in which the second and third equations from (28) have been used to write the magnetic eld in terms of the electric eld. This is the diagonalization procedure, mentioned below (19). Notice that we have used the occurs in the third adjoint states for (29), for which reason integral instead of . As a consequence of eliminating the magnetic eld from the equation, it can be seen in (29) that has a nonvafor a lossless nonreciprocal medium, where nishing real part, an extra integral over the boundary remains in the time-correlation type reciprocity theorem compared to an ordinary anisotropic and reciprocal lossless medium. For a , and, bianisotropic reciprocal lossless medium hence, the second integral on the right-hand side of (29) vanishes. In that case (29) reduces to the same form as for an ordinary anisotropic reciprocal lossless medium. We take the medium parameters in the two states and the same and that of the medium adjoint to the actual medium; the electric current source in state to be a line source given and a similar choice for by . The electric eld can be written in terms of the Greens and is substituted in function as (29) to obtain

Equation (31) is the Greens function extraction relation for the 2-D electric eld in a lossless bianisotropic medium with interfaces satisfying general lossless or perfect conducting impedance boundary conditions. In the high-frequency approximation, where , the rst integral in the right-hand side of (31) becomes equal to the second integral with opposite sign, while the third integral is approximately cancelled, such that only the term remains and we use as the wave with propagation velocity ignoring the bianisotropy effects. Under these approximations (31) reduces to

(32) which is the desired relation to extract the Greens function between and from the product of the complex conjugate of and sources on the Greens function between a receiver in the boundary and the Greens function between a receiver in and sources on the boundary . The effect of making these approximations are investigated by comparing results from (31) and (32). A. Lossless and PEC Impedance Boundaries For the numerical examples it sufces to investigate a simple 2-D conguration. We choose a homogeneous bianisotropic background medium in which one planar boundary occurs that . satises an impedance boundary condition, located at We place two outer boundaries such that , is dened by and is dened by with . The points and are located at the depth level and and . The medium parameters are given by , , and . For this on the boundaries choice of bianisotropy parameters and , where is given below (30) and is dened just above (32). The actual propagating wave velocity in the , in the negative positive -direction is -direction it is , while in the positive and and negative -directions they are . The boundary parameters are given by , , . We use the second derivative of a Gaussian as the source pulse, with a central frequency of 250 MHz as spectral bandwidth lter; we and . place 1024 electric dipoles separated by 5 cm on Because we model propagating waves, the usual discretization conditions apply to avoid spatial and temporal aliasing. We show results for a horizontal distance of 1 m between the two receivers. The result from summing the correlated signals reand from sources at both interfaces according ceived at to (31) is shown in Fig. 3(a), according to (31) but ignoring the third integral on the left-hand side, and according to (32) in Fig. 3(c). The left-hand sides of (31) and (32) can be computed as exact results in the time-domain and these are shown in and a receiver in for solid black lines for a source in positive times and for a source and a receiver in for

(30) . Exploiting the fact that all the Greens functions in (30) belonging to the adjoint medium are replaced with the Greens function in the actual medium and (30) is rewritten as where

(31)

SLOB AND WAPENAAR: GREENS FUNCTION EXTRACTION FOR INTERFACES WITH IMPEDANCE BOUNDARY CONDITIONS

357

negative times. The right-hand sides of (31) and (32) describe how the retrieved results should be obtained and the numerical results are shown in red dashed lines. The nite length of the source boundaries leads to break-off errors that can be seen in the direct waves that travel between source and receiver in all the plots in Fig. 3. Still the retrieved signal is quite accurate in amplitude and perfect in time. The reection event is very accurately retrieved, because the sources that are stationary for this reection event are relatively close to the receivers. To investigate the contribution from the third integral in the right-hand side of (31) we plot the result of the rst two integrals in the right-hand side of (31) in Fig. 3(b), where the direct wave in positive times has a much too low amplitude, the direct wave at negative times has a much too high amplitude, while the reected waves, at both positive and negative times, from the impedance boundary have approximately half the expected amplitude. Implementing the high-frequency approximation is expected to give larger errors in the direct waves than in the reected events, because in the approximation it is assumed that the wave eld that contributes to the retrieved events, leaves the source boundary perpendicular to this boundary. For receivers at an angle of 45 degrees with the vertical axis the waves that are stationary for retrieving the direct eld leave the boundary at 45 degrees and we expect an error in the amplitude as can be seen in Fig. 3(c), where it can of also be seen that the reection from the impedance boundary has no visible error. The asymmetry in the amplitudes of the retrieved direct waves at positive and negative times is because the propagating medium is bianisotropic. In Fig. 4 similar results are shown for a PEC boundary. For a PEC boundary the sources are only necessary on boundary , because the boundary lies below the PEC boundary, which is a perfect reector and waves sent from sources on the do not reach the points and . The retrieved boundary result using (31) is excellent (Fig. 4(a), and the PEC reection retrieved from (32) is very good with a slight amplitude error (Fig. 4(b). The direct events at negative times are now also reand this leads to interactions of trieved from sources on waves that leave the boundary in positive -direction, reect off the PEC boundary, traveling in negative -direction and and then . These interacpass through both receivers in tion are different than in the example with nonperfect reecting impedance boundary where the direct waves at positive times . For this reason the direct are retrieved from sources on events at positive times in Fig. 4(a) and (b) are slightly different from the direct events at positive times in Fig. 3(a) and (c), respectively. VII. CONCLUSION The matrix representation for Greens function retrieval in general bianisotropic media with IBC interfaces holds for nonreciprocal and dissipative media. It applies to natural and engineered media. The validity of the Greens function retrieval by cross correlation of two noise eld recordings has been extended to include media with interfaces that are characterized by jump-average conditions. The condition of the media and IBC interfaces being lossless was shown to be a sufcient condition to create new data from measured data due to noise sources on
Fig. 3. Exact results for a boundary satisfying lossless bianisotropic boundary conditions in a bianisotropic medium are shown in solid black lines for a source in x and a receiver in x for positive times and for a source x and a receiver in x for negative times, which is equal to the time-domain equivalent of the left-hand sides of equations (31) and (32); retrieved results are shown in red dashed lines, which are obtained according to the right-hand side of a) equation (31), b) same as a) but ignoring the third integral, and c) (32).

a remote closed boundary only. The result may nd applications in a wide variety of elds, ranging from microwave to optical regimes, and the most interesting application is to obtain pulse-echo data from cross correlations of noise observations

358

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

ACKNOWLEDGMENT This work is part of the research program of the Netherlands research center for Integrated Solid Earth Science (ISES).

REFERENCES
[1] E. Slob and K. Wapenaar, Electromagnetic Greens functions retrieval by cross-correlation and cross-convolution in media with losses, Geophys. Rev. Lett., vol. 34, no. 5, p. L05307, 2007. [2] L. B. Liu and K. He, Wave interferometry applied to borehole radar: Virtual multioffset reection monitoring, IEEE Trans. Geosci. Remote Sens., vol. 45, no. 8, pp. 25542559, Aug. 2007. [3] E. Slob and K. Wapenaar, GPR without a source: Cross-correlation and cross-convolution methods, IEEE Trans. Geosci. Remote Sens., vol. 45, no. 8, pp. 25012510, Aug. 2007. [4] E. Slob and K. Wapenaar, Practical representations of electromagnetic interferometry for GPR applications: A tutorial, Near Surface Geophys., vol. 6, no. 6, pp. 391402, Dec. 2008. [5] E. Slob, Interferometry by deconvolution of multi-component multioffset GPR data, IEEE Trans. Geosci. Remote Sens., vol. 47, no. 3, pp. 828838, Mar. 2009. [6] A. Kaufman and G. Keller, Frequency and transient soundings, in Methods Geochem. Geophys.. New York: Elsevier, 1983, vol. 16. [7] Z. G. Qian, W. C. Chew, and R. Suaya, Generalized impedance boundary condition for conductor modeling in surface integral equation, IEEE Trans. Microw. Theory Tech., vol. 55, no. 11, pp. 23542364, Nov. 2007. [8] O. zdemir, I. Akduman, A. Yapar, and L. Crocco, Higher order inhomogeneous impedance boundary conditions for perfectly conducting objects, IEEE Trans. Geosci. Remote Sens., vol. 45, no. 5, pt. 1, pp. 12911297, May 2007. [9] I. Hnninen and K. Nikoskinen, Implementation of method of moments for numerical analysis of corrugated surfaces with impedance boundary condition, IEEE Trans. Antennas Propagat., vol. 56, no. 1, pp. 278281, Jan. 2008. [10] M. F. Ctedra, C. Delgado, and I. G. Diego, New physical optics approach for an efcient treatment of multiple bounces in curved bodies dened by an impedance boundary condition, IEEE Trans. Antennas Propagat., vol. 56, no. 3, pp. 728736, Mar. 2008. [11] J. Komijani and J. Rashed-Mohassel, Symmetrical properties of dyadic Greens functions for mixed boundary conditions and integral representations of the electric elds for problems involving a PEMC, IEEE Trans. Antennas Propagat., vol. 57, no. 10, pt. Part 2, pp. 31993204, Oct. 2009. [12] S. Yuferev and L. Di Rienzo, Surface impedance boundary conditions in terms of various formalisms, IEEE Trans. Magn., vol. 46, no. 9, pp. 36173628, Sep. 2010. [13] P. Yla-Oijala, S. P. Kiminki, and S. Jarvenpaa, Solving IBC-CFIE with dual basis functions, IEEE Trans. Antennas Propagat., vol. 58, no. 12, pp. 39974004, Dec. 2010. [14] Y. Z. Umul and U. Yalcin, The effect of impedance boundary conditions on the potential function of the boundary diffraction wave theory, Opt. Commun., vol. 281, no. 1, pp. 2327, Jan. 2008. [15] R. B. Clipp and B. N. Steele, Impedance boundary conditions for the pulmonary vasculature including the effects of geometry, compliance, and respiration, IEEE Trans. Biomed. Eng., vol. 56, no. 3, pp. 862870, Mar. 2009. [16] D. Khaliullin and S. Tretyakov, Generalized impedance-type boundary conditions for thin planar layers of various media (review), J. Commun. Technol. Electron., vol. 43, no. 1, pp. 1225, Jan 1998. [17] V. Galdi and I. M. Pinto, Derivation of higher-order impedance boundary conditions for stratied coatings composed of inhomogeneous-dielectric and homogeneous-bianisotropic layers, Radio Sci., vol. 35, pp. 287303, 2000. [18] J. Kong, Theorems of bianisotropic media, Proc. IEEE, vol. 60, no. 9, pp. 10361046, Sep. 1972. [19] W. W. Chow, J. Gea-Banacloche, L. M. Pedrotti, V. E. Sanders, W. Schleich, and M. O. Scully, The ring laser gyro, Rev. Modern Phys., vol. 57, no. 1, pp. 61104, Jan. 1985. [20] E. Slob and K. Wapenaar, Retrieving the Greens function from cross correlation in a bianisotropic medium, Progr. Electromagn. Res.-PIER, vol. 93, pp. 255274, 2009.

Fig. 4. Exact results for a boundary satisfying PEC boundary conditions in a bianisotropic medium are shown in solid black lines for a source in x and a receiver in x for positive times and for a source x and a receiver in x for negative times, which is equal to the time-domain equivalent of the left-hand sides of equations (31) and (32); retrieved results are shown in red dashed lines, which are obtained according to the right-hand sides of equations (31) and (32).

for imaging and characterization of natural and engineered materials. We derived electric eld Greens function extraction from summing correlations of eld recordings from sources on a boundary. In the following 2-D numerical examples we have shown that the high-frequency approximation provides reasonably accurate results for reection events from lossless impedance interfaces as well as PEC boundaries. When the source array has a wide enough aperture, direct waves are well reconstructed even when the line between the two receivers is parallel to the source planes. In congurations where the bianisotropy parameters lead to an extra integral over the source boundary, the high-frequency approximation leads to kinematically correct direct waves and dynamically accurate reconstruction of reection events. This still applies when the bianisotropy parameters cannot be considered small and without needing to know the bianisotropy parameters.

SLOB AND WAPENAAR: GREENS FUNCTION EXTRACTION FOR INTERFACES WITH IMPEDANCE BOUNDARY CONDITIONS

359

[21] A. Sihvola, Metamaterials in electromagnetics, Metamaterials, vol. 1, no. 1, pp. 211, 2007. [22] I. Lindell and A. Sihvola, Generalized soft-and-hard surface, IEEE Trans. Antennas Propagat., vol. 50, no. 7, pp. 926929, Jul. 2002. [23] I. Lindell and A. Sihvola, Realization of the PEMC boundary, IEEE Trans. Antennas Propagat., vol. 53, no. 9, pp. 30123018, Sep. 2005. [24] I. Lindell, A. Sihvola, and K. Suchy, Six-vector formalism in electromagnetics of bi-anisotropic media, J. Electromagn. Waves Appl., vol. 9, no. 78, pp. 887903, 1995. [25] K. Wapenaar, E. Slob, and J. Fokkema, Reciprocity and power balance for piecewise continuous media with imperfect interfaces, J. Geophys. Res. B, vol. 109, no. 10, p. B10301, 2004. [26] D. Hoppe and Y. Rahmat-Samii, Impedance Boundary Conditions in Electromagnetics. Bristol, PA: Taylor and Francis, 1995. [27] V. Rumsey, Reaction concept in electromagnetic theory, Phys. Rev., vol. 94, pp. 14831491, 1954. [28] K. Wapenaar, E. Slob, and R. Snieder, Unied Greens function retrieval by cross correlation, Phys. Rev. Lett., vol. 97, no. 23, pp. 234301-1234301-4, 2006.

[29] K. Wapenaar, General wave eld representations for seismic modeling and inversion, Geophysics, vol. 72, no. 5, pp. SM5SM17, 2007. [30] E. Slob, D. Draganov, and K. Wapenaar, Interferometric electromagnetic Greens functions representations using propagation invariants, Geophys. J. Int., vol. 169, no. 1, pp. 6080, 2007. [31] I. V. Lindell and A. H. Sihvola, Losses in the PEMC boundary, IEEE Trans. Antennas Propagat., vol. 54, no. 9, pp. 25532558, Sep. 2006.

Evert Slob, photograph and biography not available at the time of publication.

Kees Wapenaar, photograph and biography not available at the time of publication.

360

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Electromagnetic Field of a Horizontal Innitely Long Wire Over the Dielectric-Coated Earth
Yong Jian Zhi, Kai Li, and You Tong Fang
AbstractThe simple analytical formula for the electromagnetic eld radiated by a horizontal innitely long wire over the planar surface of the earth coated with a dielectric layer is derived. It is concluded that the total eld includes four modes: the direct wave, the ideal reected wave or image wave, the lateral wave, and the trapped surface wave. In particular, both the trapped surface wave and the lateral wave are examined in detail. From analysis and computations, it is demonstrated the trapped surface wave, which is determined by the sum of residues of the poles, can be excited efciently. When both the innitely long wire and the observation point are on or near the planar surface, the trapped surface wave is dominant. Index TermsLateral wave, thin innitely long wire, trapped surface wave.

I. INTRODUCTION HE problem of the electromagnetic eld radiated by a horizontal innitely long wire has been intensively investigated for many years because of its extensive use in electric power transmission and telephone communication systems. In the earliest work by Carson [1], the distributed parameters of a quasi-TEM transmission line on the problem were analyzed and calculated specically. Lately, based on Carsons theory, this problem was revisited by other investigators [2]. Early work on this problem was well summarized in the book by Sunde [3]. In the mid 1950s, an exact model equation was derived for the electromagnetic eld of a thin innitely long wire over the planar surface of the earth [4]. In this work, a quasi-static and asymptotic expansion of the exact modal solution is used to treat the wave propagation along the planar surface. At the same time, an exact solution to the problem was also carried out in Russia [5], [6]. Their works were discussed specically by Kuesters et al. [7], [8]. For the solution of this problem, an important progress was made by Wait [9]. In this well-known work, with the extension of Kikuchi addressed in [4], an exact modal equation was presented for the wave propagation of a thin innitely long wire above the earth. In 1977, Wait extended his work to the general case of an arbitrary number of wires above the earth with
Manuscript received October 13, 2010; revised April 09, 2011; accepted June 25, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the National Science Foundation of China under Grant 60971057 and Grant 50877070, and in part by the Innovation fund of State Key Lab of Millimeter Waves under Grant K201010. Y. J. Zhi and Y. T. Fang are with the School of Electrical Engineering, Zhejiang University, Hangzhou 310027, China, and also with the State Key Laboratory of Millimeter Waves, Southeast University, Nanjing 210096, China. K. Li is with the Department of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China, and also with the State Key Laboratory of Millimeter Waves, Southeast University, Nanjing 210096, China (e-mail: kaili@zju.edu.cn). Digital Object Identier 10.1109/TAP.2011.2167917

an arbitrary number of layers [10]. In the past few decades, this problem a wire above earth was visited by many investigators and remarkable progress has been achieved [11][19]. A good summary of the research ndings can be found in the invited paper by Olsen et al. [20]. Corresponding to the problem a wire above earth, the electromagnetic eld radiated by horizontal or vertical electric dipole in stratied media was also investigated widely over a century. A series of papers were published and the details of the research ndings are well summarized in the three classic books [21][23]. In the past decade, the existence or nonexistence of the trapped surface wave for the electromagnetic eld radiated by a dipole over a dielectric-coated lossy underling medium continued to be a controversial subject [24][37]. In early 1990s, the approximate formulas were derived for the electromagnetic elds of vertical electric dipole and horizontal electric dipole in the presence of a three-layered region [23][27]. In their works, it was explained that the trapped surface wave could be excited efciently, but the terms of the trapped surface wave were neglected in their solution [23]. In 1998, Wait [28] wrote a comment on the paper by King and Sandle [27] and claimed that the trapped surface wave, which varies as in the far region, was neglected in their solution. Further works by Mahmoud [30] and Collin [31], [32] supported the conclusions reached by Wait [28]. To clarify the controversies on the trapped surface wave, the old problem was reinvestigated in the past several years [33][38]. From [31][38], it was concluded that the trapped surface wave, which were contributed by the sums of residues of the poles, can be excited efciently by a dipole source in the presence of a three-layered region. Those new works were summarized in the recent book by Li [39]. The new developments in [28][39], naturally, rekindle the interest in the study on the problem a wire above multi-layered earth. earth. It is predicted that the trapped surface wave can also be excited efciently for the electromagnetic eld radiated by a wire in the presence of a multilayered region. In what follows, we will attempt to treat analytically the electromagnetic eld of a horizontal innitely long wire over the earth coated with a dielectric layer and some new results and conclusions will be carried out. II. ELECTROMAGNETIC FIELD OF A HORIZONTAL INFINITELY LONG WIRE OVER THE DIELECTRIC-COATED EARTH A. Statement of the Problem The geometry under consideration is shown in Fig. 1, where a horizontal innitely long wire is located at the height above is the space above the dielectric-coated earth. Region 0

0018-926X/$26.00 2011 IEEE

ZHI et al.: ELECTROMAGNETIC FIELD OF A HORIZONTAL INFINITELY LONG WIRE OVER THE DIELECTRIC-COATED EARTH

361

(8) (9) (10) where (11)

Fig. 1. Geometry of a horizontal innitely long wire at height of dielectric-coated earth. over the

(12)

the dielectric layer characterized by the permeability and uniform permittivity . Region 1 is the dielectric layer characterized by the permeability , relative permittivity , and conductivity . Region 2 is the rest space characterized by the permeability , relative permittivity , and conductivity . The wave numbers in the three-layered regions are (1) (2) Use is made of a time dependence , the integrated formulas of the electromagnetic eld in the air radiated by a horizontal innitely long wire over a dielectric-coated earth are written in the following forms [22]:

(13) (14) (15)

(16) In the above two formulas, and are the modied Bessel functions of orders 0 and 1, repsectively. The third terms in (8)(10) are expressed as follows: (17) (18) (19)

(3)

Next, the main task is to evaluate the above three integrals by using analytical techniques. B. Trapped Surface Wave Considering the three integrals in (17)(19), it is necessary to shift the contours around the branches at , and . The conguration of the poles and the branch cuts is shown in Fig. 2. Let us examine the following pole equation: (20)

(4)

(5) where the factor is expressed as (6)

(7) The rst and second terms in (3)(5) stand for the direct wave and the ideal reected wave or image wave, respectively. The rst and second terms in (3) have been evaluated in the book by Wait [22]. The rst and second terms in (4) and (5) can also be obtained readily by using similar technique. Then, we write

Let , the earth in Region 2 can be characterized by the parameter . It is found that for the perfect conducting case . By examining the pole equation of (20), it is seen that the pole equation has no root when . When , the pole equation of (20) has one root. In general, when the condition for the dielectric layer is satised, the pole equation of (20) has roots. Correspondingly, the integrand in (17)(19) has poles. When the earth is regarded as an imperfect conductor, will be a complex number

362

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

is the root of the pole equation (20) in the perfect where conducting case. When , the root can be obtained readily by solving the simplied pole equation. It is noted that, for the integral in (26), the integral contour is in the fourth quadrant, and , and should satisfy the following condition: (27) By Newtons iteration method from the pole equation (20) or by numerical integration from (26), the roots can be obtained readily. It is noted that the roots are singular poles in the complex plane. By now, the terms of the trapped surface wave can be written in the following forms:
Fig. 2. Conguration of the poles and branch lines.

with its magnitude being less than 1 and phase being from 0 to 45 . With the substitution into (20), the pole equation of (20) can be rewritten in the form

(28)

(29)

(21) Then, we have (22) where (23) (31) From (27)(30), it is seen that the term of the trapped surface wave is contributed by the sums of residues of the poles. It is noted that the wavefront of the trapped surface wave radiated by a of a horizontal innitely long wire over the dielectric-coated earth is always a plane, which is unchanged as the dispersion of the trapped surface wave. So that the amplitude of the trapped surface wave has no dispersion losses. C. Lateral Wave When , besides the poles , the above three integrands have the branch cuts at . As shown in Fig. 2, there are three branch lines at , and . Obviously, the integrand is even function to . It is found that, for the three integrals in (17)(19), the evaluations of the integrations along the branch line are zero. Taking into account that the wave number in the earth has a large positive imaginary part, the integrals in (17)(19) along the branch cut can be neglected. In the next step, we will evaluate the contributions of the integrations along the branch line (32) where (30)

Then, we have

(24) (25)

When integration

, the th root

can be obtained by numerical

(26)

ZHI et al.: ELECTROMAGNETIC FIELD OF A HORIZONTAL INFINITELY LONG WIRE OVER THE DIELECTRIC-COATED EARTH

363

(33) (34) and , the dominant For far-eld region, contribution of the integration in (32) along the branch line comes from the vicinity of . Let , we have (35) On the two sides of the branch line from to . Then, as follows: , the phase of changes and are approximated (46)

D. Final Formulas for the Electromagnetic Field Components Using the above results obtained and the available results of direct wave and ideal reected wave in the book by Wait [22], the nal formulas of the three components , and are written readily. They are

(36) (37) With substitution (35)(37) into (32), we write

(38) where (47)

(39)

(40) (41) (42) The Fresnel integral is dened by (43) where lowing forms: . Then, (32)(34) can be written in the fol-

(48)

(44)

(45)

(49)

364

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 3. Real part of MHz,

versus the thicknesses of the dielectric layer with , and S/m.

Fig. 4. Imaginary part of with MHz,

versus the thicknesses of the dielectric layer , and S/m.

From the above derivations, it is seen that the total eld is composed of direct wave, ideal reected wave or image wave, trapped surface wave, and lateral wave. III. COMPUTATIONS AND CONCLUSION Following the above derivation and analysis, it is concluded that the trapped surface waves are determined by the sum of residues of the poles. When the condition is satised, there are modes of the trapped surface waves to propagate along the air-dielectric boundary. When is always a positive imaginary number, and the trapped-surface-wave terms including the factor decreases exponentially as in the direction. Because the wavefront of the trapped surface wave radiated by a of a horizontal innitely long wire over dielectric-coated earth is always a plane, the amplitude of the trapped surface wave has no dispersion losses. This characteristic is different to that of a dipole source in the presence of a three-layered region. For the case of a dipole source, the wavefront is enlarged as the dispersion of the trapped surface, the amplitude of the trapped surface wave attenuates as in the direction. The wave numbers of the trapped surface waves , being in the range from the wave number of the air to the wave number of the dielectric layer, are determined by the wave numbers , and and the thickness of the dielectric layer. If the dielectric layer is a lossy medium, the wave number of the trapped surface has a positive imaginary part. With MHz, , and S/m, the real and imaginary parts of the rst pole varying as the thickness of the dielectric layer are computed and plotted in Figs. 3 and 4, respectively. With m, MHz, , and S/m, for the electric eld , magnitudes of the total eld, the trapped surface wave, and the DRL waves, which is composed of the direct wave, the ideal reected wave, and the lateral wave, are computed at and and shown in Figs. 5 and 6, respectively. It is shown that the total eld is determined primarily by the trapped
Fig. 5. Magnitude of in V/m due to horizontal innitely long wire versus with MHz, the propagating distances at S/m, and m.

Fig. 6. Magnitude of in V/m due to horizontal innitely long wire versus with MHz, the propagating distances at S/m, and m.

surface wave when both the innite long wire and the observation point are on or near the planar surface of the dielectric-coated earth.

ZHI et al.: ELECTROMAGNETIC FIELD OF A HORIZONTAL INFINITELY LONG WIRE OVER THE DIELECTRIC-COATED EARTH

365

ACKNOWLEDGMENT The authors would like to thank Prof. W.-Y. Pan at China Research Institute of Radio Propagation, Qingdao, Shandong Province, China for many helpful discussions. Also, the authors would like to thank all editors and reviewers for helpful comments and suggestions. REFERENCES
[1] J. R. Carson, Wave propagation in overhead wires with ground return, Bell Syst. Tech. J., vol. 5, pp. 539554, 1926. [2] W. H. Wise, Propagation of HF currents in ground return circuits, in Proc. Inst. Elect. Eng., London, U.K., 1934, vol. 22, pp. 522527. [3] E. D. Sunde, Earth Conduction Effects in Transmission Systems. New York: Dover, 1968. [4] H. Kikuchi, Wave propagation along an innite wire above ground at high frequencies, in Proc. Electrotech. J., 1956, vol. 2, pp. 7378. [5] A. A. Pistolkors, On the theory of a wire near the interface between two media, Dokl. Akad., vol. 86, pp. 941943, 1952, Nauk, USSR, (in Russian). [6] A. A. Pistolkors, On the theory of a wire parallel to the plane interface between two media, in Radiotek., Moscow, Russia, 1953, vol. 8, pp. 818. [7] E. F. Kuester and D. C. Chang, Modal Representation of a Horizontal Wire Above a Finitely Conducting Earth Dept. Elect. Eng., Univ. Colorado, Boulder, CO, Scientic Rep. 21, RADC-TR-76-287, 1976. [8] E. F. Kuester, D. C. Chang, and R. G. Olsen, Modal theory of long horizontal wire structures above the earthPart I: Excitation, Radio Sci., vol. 13, no. 4, pp. 605613, 1978. [9] J. R. Wait, Theory of wave propagation along a thin wire parallel to an interface, Radio Sci., vol. 7, pp. 675679, 1972. [10] J. R. Wait, Excitation of an ensemble of J parallel cables by an external dipole over an M layered ground, AE, vol. 31, pp. 489493, 1977. [11] R. G. Olsen and D. C. Chang, Current induced by a plane wave on a thin innite wire near the earth, IEEE Trans. Antennas Propag., vol. AP22, no. 4, pp. 586589, Jul. 1974. [12] D. C. Chang, J. R. Wait, and R. G. Olsen, Propagation characteristics of a horizontal conducting wire located above or buried in the earth, in Proc. URSI/USNC Symp., Atlanta, GA, 1974. [13] D. C. Chang and R. G. Olsen, Excitation of an innite antenna above a dissipative earth, Radio Sci., vol. 10, pp. 823831, 1975. [14] R. G. Olsen and D. C. Chang, Analysis of semi-innite and nite thin-wire antennas above a dissipative earth, Radio Sci., vol. 12, pp. 699707, 1977. [15] E. F. Kuester and D. C. Chang, Propagating modes along a thin wire located above a grounded dielectric slab, IEEE Trans. Microw. Theory Tech., vol. MTT-25, pp. 10651069, 1977. [16] A. Semlyen, Ground return parameters of transmissions lines an asymptotic analysis for very high frequencies, IEEE Trans. Power App. Syst., vol. 100, no. 3, pp. 10311038, Mar. 1981. [17] R. G. Olsen, E. F. Kuester, and D. C. Chang, Modal theory of long horizontal wire structures above the earthPart 2: Modes, Radio Sci., vol. 13, pp. 615623, 1978. [18] R. G. Olsen and M. Wu, High frequency propagation losses on an open wire transmission line above dissipative earth, IEEE Trans. Broadcast., vol. BR-34, pp. 292300, Jun. 1988. [19] F. Rachidi, C. A. Nucci, and M. Lanoz, Transient analysis of multiconductor lines above lossy ground, IEEE Trans. Power Del., vol. 14, no. 1, pp. 294302, Jan. 1999. [20] R. G. Olsen and J. L. Young, Electromagnetic wave propagation on a thin wire above earth, IEEE Trans. Antennas Propag. , vol. 48, no. 9, pp. 14131419, Sep. 2000. [21] A. Banos Jr., Dipole Radiation in the Presence of a Conducting HalfSpace. Oxford, U.K.: Pergamon Press, 1966. [22] J. R. Wait, Electromagnetic Waves in Stratied Media, 2nd ed. New York: Pergamon Press, 1970. [23] R. W. P. King, M. Owens, and T. T. Wu, Lateral Electromagnetic Waves: Theory and Applications to Communications, Geophysical Exploration, and Remote Sensing. New York: Springer-Verlag, 1992. [24] R. W. P. King and S. S. Sandler, The electromagnetic eld of a vertical electric dipole over the earth or sea, IEEE Trans. Antennas Propag., vol. 42, no. 3, pp. 382389, Mar. 1994. [25] R. W. P. King, The electromagnetic eld of a horizontal electric dipole in the presence of a three-layered region, J. Appl. Phys., vol. 69, no. 12, pp. 79877995, 1991.

Fig. 7. Magnitude of in V/m due to horizontal innitely long wire versus m with MHz, the propagating distances at S/m, and .

Fig. 8. Magnitude of in V/m due to a horizontal innitely long wire versus m with MHz, the propagating distances at S/m, and .

m, MHz, S/m, and , for the electric eld , magnitudes of the total eld, the trapped surface wave, and the DRL waves, are computed and plotted in Fig. 7, respectively. On contrast of Figs. 6 and 7, it is seen that the trapped surface wave is affected signicantly by the conductivity of the earth. When the conductivity of the earth is low, the electric eld attenuates rapidly. Similar to those shown in Figs. 57, the corresponding results are computed at m and shown in Fig. 8. It is concluded that the total eld is determined by the DRL waves when the innite long wire or the observation point is away from the planar surface of the dielectric-coated earth.

With

366

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[26] R. W. P. King, The electromagnetic eld of a horizontal electric dipole in the presence of a three-layered region: Supplement, J. Appl. Phys., vol. 74, no. 8, pp. 48454548, 1993. [27] R. W. P. King and S. S. Sandler, The electromagnetic eld of a vertical electric dipole in the presence of a three-layered region, Radio Sci., vol. 29, no. 1, pp. 97113, 1994. [28] J. R. Wait, Comment on The electromagnetic eld of a vertical electric dipole in the presence of a three-layered region by Ronold, W. P. King and Sheldon S. Sandler, Radio Sci., vol. 33, no. 2, pp. 251253, 1998. [29] R. W. P. King and S. S. Sandler, Reply, Radio Sci., vol. 33, no. 2, pp. 255256, 1998. [30] S. F. Mahmoud, Remarks on The electromagnetic eld of a vertical electric dipole over the earth or sea, IEEE Trans. Antennas Propag., vol. 46, no. 12, pp. 17451946, Dec. 1999. [31] R. E. Collin, Hertzian dipole radiation over a lossy earth or sea: Some early and late 20th century controversies, IEEE Antennas Propag. Mag., vol. 46, no. 2, pp. 6479, Apr. 2004. [32] R. E. Collin, Some observations about the near zone electric eld of a hertzian dipole above a lossy earth, IEEE Trans. Antennas Propag., vol. 52, no. 11, pp. 31333137, Nov. 2004. [33] H. Q. Zhang and W. Y. Pan, Electromagnetic eld of a vertical electric dipole on a perfect conductor coated with a dielectric layer, Radio Sci., vol. 37, no. 4, pp. 13-113-7, 2002. [34] H.-Q. Zhang, K. Li, and W.-Y. Pan, The electromagnetic eld of a vertical dipole on the dielectric-coated imperfect conductor, J. Electromag. Waves Appl., vol. 18, no. 10, pp. 13051320, 2004. [35] H.-Q. Zhang, W.-Y. Pan, K. Li, and K.-X. Shen, Electromagnetic eld for a horizontal electric dipole buried inside a dielectric layer coated high lossy half space, Progress Electromag. Res., vol. PIER 50, pp. 163186, 2005. [36] K. Li and Y. Lu, Electromagnetic eld generated by a horizontal electric dipole near the surface of a planar perfect conductor coated with a uniaxial layer, IEEE Trans. Antennas Propag., vol. 53, no. 10, pp. 31913200, Oct. 2005. [37] L. Liu and K. Li, Radiation from a vertical electric dipole in the presence of a three-layered region, IEEE Trans. Antennas and Propag., vol. 55, no. 12, pp. 34693475, Dec. 2007. [38] Y. H. Xu, W. Ren, L. Liu, and K. Li, Trapped surface wave and lateral wave in the presence of a four-layered region, Progress Electromag. Res., vol. PIER 82, pp. 271285, 2008. [39] K. Li, Electromagnetic Fields in Stratied Media. New York: Springer-Verlag, Mar. 2009. [40] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products. New York: Academic, 1980.

Yong Jian Zhi was born in Shangshui County, Henan, China, on July 15, 1982. He received the B.Sc. degree in electrical engineering and automation from Henan Polytechnic University, in 2006. Currently, he is working towards the Ph.D. degree at the School of Electrical Engineering, Zhejiang University, Hangzhou, China. His research interests include electromagnetic wave theory and applications.

Kai Li was born in Xiao County, Anhui, China, on February 10, 1968. He received the B.Sc. degree in physics from Fuyang Normal University, Anhui, China, in 1990, the M.Sc. degree in radio physics from Xidian University, Xian, Shaanxi, China, in 1994, and the Ph.D. degree in astrophysics from Shaanxi Astronomical Observatory, the Chinese Academy of Sciences, Shaanxi, China, in 1998, respectively. From August 1990 to December 2000, he was with the faculty of China Research Institute Radiowave Propagation (CRIRP). From January 2001 to December 2002, he was a postdoctoral fellow at Information and Communications University (ICU), Taejon, South Korea. From January 2003 to January 2005, he was a research fellow in the School of Electrical and Electric Engineering, Nanyang Technological University (NTU), Singapore. Since January 2005, he was a Professor with the Department of Information Science and Electronic Engineering, Zhejiang University, Hangzhou, China. His current research interests include classic electromagnetic theory and radio wave propagation. Dr. Li is a senior member of Chinese Institute of Electronics (CIE) and a member of Chinese Institute of Space Science (CISS).

