Professional Documents
Culture Documents
COPYRIGHT
It is the IETE policy to own the copyright of the scientific and technical papers it publishes on behalf of
the authors and their employers, and to facilitate the appropriate reuse of this material by others. Authors
are required to sign an IETE copyright transfer form before publication. A copy of this form is available
in the most recent January issue of IETE publications, and online in IETE website:www.iete.org. IETE
retains the authors’ and their employers’ right to reuse their material for their own purposes.
IETE TECHNICAL REVIEW
Published bimonthly by the Institution of Electronics and Telecommunication Engineers
Special Issue
on
‘Interactive ECTs - A Focus on Rural India’
CONTENTS
192 Guest Editorial 271 Building a Strong Nation - The ECT way
B L Deekshatulu R Sreehari Rao
257 Texture Feature Matching Methods for 335 Interactive Education through ECT - A
Content based Image Retrieval Focus on Rural India
Ivy Majumdar and B N Chatterji S Arumuga Perumal
Note : The Institution of Electronics and Telecommunication Engineers assumes no responsibility for the statements and
opinions expressed by individual authors and speakers.
191
Guest Editorial
Electronics & Communication Technologies (ECTs) is a vast canvas covering several inter connected
topics and technologies, almost everything we need in our daily life: digital communications, VLSI, tele-medicine,
tele-education, networking, signal processing, information for all etc. The theme of this special issue is
“Interactive ECTs — A focus on rural India”. Of the many topics this issue could cover, IETE received 11
invited papers from experts in this field, besides three contributed/short articles.
Prof Jhunjhunwala et al, in their paper “Broad band to empower rural India” mentions about the urban/rural
divide and that the TV has made known this difference to the rural masses as to where they stand. The paper
presents many Broadband wireless technologies that could alleviate this gap to a great extent by providing higher
data speeds and system gain.
Dr V U Reddy et al, in their paper, deal with the spatial multiplexing aspects of Multiple Input Multiple
output (MIMO) communication systems-an exciting development. Various architectures are discussed. Also,
brought out how fading, considered usually as undesirable phenomena, can be exploited to attain high spectral
efficiencies under MIMO.
Advances in Satellite communications, Remote sensing and GIS (Geographic Information System) have
provided solutions to various national development programs, such as disaster management, tele-education, tele-
medicine, creative 3D image representations of human organs etc, besides land use planning and integrated
rural/urban development. Mr A S Manjunath et al discussed how advanced sensor technologies and data
processing techniques have made this possible.
Rural broadband access is developed with the help of the widely available and highly cost reduced WiFi chip
sets. Prof Anurag Kumar et al, discussed in detail, the design issues related to Medium Access Control (MAC)
for packet voice telephony and for Internet access.
Dr Lal Kishore has given a review of the various VLSI technologies, research work currently in vogue, and
the future perspective. Hosting of Websites/portals with information useful to rural areas is proving to be a boon.
VLSI has its significant contribution in this.
Remote sensing with its high resolution and multi spectral sensors is providing valuable data/ information for
land cover classification. Pre processing and data processing techniques have also commensurately advanced
for analyzing such data. Prof Subhasis Chaudhuri et al, describe a temporally adaptive partially unsupervised
classification technique for remote sensing data.
Retrieving images, similar to a given query image, from a huge image data bank is a challenging task. This
is because, ultimately human perception / understanding is involved in the interpretation of the word “similar” for
validation/ evaluation of the retrieved images. Prof B N Chatterjee et al have presented, with a few examples,
a host of techniques currently available in content-based image retrieval (CBIR).
Government’s vision to make India a developed nation by 2020 is not a simple task. Dr Srihari Rao
enumerates the various ECT technologies that empower the rural people and transform them to a knowledge
powered PURA (providing urban amenities to rural areas).
Dr Surendra Pal et al, describe in their paper “Satellite Technology utilization for rural and urban India” that
space communication has changed society beyond imagination. Many technological reversals have been seen:
like the telephone which should have been on the wired network has become wireless, while the TV which was
on wireless earlier, now works on cable. Satellite technology connects the total country irrespective of location.
This paper describes also the services provided through satellites.
192
GUEST EDITORIAL 193
Testing in VLSI is important and complex. Prof Bhargav Bhattacharya et al developed a programmable
built in self-test for embedded memory cores for detecting neighborhood pattern sensitive faults including static
and active. A programmable BIST (built in self test) architecture is also designed.
Dr V Jayaraman et al, in their paper on “Use of GIS and MIS in participatory watershed development”
discussed about the Earth Observation (EO) inputs and spatial data modeling for integrated land and water
resource development at grass roots level. EO inputs are judicious mix of GIS and MIS (management
information system). Implementation and monitoring through webGIS tools in a project mode in an area in
Karnataka, has brought about significant impacts on natural resource conservation and living conditions in the
area.
Anuradha et al, in their contributed paper, discuss the simulation of Nakagami-m fading in wireless channels
using generalized SC DS-CDMA and MC DS-CDMA schemes, using binary phase shift keying modulation
techniques. It is observed that bit error rate decreases with increasing m for both SC DS-CDMA and MC DS-
CDMA.
This volume ends with two short papers, one on design of a FIR filtering core for high-speed application, and
another on interactive education through ECT for rural India.
I feel what we covered in this volume is only a drop in the ocean of the subject matter. We thank immensely
all the contributors of this special issue. Hope the reader gets a flavor of the ever changing/ growing ECT, and
the directions in which it is heading to reach the benefits to rural India.
Thanks to IETE for giving me this opportunity to be a guest editor and to interact with the many experts. The
publication wing of IETE headquarters have done an excellent job of getting this volume printed in time.
Prof B L Deekshatulu
Council Member & Distinguished Fellow IETE,
Guest Editor, IETE Technical Review,
Visiting Professor University of Hyderabad.
Residence Address : 10-3-123/3, East Maredpally,
Secunderabad 500 026 (AP).
Email deekshatulu@hotmail.com
Mobile: 9908499081
Guest Editor
Dr B L Deekshatulu has FASc, FNAE, FNASc, Distinguished Fellow IETE etc. Fellow
distinguished himself through his IEEE (USA), Fellow Third World Academy of Sciences (Italy).
research and technological contributions His current interests are Remote Sensing Data Analysis, Digital
in the field of Control systems, Digital Image Processing and Neural Networks.
Image Processing and Remote Sensing.
Dr Deekshatulu obtained BSc (Engg - He is a recipient of many awards such as the Bharat Ratna,
Electrical) degree in 1958 from the Sir M Visveswaraya Award for “Outstanding Engineer” in
Banaras Hindu University and ME and 1984, NRDC Invention Awards in Jan 1986 and in Aug 1993,
PhD degrees from the Indian Institute and Dr Biren Roy Space Science Award in 1988, “Padamsri”
of Science (IISc), Bangalore. He was medal in Jan 1991 by President of India, Brahm Prakash Medal
awarded Martin Foster Medal by IISc, for Best PhD thesis. He for significant contributions to Engineering Technology, Om
joined as Lecturer in 1964 at IISc., Bangalore and became Professor Prakash Bhasin Award for Science and Technology for 1995.
in 1970 and continued in that position till 1976. Dr Deekshatulu Received “Sivananda Eminent Citizen Award” from Vice
visited USSR during 1968 as a Government of India delegate for President of India in Dec 1998 and Gold Medal from Indian
setting up School of Automation at IISc, Bangalore. Geophysical Union 1997. He was Chairman, National Committee
on International Geosphere, Biosphere Programme (IGBP) during
Dr Deekshatulu worked as Visiting Scientist at the IBM 1994-97, Chairman of Remote Sensing Applications Missions,
Watson Research Centre, York Town Heights, New York, and India 1987-96. Awarded Boon Indrambarya Gold Medal by
at the Environmental Research Institute of Michigan during Thailand Remote Sensing and GIS Association in November
1971-72 on Digital Image Processing and Remote Sensing. He 1999 in Hong Kong for contributions to Remote Sensing. Awarded
designed and fabricated for the first time in India, Grey scale and 2002 Aryabhatta Award by Astronautical Society of India, for
color Drum Scanners for Computer Picture processing which lifetime contributions to Remote Sensing. Awarded Distinguished
has subsequently won him and his group a NRDC Award. Alumni Award from IISc, Bangalore.
Dr Deekshatulu joined the National Remote Sensing Agency,
Hyderabad as Head, Technical Division in April 1976 and Dr Deekshatulu was a UN / FAO consultant in Beijing
became Director in January 1982 - promoted to the grade during November, 1981. He was the Government representative
“Outstanding Scientist” in July 1989 and “Distinguished in the UN/ESCAP/RSSP Directors’ meetings and Inter
Scientist” (grade of Secretary) in July 1995 and retired from Governmental Consultative Committee meetings from 1985-95.
NRSA in October 1996. He has been responsible for the He was a UN/ESCAP Senior Consultant during September-
upbringing of the National Remote Sensing Agency in all its November, 1996. He was Director of Centre for Space Science
facets and for executing National and State level projects in and Technology Education in Asia and the Pacific (CSSTE-AP),
many disciplines of Remote Sensing applications. Affiliated to the United Nations, IIRS Campus, Dehra Dun,
from November 1995 to April 2002.
He has over 130 research publications to his credit. He has
guided 12 PhD Scholars and over 60 MTech. Students’ Currently, Dr Deekshatulu is a Visiting Professor in the
dissertations. He has visited 27 countries in the world. He is Dept of Computer & Information Sciences, University of
Fellow of many Scientific/Engineering Academies such as FNA, Hyderabad pursuing research and teaching in Image pro.
194
INVITED PAPER
in rural India may be the only answer to this problem in to rural areas is not the cost of electronic equipment,
the short run. but is actually due to the following:
Telecom technology has advanced very rapidly. i) The most significant cost component is the
Even though only a small percentage of villages have site preparation and the erection of the tower.
reliable telecommunications connectivity today, the Infrastructure like roads and electricity may
situation is changing fast. With the rural thrust, it is have to be first set up. The towers are about
reasonable to expect that most villages in India will 40 m tall, and require considerable amounts of
have mobile coverage as well as broadband Internet expensive steel.
connection within the next three to four years. ii) The second highest contributor to the cost is
the electrical power infrastructure – RF cables
3. THE TELECOM SITUATION TODAY running to the top of the tower, the power
amplifiers, RF filtering and the transceivers.
The mobile revolution of the last five years has Roughly 55% of the cost of the base station
seen base stations sprouting in most towns, owned and equipment is in these RF components [4].
operated by multiple operators, including the state-
iii) The maintenance of cell site infrastructure
owned company BSNL. The base stations of BSNL,
requires local personnel who should be trained
as well as those of the new operators are also
to deal with the problems that arise in wireless
networked using optical fibre laid in the last five years.
equipment.
There is a lot of dark fibre, and seemingly unlimited
scope for bandwidth expansion. iv) Availability of ultra-low cost (ULC) mobile
phones at costs below Rs 1500/- with financing
The solid telecom backbone that knits the country packages.
ends abruptly at the towns and larger villages. Beyond
v) Proper distribution infrastructure for phones,
that, cellular coverage extends mobile telephone
SIMs, spares and accessories in the remote
connectivity only up to a radius of 5 km, and then the
areas, and availability of basic training to users
telecommunication services simply peter out. Cellular
so that they can use the phones properly.
coverage can and will grow, but this will depend on the
rate at which infrastructure costs and operating costs vi) Billing and collection infrastructure for pre-
drop, and the rate at which rural incomes rise. Fixed and post-paid subscribers.
wireless telephones have been provided in tens of
thousands of villages, but it would be safe to conclude If one accepts these as the real bottleneck, then it
that the telecommunications challenge in rural India is immediately evident that as soon as there is sufficient
remains the “last ten miles”. This is particularly true if GSM voice coverage across India, we are already
one were to include broadband Internet access in past the key hurdles for upgrading. The cell sites and
one’s scope, since the wireless technologies currently towers are set up and maintenance, distribution, user
being deployed can barely support dial-up speeds. training and billing/collection infrastructure put in place.
This then is the rural India in search of appropriate One cannot afford to deploy any new cell sites,
broadband wireless technology: characterized by fat but only add electronic equipment at existing cell sites.
optical-fiber POPa within 15-20 km of most villages, To deploy 3G equipment at a cell site, the Node B has
fairly homogenous distribution of villages in the plains, to be installed (instead of, or in addition to, the GSM
poor rural cellular coverage, and low incomes. The BTS). The cost of Node B equipment has been falling
last aspect makes the provision of basic by approximately 40% each year over the last 4 years.
telecommunications as well as broadband internet Taken together with the fact that 3G offers more
services all the more urgent, since ICT is an enabler capacity than GSM, the 3G Node B is just 50% more
for wealth creation. expensive today than the GSM BTS to deploy the
same voice capacity [4]. It has already been seen that
3.1. GSM and CDMA dominate today but 55% of the cost of base station equipment is in the RF.
3G will be available in time Since a single 3G channel of 5 MHz replaces many
GSM channels of 200 kHz required achieving the
Before we look at broadband technologies for same capacity – the RF costs of 3G systems should,
rural India, let us take a look at mobile technologies of over time, be lower than that of GSM systems. Thus
today (we will mostly focus on GSM, though CDMA 3G will eventually lead to cheaper equipment than
systems are also present in India today). It may not be GSM, resulting in mobile broadband infrastructure in
readily apparent that the bottleneck in rolling out services India.
ASHOK JHUNJHUNWALA et al : BROADBAND TO EMPOWER RURAL INDIA 197
4. BROADBAND FOR RURAL INDIA must usually be at least 40 m high even for line-of-
sight deployment, as trees have a height of 10-12 m
When considering any technology for rural India, and even in the plains, one can expect a terrain variation
the question of affordability must be addressed first. of at least 20-25 m over a 15-20 km radius. Taller
Given the income levels, one must work backwards to Base Station towers will help, but the cost goes up
determine the cost of an economically sustainable exponentially with height. A shorter tower will mean
solution. Approximately 200 households in a typical that the subscriber-end installation will need a 20 m
village having disposable incomes can spend on an mast. At around Rs 20,000/-, this is substantially costlier
average Rs 50-100/- per month for telephony and data than a pole, even if the mast is guyed and not self-
services. Assuming an average of two public kiosks standing. The cost of 250-300 masts of this type is
per village, the revenue of a public kiosk can be of the very high compared to the incremental cost of a 40 m
order of Rs 5000/- per month. Apart from this, a few tower over a 30 m one. With 40 m towers, poles are
wealthy households in each village can afford private sufficient at the subscriber-end, and only rarely need
connections. After providing for the cost of the terminals, to be more than 12 m high.
it is estimated that, a cost of at most Rs 15000/- is
In summary, for a cost-effective solution, the system
sustainable for the connection. This includes the user
gain should be of the order of 150 dB (at least for the
equipment, as well the per-subscriber cost of the
lower bit rates), a 40 m tower should be deployed at
network equipment and infrastructure (towers) linking
the fiber POP, and roof-top antennas with 6-12 m
the user up to the optical fiber POP.
poles at the subscriber-end. The cost per subscriber of
the tower and pole (assuming a modest 300 subscribers
4.1. Coverage, system gain, and cost of
per tower) is Rs 3000/-. This leaves about Rs 12000/-
towers
per subscriber for the wireless system itself, inclusive
It has been mentioned that one needs to cover a of both the infrastructure and terminal sides.
radius of 15-20 km from the fibre POP using wireless
technology. The ‘system gain’ is a measure of the link 4.2. What constitutes broadband?
budget available for overcoming propagation loss and
The Telecom Regulatory Authority of India has
penetration losses (through foliage and buildings), while
defined broadband services [4] as those provided with
still guaranteeing satisfactory system performance.
a minimum data rate of 256 kbps. At this bit-rate,
Mobile cellular telephone systems have a system gain
browsing is fast, video-conferencing can be supported,
typically of 150-160 dB, and achieve indoor penetration
and applications such as telemedicine and distance
within a radius of about 5 km. They do this with Base
education using multi-media are feasible. There is no
Station towers of 40 m height, which cost about Rs
doubt that a village kiosk could easily utilize a much
500,000/- each. If a roof-top antenna is mounted at the
higher bit-rate, and as technology evolves, this too will
subscriber-end at a height of 6 m from the ground,
become available. However, it is important to note
coverage can be extended up to 15-20 km. When the
that even at 256 kbps, since kiosks can be expected to
system gain is lower at around 135 dB, as in any line-
generate a sustained flow of traffic, 300 kiosks will
of-sight system, the coverage is limited to approx 10
generate of the order of 75 Mbps. This is a non-trivial
km and the antenna-height at the subscriber-end has
level of traffic to carry over the air in each Base
to be 10 m in order to clear the tree tops. This
Station, even with a spectrum allocation of 20 MHz.
increases the cost of the installation by about Rs 1000/
- per connection.
4.3. Suitability of broadband wireless
Thus, roof-top antennas in the villages are a must technologies
if one is to obtain the required coverage from the fiber
POP. A broadband wireless system will also need a One of the pre-requisites for any wireless
system gain of around 150 dB if it is to be deployed technology for it to cost under Rs 12,000/- is that it
with 6 m poles. This system gain may be difficult to must be a mass-market solution. This will ensure that
obtain at the higher bit-rates supported by emerging the cost of the electronics is driven down by volumes
technology, and one may have to employ taller poles in and competition to the lowest possible levels. As an
order to support higher bit-rates at distant villages. example, both GSM and CDMA mobile telephone
technologies can today meet the above cost target,
There is an important relationship between (however, an even lower cost is needed for a non-
coverage and the heights of the towers and poles, and broadband technology since the services provided are
thus, indirectly, their cost. The Base Station tower limited).
198 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
The third-generation cellular telephone technologies that increase the bit-rate by three times, while being
will probably continue to meet this cost target while backward compatible to the DECT standard. Thus,
offering higher bit-rate data services. However, they the spectral efficiency goes up three times when
will not be able to provide broadband services as compared to conventional DECT. Additionally, dual-
defined above (as at most they will provide 64 kbps to polarization antennas have been used to exploit
each user). polarization isolation while till operating within the
DECT MAC framework, and further double spectral
If one were to turn one’s attention to recently efficiency. More importantly, all this has been done
standardised broadband technologies such as while retaining the use of the low-cost DECT chipsets.
WiMAX-d (IEEE 802.16d) [5], it is found that at
present, the volumes are low and costs high. Of these, The system gain in Broadband corDECT for 256
WiMAX-d has a lower system gain than the others kbps service is 125 dB. This can be increased by a
(which are all around the required 150 dB). All of few dB, where required, by increasing the antenna
them will give a spectral efficiency of around 2 bps/ gain at the subscriber-end (which is currently 11 dBi).
Hz/cell (after taking spectrum re-use into account), This is sufficient for 10 km coverage under line-of-
and thus can potentially deliver 40 Mbps at each Base sight conditions (40 m tower for BS and 10-12 m pole
Station with a 20 MHz allocation. However, high cost at subscriber side). A repeater is used, as in the
due to low volumes is the inhibitory factor with these corDECT system, for extending the coverage to 25
technologies. km. The corDECT system, and now the broadband
corDECT system, both meet the rural price-
It is likely that one or more OFDMA-based performance requirement comfortably, but with the
broadband technologies will become widely accepted additional encumbrance of 10-12 m poles and one
standards soon. One such technology is WiMAX-e level of repeaters. The first-generation technology is
(IEEE 802.16e) [6] that is emerging rapidly. These proven in the urban and rural Indian environment, and
will certainly have a higher spectral efficiency, and much is known about how to deploy it successfully.
more importantly, if they become popular and successful, The Broadband corDECT system works with the
the cost will be low. However, it will be several years same deployment strategy. It is being deployed in
before a widely-adopted technology derives the benefits significant numbers beginning 2007.
of market size and the cost drops to affordable levels
for rural India. The obvious question is whether there 4.4.2. WiFi rural extension (WiFiRe) [8]: A
are alternatives in the interim that meet our performance new WiFi-based wide-area rural
and cost objectives. broadband technology
In recent years, there have been some sustained
4.4. Near-term Technologies efforts to build a rural broadband technology using
WiFi chipsets. The WiFi bit rates go all the way up
4.4.1. Broadband CorDECT technology to 54 Mbps. The system gain is about 132 dB for
11 Mbps service, and as in corDECT, one requires a
One of the earliest and most widely deployed 40 m tower at the fibre POP and 10-12 m poles at the
examples of such re-engineering is the corDECT subscriber-end. The attraction of WiFi technology is
Wireless Access System developed in India [7]. It the de-licensing of spectrum for it in many countries,
provided toll-quality voice service and 35/70 kbps including India. In rural areas, where the spectrum is
Internet access to each subscriber without the hardly used, WiFi is an attractive option, provided its
bandwidth having to be shared. The next-generation limitations when used over a wide-area are overcome.
Broadband corDECT system has also been launched
recently, capable of supporting 70 Mbps per cell with Various experiments with off-the-shelf equipment
a 5 MHz bandwidth (supporting 144 full-duplex 256 have demonstrated the feasibility of using WiFi for
kbps connections simultaneously). With this system, long-distance rural point-to-point links. The more serious
each subscriber will get 256 kbps dedicated Internet issue with regard to the 802.11 standard is that the
access, in addition to toll-quality telephony. These commonly supported MAC protocol is a Carrier Sense
systems are built around the electronics of the European Multiple Access (CSMA) protocol suited only for a
DECT standard, which was designed for local area LAN deployment. When standard WiFi equipment is
telephony and data services. used to set up a wide-area network, medium access
efficiency becomes very poor, and spectrum cannot
Broadband corDECT incorporates added be re-used efficiently, even in opposite sectors of a
proprietary extensions to the DECT physical layer base station.
ASHOK JHUNJHUNWALA et al : BROADBAND TO EMPOWER RURAL INDIA 199
A solution for this problem is to replace the MAC services by Next generation Mobile Networks
protocol of 802.11 with a MAC protocol more suited (NGMN) initiative in 2006. A Technical Report (TR
to wide-area deployment. Such a new MAC, christened 25.913) provides detailed requirements, which include
WiFiRe, has indeed been defined [8] carefully such downlink peak date rate of 100 Mbps within a downlink
that a low-cost WiFi chipset can still be used, and the spectrum of 20 MHz using the OFDMA technology.
in-built WiFi MAC in it can be by-passed. The new The uplink peak data rate is expected to be 50 Mbps
MAC can be implemented on a separate general- with a 20 MHz uplink spectrum using SC-FDMA
purpose processor with only a modest increase in technique. It is proposed to support 200 users in active
cost. With the new WiFiRe MAC, it is estimated that state in each cell. The users are expected to get high
using a single WiFi carrier, one can support about 25 performance even with mobility as high as 120 Kmph.
Mbps (uplink + downlink) per cell. This would be MIMO is expected to be used and an enhanced
sufficient for about 100 villages in a 10-15 km radius. multimedia service is expected to be a part of the
Repeaters, possibly operating on a different frequency, standard.
will be needed for covering more villages over a
greater distance. 4.5.3. 3G-PP2 [10]
deliver broadband to all villages. The bit-pipes would 5. IEEE Standard for Local and metropolitan area
be there. The challenge beyond would be use the bit- networks, Part 16: Air Interface for Fixed Broadband
Wireless Access Systems, IEEE Std 802.16™-2004
pipes to transform the rural economy.
(Revision of IEEE Std 802.16-2001).
REFERENCES 6. Draft IEEE Standard for Local and metropolitan area
networks “Part 16: Air Interface for Fixed and
Mobile Broadband Wireless Access Systems, IEEE
1. Ashok Jhunjhunwala, Next Generation Wireless For
P802.16e/D5, September 2004.
Rural Areas, IJRSP, Special Issue on Rural Wireless
Communication, vol 36, no 3, June 2007. 7. corDECT, Wireless access system, Technical report
of Midas Communication Technologies, Chennai,
2. Bhaskar Ramamurthi, Broadband wireless India, December 2000.
technology for rural India, IJRSP, Special Issue on
Rural Wireless Communication, vol 36, no 3, June 8. Paul Krishna, Varghese Anitha, Iyer Sridhar,
2007. Ramamurthi Bhaskar & Kumar Anurag, WiFiRe:
Rural Area Broadband Access using the WiFi PHY
3. G Venkatesh & Ashwin Ramachandra, Can 3G and a new MAC, IEEE Commun Mag (USA), Jan
technologies benefit rural India?, IJRSP, Special 2007 (accepted).