You Tong Fang received the B.Sc. degree and Ph.D. degree in electrical engineering from Hebei University of Technology, Hebei, China, in 1984 and 2001, respectively. Since December 2006, he has been a Professor with the School of Electrical Engineering, Zhejiang University, Hangzhou, China. His current research interests include the design of electrical machines and electromagnetic theory.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

367

Decomposable Medium Conditions in Four-Dimensional Representation

Ismo V. Lindell, Life Fellow, IEEE, Luzi Bergamin, and Alberto Favaro
AbstractThe well-known TE/TM decomposition of time-harmonic electromagnetic elds in uniaxial anisotropic media is generalized in terms of four-dimensional differential-form formalism by requiring that the eld two-form satises an orthogonality condition with respect to two given bivectors. Conditions for the electromagnetic medium in which such a decomposition is possible are derived and found to dene three subclasses of media. It is shown that the previously known classes of generalized Q-media and generalized P-media are particular cases of the proposed decomposable media (DCM) associated to a quadratic equation for the medium dyadic. As a novel solution, another class of special decomposable media (SDCM) is dened by a linear dyadic equation. The paper further discusses the properties of medium dyadics and plane-wave propagation in all the identied cases of DCM and SDCM. Index TermsBianisotropic media, differential forms, electromagnetic theory, eld decomposition.

any time-harmonic eld with time dependence uniquely decomposed in two parts,

can be (5)

satisfying (6) This property was probably rst published by Clemmow in 1963 [3]. In the case the decomposition can still be made but it is not unique. Actually, such a medium can be transformed to an isotropic medium through a suitable afne transformation [2]. The TE/TM decomposition in isotropic media has a longer history [4]. TE/TM decomposition in the uniaxial medium can be simply demonstrated for a plane wave. In fact, because the elds of a plane wave in any medium satisfy and , they also satisfy

I. INTRODUCTION A. TE/TM Decomposition

HE most general linear electromagnetic medium (bi-anisotropic medium) can be expressed in terms of four medium dyadics in the three-dimensional Gibbsian vector (engineering) representation as [1], [2] (1)

(7) in the uniaxial medium (2) and (3). Thus, assuming (4), the plane wave must be either a TE wave or a TM wave with respect to the axial direction dened by the unit vector . Since any electromagnetic eld outside the source region can be decomposed in a spectrum of plane waves, each component of which is either a TE wave or a TM wave, the eld can be uniquely decomposed in TE and TM parts. The same principle is valid also for more general decompositions. Thus, it is sufcient to consider the decomposition problem for plane waves only. B. Generalized Decomposition The TE/TM decomposition theory has been generalized to media where the elds can be decomposed to two parts satiswhere and are two given fying either or vectors [5]. Even more generally, [6] analyzes the occurrence of or where are four given vectors. This last decomposition was shown to be possible in bi-anisotropic media with Gibbsian medium dyadics of the form (8) (9) (10)

where the four eld vectors are elements of the vector space . The number of free parameters is in the general case. It is well known that, in a uniaxial anisotropic medium dened by medium dyadics of the form (2) (3) and satisfying (4)
Manuscript received February 05, 2011; revised June 10, 2011; accepted July 15, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was partly supported by the Academy of Finland. I. V. Lindell is with the Department of Radio Science and Engineering, Aalto University, Espoo, Finland (e-mail: ismo.lindell@aalto.). L. Bergamin was with the Department of Radio Science and Engineering, Aalto University, Espoo, Finland and also with KB&P GmbH, Fliederweg 10, 3007 Bern, Switzerland (e-mail: luzi.begamin@kbp.ch). A. Favaro is with the Department of Physics, Imperial College London, London SW7 2AZ, U.K. (e-mail: alberto.favaro04@imperial.ac.uk). Digital Object Identier 10.1109/TAP.2011.2167937

0018-926X/$26.00 2011 IEEE

368

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

(11) where , are arbitrary vectors, , are arbitrary scalars and is an arbitrary dyadic. Media dened by (8)(11) have been called decomposable media [6]. The paper [7] demonstrates that yet another scalar parameter can be added to the denitions of the medium dyadics. In this case the denitions (8)(11) are generalized (after some manipulations) to the form

in terms of a quadrivector . The modied medium dyadic maps two-forms to bivectors, elements of the space . Summary of denitions and operational rules for differential forms, multivectors and dyadics applied in this paper can be found in the appendices of [11][13] and, more extensively, in the book [10]. The medium dyadic can be expanded in four three-dimensional dyadic components by separating terms involving the temporal vector , temporal one-form , or both, as (21)

(12) (13)

which corresponds to the matrix representation (22) One can represent the modied medium dyadic by using Gibbsian medium dyadics as

(14)

(15) C. Four-Dimensional Formalism Remarkably, the conditions (12)(15) of the decomposable medium can be formulated in a compact way by applying the four-dimensional differential-form formalism. The present paper thus starts directly from the four-dimensional denition of decomposable media. It is well known that the Maxwell equations (16) and the medium equation (17) have a simpler appearance in the four-dimensional differentialform representation in comparison with the three-dimensional Gibbsian vector formalism [8][10]. Here the electromagnetic elds are represented by two-forms , , elements of the space , whose expansions in terms of three-dimensional two-forms , and one-forms , are (18) Here, is the source three-form, (19) consisting of three-dimensional charge three-form and current two-form . In the basis of one-forms , , denotes the temporal basis one-form. The medium dyadic maps two-forms to two-forms and corresponds to a 6 6 matrix. It is often simpler to consider the modied medium dyadic dened by (20) which corresponds to the matrix representation

(23)

(24) The matrix is the same as in the expression (1) which involves Gibbsian vectors. The two sets of 3D medium dyadics have the relations [10] (25) (26) (27) The four-dimensional formalism allows simple denition of important classes of electromagnetic media. For example, if the modied medium dyadic can be expressed in terms of the double-wedge square of some dyadic (which need not be symmetric) as (28) the corresponding three-dimensional Gibbsian dyadics satisfy relations of the form [10], [14] (29) for some scalar . Thus, and are multiples of the same dyadic while and may be any antisymmetric dyadics. In [14], a medium dened by (28) was called a Q-medium for brevity. Such a medium is known to have the property of being non-birefringent to propagating waves. Thus, media in this class can be conceived as generalizations of isotropic media. Also, the class known as transformation media [16][18] largely coincides with the class of Q-media with a symmetric dyadic .

LINDELL et al.: DECOMPOSABLE MEDIUM CONDITIONS IN FOUR-DIMENSIONAL REPRESENTATION

369

Generalizing the denition (29) by adding a term (30) where , are two bivectors, leads to the class of generalized Q-media, dened in [19]. One can show that, for this kind of media, the three-dimensional medium dyadics must be of the form (8)(11), i.e., any generalized Q-medium is actually a decomposable medium. However, this cannot be inverted, because the more general set of conditions (12)(15) for cannot be achieved with medium dyadics of the form (30). At this stage it is not obvious how to generalize (30) to correspond to the conditions (12)(15). D. Hehl-Obukhov Decomposition In many applications a decomposition of the medium dyadic based on its symmetry properties is often useful. As shown by Hehl and Obukhov [9] (following the symmetry properties discussed by Post [20]), the most general medium dyadic can be uniquely decomposed in three components as [9], (31)

in any medium. Dening the dot product between two twoforms , or two bivectors , as (36) (37) the four-form conditions (35) can be expressed as the scalar conditions

(38) Thus, for any medium dyadic , the two-form wave satises a condition of the form of any plane

(39) for arbitrary scalars , , . For the modied medium dyadic this condition becomes (40)

called principal (1), skewon (2) and axion (3) parts of

. The because we have

axion part is a multiple of the unit dyadic mapping any two-form to itself while the other two parts are trace free. The skewon part is dened so that the corresponding modied medium dyadic is antisymmetric, while is trace free and is symmetric. the principal part The decomposition (31) motivates an intuitive nomenclature; for example a medium dened by is called a principal is called a skewon-axion medium and one with medium. II. DECOMPOSABLE MEDIUM (DCM) A. Plane-Wave Conditions Let us now formulate the decomposition property in terms of four-dimensional quantities. Assuming plane-wave elds (32) the Maxwell (16) for the wave one-form amplitude two-forms , become and electromagnetic

(41)

B. Condition for the Medium Dyadic Let us now assume that, given two bivectors , , the medium has the property that any plane wave satises either or . Such waves can be respectively called A-waves and B-waves and the medium, in analogy with the medium dened by (8)(11), can be called by the general name decomposable medium. Thus, any plane wave in such a medium is required to satisfy (42) for two given bivectors , . Following [6] let us now dene the class of decomposable media by combining (40) and (42) and requiring that the following condition be satised for all two-forms : (43) Since the left-hand side is zero for all media, this warrants that (42) is satised when the medium is such that (43) is satised. Requiring that this be valid for any two-forms implies that the symmetric parts of the dyadics in brackets on both sides must be the same. Redening the coefcients we can write the condition in the form (44)

(33) These imply the following representations in terms of potential one-forms , , (34) Thus, the electromagnetic two-forms of any plane wave satisfy the orthogonality conditions (35)

370

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Although (44) is obviously enough to dene a large class of media, it is not enough for claiming that this class covers all media for which the decomposition condition (42) is satised. The latter question must be left for the topic of further research. To nd solutions for the condition (44), we must separate two cases: either or . The case requires solving a symmetric quadratic dyadic equation and the corresponding class of media can be called that of (proper) decomposable media or DCM for brevity. In the case the quadratic dyadic equation is reduced to one of the rst order. It will require separate consideration and actually denes a distinct class of media called that of special decomposable media or SDCM. III. SOLUTIONS TO THE MEDIUM CONDITIONS A. DCM Assuming in (44) we can set without losing generality. In this case the DCM condition (44) can be expressed in compact form as (45) by dening (46) (47) (48) (49) The condition (45) is a quadratic dyadic equation, whose solutions are derived in the Appendix (cf. (122)). Accordingly, two subclasses of DCM are obtained. The rst class assumes that there exist a dyadic such that is a multiple of . The second class assumes that there exist a dyadic so that is a multiple of . Let us consider these two cases separately and respectively call them QDCM and PDCM. This nomenclature is chosen in the light of the obvious relation with Q-media [14] and P-media [15]. Redening , the QDCM solution of (45) as obtained from (46)(49) must be of the general form (50) or (51) for some normalized dyadic , bivectors , and scalars . Thus, the denition (48) is more explicitly ,

dened by (51) satisfy the decomposition property (42) (see Section IV-A). The three-dimensional Gibbsian medium dyadics corresponding to the general QDCM can be expressed in the general form (12)(15). In fact, adding an axion term with to the medium dyadic of the generalized Q-medium (30) analyzed in [19], it can be shown to correspond to the more general class (50) of decomposable media (12)(15). The second possibility in (45), that of PDCM, yields the following forms for the decomposable medium, (53) or (54) for some normalized dyadic , bivectors . The denition (48) in this case is , and scalars ,

(55) , one can easily solve (55) for in terms Setting again of given and . For the medium coincides with one called the generalized P-medium whose basic properties have been studied in in [15]. Thus, the PDCM solution coincides with the generalized P-medium extended by an arbitrary axion component. Based on the expansions of the dyadic and the bivector product in 3D components as (56) (57) the 3D components of the medium dyadic in the PDCM take the form [15] expressed as (21)

(58) (59) (60) (61) have a quite restricted One can note that the dyadics and form. In the case they actually do not have an inverse, which is in contrast to the QDCM case. Actually, P-media and Q-media can be transformed to one another through Hodge duality [15]. The same property is also valid for the generalized Q- and P-media. B. SDCM Let us now consider the special case of the condition (44) simplied by and , which yields the following rst-order dyadic equation, (62)

(52) Given the bivectors and one can easily solve (52) for . It is also easy to verify that plane-wave elds in a medium

LINDELL et al.: DECOMPOSABLE MEDIUM CONDITIONS IN FOUR-DIMENSIONAL REPRESENTATION

371

Equation (62) can be interpreted so that the symmetric part of must be a multiple of . Redening the dyadic , the modied medium dyadic must thus be of the form (63) is an arbitrary antisymmetric dyadic. It is where known that any antisymmetric dyadic mapping two-forms to bivectors can be expressed in terms of a trace-free dyadic as [21] (64) It is now easy to verify that a medium dened by (63) satises the decomposition condition (42). In fact, any plane wave in such a medium satises

where

is the spatial unit dyadic, we have

(70) Further, we can expand (71) and the one-forms where the vectors , In terms of these we can write , are spatial.

(72) Inserting the expansions in (67) and equating with (21) we can identify one set of three-dimensional medium dyadics as (65) Alternatively, we can replace the solution (63) by (66) which corresponds to (67) Without losing the generality, the bivectors can be assumed to satisfy , whence the last one of the three terms in (67) is trace free. In this case the three terms correspond to the respective axion, skewon and principal parts of the medium dyadic . While the axion and skewon parts may be arbitrary, the principal part is restricted to be of the simple form as dened by the two bivectors and . Since the principal part is not complete, i.e., it does not have an inverse, some trouble in interpreting the medium in terms of three-dimensional medium dyadics may be expected. If is chosen in (66) and (67), SDCM reduces to a simplied class of media, previously called that of doubly-skew media [22]. C. 3D Expansions for SDCM Because SDCM denes a novel class of decomposable media, it is interesting to nd its denition in terms of three-dimensional medium parameters. Let us expand the trace-free dyadic of (64) as (68) is a spatial dyadic, where one-form. Applying a spatial vector and a spatial This implies (79) in the above 3D representations. Limiting to the most general vanishes, we case, i.e., that none of the quantities , , , must have and for some scalar . Thus, in this case we can write (80) (69)
1Here

(73) (74) (75) (76) It can be seen that the dyadics and may have arbitrary antisymmetric parts while their symmetric parts are incomplete consisting of only two dyads. The decomposition (42) can now be written for the 3D elds as1 (77) The 3D Gibbsian-dyadic representation corresponding to (73)(76) can be obtained through the transformation rules (25)(27). It turns out that their analytic expressions become quite extensive and they are omitted here. D. Example of SDCM As an example, let us consider SDCM with vanishing magnetoelectric parameters: (78)

stands for the magnetic two-form and not for the bivector.

372

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

(81) and the corresponding Gibbsian dyadics become (82) (83) The medium dened by these expressions is characterized by both electric and magnetic gyrotropy. For example, the permittivity dyadic (83) can be approximately realized by magnetoplasma in a high static magnetic eld for low frequencies [23].

and applying the rules valid for normalized

, (91) (92)

Equation (90) can be expanded as

(93) Equation (87) yields a scalar dispersion relation which splits in two quadratic equations as (94) (95) Actually, (94) corresponds to the A-wave and (95) to the B-wave as will be shown in Section V. B. PDCM

IV. DISPERSION EQUATIONS To verify the decomposition of plane waves let us derive the dispersion equation for the general plane wave in all of the previous medium cases. The equation for the potential one-form can be obtained by starting from (33), (34) which yield (84) This can be expressed as (85) where is the dispersion dyadic. The axion part of does not contribute because (86)

Neglecting again the axion term, PDCM coincides with the generalized P-medium whose dispersion equation was derived in [15]. Omitting the details, quite similar to those of the QDCM, the dispersion equation can be split in two equations which are of the form (96)

and we can omit it in all medium cases. One may note that (85) implies (87) which will be applied in the sequel. A. QDCM In the case of QDCM the dispersion dyadic equals that of the generalized Q-medium, (88) This case has been analyzed in [19], but let us retrace the steps for convenience. For simplicity, let us dene the vectors (89) Expanding (90) . The dyadic is simple since it satises is antisymmetric and it can be shown to satisfy where the bivector dened by

(97)

C. SDCM For SDCM the dispersion dyadic equals (98) Expressing the general antisymmetric dyadic as in (64), where may be any trace-free dyadic, we can write (99)

(100)

(101)

LINDELL et al.: DECOMPOSABLE MEDIUM CONDITIONS IN FOUR-DIMENSIONAL REPRESENTATION

373

Dening for simplicity the vectors (102) we can expand

Since the potential one-form is not unique, one can choose an additional condition (Lorenz condition) without changing the eld two-form . Choosing the condition (110) (109) is reduced to (111)

(103) Applying the orthogonality , we obtain (104) and similarly for , whence (105) The last term of (103) vanishes due to

Now the solution corresponding to the dispersion condition (94) leads to (112) because of the denition (47). Thus, (94) corresponds to the A-wave. To verify that the eld corresponding to the dispersion (95) satises the B-wave condition appears more compliand while assuming the cated. Starting from condition (110), from (111) we see that the potential must be of the form (113) where is some scalar coefcient. The condition (110) then requires that satisfy . Invoking (52) and the rule

(106) Thus, we are left with (107) which vanishes due to (87). Thus, we must have either or satised by . In conclusion, in the SDCM case, the fourth-order dispersion equation splits in two second-order equations (108) The plane-wave propagation depends on the metric dyadics and belonging to the space . The medium has no birefringence if the symmetric parts of these two dyadics are multiples of one another. When in addition is a multiple of , the medium coincides with the doubly-skew medium of [22]. V. PROPERTIES OF THE PLANE-WAVE FIELDS Let us now check whether the plane-wave elds satisfy the decomposition properties associated to the corresponding media. A. QDCM Applying (50) we can write for the QDCM case which equals C. SDCM (109) due to the denition (55).

(114) after a few algebraic steps we obtain

(115) Comparing with (95), the corresponding solution can be identied as the B-wave. B. PDCM Omitting again some details, in [15] it has been shown that the eld corresponding to the dispersion (97) satises the condition (116) i.e., it represents an A-wave. Similarly, the solutions of the dispersion (96) correspond to the eld condition (117)

For the SDCM case, the equation for the potential one-form (85) with expressions from Section IV-C inserted and the axion

374

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

term omitted, yields

(124) (125) (126) (118)

yields Multiplying by leads to the condition

. Multiplying by (119)

from which we obtain the relations (120) (121) This shows us that the plane waves corresponding to the two dispersion (108) are respectively A-waves and B-waves. VI. CONCLUSION In this paper we have applied four-dimensional differentialform formalism to dene media in which eld two-forms can be decomposed in two parts: the A-elds and the B-elds: . The decomposed elds are dened in terms of two given bivectors and so that they satisfy and . Media with such a property are called decomposable media (DCM) for brevity. It is shown that a large class of four-dimensional DCM medium dyadics must satisfy an equation which is either linear or quadratic. The media satisfying the linear equation are called special decomposable media (SDCM). Those satisfying the quadratic equation are shown to fall in two subclasses which equal those previously known as generalized Q-media and generalized P-media with an added axion term. Such subclasses have been dubbed QDCM and PDCM, repectively. Basic properties of all three classes of decomposable media are discussed in the paper. The properties of QDCM and PDCM are based on earlier studies on generalized Q- and P-media while the SDCM class appears to be a novel generalization of the class of non-birefringent doubly-skew media. Dispersion equations for a plane wave are derived for all classes of media and their correspondence to the decomposition of the eld are demonstrated. In the Appendix it is briey shown that a medium dyadic satisfying the quadratic equation must belong to either the class of Q-media or that of P-media. APPENDIX QUADRATIC DYADIC EQUATION Let us consider the following quadratic equation for the , medium dyadic (122) . Inserting the three-dimensional expansion in the case (21), (122) equals the set of four 3D conditions (123)

Since (124) equals (125) transposed, we can ignore (125). Now one can show that, out of the two pairs of dyadics , and , , exactly one dyadic in each pair possesses a 3D inverse. and assume To prove this, let us rst consider the pair , that neither of the dyadics possesses an inverse. Choosing the reciprocal bases , in a suitable manner, we can expand (127) (128) and , , are one-forms of which the latter where , three are linearly dependent. These substituted in (123) yields a dyadic equation between the one-forms (129) Assuming that the same subspace as scalar , we have , (129) implies that and occupy and , and that furthermore, for some (130) The relation now becomes (131) Substituting this in (124), the resulting equation (132) does not have any solutions since the left-hand side has no inverse while the right-hand side has one for . Thus, the have an inverse is incorrect. assumption that neither nor Releasing the constraint , the case , can be shown to lead to the same result. must possess an inThus, at least one dyadic of the pair , verse. The same conclusion is valid for the pair , . To demonstrate that exactly one dyadic of each pair possesses an inverse, we notice that the left-hand sides of (123) and (126) are symmetric dyadics, whence all four terms must be antisymmetric dyadics. Because 3D antisymmetric dyadics can be expressed in terms of some vector or one-form as (133) (134) the dyadics in (123) and (126) must have the general form (135) (136) Since antisymmetric 3D dyadics do not have an inverse, neither do the left-hand sides of (135) and (136). This concludes the proof. As a conclusion, there are exactly two invertible dyadics

LINDELL et al.: DECOMPOSABLE MEDIUM CONDITIONS IN FOUR-DIMENSIONAL REPRESENTATION

375

among the four dyadics. Let us split the problem of nding solutions to (122) by considering the four possible cases separately. A. Case 1: Assuming that and possess inverses, the other two dyadics can be solved from (135) and (136) as (137) (138) Dening the dyadic (139) the inverse of which exists due to the above assumption, we can express (124) as a condition for the dyadic , (140) Now it is easy to check that the solution of (140) must be of the uniaxial form (141)

Dening the dyadic (148) (124) can be shown to yield (140) in spite of the different definition of . Thus, the expression (141) can be applied. Let us for simplicity dene the additional dyadic (149) in terms of which we can express from (148), (146) and (147) the dyadics (150) (151) (152) Substituting these in (21), the modied medium dyadic becomes

(153) where the parameter and the scalar of (122) are related as (142) At this point we can construct the medium dyadic (122) for Case 1 from (139), (137) and (138) as satisfying This can be expressed in the compact form

(154) (155)

(143) and it can be further expressed in the compact form

Media dened by modied medium dyadics of the form for some dyadic have been called Q-media [14]. For , , , , the Gibbsian isotropic medium [10], eqn (5.42), is obtained which also includes the vacuum medium. C. Case 3: As a third case we consider the possibility that inverses. Equation (124) now becomes and have

(144) (145) for some Media dened by medium dyadics of the form dyadic have been called P-media [15]. Obviously, the pure axion (PEMC) medium case for B. Case 2: Assuming that and possess inverses, the other two medium dyadics can be expressed as (146) (147) , , is obtained as a special , .

(156) Denoting this time (157) which has an inverse within the limits of Case 3, the condition (156) takes the form (158)

376

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Multiplying by

we obtain (159)

whence (158) becomes (160) , this condition leads to an Because of the assumption impasse: the dyadic on the right-hand side has an inverse while that on the left-hand side has not. Thus, the original assumption that and possess inverses is obviously invalid. D. Case 4: As the nal case we consider the possibility that have inverses. Equation (124) now becomes and

(161) Since this is of the same form as (156), the conclusion must be have inverses cannot be similar: the assumption that and valid for to satisfy (122) with . In conclusion, there are only two classes of solutions to the quadratic (122), that of Q-media and that of P-media. REFERENCES
[1] J. A. Kong, Electromagnetic Wave Theory. Cambridge, MA: EMW Publishing, 2005, pp. 138138. [2] I. V. Lindell, Methods for Electromagnetic Field Analysis. New York: Wiley, 1995, pp. 5454. [3] P. C. Clemmow, The theory of electromagnetic waves in a simple anisotropic medium, Proc. IEE, vol. 110, no. 1, pp. 101106, 1963. [4] T. J. Bromwich, Electromagnetic waves, Phil. Mag., vol. 38, pp. 143164, 1919. [5] A. Kujawski and S. Przezdziecki, Necessary conditions for the splitting of electromagnetic elds into TE and TM constituents in a class of anisotropic media, Bull. Acad. Polon. Sci., Math. Astr. phys., vol. 21, no. 10, pp. 955962, 1973. [6] I. V. Lindell and F. Olyslager, Generalized decomposition of electromagnetic elds in bi-anisotropic media, IEEE Trans. Antennas Propag., vol. 46, pp. 15841585, 1998. [7] F. Olyslager and I. V. Lindell, Field decomposition and factorization of the Helmholtz determinant operator for bianisotropic media, IEEE Trans. Antennas Propag., vol. 49, pp. 660665, 2001. [8] G. A. Deschamps, Electromagnetics and differential forms, Proc. IEEE, vol. 69, no. 6, pp. 676696, 1981. [9] F. W. Hehl and N. O. Yu, Foundations of Classical Electrodynamics. Boston: Birkhuser, 2003. [10] I. V. Lindell, Differential Forms in Electromagnetics. New York: Wiley, 2004.

[11] I. V. Lindell and H. Walln, Wave equations for bi-anisotropic media in differential forms, J. Electromagn. Waves Appl., vol. 16, no. 11, pp. 16151635, 2002. [12] I. V. Lindell, Differential forms and bi-anisotropic media, Electromagnetics, vol. 26, no. 34, pp. 191201, 2006. [13] I. V. Lindell, Differential forms and electromagnetic materials, in Theory and Phenomena of Metamaterials, F. Capolino, Ed. Boca Raton: CRC Press, 2009, pp. 4.14.16. [14] I. V. Lindell and H. Walln, Differential-form electromagnetics and bi-anisotropic Q-media, J. Electromagn. Waves Appl., vol. 18, no. 7, pp. 957968, 2004. [15] I. V. Lindell, L. Bergamin, and A. Favaro, The class of electromagnetic P-media and its generalization, Prog. Electromag. Res. B, vol. 28, pp. 143162, 2011. [16] U. Leonhardt, Optical conformal mapping, Science, vol. 312, pp. 17771777, 2006. [17] J. B. Pendry, D. Schurig, and D. R. Smith, Controlling electromagnetic elds, Science, vol. 312, pp. 17801780, 2006. [18] U. Leonhardt and T. G. Philbin, General relativity in electrical engineering, New J. Ph., vol. 8, pp. 247247, 2006. [19] I. V. Lindell and H. Walln, Generalized Q-media and eld decomposition in differential-form approach, J. Electromagn. Waves Appl., vol. 18, no. 8, pp. 10451056, 2004. [20] E. J. Post, Formal Structure of Electromagnetics. Mineola, N.Y.: Dover, 1997. [21] I. V. Lindell, The class of bi-anisotropic IB-media, Prog. in Electromag. Res., vol. 57, pp. 118, 2006. [22] L. Bergamin, Doubly skew: A new class of non-birefringent media, in Proc. URSI Int. Symp. Electromag. Theory, Berlin, 2010, pp. 304307. [23] K. C. Yeh and C. H. Liu, Theory of Ionospheric Waves. New York: Academic Press, 1972. Ismo V. Lindell (S68M69SM83F90LF05) was born in 1939 in Viipuri, Finland. He received the Ph.D. degree from the Helsinki University of Technology, Espoo, Finland, in 1971. Currently, he is Professor Emeritus of Electromagnetic Theory at the Department of Radio Science and Engineering, Aalto University, Espoo, Finland. He has authored or coauthored more than 270 refereed scientic papers and 14 books, for example, Methods for Electromagnetic Field Analysis (IEEE Press, 2002), Electromagnetic Waves in Chiral and Bi-Isotropic Media (Artech House, 1994), Differential Forms in Electromagnetics (IEEE Press, 2004) and Long History of Electricity (Gaudeamus, 2009, in Finnish). Dr. Lindell received the IEEE S.A. Schelkunoff prize (1987), the IEE Maxwell Premium (1997 and 1998), the URSI van der Pol gold medal (2005) and the State award for public information (2010).

Luzi Bergamin was born in 1973 in Bern, Switzerland. He received the Ph.D. degree from the University of Bern, in 2001. He currently works on environmental physics at KB&P GmbH, Bern, Switzerland. Previously, he was a Research Assistant at Vienna University of Technology, Austria (20012005); the European Space Agency ESA, Noordwijk, The Netherlands (20052008); and Aalto University, Espoo, Finland (20092010).

Alberto Favaro was born in 1985 in Treviso, Italy. He received the M.Sci. degree (with First Class Honors) from Imperial College London, U.K., in 2008, where he is currently working toward the Ph.D. degree.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

377

Communications
Novel UHF RFID Tag Antenna for Metallic Foil Packages
Jeongki Ryoo, Jaeyul Choo, and Hosung Choo
AbstractA low-cost wideband antenna for radio-frequency identication (RFID) tags for use with metallic foil packages is designed in this communication. An proximity inductively coupled feed structure is employed to perform wideband impedance matching between a slot on a metallic foil package and a small rectangular feeding loop connected to a microchip. The results show a measured half-power bandwidth of 32.2% (799 MHz1,093 MHz) and a reading distance of over 10 m. Index TermsLow-cost, metallic foil package, UHF RFID tag, wide bandwidth.

I. INTRODUCTION In ultra-high frequency radio-frequency identication (UHF RFID) systems, the reader transmits a microwave signal to a tag that is composed of an antenna and a microchip. The microchip then responds to the reader using a modulated backscattered signal by varying its input impedance. The tag antenna in UHF RFID systems is one of the most important components, and should be able to overcome performance degradation caused by nearby dielectric or metallic materials. However, underlying metallic objects such as metallic foil packages are strong determinants of antenna impedance change, resulting in poor impedance matching and reduced reading distance. To overcome this problem, some tag antennas that can be mounted on metallic objects have been proposed [1][4]. Such antennas usually employ a thick dielectric inner layer between a tag and the metallic surface, and this bulky structure increases fabrication costs and restricts the range of applications. Recently some researchers have proposed a slot-type tag that uses the conducting surface of foil packaging as a slot radiator [5], [6]. These slot-type antennas are much more efcient than dipole-type antennas since the currents spread throughout the broad surface of the foil packaging sheet. However, a single slot antenna has a narrow bandwidth, and thus its radiating performance can be easily changed by dielectric objects in the foil packaging. Moreover, a single-slot tag antenna is not suitable for mass production due to the difculty of directly connecting the microchip to a metallic foil package [5][7]. In this communication, we propose a novel inductively coupled slot antenna to be used with metallic foil packages. The tag antenna is composed of a slot radiator, which is made by cutting a slit in the metallic foil surface of the package, and a loop feeder, to which the tag chip is attached at the center. This inductively coupled feeding structure enables detuning of the resonant frequency and impedance without changing the slot dimensions in the packaging foil. In addition, the inductive
Manuscript received August 30, 2010; revised May 05, 2011; accepted June 16, 2011. Date of publication September 19, 2011; date of current version January 05, 2012. This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2010-013-D00055). J. Ryoo is with the RFID Research Lab., LS Industrial Systems Co., Ltd., Gyeonggi-do 431-080, Korea (e-mail: jkryoo@lsis.biz). J. Choo is with Department of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 305-701, Korea (e-mail: jychoo@kaist.ac.kr). H. Choo is with the School of Electronic and Electrical Engineering, Hongik University, Seoul 121-791, Korea (e-mail: hschoo@hongik.ac.kr). Digital Object Identier 10.1109/TAP.2011.2167911

Fig. 1. Proposed RFID tag antenna structure: (a) conguration of the tag, (b) proximity inductively coupled feed structure.

coupling between the slot and the loop feeder provides a broadband property, and results in good impedance matching with the various contents of the metallic foil packages. We built the tag antenna on an actual metallic foil package and measured its performance, verifying that the proposed tag antenna can be used in item-level RFID applications for metallic foil-packaged goods at low cost. II. ANTENNA DESIGN AND OPTIMIZATION Fig. 1 shows the overall geometry and design parameters of the proposed tag antenna, which is composed of a slot radiator and a feeding loop connected to a microchip. Both the slot and the feeding loop are etched onto a substrate of polyethylene (PET; "r = 3:9, tan  = 0:003, thickness = 50 m) for exibility and to minimize cost. The measured input impedance of the tag chip (Impinj, Monza3) is 20 0j170 at 912 MHz, and the threshold power level of the tag chip (Pchip min: ) is 015 dBm. The small rectangular feeding loop is placed on top of the slot radiator, and the PET layers (H1 = 50 m and H2 = 50 m) are lled between the loop feeder and the radiator as shown in Fig. 1(a). The design parameters for the tag antenna are perimeter (Lf 2 Wf ),

0018-926X/$26.00 2011 IEEE

378

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 3. Simulated and measured input impedance.

shown in Fig. 2(a), because Dh:of f set changes the coupling coefcient between the slot and the loop feeder, similar to an inductively coupled antenna [8]. To make the antenna operate at the UHF RFID frequency used in South Korea, we used a Dh:of f set of 6.5 mm when the input impedance of the tag chip is 20 0j170 at 912 MHz. We then change the value of Dv:of f set as shown in Fig. 2(b). The impedance contour diameter of the -shape is changed according to the value of Dv:of f set , resulting in variation in the quality factor (Q) and the bandwidth of the antenna. Based on the aforementioned parametric study, we obtain the optimized design parameters as follows: Lg = 200 mm, Wg = 160 mm, Ls = 116:4 mm, Ws = 14:6 mm, Lf = 11:3 mm, Wf = 27 mm, Llf = 1 mm, and Wlf = 3:7 mm. III. RESULT To verify the performance of the proposed tag experimentally, the input impedance of the fabricated antenna is measured by using a coaxial balun probe and an Agilent E5071C vector network analyzer. Fig. 3 represents the measured and simulated antenna impedance, as well as the conjugate impedance of the tag chip. The simulated impedance agrees well with the measurement, and their values are close to the conjugated chip impedance in the frequency range of interest. Particularly, the slope of the reactance curve appears to be parallel to that of the conjugated chip in the range of 860 MHz to 960 MHz, which results in broadband matching in the worldwide UHF RFID frequency band. Fig. 4 shows the derived power reection coefcient curves based on the measured chip impedance and both measured and simulated antenna impedances. In the same gure, we also added the measured read range conducted in an anechoic chamber. The simulated half-power bandwidth (power-reection coefcient < 03 dB) is 298 MHz (32. 7%), from 798 MHz to 1,096 MHz and the measurement is 294 MHz (32.2%), from 799 MHz to 1,093 MHz. Such a broad bandwidth makes it applicable for used in various target materials with different dielectric constants. In addition to bandwidth, to examine the performance change depending on various target objects, we simulated the read range when 5-mm-thick dielectric materials were placed underneath the proposed tag antenna as shown in Fig. 5. The relative dielectric constant for the target changed from 1 to 9. Then the read range Rtag was calculated using (1) as follows, with the simulated radiation gain, impedance mismatch, and a reader with 36 dBm EIRP [9]:

Fig. 2. Impedance on Smith chart: (a) varying . (b) varying at a xed .

D = 3:65 mm 6:5 mm

at a xed

line width (Llf ; Wlf ), position (Dh:of f set ; Dv:of f set ), and slot size (Ls 2 Ws ). The feeding loop, which resonates in combination with the capacitive microchip, induces currents on the surface of the slot by using proximity inductive coupling. The dimensions of the slot (Ls 2 Ws ), the feeder loop (Llf ; Wlf ; Lf 2 Wf ), and the horizontal position of the loop (Dh:of f set ) mainly determine the resonance frequency, while the vertical position of the loop (Dv:of f set ) controls the quality factor (Q) of the antenna. To interpret the impedance characteristic of the proposed antenna, we investigate the antenna impedance by using a commercial full-wave EM simulator (Ansoft HFSS) according to the position of the feeding loop. Fig. 2 shows the computed antenna impedance on a Smith chart, ranging from 612 MHz to 1,212 MHz, by varying the Dh:of f set and Dv:of f set , respectively. In the same gure, we added 03 dB and 06 dB received power coefcient locus to observe the impedance characteristics more easily. As Dh:of f set increases at a xed Dv:of f set of 3.65 mm, the antenna impedance rotates in counterclockwise, as

tag

 P  P  P = 4
= 4

EIRP tag

EI RP

0 02 Eff D
P
tag tag chip min :

tag

(1)

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

379

Fig. 4. Power-reection coefcient of the proposed antenna.

[4] L. Mo, H. Zhang, and H. Zhou, Broadband UHF RFID tag antenna with a pair of U slots mountable on metallic objects, Electron. Lett., vol. 44, no. 20, pp. 11731774, Sep. 2008. [5] X. Zeng, J. Siden, G. Wang, and H. E. Nilsson, Slots in metallic label as RFID tag antenna, in Proc. IEEE Antennas and Propagation Society Int. Symp., Jun. 2007, pp. 17491752. [6] L. Ukkonen, M. Schaffath, L. Sydanheimo, and M. Kivikoski, Analysis of integrated slot-type tag antennas for passive UHF RFID, in Proc. IEEE Antenna and Propagation Society Int. Symp., Jul. 2006, pp. 13431346. [7] K. H. Lin, S.-L. Chen, and R. Mittra, A capacitively coupling multifeed slot antenna for metallic RFID tag design, IEEE Antennas Wireless Propag. Lett., vol. 9, pp. 447450, Apr. 2010. [8] H. Choo and H. Ling, Design of electrically small planar antennas using inductively coupled feed, Electron. Lett., vol. 39, no. 22, pp. 15631565, Oct. 2003. [9] P. V. Nikitin and K. V. S. Rao, LabVIEW-based UHF RFID tag test and measurement system, IEEE Trans. Ind. Electron., vol. 56, no. 7, pp. 23742381, Jul. 2009.

Design of a Broadband All-Textile Slotted PIFA

Ping Jack Soh, Guy A. E Vandenbosch, Soo Liam Ooi, and Nurul Husna Mohd Rais

Fig. 5. Simulated read range by varying underlying materials.

where minimum operating power Ptag is expressed by the power-reection coefcient 0tag , the radiation efciency Efftag and the directivity Dtag of the proposed tag, and PEIRP is the transmitting power of reader. Although there exist frequency shifts of maximum read range to the lower frequency band, good read range from 12 m to 15.1 m was observed at 912 MHz.

AbstractA new broadband textile based PIFA antenna structure designed for wireless body area network (WBAN) applications is presented. The new topology can be directly integrated into clothing. The study starts by considering three different materials: exible copper foil, and ShieldIt Super and pure copper polyester taffeta conductive textiles. Bandwidth broadening is successfully achieved by implementing a novel and simple slot in the radiating patch. The measured reection coefcient and radiation characteristics agree well with simulations. Moreover, radiation characteristics and bandwidth show satisfactory immunity against detuning when operating on-body, especially when placed on the back. To our knowledge, the proposed structure is the rst fully fabric based slotted PIFA to be reported in open literature with high bandwidth (more than 46%) and reasonable gain (ca. 1.5 dB), to be used for multiple applications in the frequency band of 1.8 to 3.0 GHz. Index TermsBroadband antennas, conformal antennas, planar inverted-f antennas (PIFA), textile antennas.

IV. CONCLUSION We proposed a novel UHF RFID tag based on proximity inductively coupled feeding with the slot radiator for metallic foil packages. The fabricated tag antenna showed broad matching characteristics of 294 MHz (32.2%), from 799 MHz to 1,093 MHz. The measured reading distance was over 10 m at 36 dBm EIRP in the UHF RFID frequency band. The results veried that the proposed low-cost tag antenna can be used with a metallic foil package with various inner materials.

I. INTRODUCTION Due to the increasing demand for multi-functional, multi-band wireless operation and consumer-centric technology, textile antennas have
Manuscript received October 05, 2010; revised May 20, 2011; accepted July 15, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the Malaysian Ministry of Higher Education (MOHE). P. J. Soh is with the Telecommunications & Microwave Research Division, Department of Electrical Engineering (ESAT-TELEMIC), Katholieke Universiteit Leuven, 3001 Leuven, Belgium, on leave from the School of Computer and Communication, Universiti Malaysia Perlis (UniMAP), 02000 Kuala Perlis, Malaysia (e-mail: pingjack.soh@esat.kuleuven.be). G. A. E. Vandenbosch is with the ESAT-TELEMIC, Katholieke Universiteit Leuven, 3001 Heverlee, Belgium (e-mail: guy.vandenbosch@esat.kuleuven. be). S. L. Ooi and N. H. M. Rais are with the School of Computer and Communication Engineering, Universiti Malaysia Perlis (UniMAP), 02000 Kuala Perlis, Malaysia (e-mail: sooliam@gmail.com; nwhosena@gmail.com). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167950

REFERENCES
[1] C. Cho, H. Choo, and I. Park, Design of planar RFID tag antenna for metallic objects, Electron. Lett., vol. 44, no. 3, pp. 175177, Jan. 2008. [2] S. Weigand, G. H. Huff, K. H. Pan, and J. T. Bernhard, Analysis and design of broad-band single-layer rectangular U-slot microstrip patch antennas, IEEE Trans. Antennas Propag., vol. 51, no. 3, pp. 457468, Mar. 2003. [3] S. L. Chen and K. H. Lin, A slim RFID tag antenna design for metallic object applications, IEEE Antennas Wireless Propag. Lett., vol. 7, pp. 729732, Nov. 2008.

0018-926X/$26.00 2011 IEEE

380

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

been receiving growing attention. Future wearable systems should be unobtrusive, exible, and operating with minimal degradation in proximity to the human body. These antennas have to meet bandwidth, efciency and safety requirements, while being consistent with low cost manufacturing techniques. Moreover, in wearable applications, at surfaces cannot be guaranteed. Thus, an important antenna requirement is its ability to work with good robustness against environmental, positional and location changes when being worn. Flexible textile antennas are ergonomically more suitable for on-body operation compared to conventional metallic structures, e.g., rigid copper plates or tapes which are worn. The motivation of choosing a PIFA structure is the fact that wider band operation is obtained in the presence of a ground plane. Also, antennas incorporating a ground plane will be less affected by the presence of the human body. Antenna structures without ground plane experience serious changes in resonance frequency, depending on the distance from the body. A wearable PIFA was rst published by Salonen [1], fabricated using conductive plate, and consequently implemented on a exible FlexPIFA dielectric substrate for operation at 2.45 GHz [2]. Fully textile antennas of various congurations featuring ground planes were also developed for emergency rescue personnel and covert operations [3][11]. Incorporation of relatively small or suspended ground planes to textile antennas was also investigated in [11], [12]. However, none of the later publications (except for [3]) considered the 3D PIFA topology as the basis for their design. In conventional PIFAs, parameters variations such as shorting plate, feed line size, height, shorting pin locations and ground plane size have been found to be signicant PIFA parameters. On the other hand, slotting the structure has been identied as an effective method of bandwidth broadening. Compact broadband PIFAs with more complicated slots are not suitable for fabrication using manual cutting tools, whereas the proposed structure simplies manufacturing and reduces proneness to dimensional errors. The novelty of this work is the following. A PIFA topology is implemented, introducing a slot on the radiator, and utilizing a shorting wall instead of pins at a well-chosen location to broaden the bandwidth. The introduction of the slot modies the currents on the patch, resulting in a larger bandwidth compared to recent textile antennas in [3][11] designed for operation in the 2.45 GHz ISM band. The proposed antenna structure is applicable in the Industrial, Scientic and Medical (ISM) band (2.402.48 GHz), and also in various newer applications, including International Mobile Telecommunication-2000 (IMT-2000) bands (1.922.69 GHz) and Long Term Evolution-Time Division Multiplexing (LTE-TDD) systems (1.902.62 GHz). Most importantly, the structure maintains fabrication simplicity, enabling an all-textile hardware realization. II. CONDUCTIVE TEXTILE MATERIAL Three types of conducting material are used in this study, namely a 0.035 mm thick copper foil tape (CT), and two types of conducting textiles. The conducting textiles are ShieldIt Super (SH) and Pure Copper Polyester Taffeta Fabric (PCPTF), both from LessEMF Inc. Both textiles possess high conductivities, i.e., surface resistivity, Rs < 0:05
=sq. SH is a ripstop, woven polyester textile coated with copper and nickel. The thickness, t, is 0.17 mm, its estimated weight is 230 g=m2 . The second textile, PCPTF, is plain woven and coated using pure copper. It has a thickness, t of 0.08 mm, an estimated weight of 80 g=m2 . The parameter of prime importance is the equivalent conductivity of the textile used. This parameter ultimately determines the efciency and gain of the antennas. The homogenized conductivities were calculated based on the surface resistances provided by the manufacturer.

Fig. 1. Dimensions of the proposed slotted PIFA structure: (a) top view, (b) perspective view.

Fig. 2. Fabricated prototype for (a) Plain CT PIFA, (b) Slotted CT PIFA, (c) Plain PCPTF PIFA, (d) Slotted PCPTF PIFA.

III. ANTENNA DESIGN The proposed PIFA structure consists of two conductive layers shorted by a wall, as shown in Fig. 1. Its substrate, a 6 mm thick felt fabric, is placed between ground plane and radiating patch. It is assumed that the substrate has a relative permittivity "r of 1.43 and loss tangent (tan  ) of 0.025 at 2.4 GHz [6]. The fabricated prototypes are shown in Fig. 2. To realize a full textile fabrication, some high level requirements are imposed. (1) The whole structure has to be fabricated out of a single piece of textile. This avoids additional interconnections using glue/ epoxy, and maintains a consistent surface conductivity. (2) The structure has to be simple and as symmetrical as possible, to ease fabrication. (3) Hard-to-cut shapes/corners should be avoided to minimize fabrication inconsistencies. These high level requirements can be translated into fair requirements for the dimensions of the structure. (1) The length of the radiator (RL ) is about 1.25 to 1.5 times the width of the radiator (RW ) to produce a rectangular radiator shape. (2) The width of the radiator (RW ) and the ground plane width (GW ) are identical and are aligned. This eases the fabrication using a single piece of textile.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

381

TABLE I DIMENSIONS OF THE DESIGNED PIFAS

Fig. 3. Center frequency (f = f tical feed position offset (1f ).

), bandwidth (BW ), and S

versus ver-

(3) The shorting wall width ( W ), which is placed along the length of the radiator, is about a quarter of the radiator length. (4) The probe feed is placed near the center of the radiator. The feeding position along the radiators length and width ( L and W ) are about 0 5 L and 0 5 W , respectively. CST Microwave Studio was used in the design of the structures. Ideally, the bandwidth ( ) should be centered at a frequency of 2.45 GHz, which is calculated using an experimentally derived equation, for a plain, edge-fed structure in [13]:

: R BW

: R

fc = 3RW +5:6RL +3:7h0 3Wcf 0 3:7Ww 04:3 f 2 + f 2 : W L

0 25

(1)

by dimensional variations. This will be detailed out in the next section. The fabricated dimensions are summarized in Table I. IV. FABRICATION TOLERANCES

Assuming RL = 1:25 RW , fW = 0:5 RW , fL = 0:5 RL , WW = : RL and xed dimensions of h = 6 mm, and Wf = 1 mm, the application of (1) yields RW = 19:11 mm and RL = 23:89 mm. Keeping in mind the fabrication tolerance of 0.5 mm, the calculated RL and RW are then simplied to be 24 mm 2 19 mm, respectively. A ground plane of similar size (GL 2 GW = 50 mm 2 19 mm) and a shorting wall width (WW ) of 6 mm is also introduced to complete the PIFA structure. Its center frequency, fc and BW are directly affected

Fig. 4. Center frequency (f = f ), bandwidth (BW ), and S izontal feed position offset (1f ).

versus hor-

Although the structures designed are as simple as possible, standard fabrication technology cannot be used. This implies larger inaccuracies, in the order of a few tenths of a millimeter. Thus it is important to study this through a parameter tolerance analysis, which observes the 010 dB and the 11 at its c . This is carried out while changing the parameters of the slotted and plain SH PIFAs in steps of 0.5 mm. In Fig. 3, a steady increase of c can be observed as the feed ( W ) is . The 11 s of both moving inwards, with simultaneous increase of basic and slotted prototype are quite stable, producing levels below 015 dB. In Fig. 4 moving the feed ( L ) horizontally towards the right, away from the shorting wall, also triggers an increase of c , with a de. The minimum 11 , however, increases to levels around creasing 010 dB, which is still just acceptable. In Figs. 5 and 6, c is seen to decrease as the length ( L ) or width ( W ) of the radiator is increased. In Fig. 7, c is seen to increase with increasing shorting wall width ( W ). The corresponding 11 stays at a satisfactory level of about

BW

f f

BW

BW

f f

020 dB. The slotted PIFAs decreases with an increasing W , while the opposite can be said for the plain PIFAs. A crossing point is found at W = 4 mm, the point where both structures produce s. similar , neither Even though the slot is critical to improve the overall the slot length ( W ), neither slot width ( L ), nor slot position ( P ) is , and 11 . When W , L , and P are signicantly affecting c , for both slotted and plain antennas are consistently about varied, 880 MHz, with a minimum 11 of between 025 to 035 dB. The nal conclusions of the tolerance study are the following. In general, both basic and slotted PIFAs exhibit similar behavior for an identical parameter variation. The parameters with most impact are the radiators width and length ( W and L ), the feeding positions ( L and W ), and the shorting wall dimension ( W ). Although the slot is , the c is not very sensitive to it, indicating primary in enhancing that the design is quite robust with respect to the slot. The presence of for most structures, even with varying dimenthe slot increases sions, due to the modication of the currents on the patch. The tolerance study shows that adding the slot to the simple basic structure broadens signicantly, resulting in a robust overall topology. the

W BW

BW

BW

f BW S

S S

BW

S S

BW BW

BW

382

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 5. Center frequency (f = f diator length (R ).

), bandwidth (BW ), and

versus ra-

Fig. 7. Center frequency (f shorting wall width (W ).

), bandwidth (BW ), and

versus

Fig. 8. Measured S Fig. 6. Center frequency (f = f diator width (R ).

), bandwidth (BW ), and

for all PIFA prototypes.

versus raTABLE II SIMULATED AND MEASURED S FOR PLAIN AND SLOTTED PIFAS

V. RESULTS AND DISCUSSION The PIFAs were prototyped using different conducting materials to operate at fc = 2:45 GHz. Fabrication started by cutting a single piece of textile or CT into the desired shape. Then, the textile is folded and fastened over a 6 mm thick felt fabric to form the PIFAs. A hole was created, enabling the insertion of a 50
SMA connector. The goal was to determine the differences between conducting textiles (PCPTF and SH) on the one hand, and CT on the other hand. Fastening CT and SH textile to the substrate is easy, as both materials came with adhesive reverse sides. The PCPTF textile, on the other hand, has to be stitched onto the substrate along its perimeter. A silver conductive epoxy, model 8331 from MG Chemical was used to connect the SMA connector without soldering. Fabrication using this simple method provides an estimated accuracy of ca. 0.2 mm. The plain versions of the PIFAs are evaluated rst. It can be seen from Table II and Fig. 8 that the discrepancy between the calculated and measured fc for all three prototypes is less than 1%. This fc is the mid point in between the two 010 dB matching frequency points. The measured BW s are also consistent with the simulations. Measurements basically produced between 5% and 14% larger BW s compared to simulations. In general, no signicant differences have been found

among the three prototypes, due to the high similarity in terms of surface resistivity and conductor thickness. In a second run, three different slotted PIFAs were fabricated and measured. A larger frequency shift is observed here; about 1% for the slotted CT PIFA and about 2% for the slotted textile antennas. Similar as in the case of the plain prototypes, slotted PIFAs also show a measured BW larger than predicted by the simulations. Again, CT PIFA produced the least BW change, due to its mechanical robustness and

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

383

TABLE III SUMMARY OF SIMULATED AND MEASURED GAIN AND EFFICIENCY IN FREE SPACE

the homogeneity of the conductivity. SH and PCPTF slotted antennas show a 10% and 20% BW enlargement, respectively. Also notice that for both the slotted textile antennas, the extension of the BW occurs at the lower frequencies. This bandwidth widening may be caused by two factors: (1) manufacturing inaccuracies, and (2) the use of a lossy substrate. The simulated and measured gain and efciency are given in Table III. Measurements were done in a Satimo-64 anechoic chamber. All PIFAs generated simulated gains of about 2 dB and efciencies of about 80%. Measured gain levels are between 1.5 and 2.2 dB, while measured efciencies in free space (FS) are between 76% and 86%. In general, the CT PIFAs produced better efciency and gain compared to both textile antennas, due to the higher conductivity (theoretically about 200 times better). Slotted PIFAs tend to produce a slightly higher measured gain and efciency in comparison to the plain version. Radiation patterns for the PIFAs are shown in Fig. 9. The simulated and measured co-polarized components in both ' = 0 and ' = 90 cut show an excellent agreement, generating a quasi omnidirectional pattern in the azimuth plane. CT PIFAs showed the least simulationmeasurement difference, as expected. This is again due to the conductor purity/homogeneity and mechanical robustness, compared to the more exible textiles.

Fig. 9. Simulated and measured radiation pattern for (a) plain CT PIFA, (b) slotted CT PIFA, (c) plain SH PIFA and (d) slotted SH PIFA. Legend: (----) ; (- -) measured ' ; () simulated ' ; (-o-) simulated ' . measured '

=0 1 = 90

=0

= 90

VI. PROXIMITY OF HUMAN BODY UPPER TORSO The presence of a users body alters the BW , introduces fc shift and an unavoidable efciency loss, due to its high "r value [6]. For these reasons, the proposed PIFAs are tested on-body by placing them at three distinct locations (back (B ), chest (C ), and shoulder (D)), see Fig. 10. These on-body locations are chosen specically to minimize antenna bending. A 10.3 mm ground plane-to-skin distance, which is about =12, is introduced, and both horizontal (HP) and vertical polarizations (VP) are considered. The CST Hugo voxel model is utilized to co-simulate a ShieldIt PIFA. The measurement was then carried out in an anechoic chamber, on a male test subject (with height of 1.78 m and weight of 80 kg). In practice, a 9.1 mm felt spacer was placed between the users eece jacket (which is 1.2 mm thick) and the PIFAs ground plane. All antennas were then measured at the three locations, both at HP and VP. As expected, BW is signicantly affected, as shown in Table IV. Simulations predicted that the slotted PIFAs will exhibit a larger BW change in comparison to the plain prototypes. Placement on the shoulder and back, both HP and VP, is producing less than 5% change for plain and slotted PIFAs in simulations. Measurements conrmed that both shoulder and back are suitable, when evaluated in terms of BW change. The measured plain PCPTF PIFA provides the least BW change at all locations

Fig. 10. PIFA placed on (a) Hugo voxel model (chest and shoulder); (b) human subject (back).

TABLE IV MEASURED BANDWIDTH (MHz) IN PROXIMITY OF HUMAN UPPER TORSO

(only 10%) in comparison to FS, while its slotted version exhibits a larger BW change of about 17%.

384

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE V EFFICIENCY AND GAIN IN PROXIMITY OF HUMAN UPPER TORSO

Table VI presents a comparison of the textile antenna selected from this work and previous work employing exible conductors or textile materials. It can be seen that our antenna has by far the largest bandwidth for non-ultrawideband (UWB) applications, at the cost of a reduced gain. However, we believe that due to the changing position of the human body in practice, designing a more omnidirectional antenna with a lower gain is certainly of more interest. VII. CONCLUSION A novel, compact, broadband and simple PIFA fabricated using conductive textiles, designed to suit wireless body area network applications, is proposed and discussed. Simulated and experimental results indicate that the slotted antennas are able to operate with a satisfactory reection coefcient below 010 dB and a bandwidth of up to 1200 MHz in free space and 1300 MHz on body. The slotted PIFAs are superior over the non-slotted ones implemented using the same materials and radiator topology. Considering the very difcult non-standard measuring environment, measured reection coefcients and radiation pattern characteristics of both textile and metallic antennas agree reasonably well with simulated ones. On-body operation shows a satisfactory efciency level when placed on the users back.

TABLE VI BANDWIDTH (MHz) AND GAIN (dB) COMPARISON AGAINST OTHER TEXTILE/FLEXIBLE ANTENNAS DESIGNED FOR 2.45 GHz ISM BAND

REFERENCES
[1] P. Salonen, L. Sydanheimo, M. Keskilammi, and M. Kivikoski, A small planar inverted-F antenna for wearable applications, in 3rd Int. Symp. on Wearable Computers Digest, 1999, pp. 95100. [2] P. Salonen, M. Keskilammi, J. Rantanen, and L. Sydanheimo, A novel bluetooth antenna on exible substrate for smart clothing, in Proc. IEEE Int. Conf. on Systems, Man, and Cybernetics, 2001, vol. 2, pp. 789794. [3] P. J. Soh, G. A. E. Vandenbosch, S. L. Ooi, and M. R. N. Husna, Wearable dual-band Sierpinski fractal PIFA using conductive fabric, Electron. Lett., vol. 47, pp. 365367, 2011. [4] P. Salonen and L. Hurme, A novel fabric WLAN antenna for wearable applications, in Proc. IEEE Antennas and Propagation Society Int. Symp., 2003, vol. 2, pp. 700703. [5] C. Hertleer, H. Rogier, L. Vallozzi, and L. Van Langenhove, A textile antenna for off-body communication integrated into protective clothing for reghters, IEEE Trans. Antennas Propag., vol. 57, pp. 919925, 2009. [6] C. Hertleer, A. Tronquo, H. Rogier, L. Vallozzi, and L. Van Langenhove, Aperture-coupled patch antenna for integration into wearable textile systems, IEEE Antennas Wireless Propag. Lett., vol. 6, pp. 392395, 2007. [7] A. Tronquo, H. Rogier, C. Hertleer, and L. Van Langenhove, Robust planar textile antenna for wireless body LANs operating in 2.45 GHz ISM band, Electron. Lett., vol. 42, pp. 142143, 2006. [8] B. Qiang and R. Langley, Wearable EBG antenna bending and crumpling, in Proc. Antennas and Propagation Conf., Loughborough, 2009, pp. 201204. [9] D. L. Paul, M. Klemm, C. J. Railton, and J. P. McGeehan, Textile broadband E-patch antenna at ISM band, Proc. IET Seminar on Antennas and Propagation for Body-Centric Wireless Communications, pp. 3843, 2007. [10] I. Locher, M. Klemm, T. Kirstein, and G. Troster, Design and characterization of purely textile patch antennas, IEEE Trans. Adv. Packaging, vol. 29, pp. 777788, 2006. [11] H. M. R. Nurul, F. Malek, P. J. Soh, and G. A. E. Vandenbosch, Dualband bench feed textile antenna, in Proc. Int. Symp. on Antennas and Propagation (ISAP), Macau, China, 2010, pp. 709712. [12] E. C. Lee, P. J. Soh, N. B. M. Hashim, G. A. E. Vandenbosch, V. Volski, I. Adam, H. Mirza, and M. Z. A. A. Aziz, Design and fabrication of a exible Minkowski fractal antenna for VHF applications, in Proc. 5th Eur. Conf. Antennas and Propagation (EUCAP), 2011, pp. 521524. [13] H. T. Chattha, H. Yi, Z. Xu, and L. Yang, An empirical equation for predicting the resonant frequency of planar inverted-F antennas, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 856860, 2009.

The dielectric coupling to the body is expected to introduce a slight fc down-shifting for all antennas. Analyzing the simulated fc , an average shift of 5% is found for both PIFA topologies at all three locations in comparison to FS. Measurements, on the other hand, indicate a maximum shift of about 7.5%. The location which is the least affected by the choice of orientation is the chest, generating a consistent shift of 2% for plain and 3% for slotted PIFAs. In short, the placement of the proposed antennas on a human user preserves its wideband characteristics and invokes a fc degradation, which is relatively small compared to the BW change. Although the measured PIFAs function properly at 2.45 GHz at all locations and orientations, the best location for the antenna is found to be on the back of the upper torso, consistent with [5]. Other than generating the best efciency of 60%, this location is also unlikely to cause antenna bending in the event of movement and pose variations. This is also evident when analyzing it in terms of BW change and fc detuning in Table V. On the contrary, the PIFAs placement on the chest produced a lower efciency level of 40%. It is also observed that the gain/efciency performance depends more on the location rather than on its orientation. Slotted PIFAs also perform better than plain PIFAs when placed at the same on-body location.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

385

An Ultrawideband (UWB) Slotline Antenna Under Multiple-Mode Resonance

X. D. Huang, C. H. Cheng, and L. Zhu

AbstractA novel ultrawideband (UWB) antenna based on a multiple-mode slotline radiator is proposed and developed. This slotline radiator/resonator is formed on the ground plane of a microstrip line and it is asymmetrically fed in right angle by a shorted strip conductor. As the feeding position and port impedance are properly selected, an operating 2 0) of 3.6 to 10.5 GHz with four reection zeros band (VSWR is realized. A practical 50-
microstrip-fed slotline resonator antenna is then constructed with a tapered line transformer. After its radiation performance is optimally designed, a prototype UWB antenna is fabricated and measured. The measured impedance bandwidth has successfully 2 3) as predicted achieved a fractional bandwidth of 97.9% (VSWR in theory, over which the antenna peak gain varies from 0.5 to 4.5 dBi. Index TermsSlotline radiator and multiple-mode resonance, ultrawideband (UWB) antenna. Fig. 1. Conguration of proposed microstrip-fed slotline antenna that is asymmetrically fed at an offset distance (L ). (a) Slotline radiator: 3D-structure, (b) Detailed view of the feeding structure.

I. INTRODUCTION Microstrip fed slot antennas have been extensively employed in modern communication systems due to many advantageous features, e.g., low cost and ease of implementation. The rst microstrip fed slot/slotline antenna was proposed by Yoshimura in 1972 [1]. As the complementary part of dipole antenna [2], the traditional slot antenna also suffers from narrow operating bandwidth due to its single resonance nature. So far, two well-known straightforward methods have been developed and employed to enhance the frequency bandwidth of operation. One is to decrease the quality factor of the radiating mode by widening the radiating slot (only one radiating mode is used) [2]; the other is to use the multiple-resonance [3][9] or multiple-slot concepts [5], [10]. Comparing with each other, the rst solution is very simple but the extent of bandwidth enhancement is very limited due to the single-resonance nature; the second solution may provide over 2:1 (VSWR < 2:0) impedance bandwidth if the feeding topology and the shape of the radiating slot are properly selected [5][9]. In fact, the majority of the ultrawideband (UWB) antennas can be categorized into the so-called wide-slot antennas [6][9], which usually could achieve 3:1 (VSWR < 2:0) or even wider impedance bandwidth. These wide-slot antennas actually operate under the multiple-mode resonance according to their frequency responses [6][9]. One of the main shortcomings of these wide-slot antennas is their high cross-polarization degree at some frequencies. It is because in these antenna congurations, two orthogonal radiating modes in close resonant frequencies could be excited simultaneously [5]. In this context, the width and the length of the slot are nearly equal, thus it is quite complicated to predict the resonant modes for the radiating slot with irregular shape. Due to no appearance of unied physical model, these wide-slot antennas, placed underneath a strip conductor feeder, have been usually designed with virtue to the time-consuming cut-and-try method via full-wave simulators.
Manuscript received February 14, 2010; revised June 03, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by the National Natural Science Foundation of China (NSFC) under Grant 61071020. X. D. Huang and C. H. Cheng are with the College of Electronic Science & Engineering, Nanjing University of Posts & Telecommunications, Nanjing, Jiangsu, China (e-mail: xdhuang@njupt.edu.cn; chengch@njupt.edu.cn). L. Zhu is with the School of Electrical & Electronic Engineering, Nanyang Technological University, Singapore (e-mail: ezhul@ntu.edu.sg). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167899

In the meantime, a few feasible methodologies have been reported in [3][5] to widen the operating bandwidth of the traditional narrow slot antennas to a certain degree, by introducing one or more external resonances. By introducing a quarter-wavelength resonator on the microstrip feeding line [3], a slot antenna with a width-to-length ratio (WLR) of 0.208 achieves 32.0% impedance bandwidth. Moreover, by producing one or more ctitious short points on a single radiating slot with a high-impedance feeding line, the bandwidth could be widened to 37.0% in [4] and 57.1% in [5]. The slot in these antennas is narrow, i.e., the width is much smaller that its length, so that the slot radiator itself is a quasi 1-D resonator in an open environment. As is well known nowadays, the resonant modes of a 1-D resonator or radiator can be easily predicted using simple transmission line theory. Therefore, the analysis of the antenna with simple 1-D conguration is much easier than that of the antenna with complicated 2-D or 3-D conguration. In particular, only the magnetic current density on the narrow slot area needs to be numerically discretized in simulation, resulting in a great reduction in computational time and memory. Besides, due to the narrow radiating slot, the low cross polarization level could be retained within its operating frequency band [5], which is important for most applications of an antenna, inclusive of UWB communication [11], [12]. To our best knowledge, this type of wideband antennas with single-mode operation can not be directly applied to meet the requirement of the UWB-specied operating bandwidth, i.e., 110% fractional bandwidth. By employing a narrow radiating slot, a novel UWB antenna is presented and implemented in this work. The proposed antenna is characterized and designed via an EM simulator to realize an almost 100% fractional bandwidth (VSWR < 2:0). In this context, the WLR of the radiating slot/slotline is 0.175. The wideband radiation of the narrow slot is theoretically explained by using the concept of multiple mode resonance (MMR), which is at present popular in the design of a wideband lter [13]. By simultaneously exciting its multiple resonant modes of the slotline radiator, the radiation is generated in a wide frequency range dominated by these resonant modes. This antenna is constituted in geometry by asymmetrically feeding this slotline resonator at a proper position with a short-circuited microstrip line. In our study, this slotline antenna is at rst investigated to demonstrate emergence of four resonant modes, aiming to achieve a 3:1 wideband of radiation with four reection zeros in theory. Next, an actual MMR-based UWB slotline antenna with 50-
feeding line

0018-926X/$26.00 2011 IEEE

386

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I

Fig. 2. VSWR versus frequency under varied port impedances (L : ; L : ; W : ; W : ; L : mm).

=70

=50

=10

=04

= 15 0

= 40 0

: ;

Fig. 3. VSWR versus frequency under different congurations (L : ; W : mm). W : ; L

=70

=50

=04

= 40 0

: ;

50 20 0

Fig. 4. Simulated radiation patterns at different frequencies of the proposed : mm, W : mm, L antenna with different feeding positions. (L : mm). (a) Radiation patterns of a center fed antenna (L : mm, W : mm). : mm), (b) Radiation patterns of an offset fed antenna (L

=10

= 40 0

=70 = 15 0

= =

is proposed and studied with a tapered impedance transformer, and its performance is designed using the IE3D simulator. The proposed antenna is nally fabricated on a substrate of Rogers 6002 with a relative permittivity of 2.94 and height of 0.762 mm. The measured results have evidently validated the predicted ones. II. DESIGN OF PROPOSED ANTENNA Fig. 1 shows the conguration of the proposed UWB slotline antenna with a microstrip-line feeder. Similar to its traditional counterpart in geometry [1], the proposed slotline radiator is fed by a shorted microstrip line that is constructed on the opposite side of the substrate. The slot itself is designed with stepped-impedance shape, aiming to properly allocate its rst few resonant frequencies in the desired UWB range. The

slot conguration is similar to the ones reported in [14], [15], where the radiating slot is etched at its center to save its overall circuit area. As was extensively discussed in early literature [16], a shorted microstrip line crossing a narrow slotline in right angle can offer strong coupling strength over a wide frequency range. It can be understood from [2] that as the slotline width increases, an overall electrical radiation aperture of this antenna is broadened, resulting to decrease the quality factor of all the radiating modes. On the one hand, the slotline region in middle with a short length (Region 2 in Fig. 1(b)) near the feed point is selected with a narrow width in order to achieve a tight coupling between the microstrip line and slotline. On the other hand, the remaining slotline region in two sides with much longer length (Region 1) is set as a wide width so as to enlarge the expected radiation aperture.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

387

Fig. 5. Simulated S-parameters and physical dimensions of the designed tapered-line impedance transformer.

Fig. 8. Simulated and measured VSWR of proposed antenna.

Fig. 6. Simulated VSWR of proposed antennas with/without impedance transformer.

Fig. 9. Simulated and measured antenna gain of proposed antenna.

Fig. 7. Photographs of the fabricated UWB slotline antenna. (a) Top view, (b) Bottom view.

The optimization on the parameters of W;L ;W ;W ;L , and Zport is executed using the Zeland IE3D software. In this aspect, the overall slotline length is xed as L = 40:0 mm to make sure that the rst resonance of the slot is near 3.0 GHz. The goal is to get an antenna with the impedance bandwidth (VSWR <= 2:0) of 3.1 to 10.6 GHz. To do it, the port impedance (Zport ) is also tuned in order to obtain a better match in such a wide frequency range. After the optimization, a set of the parameters is derived as L = 40:0; W = 7:0; L = 5:0; W = 1:0; W = 0:4; L = 15:0 mm and Zport = 160
. With the
c c f o c c f o

selection of these parameters, the rst four resonant frequencies of the slotline radiator or resonator can be derived using the two-port weak coupling method [17]. They are found around 2.7, 5.9, 9.0 and 10.6 GHz, respectively. Fig. 2 indicates the simulated VSWR curves of the proposed slot antenna. As shown in Fig. 2, the VSWR values and curves are strongly dependent on the selected port impedance. Three or four reection minima could be observed in all the listed cases. In

the case of 200

, the four reection minima are located at 3.5, 5.7, 8.5 and 10.1 GHz, respectively. When the port impedance is reduced to 120
, the number of minima is reduced to three at the frequencies of 4.4, 7.5 and 9.9 GHz, respectively. Of the ve graphs in Fig. 2, the graph with the port impedance of 160
achieves the VSWR less than 2.0 in a wide frequency band of 3.6 to 10.5 GHz or 97.6%. In this case, the WLR of the constructed slotline radiator equals to 0.175. As tabulated in Table I, the proposed slotline antenna achieves much wider impedance bandwidth than the reported results in [3][5] under the almost same WLR. Such a wideband radiation behavior can be understood by using the concept of MMR [13] with resorting to the simulated VSWR curve. The rst few resonant modes in the slotline radiator are concurrently excited to radiate electromagnetic power in a wide frequency band that is covered by these excited resonant frequencies. For reference, in [14], the achieved fractional bandwidth is 100% by using a slot with a much larger WLR (0.285, from Fig. 1 in [14]) and a metallic back panel in addition. Fig. 3 plots a few sets of VSWR under varied slot antenna congurations, i.e., uniform slot with offset feeder, stepped slot with an offset feeder and stepped slot with a center feeder, respectively. From the gure, it is clear that the slot shape dominates the wideband VSWR response. Comparing with the stepped slot antenna, the VSWR of the antenna with a uniform slot is relatively stable in low operating frequency, but it seems to be highly affected in the high band. With a uniform slot, the slot width ( c ) at the feed point is found to be comparable with the wavelengths at high frequencies. In another case, the antenna with the center feeder has good VSWR performance in its low band (3.0 to 8.0 GHz) except a VSWR peak of about 4.0 at 9.2 GHz.

388

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 10. Measured and simulated radiation patterns at different frequencies of proposed UWB slotline antenna.

Of them, the proposed antenna with offset feeder and stepped slot has the best performance in terms of VSWR as illustrated in Fig. 3. Now, lets investigate the radiation patterns of the two antennas with center/offset feeders. Fig. 4(a) indicates that the radiation pattern of a center fed slot antenna is symmetrical as a result of its symmetrical structure. Its H-plane pattern achieves its peak at 3.5 and 6.0 GHz at the broadside, and becomes split into two beams at 10.0 GHz. This phenomenon is the similar to that happened for a center fed dipole antenna [2]. As shown in Fig. 4(b), the radiation pattern of the offset fed slot antenna is asymmetrical with respect to the broadside direction due to its offset feeding scheme. The radiation pattern at 6.0 GHz is split to two beams since the magnetic current distributes as a full-cycle sinusoid. The unbalance of the two split beams is due to the discontinuity of magnetic current distribution in the slot region 2 (Fig. 1(b)). Besides, the cross polarization degree retains low in the two cases.

III. EXPERIMENTAL VERIFICATION In order to implement a practical antenna that can be tested via 50-
standard equipment, a tapered-line transformer in Fig. 5 is proposed and designed to transform 160-
(local port impedance in Fig. 2) to 50-
(main feeding line). As shown in Fig. 5, the simulated jS11 j is better than 014 dB from 2.0 GHz to 13.0 GHz, whereas the maximum jS21 j is 00:22 dB. By installing this transformer to the proposed slotline antenna, a testable 50-
microstrip-fed slotline antenna can be formed. Fig. 6 plots the simulated VSWR of this actual 50-
fed antenna as compared to that of the initial 160-
fed antenna. As this transformer is installed, the operating band of the proposed slotline antenna is slightly shifted down whereas the operating bandwidth is almost unchanged. Next, a prototype 50-
fed slotline antenna with a tapered-line transformer and nite ground plane of 90.0 2 107.0 mm2 is fabricated and

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

389

its top-/bottom-side views are given in Fig. 7. The measured VSWR is plotted in Fig. 8 against the predicted results for the antennas with the nite (actual) and innite (ideal) ground plane. From the gure, it can be observed that the two predicted VSWR curves of the slotline antennas with nite- and innite ground planes are almost the same as each other. It is indicated that the nite ground has no effect on the input impedance of this antenna if its overall size is sufciently large. Over the realized UWB band of 3.2 to 10.8 GHz, the measured VSWR is less than 2.3 and it is slightly higher than 1.8 obtained in simulation. This small discrepancy is mainly attributed by certain unexpected tolerance in fabrication. The simulated and measured antenna peak and average gains at its H-plane are plotted in Fig. 9 as a function of frequency. The peak gains vary from 0.5 to 4.0 dBi in the realized impedance bandwidth whereas the average gains vary from 03:0 to 1.3 dBi. Fig. 10 plots the simulated and measured radiation patterns at three individual frequencies, i.e., 3.5, 6.0 and 10.0 GHz, and they are in reasonably good agreement with each other. At the high-end frequency near 10.0 GHz, the H-plane radiation pattern is split into a few radiation beams and the cross polarization in the E-plane pattern is increased. It is primarily attributed by the fact that the slotline resonator becomes electrically large at high frequencies and many other high-order resonant modes are excited in the wide slotline resonator. This cross polarization can be decreased at high frequencies by narrowing the radiating slot at the cost of narrow impedance bandwidth. This undesired phenomenon can not be abolished in physics and also exists in many other UWB antennas, e.g., a planar monopole antenna [18]. IV. CONCLUSION In this communication, a novel UWB microstrip-fed slotline antenna has been presented, designed and implemented under the concept of multiple-mode resonance of a single resonator. Our extensive study has rstly exhibited that its wide bandwidth of 3:1 or 100% is realized by concurrently exciting the rst four resonant modes of a single slotline radiator that is fed by the microstrip line at a proper offset position. With reference to the 160-
port impedance, an initial slotline antenna is designed to achieve a wide operating band with a fractional bandwidth of 97.9%. By installing a tapered-line impedance transformer, an actual 50-
microstrip-fed UWB slotline antenna has been developed by using four excited resonant modes. Measured results of the proposed antenna are in good agreement with the predicted ones and they have demonstrated a 97.9% operating bandwidth, over which the VSWR is less than 2.3 and the antenna gain varies from 0.5 to 4.0 dBi, as like other UWB antennas.

[10] W. J. Lui, C. H. Cheng, and Y. Cheng, A novel broadband multislot antenna fed by microstrip line, Microw. Opt. Technol. Lett., vol. 45, no. 4, pp. 5557, Apr. 2005. [11] J. Langley, P. Hall, and P. Newham, Novel ultra-wideband vivaldi antenna with low cross polarization, Electron. Lett., vol. 29, no. 23, pp. 20042005, 1993. [12] K. Ryu and A. Kishk, UWB dielectric resonator antenna with low cross-polarization, in Proc. IEEE Radio and Wireless Symp., Jan. 2010, pp. 552554. [13] L. Zhu, S. Sun, and W. Menzel, Ultra-wideband (UWB) bandpass lters using multiple-mode resonator, IEEE Microw. Wireless Compon. Lett., vol. 15, no. 11, pp. 796798, Nov. 2005. [14] G. Robert-Pierre Mari, Wide Band Slot Antenna, U.S. Patent 3 031 665, Apr. 24, 1962. [15] Y. Liu, Z. Shen, and C. L. Law, A compact dual-band cavity backed slot antenna, IEEE Antennas Wireless Propag. Lett., vol. 5, no. 1, pp. 46, 2006. [16] K. C. Gupta, R. Garg, I. Bahl, and P. Bhartia, Microstrip Lines and Slotlines, 2nd ed. Norwod, NJ: Artech House, 1996. [17] G. Matthaei, L. Young, and E. M. T. Jones, Microw. Filters, Impedance-Matching Networks, and Coupling Structures. Norwood, MA: Artech House, 1980. [18] S. H. Choi, J. K. Park, S. K. Kim, and J. Y. Park, A new ultra-wideband antenna for UWB applications, Microw. Opt. Technol. Lett., vol. 40, no. 4, pp. 399411, May 2004.

A Compact Hepta-Band Loop-Inverted F Recongurable Antenna for Mobile Phone

Yue Li, Zhijun Zhang, Jianfeng Zheng, Zhenghe Feng, and Magdy F. Iskander
AbstractA folded loop-inverted F recongurable antenna for mobile phone applications is designed and the obtained results are discussed in this communication. It is shown that loop antenna mode and an inverted F antenna (IFA) mode are controlled by only one p-i-n diodes with simple bias circuit. The impedance can be matched by adopting a matching bridge for both the loop and IFA modes. In a compact volume of 60 5 5 mm , the proposed antenna operates in hepta-band, including GSM850, GSM900, GPS, DCS, PCS, UMTS and WLAN, with the return loss lower than 6 dB. A prototype of the proposed antenna is fabricated, measured, and obtained results including return loss, efciency and gain, are presented. Index TermsHandset antennas, impedance matching, Mobile antennas, multiple band antennas, recongurable antennas.