Issue on Rural Wireless Communication, vol 36, no
3, June 2007. 9. http//www.3gpp.org/ftp/pcg/Beijing workshop
presentation
4. http://www.dotindia.com/ntp/broadbandpolicy
2004.html 10. www.3g-pp2/CDG Press Release (Aug 15, 2006).
Authors
Ashok Jhunjhunwala is Professor Advisory Committee. His research interests are :
of the Department of Electrical Telecommunications and Wireless Systems; Technologies for
Engineering, Indian Institute of Rural Areas.
Technology, Chennai, India and was
department Chair till recently. He
received his BTech degree from IIT, * * *
Kanpur, and his MS and PhD degrees
from the University of Maine. From Bhaskar Ramamurthi, a founding
1979 to 1981, he was with Washington member of TeNeT, holds an MS (1982)
State University as Assistant Professor. and a PhD (1985) degree in Electrical
Since 1981, he has been teaching at IIT, Madras. Engineering from the University of
California, Santa Barbara, USA. In 1986,
Dr Jhunjhunwala leads the Telecommunications and he joined the faculty at the Department
Computer Networks group (TeNeT) at IIT Madras. This group of Electrical Engineering, Indian Institute
is closely working with industry in the development of a number of Technology Madras (IIT-M), his
of Telecommunications and Computer Network Systems. TeNeT alma mater. He had earlier graduated from
group has incubated a number of technology companies which IIT-M in 1980 with a BTech in
work in partnership with TeNeT group to develop world class Electronics Engineering. He is Dean (Planning) of IIT Madras
Telecom and Banking products for Rural Markets. since 2005, the youngest faculty member ever to become a dean
at IIT Madras.
Dr Ashok Jhunjhunwala has been awarded Padma Shri in
the year 2002. He has been awarded Shanti Swarup Bhatnagar
Award in 1998, Dr Vikram Sarabhai Research Award for the Image Coding using Vector Quantization, an area in which
year 1997, Millennium Medal at Indian Science Congress in the he did some of his early work, formed the subject of his doctoral
year 2000 and H K Firodia for “Excellence in Science & thesis. Soon after obtaining his doctorate, he joined the AT&T
Technology” for the year 2002, Shri Om Prakash Bhasin Bell Laboratories, USA, where he worked on problems in indoor
Foundation Award for Science & Technology for the year 2004, wireless communications.
Awarded Jawaharlal Nehru Birth Centenary Lecture Award by
Awarded University of California, Regents Fellowship for
INSA for the year 2006 and IBM Innovation and Leadership
1980-81 and 1981-82. The paper titled “Perfect-Capture ALOHA
Forum Award by IBM for the year 2006. He is a Fellow of
for Local Radio Communications” selected for reprinting in IEEE
INAE, IAS, INSA and NAS.
Press book on Key Papers in Multiple Access Communications.
Dr Jhunjhunwala is a Director in the Board of SBI. He is Joined the Fellowship of Indian National Academy of Engineering
also a Board member of several companies in India, including in the year 2000, awarded Vasvik Award for Electronic Sciences
TTML, BEL, Polaris, 3i Infotech, Sasken, Tejas, NRDC, and and Technology in the year 2000, awarded Tamil Nadu Scientist
IDRBT. He is member of Prime Minister’s Setup Scientific Award 2003 for Engineering and Technology.
ASHOK JHUNJHUNWALA et al : BROADBAND TO EMPOWER RURAL INDIA 201
Member of the Board of Directors in Midas company developing all components of mobile phone technology.
Communications Technologies, N-Logue Communications, In this role, he was responsible for an R&D team of 75 engineers
Valued Epistemics (P) Ltd and Hon Director and Member, developing GPRS/EDGE handset technology, with an annual
Governing Council, Centre of Excellence in Wireless budget of US $20 million. David has actively provided technical
Technology (CEWiT), Chennai. inputs to Ericsson efforts in 3G standardization, and has been
presenting extensively on 3G and cellular evolution. David’s
He is a member of National Frequency Allocation Review technical areas of expertise include 3G/3.5G cellular systems,
Committee 2001, QoS Specifications Committee of TRAI for broadband wireless systems, and DSP techniques for wireless
ILD VoIP Services and member Sectional Committee of INAE. communications.
His research interests: Modulation and coding for Mobile His technical contributions at Ericsson have resulted in 32
Communications; Wireless Communication Networks; Design US patents, 10 journal papers and 26 conference publications. In
and Implementation of Wireless Local Loop Systems. 1999, David received the “Ericsson Inventor of the Year” award,
the highest technical recognition within Ericsson. In addition, he
has received numerous management awards for contributions in
* * * the areas of technical management and intellectual property
rights. He has served as a technical expert for Ericsson in
David Koilpillai received the technology-related litigations.
BTech degree from IIT Madras, and the
In November 2003, David was elected as Fellow of the
MS and PhD degrees (all in Electrical
Indian National Academy of Engineering.
Engineering) from California Institute
of Technology, Pasadena, USA. As part of IITM’s TeNeT group, David’s research and
teaching focus is in the areas of DSP applications in wireless
In June 2002, David joined IIT
communications. He is involved in research and consultancy
Madras as a Professor of Electrical
projects in wireless and cellular systems. Some of David’s projects
Engineering, and is a member of the
are listed below:
TeNeT group. During January 2007-
July 2007, David is on leave from IITM and is working as the • DSP applications in cellular systems (WCDMA, HSPA,
Chief Scientist, Centre of Excellence in Wireless Technology LTE, CDMA 2000)
(CEWiT), a public-private R&D initiative of the Govt of India. o Algorithms for improved PHY performance – CDMA
At CEWiT, David is leading a national project – Broadband and OFDM
Wireless Consortium of India (BWCI), focusing on emerging
broadband wireless standards. Prior to joining IITM, David was o DSP techniques for software radio applications
at Ericsson USA for twelve years, where he held different o Cost-optimized GSM / EDGE basestation
technical and managerial positions. In 2000, he became the
Director of the Ericsson’s Advanced Technologies and Research
Department at RTP, North Carolina and a member of the global
management team of Ericsson Mobile Platforms, an Ericsson * * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 203-213
Under rich scattering environment, Multiple Input Multiple Output (MIMO) systems have
the potential to achieve capacities inconceivable by Single Input Single Output (SISO)
systems. This makes it one of the most exciting developments to have occurred in wireless
communications. Within a short duration, it has matured from a reasearch topic into a
technology to find a place in upcoming wireless communication standards. This paper
focuses on the spatial multiplexing aspects of MIMO. Starting from AWGN channels, we bring
out how fading, which is typically an undesirable phenomenon, can be gainfully exploited to
attain large capacities under MIMO configuration. We first briefly discuss some MIMO
architectures initially designed to realize a large portion of the theoretical MIMO capacity,
namely V-BLAST (Vertical Bell Labs Layered Space-Time) and D-BLAST (Diagonal BLAST),
and then describe briefly the conventional V-BLAST receivers and a receiver that attains
near optimal performance while keeping the complexity low. The combination of Orthogonal
Frequency Division Multiplexing (OFDM) and MIMO is introduced as a simple method to apply
MIMO communication in a delay spread environment. Taking the example of Wireless Local
Area Network (WLAN) standard IEEE 802.11n , we see how the research developments have
been incorporated in a practical system.
203
204 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Note from (1) that the two resources that control the Consider a MISO system with M t transmit
capacity are power and bandwidth. For fixed bandwidth antennas and a single receiving antenna. If we constrain
W, as power increases the total transmitted power to P as before and allot a
transmit power of P/Mt to each antenna, the capacity
Cawgn of MISO system, extending the SISO AWGN result,
¾ ¾ » log SNR bits/s/ Hz, (3) will be
W
where hi denotes the channel vector in ith fade where nmin = min (Mt, Mr) li2 is the ith eigenvalue
realization. In arriving at the above capacity, the same P
power allocation over each fade interval is assumed. of HH† and we substituted SNR for ¾ ¾ . Equation
N0W
When the transmission extends over large number of (14) conveys that the overall capacity of a deterministic
fade realizations, we consider it as a fast fading MIMO channel can be viewed as the sum of individual
scenario, and the diversity is obtained from the multiple capacities of nmin parallel channels. This capacity is
independent fade realizations. maximized when all the eigenvalues are equal. Noting
206 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
that singular values of H are the square roots of the fading, far from being disadvantageous, is essential to
eigenvalues of HH† and condition number of a matrix make high data rate communication possible.
with equal singular values is unity, the result (14)
states that a MIMO channel with lower condition 3. MIMO CAPACITY ACHIEVING
number will yield larger capacity. ARCHITECTURES
The deterministic case can be extended to the We now discuss briefly the transmit/receive
fading case in a straight forward manner assuming architectures that can achieve the capacity promised
only receiver CSI. We will also assume that the channel by MIMO system. We assume only receiver CSI as
tap variances are such that the average received this is the more practical scenario. Recall that a MIMO
power at each receiving antenna is same as the total channel decomposes into nmin parallel channels (see
transmit power P. The ergodic capacity of the (14)). This means that the total capacity is the sum
frequency flat fading MIMO channel (i.e., the channel capacity of the individual parallel channels, and outage
between each transmit-receive antenna pair has no occurs for a slow fading channel only when
delay spread) is given by [2]
é æ nmin æ SNR öö ù
é æ 1 öù Pout (R) = P ê ç S log ç1+ ¾ ¾ li2 ÷÷ < R ú
C = max E êlog det ç IMr + ¾ ¾ ¾ HK sH† ÷ú ë è i=1 è Mt øø û
Ks:tr{Ks}<P ë è N0W øû
(18)
bits/s/Hz, (15)
Thus, the outage depends not on the capacity of a
where tr {Ks} gives total transmit power from all single parallel channel but on the combined capacity of
antennas. It was shown by Telatar [2] and Foschini all the parallel channels. In the slow fading scenario,
[3] that under Rayleigh fading, the capacity of MIMO the capacity depends on individual realizations and
channels (ergodic capacity or outage capacity) scales some of these channel paths between transmitter and
linearly with minimum of transmit/receive antennas. receiver could be in fade. Then, the performance will
Thus, at high SNR, a 3 dB increase in power (and be dictated by these channels unless the data is coded
hence SNR) translates to nmin bits/s/Hz increase in over other channel paths also. It is with this perspective
spectral efficiency. To see this, consider a practical that Foschini [8] proposed the D-BLAST algorithm
scenario with Mr >> Mt and assume equal power from where each data stream passes through all transmit
all transmit antennas subject to the total power constraint. antennas so that the data passes through all channel
For the Rayleigh flat fading case, for each realization paths thereby ensuring full diversity. This is explained
of H, the columns of H tend to be orthogonal and for below for two transmit antennas case.
such case all the eigenvalues of HH† are equal to Mr.
Then, (15) reduces to Each segment of data stream is coded into two
blocks, A and B, and transmitted as shown in Fig 1. In
æ SNR ö the first time interval, only 1A is transmitted from the
C = nmin log ç 1 + ¾ ¾ M r ÷ bits /s / Hz, (16) first antenna. Soft decoding is performed directly as
è Mt ø
no transmission is present on the other antenna. In the
which shows that capacity scales with nmin. Thus, next time interval, 1B is transmitted from second
MIMO offers spatial multiplexing. antenna and 2A is transmitted from first antenna (see
Fig 1). This diagonal alignment of code words gives
To emphasize the significance of fading channels rise to the name D-BLAST (Diagonal BLAST). At
in the MIMO communications, let us look at the the receiver, 1B is soft decoded by treating 2A as an
capacity of an AWGN channel. In this case, for each interference to be suppressed. Finally, 1A and 1B soft
realization, all the channel taps will be equal and H outputs are used to decode segment-1. Once this
becomes a unit rank matrix, implying that there is only decoding is done, 1B is cancelled out from the data
one non-zero eigenvalue. Hence, from (14), received during second time interval. The situation for
segment-2 is similar to that when segment-1 started.
æ SNR ö The process is repeated. An MMSE (Minimum Mean
C = log ç 1 + ¾ ¾ ¾ l12 ÷ bits /s / Hz, (17)
è Mt ø Square Error) based D-BLAST achieves outage
capacity for any value of outage probability. However,
where l12 is the only non-zero eigenvalue. That is, in D-BLAST suffers from error propagation and requires
AWGN scenario, the capacity does not scale linearly strong codes to prevent it. Also, for small number of
with nmin. Thus, in the MIMO case, the presence of data segments to be transmitted, there is a wastage of
KRISHNAN & REDDY : MIMO COMMUNICATIONS 207
diversity order for the next decoded data will improve Table 1 shows the complexity of the various
by one assuming no error propagation. This procedure decoders for different constellation sizes in terms of
can be repeated iteratively till all the data are decoded. the number of complex multiplications involved. In
To minimize error propagation, the data layer with [11], the authors also propose a further modification to
maximum post-detection SNR (the layer most probable reduce complexity further (though at the expense of
to be decoded correctly) is decoded at each stage of performance) by modelling the problem in an equivalent
the iteration. The method can be applied to the MMSE real domain. These methods are named ZF-SIC-AXIS-
receiver also. SEARCH and MMSE-SIC-AXIS-SEARCH,
respectively. The performance of these algorithms is
4.4. Low Complex Near ML Receiver for V shown in Fig 3. Note that the plots of ZF-SIC-SEARCH,
BLAST MMSE-SIC-SEARCH and ML overlap, and therefore,
are not distinguishable.
Though ZF-SIC and MMSE-SIC algorithms can
be implemented with low complexity (see [10]), the All the theory we have considered so far has
performance of these receivers is poor under slow assumed frequency flat fading channels. The real
fading. On the other hand, the ML decoder exploits world channels, on the contrary, are typically delay
the entire receive diversity but its computational spread channels. For example, see [14] which specifies
complexity is prohibitive. For example, for a system
with Mt = Mr, the diversity of ZF and MMSE algorithms
is only 1. Though SIC technique does improve the TABLE 1. Complexity of ML, ZF-SIC, ZF-SIC
performance, the overall performance is limited by SEARCH for 4- and 16-QAM
error propagation. Hence, a lot of effort has gone into constellations
finding receivers with performance close to ML, but
with complexity comparable to that of ZF-SIC or M Complex Multiplications, 4-QAM
(16-QAM), Mt = Mr = M
MMSE-SIC. We now present one such algorithm
proposed in [11]. Although developed independently, - ML ZF-SIC ZF-SIC-SEARCH
the algorithm falls under the class of algorithms proposed 5 5 3
in [12] and also it is similar to [13]. gM M2 ¾ M3 ¾ M3 + ¾ g M 2
2 2 2
a set of indoor channel models. A simple way of in the frequency domain ensures that adjacent coded
overcoming this problem is to combine MIMO with data see independent channel fluctuations. Thus, the
OFDM which results in a frequency flat channel in interleaver, in some sense, attempts to realize the
each subcarrier bin. Spatial multiplexing is then applied benefit of a fast fading channel by taking advantage of
separately in each subcarrier bin. Thus, (12) can be fading across subcarriers.
rewritten for each subcarrier index m as
In the SISO case, it was shown by Zehavi ([16]
r(m) = H (m) s (m) + n (m) (20) and later by Caire [17]) that under fast fading
conditions, bit interleaved coded modulation (BICM)
To simplify the notation, we will drop the subcarrier is a good choice. The technique involves placing an
indices and use the same notation as in (12). It is interleaver at the bit level rather than at symbol level,
assumed that all the techniques discussed above are as done in symbol level interleaving. The bit level
applied in each subcarrier bin. interleaving helps to achieve better diversity as each
bit undergoes independent fading, whereas in a symbol
5. A PRACTICAL IMPLEMENTATION - level interleaver, only each symbol (group of bits)
IEEE 802.11N undergoes independent fading. Thus, in BICM , the
encoder is followed by an interleaver and this is
The IEEE 802.11n standard [5] is a MIMO followed by symbol modulation. This method is
extension for the WLAN standard [15]. In this section, employed in WLAN standard IEEE 802.11a [15].
we describe briefly the transceiver for 802.11n in BICM can also be extended to the MIMO case in a
relation to the MIMO architectures discussed in the straight forward manner. With these ideas, the
previous section. Typical channels, encountered by a baseband transmitter block diagram of 802.11n for
practical 802.11n system, are quasi-static or slow the case of 2 transmit antennas is as given in Fig 4.
fading, and as pointed out earlier, diversity becomes Only blocks relevant to spatial multiplexing are shown
important for good performance in the slow fading for simplicity. The BICM structure is evident from
scenario. Fortunately, with OFDM, the fluctuations in the transmit block. The interleaver takes advantage
frequency (arising out of the delay spread channel) of the fading across the subcarriers and the single
can be utilized to bring in additional diversity. An Forward Error Correction (FEC) coder (a
interleaver can be used to re-arrange the data loading convolutional encoder) ensures that the data goes
onto the subcarriers so that adjacent coded data are through all transmit paths so that transmit diversity is
transmitted on subcarriers far apart. If the coherence not lost. The corresponding baseband receiver is
bandwidth of the channel is small enough, interleaving shown in Fig 5.
Interleaver QAM
IFFT
Mapper
Convolutional stream
Encoder parser
Interleaver QAM
IFFT
Mapper
FFT
DeInterleaver
MIMO
Decoder, Soft
DeMapper Viterbi
Deparsing
FFT and soft Decoder
metric
calculation
DeInterleaver
FFT
g
¾ such pairs will be identified and the computation of
2
|| r – Ha ||2 will be performed only for these pairs. As
computations are done for bit values {0,1} and for i Î
{1,2}, the total number of distance metric calculations
will be 2g, thus significantly reducing the complexity
compared to the direct approach which would involve
g2 distance metric computations.
This can now be treated as a set of independent Fig 6 IEEE 802.11n performance at 130 Mbps with ZF and
ML based soft decoders
equations (ignoring the correlation between the noise
elements). Thus, the scenario becomes similar to the
SISO case except that the noise variances should be
accounted carefully. The covariance matrix of the
Channel estimation is done in the frequency domain. It
noise is E(ññ†) = s2 (H†H)–1. The LLR metric for the
can be seen that the ZF based soft decoder performs
kth bit of the ith layer is given by [19]
within 3 dB of the ML based soft decoder, while its
~
| ri – si|2 complexity is much lower.
LLR = min ¾ ¾ ¾ ¾ ¾ (24)
ai:ai Î Alk s2[(H†H)–1]ii 6. CONCLUSION
~ ~
where ri corresponds to the ith element of r. These In this paper, we have attempted to provide the
calculations are exactly similar to SISO case, and underlying motivation for MIMO communications.
hence, the low complexity approximations in [18] are Specifically, we have brought out how fading (rich
also valid for this case. A similar formulation for scattering environment) can be gainfully exploited to
MMSE case is given in [20]. As the LLR metric is achieve spectral efficiencies which are inconceivable
computed on individual streams, unlike in the ML case with a SISO communication. We also looked at
(section 5.1) where it is computed using all streams architectures proposed to achieve significant portions
jointly, the complexity is much lower. of this huge capacity - D-BLAST and its simpler
For the ZF-SIC and MMSE-SIC, though superior version V-BLAST. The typical V-BLAST receivers
in performance to ZF and MMSE, it is not are described as well as a low complexity receiver
straightforward to obtain good soft metrics due to the that bridges the gap between linear receivers, like ZF-
possibility of error propagation among the decoded SIC and MMSESIC, and the ML decoders for small
data. number of transmit antennas. Simulation results are
provided to demonstrate the performance of various
The performance of 802.11n with 20 MHz decoders.
bandwidth for data rate of 130 Mbps, with 2 transmit
and 3 receive antennas is shown for ML and ZF based We have described IEEE 802.11n [5] as a practical
soft decoders in Fig 6. The spectral efficiency is system where MIMO ideas are utilized. In this context,
approximately 6 bits/s/Hz. The performance is shown we have also presented soft Viterbi techniques [15]
for tgn-b and tgn-e channels ([14]). No doppler is for typical MIMO decoders. We have included some
introduced in the simulations. Training is performed to simulation results to illustrate the performance of
obtain symbol lock and carrier frequency offset. 802.11n with indoor channel models.
212 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Authors
V U Reddy joined IIT, Madras as He is a Fellow of the IETE, the Indian Academy of Sciences,
an Assistant Professor in 1972, and the Indian National Academy of Engineering, the Indian National
moved to IIT, Kharagpur as a Professor Science Academy, and the IEEE. He was the Chairman of Indian
in 1976. He held a Visiting Professorship National Committee for International Union of Radio Science
with Electrical Engineering at Stanford (URSI) (1997-2000). He was a Member of the Editorial Boards
University during 1979-82. On the of SADHANA (1992-95), a Proceedings of the Indian Academy
invitation from Osmania University, he of Sciences, the Indian journal of Engineering and Materials
joined the University in April 1982 as a Sciences (1995-98), and the Proceedings of IEEE (2000-2001).
Professor and Project Director of He received S K Mitra Memorial Award 1989 from IETE for the
Research & Training Unit for best research paper.
Navigational Electronics; he was its Founding Director. In 1988,
he moved to Indian Institute of Science (IISc), Bangalore as a * * *
Professor of Electrical Communication Engineering (he was the
department chair during 1992-1995). Since 1986, he held several Kalyana Krishnan received his
short-term appointments at Stanford University and University BTech degree from NIT, Kozhikode
of Iowa. After retiring from the IISc in August 2001, he joined (Kerala) in the year 2000 and ME degree
Hellosoft India Pvt. Ltd as Chief Technology Officer. During his in Telecom from Indian Institute of
tenure at Hellosoft, he worked on WLAN projects involving Science, Bangalore in the year 2003. Since
802.11b and 11a standards. In June 2003, he joined International 2003, he has been working in Hellosoft
Institute of Information Technology (IIIT), Hyderabad, as Pvt. Ltd, Hyderabad on WLAN
Microsoft Chair Professor where he started wireless standards. His focus areas include OFDM
communications research center. He returned to Hellosoft as the and MIMO technologies.
Chief Scientist in December 2005. His present work is focused
in synchronization and channel estimation in wireless systems,
and low complexity decoders for MIMIO systems.
The advances in Satellite Communication and Remote sensing has led to the
development of solutions for various national development programs like National Natural
Resource Management (NNRMS), disaster management, tele-education and e-Health care
using telemedicine. The advances in sensor technologies and image processing have
resulted in sensors with increased spatial, spectral and radiometric resolutions and provide
images with rich details for remote sensing applications.
India is a vast country with 29 states and 6 union territories with more than 1 billion
population, which is predominantly rural and distributed at distant geographical locations
apart from the high-density urban areas. Ensuring basic minimum health care to masses
living in remote rural areas is a top priority for any government. Tele-medicine using the
satellite communication as well as Fiber Optic communications is playing an important role
to address health care requirements of people living in remote areas, rural areas, soldiers in
battlefields etc. The Image processing techniques and advances in Database management of
images and text made it possible to create 3D virtual images of human organs by using
imaging techniques from MRI, CT & Ultrasound.
The universalisation of education has become the top priority, especially for the
developing countries. But the extension of quality education to remote and rural regions
becomes a Herculean task for a large country like India with multi-lingual and multi-cultural
population separated by vast geographical distances, and, in many instances, inaccessible
terrain. Satellites can establish the connectivity between urban educational institutions with
adequate infrastructure imparting quality education and the large number of rural and semi-
urban educational institutions that lack the necessary infrastructure.
This paper describes utilization of Satellite Communication and Remote sensing with
advanced techniques in providing solutions to areas of primary importance in rural
development such as harnessing of natural resources, education and healthcare.