REFERENCES
[1] Y. Yoshimura, A microstripline slot antenna, IEEE Trans. Microw. Theory Tech., vol. 20, no. 11, pp. 760762, Nov. 1972. [2] J. D. Kraus and R. J. Marhefka, Antennas: For All Applications Third Edition. New York: McGraw-Hill, 2002. [3] L. Zhu, R. Fu, and K. L. Wu, A novel broadband microstrip-fed wide slot antenna with double rejection zeros, IEEE Antennas Wireless Propag. Lett., vol. 2, pp. 194196, 2003. [4] N. Behdad and K. Sarabandi, A multiresonant single-element wideband slot antenna, IEEE Antennas Wireless Propag. Lett., vol. 3, pp. 58, 2004. [5] N. Behdad and K. Sarabandi, A wide-band antenna design employing a ctitious short circuit concept, IEEE Antennas Wireless Propag. Lett., vol. 53(II), no. 1, pp. 475482, Jan. 2005. [6] M. Kahrizi, T. K. Sarkar, and Z. A. Maricevic, Analysis of a wide radiating slot in the ground plane of a microstrip line, IEEE Trans. Microw. Theory Tech., vol. 41, no. 1, pp. 2937, Jan. 1993. [7] Y. W. Jang, Broadband cross-shaped microstrip-fed slot antenna, Electron. Lett., vol. 36, pp. 20562057, 2000. [8] Y. F. Liu, K. L. Lau, Q. Xue, and C. H. Chan, Experimental studies of printed wide slot antenna for wide-band applications, IEEE Antennas Wireless Propag. Lett., vol. 3, pp. 273275, 2004. [9] J. Y. Sze and K. L. Wong, Bandwidth enhancement of a microstripline-fed printed wide-slot antenna, IEEE Trans. Antennas Propag., vol. 49, no. 7, pp. 10201024, Jul. 2001.

I. INTRODUCTION With the rapid development of wireless communication technology, multi-band and multi-functional antennas are widely studied and adopted for the mobile phone applications. There is a signicant demand to integrated more wireless services to small volume mobile handsets. Wide bandwidth and compact structure are important
Manuscript received May 09, 2011; revised June 22, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work is supported in part by the National Basic Research Program of China under Contract 2010CB327402, in part by the National High Technology Research and Development Program of China (863 Program) under Contract 2009AA011503, the National Science and Technology Major Project of the Ministry of Science and Technology of China 2010ZX03007-001-01, and in part by Qualcomm Inc. Y. Li, Z. Zhang, J. Zheng, and Z. Feng are with State Key Laboratory on Microwave and Digital Communications, Tsinghua National Laboratory for Information Science and Technology, Department of Electronic Engineering, Tsinghua University, Beijing 100084, China (e-mail: zjzh@tsinghua.edu.cn). M. F. Iskander is with HCAC, University of Hawaii at Manoa, Honolulu, HI 96822 USA (e-mail: iskander@spectra.eng.hawaii.edu). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167949

0018-926X/$26.00 2011 IEEE

390

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

requirements for the internal multi-band antenna design, such as monopole antennas [1], [2], IFA [3][5], as well as the many other designs described in [6][13]. Recently, a number of publications, however, suggested that the loop antennas could also provide promising solution for mobile phone applications [6][13]. To arrange a loop antenna in a small volume with good impedance matching, however, is a challenging task as discussed in many publications. A tuning pad is utilized as described in [8], [9]. The tuning pad shows different effects to different operating modes of the loop and higher modes of operation can be tuned for wider bandwidths. However, integrating more frequency bands to an internal antenna with small volume continues to be still a challenge. Frequency recongurable antenna represents an effective solution for multi-band antenna designs. Different operating modes with different frequencies are recongured using the same antenna pattern and without extra space. Wide bandwidth is achieved by combing the overall working frequency bands. In this communication, a folded loop antenna is presented for mobile phone applications. The proposed antenna can be recongured to IFA by breaking the loop, using only one p-i-n diode and with a simple bias circuit. Besides adopting the tuning pad proposed in [8], [9], a matching bridge is also utilized in both loop and IFA modes of the proposed antenna to achieve wider bandwidth. Hepta-band coverage is realized in a compact volume of 60 25 2 5 mm3 , including GSM850, GSM900, GPS, DCS, PCS, UMTS and WLAN. Prototype was also built and the measured results show acceptable radiation efciency and gain of the proposed antenna, even after considering the effect of p-i-n diode. II. ANTENNA DESIGN AND CONFIGURATION Fig. 1 shows the geometry of the proposed recongurable antenna, the overall dimensions are 60 2 5 2 5 mm3 . The folded loop antenna (uniform width of 1 mm) is supported by foam with the permittivity close to that of air. The antenna structure is arranged just outside the ground plane area and above the main board. The main board is made of FR4 substrate ("r = 4:4; tan  = 0:01), with the thickness of 0.8 mm. A 100 3 60 mm2 metal ground is printed on the backside of the main board, connecting the shorting point of the loop through a via-hole. A 50
microstrip line is arranged on the front side of main board and is connected to feeding point of the loop. A shorting bridge (blue color in Fig. 1), which has the same width as the loop (1 mm), connects the feeding point with the shorting point. The position of p-i-n diode is illustrated in Fig. 1(b). When the p-i-n diode is ON, the antenna works in a typical loop mode. When the p-i-n diode is OFF, on the other hand, the loop will be broken into two IFA, and the left part (the one without tuning pad) is the main working branch. As a result, the operating modes of the proposed antenna can be switched by the state of p-i-n diode. The measured return losses of both working modes are shown in Fig. 2. These data agree well with simulation results as will be shown in Section III. Hepta-band is covered with 06 dB (VSWR = 3 : 1) by combining the bandwidths of two modes. For the loop mode (solid line), the achieved bands are 790870 MHz and 14902225 MHz, covering GSM850, GPS, DCS, PCS and UMTS bands. For the IFA mode (dash dotted line), the achieved bands are 845980 MHz and 22402565 MHz, covering GSM900 and WLAN bands. Analysis on the impedance matching of both modes is described in Section III. III. WORKING MODE ANALYSIS A. Loop Mode Typical loop antennas and their applications in the mobile phones area are systematically discussed by Wong [7][10]. Three resonant modes, including 0.5-wavelength mode, one-wavelength mode and 1.5-wavelength mode, are usually utilized. The tuning pad [8], [9] is a good method for impedance matching and with different effects to different modes. The 0.5-wavelength mode is tuned for lower band, and the one-wavelength and 1.5-wavelength modes are tuned together to cover the higher bands. In order to achieve wider bandwidth, a matching bridge is added between the feeding and shorting points
Fig. 1. Geometry and dimensions of the proposed antenna. (a) 3-D view. (b) Detailed dimensions in the planar view.

Fig. 2. Measured return loss of loop and IFA modes of the proposed antenna.

of the loop as shown in Fig. 1. The simulated return loss of the loop mode with and without the matching bridge is illustrated in Fig. 3. Simulations were made using the Ansoft High Frequency Structure Simulator (HFSS) software. As it may be seen, the inclusion of the matching bridge resulted in improving the bandwidth, thus covering the GSM850, GPS, DCS, PCS and UMTS bands. The matching method is shown on the Smith Chart of Fig. 4. The matching bridge works as a shunt inductor at the feeding point of loop. As studied in [14], a shunt inductor is able to move the impedance curve alone the equal admittance circle. The susceptance introduced by shunt inductor is 1=j ! L, where ! is angular frequency and L is the equivalent inductance of the matching bridge. For the lower band, the susceptance is larger than that of the high band. As a result, the impedance curve moves further away from the matching center and towards the lower frequencies. Similar to the tuning pad in [8], [9], the matching bridge also has different effect that depends on the frequency. For the lower band, shown in Fig. 4(a), the frequency shifts to lower band with similar bandwidth. However, more band moves into the VSWR 3:1 circle is shown in Fig. 4(b) and as a result the bandwidth is enhanced.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

391

Fig. 3. Simulated return loss of loop mode of the proposed antenna with or without the matching bridge.

Fig. 5. Simulated return loss of IFA mode of the proposed antenna with or without the matching bridge.

Fig. 6. Photo of the proposed antenna with bias circuit.

in Fig. 5. For the dash dotted line, we can see two resonant frequencies appearing in both lower band and high band, but with unacceptable bandwidth. The narrow bandwidths for two monopoles are mainly due to the compact volume without extra matching structure. For the two IFA results shown in the solid line, one of the two former resonant frequencies has been matched by adding the matching bridge. Therefore, two more bands of GSM950 and WLAN have been achieved based on the matched loop mode. IV. ANTENNA FABRICATION AND MEASUREMENT RESULTS To demonstrate the validity of the presented matching strategy, the proposed antenna with p-i-n diode was fabricated and tested, as shown in Fig. 6. The detailed diagram of bias circuit of p-i-n diode is shown in Fig. 7. The selected p-i-n diode is Philips BAP64-03 silicon PIN diode, with good performance up to 3 GHz [15]. When the p-i-n diode is forward-biased, it works as a series resistance. In the frequency band of 0.52.5 GHz, the insertion loss introduced by p-i-n diode is 0.10.2 dB at its typical bias current of 10100 mA. When the p-i-n diode is reverse-biased, it is equivalent to a series capacitance of approximately 0.45 pF, the isolation in the required band is better than 015 dB. Therefore, less number of p-i-n diodes will reduce the insertion loss and improve the performance of the systems. One is the minimum number of diode for two switchable states. In the bias circuit, a capacitor (Cb1 ) is used between the port and loop antenna for DC blocking; another DC block capacitor (Cb2 ) is used between the feeding and shorting points; an inductor (Lb ) is used for RF choking; another capacitor (Cs ) is used between Vcontr: and the ground in order to short the RF signal leaked from Lb ; a resistor (Rb ) is used to control the bias current. The bias voltage for Vcontr: is 3 V, supplied by two AA batteries. The values of each component in the bias circuit are: Cb1 = 120 pF, Cb2 = 120 pF,

Fig. 4. Smith Chart of the loop mode with or without the matching bridge. (a) Lower band. (b) Higher band.

B. IFA Mode In order to cover more bands by the internal antenna in the same volume, another mode is developed based on the loop structure. By cutting the loop, the antenna can be divided to two IFAs with the matching bridge, or two monopoles without the matching bridge. The comparison of return loss between two IFAs and two monopoles are illustrated

392

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

measured and shown in Fig. 9. For the GSM band, the gain varies in the range of 0.221.10 dBi; for the GPS, DCS, PCS, and UMTS bands, the gain ranges from 0.43 dBi to 3.13 dBi; and for the WLAN band, gain of 2.122.41 dBi is achieved. The results indicate the performance improvement by adopting switching mechanism. V. CONCLUSION A folded loop antenna with recongurable characteristic is proposed in this communication for multiband applications for mobile phones applications. Two strategies are designed for bandwidth enhancement. The rst one is the implementation of the loop-IFA modes switch. The IFA mode is achievedby cutting the loop structure without other changes. Another two resonant frequencies are added for wider bandwidth. Only one p-i-n diode is utilized for modes switching, thus avoiding the complex bias circuit and additional insertion losses. The second one is the matching bridge between feeding and shorting point. For loop mode, different effects take place for lower and higher bands; but for IFA mode, both lower and higher modes are matched. By adopting the matching bridge, the bandwidths of two modes are both improved. As a result, for a typical mobile phone with the ground size of 100*60 mm2 , hepta-band coverage is achieved in the compact volume of 60 2 5 2 5 mm3 , including GSM850, GSM900, GPS, DCS, PCS, UMTS and WLAN bands. The proposed antenna shows good radiation performance including the efciency and gain. The insertion loss of p-i-n diode is also discussed. High quality switch is a promising solution for recongurable antenna designs for mobile phone applications.

Fig. 7. Diagram of bias circuit of p-i-n diode.

Fig. 8. Measured radiation efciency of loop and IFA modes of the proposed antenna.

REFERENCES
[1] F. H. Chu and K. L. Wong, Simple folded monopole slot antenna for penta-band clamshell mobile phone application, IEEE Trans. Antennas Propag., vol. 57, pp. 36803684, Nov. 2009. [2] K. L. Wong and S. C. Chen, Printed single-strip monopole using a chip inductor for penta-band WWAN operation in the mobile phone, IEEE Trans. Antennas Propag., vol. 58, pp. 10111014, Mar. 2010. [3] C. H. Wu and K. L. Wong, Ultrawideband PIFA with a capacitive feed for penta-band folder-type mobile phone antenna, IEEE Trans. Antennas Propag., vol. 57, pp. 24612464, Aug. 2009. [4] Y. Li, Z. Zhang, W. Chen, Z. Feng, and M. F. Iskander, A quadband antenna with recongurable feedings, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 10691071, 2009. [5] K. R. Boyle and P. G. Steeneken, A ve-band recongurable PIFA for mobile phones, IEEE Trans. Antennas Propag., vol. 55, pp. 33003309, Nov. 2007. [6] Y. S. Shin and S. O. Park, A compact loop type antenna for Bluetooth, S-DMB, Wibro, WiMax, and WLAN applications, IEEE Antennas Wireless Propag. Lett., vol. 6, pp. 320323, 2007. [7] Y. W. Chi and K. L. Wong, Internal compact dual-band printed loop antenna for mobile phone application, IEEE Trans. Antennas Propag., vol. 55, pp. 14571462, May 2007. [8] K. L. Wong and C. H. Huang, Printed loop antenna with a perpendicular feed for penta-band mobile phone application, IEEE Trans. Antennas Propag., vol. 56, pp. 21382141, July 2008. [9] Y. W. Chi and K. L. Wong, Compact multiband folded loop chip antenna for small-size mobile phone, IEEE Trans. Antennas Propag., vol. 57, pp. 37973803, Dec. 2008. [10] Y. W. Chi and K. L. Wong, Quarter-wavelength printed loop antenna with an internal printed matching circuit for GSM/DCS/PCS/UMTS operation in the mobile phone, IEEE Trans. Antennas Propag., vol. 57, pp. 25412547, Sep. 2009. [11] C. H. Ku, H. W. Liu, and S. Y. Lin, Folded dual-loop antenna for GSM/DCS/PCS/UMTS mobile handset applications, IEEE Antennas Wireless Propag. Lett., vol. 9, pp. 9981011, 2010. [12] K. L. Wong, W. Y. Chen, and T. W. Kang, On-board printed coupled-fed loop antenna in close proximity to the surrounding ground plane for penta-band WWAN mobile phone, IEEE Trans. Antennas Propag., vol. 59, pp. 751757, Mar. 2011. [13] D. H. Lee, A. Chauraya, Y. Vardaxoglou, and W. S. Park, A compact and low-prole tunable loop antenna integrated with inductors, IEEE Antennas Wireless Propag. Lett., vol. 7, pp. 621624, 2008. [14] Y. Li, Z. Zhang, W. Chen, Z. Feng, and M. F. Iskander, A switchable matching circuit for compact wideband antenna designs, IEEE Trans. Antennas Propag., vol. 58, pp. 34503457, Nov. 2010. [15] Datasheet of BAP64-03 Silicon PIN Diode [Online]. Available: http:// www.nxp.com/documents/data_sheet/BAP64-03.pdf

Fig. 9. Measured gain of loop and IFA modes of the proposed antenna.

Lb = 120 nH, Cs = 470 pF and Rb = 46

with bias current is 65

mA. The bias voltage is controlled by a single-pole two-throw switch on the back side of ground plane. The performance of the proposed prototype antenna is measured, including the return loss, radiation pattern, efciency, and gain. The measured return loss of both two modes is shown in Fig. 2. By combining the two operation modes, the proposed antenna covers the bands of GSM850, GSM900, GPS, DCS, PCS, UMTS and WLAN with the return loss lower than 06 dB. Especially for GPS operation, linear polarization antenna is adopted instead of circularly polarization antenna in the mobile phone applications, for its unexpected orientation and compact dimension. Circularly polarization antenna is usually utilized in the base station and vehicular environment with a xed orientation and unlimited volume. The measured radiation efciencies for loop and IFA modes are shown in Fig. 8. By combing the curves of two modes, the improvement of efciency is clearly observed. For the GSM band, the efciency is better than 64.7%; for the GPS, DCS, PCS, and UMTS bands, the efciency is better than 47.4%; and for the WLAN band, 62.8% efciency is achieved. The efciency can be improved by using high quality diodes in the practical applications. The gain is also

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

393

Hybrid Monopole-DRAs Using Hemispherical/ Conical-Shaped Dielectric Ring Resonators: Improved Ultrawideband Designs
Debatosh Guha, Bidisha Gupta, and Yahia M. M. Antar

AbstractTwo geometries of dielectric ring resonator (DRR) have been investigated as new variants for designing ultrawideband hybrid monopoleDRAs (dielectric resonator antennas) with improved features. They are simply hemispherical and conical in shape and have been implemented individually. Hybrid monopole-DRA is actually a combined conguration using a grounded monopole surrounded by a small DRR touching the same ground plane. Compared to an earlier version using cylindrical DRR of same dielectric material, the proposed ones promise for about 25% larger impedance bandwidth maintaining identical gain and radiation properties. As much as 126% impedance bandwidth with consistent monopole type radiation and 24 dBi peak gain has been demonstrated using a set of proto. Physical insight in to the types shaped from dielectric rod with ultrawideband operation along with a comprehensive design guideline is presented.

= 10

Index TermsDielectric resonator antenna, hybrid dielectric resonator antenna, ultrawideband (UWB) dielectric resonator antenna, UWB monopole antenna.

I. INTRODUCTION A hybrid conguration of an electric monopole (EMP) and a dielectric ring resonator (DRR) using a common ground plane (GP) plays signicant role in designing an ultrawideband monopole antenna. This was rst indicated by a Canadian research group in 2002 [1]. They experimentally studied a simple geometry of a quarter wave monopole surrounded by a small annular-shaped DRR [1]. Its optimized design, obtained through simulated data, showed possibility of achieving ultrawide impedance bandwidth (100% with S11 < 010 dB), which was reported [2] and patented [3] in 2005. Physical insight into the radical change in impedance behavior of a monopole caused by a dielectric resonator loading was developed in 2006 [4]. An efcient design guideline, described in [4], attracted different researchers in developing newer geometries [5][7]. In [5], Yagi-like structure has been explored showing 87% impedance bandwidth. The shape of the monopole was changed to T-like structure in [6] to enhance the bandwidth up to 112% and a contemporary investigation [7] used inverted cone-shaped DRR along with modied ground plane geometry to obtain almost similar impedance bandwidth. Top-loading using a metallic disk on top of a coax-fed DRR was experimentally examined earlier [8] to explore tunability of monopole-like DRA. A different conguration using composite dielectric bodies was examined [9] to improve impedance bandwidth of monopole-type DRA.
Manuscript received March 25, 2011; revised June 08, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. D. Guha is with the Institute of Radio Physics and Electronics, University of Calcutta, Kolkata 700 009, India (e-mail: dguha@ieee.org). B. Gupta is with the ETE Department, Girijananda Chowdhury Institute of Management and Technology, Guwahati 781017, India (e-mail: gupta_bidisa@yahoo.co.in). Y. M. M. Antar is with the Electrical Engineering Department, Royal Military College of Canada, Kingston, ON K7K 7B4, Canada (e-mail: antar-y@rmc.ca). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167948

Recently a DRR of pawn-like shape, realized as a combination of two geometries: a hemispherical DRR (HDRR) placed on top of a conical DRR (CoDRR), has been investigated [10]. The aim was to improvise the DRR shape to achieve favorable interactions and adjustment of electromagnetic (EM) elds between the DRR and the EMP. Compared to a cylindrical DRR [1][4], a hemisphere, by virtue of its shape, offers smoother transition of the domain of EM interactions. The geometry of a cone also offers a smooth tapering in the interaction zone. Instead of using the combination, either a hemispherical or a conical DRR alone is found to be efcient enough to provide similar ultrawide operating bandwidth as is recently indicated by the present authors [11]. From practical point of view, the proposed geometries are important as they involve simpler DRR shapes, which are easier to design and fabricate employing lesser amount of material, machining, cost and weight. This communication addresses thorough systematic investigation leading to optimized designs of both the antenna geometries. They are conrmed through measurements using two sets of prototypes. Prior to fabrication, tolerance of critical DRR parameters have also been examined. Occurring of multiple resonances has been investigated leading to a clear physical insight into individual resonant modes. The study helps in proposing an efcient design guide line. Reliability of the proposed design has also been veried. About 120126% impedance bandwidth with consistent monopole-like radiation over the entire operating band has been experimentally demonstrated. As much as 4 dBi peak gain is experimentally obtained. Proposed antenna geometries, their characteristics, designs, and experimental studies are presented.

II. HYBRID MONOPOLE-DRA GEOMETRIES Proposed hybrid congurations of monopole DRAs are shown in Fig. 1. A vertical monopole of length l is tted with a circular ground plane of diameter d. A hemispherical dielectric resonator (relative permittivity "r , radius a) is drilled through its center to form a cylindrical hole of radius b. Thus a hemispherical dielectric ring is formed and this is placed on the ground plane surrounding monopole. The spacing between monopole surface and dielectric body is determined by the parameter s = b 0 r; r being the radius of the monopole. The geometry is shown through Figs. 1(a) and (b). The cross sectional view of conical DRA geometry is shown in Fig. 1(c). The radius of the base of the cone is a and that of the central cut is b. The height of the cone may be determined in terms of a and b parameters as a2 =(a 0 b), which after making the central hole reduces to effective height h  a. The top views of both the geometries are identical. The centrally cutout hole in either dielectric body not only provides a space for accommodating the vertical monopole, but also plays a signicant role in coupling the electromagnetic elds between the monopole and dielectric resonant structure.

III. HYBRID MONOPOLE USING HEMISPHERICAL DRR (HDRR) This investigation was started with design parameters as used in [4] and just replacing cylindrical DRR by a hemispherical DRR. Then a thorough parametric study has been performed to achieve optimum operating bandwidth and also to examine the fabrication tolerance of the practical structure. A physical insight into the resonances of the hybrid structure resulting in ultrawide impedance bandwidth has been developed. Detailed discussions along with a set of representative results are provided below.

0018-926X/$26.00 2011 IEEE

394

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Simulated S vs frequency of hybrid monopole DRAs. " = 10; a = 4:2; b = 1:3; l = 10; r = 0:65 and height of cylindrical DRR [4] = 4:4 (all dimensions in mm).

Fig. 1. Schematic diagrams of proposed hybrid monopole-DRAs. (a) Top view for either structure. (b) cross-sectional view for HDRR-loaded monopole. (c) cross-sectional view for CoDRR-loaded monopole. Fig. 3. Simulated S versus frequency of a hybrid monopole using HDRR with varying radius. b = 1:3 mm, l = 10 mm, r = 0:65 mm, Parameters as in Fig. 2.

A. Parametric Studies and Optimized Design Fig. 2 compares two simulated S11 versus frequency for an electric monopole (l = 10 mm and r = 0:65 mm) when it is successively loaded with a cylindrical and a hemispherical DRR, respectively, both being made up of identical material ("r = 10) and of same radius a = 4:2 mm. The monopole using HDRR shows about 12% extra impedance bandwidth indicating a promise for much improved optimized design. This indeed needs optimization of antenna parameters as provided below. Unlike cylindrical DRA, the hemispherical structure needs only one dimensional parameter a and its effect on the input impedance is examined in Fig. 3. For a set of antenna parameters, indicated in Fig. 3, the values of a varying between 4.55.0 mm appear to be the best choice from bandwidth point of view. Fabrication tolerance of about 0.5 mm is evident. Taking a = 5 mm, the monopole has been optimized in Fig. 4. The length of the monopole determines its resonances and details about this will be discussed in Section III.C. However for the present conguration, l = 10 to 11 mm appears to be optimum choice. The coupling between the monopole and the DRR is controlled by the parameter s or b (= r + s) related to the cutout cylinder. Larger values of b deteriorate the performance. Its optimum choice is found to be in between 1.5 mm and 2 mm for the present case. The parameter b(= r + s) is also examined and any value of b lying between 1.3 and 2.0 mm is found to work well indicating acceptable tolerance in manufacturing the device. Probe radius r = 0:65 mm is considered as a standard one for commercial SMA probes. Variation of r value up to about 50% is acceptable.

Fig. 4. Simulated S versus frequency for different probe length l; a = 5 mm, other parameters as in Fig. 2.

The monopole-DRA radiates vertically polarized elds, which is azimuthally symmetric. Its radiation characteristics over elevation angles are discussed later. The UWB antenna radiates consistently over the entire bandwidth, which is examined in Fig. 5. Maximum gain occurring around 30 70 away from bore sight are plotted against frequency. It

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

395

Fig. 5. Maximum gain versus frequency for optimum HDRA and monopole ;a ;r : ;b : ;d ;l , (all parameters. " dimensions in mm).

= 10 = 5 = 0 65 = 1 82 = 70 = 11

Fig. 7. Measured return loss characteristic of the prototype shown in Fig. 6.

Fig. 6. Prototype fabricated from Eccostock HiK material with " : ;r : ;b : ;l ;d (all dimensions in mm).

5 04 = 0 65 = 1 82 = 11 = 70

= 10. a =
Fig. 8. Measured

varies from 3.7 to 6.0 dBi. Two measured data points are also included in Fig. 5. B. Experimental Studies and Verication A hemispherical dielectric resonator has been shaped from Eccostock HIK material with "r = 10. A 70 mm diameter brass plate, centrally itted with a PE 4128 SMA probe, has been used as the monopole structure. The hybrid conguration of the prototype is shown in Fig. 6. The prototype has been measured using Agilents E8363B network analyzer to study its resonance characteristics. For the radiation measurements, Agilents E8257D signal generator (250 KHz40 GHz) has been used as a broadband source connected to the transmitting horns and Agilents E4418B power meter tted to the antenna under test for recording the received power. Fig. 7 shows measured S11 trace of the prototype with start and stop frequencies as 5 GHz and 25 GHz, respectively. Considerable impedance matching has been obtained over 521 GHz indicating more than 4:1 ratio and about 126.5% bandwidth (S11 < 010 dB). Fig. 8 compares this measured result with the simulated data obtained using [12]. The measurement closely follows the nature of the simulated curve, but shows a relative shift towards the lower side of the spectrum. Small amount of capacitive loading during its measurement is apparent, which causes shift in S11 values to the lower frequency. Sometimes it becomes unavoidable due to imperfections in some interconnects between commercial probes, SMA connectors and adapters used. Fig. 9 shows measured radiation patterns of the prototype obtained at two different frequencies in X- and Ku-bands, respectively. They closely agree with the simulated data, incorporated in the gure for comparison. Nearly 4 dBi peak gain is evident at either frequency. In

of the prototype of Fig. 6 compared with simulated data.

measurements, a large ground plane with d = 70 mm has been used, which is about 1.5 times the wavelength corresponding to the rst resonance (1 ). Smaller and also practical ground plane may be used and value of d as small as 0:61 is quite acceptable without causing degradation of radiation patterns, peak gain and impedance bandwidth. C. Physical Insight Into the Ultrawide Bandwidth Understanding the antenna behavior, particularly when a simple loading results in a dramatic change in the impedance matching over ultrawide frequency band is very important for various reasons. For the present HDRR loaded hybrid monopole, S11 versus frequency plot reveals four distinct minima as shown in Fig. 10. These indicate four resonances around 6.8, 13, 16 and 20 GHz, respectively. If we look at Fig. 2, one additional resonance due to HDRR is apparent. At the rst step, let us examine two more individual cases: (i) a standalone monopole (removing HDRR from Fig. 6) and (ii) a centrally fed HDRR (reducing the monopole height to 2.5 mm), respectively. Their individual resonance characteristics are incorporated in Fig. 10. The monopole alone shows two resonances near 6.5 and 19.5 GHz, respectively. These are corroborated by the theoretical calculations. The monopole with l = 11 mm should theoretically resonate with the fundamental mode (l = =4 near 6.8 GHz and then with the rst higher mode (l = 3=4) near 20.45 GHz. They indeed correspond to two extreme resonances due to the hybrid structure and one may, therefore, surmise that two extreme resonances in the monopole-DRA are caused by the monopole itself.

396

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 11. Simulated versus frequency for two hybrid monopole-DRA strucmm. tures. Parameters as in Fig. 6, CoDRA h

=5

Fig. 9. Measured radiation patterns of the prototype of Fig. 6 compared with simulated data. (a) 10 GHz, (b) 17 GHz.

Fig. 12. Prototype fabricated from Eccostock HiK material with " . h : ;a : ;r : ;b : ;l ;d (all dimensions are in mm).

= 5 01 = 5 045 = 0 65 = 1 825 = 11 = 70

= 10

It is important to note that the hemispherical geometry adds an additional resonance over and above [4] and results in about 2226% additional bandwidth. IV. HYBRID MONOPOLE USING CONICAL DRR (CODRR) Hybrid monopole-DRA using HDRR is examined above. If the HDRR is replaced by an identical (same radius and material) CoDRR, its impedance bandwidth remain exactly the same. This is shown in Fig. 11. Resonant characteristics as examined in Fig. 10 are also revealed. Fig. 12 shows a prototype, which uses the same monopole as in Fig. 6 and a physically realized CoDRR shaped from the same Eccostock HiK material. The measured S11 values are shown in Fig. 13 along with the simulated data. Close agreement between them is revealed. Measured co-polar radiation patterns obtained at 10 GHz are shown in Fig. 14. This shows peak gain values close to that in Fig. 9. V. COMMON DESIGN GUIDELINE AND VERIFICATION Based on the above studies and thorough understanding of the resonant modes, a design guideline is proposed which is valid for either DRR shape. The guideline provides a set of design parameters to start with and a designer should feel free to change or optimize some of the values to achieve desired or improved characteristics.

Fig. 10. Simulated reection coefcient of standalone monopole fed DRA and isolated probe-fed DRA. a l ; : mm, r : mm, " .

= 11 probe length = 2 5

= 0 65

= 10

= 5:01

monopole, mm,

The centrally fed hemispherical DRR shows a trace of resonance around 17 GHz. This is due to TM101 mode in the HDRR and corresponds to the third resonance (near 16 GHz) of the hybrid antenna. Finally we are left with the resonance occurring near 13 GHz. This is indeed caused by the monopole having effective length le < l. Electric elds in monopole are strongly coupled with the DRR resulting in restricted monopole current up to a certain height. Such an observation has been previously documented in [4]. In here, simulated portray gives an approximate estimate like le  0:5l = =4, which results in 13.6 GHz. This is very close to the actual value, i.e., 13 GHz.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

397

Fig. 13. Measured S data.

of the prototype of Fig. 12 compared with simulated

Fig. 14. Measured radiation patterns of the prototype of Fig. 12 compared with simulated data. f = 10 GHz.

Fig. 15. Theoretical designs provided in Table II veried using simulated S values. (a) Hybrid monopole using HDRA; (b) hybrid monopole using CoDRR.

Re(k a) OF THE TM

TABLE I MODE OF A HEMISPHERICAL DRA

Radius(r) : s  r  s=2; with 0:0191  s  0:0421

(2) (3)

where, s is the spacing between monopole and inner boundary of the DRR (Fig. 1). (iii) HDRR and CoDRR Parameters: Radius a can be determined following the design procedure [13] as

a(in cm) = 4:7713 2 Re(k0 a)=fr(in GHz)

(4)

A. Design Steps (i) Frequency of Operation: If the operating frequency is specied by its lower and upper values as fL and fH , respectively, then for the present hybrid monopole DRA, they are related as fH  4fL (ii) Dimension of Monopole: The rst S11 minimum relates the dominant mode resonance due to the monopole and it occurs near f1  1:25fL . Its corresponding wavelength is 1 ( = c=f1 , c being the velocity of light in free space). First higher mode due to monopole resonates at f2 = 3f1 .

where, fr = 0:5(f1 + f2 ). The quantity Re(k0 a), i.e., the real part of k0 a is a function "r [13] and some useful values for a set of "r s are provided in Table I. A designer, therefore, enjoys freedom to choose "r to realize an antenna, but values above 20 do not present a suitable choice. Higher "r causes higher Q for TM101 mode, which in turn degrades the antenna performance in terms of bandwidth as well as radiation. Height of CoDRR: h = a. B. Design Verications The above guideline is followed to determine the antenna parameters for different frequency ranges for a test. They are furnished in Table II for both HDRR and CoDRR loaded monopoles and are implemented through the simulation tool [12]. Resulting S11 versus frequency plot is shown in Fig. 15. The plots are self explanatory and conrm the reliability of the proposed design. However, design #2 in Fig. 15(b) needs further optimization since S11 value overshoots S11 < 010 dB limit by about 1.0 to 1.5 dB over a small range of frequency 1821.5 GHz.

Monopole Length (l) : l = 1 =4;

(1)

398

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE II HDRA/CODRA PARAMETERS DETERMINED USING DESIGN GUIDELINES " ;h a; d : 

= 10 =

06

[11] D. Guha, B. Gupta, and Y. M. M. Antar, Hybrid monopole-DRA: New geometries for improved ultra-wideband operation, presented at the IEEE Antennas and Propagation Symp., 2010. [12] Ansofts High Frequency Structure Simulator (HFSS) Version 11.1. [13] A. Petosa, Dielectric Resonator Antenna Handbook, 1st ed. Norwood, MA: Artech House, 2007, ch. 2.

A Half Maxwell Fish-Eye Lens Antenna Based on Gradient-Index Metamaterials

VI. CONCLUSION Hybrid monopole-DRAs explored in this communication are the most improved ones in terms of the operating bandwidth, ease of designing and fabrication. Either of the geometries can be used, but the feasibility in terms of their machining may be another aspect to choose the DRA shape. Both are equally suitable for their considerable fabrication tolerance. Surface nish of both the dielectric and metal bodies should be taken care of, particularly if the design frequency goes beyond X-band. Proposed design guideline should be useful in yielding practical antennas. Considerably compact monopole having nearly 126% or 4:1 operating bandwidth with average 4 dBi peak gain should nd a wide range of applications starting from wideband EM sensor to UWB communications. ACKNOWLEDGMENT The authors are thankful to the Reviewers for their constructive suggestions leading to improving the quality of the communication. They are also thankful to A. Banerjee, a member of this research group, for her unstinted help and cooperation. Zhong Lei Mei, Jing Bai, Tiao Ming Niu, and Tie Jun Cui

AbstractAn X-band half Maxwell sh-eye lens antenna is designed based on the effective medium theory and metamaterials technology. The non-resonant I-shaped metamaterials have been chosen to realize the lens antenna due to the broadband and low-loss features. The laboratory prototype has been fabricated and measured using a two-dimensional near-eld microwave scanning apparatus, where the experimental results agree with full-wave numerical simulations. Index TermsLens antenna, Maxwells sh-eye lens, metamaterials.

I. INTRODUCTION Metamaterials are composed of periodic or pseudo-periodic subwavelength structures with electric and/or magnetic responses. The electromagnetic (EM) parameters of metamaterials can be articially tailored with different unit-cell geometries and dimensions, rather than their intrinsically chemical components. Since the negative refraction was demonstrated experimentally in the microwave regime using composite metamaterials made of split-ring resonators and wires [1], metamaterials have attracted enormous attention in the science and engineering communities all over the world. Among various metamaterials proposed, gradient index (GRIN) metamaterials [2], [3] play an important role. These metamaterials usually work at the non-resonant region of the constitutive unit cells and hence are inherently broadband and less lossy. Due to such a feature, GRIN metamaterials have been used for the realization of traditional Luneburg lens [4], [5] and similar devices. Like the Luneburg lens, the Maxwell sh-eye (MFE) lens is a typical GRIN device proposed in 1854 [6]. This lens is usually given as an example of a geometric-optics (GO) perfect imaging instrument, which can
Manuscript received September 14, 2010; revised June 11, 2011; accepted July 02, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported in part by a Major Project of the National Science Foundation of China under Grants 60990320 and 60990324, in part by the 111 Project under Grant 111-2-05, and in part by the National Science Foundation of China under Grants 60871016, 60901011, and 60921063. The work of Z. L. Mei was supported by the Open Research Program Funds of State Key Laboratory of Millimeter Waves (No. K201115) and Chunhui Project (No. Z2010081). Z. L. Mei, J. Bai and T. M. Niu are with the School of Information Science and Engineering, Lanzhou University, Lanzhou 730000, China (e-mail: (meizl@lzu. edu.cn). T. J. Cui is with the State Key Laboratory of Millimeter Waves, Department of Radio Engineering, Southeast University, Nanjing 210096, China (e-mail: (tjcui@seu.edu.cn). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167914

REFERENCES
[1] A. Ittipiboon, A. Petosa, and S. Thirakoune, Bandwidth enhancement of a monopole using dielectric antenna resonator loading, in Proc. ANTEM Conf., Montreal, Canada, Aug. 2002, pp. 387390. [2] M. Lapierre, Y. M. M. Antar, A. Ittipiboon, and A. Petosa, Ultrawideband monopole/dielectric resonator antenna, IEEE Microw. Wireless Comp. Lett., vol. 15, no. 1, pp. 79, Jan. 2005. [3] A. Ittipiboon, A. Petosa, S. Thirakoune, D. Lee, M. Lapierre, and Y. M. M. Antar, Ultra Wideband Antenna, U.S. Patent no. 6940463 B2, Sep. 2005. [4] D. Guha, Y. M. M. Antar, A. Ittipiboon, A. Petosa, and D. Lee, Improved design guidelines for the ultra wideband monopole-dielectric resonator antenna, IEEE Antennas Wireless Propag. Lett., vol. 5, pp. 373377, 2006. [5] S. Thirakoune, A. Petosa, and A. Ittipiboon, Yagi-like DRA-loaded monopole, in Proc. Int. Conf. Electromagnetics in Advanced Applications ICEAA, 2007, pp. 891894. [6] S. Ghosh and A. Chakrabarty, Ultrawideband performance of dielectric loaded T-shaped monopole transmit and receive antenna/EMI sensor, IEEE Antennas Wireless Propag. Lett., vol. 7, pp. 358361, 2008. [7] M. N. Jazi and T. A. Denidni, A new hybrid skirt monopole dielectric resonator antenna, in Proc. IEEE Antennas and Propagation Society Int. Symp., Jul. 2008, pp. 14, Abstracts. [8] Z.-N. Chen, K.-W. Leung, K.-M. Luk, and E. K.-N. Yung, Effect of parasitic disk on a coaxial probe-fed dielectric resonator antenna, Microw. Opt. Technol. Lett., vol. 15, no. 3, pp. 166168, 1997. [9] S. H. Ong, A. A. Kishk, and A. W. Glisson, Wideband disc-ring dielectric resonator antenna, Microw. Opt. Technol. Lett., vol. 35, no. 6, pp. 425428, 2002. [10] D. Guha, B. Gupta, and Y. M. M. Antar, New pawn-shaped dielectric ring resonator loaded hybrid monopole antenna for improved ultrawide bandwidth, IEEE Antennas Wireless Propag. Lett., vol. 8, pp. 11781181, 2009.