Natural resources can be estimated and monitored 1.1. Role of Remote Sensing in Natural
using Remote Sensing Satellite data and this has been resources utilisation
useful in areas such as water resources, agriculture,
Any country’s development depends on its ability
Paper No 125-B; Copyright © 2007 by the IETE. to harness the natural resources like land, water,
215
216 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
forests etc. As India has an agriculture-oriented vertical networking, data formats, standards taken up
economy, the inputs provided in this area are invaluable to reap the full benefits of these organised data bases
to the farmers. The Remote sensing applications are at various levels. NNRMS Portal serves as the front-
diverse, and can be broadly classified into the following end for the NRR, enabling the users to interact and
areas: (1) Agriculture (2) Land and soils (3) Forestry obtain the needed data for their applications.
and ecosystem (4)Water resources (5) Geosciences
(6) Infrastructure and town planning (7) Disaster Under NRC, the project of periodic inventory of
management. land use/land cover at 1:250,000 scale has been taken
up on annual basis using IRS AWiFS data. Assessment
In a country like India, which has diverse spreads has been successfully completed for first cycle and
in terms of geography and population, it is essential to second cycle. The detailed inventory of 7 natural
put satellite data to effective use and ensure that the resources themes (land use/land cover, soil,
advantages percolate to the grassroot levels for overall geomorphology, vegetation, snow/glacier, land
development. Management of natural resources, degradation, wetlands) at 1:50,000 scale using IRS
developmental planning, environmental monitoring and LISS – 3 data has also been initiated. The pilot project
disaster management are some of the key applications of Large Scale Mapping is nearing completion, and the
of space borne Remote Sensing. operational activity using Cartosat data has been initiated
for selected areas.
1.2. Natural Resources Repository (NRR) -
A space based spatial infrastructure The NRR databases have been used by district
programme under NNRMS and state officials for a variety of applications such
as, watershed development, wasteland development,
The National Natural Resources Management integrated land and water resources development
System (NNRMS), for which Dept of Space is the planning, road connectivity analysis, rural amenities
nodal agency, has the objective of ensuring optimal (schools, health Centres etc) planning, disease
management/utilisation of natural resources by infestation analysis, etc. It serves as a backbone
integrating information derived from remote sensing database for supporting the developmental needs.
data with conventional techniques. A large number of
remote sensing application projects in the fields of 1.3. The Village Resource Centre (VRC)
agriculture, forestry, environment, geology, ground Programme
water, disasters (flood, drought, earthquake & landslide),
etc., are being carried out under the aegis of NNRMS. The VRC programme is a unique societal application
conceived by Dept of Space, wherein the capabilities
With a large amount of spatial information
of satellite communication and earth observation are
generated under NNRMS, one of the important
integrated, to reach the variety of services emanating
requirements is to archive, retrieve and serve this
from the space systems and other IT tools directly to
information for various applications. Establishing and
the rural communities. The VRCs promote a need
maintaining a national Natural Resources Repository
based single window delivery system for the services
(NRR) - an infrastructure of distributed GIS data
in the areas of education, healthcare, agriculture,
servers, linked through high-speed network and making
weather, environment, disaster resilience, and livelihood
multi-scale natural resources information accessible
support to the rural population. It is also aimed to
through a NNRMS gateway Portal thus becomes
empower them towards improving their quality of
critical. The main aim of the NNRMS-NRR would be
lives. With the support of the local government and
to create and maintain a systematic archive of all the
enhanced participation by the communities, during the
digital spatial data holdings of thematic and base maps
course of time, the VRCs are also envisaged to provide
generated using remote sensing images and promote /
the e-government services, and act as a village level
encourage its use for government, business and societal
helpline.
needs.
As a part of the NRR programme as shown in Fig Space-enabled Village Resources Centre
1, the Natural Resources Census (NRC) project (VRC)
addresses periodic inventory of land-use / land cover;
bringing out Large-Scale Maps at 1:10,000 scale using VRCs, will be configured with satellite bandwidth
the high-resolution satellite remote sensing data; It based communications backbone, which will be multi-
also involves creating a Natural Resources Data Base tasked to enable a variety of community-centric
(NRDB) architecture taking care of the horizontal and services like tele-medicine, tele-education, spatial
A S MANJUNATH et al : ROLE OF SATELLITE COMMUNICATION 217
information for management of natural resources, institutions, who have strong and committed presence
disaster management through participatory approach, in the grassroots. The satellite connectivity and
and weather based farmers’ advisories. The VRCs bandwidth, minimal telemedicine and tele-education
have a interactive VSAT (Very Small Aperture facilities, and available/ customized spatial information
Terminal) based video-conferencing and information on natural resources along with indigenously developed
exchange facility, linking them with selected knowledge query system are provided by DOS. The task of
institutions/ centres/ specialty hospitals, etc., for housing, managing and operating the VRCs, with all
providing the above services. The services to be relevant contents are taken care by the partner
delivered by the VRCs are depicted in Fig 2. agencies.
The VRCs are being set up in large villages/ Till date, 180 VRCs have been set up in 11 States,
block/ taluk Hqs., in association with selected partner and another 160 VRCs are being set up. The various
areas addressed/ services provided by these VRCs and operated by the state governments and utilized by
include: agriculture development, fisheries development, all their departments. NGOs are visualised as other
supplementary teaching, adult & computer literacy, major users. Development and Educational
alternate livelihood related vocational training, marketing Communication Unit (DECU) is the nodal agency.
of agro-products, micro-finance/ enterprises, livestock
management, healthcare, etc. Encouraged by the Gramsat Pilot Projects are currently operational in
positive impact being made by the VRCs, plans are Madhya Pradesh, Gujarat, Karnataka, Orissa and
being drawn to set up VRCs in all the rural/ semi- three districts of Rajasthan. Planning and project
urban blocks of the country during the 11th Five Year implementation for other states like Meghalaya and
Plan period. other North-east Region, Goa, Andaman and Nicobar
Islands, Himachal Pradesh, Chattisgarh and Jharkhand
1.4. Satellite for rural development are underway. Presently there are seven teaching
(Gramsat Programme) ends (Uplink and Studio) and more than 2000 Direct
Reception Set (DRS) terminals set up by various state
The Gramsat Programme (GP) intends to governments, NGOs and other educational institutions.
provide a communications network at the state level
connecting the state capital to districts and blocks and 1.5. Geo Information in Disaster Manage-
enabling a reach to villages. The networks provide ment
computer connectivity, data broadcasting and TV
broadcasting facilities having applications like e- The remote sensing inputs have been used for
Governance, Natural Resource Information Service many disasters including drought, flood, earthquake,
(NRIS), Development Information, Tele-conferencing, cyclone, landslides, volcanoes, avalanches, forest fire,
Disaster Management and Rural/ education crop pest / diseases, etc., Similarly, cyclone and forest
Broadcasting. The GP aims to cover/connect the entire fires can be monitored and assessment of damage can
country – its each and every district and taluka in a be made. The information required for decision making
phased manner. Central agencies, the concerned state during drought is diverse and spatial/temporal in nature.
governments, their user departments, NGOs, etc. are But, synergistic coupling of remote sensing inputs with
the partners. The networks will be ultimately owned conventional systems and space communications, in
well-knit multi-energy interface, offers better Development Planning, Irrigation Command Monitoring,
operational services and decision support. For Forest Resources Assessment, Potential Fishing Zone
operationally integrating space inputs and services on Mapping and several others, have been realized by the
a timely basis Department of Space has launched a Deptt of Space. The databases of these projects can
major programme called Disaster Management Support be maintained and can be made accessible to users of
Programme (DMSP). interest for better planning.
Under DMSP, Decision Support Centre (DSC) of While remote sensing data is being utilised to
NRSA is providing remote sensing based services in prepare thematic maps/information on various resources,
near real time for monitoring, mapping and management thrust has been on integrating these information with
of natural disasters viz. Drought, Forest Fires, conventional data sets towards generating action plans
Landslides, Earthquakes, Floods and Cyclones using for sustainable development at local level. Integrated
satellite based information along with other ancillary Mission for Sustainable Development has helped in
and ground based information. Decision Support Centre preparation of location-specific action plan. Besides
at NRSA as a single-window operational service these, Environment Impact Assessment Studies like
provider for natural disaster management and support Forest Fire Mapping and Monitoring, Assessing the
the International Charter on Space and Major Disasters, Impact of Power Plants/Mining Areas, Atmospheric
as a signatory. The major activities and services provided Aerosol Studies, Coastal Zone Studies, all these form
by DSC are shown in Fig 3. an integral part of the resource planning. The concept
of e-governance with relevant inputs feasible from
1.6. Remote Sensing and GIS for e- remote sensing is depicted in Fig 4.
Governance
2. TELE-EDUCATION USING SPACE
National projects using remote sensing data, like TECHNOLOGY
Drought Monitoring, Crop Acreage and Production
Estimation, Mapping and Damage Assessment of The pivotal role of education as an instrument of
Disasters like Floods, Cyclones, Landslides, etc., social change by altering the human perspective and
Snowmelt Runoff Forecasting, Wasteland Mapping, transforming the traditional mindset of society is well
Urban Studies, Soil Mapping and Land Degradation recognised. The universalisation of education has
Studies, Geological Surveys, Mineral Targeting, Drinking become the top priority, especially for the developing
Water Potential Zone Mapping, Watershed countries. But the extension of quality education to
IRS Satelites
Toporgaphics Data Storage
and Existing & Retireval
Maps & Maps
Census Data
Data from
Census Modelling &
Network
Analysis
remote and rural regions becomes a herculean task services, a need was felt to launch a satellite dedicated
for a large country like India with multi-lingual and for educational service and ISRO conceived the
multi-cultural population separated by vast geographical EDUSAT Project in October 2002.
distances, and, in many instances, inaccessible terrain.
EDUSAT shown in Fig 5 is the first exclusive
Since independence, India has seen substantial increase
satellite for serving the educational sector. It is specially
in the number of educational institutions at primary,
configured for audio-visual medium, employing digital
secondary and higher levels as well as the student
interactive classroom and multimedia multicentric
enrolment. But the lack of adequate rural educational
system. The satellite has multiple regional beams
infrastructure and non-availability of good teachers in
covering different parts of India. The potential uses of
sufficient numbers adversely affect the efforts made
the EDUSAT are depicted in Fig 6.
in education.
EDUSAT is primarily meant for providing
Satellites can establish the connectivity between
connectivity to school, college and higher levels of
urban educational institutions with adequate
education and also to support non-formal education
infrastructure imparting quality education and the large
including developmental communication. ISRO will
number of rural and semi-urban educational institutions
provide technical and managerial support in the
that lack the necessary infrastructure. Besides
replication of EDUSAT ground systems to
supporting formal education, a satellite system can
manufacturers and service providers. The ground
facilitate the dissemination of knowledge to the rural
infrastructure to meet the country’s educational needs
and remote population about important aspects like
will be built and during this period, EDUSAT will be
health, hygiene and personality development and allow
able to support about 25 to 30 uplinks and about 5000
professionals to update their knowledge base as well.
remote terminals per uplink. The indigenous realisation
Thus, in spite of limited trained and skilled teachers,
and launch of EDUSAT has provided a substantial
the aspirations of the growing student population at all
capability for countrywide distance education in India.
levels can be met through the concept of tele-education.
The forerunner for this concept was the SITE (Satellite
3. e-HEALTH CARE USING TELEMEDICINE
Instructional Television Experiment), telecasted
progammes directly to over 2400 villages during 1975-
76, Subsequently in 1983 INSAT was used for similar India is a vast country gifted with ancient historic
tasks. background and geographically nature has provided
with all varieties like the mountain regions, deserts,
In the 1990s, Jhabua Developmental green planes and far-flung areas in the northeast and
Communications Project (JDCP) and Training and the offshore islands of Andaman and Lakshadweep.
Developmental Communication Channel (TDCC) For our population, which is predominantly rural and
further demonstrated the efficacy of tele-education. distributed at distant geographical locations apart from
With the success of the INSAT based educational the high-density urban areas, to provide the basic
minimum health care has been one of the priorities for Telemedicine as shown in Fig 7 is a confluence of
the health administration all along. communication technology, the information technology,
medical engineering and medical science. The
In today’s world with several advances made in
Telemedicine system consists of customized hardware
the medical field still the benefits are available to the
and software at both the Patient & Doctors end with
privileged few, residing mainly in the urban areas.
some of the diagnostic modalities like ECG, X-ray &
With the advent of communication technology especially
pathology provided at the patient end. They are
the SatCom combined with Information Technology,
connected through a communication backbone, which
we have means to benefit from the advance medical
could be VSAT or ISDN lines or leased lines or even
sciences even to the remote and inaccessible areas. It
a Telephone lines.
is known that 75% of the qualified consulting doctors
practice in urban centres and 23% in semi urban ISRO’s Telemedicine initiative has been broadly
centre (towns) and only 2% operate from rural areas, divided into the following areas:
where as the vast majority of population live in the
rural areas. Hospital beds/1000 people are 0.19 in a) Providing Telemedicine Technology &
rural as compared to 2.2 in urban areas. connectivity between remote/rural hospital and
Telemedicine helps patients in distant and remote Super Speciality Hospital for Teleconsultation,
areas to avail timely consultations from specialist Treatment & Training of doctors, nurses &
doctors without going through the ordeal of travelling technicians.
long distances at large expense. The facility caters b) Providing the Technology & connectivity for
normally for transmission of patient’s medical images, Continuing Medical Education (CME) between
records, output from medical devices, besides live Medical Colleges & Post Graduate Medical
two-way audio and video conferencing. With the help Institutions/Hospitals.
of these, a specialist doctor advises the local doctor or
a paramedic at the patient’s end, on line, on medical c) Providing Technology & connectivity for
care. In the context of distant and rural areas, the Mobile Telemedicine units for rural health
telemedicine-based medical care is also highly cost camps especially in the areas of ophthalmology
effective. and community health.
222 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
III SAT
Mobile Service
d) Providing Technology & connectivity for for demonstrating further ‘technology packaging’ and
service during disasters like Flood, Cyclone, implementation efforts - by way of ‘Point-to-Multipoint’
and Earthquake, etc. and ‘Multipoint-to-Multipoint’ type of connectives -
has been realized. These efforts would also facilitate
Further, towards reaching speciality healthcare to imparting training to local doctors and paramedics
larger sections of the society, across the country, and working in rural areas, besides supporting Continuing
in various faculties of medical specialities; the need Medical Education (CME) efforts.
ISRO has identified “Health Communication malaria etc., Remote Sensing provides useful indicators
Support Network” as an important activity. In this for identifying the presence or impact of the vectors,
connection, the types of services which are planned to which can be analysed to provide an effective solution.
be initiated are Tele-medical-education, Tele-
consultation, Tele-referral and Tele-health. ISRO has CONCLUSION
plans to integrate telemedicine with tele-education
wherever applicable and the Village information Kiosks/ The paper on “Role of Satellite Communicatgion
Community Centres to reach out to more rural areas & Remote Sensing in Rural Development” describes
of the country. The various tele-medicine links the efforts by the dept of Space in harnessing Space
established by ISRO are shown in Fig 8. Technology and Image processing and GIS in areas
that influence rural development viz, natural resources
3.1. Image processing in e-Healthcare utilisation, education and health care etc. As the current
trend is integration of technologies for enhanced utility,
Applications of telemedicine are found in every the space technology can be made integral part in
area of specialisation. Teleradiology is the most common various phases of rural development for the country’s
application followed by cardiology, dermatology, overall development.
psychiatry, emergency medicine, home health care,
pathology, and oncology, etc. The technological basis ACKNOWLEDGEMENTS
and the practical issues are highly variable from one
clinical application to another. The image processing We are grateful to Dr K Radhakrishnan, Director,
NRSA for his encouragement in bringing out this
methods like contrast enhancement, segmentation, publication.
registration, clustering, mosaic, 3D constructions,
rendering, etc., are applied for various medical images. REFERENCES
Other image processing techniques such as Image
acquisition, digitization and image compression are 1. http://www.isro.org
used effectively for image transmission. 2. http://www.nnrms.gov.in
3. Dr. Rao, K M M, et al, Emerging Technologies and
3.2. Remote sensing and GIS for vector
Remote Sensing with GIS for effective e-
borne Diseases Governance, Proceedings of National Conference
on e-Governance, 10th -12th September, 2004, IETE,
Vector-borne diseases have been the most Hyderabad-7.
important health problems worldwide for many years 4. Adam William Darkins, Margaret Ann Cary
and still represent a constant and serious risk to a large Telemedicine and Telehealth – principles, policies,
part of the world’s population. Satellite Remote Sensing performance and pitfalls, Springer publishing
techniques are used to monitor factors affecting disease company, 2002.
transmission. Both spatial and temporal changes in 5. Marlene Maheu, Pamela Whitten, Ace Allen, E-
environmental conditions are important determinants Health, Telehealth, and Telemedicine, Wiley
of vector-borne disease transmission. The satellite publishers, 2001.
images are used for identifying these changes, and 6. David L Verbyla, Satellite Remote Sensing of Natural
able to define and predict areas and periods of high Resources, CRC Publishers,1990
spreading. Satellite images, digitized land use maps 7. ISPRS proceedings – Istambul 2004, Volume XXXV,
and Geographic Information Systems (GIS) are used Parts B1-B6.
for predicting changes in habitats of disease vectors,
8. Proceedings of the National Workshop on Flood
and high resolution satellite data is used to characterize Disaster Management – space Inputs, Organised by
immature habitats of vector borne diseases like cholera, NRSA, ITC Netherlands, 2004.
224 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Authors
A S Manjunath obtained his BE He has been honoured with NRDC award from President of
degree from Bangalore University India for meritorious invention of Color Photowrite System;
(1970) and MTech from IIT, Madras Best Import Substitution and Best Invention awards for
(in 1972). He then joined Electronics technological developments from Exhibition Society, Hyderabad;
Systems Division of Space ‘The Outstanding Young Indian (TOYI)’ Award-1991 by the
Applications Centre, Ahmedabad in Indian Junior Chamber Secunderabad Chapter, AP State Chapter
1972 where he worked for the Satellite and IJC National body for his contribution and achievement in
Instructional Television Expt. (SITE). scientific & technological development; 2 patents; several design
He joined the National Remote Sensing articles and publications in international technical journals. He
Agency in 1975 and since then he is was selected in the final panel for Indian Astronaut to fly on-
with NRSA. He has worked in the areas of Image Processing, board US space shuttle ‘Challenger’ in 1986.
Microwave Remote Sensing, Artificial Intelligence, Data
Quality Evaluation and Quality Assurance activities at NRSA. He is a Senior Member of IEEE (for past 19 years) and life
He developed a real aperture X band Side Looking Airborne member of ISRS.
Radar System. Presently as Dy Director (Data Processing
Area), he is handling the data processing activities which * * *
include Product Generation, Data Quality Evaluation, Product
S Rajendra Kumar obtained his
Quality Control, Microwave Remote Sensing, Special Products
Diploma in Electronics &
generation, Imaging systems, QA activities, data archival,
Communication Engineering in 1974. He
Software development, hardware development, etc. He has
worked in Electronic Research Laboratory
made significant contributions in these areas of activities at as a technical assistant in the field of
NRSA. His research interests are in the image processing and VHF/UHF receivers for about five years.
microwave remote sensing areas.
He joined the National Remote Sensing
Agency in 1989 and since then he is with
* * * NRSA. He has worked in the areas of
D S Jain was born in December, Digital Film Recorders Development.
1953 in a village Mandalgarh in Rajasthan Trained on Digital film recorders at Canada in 1984 and 1987. He
State. He has done his BE (Hons) was also recipient of NRDC award for design and development
(Electronics in 1975 and ME of small format digital film recorder. Presently he is Head, Special
(Electronics) in 1977, both with product development Section involved in Image Processing,
distinction, from BITS, Pilani. Special product generation, S/w and H/w development activities
and direct writing large format film recorder operations.
During 1977-79 he worked as
Scientist at Central Electronics
Engineering Research Institute (CEERI), * * *
Pilani where he was involved in design and development of
process control instruments for sugar industry which is patented R V G Anjaneyulu obtained his
in India. BTech degree from JNTU University,
Hyd (1996) and MTech from Osmania
Since 1979 he is working at National Remote Sensing Agency University, Hyd (in 2000). He worked
(Department of Sapce, Govt of India), Hyderabad. Initially he as a technical officer in ECIL, Hyderabad
was involved in development of several innovative technologies for about two years. He joined the
used for Remote Sensing applications such as Color Photowrite National Remote Sensing Agency in 1998
System, Satellite Image Processing System (SIPS), Additive and since then he is with NRSA. He has
Color Viewer, Dual Densitometer, Image Analyser etc. These worked in the areas of Digital Film
technologies were subsequently transferred to the Industry. Recorders maintenance, Image
Processing, Special product generation, S/w and H/w development
At NRSA he has held several key positions in past on such
activities at NRSA making significant contributions in these
as Project Director, ASDF; Deputy Project Director, IRS-1B;
areas. His research interests are in the areas of image processing
Operations Director, IRS-1B; Deputy Project Director, Cartosat-
namely merging methods improvement / development, Natural
1 etc. Presently he is Engineer ‘G’ and Group Director (Imaging
Systems and Special Products) wherein he is responsible for color composite generation for the IRS images etc.
development and generation of Value Added customized Products
from Remote Sensing data acquired from various Satellites. * * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 225-242
ANURAG KUMAR
Department of Electrical Communication Engineering,
Indian Institute of Science, Bangalore, 560 012, India.
email: anurag@ece.iisc.ernet.in
Fig 1 WiFiRe network configuration. The figure on the left shows a deployment with three sectors, and the figure to
the right shows a tall tower carrying several BSs, with sector antennas, and several STs in a sector, with lower
height directional antennas
multisector TDM (TDD-MSTDM) scheduling base station should either all be in transmit mode or all
of slots. Time is divided into frames, which in receive mode and the transmissions should satisfy
contain traffic slots. The set of slots in a some power relations. These ideas have been made
frame is partitioned into contiguous uplink and use in the spatial reuse model that we discuss.
downlink segments. During the downlink
Shetiya [5] considers the joint routing and
segment, in each slot, one or zero transmissions
scheduling problem in WiMax mesh networks. A
can take place in each sector; and similarly in
dynamic programming problem is formulated to
the uplink segment. Because of site and
maximize the throughput, and is found to be
installation dependent path loss patterns, and
computationally complex to solve. Heuristics are used
because of time varying traffic requirements,
to attain a near optimal solution by considering the
the schedule will need to be computed on-line.
routing and scheduling problems separately.
The objective of this work is to abstract out the
basic scheduling problem, to develop a mathematical B. Preview of Contributions
formulation for the problem, to provide some scheduling
algorithms, and to provide a capacity assessment of We begin by developing a model for antenna
the MAC architecture. coverage and spatial reuse in a single channel multi-
sector operation. It is seen that in multi-sector operation,
A. Related Literature depending on the path loss, receive sensitivity and the
antenna directivity, the number of simultaneous
Bhagwat et al [3] have discussed issues related transmissions can be 3 or 4. We then set up an
to using 802.11 family of wireless technologies for abstraction for the TDD, single channel multi-sector
long distance transmission in rural environment, such scheduling problem.We begin by developing capacity
as the quality of 802.11 PHY performance outdoors, bounds for fair rate allocation, sum rate and sum of log
range extension, spectral vs. cost efficiency. The rate.This analysis also shows how the sectors should
authors provide details of the 802.11-based mesh be angularly oriented.We then develop a scheduling
network deployed in the Digital Gangetic Plains Project methodology for voice and data traffic.
providing voice and data services to villages. Raman
and Chebrolu [4] discuss the issues in using CSMA/ C. Overview of the Paper
CA in networks including long distance links. CSMA/
CA is designed to resolve contention in the indoor Section 2 sets up the notations used through the
environment, but is inefficient in long distance point to rest of the report and also explains the voice and TCP
point links. The authors provide a new MAC for mesh traffic models used. Section 3 explains the model used
networks synTX/synRX, which in the context of our to characterize the interference in the network by
problem translates to saying that the antennas at the disallowing transmission in some regions and by limiting
VARGHESE & KUMAR : DESIGN OF A TDD MULTISECTOR TDM MAC 227
the total number of simultaneous transmissions. Section B. Directional Antennas and Intersector
4 provides bounds on the capacity of the system. The Interference
optimum antenna positioning can be obtained based on
these bounds. Section 5 gives the scheduling problem The radiation pattern of a typical antenna used in
in hand. A dynamic programming problem formulation the deployment is shown in Fig 2. Based on the
of the problem is given in Section 5C. In this section, antenna pattern, we can divide the region into an
we also propose a greedy heuristic scheduler for association region, a taboo region and a limited
uplink and downlink. Section 5G gives a scheduling interference region with respect to each BTS.
algorithm to improve the fairness among users.