0018-926X/$26.00 2011 IEEE

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

399

Fig. 1. The refractive index prole of the 2D metamaterial HMFE lens antenna. (a) Ideal distribution. (b) Discretized distribution.

transform a point source at the lens surface into a focus at the diametrically opposite side of the lens. However, due to the lack of signicant applications, MFE lens remains a theoretical model in the laboratory for a long time. In recent years, Fuchs et al. suggested to use half of the MFE lens as a highly directive antenna, which is termed half Maxwell sh-eye lens (HMFE). This gives MFE a nontrivial practical application in the antenna design [7][9]. In their implementations, the HMFE lens antennas are approximately made of concentrically layered media with different permittivity values, which is the major difference from our design. In this communication, we design a two-dimensional HMFE (2D) lens antenna using GRIN metamaterials. Section II presents the detailed design process using I-shaped unit cells [10], [11]. In Section III the lens antenna is implemented and tested in a 2D near-eld scanning apparatus [12]. We summarize the work in Section IV. II. LENS ANTENNA DESIGN The 2D HMFE lens is innitely long in the z direction, and the refractive prole in the xoy plane is the same as that of the equatorial plane in the MFE lens. As a result, the refractive index distribution is characterized by the following function of spatial position
n(r ) = n0 1 + (r=R)2

Fig. 2. Design of the 2D metamaterial HMFE lens antenna. (a) Unit cell geometry of the metamaterial lens antenna and the excitation setup. (b) The fabricated HMFE lens antenna.

(1)

where n0 represents the refractive index at the lens center and R is the radius of the MFE lens, while r denotes the distance from the center of the lens in the cylindrical system. In order to design a 2D HMFE lens antenna in the X band using metamaterials, we set the two constants in (1) as n0 = 2 and R = 48 mm, respectively. And thus the refractive index of HMFE lens varies radially from 2 at the center to 1 at the surface, as illustrated in Fig. 1(a). The lens is rst divided into 4-by-4 mm square grids, where refractive index is given by the analytic formula at the center of the grid. The approximate prole is shown in Fig. 1(b). The key problem is to map the refractive index at each point to metamaterial unit cells with different geometrical sizes. Due to the broadband and low-loss property, I-shaped unit cells are employed in our design. The detailed structure is demonstrated in Fig. 2(a) together with the excitation setup in the numerical simulation. In the design, F4B is chosen as the PCB substrate, whose permittivity is 2.65 with the loss tangent 0.001 at 10 GHz. The substrate thickness is 0.25 mm. The I-shaped copper pattern lies on one side of the substrate, with a thickness of 0.038 mm. Note that the side length of the unit cell is 4 mm, about 1/7.5 of the working wavelength, hence we can use the effective medium theory to analyze the problem. The geometrical mapping process contains three stages. First, the commercial software, Ansoft HFSS, is utilized to calculate the S scattering parameters for a single cell with a xed arm length using periodic boundary conditions. This process is repeated for different arm lengths, which varies from 0.8 to 3.5 mm in our design. Second, the well-accepted retrieval method is used to get the effective constitutive

Fig. 3. The relationship between the geometry of the I-shaped structure and the effective refractive index.

parameters corresponding to different as [13]. As a consequence, the polynomial expression for the effective refractive index versus a is determined using the curve-tting technique (shown in Fig. 3). Finally, the cell geometry for each grid is found based on the above polynomial expression using a root nding algorithm. It should be noted that the above process is not rigorously accurate since the designed antenna is not a real periodic structure. However, since the arm lengths between adjacent metamaterial cells vary gradually, the quasi-periodic model can be validated. And this has been conrmed by measured results [10][11]. For ease of fabrication and assembling, the unit cells on the same line parallel to the at interface are fabricated on a single strip of printed circuit board (PCB). The fabricated HMFE lens antenna is shown in Fig. 2(b). The radius of the HMFE lens is 48 mm with a thickness of 12 mm (3 unit cells in the z direction). The PCB strips are fabricated using the lithography technology, and the supporting frame is a handmade foam structure with permittivity close to one. Fig. 3 demonstrates the retrieved EM parameters for different arm lengths at 10 GHz. It can be seen that the refractive index varies from 1.06 to 2 when a changes from 0.8 to 3.44 mm. Therefore the I-shaped unit cells are sufcient for the realization

400

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

of the HMFE lens. Moreover, the imaginary part of the material parameters is approximately zero, indicating the very small loss. Note that in the simulation, we x p = 4 mm and w = 0:3 mm. III. RESULTS AND DISCUSSIONS In this section, we give the simulation and measurement results for the HMFE lens antenna. The simulations are made using the nite-element based commercial software, COMSOL MULTIPHYSICS, and the measurements are performed using a 2D eld-mapping equipment, the 2D-Mapper [12], which is actually a parallel-plate waveguide system. The height of the waveguide is about 13 mm, just enough to hold the lens antenna. The 2D point source (or line source) is implemented using a probe on the bottom side. The HMFE lens antenna is placed on the bottom metal plate, and another probe located on the top plate scans above the lens top surface to measure the near-eld distribution. In Fig. 4, we give both simulated and measured electric-eld distributions for the HMFE lens antenna. Fig. 4(a) shows the simulation result at 10 GHz, demonstrating the cylindrical waves passing through the designed structure. In the simulation, the lens is fed by a current line source near the surface, which is surrounded by perfectly matched layers (PMLs) in three directions except in the main radiation direction. The conguration is similar to our experimental setup. Undoubtedly, after passing through the lens, the circular wavefronts are attened and the cylindrical waves are transformed into quasi-plane waves, which suggests a highly directive radiation. The measured electric-eld distributions at 10 and 12 GHz are shown in Figs. 4(b) and (c) respectively. Similar transformation effects from cylindrical waves to plane waves are clearly observed. However, the attening effect is more obvious at higher frequencies, which implies a higher directivity and smaller main-lobe width. The discrepancies between simulation and measured data are mainly attributed to the non-ideal cell arrangement in the device. The foam structure fabricated by hand will also generate some size deviation in the square grids. Since the effective refractive index of each cell is closely related to the geometrical size, the refractive index distribution of the HMFE lens antenna may not be strictly satised. This will lead to the small change of the eld patterns. Besides, treating the quasi-periodic structure as a periodic one can also produce some errors. Moreover, in our experimental setup, there is an air gap between the top metal plate, the surface of the antenna structure and the microwave absorbing materials, which will cause energy leakage in the ambient environment, and this is more obvious in directions other than the peak radiation direction. We remark that the measured results at 8, 9 and 11 GHz (not shown here for the sake of clarity) illustrate similar near-eld patterns. Based on the near-far-eld transformation [11], [14], the far-eld radiation patterns have been calculated using the measured data. As shown in Fig. 5, in the peak radiation direction, the simulated data for 10, 11 and 12 GHz (red, green and blue solid curves) almost coincide. The side lobes are all below 014 dB. For the measured data, the 10 GHz curve (red dash-dotted curve) has the largest main beam width and can be clearly identied, while others (green dashed and blue solid and dotted curves for 11 and 12 GHz) agree well with the simulated results in the peak direction. Even for the 10 GHz data, the disagreement above 05 dB is still very small. The maximum directivity of antenna is also calculated using the expression given in [15],

Fig. 4. The near-eld distributions of electric elds inside and outside the metamaterial HMFE lens antenna. (a) The simulation result at 10 GHz, in which the HMFE lens is surrounded by PMLs. (b),(c) The measurement results at 10 and 12 GHz. Unit: mm.

D0

= 4

where F (; ) is the radiation intensity. For the simulation case at 10 GHz, the computed directivity equals 29 dB. Whereas for the measured result, the directivity at 10, 11 and 12 GHz is approximately 23, 25 and

2  0 0

F (; )jmax F (; ) sin  d d

(2)

26 dB, respectively. According to our understanding, two factors exist to determine the performance of the designed antenna, one is the effective medium theory, which requires longer working wavelength and is approximately met in our design; the other is the theory of geometrical optics, which demands shorter working wavelengths and is the cornerstone for the HMFE. According to previous work from our and other groups [10], [11], the unit cell applied in the design works well in the X band. This means the refractive index prole can be approximately met in this band too. Hence the GO theory dominates in this case, which prefers higher frequencies. As a result, the ideal eld distribution at the lens aperture (e.g., the simulation result) is more accurately realized at higher frequencies. And this directly leads to better far-eld performances, as is shown in the gure. Qualitatively speaking, the directivity for aperture antennas is inversely proportional to the square of the wavelength [15], and this also explains the measurement result in Fig. 5. It is shown that the HMFE lens antenna has a good directive behavior. However, comparing the simulated and measured far-eld radiation patterns, large disparities are observed around side lobes. As mentioned earlier, this is attributed to the eld leakage in the experimental setup. In our simulations, PMLs are used to enclose the point source except for the peak direction. While in the measurement, we use microwave absorbing materials. Since the height of the lens antenna is

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

401

ping device and the measured results have good agreement to numerical simulations.

REFERENCES
[1] R. A. Shelby, D. R. Smith, and S. Schultz, Experimental verication of a negative index of refraction, Science, vol. 292, pp. 7779, Apr. 2001. [2] D. R. Smith, J. J. Mock, A. F. Starr, and D. Schuring, Gradient index metamaterials, Phys. Rev. E, vol. 71, no. 3, p. 036609, Mar. 2005. [3] T. Driscoll, D. N. Basov, A. F. Starr, P. M. Rye, S. Nemat-Nasser, D. Schuring, and D. R. Smith, Free-space microwave focusing by a negative-index gradient lens, Appl. Phys. Lett., vol. 88, no. 8, p. 081101, Feb. 2006. [4] H. F. Ma, X. Chen, X. M. Yang, H. S. Xu, Q. Cheng, and T. J. Cui, A broadband metamaterial cylindrical lens antenna, Chinese Sci. Bull., vol. 55, no. 19, p. 2066, 2010. [5] Q. Cheng, H. F. Ma, and T. J. Cui, Broadband planar Luneburg lens based on complementary metamaterials, Appl. Phys. Lett., vol. 95, no. 18, p. 181901, Nov. 2009. [6] J. C. Maxwell, Solutions of Problems, problem no. 2, Math J., vol. 8, p. 188, 1854. [7] B. Fuchs, Q. Lafond, S. Rondineau, M. Himdi, and L. L. Coq, Design and characterisation of half-Maxwell sh-eye lens antenna in 7681 GHz band, Electron. Lett., vol. 42, no. 5, pp. 261263, Mar. 2006. [8] B. Fuchs, O. Lafond, S. Rondineau, and M. Himdi, Design and characterization of half Maxwell sh-eye lens antennas in millimeter waves, IEEE Trans. Microw. Theory Tech., vol. 54, no. 6, pp. 22922300, Jun. 2006. [9] B. Fuchs, O. Lafond, S. Palud, L. L. Coq, M. Himdi, M. C. Buck, and S. Rondineau, Comparative design and analysis of Luneburg and half Maxwell sh-eye lens antenna, IEEE Trans. Antennas. Propag., vol. 56, no. 9, pp. 30583062, Sep. 2008. [10] R. Liu, C. Ji, J. J. Mock, J. Y. Chin, T. J. Cui, and D. R. Smith, Broadband ground-plane cloak, Science, vol. 323, pp. 366368, Jan. 2009. [11] H. F. Ma, W. X. Jiang, X. M. Yang, X. Y. Zhou, and T. J. Cui, Compact-sized and broadband carpet cloak and free-space cloak, Opt. Express, vol. 17, no. 22, pp. 1994719959, Oct. 2009. [12] B. J. Justice, J. J. Mock, L. H. Guo, A. Degiron, D. Schuring, and D. R. Smith, Spatial mapping of the internal and external electromagnetic elds of negative index metamaterials, Opt. Express, vol. 14, no. 19, pp. 86948705, Sep. 2006. [13] D. R. Smith, S. Schultz, P. Markos, and C. M. Soukoulis, Determination of effective permittivity and permeability of metamaterials from reection and transmission coefcients, Phys. Rev. B, vol. 65, no. 19, p. 195104, Apr. 2002. [14] H. F. Ma, X. Chen, H. S. Xu, X. M. Yang, W. X. Jiang, and T. J. Cui, Experiments on high-performance beam-scanning antennas made of gradient-index metamaterials, Appl. Phys. Lett., vol. 95, no. 9, p. 094107, Sept. 2009. [15] C. A. Balanis, Antenna Theory: Analysis and Design, 3rd ed. New York: Wiley, 2005, p. 50, 687. [16] R. Liu, Q. Cheng, J. Y. Chin, J. J. Mock, T. J. Cui, and D. R. Smith, Broadband gradient index microwave quasioptical elements based on non-resonant metamaterials, Opt. Express, vol. 17, no. 23, pp. 2103021041, Nov. 2009.

Fig. 5. Normalized far-eld radiation patterns for the 2D HMFE lens antenna at 10 GHz (red), 11 GHz (green), and 12 GHz (blue). Solid curves represent simulation results and other curves with the same color correspond to measured data.

Fig. 6. Simulated far eld patterns for the 2D HMFE lens with PMLs (red solid curve) and with a corner reector (black dashed curve). The inset shows HMFE with a 90 corner reector.

slightly smaller than the distance between top and bottom plates for the scanning process, the energy will surely leak into the ambient environment. That explains why the calculated side lobes are higher than the simulation results. In real applications, we can place a perfectly electric conducting (PEC) corner reector behind the line source to restrain the back radiation and suppress the side lobes. By doing this improvement, we obtain the directivity of 33 dB. The lens antenna will thus achieve both high directivity and radiation efciency. The simulated far-eld patterns are given in Fig. 6, which conrms our arguments. Finally, we discuss the improvement of the 2D HMFE lens antenna. In the design, we focus on the realization of the gradient refractive index in the lens structure, but few efforts are made on the impedance matching of the nal device. Due to the mismatch of impedance, there exist small reections at the planar interface. To reduce the reection, extra metamaterial layers can be added outside the interface to obtain a smooth refractive index transition and reduce the impedance contrast between the lens and free space. Alternately, we can choose modied I-shaped unit cells [14], which can provide impedance-matched gradient-index metamaterials and lead to better performances. The extension to the 3D case can be easily realized with layered metamaterials along the optical axes direction. In this regard, the closed-square ring (CSR) unit cells are good candidates due to their isotropic properties inside each layer [16]. In order to achieve a highly directive behavior, the horn antenna or circular waveguide can be used to feed the lens antenna, as was used by Fuchs et al. [7][9]. IV. CONCLUSION In conclusion, we have designed a 2D HMFE lens antenna using gradient index metamaterials. The I-shaped unit cells are utilized in the design. The fabricated prototype has been tested using the 2D eld-map-

402

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Dual-Polarized Planar Feed for Low-Prole Hemispherical Luneburg Lens Antennas

Andrew R. Weily and Nasiha Nikolic

AbstractA dual-polarized planar feed antenna suitable for illuminating a hemispherical Luneburg lens on a ground plane operating at Ku band is presented. Compared to a waveguide feed, the planar antenna has the advantage of reducing the overall prole of the lens system. The planar feed is realized using a 2 2 array of stacked patches, and orthogonal polarizations of the patches use different coupling methods to increase isolation between the two ports of the device. Simulated S-parameters, radiation patterns, gain, directivity, and efciency of the planar feed are compared with measured results. Measured radiation patterns and gain are also presented for a commercial Luneburg lens illuminated by the planar feed, to characterize the combined lens/feed performance and to validate the utility of the planar feed design. Index TermsArray antenna, Luneburg lens, microstrip antenna, planar antenna, Satcom-on-the-move (SOTM).

Fig. 1. Hemispherical Luneburg lens on a ground plane with: (a) waveguide feed, and (b) planar microstrip feed.

I. INTRODUCTION Mobile satellite communication systems operating in Ku band typically require high gain antennas, with a low prole, that are capable of scanning azimuth, elevation and the angle of linear polarization. Known as Satcom-on-the-move (SOTM), most commercial systems are based on reector antenna or planar array designs [1]. However, the hemispherical Luneburg lens on a ground plane has several advantages as a low-prole vehicle-mounted scanning antenna for SOTM applications. Several authors have noted that the hemispherical lens can almost double the effective aperture height compared to a reector solution [2], [3], which gives a more directive beam in the elevation plane. The lens has the added advantage that multiple feeds can be used, giving the potential to access multiple satellites, and compared to electronically steerable array antenna solutions [1] the lens reduces gain loss for low elevation angles. A four-element array of hemispherical Luneburg lenses has been described for Ka band receive-only communications to an aircraft [2]. A dual-beam single hemispherical lens concept, where the lens is fabricated from two dielectric layers, has recently been reported for two-way satellite communications to moving trains [4]. For high elevation angles, the prole of hemispherical Luneburg lens systems is increased by the length of the waveguide feed horn. An alternative to a waveguide feed that reduces the prole for high elevation angles is a planar microstrip antenna [5][7]. Fig. 1 compares the overall system height of two lens antennas with a horn feed and planar feed, for elevation angles of 90 . It is clear from this drawing that a reduction in total height can be obtained through the use of a planar feed. In this communication, a dual-polarized planar array feed is described that is suitable for illuminating a hemispherical Luneburg lens. The SOTM application requires the feed to be a two-port device, with the transmit port matched from 1414.5 GHz and the receive port matched from 12.2512.75 GHz. For the Ku band system of interest, the polarizations of the two ports should be orthogonal and linear. A
Manuscript received May 18, 2010; revised February 24, 2011; accepted May 30, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the CSIRO ICT Centre, Epping, NSW 1710, Australia (e-mail: Andrew.Weily@csiro.au). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167941

further design aim is to achieve high isolation between the transmit and receive bands, particularly over the transmit band, and the cross polarization levels of the radiation patterns should be low at boresight. To achieve these design aims, we introduce a compact stacked patch array conguration combining edge and single aperture-coupled feeding. The use of edge-feeding with a double aperture coupled feed has been reported in [8], but this conguration is not suitable for a closely spaced array at Ku band since there is not enough room for the microstrip feed network between elements. A closely spaced array is required for good aperture efciency in the feed application. Theoretical results for an aperture coupled microstrip feed have been described in the literature for a half Maxwell sheye lens [9], but the feed had only one polarization and no measured results were reported for the lens/planar microstrip feed combination. We present the design and measured performance of a dual-polarized planar array feed with a small hemispherical Luneburg lens suitable for SOTM applications. In Section II the feed parameters and measured performance are discussed. Then in Section III measured performance of the planar feed illuminating a 185 mm radius hemispherical Luneburg lens is presented. Finally in Section IV, overall conclusions are given. II. DUAL-POLARIZED PLANAR MICROSTRIP FEED The design aim for the planar feed is to create a dual port device, with port 1 matched over the transmit frequency band of 1414.5 GHz and port 2 matched over the receive frequency band of 12.2512.75 GHz. The polarization of the radiation generated from the two ports should be orthogonal and linear, with low boresight crosspol. Isolation between ports should also be high to minimize the amount of transmit band power leaking into the receiver. Assuming an f =D = 0:6 for a commercially available Luneburg lens [10], then for efcient illumination of the lens the relative power of the feed radiation patterns should be 010 dB or lower at the edge of the lens, or 656 , in both principal planes. To further improve the lens/feed prole the f/D can be reduced through optimization of the Luneburg lens shell radii and dielectric constants [11], [12]. A multilayer microstrip patch array feed was chosen to meet the multiple design aims of bandwidth, low cross-polarization and high isolation between orthogonal polarizations. A stacked patch element has the benet of a wide impedance bandwidth, and extra degrees of freedom for optimization compared to a single patch on a thick substrate (i.e., width and length parameters of the extra patch, as well as separation between patches). The stacked patch has the further advantage that the microstrip feed network may be placed on a thin substrate (the lower or driven patch) that creates lower spurious radiation and hence lower cross-polar levels compared to a single patch on a thick substrate. Two different techniques of exciting the orthogonal polarizations of the patches were chosen to increase the isolation between transmit and receive bands: edge-feeding and single aperture-coupled feeding. The array was designed with CST Microwave Studio [13] and uses three layers of 0.508 mm thick Rogers RO6002

0018-926X/$26.00 2011 IEEE

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

403

Fig. 2. Exploded perspective view showing the various layers of the planar microstrip antenna.

Fig. 3. Conguration of the 1st layer of RO6002 for the planar microstrip antenna.
("r = 2:9460:04; tan  = 0:0012) substrate, two layers of Rohacell 51 ("r = 1:07; tan  = 0:0041), one layer of CuClad 6700 ("r = 2:32; tan  = 0:0025) and a metal reecting plate as shown in the exploded perspective view of Fig. 2. It consists of a 2 2 2 array of stacked microstrip patches, with Rohacell 51 foam placed between the substrates of the upper and lower patches to increase the impedance bandwidth. Edge feeding has been used to excite the vertically polarized component of the patch array, while the horizontal polarization is excited using a single aperture coupled to a microstrip line. Isolation is also enhanced due to port 1 mismatch over the receive band, and port 2 mismatch over the transmit band. The purpose of the reecting plate is to increase the front-to-back ratio, since there is some back radiation from the slot of the aperturecoupled conguration. If the reector is removed, then the front-toback ratio is reduced by 10.3 dB at 12.5 GHz and 10.4 dB at 14.25 GHz and spillover efciency will be lower. Fig. 3 shows the conguration and parameters of the electromagnetically coupled patches etched on the 1st layer of the RO6002 substrate. It includes the parameters of length and width of the patches, L1 and W1 , as well as the centre-tocentre spacing of the array elements, Ax and Ay . Two pockets are milled into the sides of the 1st substrate layer to accommodate SMA

Fig. 4. Conguration of the 2nd and 3rd layers of RO6002 for the planar microstrip antenna: (a) driven patches on the top of the 2nd layer, (b) apertures between the 2nd and 3rd layer, and (c) feed network on the bottom of the 3rd layer, including the relative position of the apertures shown as dashed lines.

connectors, and four holes with a diameter of 3 mm are drilled in the corners to enable accurate alignment between the multiple layers. Fig. 4(a) depicts the driven patches etched on top of the 2nd layer of RO6002 substrate, showing the length and width parameters, L and W, as well as the edge fed microstrip lines that excite the vertical polarization. To tune the operating frequency of the vertically polarized mode of the stacked patch the parameters W and W1 are adjusted, while

404

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE I PARAMETERS OF THE PLANAR FEED ANTENNA

Fig. 5. Measured and computed S-parameters for the planar feed antenna.

the horizontally polarized mode is tuned by varying L and L1 . The apertures used to couple to the horizontal polarization of the stacked patches are shown in Fig. 4(b), and are characterized by the parameters slot length and slot width, sl and sw . Finally the microstrip feed networks used to combine and correctly phase the array elements are depicted in Fig. 4(c). This gure displays the stub-length parameter, Lstub , which is important for tuning the impedance of the aperture-coupled stacked patch. Two vias, or plated through holes, with diameters of 200 m have been used to couple between the metallization on the top of the 2nd substrate layer to the bottom of the 3rd layer. Two circular apertures with diameters of 1.28 mm are included in the metallization between the 2nd and 3rd substrate layers to ensure high transmission through the vias. Examination of Figs. 3 and 4 shows that a symmetric feeding arrangement has been used to excite the two polarizations of the patch array (i.e., the edge feed location for the upper pair of patches in Fig. 4(a) is on the opposite side to the lower pair). This method was used because it gives better symmetry in the radiation patterns [14]. However, it requires a 180 phase shift to be introduced into the feed network to ensure the patches are fed in phase. The dimensions used for the main parameters of the array feed are listed in Table I. A prototype antenna was fabricated and tested. The measured and CST simulated S-parameters for the prototype array are presented in Fig. 5. There is reasonable agreement between measurement and simulation, and the measured isolation is better than 30 dB over the transmit band. The discrepancies between measurement and simulation are most

Fig. 6. Measured and computed radiation patterns for port 2 of the planar feed , and (b)  . at 12.5 GHz: (a) 

=0

= 90

likely due to tolerances in the fabrication, particularly the via diameter, non-ideal connectors and slight misalignment between the various substrate layers. Simulations in Fig. 5 used a value of "r = 2:90 for the substrate dielectric constant as well as including the CuClad 6700 material, and the coaxial section of the connectors has also been modelled to improve accuracy. Radiation patterns were measured in a spherical neareld chamber, and are compared with simulated results at 12.5 GHz and 14.25 GHz, in Figs. 6 and 7, respectively. Referring to Figs. 3 and 4, the  = 0 plane corresponds to the x-z plane while the  = 90 plane corresponds to the y-z plane of the antenna. There is a discrepancy between the measured and simulated cross-polar patterns at 14.25 GHz, however, the measured cross-polar levels at boresight remain below 025 dB. The nature of the boresight polarization was analyzed through simulation. It was found that the minimum boresight cross-polar level occurs when the reference polarizations are +1 from the principal planes of the antenna. If the antenna is rotated around the z-axis by 1 , then the cross-polar level at boresight is reduced by 1.4

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

405

TABLE II MEASURED AND COMPUTED DIRECTIVITY, GAIN AND EFFICIENCY OF THE DUAL-POLARIZED PLANAR FEED

Fig. 8. Photographs of the prototype planar feed antenna.

Fig. 7. Measured and computed radiation patterns for port 1 of the planar feed , and (b)  . at 14.25 GHz: (a) 

=0

= 90

dB. A photograph of the fabricated prototype and various layers is presented in Fig. 8. The measured and simulated directivity, gain and efciency for the prototype are listed in Table II. III. HEMISPHERICAL LUNEBURG LENS WITH PLANAR FEED To test the performance of the planar feed illuminating a hemispherical Luneburg lens, a commercial lens from Sumitomo has been used [10]. The commercial lens consists of 8 concentric dielectric layers and has an outer radius of 185 mm. The f/D was not specied by the

supplier, but was determined experimentally to be 0.6. This lens was placed on an aluminium ground plane with dimensions of 490 2 700 mm and included a mechanism for varying the feed distance from the lens centre and the feed elevation angle. For the azimuth radiation patterns, the planar feed was located at an elevation angle of 45 . The lens orientation and measured patterns at 12.5 GHz and 14.25 GHz are shown in Fig. 9. This gure shows that the cross-polar radiation level of lens/planar feed combination is dominated by the cross-polar level of the planar feed, which was worse at 14.25 GHz. Elevation radiation patterns were measured for various beam angles between 15 and 75 , where each beam angle corresponds to a different feed location. The lens orientation is shown in Fig. 10(a), while the measured elevation patterns for 12.5 GHz and 14.25 GHz are shown in Figs. 10(b) and (c), respectively. There is a small amount of gain loss at 15 due to the nite size of the ground plane, while at 75 gain loss is due to blockage by the feed and feed support structure. The system has the advantage that the beam shape remains quite constant as the beam is scanned in elevation. Measured gain and aperture efciency for the lens illuminated by the planar feed and a typical horn feed, for elevation angles of 45 , is compared in Table III. The horn antenna used for the measurements in Table III was a conical horn that included a circular-to-rectangular waveguide transition and a single concentric choke ring at the aperture to improve pattern symmetry and cross-polar performance. The drop in efciency of the lens/planar feed system compared to the lens/horn feed in Table III is due to the combination of higher ohmic loss and lower spillover efciency of the planar feed. Spillover efciency is lower than the horn antenna due to the lower front-toback ratio and higher sidelobe levels for non-illuminating angles of the planar feed. A slightly poorer taper efciency and phase efciency also contribute to the lower performance, though to a lesser extent. Future work will examine improving the spillover efciency of the planar feed through the use of a =4 choke in the sidewall of the antenna, similar to that described in [15]. Unfortunately the reduced aperture efciency of the lens/planar feed partly negates the space saving, compared to a lens/horn system with higher aperture efciency. Given the maximum height of the

406

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE III MEASURED GAIN AND APERTURE EFFICIENCY OF THE LUNEBURG LENSPLANAR VS HORN FEED

Fig. 9. Measured co-polar and cross-polar radiation patterns in the azimuth plane for the Luneburg lens fed by dual-polarized planar feed: (a) orientation of lens antenna for pattern measurement, (b) port 2 at 12.5 GHz, and (c) port 1 at 14.25 GHz. Note that the planar feed is at an elevation angle of 45 .

Fig. 10. Measured radiation patterns in the elevation plane for the Luneburg lens fed by dual-polarized planar feed: (a) orientation of lens antenna for pattern measurement, (b) port 2 at 12.5 GHz, and (c) port 1 at 14.25 GHz.

lens/planar feed it is possible to calculate the gain of a horn fed lens that ts into the same prole. Assuming f =D = 0:6 for both cases, a planar feed height = 7:5 mm and lens radius = 185 mm, then the overall prole above the ground plane is 229.5 mm. Given that the dual-polarized horn feed length is 50 mm then the lens diameter that will t the prole of 229.5 mm, with f =D = 0:6, is 149.6 mm. Hence

the gain of the equivalent prole lens/horn, assuming an aperture efciency of 72% at 12.5 GHz, is 30.4 dBi or 0.3 dB less than the lens/planar feed. A dual-polarized 2 2 2 array of stacked microstrip patches has been presented. It has been shown to be suitable for illuminating a hemispherical Luneburg lens operating at Ku band. Measured results conIV. CONCLUSION

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

407

rm that the planar feed has acceptable beam symmetry, low boresight cross-polar levels, efciency of 85%, and port-to-port isolation better than 30 dB across the transmit band. The planar feed is capable of reducing the prole of the overall lens system by up to 40 mm, compared to a dual-polarized waveguide feed design that has equivalent isolation between ports. Measured gain of the feed/lens combination at an elevation angle of 45 is 30.7 dBi at 12.5 GHz, and 31.3 dBi at 14.25 GHz. A further advantage of the planar feed is the potential for easy integration with a low-noise block. The planar feed/lens system has applications in low-prole SOTM systems. ACKNOWLEDGMENT The authors thank F. Ceccato and K. Smart for assisting with the measurement of the lens antenna, and I. Kekic for fabricating the lens ground plane and test structure.

TM Scattering by Perfectly Conducting Polygonal Cross-Section Cylinders: A New Surface Current Density Expansion Retaining up to the Second-Order Edge Behavior
Giulia Coluccini, Mario Lucido, and Gaetano Panariello
AbstractIn the analysis of scattering by perfectly conducting cylinders with polygonal cross-section by means of surface integral operator formulations, fast convergence can be achieved by expanding the surface current density on each side with basis functions factorizing the correct behavior of the elds on the wedges. Usually, the factorized edge behavior is chosen to be coincident with the rst order behavior prescribed by Meixners theory. However, it could not be the correct one and, consequently, the convergence of the method becomes increasingly slow as the theoretical behavior differs from the real one. This phenomenon is particularly notable when one or more of the predicted singularities unexpectedly disappear. To overcome this problem, in this work the analysis of TM scattering is made by introducing a new expansion devised so that only the rst two terms are responsible for the reconstruction of the singularities while the remaining part of the expansion factorizes the second order edge behavior. Actually, the proposed expansion outperforms the one introduced by the authors in a previous work factorizing the rst order edge behavior even when this is the correct one. Index TermsEdge behavior, electromagnetic scattering, expansion basis.

REFERENCES
[1] J. DeBruin, Control systems for mobile Satcom antennas, IEEE Control Syst. Mag., pp. 86101, Feb. 2008. [2] M. Rayner, Datron/Transco Inc., Use of the Luneburg lens for low prole applications, in Microwave Product Digest, Dec. 1999, pp. 113. [3] J. Thornton, Wide-scanning multilayer hemispherical lens antenna for Ka band, IET Proc. Microw. Antennas Propag., vol. 153, no. 6, pp. 573578, Dec. 2006. [4] J. Thornton, A. White, and G. Long, Multi-beam scanning lens antenna for satellite communications to trains, Microw. J., vol. 52, no. 8, pp. 5670, Aug. 2009. [5] Handbook of Microstrip Antennas, J. R. James and P. S. Hall, Eds. London, U.K.: Peter Peregrinus, 1989. [6] D. M. Pozar and D. Schaubert, Microstrip Antennas: The Analysis and Design of Microstrip Antennas and Arrays. New York: IEEE Press, 1995. [7] R. Garg, P. Bhartia, I. Bahl, and A. Ittipiboon, Microstrip Antenna Design Handbook. Norwood, MA: Artech House, 2001. [8] X.-L. Liang, S.-S. Zhong, and W. Wang, Design of a dual-polarized microstrip patch antenna with excellent polarization purity, Microwave Opt. Technol. Lett., vol. 44, no. 4, pp. 329331, 2005. [9] B. Fuchs, S. Palud, L. Le Coq, O. Lafond, M. Himdi, and S. Rondineau, Scattering of spherically and hemispherically stratied lenses fed by any real source, IEEE Trans. Antennas Propag., vol. 56, no. 2, pp. 450460, Jun. 1998. [10] Sumitomo Electric USA, Inc. [Online]. Available: www.sumitomoelectricusa.com [11] H. Mosallaei and Y. Rahmat-Samii, Nonuniform Luneburg and twoshell lens antennas: Radiation characteristics and design optimization, IEEE Trans Antennas Propag., vol. 49, no. 1, pp. 6069, Jan. 2001. [12] N. Nikolic and A. R. Weily, Realistic source modeling and tolerance analysis of a Luneburg lens antenna, in Proc. IEEE AP-S Int. Symp., Jul. 2010, pp. 14. [13] CST. Darmstadt, Germany, CST Microwave Studio 2006 User Manual [Online]. Available: www.cst.com [14] J. Huang, Dual-polarized microstrip array with high isolation and low cross-polarization, Microwave Opt. Technol. Lett., vol. 4, no. 3, pp. 99103, 1991. [15] A. A. Kishk and L. Shafai, Optimization of microstrip feed geometry for prime focus reector antennas, IEEE Trans Antennas Propag., vol. 37, no. 4, pp. 445451, Apr. 1989.

I. INTRODUCTION In the analysis of scattering by objects with wedges the behavior of the elds at edges plays a key role because the fast evaluation of the solution depends on the choice of the functional space to which the solution has to belong. It is well-known that, by means of Meixners theory [1], the edge behavior of the elds on a perfectly conducting wedge can be obtained by expanding them in power series of the distance from the wedge itself and retaining the dominant term when the distance approaches zero. However, the coefcients of the power series are functions of the specic electromagnetic problem at hand and it has been demonstrated that, for particular congurations of the geometry of the scatterers and of the incident eld, the singularities of the transversal component of the elds, corresponding to the rst order term of Meixners expansion, can disappear [2]. It must be concluded that, in general, the correct behavior of the elds on the wedges could not be easily established a priori. In the literature dedicated to scattering by perfectly conducting cylinders with polygonal cross-section by means of surface integral operator formulations, it has been widely observed that a fast convergence can be achieved by expanding the surface current density with basis functions factorizing the correct behavior of the elds on the wedges and discretizing the operator by means of Nystrm [3], collocation [4] or Galerkin like scheme [5][12], just for an example. Obviously, the convergence of the method is increasingly slow as the factorized edge behavior differs from the real one, and this phenomenon becomes particularly notable when one or more of the predicted singularities unexpectedly disappear. In order to overcome this problem, in this work the analysis of TM scattering by a perfectly conducting polygonal cross-section cylinder,
Manuscript received December 27, 2010; revised May 20, 2011; accepted July 20, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The authors are with the D.A.E.I.M.I., University of Cassino, 03043 Cassino, Italy (e-mail: g.coluccini@unicas.it; lucido@unicas.it; panariello@unicas.it). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167924

0018-926X/$26.00 2011 IEEE

408

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

In this regard, the following set of analytically Fourier transformable orthonormal functions factorizing the rst order edge behavior of the unknown current density on the generic side of the cylinder surface has been used as expansion basis in [10][12] with excellent results

'(;) n

x a

0x a

1+

x a

( Pn;) (x=a) ( n;)

x a

(2)

with n = 0; 1; . . ., where 2a is the dimension of the side, 5( 1 ) is the unitary rectangular window
Fig. 1. Geometry of the problem.

 (;) =
n

formulated as electric eld integral equation (EFIE) in the spectral domain, is made by introducing a new expansion basis devised so that only the rst two terms are responsible for the reconstruction of the singularities, while the remaining part of the expansion factorizes the theoretical second order edge behavior of the surface current density on the wedges. In this way, the coefcient associated to the singularity on a wedge vanishes when the singularity itself disappears. It follows that the expansion proposed here is faster convergent than the one presented in [10] that, instead, factorizes the rst order edge behavior on the wedges. Actually, the reported numerical results reveal that the expansion proposed here outperforms the one in [10] even when the correct edge behavior coincides with the predicted one. Moreover, as will be shown, the new expansion guarantees signicantly lower computation times, for any xed number of expansion functions (greater than two) and relative accuracy in the numerical evaluation of the integrals of the scattering matrix. Therefore, it can be concluded that the gain in terms of computation time by using the expansion proposed here with respect to the one in [10] increases more and more as higher is the accuracy required for the solution. II. FORMULATION OF THE PROBLEM In Fig. 1, a perfectly conducting polygonal cross-section cylinder dipped in a homogeneous and isotropic medium is depicted. A coordinate system (x; y; z ) is introduced so that the z axis coincides with the cylinder axis. The L sides of the polygonal cross-section are numbered clockwise and on the ith side, of dimension 2ai , a local coordinate system (xi ; yi ; z ) is introduced so that the yi axis is orthogonal to the ith side and the origin is at the centre of the side itself. A TM polarized plane wave impinges onto the cylinder orthogonally with respect to the cylinder axis inducing a longitudinal current density on the cylinder surface. Therefore, the only non-vanishing component of the scattered electric eld is [10]

0x a

1+

x a

( Pn;)

x a

dx

a2++1 0(n +  + 1)0(n +  + 1) n!(2n +  +  + 1)0(n +  +  + 1)

(3)

and Pn ( 1 ) is the Jacobi polynomial of order n and parameters ;  which, suitably chosen, allow to factorize the behavior of the unknown functions along the edges. In general such behavior cannot be easily predicted and the expansion proposed in [10] could become slowly convergent in the cases in which the correct edge behavior differs from the factorized one. This phenomenon becomes particularly notable when one or more of the predicted singularities unexpectedly disappear. In order to overcome this problem, a new expansion reconstructing the singularities and the second order behavior of the current at edges will be devised. To this end, the expansion of the elds near a perfectly conducting wedge, as introduced by Meixner in [1] and reviewed and further developed in [13][15], will be overviewed in the following subsection.1 A. The Edge behavior of the Electromagnetic Fields on a Perfectly Conducting Convex Wedge In the time-varying case and for orthogonal TM incidence, the longitudinal component of the electric eld and the radial component of the magnetic eld in the proximity of a perfectly conducting convex wedge (see Fig. 2) are expanded as follows [14]:

(;)

Ez (r; ') =
s

rt

+1
i

Hr (r; ') =
s

=0 +1 ( rt 01 ri hrs;i) (') i=0

( ri ezs;i) (')

(4a) (4b)

Ez (x; y; z ) = 0

!
2

+1

=101

e0j jy (x;y)j k 0u 0jux (x;y) ~ p2 2 e du (1) Jiz (u) k 0u

where the exponents ts can be determined by means of the subsequent procedure. Starting from the Helmholtzs equation

r2 E (r; ') + k2 E (r; ') = 0

t z z

(5)

where ! is the angular frequency,  is the magnetic permittivity and k ~ is the wave number of the medium, and Jiz ( 1 ) is the Fourier transform of the current on the ith side of the cylinder with respect to xi . A system of L integral equations is obtained by imposing the total electric eld to be zero on the cylinder surface. III. PROPOSED SOLUTION In order to reduce the obtained system of integral equations to a linear system of algebraic equations, Galerkins method can be applied. A key point is the choice of the basis functions. It is well-known that a fast convergence can be achieved by using expansion basis reconstructing the correct edge behavior of the unknowns [5][12].

the following recurrent system of ordinary differential equations for the coefcients of the expansion in (4a) can be readily obtained

d2 (s;0) 2 (s;0) (') = 0 e (') + ts ez d'2 z d2 (s;1) 2 (s;1) (') = 0 e (') + (ts + 1) ez d'2 z

(6a) (6b)

1It is worth noting that Meixners theory does not have general validity. Indeed, Bach Andersen and Solodukhov [13] demonstrated that Meixners expansion is inconsistent for penetrable wedges for which a more general representation including the powers of the logarithmic of the distance from the edge has to be considered [14], [15]. However, just as noted by Bach Andersen and Solodukhov [13], and Makarov and Osipov [14], Meixners theory remains fully valid for perfectly conducting wedges.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

409

formula (10) justies the following new representation for the current on the ith side of the cylinder in Fig. 1

0 Jiz (xi ) = Ci '01

( ; )

0 xii a
x a
=

+ ( ; ) +Ci '01

xi ai

 +Jiz (xi )

(16)

where the rst two functions

Fig. 2. Perfectly conducting convex wedge.

'(; ) 01

( a 0; )

'(;) 0

x a

(17)

d2 (s;i) 2 (s;i) e (') + (ts + i) ez (') d'2 z 2 (s;i02) (') for i  2: = 0k ez

(6c)

For r ! 0, the boundary conditions for the electric eld, i.e., the vanishing of the tangential component of the electric eld on the sides of the wedge, can be directly imposed on the function solution of (6a)
( ezs;0) (') =

are the only responsible for the reconstruction of the singularities of 0 0 the elds at edges. In fact i ; i and i ; i are respectively related to the rst and second order behavior of the current at edges, while  the residual function Jiz (xi ) is expanded by using the functions (2) factorizing the second order behavior of the current at edges. Moreover, the property (14), that can be easily rewritten as
x

(s;0)

cos(ts ') + 

(s;0)

sin(ts ')

(7)

lim !a (ai

Jiz (xi ) 0 xi )

lim !0a

Ji+1z (xi+1 ) (ai+1 + xi+1 )

(18)

obtaining the following homogeneous system of algebraic equations

(s;0) = 0  (s;0) sin(ts (2 0

sin(ts (2

)) = 0

(8)

6 + 0 leads to the relation Ci = Ci+1 , i.e., the unknown coefcients Ci are associated to functions dened on two adjacent sides. Therefore, by introducing the following function:
i xi01 xi ; ai01 ai
' 

that admits nontrivial solutions if and only if

)) = 0

, ts = 2s 0

(9)
=

(x

=a

for yi01 = 0

with s > 0 in order to guarantee the energy to be nite in any nite region including the edge. Therefore, by reordering the series in (4b) for increasing powers of r , it is obtained

+ 
'

(0x =a ) + 

(19)
for yi = 0

Hr (r; ') = rt

01 h(1;0) (') + rt 01 h(2;0) (') + 1 1 1 r r

(10)

where
a

where 01=2  t1 0 1 < 0 and 0  t2 0 1 < 1, i.e., only the rst term diverges on the wedge. Moreover, being from Maxwells equations
( hrs;i) (') =

(; 01 ) =

0a
2

0x a

1+

x a

'(; ) 01

x a

dx

d ( 0 j! d' ezs;i)(')

1

(11)

( a 0; )

(20)

it can be immediately obtained

(1 hr ;0) (') = t1  (1;0) cos(t1 ')

formula (16) can be rewritten as follows (12)

i Jiz (xi ) = J01 i

xi01 xi ; ai01 ai xi xi+1 ; ai ai+1

(
;

hence, the following relation can be achieved

(1 hr ;0) (0) = (1 hr ;0) (2

i+1 i+1 +J01 

+1

+
n=0

i Jn 'n

( ; )

xi ai

(21)

(13)
)

where
i J01

that leads to the property

Hr (r; 0) lim r!0 r t 01 ts

Hr (r; 2 0 = lim r!0 rt 01

(14)

0 Ci

01

01

( ; ) 2

(22)

For the sake of completeness, it is worth noting that also the solution = 0, corresponding to nontrivial solutions of the transverse electromagnetic problem, should be considered. However, since it can be (0;0) (0;1) demonstrated that hr (') = hr (') = 0, i.e., the rst non-vanishing term is an O(r) for r ! 0, the ordering of the terms in the expression (10) does not change. B. Devising an Expansion Basis Remembering the relation between surface current density and magnetic eld on the cylinder surface

C. Asymptotic behavior of the Fourier Transform of the Expansion Functions As will be shown in the next Section, the expansion in (21) is always faster convergent than the one in [10]. As can be expected, a lower number of expansion functions is needed to achieve the required accuracy for the solution. What so far has not been revealed is that the new expansion (21) guarantees even lower computation times, for any xed number of expansion functions (greater than two) and relative accuracy in the numerical evaluation of the integrals of the scattering matrix. The following considerations are intended to understand this additional positive result.

Jiz (xi ) = 0Hx (xi ; yi

= 0)

(15)

410

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

In [6][12] it has been demonstrated that the coefcients of the scattering matrix, obtained by means of Galerkins method applied to integral equation formulations in the spectral domain, can be always reduced to improper single integrals involving the Fourier transform of the basis and test functions. It can be concluded that the convergence rate of such integrals inevitably depends on the asymptotic behavior of the Fourier transform of the expansion functions. It is well-known that, by means of Watsons lemma [16], it is possible to relate the behavior of the expansion functions at edges with the asymptotic behavior of their Fourier transforms, obtaining

01 where 6 are suitable parameters. Therefore, the integrals involving

expansion functions that diverge at an edge reveal to be slower convergent than the ones in which the expansion functions do not diverge at all. IV. NUMERICAL RESULTS This section is devoted to analyze the convergence of the proposed expansion in comparison with that presented in [10]. To this end, the following normalized truncation error is introduced

1 2

+1 x jxu juj!+1 0 e0jua ejua '(; ) e dx   +1 + + +1 n a u u

(23)

err(N ) =

where k 1 k is the usual euclidean norm and JM is the vector of all the expansion coefcients of the current on all the sides evaluated with M terms on each side. All the simulations are performed on a laptop equipped with an Intel Core 2 Duo CPU T9600 2.8 GHz, 3 GB RAM, running Windows XP and the integrals of the scattering matrix evaluated by means of an adaptive Gaussian quadrature routine with a relative accuracy of 10 10 unless otherwise specied. In Fig. 3(a) and (b) the normalized truncation error and the surface current density are plotted for the perfectly conducting triangular crosssection cylinder dipped in vacuum sketched in the inset when two TM polarized plane waves impinge onto the cylinder such that the singularity in the current on the wedge C is cancelled out. This case is considered non predictable, i.e., it is assumed that the current diverges on all the wedges. It is immediate to note that the expansion in [10], that has to reconstruct the current vanishing on the wedge C with functions that diverge on the same wedge, converges slower than the new expansion introduced here. This phenomenon can be further understood by observing Fig. 3(c), where the coefcients of the surface current density on the side CA are reported: it can be clearly noticed that by using the expansion in (21) in which each singularity is associated only to one expansion function, the disappearance of the singularity at edge C simply leads to the vanishing of the coefcient associated to the singularity itself. Actually, the expansion in (21) reveals to be faster convergent than the one introduced in [10] even when the correct edge behavior coincides with the predicted one (the rst order behavior prescribed by Meixners theory), as shown in Fig. 4(a) and (b) in which the normalized truncation error and the surface current density are plotted for the same cylinder of the previous example when a single TM polarized plane wave impinges onto the cylinder. On the other hand, this conduct was expected because the expansion in (21) reconstructs the rst and second order behavior of the current at edges while the one presented in [10] factorizes only the dominant behavior of the current at edges. It is well-known that more accurate solutions can be obtained by properly increasing the number of expansion functions and/or the accuracy in the numerical evaluation of the integrals of the scattering matrix. It leads to an inevitable increase in computation time. For this reason, it is important to point out that the new expansion introduced here is

kJN +1 0 JN k kJN k

(24)

Fig. 3. For the cylinder in vacuum and the incident eld sketched in the inset: (a) normalized truncation error, (b) absolute value of the current density, (c) absolute value of the expansion coefcients of the current density on the side CA. jE j jE j V/m, AB = ; BC CA; B CA = .

=1

= 4

^ = 6

computationally more efcient than the one in [10] as a higher number of expansion functions is considered and greater the relative accuracy in the numerical evaluation of the integrals of the scattering matrix is. This behavior is claried in Fig. 5, in which, for the cylinder of the previous examples, the ratio between the computation time to ll the scattering matrix, as obtained by using the expansion from [10], with respect to the new expansion (21) discussed in this communication, is shown as a function of the number (N) of expansion functions and for different values of the relative accuracy (R) used in the numerical evaluation of the integrals of the scattering matrix. The comparison between the computation times shown in Fig. 5 is carried out by setting the same number of basis functions for both expansions. However, Figs. 3(a) and 4(a) show that the gain in terms of number of basis functions by using the new expansion proposed here with respect to the expansion in [10] always increases for smaller values of the normalized truncation error. It follows that the gain in computation time by using the expansion (21) with respect to the one in [10] drastically increases as higher is the accuracy required for the solution.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

411

Fig. 6. For the problem analyzed in Fig. 3, computation time to ll the scattering matrix with the expansion in [10] and the one proposed here as a function of the normalized truncation error.

Fig. 4. For the cylinder in vacuum and the incident eld sketched in the inset: (a) normalized truncation error, (b) absolute value of the surface current density. jE j V/m, AB = ; BC CA; B CA = .

=1

= 4

^ = 6

Fig. 7. For the problem analyzed in Fig. 4, computation time to ll the scattering matrix with the expansion in [10] and the one proposed here as a function of the normalized truncation error.

of the predicted singularities unexpectedly disappear and always outperforms the solution proposed by the authors in a previous work [10].

REFERENCES
[1] J. Meixner, The behavior of electromagnetic elds at edges, IEEE Trans. Antennas Propag., vol. 20, pp. 442446, Jul. 1972. [2] J. Van Bladel, Singular Electromagnetic Fields and Sources. New York: IEEE Press, 1991. [3] J. L. Tsalamengas, Exponentially converging Nystrms methods for systems of singular integral equations with applications to open/closed strip- or slot-loaded 2-D structures, IEEE Trans. Antennas Propag., vol. 54, pp. 15491558, May 2006. [4] D. Assante, S. Falco, M. Lucido, G. Panariello, F. Schettino, and L. Verolino, Shielding effect of a strip of nite thickness, Elect. Engrng., vol. 89, pp. 7987, Dec. 2006. [5] A. I. Nosich, The method of analytical regularization in wave-scattering and eigenvalue problems: Foundations and review of solutions, IEEE Antennas Propag. Mag., vol. 41, pp. 3449, 1999. [6] R. Araneo, S. Celozzi, G. Panariello, F. Schettino, and L. Verolino, Analysis of microstrip antennas by means of regularization via Neumann series, in Rev. Radio Sci. 19992002, W. R. Stone, Ed. PiscatawayHoboken, NJ: IEEE PressWiley Interscience, 2002, pp. 111124. [7] K. Eswaran, On the solutions of a class of dual integral equations occurring in diffraction problems, in Proc. Roy. Soc. London, 1990, pp. 399427, ser. A. [8] E. I. Veliev and V. V. Veremey, Numerical-analytical approach for the solution to the wave scattering by polygonal cylinders and at strip structures, in Analytical and Numerical Methods in Electromagnetic Wave Theory, M. Hashimoto, M. Idemen, and O. A. Tretyakov, Eds. Tokyo: Science House, 1993.

Fig. 5. Ratio between the computation time to ll the scattering matrix for the cylinder in vacuum sketched in the insets of the previous gures, as obtained by using the expansion from [10], with respect to the one proposed here, as a function of the number (N) of expansion functions and the relative accuracy (R) in the evaluation of the integrals of the scattering matrix.

For the problems analyzed in Figs. 3 and 4, this result is clearly shown in Figs. 6 and 7 respectively, in which the computation time to ll the scattering matrix with the expansion in [10] and the one proposed here is reported as a function of the normalized truncation error. V. CONCLUSION In this communication the analysis of TM scattering by a perfectly conducting polygonal cross-section cylinder has been done by introducing a new expansion basis for the surface current density in which only the rst two terms are responsible for the reconstruction of the singularities while the remaining part of the expansion factorizes the theoretical second order edge behavior. The reported numerical results reveal that the proposed expansion is fast convergent even when one or more

412

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

[9] M. Lucido, G. Panariello, and F. Schettino, Analytically regularized evaluation of the scattering by perfectly conducting cylinders, Microw. Opt. Technol. Lett., vol. 41, pp. 410414, 2004. [10] M. Lucido, G. Panariello, and F. Schettino, Analysis of the electromagnetic scattering by perfectly conducting convex polygonal cylinders, IEEE Trans. Antennas Propag., vol. 54, pp. 12231231, Apr. 2006. [11] M. Lucido, G. Panariello, and F. Schettino, Electromagnetic scattering by multiple perfectly conducting arbitrary polygonal cylinders, IEEE Trans. Antennas Propag., vol. 56, pp. 425436, Feb. 2008. [12] M. Lucido, G. Panariello, and F. Schettino, TE scattering by arbitrarily connected conducting strips, IEEE Trans. Antennas Propag., vol. 57, pp. 22122216, Jul. 2009. [13] J. B. Andersen and V. V. Solodukhov, Field behavior near a dielectric wedge, IEEE Trans. Antennas Propag., vol. 26, pp. 598602, Jul. 1978. [14] G. I. Makarov and A. V. Osipov, Structure of Meixners series, Radiophys. Quant. Electron., vol. 29, pp. 544549, Jun. 1986. [15] E. Marx, Logarithmic terms in elds near the edge of a dielectric wedge, in IEEE Antennas. Propag. Soc. Int. Symp. Dig., 1990, pp. 10831086. [16] D. S. Jones, The Theory of Electromagnetism. New York: Pergamon Press, 1964.

A Modication of the Kummers Method for Efcient Computation of the 2-D and 3-D Greens Functions for 1-D Periodic Structures
Sergei P. Skobelev
AbstractA new modication of the Kummers method of th order for 2 6 is proposed for efcient summation of the spectral and spatial series representing the 2-D and 3-D Greens functions, respectively, for 1-D periodic structures in homogeneous media. The modication is based on transformation of the auxiliary series consisting of asymptotic terms of the original series and subsequently subtracted from the latter into a new series which, unlike the previous one, allows its summation in closed form. As a result, there are obtained new representations of the Greens functions in question consisting of rapidly converging difference as , as well as new series whose terms decay with rate rigorous analytic expressions for the sums of the transformed auxiliary series. Some numerical examples and comparisons characterizing the effectiveness of the proposed method are also presented and discussed. Index TermsAcceleration techniques, Greens functions, Kummers method, numerical methods, periodic structures.

of Hankel functions of purely imaginary arguments. The latter modication is also less efcient than a higher order modication [12], [13] involving Hankel functions of complex arguments with real and imaginary parts satisfying a special condition proposed in [26]. However, no comparisons have been carried out between [16] from one side and [12], [13] from the other one. No comparisons have also been made between high-order modication [22] involving Macdonald functions and modication [16] based on elementary functions. Therefore, the question of what modication of the high-order Kummers method is most efcient continues to be of interest. In the present communication, a new high-order modication of the Kummers method, following a procedure similar to that presented in [4], [6], [7], [10], [11], and [16], is proposed. It is based on a new more general and compact transformation of the auxiliary series constructed of asymptotic terms of the original series. As a result, the new auxiliary series keeping the same asymptotic behavior of its terms allows its summation in closed form, and appropriate new expressions for the sums are derived. The indicated approach is also applied to modifying the Kummers method developed in [25] for calculation of the 3-D Greens function for 1-D periodic structures that allows constructing a more compact algorithm than that in [25]. The effectiveness of the proposed approach is illustrated by a few numerical examples and some comparisons. II. THE 2-D GREENS FUNCTION The 2-D Greens function for 1-D periodic structures in homogeneous media is a potential of the eld created by an innite periodic array of line sources of identical amplitude and linearly progressing phase. We introduce a rectangular Cartesian coordinate system 0xyz in free space and assume that the line sources are arranged at some identical distance from the plane z = 0, parallel to the y axis, and with spacing b along the x axis. Omitting the harmonic time dependence e0i!t , we can write the Greens function in question in the form of a well known spectral series

G(;  ) =

1 ei 00 2b q=01 0q

(1)

I. INTRODUCTION Analysis of electromagnetic wave scattering by periodic structures with using the method of integral equations for a number of practical applications [1][4] requires efcient computation of appropriate Greens functions. At present, there exist several approaches to solution of that problem [4][25]. Concerning the computations of the 2-D Greens function for 1-D periodic structures, the studies and comparisons carried out in [16] and [22] show that the most efcient approach is based on the Kummers method of high orders. Various modications of such an approach are also considered in [4], [6], [7], [10][13], [17], and [19]. The modication [16] based only on elementary functions is shown to be more efcient than the modication of the 1st order using an auxiliary series
Manuscript received January 18, 2011; revised June 04, 2011; accepted June 24, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. The author is with Company Radiophyzika, Moscow 125363, Russia (e-mail: skobelev@rol.ru). Digital Object Identier 10.1109/TAP.2011.2167928

where  = x0x0 ;  = jz 0z 0 j; (x; z ) are coordinates of the observation point, (x0 ; z 0 ) are coordinates of a reference source (usually closest to 2 the coordinate origin), q = k sin  + 2q=b; 0q = q 0 k2 = 0i k2 0 2 ;  is the angle of phasing measured from the z axis, k = q 2= is the wavenumber, and  is the wavelength. The representation (1) is attractive from the viewpoint of involving exclusively elementary functions that do not require great computer resources for their evaluation. However, if the observation point is located in the array plane or very close to the latter, the series (1) converges so slowly that becomes unusable for practical calculations. One of the ways of accelerating its convergence in the indicated cases is based on application of the well known Kummers method involving an auxiliary series consisting of asymptotic expressions for the terms of the original series. Expanding the function exp(006q  )=06q , standing in (1), in the Taylor series in terms of powers of 1=q and retaining its M rst terms, we can write the expression

ke00

06q

be7 

 M

q m=0

01 A6 m +O m+1

qM +1

(2)

0018-926X/$26.00 2011 IEEE

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

413

valid for sufciently large q , where the expansion coefcients A6 for m M = 6 are placed in Appendix A. If, following the Kummers method, we subtract the auxiliary series, consisting of the terms involving the M th order asymptotic expression (2), from the original series (1), the difference series will converge with rate 1=jq jM +1 . So, the convergence rate at appropriate choice of M , can be made high enough. However this situation still leaves the problem of evaluating the sum of the auxiliary series itself because no closed-form analytical expressions in elementary functions for the sums at m > 0 have been obtained. In the 3rd-order modication [4], [6], [7], the right-hand side of (2) is replaced by
be7 

Having determined the coefcients Pml and using (3), we can rewrite (2) in the form
ke00

be 06q =

7  M 01 B 6


m=0

1 Mm + m + O qM +1

(8)

6 with new expansion coefcients BMm . The appropriate formulas for them are given in Appendix B. The new asymptotic expression (8) allows us now to represent the Greens function (1) as a sum
G ; 

6  1 a6 1 2 + q(qa+ 1) + q(q + 1)(q + 2) + O q14 q

where

( ) = e 200 b + G+ + G0 + S + + S 0
ke00

i 00 

(9)

where coefcients a6 and a6 are explicitly given in [4], [6], [7], and 1 2 the auxiliary series composed of the indicated terms is summed up analytically. The auxiliary series in [16] is transformed into a closed-form analytical expression plus one more rapidly converging series. We propose to solve the problem in somewhat different way described below. The key feature of our approach is representation of 1=q m for 2  m  M in the form

6=

1 1 2kb q=1

06q


0

be7 

6  M 01 BMm

m=0

+m

i 
(10)

1 = M 01 Pml + O 1 qm q M +1 l=0 q + l

are difference series whose terms decay as 1=q M +1 , and

(3)
S

M 01 6 = e(i7)ku 6 B

where Pml are coefcients that we need to determine. To demonstrate one of the possible ways of how to do that, let us consider, for brevity, the simplest case of M = 2. The expression (3) in that case can be rewritten as

4

m=0

6 Mm m

(11)

are results of analytical summation of the auxiliary series in closed form. The sum

q2

1 = P20 + P21 1 + 1 q q q

01

+ O q13

0

+=

1 pq eiq

(4)

Using the Taylor series

= ln 1 2p cos + p2 + iarctg 1 p sin p cos

0 0 0

q=1

(12)

(1 + x)01 = 1
we reduce (4) to

+ x2

3 + 111

(5)

for m = 0, where p = exp(02=b) and = 2=b, is obtained with using tabulated sums 1.448.1 and 1.448.2 given in [27]. The sums + m for m > 0 are determined via (12) as
m

1 = P20 + P21 q2 q

P21 q2

+O

q3

1 pq eiq

(6)

q=1

= e pm q+m

0im

0

+0

m pq eiq q=1
q

(13)

P21 satisfy two linear algebraic equations P20

We can see from (6) that (4) will be valid if the coefcients P20 and

Some expressions for m = 1 similar to (13) are available in [10] and [20]. Note, that (13) can also be written in the form of a more compact recurrent relation

+ P21 = 0;

P21

= 1
0

(7)

m

+ = e p m01

0i

1:

(14)

where P21 is determined directly. Substituting the value of P21 in the rst equation, we obtain P20 = 1. This version of the 2nd order Kummers method was originally applied in [10] and [11]. The coefcients Pml in (3) for other values of M are determined in a similar way. With using (5), the problem for each m standing in (3) in the range 2  m  M is reduced to a system of linear algebraic equations of M th order. In fact, PM 0 is present only in the rst equation. Therefore, we can rst solve other M 0 1 equations that do not contain PM 0 , and after that we can determine PM 0 from the rst equation.

0 + Finally, m = m3 for 0  m  M 0 1, where the asterisk denotes complex conjugation. Thus, we have obtained a new representation of the 2-D Greens function for 1-D periodic structures (9) consisting of rapidly converging series of elementary functions (10) and closed-form rigorous analytical expressions (11). Note that (9) contains only one rapidly converging series for each case of positive and negative indices of summation while the approach [16] gives two series with the same 1=qM +1 convergence rate.

414

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

III. THE 3-D GREENS FUNCTION The 3-D Greens function for 1-D periodic structures is a potential of the eld created by a linear periodic array of point sources of identical amplitude and linearly changing phase. Let the sources be arranged with spacing b along a straight line parallel to the x axis. Then the Greens function can be represented by a well known innite spatial series

where = kb 6 U . The expressions (20) and (21) are directly obtained from (12), (13), and (14). The coefcients BMm are determined by formulas given in Appendix B where coefcients Am corresponding to the 3-D case should be calculated now according to expressions given in Appendix C.

IV. COMPARISONS AND DISCUSSION To evaluate the effectiveness of the method proposed above, let us compare (9) to another modication of the Kummers method where the 2-D Greens function (1) is represented in the form

G(; ;  ) =

where Rn =

corresponding to n = 0; (x; y; z ) are coordinates of the observation point, and U is the phase difference for adjacent sources. Following [25], we expand exp(ikR n )=R n in a Taylor series over powers of 1=n with retaining M rst terms

0 2 2 2 2 0; + ; =  +  ;  = x 0 ;  = 0 ;  = z 0 z(0; 0xnby)0 ; z0 )arecoordinates of the referencexsource y0y ( 6 6

1 eikR 4 n=01 Rn
1

inU

(15)

eikR R n

M eik(nb ) b

01 A6 m

m=0

nm+1

+O

nM +1

(16)

where appropriate expressions for the expansion coefcients are available in [25]. However, for completeness, coefcients Am are also presented in Appendix C in an alternative form. The expression (16) is used in [25] in the auxiliary series subtracted from (15) with forming a rapidly converging difference series. Evaluation of the auxiliary series itself is performed in [25] with use of a few sets of approximate expressions providing, nevertheless, high accuracy of calculations in appropriate regions of the argument values. The authors of [25] have demonstrated high effectiveness of their approach. However, since each region of the argument values requires its own set of expressions for calculation, such an algorithm looks rather cumbersome. To make it more compact, we propose here to transform (16) into a form similar to (8). After doing that, we represent the Greens function (15) as
eikR + +G +G 4R0

1 e00  00  0 Re e 0aq ei  2b q=01 0q 1 inkb sin  1 (1) Im H0 (ka rn ) (22) 04 e n=01 2 2 0 ka ; ka = k1 + ik2 is an auxiliary complex where 0aq = q (1) wavenumber with k2 > 0; H0 ( 1 ) is the Hankel function of the rst kind and zero order, and rn = ( 0 nb)2 +  2 .
G(;  ) =
1

G(; ;  ) =

0 + S+ + S0

(17)

where

6=

1 4

1
n=1

eikR R n

e 6 0

are difference series whose terms decay with the rate 1=nM +1 , i.e., the same as that in [25], and

ik(nb7) M 01 B 6 Mm e6inU b n+m m=0

(18)

6 = e7ik M 01 BMm m 6 6
4b

m=0

(19)

are rigorous sums of the auxiliary series with expressions

0 = m =

1
n=1

ein n

0 ln

2 sin

6
2

+i

0 6
2

(20)

n=1 0i =e

ein im =e n+m

0
1

0

60

m n=1

6 m01 0

ein n

m>0

(21)

The case when the auxiliary wavenumber is purely imaginary has been considered in [16], [19], and [22]. The Hankel function in this (1) case is expressed as H0 (ik2 rn ) = 0(2i= )K0 (k2 rn ) where K0 ( 1 ) is the real-valued Macdonald function. Due to the behavior of the latter [27], the second (spatial) series in (22) converges exponentially, while the terms of the rst (spectral difference) series in (22) in the worst case  = 0 decay with rate 0q 3 as jq j ! 1. A more effective approach is based on use of a complex wavenumber 2 2 whose components satisfy the condition k1 0 k2 = k2 proposed originally in [26] for efcient computation of the Greens function for rectangular waveguide. The representation (22) with such a special complex wavenumber has been used in [12] and [13]. Keeping the exponential convergence of the second series, this approach provides the convergence of the rst series in (22) with rate 0q 5 . Further modications providing even faster convergence have been proposed and studied in [24]. In all cases there exists an optimum value of k2 minimizing the computation time of (22) depending on the parameters of the problem. Some values of the CPU time t required for computation of the 2-D Greens function represented by (9) and (22) at specied values of relative error " are presented in Table I. The parameters of the problem are taken from [19] for the case (a) and from [22] for the cases (b) and (c). The FORTRAN-90 codes developed for the computations use real and complex variables of double precision. The relative errors of summation of the series in (9) and (22) are determined with respect to each sum as a whole in case (a) like, for instance, in [18], and with respect to the zero-order term in each series in cases (b) and (c) like in [22]. The results for the CPU time are normalized to the minimum value tref determined over all the results. The Table also contains total numbers of the terms summed up for achieving the specied accuracy. The rst value in each cell for (22) corresponds to the spectral difference series while the second one corresponds to the spatial series containing the Hankel functions. According to Table I, the modication (9) of the 6th order is fastest. However, its advantage over the modication (22) using 2 k2 + k2 + ik2 decreases with increasing the spacing b, and, ka = as the check has shown, the CPU time for both approaches becomes approximately equal already at b of about 6-to-7 wavelengths. Since the modication (22) with ka = ik2 has already been considered in the literature, we can make indirect comparisons of our modication (9) with some other approaches.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

415

TABLE I RELATIVE CPU TIME FOR THE COMPUTATION OF FOR: (a) (b)

G VIA (9) AND (22) b = 0:8; kb sin  = 1:2;  = 0:4; = 0;" = 10 ; b = 0:6; = =4; = 0:5b; = 0;" = 10 ; (c) b = 2; = =4; = 0:5b; = 0; " = 10

Finally, since the convergence rate of the series (18) and the level of complexity of the expressions used in the representation (17) of the 3-D Greens function are similar to those in the Kummers method modication proposed in [25], we may expect that both these approaches are comparable over the effectiveness. APPENDIX A The coefcients of expansion (2) are determined by expressions

A6 = 1; A6 = (C1 7 u)b 0 1  6 = (C2 7 2C1 u + u2 )b2 A2 2 (C3 7 3C2 u + 3C1 u2 7 u3 )b3 A6 = 3 3 2 3 4 4 A6 = (C4 7 4C3u + 6C2u4 7 4C1u + u )b 4 b5 A6 = (C5 7 5C4u + 10C3u2 7 10C2u3 + 5C1u4 7 u5 ) 5 5 where u = sin  and 2 1 C1 = k = ; C2 = 2 + 2 ; 23 4 2 3 C3 = 6 + 34 ; C4 =  + 2 + 8 ; 24 5 3 C5 = 120 + 524 + 58
are the coefcients of expansion of function (2) divided by parameter 6 6q in the Taylor series over powers of 6 6q (see Appendix in [22] for comparison).

k=

k=

APPENDIX B The coefcients of expansion (8) are determined as

for According to [19], the Ewalds method in the problem with parameters indicated in Table I(a) is about 2 times faster than (22). The modi= 6 is more than 8 times faster than (22). Therefore cation (9) with = 6 is about 4 times faster than the Ewalds method for (9) with case (a). In the cases (b) and (c) considered in [22], the Ewalds method is faster than (22) with a = 2 about 4.9 and 7.8 times, respectively. = 6 is faster than (22) about 9 and 14.2 times, reSince (9) at spectively, the modication (9) will be about 1.8 times faster than the Ewalds method for both cases1. Relying on these results, we may expect that the modication (9) of the 6th order will also be faster than the Veysoglus method [9], since according to [14] and [19], the effectiveness of the latter is comparable with that of the Ewalds method. The Kummers method modication of the 7th order in [22] based on the Macdonald functions in cases (b) and (c) is over 5 and 9 times faster than a similar modication of the 1st order corresponding to (22) with a = 2 . Comparing the indicated values to those given above = 6 and (22) we see that the modication (9) of the 6th for (9) with order is over 1.5 times faster than the 7th-order modication [22] for both cases.

M = 2; M = 3;

B20 = A0 + A1; B21 = 0A1

3 1 B30 = A0 + 2 A1 + 2 A2 1 B31 = 02A1 0 A2 ; B32 = 1 A1 + 2 A2 2 1 B40 = A0 + 11 A1 + A2 + 6 A3 6 5 1 B41 = 03A1 0 2 A2 0 2 A3 3 1 B42 = 2 A1 + 2A2 + 2 A3 1 1 B43 = 0 1 A1 0 2 A2 0 6 A3 3

for

k ik

for

ik M

1According to one of the anonymous reviewers with reference to [21], the CPU time for the Ewalds method in case (c) in [22] corresponds to not the best splitting parameter value. Therefore, this case may require an additional discussion.

M = 4; 25 35 5 1 B50 = A0 + 12 A1 + 24 A2 + 12 A3 + 24 A4 3 1 B51 = 04A1 0 13 A2 0 2 A3 0 6 A4 3 1 B52 = 3A1 + 19 A2 + 2A3 + 4 A4 4 7 7 1 B53 = 0 4 A1 0 3 A2 0 6 A3 0 6 A4 3 11 1 1 B54 = 1 A1 + 24 A2 + 4 A3 + 24 A4 4

416

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

given above is omitted for simplicity.

M = 5, and 17 1 1 B60 = A0 + 137 A1 + 15 A2 + 24 A3 + 8 A4 + 120 A5 60 8 77 71 7 1 B61 = 05A1 0 12 A2 0 24 A3 0 12 A4 0 24 A5 59 13 1 B62 = 5A1 + 107 A2 + 12 A3 + 12 A4 + 12 A5 12 49 1 B63 = 0 10 A1 0 13 A2 0 12 A3 0 A4 0 12 A5 3 2 61 41 11 1 B64 = 5 A1 + 24 A2 + 24 A3 + 24 A4 + 24 A5 4 5 7 1 1 B65 = 0 1 A1 0 12 A2 0 24 A3 0 12 A4 0 120 A5 5 6 for M = 6. The superscript 6 at BMm and A6 in the formulas n
for APPENDIX C The coefcients of expansions (16) are determined by formulas

A6 = 1; A6 = C1 6  0 1 A6 = C2 6 2C1 + 2 2 A6 = C3 6 3C2 + 3C12 6 3 3 A6 = C4 6 4C3 + 6C22 6 4C13 + 4 4 A6 = C5 6 5C4 + 10C32 6 10C23 + 5C14 6 5 5 where  = =b; 2 1 2 C1 =  ; C2 = 2 0 2 2 ; b b  3 0 3 3 ; C4 =  4 0  2 + 3 4 C3 = 6 4 b3 24 2 8 b4 5 5 3   C5 = 120 0 524 + 58 5 b and  = ik=2.
ACKNOWLEDGMENT The author is grateful to the anonymous reviewers for their useful comments concerning reorganization of the communication, existence of papers [16], [20], and [21], as well as some misspellings.

REFERENCES
[1] N. Amitay, V. Galindo, and C.-P. Wu, Theory and Analysis of Phased Array Antennas. New York: Wiley, 1972. [2] Frequency Selective Surface and Grid Array, T. K. Wu, Ed. New York: Wiley, 1995. [3] C.-F. Yang, W. D. Burnside, and R. C. Rudduck, A doubly periodic moment method solution for the analysis and design of an absorber covered wall, IEEE Trans. Antennas Propag., vol. 41, no. 5, pp. 600609, May 1993. [4] V. P. Shestopalov and Y. K. Sirenko, Dynamic Theory of Gratings. Kiev: Naukova Dumka, 1989, (in Russian). [5] J. R. Zinecker, Efcient computation of Greens functions for periodic structures, Masters Thesis, Dept. Elect. Eng., Univ. Houston, Houston, TX, 1985. [6] Yu. A. Tuchkin and V. P. Shestopalov, Regularization of a boundary value problem of diffraction by a wavy surface with Dirichlet boundary condition, Soviet Phys. Dokl., vol. 35, no. 4, pp. 347349, 1990. [7] Yu. I. Krutin, Yu. A. Tuchkin, and V. P. Shestopalov, Regularization of a boundary value problem on diffraction by a semitransparent lattice made of bars of arbitrary cross section with the Dirichlet boundary condition, Comp. Math. Math. Phys., vol. 31, no. 6, pp. 6069, 1991. [8] S. Singh, W. F. Richards, J. R. Zinecker, and D. R. Wilton, Accelerating the convergence of series representing the free space periodic Greens function, IEEE Trans. Antennas Propag., vol. 38, no. 12, pp. 19581962, Dec. 1990.