The radial zone over which the directional gain of
the antenna is above –3dB is called the association
2. PROBLEM SETUP
region. In our analysis, we take the directional gain to
A. Some Typical System Parameters be constant over this region. Any ST which falls in this
region of a BTS antenna j is associated to the BTS j.
Typically, there will be 80 to 120 subscriber
terminals (STs) in a 15 to 20 km radius covered by a 6 The region on either side of the association region
sector system. Each station will be associated with a where the directional gain is between –3dB and –15dB
base transceiver station (BTS). The TDD-MSTDM is called the taboo region. Any ST in this region of
scheduling is done over a frame. A typical frame time BTS j causes significant interference to the transmissions
is 10ms with slot time of 32µs, giving rise to 312 slots occurring in Sector j. When a transmission is occurring
per frame. The frame is divided into downlink and in Sector j, no transmission is allowed in this region.
uplink segments in a ratio which is a design parameter. In the limited interference region the directional
During downlink transmissions, a significant amount gain of the BTS antenna is below –15dB. A single
of power from the transmitting BTS reaches other transmission in this region of BTS j may not cause
BTSs, the distance separating them being very small. sufficient interference to the transmission in Sector j.
So, when downlink transmissions are scheduled in any But a number of such transmissions may add up
one of the sectors, other BTSs cannot be in the causing the SINR of a transmission in Sector j to fall
receive mode. Therefore, downlink and uplink below the threshold required for error free
transmissions must alternate over the entire cell. It transmission.We take care of this by limiting the total
follows that the ratio of number of slots in uplink to number of simultaneous transmissions in the system
that in downlink must be the same in all sectors. This as explained in section 3.
ratio is kept constant. A beacon marks the beginning
of the frame and also carries the scheduling MAP. A As an example, for the antenna pattern shown in
total of 24 slots are needed for the beacon in every Fig 2, the association region is a 60o. sector centered
frame. at the 0o. mark, the taboo region is 30o. on either side
of this association region, and the limited interference
All links are at 11 Mbps. A slot is of 32µs. At 11 region covers the remaining 240o.
Mbps, this is 44 bytes. A VOIP packet is 40 bytes
long. Thus, assuming that the MAC overhead is 4 C. Notation and Terminology
bytes per packet, the transmission of a VOIP packet
can be done in a slot. Each transport block (TB) has a n he number of BTSs (e.g., 6); BTSs are indexed
96µs PHY overhead, i.e., 3 slots. Hence, the minimum clockwise; for Sector j the interference region
size of a transport block is 4 slots. A TB should fit into in the previous counter-clockwise sector will
an integral number of slots. An uplink TB is always for be denoted by j– and the next clockwise
one ST, but downlink TBs can be for multiple STs. sector will be indexed by j+.
There is a maximum size of TB (Tmax) which indicates n0 the number of sectors that can have
PHY limitations or may correspond to higher layer simultaneous transmissions; see section 3.
limitations.
m the number of STs (e.g., 40 or 120); the
Implications for Scheduling: Since each TB number of STs in Sector j will be denoted by
involves a 3 slot overhead, it is advantageous to use mj and the number of taboo STs in the previous
long TBs. However, this would result in starving some counterclockwise previous sector by mj– ;
STs while favoring others. Note also that, because of similarly we define mj+.
3 N the number of slots in a scheduling frame.
Tmax, there is a minimum overhead of ¾ ¾ slots per
Tmax Each slot is used either for downlink
slot. communication or for uplink communication.
228 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Fig 2 Radiation pattern of a typical BS antenna that could be used in the deployment.
The association region is a 60o. sector centered at the 0o. mark, the taboo region
is 30. on either side of this association region, and the limited interference region
covers the remaining 240o
For example, N = 312 slots, as per the of u is j. Evidently an activation vector should
numerical values provided earlier. satisfy |{j : uj > 0}| £ n0. Also, u must satisfy
A the association matrix; an m × n matrix, where the exclusion constraints given by I.
each row corresponds to an ST and each U: Maximal activation vector: If no more links
column corresponds to a BTS. The (i, j)th in an activation vector can be activated without
element of the matrix is a 1 if the ith ST is causing interference to some other
associated with BTS j. Otherwise it is 0. We transmissions scheduled in the same vector
will sometimes refer to each nonzero element then this activation vector is maximal.
of A as a link. u: Activation set: the set of all maximal activation
B(i) the BTS with which ST i, 1 £ i £ m is associated. vectors.
i.e., Ai,B(i) = 1. S: A schedule: A schedule is an N × (n + 1)
I the exclusion matrix; an m × n matrix, where matrix, with rows corresponding to slots and
each row corresponds to an ST and each columns (except the last column) corresponding
column corresponds to a BTS. The (i, j)th to sectors, where the (i, j)th entry corresponds
element of the matrix is 1 if ST i is taboo for to the link in the jth sector that is scheduled to
BTS j or i is associated with j. Otherwise it is transmit in the ith slot. If no ST in the jth
0. Note that the matrices A and I together sector can transmit in the ith slot of the frame
define the scheduling constraints. (because this will interfere with other scheduled
u: Activation vector: a 1 × n matrix, where the transmissions in the frame) the corresponding
ith element denotes which ST in Sector i is entry is 0. The last column indicates the number
transmitting. If we decide to transmit between of consecutive slots for which the activation
ST j and its BTS, say BTS i, the ith element vector is used.
VARGHESE & KUMAR : DESIGN OF A TDD MULTISECTOR TDM MAC 229
I(s): This is the set of links that can interfere with per frame in the uplink and downlink such that the
any of the links in s. desired probability of packet dropping is achieved [9,
Chapter 5].
D. Traffic Models, QoS Objectives
Assuming a mean call holding time of 3 minutes
(µ–1 = 3 minutes) and a call arrival rate of 3 calls per
A possible network architecture for a WiFiRe
hour (l = 3/60 per minute), which are typical values
deployment is shown in Fig 3. There are a number of
for a home telephone, r = l/µ = 0.15 erlangs. The
telephones and PCs connected to the WiFiRe BTS
frame time in WiFiRE is 10 ms. Given that the vocoder
through STs. Several BTSs are controlled by a single
emits one packet every 20 ms, CBR traffic requires
base station controller (BSC). The BSC is connected
one uplink slot per call in every 2 frames, and VBR
to the Internet and the PSTN through switches. All
traffic requires almost one uplink slot per call in every
telephony traffic is carried as VoIP over the WiFiRe
4 frames. So, with 2 slots reserved for voice calls per
access network. For this purpose, notice that there is a
ST per frame, the number of calls that can be supported
gateway between the PSTN and the WiFiRe network.
is Nv = 4 for CBR traffic, and Nv = 8 for VBR traffic.
For packet voice telephony, we assume that the voice
With r = 0.15 and Nv = 4, we can have 7 subscribers
coder emits a frame every 20 ms. Assuming a frame
with a probability of blocking as low as 0.02. With
time of 10 ms we need one voice packet to be
r = 0.15 and Nv = 8, we can have 24 subscribers at
transmitted in each direction (uplink and downlink),
probability of blocking 0.02. With 4 slots reserved per
every 2 frames. This also implies that if a voice packet
ST these numbers are 24 for CBR calls and 65 for
is transmitted in the frame following the one in which it
VBR traffic. Given the village economics we expect
arrives, then its delay is bounded by 20 ms. We
that just 2-4 slots per ST may be all that are required.
propose to admit only so many VoIP calls, so that the
probability of a voice packet not getting transmitted in One VOIP packet is 40 bytes. The MAC header
the slot after the one in which it arrives is small, say has been taken to be 4 bytes, so that transmission of a
0.02.We note that the slot utilization can be optimized lone voice packet can be completed in 4 slots (3 slots
by performing silence suppression before the periodically of PHY overhead + 1 slot voice packet). It is possible
arriving voice packets enter the system. This will give to send more voice packets in a single transport block
rise to an on-off packet arrival process for each VoIP without additional PHY overhead. Thus, for a single
call in each direction. The onoff process will be random call from an ST, we need 4 slots each in uplink and
(typically modeled by a Markov process). For calls downlink every 2 frames; for two calls from an ST,
between the BTS and ST i, it suffices to allocate Ci £ we need 5 slots each in the uplink and downlink every
mv,i (mv,i is the number of voice calls for ST (i) slots 2 frames and so on.
For TCP controlled data traffic (the predominant P be the power required, so that the receiver decodes
traffic over the Internet), we assume that this wireless the frame with a given probability of error, in the
access system is the bottleneck along the path. As a presence of interferers. The directional gain of the
first model, we assume that TCP packets are BTS antenna is –15dB in other sectors. Hence, the
backlogged in each direction (i.e., at the BTS and the interference power from a transmission in any other
STs) and the scheduling objective is to pack as many 3
sector would be 10–¾2 P. For decoding a frame with
TCP packets as possible into the schedule, after ensuring
less than a given probability of error, we need a SNR
that voice QoS is met. We will also consider the
of P0 at the receiver. If there are n0–1 simultaneous
problem of ensuring some form of fairness between
transmissions, the path loss factor being h, we need R
the TCP users.
such that:
3. SPATIAL REUSE MODEL
P0 ( ¾R ) –h
R
y rcv = ¾ ¾ ¾ ¾ ¾ ¾0 3¾ ¾ ¾ ¾ ¾ ³ P0
In [6], the authors prove that maximizing the
1 + (n0 – 1) 10–¾2 P0 (¾R ) –h
cardinality of independent sets used in a schedule R0
need not necessarily increase the throughput, since as –h
R
the cardinality of the set increases, the prevailing ( ¾R ) – 1
0
SINR drops, thereby resulting in an increase in the n0 £ 1 + ¾ ¾ ¾ ¾3 ¾ ¾ ¾ ¾ ¾ ¾
probability of error, decreasing the throughput. Hence 10 –¾2 P0 ( ¾RR ) –h
0
it is necessary to limit the cardinality of the independent
set used so as to satisfy the SINR requirements. i.e.,
there is a limit to the number of simultaneous To provide a margin for fading, we consider a
transmissions possible. reduced range R´ such that 10 log ( ¾R´ )–h ³ 2.3s
R
In this section we consider the problem of finding where s is the shadowing standard deviation. Thus
the maximum number of simultaneous transmissions 99% of the STs in a circle of radius R´ around the BTS
possible in different sectors in the uplink and the can have their transmit power set so that the average
downlink. We assume that there is no power control in power P is received at the BTS in the uplink.
the downlink. The BTS transmits to all the STs at the Notice that, to allow spatial reuse, the coverage of
same power. We can have static power control in the the system needs to be reduced to R´ < R0. There is
uplink. Each ST transmits to the BTS at a fixed power, thus a tradeoff between spatial reuse and coverage,
such that the average power received from different which is captured by the spatial capacity measure C =
STs at the BTS is the same. The STs near the BTS n0R´2, which has units slots × km2. We note that this
transmit at a lower power and the ones farther away measure has the same motivation as the bit metres per
transmit at a higher power. second measure introduced in [7].
A. Uplink The variation of the number of transmissions and
system capacity with coverage is as shown in Fig 4.
In the uplink, we assume that there is static power We can see that, for each h, that there is an optimal n0
control. All STs transmit at a power such that the and R´ such that the coverage is maximum.
average power received at the BTS is P times noise
The coverage for which the capacity is maximum
power. Let the maximum power that can be transmitted
by an ST be Pt times noise power. Let R0 be the can be obtained from dC
¾ = 0 where r´ = ¾R´ Thus
dr´ R0
distance such that when Pt is transmitted by an ST at
distance R0, the average power received at the BTS is we get the optimum value of r´ and h0 as
P0 times noise power, where P0 is the minimum SNR 1
required to decode a frame with a given probability of æ 2.3 s 1 + 10– ¾3 P0 ö¾
error. Also, let R be such that when Pt is transmitted r´ = ç10–¾1 0¾ ¾ ¾ ¾ 2¾ ¾ ÷h
è 1 + ¾h ø
from an ST at distance R, power received at the BTS 2
Fig 4 Variation of the number of simultaneous transmissions possible (n 0) and system capacity (C) with coverage
relative to a reference distance R0 for n = 2.3, 3, 4 and s = 0, 4, 8. Plots for s = 0, 4, 8 are shown left to right
We can examine various forms for the utility maximizing the sum rate gives high overall capacity,
function. The optimization can be done so as to but poor fairness. On the other hand, maximizing the
maximize the average rate allocated to each ST, with average rate to each ST gives good fairness, but low
the constraint that each ST gets the same average sum capacity. Maximizing Sm log ri gives a good trade
rate. This is called max-min fairness. Trying to optimize off between maximizing the system capacity and
the rates such that the rate to each ST is maximized providing fairness. It is interesting to note that in
will adversely affect the sum capacity of the system. maximum Sm log ri case, the sum capacity is higher
So, take U(r) = Smj=1 (rj ). This maximises the sum when fairness is lower and vice versa. For example, at
capacity of the system, giving preference to STs that q = 10, we can see that the sum rate is close to 4. The
are in a favourable position, causing less interference fairness index is also close to 1. So, we may choose
to other STs. This has an impact on the fairness of the this orientation to operate the network.
system. To improve the fairness, we can take the log Note that the above computation can be done off-
utility function, U(r) = Smi=1 log(r(i)); this is known to line once the ST locations are known. Then the sector
lead to what is called proportional fairness. orientations can be obtained from this analysis.
Evaluating the upper and lower bounds derived
above, we find that the bounds are close to each other. 5. SCHEDULING: PROBLEM FORMULA-
So, we only report the results from the lower bound. TION AND SCHEDULER DESIGN
Hence, we have computed maxrÎ RL(q) U(r) for various Based on the discussion up to this point, the
values of . For each we obtain a vector r of average scheduling problem we are faced with is the following.
rates. We evaluate the fairness of this vector by using First partition the frame of size N slots into a
contiguous part with ND downlink slots and an uplink
the the fairness index is given by . part with NU uplink slots, such that ND + NU = N – NB,
If the rates to different STs are equal, then fairness where NB is the number of slots required for the
periodic beacon. Typically we will have ND » NU. This
index is 1, and the index decreases if there is rate
is because data transfer traffic is highly asymmetric,
variability between the STs.
as users download a lot more than they upload. During
In Fig 5 we plot, as a function of q, the total rate downloads, long TCP packets (up to 1500 bytes) are
(left panel) over all the STs for each of the three utility received in the downlink and one 40 byte TCP ACK is
functions, and also the fairness index (right panel) (the sent in the uplink for every alternate TCP data packet
lower bounds are plotted here). It can be seen that received.
Fig 5 Variation of sum rate and fairness index with antenna orientation for different utility functions
VARGHESE & KUMAR : DESIGN OF A TDD MULTISECTOR TDM MAC 235
Now, when mv,i, 1 £ i £ m, VoIP calls are admitted slots. Then, a bound on the total number of slots
for ST i, we need to determine the number of slots Ci available for transmission would be
to be reserved in the uplink and downlink subframes
for ST i, such that the QoS targets are met for all the æ Tmax – 3 ö
voice calls. For doing this, evidently the set of vectors |U| ç¾ ¾ ¾ ¾ ÷ N
è Tmax ø
C = (C1, ...,Cm) that are feasible (i.e., can be scheduled)
needs to be known. For each A and I, there will be an if the frame size N is an integral multiple of Tmax. Note
optimal set of such vectors Copt(A, I), and for any that the maximal activation vector depends on the
practical scheduler, there will be an achievable set of antenna used, the distribution of STs, etc. If there are
admissible vectors C Î ¸ Copt(A, I). several activation vectors with with the maximum
Once the required vector of voice payload slots cardinality, we can achieve some fairness by cycling
has been scheduled, we need to schedule as many between these vectors every Tmax slots. Still, this
additional payload slots, so as to maximize the traffic throughput maximising approach may result in some
carrying capacity for TCP while ensuring some fairness STs getting starved.
between the flows.
B. The Optimal Scheduling Problem for
A. Obtaining the Activation Set Uplink
Consider a graph with the STs and BTSs as nodes As an illustration, we focus here on the uplink
and the communication links between the STs and scheduling problem. The uplink scheduling problem
BTSs as edges. An activation vector is a matching on avoids the slight complexity in the downlink problem,
this graph. [8] proposes randomized algorithms that that, in the downlink, we can combine transmissions
can be used for finding nearmaximal matchings, with from a BTS to multiple STs in the same TB.
complexity O(Number of nodes). But, the inherent An instance of a scheduling problem is defined by
graph in the problem we consider being bipartite in an association matrix A, an interference matrix I, and
nature, and the scheduling being centralized, the maximal a vector C Î Copt(A, I) of required uplink voice
matchings can be found without randomized algorithms. capacities for the STs. We formulate the uplink
The algorithm for enumerating a maximal activation scheduling problem as a constrained dynamic program
vector is as follows. over a finite horizon NU, i.e., over the indices k Î {1, 2,
..., NU}, where NU is the number of uplink slots in a
Algorithm 5.1 frame. The state of the system at the beginning of slot
k is denoted by (xk, qk), where
1) Choose a link to be included in the activation
vector. This might be based on criteria such as xk: a 1 × m vector with xki denoting the number of
(i) the link with the longest queue length, (ii) or consecutive slots for which ST i has been
the link that received the lowest average rate transmitting; clearly, x0 = (0, 0, ..., 0)
over previous window of frames. qk: a 1 × m vector, with qki being the number of
2) Eliminate all the links that can cause interference required voice slots yet to be scheduled; clearly
to transmission on the set of links chosen until q0 = C = [C1,C2, ..., Cm]
this point.
3) Choose the best link (according to the above At the beginning of slot k, k Î ¸ {0, 1, ..., NU–1} an
criterion) from the remaining set of links. activation vector uk Î ¸ u is selected. Then in the slot,
all links appearing in u are allowed to transmit, with
4) Repeat Steps 2 and 3 until there are no more
the voice queues being depleted first. The state evolves
links that can be activated, or the set contains n0
as follows.
links (the maximum that can be activated at a
time).
xk+1,i = f1(xki, uki)
Remarks 5.1: ì (xk,i + 1)I{uk,B(i)= i} xk,i + I{uk,B(i)=i} < Tmax
=í
Suppose that we consider only TCP traffic. If we î 0 xk,i + I{uk,B(i)=i} = Tmax
try to maximize the number of useful slots without any
regard to fairness, then the schedule will be to use the i.e., if ST i is scheduled in slot k, (uk,B(i) = 1) and the
vector U with maximum cardinality (U Î u) for all maximum TB length has not been reached, then we
236 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
C. On Solving the DP formulation 1) Modify the voice queue lengths to include the
overhead slots required. i.e., if an ST has a voice
In the dynamic programming formulation of the queue of 2 packets, add 3 slots of PHY overhead
scheduling problem, the state of the system can be to make the queue length 5.
VARGHESE & KUMAR : DESIGN OF A TDD MULTISECTOR TDM MAC 237
Fig 8 Variation of total rate and fairness index with averaging interval for different
values of a . The upper set of plots are of the total rate, and bottom set are for
the fairness index
240 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
3, 10o min d/l rate 172 165 140 4, 20o min u/l rate 25 9 0
max d/l rate 212 208 190 max u/l rate 157 168 160
sum d/l rate 15573 14078 12499 sum u/l rate 4833 3699 2771
packet drop u/l 0 0.0025 0.0304
considered. A larger a in the rate averaging algorithm Mbps; this is because the packing can become more
yields a smaller average throughput. efficient. For this same case, with one voice call, the
average minimum uplink data throughput is 17 kbps,
6. VOICE AND DATA CAPACITY: and the average total downlink data throughput is 3.57
SIMULATION RESULTS Mbps. Adding to this 1.41 Mbps for voice, we obtain a
total uplink utilisation of 5.18 Mbps over a nominal
The scheduling algorithm discussed in Section 5D bandwidth of 11 Mbps allocated to the uplink. Because
was implemented in a MATLAB simulation. The PHY of being smaller, the uplink frame is more inefficiently
rate is 11 Mbps. We consider a random distribution of packed.
80 STs in 6 sectors. The spatial reuse n0 of 3 or 4 has
been considered, and the taboo regions in each sector, If n0 = 3 and a taboo region of width q = 10o, the
on either side of the sector, are q = 10o, 20o, 30o. fraction of voice packets dropped is 0.29% when we
Simulation is done with all STs having the same number support 2 calls per ST and 2.29% when we support 3
of ongoing voice calls: 1, 2 or 3. One VoIP call calls per ST. With 3 voice calls per station, we can see
requires one slot every alternate frame. A voice packet that the packet drop is high, and the uplink capacities
that arrives in the system is scheduled within the next to some STs are 0. With n0 = 3, the width of the taboo
two frames. If the scheduling constraints do not allow region does not have an effect on the system capacity,
the voice packet to be transmitted within two frame since we are always able to schedule in 3 sectors.
times of arrival, the packet is dropped. In the simulation, With n0 = 4, the system capacity reduces as q
we have assumed synchronous arrival of voice packets, increases. With q = 30., we can usually schedule
i.e., if two voice calls are going on from an ST, transmissions in just 3 sectors in a slot, even though
packets for both calls arrive synchronously, in the the SINR constraints allows 4 transmissions in a slot.
same frame. The data traffic model is that all the STs
have packets to be transmitted throughout. 7. CONCLUSION
The results are shown in Table 4 and Table 5. We consider the problem of finding the amount of
Here, min d/l rate is the average of the minimum rate spatial reuse possible with a single channel multi sector
over STs in the downlink, averaged over 30 random WIFiRE system with 802.11 MAC. It has been found
deployments; max d/l rate is the average of the that there is an optimum value for the number of
maximum rate over STs in the downlink, and sum d/l simultaneous transmissions possible, so as to maximize
rate is the average of the sum of downlink rates to the the total system capacity. The number of simultaneous
STs. The same measures are also given for the uplink. transmissions is found to depend on the path loss
The packet drop u/l is the fraction of voice packets factor and the radiation pattern of the antenna. For the
dropped in the uplink, this being the bottleneck direction. antenna pattern considered, and for path loss factor
All the rates indicated are in terms of the MAC 2.3, as is applicable for rural environments, the number
payload. The PHY overhead has already been of simultaneous transmissions possible is found to be
accounted for in the calculations. 3. This can be improved by the user of antennas with
lesser back lobe radiation.
Each voice call requires a payload of 44 Bytes
every 20 ms, and hence 1.41 Mbps are utilised per Also, for a given deployment, we find bounds to
voice call, in the uplink and downlink, for 80 STs. With the system capacity, assuming full channel reuse.
a PHY rate of 11 Mbps, with n0 = 3 we have an Based on this we can find the optimum positioning of
aggregate nominal rate of 22 Mbps in the downlink the antenna, such that the log utility function of rates
and 11 Mbps in the uplink (assuming that 2/3 of the obtained by different STs is maximized. The bounds
frame time is allocated to the downlink). From the obtained here are weak since we do not consider the
table, it can be seen that with 80 STs in 6 sectors, and constraint imposed by the maximum number of
1 voice call, with a taboo region of 10o on either side of simultaneous transmissions.
each sector, and n0 = 3, each ST gets an average
minimum data throughput of 164 kbps, and the average A constrained dynamic programming problem is
total rate is 13.749 Mbps. Adding to this 1.41 Mbps, found to give the optimum schedule for the system.
we obtain about 15.16 Mbps, for a nominal downlink But, the problem is intractable due to the explosion of
bandwidth of 22 Mbps. The difference is because of state and action space. We employ a maximal weight
PHY overheads, and the inability to fill up all slots in a algorithm were the weights are the queue lengths of
frame. We notice that a second simultaneous call at the voice queue, such that each ST transmits in
each ST reduces the data throughput by less than 1 contiguous slots, so as to minimize the PHY overhead.