[9] M. E. Veysoglu, H. Yueh, R. Shin, and J. Kong, Polarimetric passive remote sensing of periodic structures, J. Electromagn. Waves Appl., vol. 5, no. 3, pp. 267280, Mar. 1991. [10] L. L. Mukhamedov, Algorithms of the method of auxiliary sources for calculation of characteristics of periodic waveguide-rod structures, Masters thesis, Moscow Inst. Phys. Technol., Russia, 1990, (in Russian). [11] S. P. Skobelev and L. L. Mukhamedov, Algorithms of the method of auxiliary sources for analysis of waveguide-dielectric periodic structures, Izvestiya VuzovRadiozika, vol. 34, no. 4, pp. 392402, 1991, (in Russian). [12] S. P. Skobelev and L. L. Mukhamedov, Calculation of characteristics of a waveguide-dielectric array excited by TM waves, Radiotekhnika i Elektronika, vol. 37, no. 7, pp. 12121219, 1992, (in Russian). [13] S. P. Skobelev and L. L. Mukhamedov, Analysis of waveguide antenna arrays with protruding dielectric elements, IEEE Trans. Antennas Propag., vol. 41, no. 5, pp. 574581, May 1993. [14] A. W. Mathis and A. W. Peterson, A comparison of acceleration procedures for the two-dimensional periodic Greens function, IEEE Trans. Antennas Propag., vol. 44, no. 4, pp. 567571, Apr. 1996. [15] S. Yu. Sadov, Computation of quasiperiodic fundamental solution of Helmholtz equation, in Proc. 2nd Int. Conf. on Advances in Difference Equations, Veszprem, 1995, pp. 551558, CRS Press, 1997. [16] B. Baekelandt, D. De Zutter, and F. Olyslager, Arbitrary order asymptotic approximation of a Greens function series, AEU Int. J. Electron. Commun., vol. 51, no. 4, pp. 224230, 1997. [17] G. S. Wallinga, E. J. Rothwell, K. M. Chen, and D. P. Nyquist, Efcient computation of the two-dimensional periodic Greens function, IEEE Trans. Antennas Propag., vol. 47, no. 5, pp. 574581, May 1999. [18] F. Capolino, D. R. Wilton, and W. A. Johnson, Efcient computation of the 2-D Greens function for 1-D periodic structures using the Ewald method, IEEE Trans. Antennas Propag., vol. 53, no. 9, pp. 29772984, Sep. 2005. [19] G. Valerio, P. Baccarelli, P. Burghignoli, and A. Galli, Comparative analysis of acceleration techniques for 2-D and 3-D Greens functions in periodic structures along one and two directions, IEEE Trans. Antennas Propag., vol. 55, no. 6, pt. 1, pp. 16301643, Jun. 2007. [20] O. Malyuskin, V. Fusco, and A. Schuchinsky, Convergence acceleration of the doubly periodic Greens function for the analysis of thin wire arrays, IET Microw. Antennas Propag., vol. 2, no. 5, pp. 410417, 2008. [21] F. T. Celepcikay, F. Capolino, D. R. Jackson, and D. R. Wilton, Choosing splitting parameters and summation limits in the numerical evaluation of 1-D and 2-D periodic Greens functions using the Ewald method, Radio Sci., vol. 43, 2008, RS6S01, doi:10.1029/2007RS003820. [22] A. L. Fructos, R. R. Boix, F. Mesa, and F. Medina, An efcient approach for the computation of 2-D Greens functions with 1-D and 2-D periodicities in homogeneous media, IEEE Trans. Antennas Propag., vol. 56, no. 12, pp. 37333742, Dec. 2008. [23] D. Van Orden and V. Lomakin, Rapidly convergent representations for 2D and 3D Greens functions for a linear periodic array of dipole sources, IEEE Trans. Antennas Propag., vol. 57, no. 7, pp. 19731984, Jul. 2009. [24] M. M. Ivanishin and S. P. Skobelev, A modication of the Kummers method for efcient computation of the Greens function for doubly periodic structures, IEEE Trans. Antennas Propag., vol. 57, no. 9, pp. 27942798, Sep. 2009. [25] A. L. Fructos, R. R. Boix, and F. Mesa, Application of Kummers transformation to the efcient computation of the 3-D Greens functions with 1-D periodicity, IEEE Trans. Antennas Propag., vol. 58, no. 1, pp. 95106, Jan. 2010. [26] M. M. Ivanishin, Use of integral equations in the problem on a cylinder inside a rectangular waveguide, Radio Eng. Electr. Phys., vol. 29, no. 12, pp. 18, 1984. [27] I. S. Gradstein and I. M. Ryzhik, Tables of Integrals, Series, and Products. New York: Academic, 1980.

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

417

A Spatial Beam Splitter Consisting of a Near-Zero Refractive Index Medium

Ruey-Bing Hwang, Neng-Chieh Hsu, and Cheng-Yuan Chin

AbstractA novel metamaterial-based beam splitter is demonstrated in this communication. The metamaterial is a three-dimensional (3 D) shnet structure arranged in a two-dimensional (2 D) square lattice. According to Snells law, the refracted wave from medium into air tends to be normal to the interface if the effective refractive index of medium is approximately zero. In light of this concept, the study presents a metamaterial with nearzero effective refractive index and an internal line source inside as excitation. Metallic ares with specic tilting angle were installed at four sides of the beam splitter to reduce reection coefcient and match the impedance of the shnet structure and free space. In addition to calculating the phase relation of eigenwaves in the metamaterial, this study measures the radiation patterns to verify its spatially beam-splitting characteristics. Index TermsMetamaterial, near-zero refractive index medium, periodic structures, spatial beam splitter.

I. INTRODUCTION Metamaterial is an articially engineered material that obtains its unusual electromagnetic properties from structure rather than its composition. Metamaterial is generally constructed by embedding specic inclusions such as periodic structures in a host medium. Previous studies have applied metamaterial in waveguides and antenna designs [1], [2]. A metamaterial with a near-zero effective refractive index can reshape the far-eld pattern of an embedded antenna. A matched zero-index slab can also transform curved wave fronts into planar ones [3]. Metamaterial made of wire medium has been studied intensively and particularly on its effective refractive index, permittivity, and permeability. A structure composed of metallic mesh wires, which has very small electrical length in the period and wire thickness, can be treated as a homogeneous medium with a low plasma frequency [4]. Regarding the optical domain, a dielectric medium embedded with metallic nanoparticles and nano-wires has zero effective permittivity, creating band gaps [5]. Moreover, split ring resonators can produce an effective negative permeability over a microwave frequency band [6]. The rst lefthanded metamaterial in microwave frequency was developed while the extraordinary refraction phenomenon was also demonstrated [7]. Metamaterials with both negative permittivity and permeability over an overlapping near-infrared wavelength range have a low loss negative-refractive-index [8], [9]. A three-dimensional optical metamaterial made of cascaded shnet structures has a negative index within wide-band range [10]. Some researchers also used the effective medium method to consider the metamaterial slab as a uniform medium. Single-mode approximation can mathematically extract the effective parameters using the reection- and transmission-coefcients of the metamaterial slab [11]. Antenna design benets from the specic properties of metamaterial. For instance, a metamaterial consisting of six identical metallic
Manuscript received September 06, 2010; revised March 17, 2011; accepted July 02, 2011. Date of publication September 19, 2011; date of current version January 05, 2012. This work was supported by the National Science Council under Contract NSC 100-2221-E-009-138. The authors are with the Department of Electrical Engineering, National Chiao-Tung University, 300 Hsinchu, Taiwan (e-mail: raybeam@mail.nctu. edu.tw). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167913

Fig. 1. Structure conguration: (a) photo of the 3 D shnet metamaterial, and (b) front view of the shnet structure.

grids with a square lattice embedded in a foam [12]. They placed a monopole source in the middle of the structure, and a metal plate on the bottom of structure to control the emission. Experimental and numerical analysis proves that this metamaterial can modify the emission of an embedded source. Researchers have proposed and analyzed an epsilon-near-zero metamaterial for tailoring the phase of radiation pattern of arbitrary sources for some canonical geometries [13]. Other researchers studied a strongly modulated photonic crystal with an effective refractive index controllable by the band structure [14]. Experimental results prove that such a metamaterial can modify the emission of an embedded source and enhance its gain and directivity [15].

II. STRUCTURE CONFIGURATION Fig. 1(a) shows the two metallic parallel plates sandwiching the shnet structure. The numbers of the unit cell along the x- and y -directions are both 5; the periods (or lattice constant) along the x- and y -directions are dx (14.2 mm) and dy (14.2 mm), respectively. Each row or column of the 3 D shnet structure was constructed by the building block depicted in Fig. 1(b). Photo lithography and chemical etching processes were used to print a jungle-gym-like pattern on the dielectric substrate RO4003 with a relative dielectric constant of 3.55 and a thickness of 0.508 mm. The same pattern was printed on both sides of the dielectric substrate. This gure also shows the dimensions of the metal strip. The metallic parallel plates were made of aluminum measuring 1.6 mm thick, and the distance between them was h (12.8 mm). The experiments in this study excited cylindrical wave with a coaxial probe in this structure to observe its beam-splitting characteristic.

0018-926X/$26.00 2011 IEEE

418

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Phase relation of waves propagating in the metamaterial.

measurements in 1970 [16]. The scattering property of a homogeneous material can be determined by calculating its z and n from reectionand transmission-coefcients. Here z refers to the wave impedance normalized to 377
inside metamaterial whereas n refers to the effective refractive index. The effective medium approach extends same idea of characterization to metamaterial when it is regarded as a homogeneous material. For better characterization, many researchers have attempted to improve the retrieval processes. Existing methods [17][20] use the scattering parameters of a metamaterial slab under the normal incidence of a plane wave. This study performs the parameter retrieval analysis on a 3 D shnet structure with a periodic arrangement. However, the retrieval process may fail for several reasons such as inaccurate estimates of the effective thickness, relatively small scattering parameters, or incorrect determination of the branch index of the refractive index n. To avoid these issues, this study chooses a robust method [11] for effective parameter extraction. The effective permittivity and permeability are calculated from  = nz and  = n=z . Since metamaterial is a passive medium, the signs of computed z and n must satisfy the requirement

III. CHARACTERIZATION OF THE METAMATERIAL A. Phase Relation When analyzing a periodic structure, the dispersion (or phase) relation of the eigen-wave supported in the medium is an important issue. The boundary conditions of the top and bottom surfaces are perfect electric conductors. Since the metamaterial is assumed to have innite 2 D periodicity, the phase differences along the x- and y -directions can be expressed as kx dx and ky dy respectively, where phase constants kx and ky must comply with the Bloch-Floquet theorem, namely, the periodic boundary condition. Therefore, it is possible to solve the eigen-value problem of the closed structure using the nite-element method. Notice that the eigen-value (frequency) obtained corresponds to a propagating mode subject to every given combination of phase constants kx and ky . After iterating the phase constants, it is possible to determine the phase relation (the relationship among frequency, kx and ky ). Fig. 2 shows a contour map of the phase relation, in which the horizontal and vertical axes represent the normalized phase constants (kx =ko and ky =ko ) along the x- and y -axis, respectively. This gure also reects the symmetric properties of the unit cell. The colored circles with smaller radii represent the phase relations in the metamaterial calculated from 7.2 GHz to 8.0 GHz with 0.2 GHz step, while the circle with larger radius is the phase relation in air for reference. Apparently, the neff of a wave in the metamaterial is much smaller than that in air. No real root (frequency) is apparent below 7 GHz due to the cutoff phenomenon (plasma-like property) of this structure. The phase relation above provides physical insights into wave propagation at the interface. For easy interpretation, we regard the metamaterial as a homogeneous medium with a refractive index neff . The incident and transmitted angles in the medium and air are denoted by i and t respectively. According to Snells law, the relation between the incident and transmitted angles is as follows.

Re z Im n

f g f g

0 0:

(2) (3)

Thus z and n can be determined independently by

z= e0jk nd = n=

0 p 0X X 6j
1 1

(1 + S11 )2 (1 S11 )2
2

0S 0S

2 21 2 21

(4) (5)

0jk nd ) [Im ln(e ko d 0jk nd ) ] + j [Re ln(e

neff sin i = sin t

(1)

Since neff 1, the transmitted angle (t ) must be near zero. This means that the transmitting wave is likely perpendicular to the interface between the two media. B. Effective Medium Approach for Extracting the Uniform Transmission Line Parameters Nicolson and Ross proposed a method for determining the permittivity and permeability of a homogeneous material using time-domain

2 2 where X = (1 S11 + S21 )=2S21 , S11 and S21 individually represent the reection- and transmission-coefcients. The terms ko and d are the free-space wave-number and slab thickness, respectively. Since this study extracts effective parameters of the shnet-type metamaterial under the normal incidence of a TEM wave in a parallel-plate waveguide of height h, the analysis can be simplied by considering only the unit cell inside a waveguide with two horizontal PEC walls and two vertical PMC walls. Solving (6) directly may lead to an ambiguous result because the logarithmic function itself is multi-valued. Therefore, the choice of branch index m is critical and affects the real part of n. Nevertheless, the continuity of both permittivity and permeability as functions of frequency is an essential condition that helps determine the branch index. This study employs an iterative method to address this problem, as noted in [11, sect. D]. The constitutive effective parameters of metamaterial are shown in Fig. 3 using (4)(6). The plasma-like dielectric function is negative for the operational frequency below 7 GHz, while the effective permeability is positive. Consequently, the wave should have a pure imaginary number of propagation constant (refractive index) as depicted in (c), exhibiting the below-cutoff phenomenon. Interestingly, the effective refractive index is zero at 7 GHz, while the impedance is extremely high. Once the wave is excited in the effective medium at 7 GHz, the high wave impedance causes a strong reection from the interface between the effective medium and air. Specically, the wave impedance inside metamaterial is difcult to match the free-pace wave impedance, and therefore we append the ared opening between shnet structure and free space as an impedance transformer. The effective medium approach can properly model the complicated structure as a

f0 f f

gg

+ 2m ]

(6)

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

419

Fig. 3. Retrieved constitutive effective parameters of the 3 D shnet metamaterial: (a) permittivity, (b) permeability, (c) refractive index, (d) normalized impedance.

Fig. 4. (a) Poynting vector distribution in the structure with metamaterial and (b) Electric-eld strength in the structure with metamaterial.

uniform dielectric medium, providing uniform transmission-line parameters when the operational wavelength exceeds the period of a periodic structure. After identifying the phase- and dispersion-relation of eigen-waves supported in the metamaterial, it is possible to observe the physical picture of wave process. By properly tuning the length of the monopole (line source) in the metamaterial, it is possible to diminish the reection between the coaxial probe and whole structure. The refractive wave should then be normal to the metamaterial surface based on previous predictions. Therefore, this study calculates the electric-eld distribution and vectorized Poynting power in the parallel-plate region to verify this conjecture. Fig. 4(a) shows that the refracted wave leaving the metamaterial (highlighted by a grid region) tends to be perpendicular to the metamaterial surface, particularly around the central part. On the other hand, the wave propagates with a small angle deviated from direction perpendicular to side faces of the shnet structure in the horizontal

plane for the reason that the effective refractive index of metamaterial is not perfect but approximate zero in our design. C. Radiation Characteristics This study demonstrates the spatial power-splitting characteristics of the original structure equipped with a ared opening on each of the four output ports. The ared opening in this design enhances the directivity of the original structure. As is well known in microwave and millimeter wave engineering, an E -plane horn antenna uses a ared opening to taper its dominant waveguide mode from the waveguide end to a large opening while maintaining eld uniformity. This study uses the same design to obtain a uniform electric-eld distribution on the ared opening. The radiation patterns along the X -Y and X -Z plane were measured in an anechoic chamber using a vector network analyzer (HP 8722 D) and standard horn antennas.

420

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

vector and electric-eld distribution in the metamaterial veries the property of near-zero refractive index. The excellent agreement between measured and calculated results conrms the design concept of this study.

REFERENCES
[1] N. Engheta and R. W. Ziolkowski, Introduction, history and selected topics in fundamental theories of metamaterials, in Metamaterials: Physics and Engineering Explorations, N. Engheta and R. W. Ziolkowski, Eds. Piscataway/Hoboken, NJ: IEEE Press/Wiley, 2006, ch. 1, pp. 541. [2] N. Engheta, A. Al, R. W. Ziolkowski, and A. Erentok, Fundamentals of waveguide and antenna applications involving DNG and SNG metamaterials, in Metamaterials: Physics and Engineering Explorations, N. Engheta and R. W. Ziolkowski, Eds. Piscataway/Hoboken, NJ: IEEE Press/Wiley, 2006, ch. 2, pp. 4385. [3] R. W. Ziolkowski, Propagation in and scattering from a matched metamaterial having a zero index of refraction, Phys. Rev. E, vol. 70, no. 4, pp. 046608046608, 2004. [4] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, Low frequency plasmons in thin-wire structures, J. Phys. Condens. Matter, vol. 10, no. 22, pp. 47854809, 1998. [5] N. Garcia, E. V. Ponizovskaya, and J. Q. Xiao, Zero permittivity materials: Band gaps at the visible, Appl. Phys. Lett., vol. 80, no. 7, pp. 11201122, 2002. [6] J. B. Pendry, A. J. Holden, D. J. Robbins, and W. J. Stewart, Magnetism from conductors and enhanced nonlinear phenomena, IEEE Trans. Microw. Theory Tech., vol. 47, no. 11, pp. 20752084, 1999. [7] R. A. Shelby, D. R. Smith, and S. Schultz, Experimental verication of a negative index of refraction, Science, vol. 292, no. 5514, pp. 7779, 2001. [8] S. Zhang, W. Fan, K. J. Malloy, S. R. Brueck, N. C. Panoiu, and R. M. Osgood, Near-infrared double negative metamaterials, Opt. Express, vol. 13, no. 13, pp. 49224930, 2005. [9] S. Zhang, W. Fan, N. C. Panoiu, K. J. Malloy, R. M. Osgood, and S. R. Brueck, Optical negative-index bulk metamaterials consisting of 2 d perforated metal-dielectric stacks, Opt. Express, vol. 14, pp. 67786787, 2006. [10] J. Valentine, S. Zhang, T. Zentgraf, E. Ulin-Avila, D. Genov, G. Bartal, and X. Zhang, Three-dimensional optical metamaterial with a negative refractive index, Nature, vol. 455, no. 7211, pp. 376379, 2008. [11] X. Chen, T. M. Grzegorczyk, B.-I. Wu, J. Pacheco, and J. A. Kong, Robust method to retrieve the constitutive effective parameters of metamaterials, Phys. Rev. E, vol. 70, no. 1, pp. 016608016608, 2004. [12] S. Enoch, G. Tayeb, P. Sabouroux, N. Gurin, and P. Vincent, A metamaterial for directive emission, Phys. Rev. Lett., vol. 89, no. 21, pp. 213902213902, 2002. [13] A. Al, M. Silveirinha, A. Salandrino, and N. Engheta, Epsilon-nearzero metamaterials and electromagnetic sources: Tailoring the radiation phase pattern, Phys. Rev. B, vol. 75, no. 15, pp. 155410155410, 2007. [14] M. Notomi, Theory of light propagation in strongly modulated photonic crystals: Refractionlike behavior in the vicinity of the photonic band gap, Phys. Rev. B, vol. 62, no. 16, pp. 1069610705, 2000. [15] B.-I. Wu, W. Wang, J. Pacheco, X. Chen, T. M. Grzegorczyk, and J. A. Kong, A study of using metamaterials as antenna substrate to enhance gain, PIER, vol. 51, pp. 295328, 2005. [16] A. M. Nicolson and G. F. Ross, Measurement of the intrinsic properties of materials by time-domain techniques, IEEE Trans. Instrum. Meas., vol. 19, no. 4, pp. 377382, 1970. [17] J. Baker-Jarvis, E. Vanzura, and W. Kissick, Improved technique for determining complex permittivity with the transmission/reection method, IEEE Trans. Microw. Theory Tech., vol. 38, no. 8, pp. 10961103, Aug. 1990. [18] D. R. Smith, S. Schultz, P. Marko, and C. M. Soukoulis, Determinas tion of effective permittivity and permeability of metamaterials from reection and transmission coefcients, Phys. Rev. B, vol. 65, no. 19, pp. 195104195104, 2002. [19] P. Marko and C. Soukoulis, Transmission properties and effective s electromagnetic parameters of double negative metamaterials, Opt. Express, vol. 11, no. 7, pp. 649661, 2003. [20] R. W. Ziolkowski, Design, fabrication, and testing of double negative metamaterials, IEEE Trans. Antennas Propag., vol. 51, no. 7, pp. 15161529, Jul. 2003.

Fig. 5. Reection coefcient (in dB) versus frequency of this spatial beamsplitting structure.

Fig. 6. Radiation patterns of the spatial beam-splitting structure on - planes at 7.5 GHz.

XZ

X -Z and

This study includes the numerical simulations to determine far-eld radiation patterns and verify the proposed design concept. Fig. 5 shows the reection coefcient having bandwidth of 0.2 GHz with VSWR less than 2. Fig. 6 shows both measured and simulated co-polarization patterns. Due to the smaller dimension along the ared-opening direction compared to the width along the x- (or y -) axis, the beam pattern has a wider beam-width along the X -Z plane. Although not shown in the gures, measurements of the cross-polarization patterns are much smaller than those of co-polarization. These gures show that the four cone-type beam patterns indeed can serve as a spatial beam splitter to distribute the input power in four directions. The excellent agreement between numerical and measured results conrm the near-zero-refractive-index properties of the metamaterial.

IV. CONCLUSION This communication develops an articial medium based on the 3 D shnet structure. Using the effective medium approach, the proposed approach regards a 2 D periodic structure as a homogeneous medium and extracts its effective parameters, including the permittivity, permeability, refractive index, and wave impedance. Interestingly, the medium possesses a near-zero refractive index. A spatial beam splitter incorporating this metamaterial was developed to distribute the input power in four directions. To determine the underlying physics of wave process in the metamaterial, this study analyzes the dispersion characteristics of the shnet structure. In addition, the Poynting

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

421

Extended Mode-Based Bandwidth Analysis for Asymmetric Near-Field Communication Systems

Youndo Tak and Sangwook Nam

AbstractAn extended mode-based analysis for near-eld coupled antennas is proposed. Based on this analysis, a method for estimating 3 dB bandwidth of near-eld communication (NFC) systems with non-identical electrically small antennas is also proposed. The estimated results are in good agreement with the results from a full EM simulation. Index TermsAddition theorem, antenna equivalent circuit, bandwidth estimation, electrically small antenna, near-eld communication (NFC), near-eld coupling.

I. INTRODUCTION According to the Shannon-Hartley theorem, the capacity of a communication system is strongly related to the 3 dB bandwidth of the system. Hence, in order to estimate the performance of a communication system, the bandwidth analysis of the system is required. When the near-elds of two antennas are overlapped, an efcient short-range communication system can be easily constructed, and the coupling of near-eld coupled antennas can be estimated by the addition theorem [1]. Especially, when an electrically small antenna (ESA) is used, the analysis of the antenna coupling is particularly easy to calculate, because the ESA predominantly generates the TE10 or TM10 mode [2], [3]. Recently, based on the addition theorem, the method for estimating the 3 dB bandwidth of the near-eld communication (NFC) system is proposed for the symmetrical case where the same small antennas are used for both the transmitting and receiving antennas [4]. In addition, it is shown that, by using the correction factor which is proposed for considering the impedance characteristics of the antenna, the estimation accuracy can be improved more than the method in [5]. However, in actual NFC applications, for example, the NFC-based e-health monitoring system which is mentioned in [6], the transmitter and receiver are placed in different circumstances, respectively. Therefore, the types or dimensions of the transmitting and receiving antennas may not be identical to each other, and thus the previously proposed method cannot be applied. Therefore, the extended coupling model for the asymmetric case, where different antennas are used for the transmitting and receiving antennas, respectively, needs to be proposed, and consequently, the mode-based bandwidth analysis has to be modied based on the proposed model. II. EXTENDED COUPLING ANALYSIS FOR NEAR-FIELD COUPLED ANTENNAS BASED ON THE ADDITION THEOREM The power ratios of each spherical modes generated by the antenna vary according to the type of antenna and its physical dimensions [7]. When the ratio of the radiated power for each spherical mode to the
Manuscript received March 11, 2011; revised June 21, 2011; accepted July 28, 2011. Date of publication September 15, 2011; date of current version January 05, 2012. This work was supported by the Korean government (MEST) under a National Research Foundation of Korea (NRF) Grant (No. 2011-0001270). The authors are with the School of Electrical Engineering and INMC, Seoul National University, Seoul 151-742, Korea (e-mail: ydtak76@gmail.com; snam@snu.ac.kr). Color versions of one or more of the gures in this communication are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAP.2011.2167946

Fig. 1. Mode-based equivalent circuit for the coupled antennas of different characteristics.

TEnm mode and the TMnm mode, respectively. In addition, when the input port of the antenna is matched, the coefcients of the normalized spherical mode functions are given by: aT E aT M
= R
rad TE rad PT E Iin = R Prad Iin rad PT M Iin = Rrad Prad Iin rad

rad PT E R = Prad Rrad (1) rad PT M rad RT M = Prad Rrad (2) where Rrad is the total radiation resistance, Prad is the total radiarad rad and PT M is the radiation power of the tion power, and PT E rad TE

total radiated power is known, the radiation resistance for each mode can be given as:

(3) (4)

rad = RT M

where Iin is the input current of the antenna, based on the denition of the generalized scattering parameters [2], [8]. If the transmitting and receiving antennas are designated as antenna 1 and antenna 2, respectively, and they are of different types or dimensions, the equivalent network for coupled antennas can be given as shown in Fig. 1, based on the network representation of the coupled antennas in [2]. In the network, each antenna is equivalently described as the transformer array, and the characteristics of each transformer are individually determined by the power ratios of spherical modes generated from each antenna.

0018-926X/$26.00 2011 IEEE

422

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Fig. 2. Equivalent circuit description of coupled non-identical antennas.

Hence, the mutual impedance components of the coupling network between some spherical modes can be given as:

III. MODE-BASED ANALYSIS OF BANDWIDTH FOR THE NEAR-FIELD COMMUNICATION (NFC) SYSTEM In the previous study, the bandwidth analysis for the NFC comprised of symmetrical antennas is proposed by using the network analysis method for symmetrical 2-port networks based on the even-odd mode analysis [4]. However, when different antennas are used for transmitting and receiving antennas, respectively, the previously proposed analysis cannot be applied. An NFC system with asymmetrical transmitting and receiving antenna can be generally described using the equivalent circuit shown in Fig. 2. In the equivalent circuit, the tuned transmitting antenna is represented as the circuit composed of Ra1 and La1 , which are the resistance and the reactance of the antenna itself, respectively, with the tuning capacitor Cs1 . Similarly, the tuned receiving antenna is represented as the circuit composed of Ra2 , La2 , and CS 2 . Because the transmitting antenna and the receiving antenna are different to each other, the components representing each antenna are differently given. In addition, the mutual impedance, which is reciprocal and composed of the real and imaginary components, is represented by L12 and R12 . From the circuit theory, the transducer power gain of the two-port network can be given as:

Z21

= I TE =
rad RT E

VT E

(ANT 2) (ANT 1)

I
1

=0

(ANT 2)

aT E

(ANT 2)

+ aT E
(ANT 1)

(ANT 2)

R rad = RT E
(ANT 1)

aT E
rad RT E
(ANT 2)

An m

nm

(5)

Z21

= I TM =
rad RT E

VT E

(ANT 2) (ANT 1)

I
1

=0

(ANT 2)

aT E

(ANT 2)

+ aT E
(ANT 1)

(ANT 2)

R rad = RT E

aT M
rad RT M

(ANT 1)

(ANT 2)

Bn m

nm

(6)

GT =

where the translation coefcients An m nm and Bn m nm are derived from the vector addition theorem [2], [9]. In addition, based on the duality of the electric eld and the magnetic eld, other parameters can be given as:

4RG RL Z21 2 2 (RG + Z11 )(RL + Z22 ) Z21 2

j j 0 j

(10)

Z21 Z21

where the source impedance and the load impedance are given as RG and RL , respectively, and the mutual impedance is given as Z21 [10]. From (10), the characteristic equation to nd the 3 dB fractional bandwidth can be given as:
4 2 ! ! ! a4 21!3 dB + a2 21! 3 dB + a1 21! 3 dB + a0 = 0 s s s

= =

rad RT M rad RT M

(ANT 1)

rad RT M rad RT E

(ANT 2) n m nm

(7)

(11)

when the resonant frequency of the antenna is given as:

(ANT 1) (ANT 2)

Bn m

nm

(8)

!s = (La1 Cs1 )00:5 = (La2 Cs2 )00:5

sL Qe1 = !R a1 = ! C1 R G s s1 G sL Qe2 = !R a2 = ! C1 R L s s2 L j!s L12 Qk1 = R G

(12)

Since the input current is common for all ports of antenna 1, as shown in Fig. 1, the mutual impedance between coupled antennas can be given as:

and the coefcients of the equation are dened in Table I with additional parameters, dened as: (13) (14) (15)

Z21 = Vout Iin

= =

n ;m

I =0 VT E

(ANT 2)

n ;m Iin

V2

n;m;n ;m

Z21

+ Z21
:
(9)

and
s Qk2 = j!RL12 L

+Z21

+ Z21

(16)

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

423

TABLE I PARAMETERS FOR PROPOSED CHARACTERISTIC EQUATION.

TABLE II POWER RATIO OF SPHERICAL MODES FOR THE SOLENOIDAL SMALL LOOP ANTENNAS AT 10 MHz

Fig. 4. Wide open-ended small helix antenna (a) physical structure; (b) reactance behavior.

Fig. 3. Solenoidal small antennas of different types (a) wide solenoidal small loop, and (b) narrow solenoidal small loop set to the receiving antenna.

TABLE III POWER RATIO OF SPHERICAL MODES FOR THE WIDE SMALL HELIX AND THE NARROW SOLENOIDAL SMALL LOOP ANTENNAS AT 20 MHz

under the condition that the correction factors are given as F1 and F2 , respectively, in order to consider the actual impedance behavior of the antenna with respect to the frequency variations for each antenna, as proposed in [4]. IV. EXAMPLES A. NFC Comprised of Different Solenoidal Antennas As an example, the proposed method is applied to an NFC comprised of two 10-turn solenoidal small loop antennas of different types. According to the ratio of the radius to the height, the antennas are classied as wide and narrow types, as shown in Fig. 3. For the transmitting antenna, the wide solenoidal loop antenna is used, whereas the narrow solenoidal loop antenna is used for the receiving antenna. In addition, the operating frequency is set to 10 MHz, and the distance between the antennas is set to 0.5 m. The power ratios of the spherical mode are evaluated using the commercial EM software, FEKO, for each antenna, as shown in Table II. B. NFC Comprised of a Helix Antenna and a Solenoidal Antenna As another example, the proposed method is applied to an NFC comprised of one 10-turn open-ended helix, shown in Fig. 4(a), for the transmitting antenna, and a solenoidal small loop antenna, shown in Fig. 3(b), for the receiving antenna. The operating frequency is set to 20 MHz, and then the power ratios of the spherical mode are evaluated by the commercial EM software, as shown in Table III. However, in contrast to the solenoidal loop, which can be equivalently described as a single parallel resonator, the open-ended helix cannot be approximated by a single parallel LC resonator, because the reactance characteristic of the helix is given as shown in Fig. 4(b). According to the reactance behavior of the helix antenna with respect to the frequency variation, the antenna can be approximated to the equivalent circuit shown in Fig. 5. Hence, in order to consider the tuned open-ended helix as a simple RLC resonator circuit, its correction factor needs to be newly derived, as proposed in [4]. Based on the analysis proposed in [4], the correction factor can be dened as

F2 =

10

! !

! !

01

01

(17)

!as and the parallel resonant frequency of the antenna itself is given as !ap .
C. Comparisons The values of the 3 dB fractional bandwidth, which are calculated by the full EM simulation and the proposed method, are compared in Table IV. The congurations of the coupled antennas are classied into four cases according to the relative position and orientation of the antennas, as shown in Fig. 6. In addition, to be compared with other

where the series resonant frequency of the antenna itself is given as

424

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

TABLE V FRACTIONAL BANDWIDTH OF NEAR-FIELD COUPLED SMALL ANTENNAS IN PARALLEL CONFIGURATION

Fig. 5. Equivalent circuit representation of open-ended helix antenna.

coupling model and the general equivalent description of the coupled antennas. Additionally, from the results, it can be shown that the estimation errors for the collinear conguration are larger than others. They are thought to be due to the interaction between higher order modes, because the coupling between TE1;61 and TM1;61 is relatively large for the collinear conguration. V. CONCLUSION
Fig. 6. Congurations of two coupled antennas (a) parallel conguration, (b) diagonal conguration, (c) collinear conguration, and (d) diagonal conguration with the receiving antenna =4-rotated counter-clockwise about the positive y-axis.

TABLE IV FRACTIONAL BANDWIDTH OF NEAR-FIELD COUPLED SMALL ANTENNAS

In this communication, an extended mode-based analysis of an NFC comprised of non-identical antennas is proposed. The coupling of antennas can be calculated using the addition theorem, and the bandwidth is analyzed based on the equivalent circuit description. The estimated values of the 3 dB bandwidth of the NFC are compared with EM simulation results. This comparison indicates that the bandwidth can be accurately estimated by the proposed method.

REFERENCES
[1] W. Wasylkiwskyj and W. K. Kahn, Scattering properties and mutual coupling of antennas with prescribed radiation pattern, IEEE Trans. Antennas Propag, vol. 18, no. 6, pp. 741752, 1970. [2] J. Lee and S. Nam, Fundamental aspects of near-eld coupling antennas for wireless power transfer, IEEE Trans. Antennas Propag, vol. 58, no. 11, pp. 34423449, 2010. [3] Y. Tak, J. Park, and S. Nam, Mode-based analysis of resonant characteristics for near-eld coupled small antennas, IEEE Antennas Wireless Propag. Lett, vol. 8, pp. 12381241, 2009. [4] Y. Tak, J. Park, and S. Nam, Mode-based estimation of 3 dB bandwidth for near-eld communication systems, IEEE Trans. Antennas Propag, vol. 59, no. 8, pp. 31313135, 2011. [5] H. C. Jing and Y. E. Wang, Capacity performance of an inductively coupled near eld communication system, presented at the IEEE Int. Symp. Antennas Propag, San Diego, CA, 2008. [6] E. Strmmer, J. Kaartinen, J. Prkk, A. Ylisaukko-oja, and I. Korhonen, Application of near eld communication for health monitoring daily life, in Proc. IEEE EMBS Annual Int. Conf., New York, 2006, pp. 32463249. [7] Y. Tak, J. Park, and S. Nam, The optimum operating frequency for near-eld coupled small antennas, IEEE Trans. Antennas Propag, vol. 59, no. 3, pp. 10271031, 2011. [8] D. M. Pozar, Microwave Engineering. New York: Wiley, 2005. [9] W. C. Chew and Y. M. Wang, Efcient ways to compute the vector addition theorem, J. Electromagn. Waves Appl, vol. 7, no. 5, pp. 651665, 1993. [10] J. O. Rollet, Stability and power-gain invariants of linear two ports, IRE Trans. Circuit Theory, vol. 9, pp. 2932, 1962.

work, the estimation results by the proposed method and the method presented in [5] are compared in Table V for the parallel conguration without rotation of receiving antenna. The results shown in Table IV indicate that the 3 dB bandwidth of the NFC that uses non-identical antennas can be accurately estimated by the proposed method, which is based on the extended mode-based

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

425

Errata
Errata to Three-Dimensional Near-Field Microwave Holography Using Reected and Transmitted Signals
Reza K. Amineh, Maryam Ravan, Ali Khalatpour, and Natalia K. Nikolova I. INTRODUCTION In [1], detailed text was missing from Fig. 2. The correct gure now follows. The IEEE regrets this error.

REFERENCES
[1] R. K. Amineh, M. Ravan, A. Khalatpour, and N. K. Nikolova, Three-dimensional near-eld microwave holography using reected and transmitted signals, IEEE Trans. Antennas Propag., vol. 59, no. 12, pp. 47774789, Dec. 2011.