242 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
For scheduling data in the system, we follow the 3. Pravin Bhagwat, Bhaskar Raman & Dheeraj Sanghi,
maximal weight algorithm, where the weights are Turning 802.11 Inside-Out, ACM SIGCOMM
Computer Communication Review, 34(1):33–38, Jan
reciprocal of the average rate obtained by each ST in
2004.
previous slots. The average considered is an exponential
4. Bhaskar Raman & Kameswari Chebrolu, Revisiting
weighted average of rates. A simple round robin MAC Design for an 802.11-based Mesh Network,
scheduler have also been considered. Scheduling Third Workshop on Hot Topics in Networks, San
examples are given for the greedy heuristic scheduler. Diego, Nov 2004.
The different schedulers considered were 5. Harish V Shetiya, Efficient Routing and Scheduling
for IEEE 802.16 Mesh network, Master’s thesis,
implemented in MATLAB and the data throughput in Electrical Communication Enigineering, IISc, 2005.
each case is obtained as the average data throughput
6. Arash Behzad & Izhak Rubin, On the Performance of
over deployments, in terms of payload slots. Graph Based Scheduling Algorithms for Packet
Deployments with different number of sectors and Radio Networks, GLOBECOM, 2003.
STs, width of taboo region have been considered for 7. Piyush Gupta & P R Kumar, The Capacity of
different voice loads, and the data throughput has Wireless Networks, IEEE Transactions on
been obtained in each case. Information Theory, IT46(2), pp 388-404, March
2000.
REFERENCES 8. Devavrat Shah, Paolo Giaccone & Balaji Prabhakar,
An efficient randomized algorithm for input-queued
1. Wi-Fi Rural Extension (Wi-FiRE), Technical report, switch scheduling, IEEE Micro, 22(1), pp 19-25,
www.cewit.org.in/docms/cewit07/review- January-February 2002.
comments-draft.pdf. 9. Anurag Kumar, D Manjunath & Joy Kuri,
2. Krishna Paul, Anitha Varghese, Sridhar Iyer, Communication networking, an analytical
Bhaskar Ramamurthi & Anurag Kumar, WiFiRe: approach, Norgan, Kaufman Networking Services,
Rural area broadband access using the WiFi PHY Elsevier 2004.
and a multisector TDD MAC, IEEE Communica- 10. Dimtri P Bertsekas, Dynamic Programming and
tions Magazine, 2007. Optimal Control, vol 1, Athena Scientific, 2001.
Authors
Anurag Kumar received BTech Kumar, Manjunath, and Kuri (Morgan-Kaufman/Elsevier).
from IIT Kanpur and PhD from Cornell
University, both in electrical
engineering. He was with Bell * * *
Laboratories, Holmdel, New Jersey, for
over six years. He is now a professor of Anitha Varghese obtained BTech
the Electrical Communications degree from the College of Engineering,
Engineering Department at the Indian Thiruvananthapuram, Kerala, in 2004
Institute of Sciences, Bangalore. His area and ME from the Electrical
of research is Communication Communications Engineering
networking, and he has recently focused primarily on wireless Department, Indian Institute of Science,
networking. He is a fellow of the IEEE, the Indian National Bangalore. She is presently working in
Science Academy (INSA), and of the Indian National Academy the Vehicular Communications and
of Engineering (INAE). He is an associate editor of IEEE Information Management group in
Transactions on Networking and IEEE Communications General Motors , ISL, Bangalore.
Surveys and Tutorials. He is a co-author of the textbook
Communication Networking: An Analytical Approach by
* * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 243-248
VLSI Technology has advanced rapidly since late 90s. Low cost, high performance
Chips designed and fabricated made strong impact in systems development and resulted in
applications to diversified fields like computers, communications, entertainment electronics,
medicine, and rural necessities. Spreading of Internet globally, making this world as a ‘global
village” and rapid strides in mobile communications have a direct bearing on the progress
made in VLSI technology. Research work is being done to further develop the technology in
new dimensions like Silicon Photonics, Organic semi-conductors, Flexible displays etc. These
developments take the technology further close to rural applications and can result in
significant improvement in the quality of life in villages. Fibre optic connectivity, hosting of
portals with useful information to rural folk, e-governance, energy conservation, wireless
sensor networks, Telephone-TV can improve the standard of living in rural areas significantly.
VLSI technology directly or indirectly is playing a vital role in this direction.
This paper reviews various developments taking place in the technology, research
work being done in this area, new devices and systems being developed and their
applications. The application of the technology and systems developed for rural areas is also
discussed. The projects implemented and possible developments to improve quality of life in
rural areas is discussed.
With the rapid strides being made in silicon device technology, efficient, low cost
devices with high performance are being manufactured. The chip design methods, making
use of the software tools are producing hardware for various applications like computers,
communications, defence, household appliances, commercial toys etc. These developments
are also making inroads into rural areas and changing the technological applications scenario
there. Hosting of websites / portals with information useful to rural areas is proving to be a
boon, for which the VLSI technology is contributing significantly.
1. TECHNOLOGY DEVELOPMENT double its present value requiring more elaborate heat
sinks. Some of the extra power consumption will
T HE International Technology Road Map for semi-
conductors, which is published periodically by the
come from gate-to-substrate and source-to-drain
current leakage that will grow larger as channel lengths
semi-conductor Industry Association (SIA, San Jose scale down to a few tens of nanometers. Approximately
Califf) gave projections for 2003 Technology node to 9 billion transistors on a chip may sound like a dream,
90 nm and to 6 nm by 2019. Technology node refers but it is already being made a reality. The packing
to the set of processes needed to print the smallest density of cells in human brain is 107 cells / cm3.
feature. According to this road map, high performance
ICs will contain more than 8.8 billion transistors in an As the channel length (the distance between the
area of 280 mm2 by 2016. This is more than 25 times Source and Drain) decreases, the transistor dimensions
the packing density of chips built with 130 nm feature also continue to shrink. A short channel means faster
size. Typical feature size, which is also referred to as transistor switching, because the charge carriers have
line widths, will shrink to 11 nm by 2014. Power a shorter distance to travel. The voltage on the drain
dissipation on high performance micro-processors will begins to lower the energy barrier in the channel
reducing the threshold voltage and freeing the carriers
to flow even when there is no voltage on the gate. This
Paper No 124-C; Copyright © 2007 by the IETE. is what in essence called “Short Channel effect” [1].
243
244 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
To avoid short channel effect, transistor hafnium dioxide gate insulation is combined with a
performance is to be sacrificed to some extent, and strained – silicon substrate, the mobility is found to be
tolerate some increase in power consumption. Reducing 60% higher. A high dielectric material replacement to
the thickness of the depletion region under the gate by SiO2 in the MOSFET structure reduces gate leakage
increasing the doping in the channel, maintains gate while strained silicon increases the performance of a
control, but it also reduces carrier mobility. If the transistor by material innovation. If these two
thickness of Silicon dioxide gate insulation above the advantages could be combined together without
channel is reduced to give gate, better control over the inferring, it is advantageous.
channel, the thinner oxide lets more current leakage
between the gate and substrate, driving up power The use of high – k dielectrics plus metal gates
consumption. This is tackled by having complex doping plus strained silicon plus increasingly complex doping
profiles but the device engineers are running out of profiles will extend life of the planar CMOS transistor
steam. for atleast another decade. The researchers are looking
at double gate MOSFETS. In double-gate devices, the
The gate length of a MOSFET, which is an indicator gate is on both sides of the channel giving much tighter
to gauge, the size of CMOS transistor, was 50 nm in control of the transistor’s ON and OFF states. Fin
2002, and is projected to be 9 nm by 2015. To improve FET structure is becoming popular and the advantage
the performance, silicon is to be mixed with a semi- is, channel is undoped. So the channel length shrinks.
conductor like Germanium to produce strained Another advantage is Fin can be made extremely thin.
crystalline structure to allow electric charge carriers’ Power consumption is lower, because there is no
move faster. To reduce leakage current which pushes leakage path for charge carriers to flow along, between
up power consumption, gate oxide is to be made of Source and Drain, when the device is OFF. With
materials with more than eight times the dielectric narrow line width, and small source and drain junctions,
constant (k) of SiO2. the resistance in series with the transistor’s channel
will increase, increasing power consumption and
Some of the materials are: Nitrided Oxides, Al2O3,
degrading the performance. Engineers have to solve
CeO, SrTiO3, Ta2O5, ZrO2 MgAl2O4 .
these problems in future. When the dimensions of
For better control of transistors’s switching states, CMOS transistor can’t be scaled down further, new
the gate of the MOSFET can be of metal instead of devices like nanotubes, single electron transistors, super
Polysilicon. If more than one gate is employed, power conducting transistors and molecular transistors will
consumption will be reduced and there will be better have to emerge [2].
control over ON-OFF action of the transistor.
2. MODERN FABS
In a pure silicon crystal, the distance between a
silicon atom and its nearest neighbours is the same in The 32 bit / 64 bit Microprocessors are premier
all three directions. But in the strained silicon layer, the markers of technological developments. A postage
atomic separations in the wafer plane are different stamp size chip will be having a billion transistors.
from those in the perpendicular direction. This change Hundreds of such chips are manufactured on a silicon
in the crystal symmetry changes the energy band Wafer of a dinner plate size. In a single day, state-of-
structure in the conduction and valence bands. This the-art fab can make nearly 100 trillion transistors,
change is to reduce electron – hole collisions with roughly 250 times the number of stars in the Milky
phonons (vibrations of atoms in crystal), the so called Way Galaxy. The device engineering is to be
scattering that slows the carriers down. appreciated for this technology marvel of the 20th
Silicon-on-Insulator (SOI) Wafers are being widely century. The fabs have a material handling system
used in semi-conductor manufacturing for the past that uses software – controlled robots and monorails
several years. SOI Wafers have a layer of Silicon to transport Wafers, etching chambers, Wafer polishers,
dioxide insulator buried under the device layer, to photolithographic steppers etc. Manufacturing software
reduce junction capacitance and so speedup transistor tools are used that track Wafers to the right tools, at
switching. SOI gets rid of lot of junction area to the right time and process control systems to manage
reduce junction capacitance. To reduce leakage and the chip making recipes the mix of gases, chemicals,
power consumption, the present gate insulation of metals and semi-conductors that constitute chip. The
SiO2 is to be replaced with a material having a larger Automated Manufacturing Technology or AMT,
dielectric constant (k), ‘A materials’ dielectric constant monitors, and controls the hundreds of steps, Wafers
“k” value is a measure of the extent to which it must pass through, on their way to become Pentium,
concentrates electric field lines. It is found that when Itanium and core 2 Duo processors etc. The Wafer
K LAL KISHORE : TRENDS IN VLSI TECHNOLOGY 245
used for manufacturing chips is a mirror – polished densities at 40 GHzs of 10W / mm with 40 V drain
disk of 99.9999999 percent pure silicon, 300 nm bias.
diameter and less than 1 mm thick.
Future wireless challenges include signal isolation
The AMT suite of programmes have been improved and software defined radio (SDR) SDR presents many
tremendously over the last couple of decades. Software issues such as analog-to-digital (ADC) performance;
called ‘THE GRID’ developed by Intel’s Logic transmitter solutions and cost. Implementation of high
Technology Development Unit allows various machines - k (dielectric constant) gate stack materials in low
in the fab and number of applications that make up the standby power applications should be achievable. But
AMT suite to communicate with each other to fully implementation of these materials in high performance
automate fab operations. The GRID is basically a (HP) Logic and Low Operating applications is still
giant electronic bulletin board, where machines and considered a difficult challenge. Introduction of 450
Wafers inside the fab announce their respective states nm wafers in 2012 is still considered a difficult challenge
and locations. AMT programs are written in a variety facing numerous issues [4].
of programming languages including C, C++ etc and
the common language is called extensible Markup 3. SILICON PHOTONICS
Language or XML. Each program posts messages
encoded in XML on the grid for other programs to see With the emerging Nano Technology in chips, the
and, if necessary act upon. transit time of electrons between transistors in chips
will get reduced but, the speed between chips in a
Software has automated all routine tasks needed system is not that fast. This bottleneck is being tackled
to run a fab. Software being used in fabs has freed our by Luxtera Inc. researchers. They have sculpted a
engineers to investigate ways to work with silicon laser waveguide and modulator into a tightly coupled
through the 22-nm generation of chips. In the near silicon package that can be hooked up easily. They
future, Carbon nanotubes function as transistors and claim that their chips handle 10 gigabits per second
light – emitting semi-conductors for optical outer and even higher speeds are in the offing. Then
connects [3]. broadband can become cheap and a feature film can
The Semi-conductor Industry became a global be downloaded in minutes [5].
industry in the 1990s. Chip manufacturing units and
assembly facilities were established in different parts 4. PHOTONS INSIDE THE COMPUTER
of the world. So International Technology Road map
for semi-conductors (ITRS) is created in the late 90s. The movement of data in a computer is in the
The Semi-conductor Industry Association (SIA) congested core of the microprocessor, the bits ‘fly’ at
extended invitation to Europe, Japan, Korea and Taiwan a very high speed. But the copper that links one
to co-operate on ITRS. This was done at the world processor to another and one circuit board to the next
semi-conductor Council in April 1998. Full revisions of slows down the movement. In a Pentium 4 processor
ITRS were produced in 1999, 2001, 2003 and 2005. operating at 2-4 GHz, the data travels on a bus operating
ITRS updates were produced in even-numbered years, only at 400 MHzs. Many researchers say that soon,
2000, 2002, 2004 and 2006. many of the copper connections in computers will
yield to high-speed optical inter connects, in which
The boundary between Silicon based and III - V photons rather than electrons will pass signals from
semi-conductors continues to move to higher board to board or chip to chip or even from one part of
frequencies with time. Parameters such as noise figure, a chip to another. An electrical signal from the processor
output power, power-added efficiency; linearity and would modulate a miniature laser beam which would
ultimately cost will become more important. For CMOS, shine through the air or a wave guide to a photo
long term prediction of device RF and noise conductor, which would in turn pass the signal on to
performance becomes more uncertain with the the electronics. Though at the moment it is expensive
introduction of metal gate electrodes, high permittivity to communicate with light than with electric current, in
(high - k) gate dielectrics and new device structures the near future only optical technologies will be able to
such as fully depleted or double gated Silicon - on - keep up with the elements of even more powerful
Insulator (SOI). microprocessors. Compact optical I/O devices will be
used within the computer. At high frequencies, the
For millimeter wave applications, InP- based RF wire, series inductance becomes more important than
transistors have demonstrated very high frequencies its resistance as an impeding factor, which also puts a
and GaN transistors have demonstrated record power limit on the rate at which the trace can transmit pulses.
246 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
These parasitic factors depend heavily on the geometry back planes. Plastic is a better substrate material
of the wire, its length and cross sectional area (c.s.a). because it is light, bendable and rugged. Fortunately,
So a wires’ ultimate bit rate turns out to be proportional many organic semiconductors are like amorphous
to its cross section, but falls with the square of its silicon.
length. So a thinner and longer wire means - a lower
Micro contact printing, laser and inkjet printing
bit rate. Transition time limitations can be fought by
techniques are being used for organic semi-conducting
driving the line harder, but it is not a good solution. It
materials. Thermal printing technology of Dupont is
adds noise, increases power requirements and
also being adapted to organic electronics manufacture.
aggravates already serious thermal management
First a large plastic sheet is coated with an organic
problems. If the wires are made thicker, they occupy
compound sensitive to light from a laser used in the
more space. Photons don’t suffer from these limitations.
printing process. Then a layer of conductive polymer
The exact point which economics of moving data
that will form the circuit’s gate electrodes and inter
dictate optical inter connects is debatable. When
connects is made. The coated plastic sheet is then
Microprocessors reach speeds of 10 GHzs, their inability
processed to a second piece of plastic that will act as
to communicate quickly with computer memory or
the circuit substrate, and the two are run through an
with other chips in a multiprocessor system will start
apparatus that acts like an industrial laser printer.
to stifle their usefulness [6].
Wherever the laser strikes, the sensitive compound
5. ORGANIC SEMI-CONDUCTORS – evaporates pushing the conductive polymers onto the
FLEXIBLE DISPLAYS circuit substrate, where it strikes. OLE D displays
have light weight, superior image quality and consume
Organic electronic materials can help in the low power. Still quite a lot of fundamental research
development of ICs with plastics. These use semi- remains to be done, including work on organic
conducting and sometimes conducting materials that semiconductors, having greater charge carrier mobilities
are made of molecules containing carbon, mostly in and operating lifetimes. The mobility of charge carriers
combination with hydrogen and oxygen. Such devices helps define a transistor switching speed. But it is far
are slower than silicon devices, but much cheaper. lower in organic semiconductors than in the crystalline
The technology is developed to produce circuits with silicon used in ICs. Researchers could overcome the
hundreds of transistors printed on plastics, sensors, problems and limitations invocated with amorphous
memories and displays that can be bent like paper. silicon devices. The same is also expected to be true
The present R&D work is aiming at cheap, flexible, of organic transistor [7].
flat panel displays. The conventional LCD displays
are made of glass and are heavy. If the organic semi- 6. RURAL PERSPECTIVE
conductors technology is fully developed, it is capable
of creating arbitrarily large displays on light weight Fibre optic connectivity to rural areas can bring
flexible plastic substrates, the use of flat panel displays the impact of ‘e-revolution’ to villages. Websites can
could be revolutionized. Researchers are aiming at be created keeping in mind the needs of rural people in
displays printed onto rolls of plastic which could be the areas of Agriculture, Health, e-governance, primary
unfurled, processed, and cut up into devices of any education, Rural Energy etc.
size. Such devices will have more adaptability for rural
The farming community must be kept informed
applications and electronics and information technology
about the changes and new developments in the
can penetrate faster into rural areas. Two types of
agricultural sector. A website with a platform for
devices are under development:
farmers, co-operatives, form a machinery vendors,
1) Organic Field effect transistors (OFETs) fertilizer and chemical companies’ agronomists and
2) Organic Light Emitting Diodes (OLEDs) farm advisors will be useful to the rural people. Thus a
knowledge society in agriculture can be created.
Because polymers can be solution printed like ink,
researchers are looking to the printing industry for Medical facilities in rural areas are very poor.
technology. This research work requires a different Child mortality rate is high. A website which gives
set of lab skills from silicon or III - V compounds. To information on different stages in the life of a child
carryout the research works expertise in organic beginning with pregnancy to adolescence, nutrition
chemical synthesis novel manufacturing techniques food to be taken, vaccinations to be given, information
and device physics is needed. Bell labs have produced about public Health Centers, schemes related to rural
useful n-type organic semiconductors, complementary health and sanitation, will be useful in improving the
circuits, and plastic backed active-matrix organic display quality of life of rural people.
K LAL KISHORE : TRENDS IN VLSI TECHNOLOGY 247
A website on e-governance providing link to light, wind, relative humidity, and rainfall can be collected
Governments’ schemes and services for the people by a network of weather sensors, embedded in
can help villagers with easy access to the government. communication units for Environmental Monitoring. In
Princeton’s’ Zebra net Project, a dynamic (B9) sensor
A portal on providing information products on
network has been created by attaching special collars
children going to schools, girl child education, IT literacy
equipped with a low power GPS system to the necks
will also be very much useful. On Line discussions in
of Zebras to monitor their moves and their behaviour.
regional languages on current topics and education
More or less on similar lines, cattle health, and veterinary
will help the rural children in a big way.
medical operations can be carried out. Usage of water
Energy conservation is universally important. The for agricultural purposes, humidity level in the soil,
importance of this is to be explained to the rural people usage of pesticides, information about weather
in their own language to educate them, to bring conditions, agricultural product – all such applications
awareness, and to make them participate in energy for rural areas can be implemented through such
conservation. A portal to motivate villagers to take up networks, in which VLSI chips and embedded systems
clean energy production through bio-gas, usage of play vital role. Wireless – Networked, sensor –
solar energy for home and village lighting, pump water enhanced toys and other class room objects supervise
technologies can help rural people. This can help the learning process of children and allow un-obstructive
families in a remote village find ways to improve monitoring by the teacher [9].
access and use energy for livelihood improvement in
their place of living [8]. In rural areas where medical facilities are not
available, sensors can be used to capture vital signs
6.1. Wireless Sensor Networks from patients in real-time and relay data to hand held
computers carried by medical personnel and wearable
Wireless networks are large number of small sensor nodes can store patients’ data. Such an
sensing self-powered nodes which gather information infrastructure designed to support wireless medical
or detect special events and communicate in a wireless sensors, PDAs, PCs and other devices that may be
fashion. The processed data is handed over to a base used to monitor and treat patients in various medical
station. Sensing of data, processing of data and situations (B 13). Smart houses can also be established
communicating the data or information, the combination wherein wireless sensor and activator networks
of all is done in a timing device. This gives scope for integrated within buildings could allow distributed
many applications. Recent Advances in LOW POWER monitoring and control, improving living conditions and
VLSI, embedded systems, computing, communication reducing the energy consumption, by controlling
hardware and in general the convergence of computing temperature, air flow etc. Fire accidents and loss due
and communications are making this technology a to such mishaps can be minimized, where facilities are
reality. minimum in rural areas.
Earth Sensing satellites are used to predict weather,
Applications
cyclones and hurricanes. Loss of lives, property and
other calamities can be minimized. The degree and
Sensor networks can be applied to:
precision of weather prediction increases with VLSI
1. Agriculture chip, of efficient and power computing capabilities.
2. Environmental Monitoring
3. Micro-Surgery, Medicine 6.2. Broad Band Over Power Lines
4. Child education
The idea is simple and the consumer gets access
5. Surveillance to a new broadband vendor and has the house wired
6. Warfare. for data with no additional effort. For rural applications,
this technology is particularly suitable and economical.
VLSI Chips play an important role in the hardware The cost of infrastructure required will be very much
to be provided for the networks. The nodes send the reduced and high technology can be taken to rural
data to a base station. If habitat monitoring is to be areas quickly. Police departments, amateur radio
done in a forest area, as they are sensitive to human operators are opposing this idea saying that this will
presence, using sensor network provides a non-invasive broadcast into their own bands. Google, Intel, Cisco
approach. Similarly data pertaining to air, temperature, are some giants working in this area.
248 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Author
Lal Kishore obtained MTech, and Circuits and Electronic Circuit Analysis. Two more text books
PhD, degrees from Indian Institute of on I C Applications and Electronic Measurements are in print.
Science Bangalore. He has more than 30 He is a Fellow of IETE, Member IEEE, ISTE and ISHM. He has
years of teaching experience, and has conducted refresher course on VLSI at the UGC Academic Staff
more than 60 publications to his credit. College, Hyderabad 8 times. He has implemented number of
He is guiding number of Research Research Laboratory Development Projects of DST, MHRD
Scholars in the area of VLSI Design, and AICTE. He is closely associated with various activities of
Technology and Micro Electronics. He IETE, Hyderabad Centre. He held number of administrative
had won the First Bapu Seetharam posts in Jawaharlal Nehru Technological University, Hyderabad,
Memorial Award for Research work from including Principal of University Engineering College, and
Institution of Electronics and Telecommunication Engineers presently Registrar of the University.
(IETE), New Delhi in the year 1986. He had also received Best
Teacher Award from the Government of Andhra Pradesh in the
year 2004. He wrote 2 text books on Electronics Devices & * * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 249-256
Remote sensing is being increasingly used over the last few decades as a powerful tool
for monitoring, study and analysis of the surface of the earth as well as the atmosphere. In
this paper we shall consider temporally adaptive pattern recognition techniques for land-
cover classification in multitemporal and multispectral remote sensing images. The technique
comprises of pre-processing using global and classwise probability density function (PDF)
matching for temporally adapting the statistics before classification. We focus on the utility
of these techniques in generating improved partially unsupervised land-cover classifiers and
their comparative study.
the supervised and unsupervised approaches [1]. histogram specification) which performs a matching
of shapes of the histograms by means of the use of
With reference to multi-temporal analysis, partially
their cumulative histograms [4]. This method is applied
unsupervised classification schemes carry a lot of
to multispectral images band-by-band.
importance. The term “partially unsupervised” is used
to indicate that although no ground truth is available Let X1 and X2 denote two multispectral images
for the specific image to be classified, some training with N channels and captured at time instants and t1
data that exists for another image of the same and t2 respectively. One of these images is considered
geographical area (but captured at some different as the source and the other as the target in terms of
instant of time) and this is utilized for generating the their histograms. Each image can be represented as a
classifier. In this case, classifiers are first trained on collection of pixels each being represented by a vector
the existing training set, whose parameters are then with N components corresponding to the intensity
updated using an assumed distribution of the second levels in N spectral bands. Hence, for each image, we
image [1]. The popular technique used is updating the can get N marginal distributions p1 (Xb) and p2 (Xb),
parameters of the previously trained classifier using one for every spectral band (b = 1, 2, ... N). For a
the distribution of the test image by Expectation- single dimension or band, the problem can be
Maximization (EM) algorithm [2]. It shows promising represented as finding a monotone mapping function
improvement in classification accuracies after retraining T(Xb) [4,5] such that:
but faces the limitation that it suits mostly to image
data following Gaussian distributions. Another method T(Xb) = C2–1 [C1 (Xb)] (1)
to improve classification accuracies in partially
unsupervised classifiers is the use of Markov Random where C1 and C2 are the cumulative PDFs of the
Field based approach which also considers the source and target images, respectively. This is the
contextual information [3]. In general, these methods standard band-by-band histogram matching (or PDF
aim at adapting the parameters of an existing classifier matching) problem which can be solved using discrete
to the statistics of the image to be classified. The lookup tables. Note that the terms PDF matching,
concept of matching of the statistics is elaborated in histogram matching, PDF transfer are used
the subsequent section. interchangeably in this paper. The above procedure
can be applied N times independently for N bands.
3. MATCHING OF STATISTICS This results in bringing the individual distributions of
the bands as closer as possible to each other.
The main issue in case of multi-temporal images is
that the images (even of same geographical area) B. Classwise PDF matching
acquired at different times are characterized by
differences in statistics owing to the change in It is worth noting that till this point we are discussing
atmospheric or ground conditions or differences in about the histograms in one band of the entire image.
sensor calibration at the acquisition time [1]. These But the entire image comprises of some “M” different
issues become even more prominent in case of images land-cover classes. Depending on soil and vegetation
captured during different seasons. For this reason, the condition, these classes exhibit different spectral
classifiers developed for the image of the same properties, i.e., they respond with different reflectance
geographical area tend to exhibit unacceptable for different part of the spectrum. It is precisely this
classification accuracies if applied directly to the image property that is exploited in distinguishing one class
of interest. Hence is the need for matching of statistics from the other and hence in classification. Thus, the
among the training and test images as an additional problem of classification demands an analysis of both
pre-processing step. the spectral and the statistical properties of various
classes. This issue can be handled more effectively if
A. Histogram matching technique we consider the histogram in each band as comprising
of “M” individual histograms, each corresponding to a
One way to adapt the statistics is the use of image class, i.e., modeling the distribution in each band as a
processing techniques for matching of statistical mixture density. In practice, it is assumed in many
properties of the multi-temporal images at a pre- applications that every class follows a Gaussian
processing stage. The simplest technique that can be distribution. This assumption of model has been
employed to match the probability density function in observed to suit well to satellite image data. Moving
each spectral band of the two images is explained beyond this assumption, in the presented approach of
here. This method uses histogram matching (or partially unsupervised classifier, there is no need of
INAMDAR & CHAUDHURI : R EMOTE SENSING IMAGES 251
any assumption on the nature of the distributions at the The ground truth has been made available to us
PDF matching stage, though assumed so in the later for all the three images. But for a given experiment of
stages of Maximum-Likelihood classification. The partially unsupervised classification, we assume that
histogram matching method in single dimension will the training data are available only for one image, to be
match the shape of any source histogram to any other used as training image. The test data set corresponding
shape of the target. In multi-temporal images, the to the other two dates is just used for validation of
difference in the shapes of the source and the target classification results.
histograms totally depends on the temporal changes
The accompanying ground truth information was
that occur in the spatial distribution, vegetation,
used to derive training and the test sets. The area
atmospheric conditions, air and soil moisture content.
under consideration is characterized by five land-cover
One point that should be again noted here is that every
classes viz wheat field, urban area, water body, rice
class may react differently to the temporal changes.
field and forest. Please note that the dark background
Hence, it may be more advantageous if we can derive
in Fig 1 is not a part of the image to be considered for
the PDF matching transfer function in eqn (1) differently
classification purposes. The corpus of training and test
for each class rather than a global or single transfer
data is as shown in Table 1.
function for the entire image in a particular band.
In the framework of the presented classification
problem, the aforementioned discussion can be neatly
represented as two experiments, viz. the global TABLE 1: Training and test sets corresponding to
bandwise histogram matching (GB) and the classwise images acquired at T1, T2 and T3
bandwise PDF matching (CB) as preprocessing
Land-cover No. of samples
techniques. The subsequent section will provide an Class
elaborate discussion on the related issues and Training Test Set
implementation of these two methods. Set
at the classification stage. This assumption has been matching is performed individually on each pair of
found to be reasonable. The final classification is done source and target classes. Let the modified source
using maximum likelihood (ML) rule in which every image in this case be represented as X1CB-ML. It is
class is modeled by Gaussian parameters. expected that the CB method should bring the
distribution of each class closer to each other than the
C. Histogram matching GB method. However, the performance of ML
classifier is expected to be poor without retraining and
In the first set of experiments, the image at time the method is not expected to yield accurate results. In
T1 was considered as the source (and the training the next variation for CB method, instead of forming
image) and the test image, either T2 or T3, was the target classes using the direct ML classification
considered as the target. Similarly, in the next set of result, they were formed using the results of ML
experiments the image at time T3 was chosen as classification after GB method of pre-processing. The
source and the test image, either T2 or T1, was steps are as follows:
considered as the target. The aim was matching the
statistics of the source image to that of the target. 1) Choose source (training) and target (test)
Initially bandwise histogram matching was performed images.
globally (GB method) using the histogram of the entire 2) Form the source clusters by grouping together
image in a particular spectral band. Probability density the pixels labeled as belonging to the same
functions or the histograms were constructed from class in the training set.
band 1 and 2 of the source and targets and the 3) Perform global bandwise histogram matching
cumulative distribution functions were found out. Then, and get modified source X1GB .
for every pair of source and target image, two histogram 4) Train the ML classifier using training data on
matching problems were solved corresponding to each X1GB.
spectral band according to eqn (1). Let the modified
5) Perform classification on the test set from
source image be represented by X1GB.
target image.
As discussed in the previous section, the bandwise 6) Group all the pixels from the test set classified
histogram matching was also performed locally or in into same class and form “M” target clusters.
other words on each class separately, i.e. classwise
bandwise matching (CB method). The critical issue As before, we have “M” source and “M” target
here was defining source and target clusters or classes. classes with the correspondence known between them.
In case of the source image, which is also the training Bandwise histogram matching is performed individually
image, forming clusters is trivial. The pixels labeled as on each pair of source and target classes similar to the
belonging to the same class were grouped together to previous case. Let the modified source image in this
form a cluster. But the target clusters were needed to case be represented as X1CB-GB.
be formed from the test images which have no In order to evaluate the effectiveness of this
information available for them. This problem was solved bandwise matching technique in bringing close the
as follows: distributions of source and target clusters, we calculated
1) Choose source (training) and target (test) the Bhattacharya distances between the source and
images. the target clusters before and after the modification
2) Form the source clusters by grouping together using the presented technique. The Bhattacharya
the pixels labeled as belonging to the same distance between the distributions of any class w i from
class in the training set. the source and the target (say, p1 (X | w i) and p2( X | w i
)) is calculated as follows:
3) Train the ML classifier using training data
from X1 .
4) Perform classification on the test set from
Bi = – ln òx p1 (X | w i) p2 (X | w i) dX (2)
target image.
As already mentioned, it is reasonable to assume
5) Group all the pixels from the test set classified a Gaussian distribution for the classes at the
into same class and form “M” target clusters. classification stage. Hence, the distance calculation
Now, we have “M” source and “M” target classes can be simplified as
with the correspondence known between them. The
problem of bandwise histogram matching now becomes 1 æ Si,1 + Si,2 ö –1
that of “M” independent problems. Bandwise histogram Bi = ¾ (µi,1 – µi,2)t ç ¾ ¾ ¾ ¾ ÷ (µi,1 – µi,2) +
8 è 2 ø
INAMDAR & CHAUDHURI : REMOTE SENSING IMAGES 253
(a) November 1999 (T1) (b) January 2000 (T2) (c) March 2000 (T3)
Fig 1 Band 2 of the multitemporal image data set used for experiments
TABLE 2: Bhattacharya distances for various classifiers trained on November 1999 dataset (T1)
TABLE 3: Bhattacharya distances for various classifiers trained on March 2000 dataset (T3)
TABLE 4: Classification accuracies for classifiers trained on November 1999 dataset (T1)
TABLE 5: Classification accuracies for classifiers trained on March 2000 dataset (T3)
more accurate than the ML classifier used in CB-ML classwise PDF matching more closely matches the
method. Overall results indicate that the modified training statistics of the classes in the source and target images
images which are closer in statistics to the target than the global PDF matching. Hence it also turns out
images to be classified result in improved classifiers to be more effective for generating modified training
which are “adapted” to the test images from the images which help generate more accurate classifiers.
temporal series. In other words, the methods used are The advantages of this method include elimination of
“temporally adaptive”. the need of retraining, and the low computational
complexity. The main drawback of this method is that
The results of classification are shown in Table 4 it considers each spectral band separately, neglecting
and Table 5. The average accuracy of classification the correlation among them. Our current work focuses
and the classwise accuracies are presented. It is on it.
worth noting that the classwise transfers improve the
classification accuracy of each class significantly and ACKNOWLEDGEMENTS
hence also the average accuracy of classification.
Further, the use of this procedure on multitemporal The authors are thankful to ISRO, India and Dr B K
images yields results of classification which are at an Mohan, IIT Bombay for making available the image data
acceptable level of accuracy and hence can eliminate set with ground truth used for this experiment. Funding
support from ISRO Cell is gratefully acknowledged.
the need for retraining using a procedure like the EM
algorithm which can be computationally exhaustive.
REFERENCES
For sake of verification, the January 2000 dataset was
also used as training set and the above mentioned four
1. L R Bruzzone & D F Prieto, Unsupervised retraining
classifiers were tested on November 1999 and March of a maximum likelihood classifier for the analysis of
2000 datasets. Results obtained were similar with multitemporal remote-sensing images, IEEE Trans.
classifiers designed with classwise transfers Geosci. Remote Sensing, vol 39, no 2, pp 456-460,
outperforming those designed with global transfer. Feb 2001.
2. T K Moon, The Expectation-Maximization
5. CONCLUSIONS Algorithm, IEEE Signal Processing Magazine, Nov
1996.
In this work, the suitability of global and classwise 3. R Cossu, S Chaudhuri & L Bruzzone, A context-
PDF matching techniques has been described with sensitive Bayesian technique for the partially
their application to partially unsupervised classification. supervised classification of multitemporal images”,
In multitemporal framework, these techniques have IEEE Geosci. Remote Sensing Letters,vol 2, issue 3,
been used to generate modified training images which pp 352-356, July 2005
are closer in statistics to the images to be classified. In 4. R Gonzalez & R Woods, Digital image processing,
general, for every test image to be classified, a modified 2nd edition, Prentice Hall, 2002.
training image can be generated which matches the 5. F Pitie, A Kokaram & R Dahyot, N-Dimensional
statistics of the test image. Thus, the training of the probability density function transfer and its
classifier is adaptive to the test image. Experimental application to color transfer, Proc IEEE ICCV,
results confirm the effectiveness of the methods. The Beijing 2006.
256 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Authors
Subhasis Chaudhuri was born resolution’ both published by Springer, NY. He has also edited
in Bahutali, India. He received his BTech a book on ‘Super-resolution imaging’ published by Kluwer
degree in Electronics and Electrical Academic in 2001. His research interests include image processing,
Communication Engineering from the computer vision and multimedia.
Indian Institute of Technology,
Kharagpur in 1985. He received MS and
PhD degrees, both in Electrical * * *
Engineering, respectively, from the
University of Calgary, Canada and the Shilpa Inamdar received her
University of California, San Diego. He Bachelor of Engineering degree in
joined the IIT, Bombay in 1990 as an assistant professor and is Electronics and Telecommunications
currently serving as the professor and head of the Dept. from University of Pune in 2003 and
He has also served as a visiting professor at the University the Master of Technology degree in
of Erlangen-Nuremberg, Germany and the University of Paris Electrical Engineering from I IT Bombay
XI. He is a fellow of the Alexander von Humboldt Foundation, in 2007. She has served as a Research
Germany, the Indian National Academy of Engineering, and the Assistant for the Indo-Italian project on
Indian Academy of Sciences. He is the recipient of Dr Vikram “Advanced Techniques for Remote
Sarabhai Research Award for the year 2001 , and the Prof SVC Sensing Image Processing” from July
Aiya Memorial Award and the Swarnajayanti Fellowship both 2004 - June 2007. Her areas of interest include pattern recognition
in 2003. He received the S S Bhatnagar Prize in engineering and remote sensing image processing.
sciences for the year 2004.
AVIJIT KAR
Jadavpur University, Kolkata 700 032, India.
email: ivy_majumdar@yahoo.co.in; bnchatterji@gmail.com; dr.avijit.kar@gmail.com
Texture features are widely used for matching in content based image retrieval. Since
mid nineteen nineties we find a lot of R & D activities in this area. In this paper an attempt has
been made to review these works. Seventy one papers were reviewed and these were
classified into spatial domain, wavelet based and miscellaneous methods. These methods
were further sub-classified on the basis of mathematical techniques used or the algorithm
applied. A very brief description of these methods is given. Finally the paper discusses
about the need for comparison of the methods and on the future directions in this area.
Retrieval stage
Image Retrieved
Query Image Digitizer Feature Matching and
Extraction image
Indexing
2.1. Histogram based Method Level Aura Matrices (GLAM) to model texture images
and matched the images using the similarity measure
Here we use the image histograms. Qin and Gao Support Vector Machine (SVM) to achieve excellent
[3], Quin et al [4] gave a matching method for texture retrieval accuracy. Guo et al [12] used a metric for
images with the help of a descriptor based on a two texture image retrieval which is the signed distances
dimensional histogram ARPIH (Angular Radial of the image from the boundary obtained using the
Partitioning Intensity Histogram ). This is determined SVM learning algorithm to obtain better accuracy
from local invariant regions of the image and has the compared to that obtained using Euclidian distance.
property of distinctiveness and robustness to The retrieval was insensitive to sample distribution
deformations. Grauman and Darell [5] described an and same result was obtained using different but visually
efficient and accurately matching method using an similar queries.
approximation to Earth Mover’s Distance (EMD)
which does not require clustering descriptors. 2.2.2. Using Ordinal Measure
2.2. Co-occurrence Matrix based method Partio et al [13-15] used a combination of co-
Co-occurrence matrix is in use for texture image occurrence matrices and ordinal measures which they
analysis for more than three decades. CBIR using co- called Ordinal Co-occurrence Matrix. The matching /
occurrence also received a good attention. We can retrieval was achieved using city block distance
further sub classify these methods into the following functions giving better performance compared to that
two categories. using gray level co-occurrence matrices.
Daubechies Semantics
Wavelet
Fuzzy Set
Similarity Matching theoratic
Using Distance
Three Step
GGD K-L Distance Hierarchic
the modal structure (ii) latent semantic indexing to Tamura et al [27] proposed several features of
obtain good results with database fabric and wrapping texture images namely coarseness, directionality,
paper images. contrast etc which are used in the IBM QBIC System
of CBIR.
2.5. Active Appearance Model based
Method 3. WAVELET BASED METHODS
Cootes et al [20] described a statistical appearance A large number of works can be seen in image
model which is obtained by combining the models of matching for CBIR with texture features using
shape and texture variations. They obtained an iterative wavelets. Efforts were made to use different wavelets
matching algorithm in which perturbations of the model and the effectiveness of these for the retrieval
parameters are related with minimization of texture performance was found out. A brief description of
error to provide improved texture match. these methods will be discussed in the following
subsections.
2.6. Neural Network based Method
3.1 Daubechies Wavelet based Method
Long et al [21] described a neural network based
method which optimizes the invariant / perceptual Wang et al [28] used Daubechies wavelet and
mappings of texture images. Gabor features of texture determined the wavelet coefficients as features and
image are mapped to an invariant space in the first used a two level multiresolution matching for retrieval
three layers. Texture is mapped into the perceptual which they called image indexing and searching to
space in the last three layers. They obtained improved obtain good results.
retrieval performance.
3.2. Method based on Similarity Matching
using Distance
2.7. Markov Random Field based Method
Laine and Fan [29] used the conventional discrete
Gimel’farb and Jain [22] used learning scheme for wavelet transform to determine the features and used
a Markov random field image model with Gibbs distances as similarity measure for matching. Chang
probability distribution. Their distance measure is based and Kuo [30] used tree structured wavelet transform
on the gray level histogram collected in accordance and distance functions for texture analysis. Unser [31]
with the structure of multiple pairwise pixel interactions used wavelet frame for texture classification and
in the sub-images to be matched. The method can segmentation. June and Scharcanski [32] used discrete
tolerate texture rotation and scale to a certain extent. dyadic wavelet transform (DDWT) to represent the
images in multiple resolutions to determine the
2.8. Other Spatial Domain Methods orientation and gray level distributions. They determined
a multi-resolution distance measure to match similarity.
In addition to the above seven types we find a Fauzi and Lewis [33] used sub-images at various
few more spatial domain models which does not fall resolutions of wavelet transform and determined the
into any of the above. Ohanian and Dubes [23] used texture features in each sub-image. In their matching
four classes of texture features namely random field algorithm they compared the distance between the
modeling, fractal geometry, co-occurrence matrices features of the query image with the features in each
and Gabor filtering separately and found that the co- sub-image thereby obtaining good accuracy. Kokare
occurrence matrices perform the best amongst these. et al [34] used Daubechies orthogonal wavelet 8-tap
Liu and Picard [24] used a new Wold model for filter coefficients as feature and gave an excellent
texture images and demonstrated that this model comparison for retrieval performance using Euclidian,
perform better compared to (i) shift invariant principal Manhattan, Chebychev, Mahalanobis, Square Chord,
component analysis and (ii) multiresolution Squared Chi-Squared, Canberra, Bray-Curtis and
simultaneous autoregressive model. Carkacioglu and weighted mean variance distance functions. Shang et
Vural [25] gave a statistical analysis of structural al [35] also used DDWT and a novel technique to
information (SASI) descriptor for texture image describe the textures with their orientation and gray
retrieval and demonstrated its effectiveness. Xu and level distribution using global spatial relationship of
Liao [26] used clustering technique for texture image colour to achieve better performance in comparison to
retrieval and used this method for medical images. that of [32].
IVY MAJUMDAR et al : TEXTURE FEATURE MATCHING METHODS 261
3.3. Generalized Gaussian Distribution and 3.8. Complex Wavelet based Method
Kullback-Leibler Distance based
Method Rivaz and Kingsbury [46] used complex wavelet
transform for texture image retrieval. They used the
Tzagkarakis and Tsakalides [36] used statistical complex wavelet transform to obtain texture features
feature and similarity measurements for CBIR. They and using a similarity matrix they obtained accuracy
used generalized Gaussian distribution (GGD) for approximately same as that obtained using Gabor
modeling wavelet coefficient and matched the textures wavelets. Hatipoglu et al [47] used dual tree complex
using Kullback-Leibler (K-L) distance between alpha- wavelet transform (DT-CWT) for texture image
stable distributions. Do and Vetterli [37, 38] also used features and obtained retrieval performance better
the GGD for modeling marginal distribution of wavelet than that obtained using real discrete wavelet transform
coefficients and a closed form K-L distance between (DWT). Kokare et al [48-50] used (i) rotated complex
GGD’s for matching thereby achieving significant wavelet filter(RCWF), (ii) dual tree complex wavelet
improvement in retrieval rates.
transform (DT-CWT) and (iii) RCWF and DT-CWT
as three different feature sets and obtained improved
3.4. Steerable Pyramid based Method
retrieval performance compared to that obtained using
Patrice and Konik [39] derived the texture features real DWT and using Gabor wavelet. They could use
on a multiresolution paradigm called steerable pyramid the method for both rotated and non-rotated database.
in wavelet based decomposition. They used several The method is effective for small size rotated, medium
similarity criteria like weighted L2 norm between two size rotated and large size rotated image databases.
feature sets and symmetrized version of K-L divergence
for matching. Finally they introduced a relevance 3.9. M-Band Wavelet based Method
feedback for improvement in the retrieval performance.
Acharya and Kundu [51] used M-Band wavelet
3.5. Gabor Wavelet based Method transform to give an adaptive and unsupervised
segmentation method for texture images. Kokare et al
Wu et al [40,41] and Manjunath and Ma [42] used [52] used the M-Channel wavelets for texture image
Gaussian function modulated by complex sinusoid called retrieval. They could achieve good retrieval accuracy
Gabor functions as features and using normalized with less computational effort than that using Gabor
Euclidean distance obtained approximately 74% wavelets.
retrieval accuracy. Ma and Manjunath [43] used the
Gabor features in a hybrid neural network for learning 3.10 Cosine Modulated Wavelet based
the similarity as feature space clustering thereby Method
improving the retrieval performance.
Koilpillai and Vaidyanathan [53] gave the concept
3.6. Spline Wavelet based Method of cosine-modulated FIR filter banks which is a special
class of unitary filter banks. Kokare et al [54,55]
Qiao et al [44] used spline wavelets to determine used this as (i) cosine modulated wavelet based and
features like mean and standard deviation of the sub- (ii) cosine modulated wavelet packet based texture
band coefficients. Their database consisted of 1,792 features for CBIR. They obtained better performance
images. It was found that linear Battle-Lemarié spline both in terms of accuracy and retrieval time as
wavelet, cubic bi-directional orthogonal spline wavelets compared to that obtained using Gabor wavelets.
and quadratic spline dyadic wavelet give better
performance compared to cosine modulated and Gabor 4. MISCELLANEOUS METHODS
wavelets.
In this section we will review some of the methods
3.7. Polar Wavelet based Method which do not fall in the category mentioned in the
previous sections. These methods either use a new
Pun [45] used polar wavelet for retrieval of texture
images. The feature extraction is the polar transform concept like fuzzy sets, neuroscience, semantics etc
followed by an adaptive row shift invariant wavelet or use a combination of concepts described in the
packet transform. During similarity matching k-nearest previous section. These methods are given in the
neighbor search is used. following subsections.
262 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
4.1. Semantic based Method with human performance. They could extract the
perceptual features from the vocabulary and their
Li et al [56] and Wang et al [57] gave a semantic match process use grammatical rules to compare the
sensitive integrated matching method which they called patterns. They considered several examples in different
SIMPLIcity. They used integrated region matching applications to illustrate the method. Nilufar and Chen
(IRM) approach in which the features of one region of [66] gave a method with visual features colour and
the first image is compared with the features of several textures using natural language. Their retrieval is using
regions of the second image for matching and a a fuzzy description of these features. The use of
similarity measure is then determined. Their method is linguistic terms in description helps in efficient retrieval
robust for variation in intensity and sharpness, rotation, from large databases.
colour distortion, cropping and shifting.