Fig. 2. Dielectric targets in free space scanned by two (at 35 GHz) horizontally-polarized ( -polarized) dipoles; dipole 1 is moving on the mm plane while dipole 2 is moving on the mm plane. The simulated -parameters are recorded in the frequency band of 25 GHz to 45 GHz for: mm and (a) two similar cuboids with sides of 3 mm, centered at and S/m; (b) four sim(0,4.5,25) mm with dielectric properties of mm, (4.5,0,11) mm, ilar cuboids with sides of 3 mm, centered at mm, and (0,4.5,27) mm and all having dielectric properties of and S/m; (c) two similar X-shape objects with square cross-sections 2 mm on a side and length of each arm 16 mm, parallel to the - plane, one centered at (0,0,27) mm with its arms along the and axes, the other one centered at (0,0,35) mm, with the arms rotated by 45 degrees with respect to the and axes, both targets having dielectric properties of and S/m; (d) two similar cuboids with sides of , centered at mm and and S/m. (4.5,0,35) mm, with dielectric properties Manuscript received November 30, 2011; accepted December 01, 2011. Date of current version January 05, 2012. The authors are with the Department of Electrical and Computer Engineering, McMaster University, Hamilton, ON L8S 4K1, Canada (e-mail: khalajr@mcmaster.ca). Digital Object Identier 10.1109/TAP.2011.2178917 0018-926X/$26.00 2011 IEEE

426

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

List of Reviewers for 2011

EVIEWING for the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION is an important service undertaken for the Antennas and Propagation Society. Those people who responded to the invitation to review, and who also submitted at least one review in the period between January 2011 and December 2011, are listed below. The names were compiled from information obtained from the ScholarOne Manuscript database. Grateful thanks are extended to all reviewers who are listed for their contribution to the IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION over the past year.
Antoniades, Marco Apaydin, Elif Apaydin, Nil Arai, Hiroyuki Arapoglou, Pantelis-Daniel Ares-Pena, Francisco Argyropoulos, Christos Arima, Takuji Armanious, Miena Arndt, Fritz Arrebola, Manuel Arunachalam, Kavitha Arvas, Ercument Athley, Fredrik Augustin, Gijo Austin, Andrew Awai, Ikuo Aydin Civi, Ozlem Azad, Mohammed Azaro, Renzo Azevedo, Joaquim B Baars, Jacob Babaeizadeh, Saeed Babolian, Esmail Baccarelli, Paolo Baczewski, Andrew Bagci, Hakan Bagri, Durgadas Baird, Christopher Bakr, Mohamed Balanis, Constantine Balling, Peter Balmain, Keith Bandaru, Subbarao Barba, Ismael Barba, Mariano Barba, Pedro Barbos, Afonso Barka, Andre Batchelor, John Baumann, Dirk Bayatpur, Farhad Bayka, Yahya Bayraktar, Zikri Bayram, Yakup Behdad, Nader Bekers, Dave Bengtsson, Mats Berenger, Jean-Pierre Bergmann, Jos Bernad, Laura Bernard, J.M.L. Bernhard, Jennifer Bernhardt, Paul Bertoni, Henry Best, Steven Bhattacharyya, Arun Bialkowski, Marek Bibby, Malcolm Biebl, Erwin Bird, Trevor S. Biro, Oszkar Blaunstein, Nathan Bleszynski, Marek Bluck Mike J. Boag, Amir Boccia, Luigi Boeringer, Daniel Bogaert, Ignace Boix, Rafael Bolomey, Jean-Charles Bolte, John Bonek, Ernst Boria, Vicente E. Bories, Serge Boriskin, Artem Borja, Carmen Boryssenko, Anatoliy Bosch, Xavier Bostian, Charles Botha, Matthys Bourlier, Christophe Bourqui, Jeremie Boutayeb, Halim Boyes, Stephen Boyle, Kevin Braaten, Benjamin Bray, Joey Bregman, Jaap Bressan, Marco Brizzi, Alessio Brockett, Timothy Brown, Gary Budaev, Bair Budimir, Djuradj Bunton, John Burfeindt, Matthew Burghignoli, Paolo Burke, Peter Burkholder, Robert Burr, Alister C Cabedo-Fabres, Marta Cahill, Robert Cai, Wei Cai Yong Camps, Adriano Capoglu, Ilker Capolino, Filippo Capozzoli, Amedeo Capsoni, Carlo Caramanica, Federico Caratelli, Diego Carlin, Matteo Carney, Scott Carozzi, Tobia Carr, Michael Carrasco, Eduardo Carrie, Guillaume Casaletti, Massimiliano Case, Joseph Castaldi, Giuseppe Casula, Giovanni Andrea Catedra, Manuel Cavallo, Daniele Cetiner, Bedri Chaharmir, Reza Chai, Mei Chair, Ricky Chakraborty, Swagato Chalas, Jeff Chambers, David Chan, Chi Chan, S. C. Chang, Dau-Chyrh Channabasappa, Eswarappa Chao, Hsueh-Yung Chatterjee, Deb Chattopadhyay, Goutam Chaudhuri, Sujeet Chavannes, Nicolas

A Abbas, Zulkiy Abbasi, Qammer Abbaspour-Tamijani, Abbas Abdel-Wahab, Wael Abdul Rahim, Sharul Aberle, James Abramovich, Yuri Abubakar, Aria Adalan, Ayse Adamiuk, Grzegorz Adams, Robert Aditya, Sheel Afsar, Mohammed Agar, Joshua Agio, Mario Ahmed, Iftikhar Ahmed, Shahid Akyurtlu, Alkim Al-Jaber, Sami Al-Tari, Muhannad Ala-Laurinaho, Juha Alatan, Lale Alayon-Glazunov, Andres Albani, Matteo Alexandridis, Antonis Ali, Mohammod Aliferis, Ioannis Alitalo, Pekka Allard, Rene Almajali, Eqab Alomainy, Akram Altintas, Ayhan Alu, Andrea Alvarez, Yuri Amano, Yoshiaki Amaya, Cesar Amendola, Giandomenico Ammann, Max Amra, Claude Anagnostou, Dimitris Anastassiu, Hristos T. Andersen, Jorgen Anderson, Stuart Andrenko, Andrey Andriulli, Francesco Anguera, Jaume Antar, Yahia

Digital Object Identier 10.1109/TAP.2011.2180569 0018-926X/$26.00 2011 IEEE

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

427

Che, Wenquan Chee, Kin Lien Chen, Canhui Chen, Chi-Chih Chen, Hongsheng Chen, Hua-Ming Chen, Ji Chen, Jiefu Chen, Kesong Chen, Kun Shan Chen, Nan-Wei Chen, Qiang Chen, Ru Shan Chen, Shih-Yuan Chen, Wen-Shan Chen, Wenhua Chen, Xiaodong Chen, Xudong Chen, Zhi Ning Chen, Zhiming Chen, Zhizhang Cheng, Qiang Cheung, S. W. Chew, Weng Chiu, Cheng-Nan Chiu, Chi-Yuk Chizhik, Dmitry Cho, Yong Heui Choi, Jaehoon Chou, Hsi-Tseng Christodoulou, Christos Chryssomallis, Michael Chu, Shuce Chu, Tah-Hsiung Chu, Yi Chuan, Lu Chuang, Huey-Ru Chung, David Chung, Kwok Chung, Shyh-Jong Chung, You Chung Ciattaglia, Matteo Cicchetti, Renato Cirstea, Silvia Claudio, Elio Clavijo, Sergio Clenet, Michel Coetzee, Jacob Colburn, Joseph Collado, Ana Collins, Brian Consoli, Fabrizio Constantinou, Constantinos Cools, Kristof Cooper, Ken Cooray, Francis Corona, Alonso Cortes-Medellin, German Costa, Filippo Costa, Jorge Costa, Mrio Costantine, Joseph Costanzo, Sandra

Costen, Fumie Cotton, Simon Craeye, Christophe Cranganu-Cretu, Bogdan Creticos, Justin P. Crocco, Lorenzo Cui, Suomin Cui, Tie Jun Cummer, Steven Czink, Nicolai D Derrico, Raffaele Durso, Michele Dault, Daniel Dauvignac, Jean-Yves Davidson, David Davidson, Tim De Hoop, Adrianus De La Rubia, Valentn De Lustrac, Andr De Menezes, Leonardo De Oliveira, Rodrigo De Zutter, Daniel Degli-Esposti, Vittorio Della Giovampaola, Cristian Demir, Metin Aytekin Demir, Veysel Deng, Fengshun Deng, Hai Deng, Tianwei Denidni, Tayeb Densmore, Arthur Derneryd, Anders Deutsch, Aina Di Giampaolo, Emidio Di Massa, Giuseppe Diaz, Rudy Dimitriou, Antonis Djordjevic, Antonije Djordjevic, Miroslav Dockery, George Donderici, Burkay Donelli, Massimo Dong, Junwei Dong, Qf Dong, Yuandan Dosopoulos, Stylianos Douvenot, Remi Dragoman, Mircea Dreher, Achim Du, Yang Duofang, Chen Dural, Gulbin Dussopt, Laurent Dvorak, Steven Dyab, Walid Dyczij-Edlinger, Romanus E Edelvik, Fredrik Ederra, Inigo Edwards, Robert

Eibert, Thomas F. Eisenstadt, William El-Shenawee, Magda Eldek, Abdelnasser Eleftheriades, George Ellingson, Steven Elmahgoub, Khaled Elnaggar, Michel Elsallal, Wajih Emiroglu, Caglar Encinar, Jose Eom, Hyo Erentok, Aycan Ergul, Ozgur Erricolo, Danilo Erturk, Vakur B. Eshrah, Islam Essaaidi, Mohamed Esteban Gonzalez, Hector Ethier, Jonathan Ettorre, Mauro F Faircloth, Daniel L.
Fakharzadeh Jahromi, Mohammad

Fujio, Shohei Fukasawa, Toru Fukusako, Takeshi Fumeaux, Christophe Fuschini, Franco Fusco, Vincent F. G Gagnon, Nicolas Galdi, Vincenzo Galli, Alessandro Gan, Yeow-Beng Gao, Jun Gao, Steven Garcia, Enrique Garg, Ramesh Gavrilovic, Minya Ge, Yuehe Gedney, Stephen Geffrin, Jean-Michel Geise, Robert Gennarelli, Claudio Genovesi, Simone Gentili, Guido Georgakopoulos, Stavros Georgiadis, Apostolos Gerini, Giampiero Getu, Beza Ghadarghadr, Shabnam Gheethan, Ahmad Ghoraishi, Mir Ghorbani, Kamran Ghosh, Saswati Gianvittorio, John Gillard, Raphael Ginn, James Giri, David Gjonaj, Erion Glazunov, Andres Alayon Gligorevic, Snjezana Godara, Lal Gogineni, Sivaprasad Gmez-Tornero, Jose Luis Gong, Xun Gonzalez, F. Sanchez Gonzalez Garcia, Salvador Gonzlez Ovejero, David Gonzlez Posadas, Vicente Gonzalez-Arbesu, Jose Maria Gonzalo, Ramon Gope, Dipanjan Gorbachev, Anatoly Goshi, Darren Goswami, Jaideva Goudos, Sotirios Goussetis, George Graglia, Roberto Gragnani, Gian Luigi Granet, Christophe Grau, Alfred Gray, Derek Grbic, Anthony Greenwood, Andrew

Falcone, Francisco Fang, Dagang Fang, Qianqian Fang, Wen-Hsien Farle, Ortwin Farr, Everett G. Farzaneh Koodiani, Sadegh Faseth, Thomas Feick, Rodolfo Feng, Yijun Feresidis, Alexandros Fernandes, Carlos A. Fernandez, Mario Fernandez-Recio, Raul Ferrer, Pere Ferrires, Xavier Fikioris, George Filipovic, Dejan S. Findeklee, Christian Fink, Patrick W. Fleury, Bernard Floury, Nicolas Foltz, Heinrich D. Foroozesh, Alireza Fort, Andrew Foster, Robert Fostier, Jan Fourn, Erwan Franchois, Ann Francis, Michael Frangos, Panayiotis Frei, Michael Freni, Angelo Frenne, Mattias Fuchs, Benjamin Fuchs, Jean Jacques Fujimoto, Kyohei Fujimoto, Mitoshi

428

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Gregson, Stuart Gremont, Boris Greving, Gerhard Grifn, Joshua Grzyb, Janusz Guan, Ning Guglielmi, Marco Guha, Debatosh Guinvarch, Rgis Guney, Kerim Guo, Lixin Guo, Y. Jay Guo, Yongxin Guo, Zhonghai Gupta, Inder Gupta, Ramesh Gurel, Levent Gurgel, Klaus-Werner Gustafsson, Mats Gustrau, Frank H Habib, Adnane M. Hadi, Mohammed Hagness, Susan Hall, Stephen Hallbjrner, Paul Haneda, Katsuyuki Hanham, Stephen Hansen, Robert Hansen, Thorkild Hanson, George Hao, Yang Hargrave, Chad Harrysson, Fredrik Hassan, Ahmed Hasse, Rainer Hateld, Michael Haupt, Randy Havrilla, Michael Hay, Stuart Hayakawa, Masashi He, Bo He, Shanhong He, Shiquan Heldring, Alex Hemmi, Chris Henault, Simon Hendrantoro, Gamantyo Hettak, Khelifa Heyman, Ehud Hirano, Takuichi Hirata, Akimasa Hirokawa, Jiro Hirose, Masanobu Hislop, Greg Hodges, Richard Hoenders, Bernard Hoffman, John Holloway, Christopher Holm, Peter Holopainen, Jari Hong, Wei

Hong, Wonbin Honma, Naoki Horii, Yasushi Hrabar, Silvio Hsu, Powen Hu, Jun Hu, Wenyi Hu, Xin Hu, Zhirun Huang, Yi Huang, Yikun Huff, Gregory Huggard, Peter Hui, Hon Tat Hum, Sean Huray, Paul Hussain, Malek Hwang, Huan-Sheng Hwang, Ruey I Icheln, Clemens Ichitsubo, Shinichi Ida, Ichirou Imbriale, William Imperatore, Pasquale Ingvarson, Per Iodice, Antonio Isernia, Tommaso Ishimiya, Katsunori Iskander, Mina Awar Islam, Md. Raqul Islam, Rubaiyat Ismail, Widad Itoh, Tatsuo Ivashina, Marianna Iwai, Hiroshi Iyer, Ashwin J Jacob, Martin Jakobus, Ulrich Jam, Shahrokh Jan, Jen-Yea Janaswamy, Ramakrishna Jandieri, Vakhtang Jeffrey, Ian Jeffs, Brian Jenn, David Jensen, Michael A. Jiang, Huiling Jiang, Lijun Jiao, Dan Jiao, Yong-Chang Jin, Cheng Jin, Jian-Ming Jin, Nanbo Jin, Ronghong Jofre, Lluis Jofre Roca, Luis Johnson, Joel Johnson, William Joler, Miroslav

Jonsson, Lars Jorgenson, Roy Joseph, Wout Joshi, Ravi Joubert, Johan Jouny, Ismail Joyner, Ken Jung, Baek Ho Jungnickel, Volker K Kabalan, Karim Kafesaki, Maria Kagoshima, Kenichi Kahn, Walter Kaifas, Theodoros Kainz, Wolfgang Kajfez, Darko Kalialakis, Chris Kalis, Antonis Kalluri, Dikshitulu
Kamarudin, Muhammad Ramlee

Kan, Hing Kanatas, Athanasios Kantartzis, Nikolaos Karasik, Boris Karedal, Johan Karimkashi, Shaya Karlsson, Anders Karlsson, Kristian Karwowski, Andrzej Kastner, Raphael Kasturi, Sreenivas Katsounaros, Anestis Kaufmann, Thomas Kaur, Guneet Kawakami, Haruo Kazemzadeh, Alireza Ke, Haixin Kefauver, William Keizer, Will Kennedy, Rodney Kerkhoff, Aaron Kesteven, Michael Khamas, Salam Khodier, Majid Kiani, Ghaffer Kidera, Shouhei Kildal, Per-Simon Kim, Hunseok Kim, Ilkyu Kim, Jaehee Kim, Jaeheung Kim, Kyungjung Kim, Oleksiy Kim, Sung Kim, Youngwook Kingsley, Nickolas Kinsey, Richard Kishk, Ahmed Kishor, Krishna Kiziltas, Gullu Klemm, Maciej

Knott, Peter Kobidze, Gregory Kodera, Toshiro Koh, Il-Suek Koh, Jinhwan Kolundzija, Branko Kong, Fannian Kononov, Victor Kooiji, B. J. Kornbau, Thomas Kosmas, Panagiotis Koustov, A. V. Koutitas, George Kowalski, M. Koyanagi, Yoshio Krishnasamy, Selvan Kubo, Shin Kucharski, Andrzej Kumar, B. Preetham Kuramoto, Akio Kurkcu, Harun Kurner, Thomas Kuwahara, Yoshihiko Kuzuoglu, Mustafa Kwon, Do-Hoon Kysti, Pekka L Lacoste, Frederic Lager, Ioan Lail, Brian Laitinen, Tommi A. Lakhtakia, Akhlesh Lamminen, Antti Landecker, Tom Langley, Richard J. Las-Heras, Fernando Lasri, Tuami Latif, Saeed Lau, Buon Kiong Lau, Jonathan Laurin, Jean-Jacques Lazaropoulos, Athanasios Le Caillec, Jean-Marc Le Coq, Laurent Leblebicioglu, Kemal Lee, Cheng-Jung Lee, Choon Lee, J. Lee, Jeong-Hae Lee, Kai-Fong Lee, Richard Lee, Teh-Hong Lee, Yee-Hui Lee, Yoonjae Leferink, Frank Legay, Herve Lehtinen, Nikolai Lei, Zhang Lempiainen, Jukka Leone, Giovanni Lesselier, Dominique Leung, Kwok

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

429

Levitas, Menachem Levy, Mireille Lewis, Dennis Lewis, Rob Lezar, Evan Li, Bin Li, Hui Li, Jia-Lin Li, Jingbo Li, Joshua Le-Wei Li, Long Li, Maokun Li, Morui Li, RongLin Li, Shaojing Li, Yang Li, Yue Li, Yujia Li, Zhaofeng Liang, Junli Liang, Ming-Cheng Linard, Martine Lier, Erik Lim, Eng Hock Lim, Sungjoon Lin, Jau-Jr. Lin, Yi-Cheng Lindberg, Peter Lindell, Ismo Ling, Hao Lippens, Didier Liseno, Angelo Litman, Amelie Litschke, Oliver Liu, Dawei Liu, Duixian Liu, Enxiao Liu, Jianguo Liu, Wei Liu, Wen-Chung Liu, Wendong Liu, Yang Liu, Yuan Liu, Zhe Liu, Zhiwen Liu, Zi-Liang Livieratos, Spiros Lizzi, Leonardo Llombart, Nuria Locatelli, Andrea Loison, Renaud Lomakin, Vitaliy Lombardi, Guido Long, Stuart Loredo, Susana Losada, Vicente Lovat, Giampiero Lovetri, Joe Lu, Caicheng. Lu, Jui-Han Lu, Mingyu Lu, Yilong Lucido, Mario

Ludick, Daniel Lui, Hoi-Shun Luini, Lorenzo Luk, K. Luo, Chong Luukkonen, Olli Luxey, Cyril Lyalinov, Mikhail Lysko, Albert M Ma, Hui Feng Ma, Tzyh-Ghuang Maaskant, Rob Maci, Stefano Mackowiak, Michal Madrid, Monica Mahmoud, Mohamed Mahmoud, Samir Mailloux, Robert Makarov, Sergey Makinen, Riku Maksimenko, Sergey Malyuskin, Oleksandr Manara, Giuliano Mangoud, Mohab Mani, Francesco Manikas, A. Manteuffel, Dirk Manteuffel, Dirk Mao, Yilin Maradudin, Alexei Marcano, Diogenes Marcos, Sylvie Marcus, Sherman Marengo, Edwin Marhefka, Ronald Marklein, Rene Markley, Loc Marrocco, Gaetano Martelly, Richard Martin, Anthony Martin, Ferran Martinez-Vazquez, Marta Martini, Enrica Martone, Anthony Maruyama, Tamami Marvin, Andy Marzetta, Thomas Marzuki, Marzuki Maslovski, Stanislav Masouri, Zahra Massa, Andrea Matekovits, Ladislau Mateo-Segura, Carolina Matolak, David Matricciani, Emilio Matsuzawa, Shin-Ichiro Mayhew-Ridgers, Gordon Mazzarella, Giuseppe Mazzinghi, Agnese Mcewan, Neil Mcgrath, William

Mclean, James Mei, Zicong Mcnamara, Derek Meaney, Paul Medina, Francisco Mehmood, Rashid Melapudi, Vikram Melde, Kathleen Mendes, Paulo Menzel, Wolfgang Merilampi, Sari Merkel, Sandra Mesa, Francisco Michalski, Krzysztof A. Michelson, David Michelusi, Nicolo Michielsen, Bastiaan Michielssen, Eric Michishita, Naobumi Migliaccio, Claire Migliore, Marco Miron, Douglas Mitri, Farid Miyamoto, Ryan Miyashita, Hiroaki Moghaddam, Mahta Mohan, Ananda Mohan, Anuraag Mohanan, P. Molisch, Andreas Mongia, Rajesh Mongiardo, Mauro Monorchio, Agostino Montisci, Giorgio Montoya, Thomas Morabito, Andrea Moraitis, Nektarios Moreira, Antonio Moreira, Fernando Morini, Antonio Morishita, Hisashi Moriyama, Toshifumi Mosca, Stefano Mouthaan, Koenraad Mrozowski, Michael Mumcu, Gokhan

Nguyen, Hoang Nickolaenko, A. Nie, Zaiping Nikitin, Pavel Nikolaou, Symeon Nikolic, Nasiha Nilavalan, Rajagopal Nishimori, Kentaro Nishimura, Naoshi Noghanian, Sima Nordholm, Sven Nosal, Samuel Nosich, Alexander Notaros, Branislav O Obukhov, Yuri Odendaal, Johann Oestges, Claude Ogawa, Koichi gc, Glge Ogurtsov, Stanislav Ohlmer, Eckhard jefors, Erik Okado, Hironori Okano, Yoshinobu Okhmatovski, Vladimir Okoniewski, Michal Oliveri, Giacomo Olsen, Robert Olver, David Omar, Abbas Ooi, Sooliam Osipov, Andrey Otero, Pablo Oughstun, Kurt Ozdemir, Caner Ozturk, Alper P Paiva, Carlos Paknys, Robert Palmer, Dev Palmer, Keith Pan, Bo Pan, Guang-Wen Pan, Helen Pan, Yong Mei Panagopoulos, Athanasios Panaretos, Anastasios Pang, Yong Paquay, Maurice Parini, Clive Park, Ikmo Park, Seong-Ook Park, Seongook Parker, Ted Parsa, Armin Pasian, Marco Pastorino, Matteo Patel, Parbhu Pathak, Prabhakar Paul, Dominique

N Nair, Naveen Naishadham, Krishna Nakano, Hisamatsu Nakhla, M. Nam, Sangwook Nanba, Shinobu Nanzer, Jeffrey Nasimuddin, Nasimmudin Navarro, Enrique Nayeri, Payam Nehorai, Arye Neto, Andrea Nevels, Robert D. Newman, Edward H.

430

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

Paulotto, Simone Pechac, Pavel Peixeiro, Custodio Peng, Zhen Pereda, Jose Pereira-Filho, Odilon Perez Fontan, Fernando Perregrini, Luca Perruisseau-Carrier, Julien Persson, Patrik Peterson, Andrew Petko, Joshua Petosa, Aldo Pham, Anh-Vu Philbin, Thomas Piazza, Daniele Pichon, Lionel Pierri, Rocco Pinchera, Daniele Pivnenko, Sergey Plicanic, Vanja Podilchak, Symon Pogorzelski, Ronald Polemi, Alessia Polimeridis, Athanasios Polivka, Milan Polycarpou, Anastasis Poon, Ada Popovic, Milica Porrat, Dana Port, Jorge Potter, Mike Pouliguen, Philippe Pozar, David Prieto-Cerdeira, Roberto Priou, Alain Psychoudakis, Dimitris Q Qi, Yunping Qiang, Rui Qing, Xianming Qiu, Cheng-Wei Quesada Pereira, Fernando Quevedo-Teruel, Oscar Quitin, Franois R Raisanen, Antti Raj, Rohith Rajagopalan, Harish Rajo-Iglesias, Eva Ramahi, Omar Ramirez, Raul Randazzo, Andrea Rao, Patnam Rao, Qinjiang Rao, Sadasiva Rao, Sudhakar Ravipati, Babu Rawat, Vineet Razavi-Ghods, Nima Rebeiz, Gabriel

Rego, Cssio Rehman, Samee Reid, David Rekanos, Ioannis Rengarajan, Sembiam Richards, Eric Richter, Andreas Ricketts, David S. Riera, Jose Manuel Rius, Juan Riva, Carlo Roblin, Christophe Rocca, Paolo Roden, J. Alan Rodrguez-Gonzlez, Juan Roederer, Antoine Rogers, David Rogier, Hendrik Romeu, Jordi Rothwell, Edward Row, Jeen-Sheen Rowell, Corbett Roy, Jasmin E. Rozanov, Konstantin Rubk, Tonny Rud, Leonid Rudish, Ron Ruvio, Giuseppe Rydberg, Anders Rylander, Thomas Ryu, Seungwoo S Sachs, Jrgen Sahalos, John Sahr, John
Sai Ananthanarayanan, Peruvemba Ranganathan

Sakakibara, Kunio Salghetti Drioli, Luca Salinas, A. Salmi, Jussi Salonen, Pekka Sanchez, Victor Sanchez-Hernandez, David Santalla, Veronica Sarcione, Michael Sarkar, Tapan Sarkis, Rmi Sarma, Achanta Sarris, Costas Sasamori, Takayuki Sato, Kazuo Sauleau, Ronan Sayeed, Akbar Sayidmarie, Khalil Scanlon, William Schaffner, James Schantz, Hans Schejbal, Vladimir Schmidt, Carsten Schmidt, Stefan Schneider, John

Schnepp, Sascha Schuchinsky, Alex Schulwitz, Lora Scire-Scappuzzo, Francesca Sebak, Abdel See, Terence Selleri, Stefano Selvanayagam, Michael Seo, Dong-Wook Sepulveda, Nelson Serebrennikov, Aleksey Serhir, Mohammed Serra, Andrea Sertel, Kubilay Sevgi, Levent Shafai, Leili Shafai, Lot Shahvarpour, Attieh Shaker, George Shaker, Jafar Shamim, Atif Shanker, B. Shao, Wei Sharaiha, Ala Sharma, Rajesh Sharma, Satish Shen, Zhongxiang Sheng, Xin-Qing Shiroma, Wayne Shlivinski, Amir Shore, Robert Siakavara, Katherine Sibille, Alain Sierra-Castaer, Manuel Sierra-Perez, Manuel Sievenpiper, Daniel Silveirinha, Mrio Simovski, Constantin R. Simpson, Jamesina Simpson, Ted Simsek, Ergn Singh, Surendra Singh, Surya Pal Sipus, Zvonimir Siqueira, Paul Siwiak, Kai Sjoberg, Daniel Sjblom, Peter Skobelev, Sergei Skrivervik, Anja Slone, Rodney Smely, Dieter Smith, David Smith, Glenn Smith, Paul Smith, Stephanie Smolders, Bart Solbach, Klaus Soldovieri, Francesco Soliman, Ezzeldin Song, Jiming Song, Wei Soras, Constantine

Sorolla, Mario Sorrentino, Roberto Sounas, Dimitrios Spencer, Quentin Spitsyn, Vladimir Staiculescu, Daniela Stancil, Daniel Sten, Johan Stephanson, Matthew Stevanovic, Ivica Steyskal, Hans Stolz, Peter Stuart, Howard Stupfel, Bruno Stutzman, Warren Suh, Seong-Youp Sullivan, Anders Sullivan, Dennis Sun, Din Kow Sun, Guilin Sun, Jwo-Shiun Sun, Mei Sun, Nian Sun, Weiwei Sun, Zhi Suntives, Asanee Sutinjo, Adrian Svantesson, Thomas Swenson, George Sydanheimo, Lauri T Tabatabaeenejad, Alireza Taboada, Jose Taove, Allen Taguchi, Mitsuo Taillefer, Eddy Takada, Jun-Ichi Takenaka, Takashi Tam, Wai-Yip Tang, Bo Tang, Philip Tang, Wenxuan Taniguchi, Tetsuki Targonski, Stephen Tarricone, Luciano Tastsoglou, Ioannis Tawk, Youssef Taylor, Graham Teixeira, Fernando Tentzeris, Emmanouil Thal, Herbert Thayaparan, Thayananthan Theron, Isak Thevenot, Marc Thiel, David Thomae, Reiner Thompson, John Tian, Ruiyuan Tobon, Luis Toccafondi, Alberto Tognolatti, Piero Toh, Wee Kian

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 1, JANUARY 2012

431

Tong, Kin Kai Tong, Meisong Topa, Antonio Topalli, Kagan Topsakal, Erdem Torres, Rafael Torrico, Saul Toso, Giovanni Toupikov, Mikhail Tretyakov, Sergei Trintinalia, Luiz Trott, Keith Trucco, Andrea Trueman, Christopher Tsang, Leung Tsiboukis, Theodoros Tsimbinos, John Tsimenidis, Charalampos Tsuji, Mikio Tsukerman, Igor Tufvesson, Fredrik Tuncer, Ozgur

Vescovo, Roberto Vigano, Maria Vinoy, Kalarickaparambil Vipiana, Francesca Virone, Giuseppe Vishvakarma, Babau Visser, Huib Volakis, John Volmer, Christian Vorobyov, Sergiy Voronkov, Maxim Vouvakis, Marinos Vrba, Jan W Wagen, Jean-Frederic Wahid, Parveen Walkenhorst, Brett Wallace, Jon Wang, Chao-Fu Wang, Chien-Jen Wang, Gang Wang, Hanyang Wang, Jianqing Wang, Johnson Wang, Junhong Wang, Le Wang, Rui Wang, Xiande Wang, Yan Wang, Yide Wang, Yong Wang, Yuanxun Ward, Jeffrey Warnick, Karl Waterhouse, Rod Webb, J. Webb, Matthew Wei, Fangzhou Wei, Guowei Wei, Kunpeng Weiland, Thomas Weile, Daniel Welch, Bryan Wells, Mike Wen, Geyi Werner, Douglas White, Daniel Whitman, Gerald Wiart, Joe Wiesbeck, Werner Wildman, Raymond Willis, Keely Wirflt, Petter Wolfe, Ct

Wong, Kainam Wong, Kin-Lu Wong, Silvester Woo, Jong-Myung Worms, Josef Wu, Bae-Ian Wu, Boping Wu, Chen Wu, Dagang Wu, Ming-Feng Wu, Qun Wu, Xuan Hui Wu, Yue Wyne, Shurjeel X Xia, Mingyao Xia, Tieqi Xiao, Shaoqiu Xin, Hao Xin, Jianguo Xu, Feng Xu, Junfeng Xu, Lisheng Xu, Luzhou Xu, Xiao-Bang Xu, Xiaojian Xu, Xin Xue, Zhenghui Y Yaghjian, Arthur D. Yakovlev, Alexander Yamamoto, Atsushi Yamamoto, Manabu Yan, Su Yanakiev, Boyan Yang, Chang-Fa Yang, Fan Yang, Guomin Yang, H.Y. David Yang, Kai Yang, Li Yang, Rui Yang, Shing Lung Steven Yang, Shiwen Yang, Xiaoling Yang, Xue-Xia Yapar, Ali Yarovoy, Alexander Yashchyshyn, Yevhen Yasumoto, Kiyotoshi Yazgan, Erdem Ye, Qiubo

Ye, Zhongfu Yegin, Korkut Yener, Namik Yeo, Junho Yeo, Tat Yeung, Sai Ho Yilmaz, Ali Yilmazer, Nuri Yin, Wen-Yan Yin, Xuefeng Ying, Lexing Ying, Zhinong Yla-Oijala, Pasi Yook, Jong-Gwan Yoon, Young Joong Young, John Young, William Yu, Chun Yu, Wenhua Yu, Yantao Yu, Yaxin Yuan, Naichang Yun, Jane Yun, Zhengqing Z Zaghloul, Amir Zang, Zhuquan Zapata, Juan Zeng, Qingsheng Zentner, Radovan Zhang, Bing Zhang, Hualiang Zhang, Shuai Zhang, Wen-Xun Zhang, Y. P. Zhang, Yan Zhang, Yan Zhang, Yong Zhao, Anping Zhao, Kezhong Zhao, Shan Zhao, Weijiang Zhao, Zhiqin Zheng, Jianfeng Zheng, Ming Zhong, Shun-Shi Zhong, Yu Zhou, Dawei Zhu, Aiming Zhu, Jiang Zieniutycz, Wodzimierz Ziolkowski, Richard Zvanovec, Stanislav Zwick, Thomas

U Uchimura, Hiroshi Ueda, Tetsuya Uher, Jaroslaw Ukkonen, Leena Ulazniuk, Tatsiana Umul, Yusuf Ur Rehman, Masood

V Valerio, Guido Valero-Nogueira, Alejandro Vallecchi, Andrea Valtr, Pavel Van Caekenberghe, Koen Van Cappellen, Wim Vant Klooster, Kees Vandenbosch, Guy Vanjani, Kiran Varadan, Vasundara Vardaxoglou, J(Yiannis) Vasiliadis, Theodore Vaskelainen, Leo Vasylyshyn, Volodymyr Vazquez Alejos, Ana Vecchi, Giuseppe Veidt, Bruce Venneri, Francesca Ver Hoeye, Samuel Verhaevert, Jo

Digital Object Identier 10.1109/TAP.2011.2181918

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION

Editorial Board Brenn Ellsworth, Editorial Assistant Michael A. Jensen, Editor-in-Chief e-mail: tap.editor@ieee.org Department of Electrical and Computer Engineering (801) 422-3903 (voice) 459 Clyde Building Brigham Young University Provo, UT 84602 e-mail: tap.editor@ieee.org (801) 422-5736 (voice) Senior Associate Editor Karl F. Warnick Dimitris Anagnostou Hiroyuki Arai Ozlem Aydin Civi Zhi Ning Chen Jorge R. Costa Nicolai Czink George Eleftheriades Magda El-Shenawee Lal C. Godara Associate Editors Yang Hao Andrea Massa Stuart G. Hay Derek McNamara Sean V. Hum Andrea Neto Ramakrishna Janaswamy George W. Pan Buon Kiong Lau Athanasios Panagopoulos Jin-Fa Lee Patrik Persson Kwok Wa Leung K.V. S. Rao Duixian Liu Shanker Balasubramaniam Satish Sharma Jamesina Simpson Mei Song Tong Jon W. Wallace Fan Yang Ali Yilmaz Zhengqing Yun Zhijun Zhang Yue Ping Zhang

INFORMATION FOR AUTHORS

MANUSCRIPTS should be specially prepared for publication in this TRANSACTIONS and submitted electronically via Manuscript Central, the manuscript administration software. The preferred format of the manuscript is in the IEEE two-column template with gures and captions included in the text. This template can be downloaded from http://www.ieee.org/web/publications/authors/transjnl/index.html. Please follow instructions given in the template. Color gures should be avoided unless the paper is intended to be published with color gures at extra cost. For details refer to Author Digital Tool Box on the IEEE website at http://wwww.ieee.org/web/publications/ authors/index.html. ELECTRONIC SUBMISSION of manuscripts is required for both initial review and for publication. To submit your manuscript electronically for review go to the Manuscript Central website at http://mcmanuscriptcentral.com/tap-ieee. For additional information about the electronic submission process and creating les for electronic submission, visit the AP TRANSACTIONS website at: http://ieeeaps.org/aps_trans/ COPYRIGHT: It is the policy of the IEEE to own the copyright to the technical contributions it publishes on behalf of the interests of the IEEE, its authors, and their employers, and to facilitate the appropriate reuse of this material by others. To comply with the U.S. Copyright Law, authors are required to sign and submit an IEEE Copyright Form before publication. This form returns to authors and their employers the full rights to reuse their material for their own purposes. Authors must submit a signed copy of this form before their manuscript can be published. The copyright form can now be completed electronically in Manuscript Central, and downloaded from the AP TRANSACTIONS website as a PDF le. PAGE CHARGES: All Authors are requested to honor page charges of $110 per printed page for the rst eight pages of a Paper and for the rst four pages of a Communication. Payment of page charges for this IEEE TRANSACTIONS, like journals of other professional societies, is not a necessary prerequisite for publication. The author will receive 100 free reprints (without covers) only if the page charge is honored. Detailed instructions will accompany the proofs. A Page Charge Form must be completed and returned to the Editor before a manuscript can be published. The page charge form, in PDF format, may be downloaded from the AP TRANSACTIONS website. CHARGES FOR EXCESSIVE LENGTH: A mandatory EXCESSIVE PAGE LENGTH charge of $200 per page is required for Papers longer than eight (8) printed pages and Communications longer than four (4) printed pages. If the author will not agree to pay mandatory charges, the author will be required to reduce the manuscript to eight printed pages or less (four printed pages for Communications) prior to nal acceptance and scheduling for publication. This reduction will occur after technical review of the manuscript. The above policy has no relationship to requests by the Editor to reduce very long manuscripts prior to technical review. PAPERS may be any length from one to about 25 typewritten double-spaced pages and are subjected to review by the Editor, Associate Editors, and their staff of reviewers. Each paper must have an abstract of about 150 words. Authors should allow approximately 12 weeks for the initial review cycle. Authors are encouraged to make their initial submission as short as possible. Because of limited space available in the TRANSACTIONS and the difculty in nding reviewers for very long papers, manuscripts (either single or multiple-part papers) that exceed 40 pages of double-spaced text and gures will not be considered. Upon being informed that your paper has been accepted for publication, be sure that the nal version is accompanied by a disk or CD. The electronic form of the paper should match the hard copy exactly and should include the references, biographies, footnotes, and gure captions. Most popular IBM-PC-compatible, Macintosh, and UNIX formats are acceptable. A more complete IEEE authors electronic guide or an IEEE\LaTeX style le can be obtained by email at help@ep.ieee.org or by fax at +1 732 562 0545. COMMUNICATIONS are narrower in scope and shorter than PAPERS. They will typically cover an extension of previous work or provide new insight on a specialized topic. As a result, the review of past work is expected to be limited to the key material. The manuscript length is normally up to four journal pages (two-column template) and should not exceed six journal pages. Page charges apply after four pages. COMMUNICATIONS are subjected to the same rigorous review as used for PAPERS. COMMENTS are brief notes containing remarks (and Authors replies) as well as corrections on previously published manuscripts. Usually, they are reviewed only by the Editor, and are published at his discretion. REVIEWS AND TUTORIALS are full length papers which need not contain original work. They primarily review the state of the art and developments leading thereto. Please contact the Editor prior to undertaking preparation of a review or tutorial. ILLUSTRATIONS should be kept to a minimum. Figures and photos should be original proofs and not photocopies. Photos must be glossy prints with no screens. Laser prints are not acceptable in place of photos or where gray scale graphics are involved. All line drawings and photos should be in black and white unless specically requested. Letters and symbols should be large enough to be readily legible when the drawing is reduced to a one-column width as much as 4:1 reduction from the original. Materials . For nal versions, electronic graphics must be provided as separate les. should be no larger than 22 28 cm PAGE PROOFS are sent to each principal author just prior to publication of the paper. Minor and typographical errors should be noted and corrected according to the instructions enclosed with the proofs. Major changes required to correct errors in the authors manuscript will not, in general, be permitted. In order to avoid the need for a major change, the author should proofread his manuscript very carefully. PUBLICATION POLICY: The author must ascertain that any clearance, company or foundation release, necessary for publication of the paper in this TRANSACTIONS has been obtained. If the paper or any modication of it has been published in the public literature, the author must indicate this in the letter of transmittal accompanying the submitted paper. NEWS ITEMS concerning the Society members and activities should be sent to the AP Magazine Editor-in-Chief, Dr. W. R. Stone, Stoneware Limited, 1446 Vista Claridad, La Jolla, CA 92037. Telephone: +1 619 459 8305.

Digital Object Identier 10.1109/TAP.2011.2181915

INSTITUTIONAL LISTINGS
The IEEE Antennas and Propagation Society is grateful for the support given by the organizations listed below and invites applications for Institutional Listings from other rms interested in the eld of Antennas and Propagation.

The charge for an Institutional Listing is $75 per issue, $375 for six consecutive issues, and $725 for 12 issues. Agency fee is not granted on Institutional Listings. Applications for Institutional Listings, should be sent to the Editor-in-Chief or Susan Schneiderman, IEEE Media, 445 Hoes Lane, Piscataway, NJ 08854, ss.ieeemedia@ieee.org. Please include name of company, company contact information, billing address, name of publication, dates of issue, artwork in the form of a pdf should be sent to Felicia Spagnoli, f.spagnoli@ieee.org, +1 732-562-6334.

Digital Object Identier 10.1109/TAP.2011.2181916