4.5. Meta Search Engine based Method
4.2. Fuzzy Set Theoretic Method
Beigi et al [67] gave the concept of a content
Fuzzy set theoretic concept is very well established based meta-search engine which they called
now and has been used in many engineering applications. MetaSEEK, in which they used visual information to
Santini and Jain [58] gave a new model called Fuzzy find images. Their method has provision for interactions
Feature Contrast (FFC) by extending the Tversky’s with several on-line image search engines and for
[59] feature contrast model. They used this concept intelligent selection and ranking according to
as a similarity measure and used it for texture image performance.
matching for CBIR. Chen and Wang [60] gave fuzzy
4.6. Hypo graph based Method
set theoretic method namely unified feature matching
(UFM) for region based segmentation. They used this Chastel and Paulus [68] described a method in
concept in their experimental SIMPLIcity content based which a digital image is matched with the help of hypo
image retrieval system for image matching. Chen et al graph. They discussed the problem of segmentation,
[61] gave an efficient and robust measure of similarity edge-detection, and noise detection. In their texture
which they called Fuzzy Integrated Region Matching classification they considered number of possible
and used it for region-based image retrieval. Their configurations, determined one-to-one maps and
measure for similarity was between two families of equivalent classes up to which mapping is possible
fuzzy sets. They tested their method on COREL based on thresholds in the different grids.
database consisting of 20,000 images and could achieve
better retrieval accuracy, speed and robustness 4.7. Neuroscience based Method
compared to IRM. Huang and Dai [62,63] gave a two
Bileschi and Wolf [69] considered neuroscience
stage CBIR system for texture images. The first stage
inspired set of visual features and used it for shape
was a fuzzy matching process based on energy
based object detection, texture understanding and
distribution pattern string (EDP). In the second stage
classification. Their method has good potential for
the composite sub-band gradient (CSG) vectors of
texture image in CBIR.
images and query image are compared after they pass
through a filter to provide good matching performance. 4.8. Modified Zernike Moments
4.3. Three Step Hierarchic Method Sim et al [70] gave a method for texture image
retrieval which is rotation, translation and scale invariant.
Vujovic and Brzakovic [64] described a three step They converted the image in Fourier domain and the
hierarchic method to match random texture pattern in descriptor consisted of (i) the power spectrum for
a large image database. They could take into account translation invariance, (ii) normalized power spectrum
the misregistration of random pattern relative to the for scale invariance and (iii) modified Zernike moments
representation in the database. They could determine for rotation invariance. They tested the method on
the probabilities of (i) two images having same pattern several databases and found that it gives better accuracy
(ii) wrong match and (iii) performance in terms of compared to Gabor, Radon and wavelet based methods
processing time. and it requires very low computational effort.
Mojsilovic et al [65] used a new distance function In the previous sections brief descriptions of a
using texture and colour features which correlates large number of methods for texture image matching
IVY MAJUMDAR et al : TEXTURE FEATURE MATCHING METHODS 263
were presented. In this section we are going to discuss Brodatz texture images of size 512 ´ 512 were taken
the results of a few of these important methods. from University of Southern California (USC) [74]
database. Then each image was divided into 16
5.1. Gray Level Co-occurance matrix based nonoverlapping subimages of size 128 ´ 128 to produce
spatial method nonrotated database D1. Small rotated database D2
was created by rotating the 13 images from Brodatz
Qin and Yang [11] in their method used Brodatz
database at 0°, 30°, 60° and 120° and then partitioning
[71] and the VisTex [72] texture databases for the
each of them into 128 ´ 128 nonoverlapping images.
experimentation. They compared their method with
Medium rotated database D3 were created by the
Ma and Manjunath’s [42,43] and Guo et al’s [73]
same method as that of the D2 by taking 40 images
method. Brodatz texture database contains 112 texture
from VisTex database [72]. Large rotated database
classes each of which is a 512 ´ 512 image. These
D4 were created by the same method as that of the
112 texture image classes were grouped into 32 clusters.
D2, D3 by taking 109 textures from the Brodatz
Each image was divided into 49 subimages of 128 ´
texture photographic album [71] and seven textures
128 pixels each with overlapping. The first 33 subimages
from the USC database of size 512 ´ 512 [74]. Do and
were used as the training set and rest of them were
Vetterli [75] used steerable wavelet domain- hidden
used for retrieval. Guo et al’s algorithm performs
Markov model (SWD-HMM) and gave the retrieval
image retrieval only with learning. On the basis of
result. Kokare et al compared their method with
retrieving images of same class as that of the query
standard DWT, dual tree complex wavelet transform
image results of Ma and Manjunath’s paper [42] tells
(DT - CWT) and SWD- HMM. Kokare
that their learning algorithm has about 90% retrieval
et al’s method combination of DT-RCWF and DT-
performance where as without learning average
CWT is the best for both rotated and nonrotated
retrieval accuracy of GLAM method is about 95%
databases. In precisely characterizing texture features
and with learning it has got 100% accuracy. For a
based on proposed scheme are more expressive than
given query image few retrieved images may not
the DWT features. In comparison with a standard real
belong to the same class as the query image but
DWT-based approach Kokare et al’s method improves
belong to the same cluster. Still they are taken as
the retrieval performance from 83.17% to 93.75%
similar to the query image because all subimages of
(database D1), 82.21% to 90.86% (database D2),
the same cluster are taken to be similar. GLAM
72.81 to 76.09% (database D3) and 64.17 to 78.93%
algorithm show best performance in terms of
(database D4). For CWT the retrieval performance
percentage of retrieving similar patterns also. GLAM
are 89.42%, 88.46%, 73.82%, 76.83% for D1, D2, D3
shows the best performance on Vistex database also.
and D4 respectively and for SWD-HMM 86.41%,
With learning it shows 100% retrieval accuracy and
86.77% for D1 and D2 respectively. As the texture
without learning the accuracy is about 91% for these
features of proposed method are robust and
textures. Samples of query and retrieved images of
efficient it shows much better performance for large
Qin and Yang’s method are shown in Fig 3.
database. Kokare et al followed the method stated by
5.2. Complex Wavelet based method B S Manjunath and W Y Ma [42] and evaluate the
performance in terms of the average rate of retrieving
Four different sets of experiment were carried out relevant images as a function of the number of top
by Kokare et al [50] using different database. 13 retrieval images. For top 116 retrieval of images DT-
Fig 3 Samples of query and retrieved images for GLAM method. (a) query image, (b) first of the forty similar
image retrieved, (c) first of the fifteen perceptually similar images of first category, (d) first of the ten
perceptually similar images of second category, (e) first of the eight perceptually similar images of
third category, (f) first of the eleven perceptually similar images of fourth category
264 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Fig 4 Samples of query and retrieved images for combination of DT-RCWF and DT-CWT method.
(a) query image, (b) first similar image retrieved, (c) and (d) two of the sixteen similar image
retrieved, (e) and (f) two of the four perceptually similar images retrieved
CWT + DT-RCWF method (96.82%) is always better shows almost the same efficacy as that of the CSG
than DWT (94.55%). Computational complexity is method. For texture discrimination gradient vector
approximately same for all the methods. In Fig 4 the generated from the four subimages LL, LH, HL, HH
query and retrieved image samples of Kokare et al’s by wavelet decomposition for an image, LL has got
method are given. the maximum information. Retrieval efficacy of
different combination of gradient vector was calculated
In the database there are 16 ground truth images
and it was observed that the combination of
for each class. DWT retrieved 20 perceptually similar
LL+LH+HL is the best in term of retrieval efficacy
images out of which only 8 were ground truth images
and computation efficiency. In their system EDP strings
of that class in the top. DT-CWT retrieved 14 ground
were used as filtering signatures to improve the speed
truth images in the top but other six images were
by keeping the retrieval efficacy same. The image
perceptually very different. On the other hand Kokare
retrieval speed may be increased by two to five times.
et al’s method retrieved16 ground truth images of that
As some of the qualified images may be filtered out
class and other 4 images were perceptually very similar
the retrieval efficacy may be reduced. However by
to that of the query image.
choosing proper fuzzy set we can have high speed as
5.3. Fuzzy Set Theoretic Method based well as high efficacy by using this method. Proposed
miscellaneous method method has better performance (92% retrieval
accuracy) than Do and Vetterli’s [38] method (78%
P W Huang and S K Dai [62] carried out their retrieval accuracy). But in terms of representation of
experiment by using 150 images of size 512 ´ 512 texture features Do and Vetterli’s method is more
taken from Brodatz album [71]. Each image was economic.Figure 5 shows the samples of query and
partitioned into 16 nonoverlapping 128 ´ 128 images. retrieved images of PW Huang and SK Dai’s method.
They compared their methods namely composite
subband gradient (CSG) method and hybrid method, 5.4. Modified Zernike Moments based
combination of CSG and energy distribution pattern miscellaneous method
string (EDP) with gradient vector method [76]. CSG
shows better retrieval efficacy (93.20%) than the Dong-Gyu Sim et al [70] carried out their
gradient method (87.75%). CSG method has more experiment using Brodatz album [71], Corel texture
powerful discriminating ability (88.51%) than the dataset CDs (size of each image is 768 ´ 512) and
gradient vector method (87.75%). Hybrid method some real texture of outdoor object taken from
Fig 5 Samples of query and retrieved images for combination of CSG and EDP string method. (a) query
image, (b)-(f) sample of retrieved images of five different categories with ranks 1, 2, 3, 4 and 5.
IVY MAJUMDAR et al : TEXTURE FEATURE MATCHING METHODS 265
Fig 6 Samples of query and retrieved images for modified Zernike moments based method. (a) Query image, (b) first
retrieved similar images, (c)-(f) perceptually similar retrieved images
Information and Communications University (ICU). classified into a number of subclasses depending on
Each of 109 Brodatz texture images (512 ´ 512) was the method used in spatial domain and the type of
partitioned into 16 nonoverlapping images (128 ´ 128). wavelet used in the wavelet domain. We find that
The rotated texture database was obtained using 30 there are few other methods which could not be
Brodatz textures and 25 ICU textures. The images classified into the above two classes which we called
were rotated with 16 arbitrary angles and textures miscellaneous methods. We have tried to give a very
rotated by 30° were taken as queries. Each image of brief idea to compare the effectiveness of some of the
Corel set was divided into four subimages of size methods. Kokare [77] has given a numerical comparison
384 ´ 286. Proposed algorithm is based on invariant of some of the wavelet based methods. A recent trend
features so its performance is better (88.6%) than is the combination of more than one method to improve
conventional method namely Gabor and wavelet based the accuracy. Although this will lead to increase in
technique (89.58%, 70.79% respectively) for translated retrieval time but with the development of VLSI
and rotated textures. Zernike moment based method technology, increase in the processor speed and due to
is scale invariant. So it shows better retrieval accuracy the development of efficient algorithms/software there
for scaled images (86.2%) than the Gabor, Radon and can be considerable improvement in the retrieval time.
wavelet based method (82.33%, 84.91%, 79.31%). It
If we look into the works on CBIR using texture
can show good result for even 50% scaled images.
images one interesting observation is that except for
Descriptor size of Zernike moments method is small
wavelets, the other image transformations were not
(22 byte) in comparison with Gabor (50) and Radon
tried or not reported even if some work was done in
(56) method. Computational complexity is also less for
some places. This may be because wavelet transform
the proposed method. For feature extraction it takes
became standard for JPEG, MPEG and several other
approximately 0.1 s on the Pentium-II 450 MHz.
image processing algorithms. Fourier transform and
which is less than the time taken by Gabor (1.7s) and
Fourier Mellin transform have excellent properties like
Radon (0.2s) but greater than wavelet based method
rotation, translation and scale invariance. Karhunen-
(0.05s). Figure 6 shows the samples of query and
Loeve transform is optimum in some sense. Similarly
retrieved images of Dong-Gyu Sim et al According to
the transform like Hadamard, Haar, Walsh, Slant etc.
the human perception the retrieved textures are very
have several interesting properties. In our opinion
similar to the query image.
these can be tried for CBIR using texture patterns and
the advantages / limitations can be found out. Also
6. CONCLUSION
there can be R & D work to determine (i) application
In this paper an attempt has been made to review specific methods, (ii) combination of more than one
the research and development work for matching method for texture matching and (iii) the computational
methods in content based image retrieval (CBIR) complexities of the methods.
systems using texture features. The two important
REFERENCES
steps in matching namely the feature extraction and
similarity determination were considered and a set of 1. Manesh Kokare, B N Chatterji & P K Biswas, A
seventy one available research works were reviewed. survey on current content based image retrieval
The performance of the system is usually specified in methods, IETE Journal of Research, vol 48, nos 3 &
terms of retrieval accuracy and retrieval time. In the 4, pp 261-271, May-Aug 2002.
opinion of the authors there are two broad classification 2. J Sklansky, Image segmentation and feature
namely (i) spatial domain methods and (ii) wavelet extraction, IEEE Trans on Systems, Man and
based methods. Each of these classes were further Cybernetics, vol 8, pp 237-247, 1978.
266 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
3. Lei Qin & Wen Gao, Image Matching Based on A Conference on Image Processing, vol 1, pp 517-520,
Local Invariant Descriptor, IEEE International 2005.
Conference on Image Processing, pp 766-769, 2005.
16. Noureddine Abbadeni, Content Representation &
4. Lei Qin, Wei Zeng, Wen Gao & Weiqiang Wang, Similarity Matching for Texture-based Image
Local Invariant Descriptor for Image Matching, IEEE Retrieval, ACM, Proceedings of MIR’03, Berkeley,
International Conference on Acoustic, Speech & pp 63-70, 2003.
Signal Processing (ICASSP 2005), pp II-1025 - II-
17. Noureddine Abbedeni, A New Similarity Matching
1028, 2005.
Measure Application to Texture Based Image
5. Kristen Grauman, Trevor Darrell, Efficient Image Retrieval, Proceedings of the 3rd International
Matching with Distributions of Local Invarient Workshop on Texture Analysis and Synthesis
Features, IEEE International Conference on (TEXTURE 2003), 2003.
Computer Vision and Pattern Recognition, San 18. Marco Carcassoni, Eraldo Ribeiro & Edwin R
Diego, CA, June 2005. Hancock, Texture recognition through Modal
6. Selim Aksoy & Robert M Haralick, Textural Features Analysis of Spectral Peak Patterns, IEEE,
for Image Database Retrieval, IEEE Workshop on International Conference on Pattern Recognition,
Content-Based Access of Image and Video pp 243-246, 2002.
Libraries, Santa Barbara, CA, pp 45-49, June 21,1998. 19. Marco Carcassoni, Eraldo Ribeiro, Edwin R
7. Selim Aksoy & Robert M Haralick, Content-based Hancock, Eigenvector Method for Texture
Image Database Retrieval Using Variances of Gray Recognition, IEEE, International Conference on
Level Spatial Dependencies, IARP International Image Processing, pp III-321-III-324, 2002.
Workshop on Multimedia Information Analysis and 20. Timothy F.Cootes, Gareth J Edwards & Christopher
Retrieval, Hong Kong, pp 3-19, August 13-14,1998. J Taylor, Active appearance models, IEEE
8. Selim Aksoy & Robert M Haralick, Using Texture in Transaction on PAMI, vol 23, no 6, pp 681-685, June
Image Similarity & Retrieval, International 2001.
Workshop on Texture Analysis in Mashin Vision, 21. Huizhong Long, Chee Wee Tan, Wee Kheng Leow,
Oulu, Finland, pp 111-117, 1999. Invariant & Perceptually Consistent Texture
Mapping for Content Based Image Retrieval, IEEE
9. Selim Aksoy & Robert M Haralick, Using Texture in
International Conference on Image Processing, vol
Image Similarity & Retrieval, M.Pietikainen, ed.,
2, pp 117-120, 2001.
Texture analysis in Machine Perception and
Artificial Intelligence, World Scientific, pp 129-149, 22. Georgy L Gimel’farb & Anil K Jain, On Retrieving
2000. Textured Images from an Image Database, Pattern
recognition, vol 29, no 9, pp 1461-1483, 1996.
10. Svetlana Lazebnik, Cordelia Schmid & Jean Ponce,
Affine Invariant Local Descriptors & Neighborhood 23. P.P Ohanian & R C Dubes, Performance evaluation
Statistics for Texture Recognition, 9th IEEE for four classes of textural features, Pattern
International Conference on Computer Vision Recognition, 25(8), 1992.
(ICCV’03), 2003. 24. F Liu & R W Picard, Periodicity, directionality &
11. Xuejie Qin & Yee-Hong Yang, Similarity Measure & randomness: Wold features for image modeling &
Learning with Gray Level Aura Matrices (GLAM) for retrieval, IEEE Trans on Pattern Analysis and
Texture Image Retrieval, IEEE Computer Society Machine Intelligence, vol 18, pp 722-733, 1996.
Conference on Computer Vision and Pattern 25. A Carkacioglu & F Y Vural SASI: a generic texture
Recognition (CVPR’04), 2004. descriptor for image retrieval, Pattern Recognition,
12. Guodong Guo, Hong-Jiang Zhang & Stan Z Li, vol 36, pp 2615-2633, 2003.
Distance-from-boundary as a Metric for Texture 26. Hui Xu & Mengyang Liao, Cluster-Based Texture
Image Retrieval, IEEE International Conference on matching for Image Retrieval, IEEE International
Acoustics, Speech and Signal Processing Conference on Image Processing, pp 766-769, 1998.
(ICASSP’01), vol 3, 2001.
27. H Tamura, S Mori & T Yamawaki, Textural features
13. Mari Partio, Bogdan Cramariuc & Moncef Gobbouj,
corresponding to visual perception, IEEE Trans on
Texture Similarity Evaluation using Ordinal Co-
System, Man, Cybernetics, 8(16), pp 460-473, 1978.
occurrence, IEEE International Conference on
Image Processing (ICIP), vol 3, pp 1537- 1540, 2004. 28. James Ze Wang, Gio Wiederhold, Oscar Firschein &
Sha Xin Wei, Content-based image indexing &
14. Mari Partio, Bogdan Cramariuc & Moncef Gobbouj,
searching using Daubechies Wavelets, Inter-
Texture Retrieval using Ordinal Co-occurrence
national Journal on Digital Libraries, pp 311-328,
Features, Proceedings of the 6th Nordic Signal
1997.
Processing Symposium - NORSIG 2004, June 9-11,
2004. 29. A Laine & J Fan, Texture classification by wavelet
15. Mari Partio Bogdan Cramariuc & Moncef Gobbouj, packet signature, IEEE Trans on Pattern Analysis
Block Based Ordinal Co-occurrence Matrices for and Machine Intelligence, vol 15, no 11, pp 1186-
Texture Similarity Evaluation, IEEE International 1191, Nov 1993.
IVY MAJUMDAR et al : TEXTURE FEATURE MATCHING METHODS 267
30. T Chang & C C Kuo, Texture analysis & 43. W Y Ma & B S Manjunath, Texture Features &
classification with tree structured wavelet transform, Learning Similarity, IEEE International Conference
IEEE Trans on Image Processing, vol 2, no 4, pp on Computer Vision and Pattern Recognition, pp
429-441, 1993. 425-430, 1996.
31. M Unser, Texture Classification & Segmentation 44. Yu-Long Qiao, Zhe-Ming Lu, Jeng - Shyang Pan, &
using Wavelet Frames, IEEE Trans Image Sheng - He Sun, Spline Wavelets Based Texture
Processing, vol 4, no 11, pp 1549-1560, 1995. Features For Image Retrieval, International Journal
of Innovative Computing, Information and Control,
32. Junc C R, Scharcanski J, Stochastic Texture Image vol 2, no 3, pp 653-658, 2006.
Retrieval & Similarity Matching, IEEE, xiii brazilian
symposium on computer graphics and image 45. Chi Man Pun, Rotation-invariant texture feature for
processing, pp 178-184, 2000. image retrieval, Computer Vision and Image
Understanding, vol 89, pp 24 - 43, 2003.
33. Mohammad F A Fauzi & Paul H Lewis, Texture-
based Image Retrieval Using Multiscale Sub-image 46. P D Rivaz & N Kingsbury, Complex wavelet features
Matching, IS and T/SPIE’s 15th Annual Symposium for fast texture image retrieval, Proceedings of IEEE
Electronic Imaging : Image and Video International Conference on Image Processing
Communications and Processing, Santa Clara, (ICIP 1999), vol 1, pp 109-113, 1999.
California, pp 407-416, 2003. 47. S Hatipoglu, S K Mitra & N Kingsbury, Image texture
34. Manesh Kokare, B N Chatterji & P K Biswas, description using complex wavelet transform,
Comparision of Similarity Metrics for Texture Image Proceedings of IEEE International Conference on
Retrieval, IEEE, conference on convergent Image Processing (ICIP 2000), vol II, pp 530-533,
technologies for Asia pacific region, TENCON 03, 2000.
pp 571-575, 2003. 48. Manesh Kokare, P K Biswas & B N Chatterji, Rotated
Complex Wavelet based Texture Features for
35. Zhao -Wei Shang, Gui-Zhong Liu & Ya- Tong Zhou,
Content Based Image Retrieval, Proceedings of
Texture Image Retrieval & Similarity Matching,
IEEE International Conference on Pattern
Proceedings of the third International Conference
Recognition (ICPR 2004), 23rd to 26th Aug, 2004,
on Machine Learning and Cybernetics, Shanghai,
Cambridge (UK).
pp 4081- 4084, August 2004.
49. Manesh Kokare, P K Biswas & B N Chatterji, Texture
36. George Tzagkarakis & Panagiotis Tsakalides, A
image retrieval using new rotated complex wavelet
Statistical Approach To Texture Image Retrieval
filters, IEEE Trans on Systems, Man and
via Alpha-stable Modeling of Wavelet
Cybernetics- part B, vol 35, no 6, pp 1168-1178,
Decompositions, WIAMIS Lisbon, Portugal, April
2005.
21-23, 2004.
50. Manesh Kokare, P K Biswas & B N Chatterji,
37. Minh N Do & Martin Vetterli, Texture Similarity Rotation invariant texture image retrieval using
Measurement using Kullback-Leibler Distance on rotated complex wavelet filters, IEEE Trans on
Wavelet Subbands, IEEE International Conference Systems, Man and Cybernetics- part B, vol 36, no 6,
on Image Processing, vol 3, pp 730-733, 2000. pp 1273-1282, 2006.
38. Minh N Do & Martin Vetterli, Wavelet-Based Texture 51. M Acharyya & M K Kundu, An adaptive approach
Retrieval Using Generalized Gaussian Density & to unsupervised texture segmentation using M-Band
KullBack-Leiblar Distance, IEEE Transaction on wavelet transform, Signal Processing, vol 81, pp
Image Processing, vol 11 no 2, pp 146-158, February 1337-1356, 2001.
2002.
52. Manesh Kokare & B N Chatterji & P K Biswas, M-
39. Blancho Patrice, Hubert Konik, Texture Similarity Band Wavelet based Texture Features for Content
Queries & Relevance Feedback for Image Retrieval, Based Image Retrieval, Proceedings of Indian
IEEE International Conference on Pattern Conference on Vision, Graphics and Image
Recognition, vol 4, pp 55-58, 2000. Processing (ICVGIP 2002), pp 173-178, Dec 2002,
40. P Wu, B S Manjunath, S D Newsam, & H D Shin, A Ahemadabad (INDIA).
Texture Descriptor for Image Retrieval & Browsing, 53. R D Koilpillai & P P Vaidyanath, Cosine-modulated
IEEE International Conference on Computer FIR filter bank satisfying perfect reconstruction,
Vision and Pattern Recognition, pp 3-7, June 1994. IEEE Trans Signal Processing, pp 770-783, April
1992.
41. P Wu, B S Manjunath, S D Newsam, & H D Shin, A
Texture Descriptor for Browsing & Similarity 54. Manesh Kokare, B N Chatterji & P K Biswas, Cosine-
Retrieval, Journal of Signal Processing: Image modulated wavelet based texture features for
Communication, vol 16, Issue 1-2, pp 33-43, Sept content based image retrieval, Pattern Recognition
2000. Letter, vol 25, no 4, pp 391-398, March 2004.
42. B S Manjunath & W Y Ma, Texture Features for 55. Manesh Kokare, P K Biswas & B N Chatterji, Cosine-
Browsing & Retrieval of Image Data, IEEE modulated wavelet packet based texture features for
Transaction on PAMI, vol 18, no 8, pp 837-842, Aug content based image retrieval, IETE Journal of
1996. Research, vol 51, no 6, pp 477-483, Nov-Dec 2005.
268 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
56. Jia Li, James Z Wang & Gio Wiederhold, IRM: 66. Sharmin Nilufar & Liang Chen, H K Kwan, A Fuzzy
Integrated Region Matching for Image Retrieval, 8th Query Based Image Retrieval System, Proceedings
ACM international Conference on Multimedia of IEEE International Conference on
2000, Los Angeles, USA, Oct 30-Nov 3, pp 147-155, Communication, Circuits and Systems, vol 2, pp
2000. 1017-1021, 2005.
57. Jams Z Wang, Jia Li & Gio Wiederhold, SIMPLIcity: 67. Mandis Beigi, Ana B Benitez & Shih-Fu Chang,
Semantics-Sensitive Integrated Matching for Picture MetaSEEK: A Content-Based Meta-Search Engine
Libraries, IEEE Transaction on PAMI, vol 23, no 9, for Images, SPIE, January 1998.
pp 947-963, September 2001. 68. S Chastel & D Paulus, Texture Identification Using
58. Simone Santini & Ramesh Jain, Similarity Measure, Image Neighborhood Hypergraphs, SPIE, San Jose,
IEEE Transaction on PAMI, vol 21, no 9, pp 871- 2004.
883, September 1999. 69. Stanley Bileschi & Lior Wolf, AUnified System For
59. A Tversky, Features of similarity, Psychological Object Detection, Texture Recognition, & Context
Review, vol 84, no.4, pp 327-352, July 1977 Analysis Based on the Standard Model Feature Set,
Proceedings of British Vision Conference (BMVC),
60. Yixin Chen, James Z Wang, A Region Based Fuzzy 2006.
Feature Matching Approach to Content Based Image
Retrieval, IEEE Transaction on PAMI, vol 24, no 9, 70. Dong Gyu Sim, Hae Kwang Kim, Rae-Hong Park,
pp 1252-1267, September 2002. Invariant texture retrieval using modified Zernike
moments, Image and Vision Computing, vol 22, pp
61. Yixin Chen, James Z Wang, Jia Li, FIRM: Fuzzily 331-342, 2004.
Integrated Region Matching for Content-based
71. P Brodatz, Textures: A Photographic Album for
Image Retrieval, 9th ACM international Conference
Artists & Designers, New York: Dover, 1966.
on Multimedia 2001, Ottowa, Canada, Sept 30-Oct
5, pp 543-545, 2001. 72. MIT Vision & Modeling Group, Vision Texture
[Online] Available: http://vismod.www.media
62. P W Huang & S K Dai, Image retrieval by Texture
mit.edu
similarity, Pattern Recognition, vol 36, pp 665-679,
2003. 73. Guo, G, S Z Li, & K L Chan, Learning Similarity for
Texture Image Retrieval, ECCV, pp 178-190, 2000.
63. P W Huang & S K Dai, Design of a two-stage content
based image retrieval system using Texture 74. University of Southern California, Signal & Image
similarity, Information Processing and Processing Institute, Rotated Textures [Online]
Management, vol 40, pp 81-96, 2004. Available: http://sipi.usc.edu/services/database/
Database.html
64. N.Vujovic & D.Brzakovic, Evaluation of an
Algorithm for Finding a Match of a Distorted Texture 75. M N Do & M Vetterli, Rotation invariant texture
Pattern in a Large Image Database, ACM characterization & retrieval using steerable wavelet-
Transaction on Information Systems, vol 16, no 1, domain hidden Markov models, IEEE Trans
pp 31- 60, January 1998. Multimedia, vol 4, no 4, pp 517-527, Dec 2002.
65. Aleksandra Mojsilovic, Jelena Kovacevic, Jianying 76. S.R Fountain, T N Tan, Efficient rotation invariant
Hu, Robert J.Safranek, Kicha Ganapathy, Matching texture features for content-based image retrieval,
& Retrieval Based on the Vocabulary & Grammar of Pattern Recognition, vol 31, pp 1725-1732, 1998.
Color Patterns, IEEE Transaction on Image 77. Manish Kokare, Wavelet Based Features Texture
Processing, vol 9, no 1, pp 38-54, January 2000. Image Retrieval, PhD Thesis, IIT Kharagpur, 2004.
IVY MAJUMDAR et al : TEXTURE FEATURE MATCHING METHODS 269
Authors
Ivy Majumdar born on 12th May reliability and e-security modelling. Blended with his profession
1972, obtained BTech degree in the year he is a photography buff. Western classical music relaxes his
of 1997 in Optics and Optoelectronics nerves.
and MTech degree in the year of 1999 in
Radio Physics and Electronics both from
Calcutta University. She worked as JRF * * *
in the Optics and Optoelectronics
Department of Calcutta University in B N Chatterji born on 10 Nov
the year 1999.Since January 2000 she is 1942, obtained BTech (Hons) (1965)
working as faculty member at B P Poddar and PhD (1970) in Electronics and
Institute of Management and Technology in the Electronics and Electrical Communication Engineering
Communication Engineering Department where she became of IIT Kharagpur. He did Post Doctoral
Assistant Professor in the month of February 2006. She has work at University of Erlangen-
registered herself for PhD degree in Jadavpur University in the Nurenberg, Germany during 1972-73.
area of Content Based Image Retrieval. Her research interests Worked with Telerad Pvt Ltd Bombay
include Pattern Recognition, Image Processing. Image Retrieval, (1965), Central Electronics Research
Soft Computing, Optoelectronics and Communication Institute Pilani (1966) and IIT Kharagpur
Engineering. as faculty member during 1967-2005. He was Professor during
1980-2005. Head of the Department during 1987-1991, Dean
Academic Affairs during 1994-1997 and Member of Board of
* * * Governors of IIT Kharagpur during 1998-2000. He has published
about 150 journal papers. 200 conference papers and four
Avijit Kar did his MSc and PhD books. He was Chairman of four International Conferences and
from IIT Kharagpur in 1980 and 1984 ten National Conferences. He has co-ordinated 25 short term
respectively. He is a Professor of courses and was the chief investigator of 24 Sponsored Projects.
Computer Engg. in Jadavpur University, He is the Fellow/ Life Member/ Member of eight Professional
Kolkata. He has supervised several PhD Societies. He has received ten National Awards on the basis of
theses and is actively involved in many his Academic/ Research contributions. His areas of interests are
R&D activities and IT related consultancy Pattern Recognition, Image Processing, Signal Processing, Parallel
for the Govt and the private sector. He Processing and Control Systems.
mainly works in computer vision with
varied application areas, engg. systems
* * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 271-276
The strength of any nation depends on the availability of resources, its industrial
strength and infrastructure base, with specific emphasis on the communication
infrastructure. India, on one hand has abundant natural resources and excellent industrial &
infrastructure base, which is mostly confined to urban and civilized society. On the other
hand, it has also a vast majority of population residing in villages, which is deprived of the
basic amenities and other growth opportunities. In this backdrop, the goal of achieving the
government’s vision to make India a ‘developed’ nation by the year 2020 is not a simple task.
This vision of transformation to a ‘developed’ India can only be realized by empowering the
rural people with the Knowledge Powered PURA (Providing Urban amenities in Rural Areas).
This is only possible through the creation of a Knowledge Society. This paper discusses the
ways and means of leveraging the ‘Electronics and Communication Technology’ towards
achieving this goal.
271
272 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
In order to improve rural economy, agriculture has 1. Information on Market Prices, Cropping
to be run as an agri-business rather than for mere Pattern, Weather Forecast, Agriculture
subsistence. This needs re-educating the farmer on Marketing (getting better prices for produce),
crop cultural practices for high yield with minimum 2. e-Governance: Access to all Forms, Copies
expenditure. Hence, innovative improvements are the of Land Records, Applications, Certificates,
need of the hour. The rural economy can be boosted if Grievance Redressal etc.
villagers have access to the following Facilities/Services:
3. Collection of bills: Electricity, Telephone, and
Capacity Building House-Tax etc.
– Education 4. e-Commerce: Insurance, e-Banking etc.
– Health 5. Telemedicine, e-Learning.
– Land distribution, Title and other Grievance The farmer needs a complete end-to-end solution
Redressal etc. from pre-plantation consulting to post harvest storage
and sales. He needs timely information on various
Income Generation
inputs like seeds, fertilisers, pesticides, farm
– Agriculture machinery and financing opportunities. He also needs
– Entrepreneurship timely solutions to irrigation risks like production risk
– Outsourcing (crop disease), rainfall risk and market risk (price
risk). In addition, he also needs complete information
Enabling Services on market trends, harvest storage procedures and
– Finance facilities, and transportation means. ECT can be
effectively used in addressing all these aspects. For
– Markets (exchanges /trading)
example, risk of crop disease can be managed by use
– e-Governance of video conferencing to connect the farmer to an
– Water Management agricultural expert and obtaining the answers before it
– Energy is too late.
– Communications and Transportation e-Governance mechanism ensures speedy service
Since most of these services can be extended to the people under one roof for all their government
through ECT, it can be effectively utilized to leverage related service requirements. It is addressed with
rural income. At the basic level, Communication electronic connectivity between the various state and
Facilities reduce the need for physical movement of central Government Departments and their respective
people, which in turn results in the saving of cost, time, databases with real time updated data exchange for
effort and energy. ECT itself will not work wonders, fully committed and transparent administration. With
but it will open up people, especially the young, to new this, all forms and applications are made available for
ideas and new worlds. Computers connected through online submission for speedier processing. Access to
ECT will make people learn new skills, which could be land records can also be made available and copies
harnessed in a myriad of different areas. of required records can be obtained online. In addition,
any required certificates from Govt Departments can
For example, farmers could use ECT to get also be obtained online. Similarly, these services can
commodity prices faster, or get information on new also be extended for online bill payments in respect of
agricultural techniques. The youth would get details on electricity, telephone, and house-tax, insurance etc.
job opportunities across the state. The district Access to e-Commerce simplifies the banking
administration could get details of problems in near transactions including loan sanctions and repayments.
real-time. The eligible could search for matrimonial
matches across adjacent villages. The voters would With telemedicine, rural PHCs equipped with ECT
communicate their concerns to the politicians and facilities may obtain timely treatment and remedial
bureaucrats electronically. The village officials could advices for any advanced or abnormal deceases. A
share governance best practices faster among their moderately learned technician or a doctor can interact
counterparts elsewhere. with the specialist by forwarding the symptoms of the
decease and the condition of the patient to get
If access to the following facilities is made available appropriate medical advice for proper treatment. Expert
through ECT, preferably with local language interfaces, advice for moving the patient to appropriate high level
it will certainly improve the quality of life of the poor medical centers and suggestions on the vacancy position
rural communities: in various nearby hospitals may also be obtained
R SREEHARI RAO : B UILDING A STRONG NATION 273
interactively online. This will be more helpful for timely The literacy rate in rural India is only about 50 per cent
action under emergency conditions such as epidemic as against 85 per cent in urban areas. India has 192
outbreaks or natural disasters. million illiterate women (mainly concentrated in rural
areas) that represent nearly one-third of all illiterate
With e-Learning, normal schools can be turned
women in the world.
into hi-tech ones with online e-lessons, to cope up with
the prevailing low teacher/student ratio of the order of With regard to child education, even though the
1:58 at primary level in rural regions of India. It can Eighty-sixth Constitution Amendment Act makes free
also be used for imparting high-level computer education and compulsory education for children between the
to rural students. e-Learning ensures equal opportunity age group of 6 to 14 years as a fundamental right, their
to students as it presents the same content and standard desperate economic conditions force the poor rural
to all, irrespective of their physical location. children to toil often in subhuman conditions, depriving
them of their most basic rights as children, i.e., education
The brighter side of India and a joyful childhood. It is pitiable to note that India
has the largest number of child workers in the world.
India is the world’s largest democracy and the As per a case filed in Supreme Court last year, there
10th largest industrial power, with solid and consistent are 100 million working children in India, nearly half of
economic growth. It has the third largest scientific and India’s child population.
technical workforce with its engineers holding top
positions in computer corporations all over the world. Social impact of the ‘Digital Divide’
In agriculture, India is the highest producer of sugar,
groundnuts, tea & fruits and is the number two in The following scenario illustrates how rural youth
production of rice, wheat, vegetables and milk. with little or no ICT skills can be marginalized in the
job market:
India has also recorded tremendous growth in
communication infrastructure. It presently holds a mobile There are kids in village schools who do not
phone customer base of about 100 million and landline even receive their textbooks on time while
customer base of about 40 million. The explosion of their counterparts in urban schools have all
Internet has also had its impact over India, which is the necessary resources to send e-mails and
evident from the whooping demand for Broadband & surf the Internet for supplementary information
Dial-up Internet connections. The current Internet for school projects. Without any doubt, all
customer base of about 10 million is expected to touch these children will be queuing up for
20 million by the year 2010. All this advancement has employment in the near future and most
been possible with the right use of Electronics and probably the ones who lack familiarity and
Communication Technology. skills in ICT will be sidelined. Children
everywhere have the same levels of curiosity.
The Darker side of India They can learn at the same quick pace of
their city brethren. For these children, the
Like in many other countries, unfortunately in availability of computer and Internet services
India too, ECT has been concentrated in the urban made all the difference.
pockets, leading to a clear digital-division between the
urban and rural population. Since timely access to This type of marginalisation from the
news and information makes remarkable difference in mainstream and frustration of the vernacular
promoting trade, education, employment, health and educated youth due to unequal opportunities
wealth, this “digital divide” threatens to exacerbate often lead to violent youth insurrections and
wide gaps between the rich and the poor. civil wars. Thus, for the overall prosperity of
any nation, creation of equal opportunities is
With about 70 per cent of Indian population residing of prime importance.
in rural areas characterized by grinding poverty and
social injustice, it is not possible to build a strong nation These circumstances lend to large-scale
without extending the fruits of the advances in exploitation of rural poor even by moderately wise
technology, in particular the ‘Information & businessman. Rural illiteracy often makes them fall
Communication Technology’, to the rural masses. prey for the dominant landlords and other financiers in
Unfortunately, the rural populations mostly comprising villages. Since village administrations (Panchayats)
of uneducated peasants and artisans are not able to are normally in the hands of influential few, it is likely
make use of these technologies to their advantage. that the benefits of the government sponsored programs
274 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
and schemes aimed at economic development and play an intermediary role with regard to empowerment
social justice will not reach the innocent poor. They and capacity building. The governments also need to
remain powerless, locked in an oppressive system of promote access to education and health care through
economic exploitation, class division, caste prejudice distance learning and telemedicine, preferably with
and pervasive corruption. While the economy of modern local language interface. The rural ICT requires special
India grows, the rural people lack access to education, efforts to create appropriate models for those who
nutrition and health care, sanitation, land and other can neither afford the Internet access nor have the
assets that could otherwise enable them to escape language capacity to understand the content. One
from the trap of poverty. As a result the poor remain example of such model is the UNESCO’s ‘Kothmale
poor and the rich grow richer. Community Radio and Internet programme’ running in
central Sri Lanka.
It is said that if a society cannot help the many
who are poor, it cannot save the few who are rich.
Radio Web Browsing
This means, the nation as a whole cannot flourish
without involving the larger rural population. On the The Kothmale Community Radio station
other hand, the rural populations do not have the broadcasts a daily ‘Radio Browsing the Internet’
necessary awareness, skills and facilities of ICT to programmes, and in this programme, the
contribute to their own development. Schools with broadcasters, supported by resource personnel,
best infrastructure, teachers who love teaching, primary browse the Internet on-air together with their
health centres, silos for storage of products and markets listeners and discuss and contextualise information
for promoting cottage industries and business, in local language.
employment opportunities for artisans are some of the
elements of PURA. Therefore, we should plan to The radio programme thus contributes to raise
promote rural prosperity by focusing on the information awareness about the Internet in a participatory
and knowledge-led rural economy. manner - the listeners request the broadcasters
to surf the WEB on their behalf and the
THE SOLUTION programme transmits information in response to
their requests. This information is explained and
One sustainable solution to this situation is to contextualised with the help of the studio guests,
reduce poverty by bridging the digital divide through for example: a local doctor may explain data on a
the process of ‘e-Learning’ among rural masses, aimed health website.
at simultaneously imparting education and information.
A kiosk is a multi-point service delivery channel in
With the current PC density of only about 4.6 per
a village for delivering the government and other
1,000 people and a telephone density of about 32 per
services to villagers through ICT. The kiosk is equipped
1,000 people, India needs to do a lot (to increase
with equipment like a computer, a digital camera and a
penetration of PCs and communication lines) for any
photo printer. The kiosk operator provides different
e-Learning project to be successful. The high cost of
kinds of services to villagers like computer education,
ownership and higher tariff levels are preventing
computer lending, printing, copying, fax, telephone,
proliferation of this technology in rural areas. Though
digital photos, insurance and other Internet-based
wireless connectivity could be considered as a cost
services. In addition, the kiosk also offers electronic
effective method to bridge the “digital divide”, the fact
delivery of all government services to the people at
that the bids for licenses under different telecom circles
nominal charges with convenience of local availability.
failed to attract investment in rural circles clearly
indicates the lack of interest of the telecom providers Involvement of Self Help Groups (SHGs) may
to cater to low volume and low paying users. also be considered as the answer for the transparent
delivery of a host of government schemes, education
Under these circumstances, the state will have to
and credit extension. In such a case the ICT based
step in and either provide additional incentives to the
village knowledge centres (internet kiosks) will be
market or cater to the needs as a publicly funded
organized and managed by the SHGs. The SHGs may
initiative. The government may also involve corporate
in turn identify some rural women and impart them
sector to come forward to fund such initiatives by
special training to run the kiosks.
luring them with tax concessions.
Internet kiosks and community information centers THE PRESENT STATUS IN INDIA
are to be planned with the involvement of NGOs that
have a strong presence in Rural India and which can The Indian government, on its part, has taken
R SREEHARI RAO : BUILDING A STRONG NATION 275
significant steps towards dissemination of information Internet Village’ by use of Information and
through a number of e-Learning projects, not only for Communication Technology. Under this program short
rural students but also for the community at large. named ‘RajIV’, it is committed to provide good
governance and revitalize the rural economy for
The ‘social development projects’ provide informal
integrated and sustained growth, to empower common
IT training to the rural illiterate masses. They focus on
man in the rural parts of the State to access all
generating network of information centres, employment
government, local-body and e-commerce services in
and investment opportunities in rural areas. Examples
an integrated manner in order to create better means
are akshaya, aqua choupal, gramin gyan kendra
of livelihood and improve the quality of life. Currently
(village knowledge center), etc.
more than 150 services are provided under this scheme.
The ‘community information services projects’
are aimed at the people who have the minimum required Contributions of Defence Research and
knowledge. They have been designed to disseminate Development Organisation (DRDO) - in
information on rural farm-gate price realization, cut Electronics and Communications
transaction costs, supply of high quality farm inputs
etc. Examples are: e-Choupal and PlantersNet, etc. Based on technologies from DLRL (Defence
Electronics Research Laboratory), Hyderabad, Convoy
Finally the ‘school based curriculum projects’ impart Jammers for VIP protection are being developed.
computer education to rural students. Examples are Any explosive devices planted nearby the roads on
SchoolNet India, Uttaranchal’s Aarohi etc. which vehicles travel can be deactivated by these
In addition to government, a number of ICT projects Jammer Systems. Technologies are also available with
are being sponsored by corporate sector. The most regard to Low Power and High Power Cell phone
notable among them are e-Choupal, i-Shakti, and e- Jammers. These can be used to thwart the efforts of
sagu etc. Today, they lead a silent revolution that terrorists who operate their destruction methodology
empowers farmers with relevant information to make by using cell phones. Muting systems are developed
their lives better. for Safety of Convoys of VIPs against Remotely
Controlled Improvised Explosive Devices.
ITC’s e-Choupal web portal brings real-time
information on weather forecasts and customized Antennas Developed by the laboratory can be
knowledge on better farming practices to the farmers’ used for different civilian communication systems. A
doorstep to improve his crop management. ITC’s e- variant of Pulse Doppler Radar, developed can be
Choupal supply chain brings good quality farm inputs used for monitoring over speeding vehicles.
at competitive prices to increase his farm yields.
DEAL (Defence Electronics Applications
HLL’s i-Shakti provides information and services Laboratory), Dehra Dun, contributes to Development
to the farmers through a portal, which has contents of Satellite Communication Systems. In the last decade
pertaining to a variety of rural issues. It enables farmers by making use of Mobile S Band ‘Sat Com’ terminal,
to have solutions to pest problems and it also enables during floods in Orissa, the disaster management was
rural women by providing a sustainable micro enterprise successfully undertaken.
opportunity to improve their living standards through
health and hygiene awareness. MM (Millimeter wave) imaging technology
developed in DEAL will be useful for identification of
e-sagu has a three-tier system consisting of farmers any armaments, if any, concealed in the clothes of
as end users. The agricultural scientists with knowledge persons.
system prepare farm advices. Coordinators as
intermediaries forward crop status through digital Based on the technologies developed in CAIR
photographs and text to Scientists and communicate (Center of Artificial Intelligence and Robotics),
their advice to the farmers. Bangalore, Robotics realized can be used for handling
Nuclear Materials.
Indian Institute of Technology (IIT)-Bombay and
its partners sponsor the web site called aAqua.org to “Light Weight Battery Powered Radar (BFSR -
provide farmers with relevant demand-driven farming Battle field surveillance Radar) can be used for
knowledge. surveillance .of large industrial and other civil
installations to detect intruders up to the low distance
Govt of Andhra Pradesh has pioneered in of around 2 km. Even a crawling man can be detected
introducing e-Governance under the name ‘Rajeev by utilizing this system. Targets such as a crawling
276 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
man, walking man, group of walking men, and moving off-shore Radar Installations.
light vehicles can be properly distinguished by this
BFSR. CONCLUSION
‘Radar Data Processing Technology’ utilizing the
state-of-the-art processors (similar to power PCs) As we are crossing the information society and
and very efficient algorithm finds its applications in leading to knowledge society, irrespective of rural or
automatic tracking of Multiple Targets in Air Traffic urban area, distance will be shortened using ECT.
Control Scenario. Removal of poverty calls for improving the quality of
life in rural places to the extent of ‘Providing Urban
Technologies Developed by LRDE (Electronic amenities in Rural Areas’ (PURA). Once this
Research and Development Establishment), Bangalore, proposition comes true, it may prompt for reverse
resulted in “Terminal Based ATC Radars”. In this migration, i.e., instead of village population coming to
case, 3D Medium Surveillance Radar System can be urban area, the urban population may come back to
upgraded with specific Radar Data Processing for villages. Such instances are becoming more common
improved and efficient handling of larger Air Traffic. in the developed countries, mainly due to the increasing
The Radar Networking Station can be utilized to inter ‘work from home’ culture. It helps to reduce the
link various ATC Radars to extend the Air Surveillance urban congestion prevalent today in most of the metro
coverage to regulate Civil Air Traffic. Transfer of cities. Such an ideal situation will lead to more uniform
Radar Air Pictures through appropriate advanced distribution of people, resources, facilities and
Digital Data Link will enable melting shore based and opportunities.
Author
R Sreehari Rao took over as These systems were productionized and inducted into services.
Director,Defence Electronics Research In view of indigenous R&D efforts of DLRL, many production
Laboratory (DLRL), Hyderabad with agencies were established in the country to cater the needs of
effect from 01 Nov 2005. He obtained services. A strong indigenous technology base has been established
his BE (Electronics and Communication in this field. He has successfully planned, steered and executed a
Engineering) from College of Engineering, number of Airborne, Ship-borne and Ground based projects and
Kakinada in 1972, MTech (Microwave programmes. Because of indigenous efforts DLRL is in a strong
& Radar Engineering) from IIT, position to meet present as well as future requirements of all
Kharagpur in 1974, and PhD from IIT, three services avoiding perpetual dependence on foreign sources.
Madras in 1995. Dr Rao joined DLRL as He has been honoured with many commendation and awards
SSO-II in 1974 and served the organization in various levels. towards his praiseworthy contributions in various programmes.
He has authored 66 technical reports and published several
In the last three decades, Dr R Sreehari Rao has made articles in National and International Journals. He is a fellow of
significant contributions to the indigenous Design and IETE.
Development of Broad band Antennas, Microwave components
and systems for all three services i.e. Navy, Army and Air force.
* * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 277-286
V S RAO, FIETE
As we become increasingly networked, our world Most Important - The Internet & World Wide
will grow smaller and bigger simultaneously. Web
Communication is at the root of the progress of every
society. Society is often described as essentially people All these need either terrestrial or satellite channels
in communication. to serve the user. The present day scenario can be
best summarized by Fig 1. Here space communication
The conventional communication tools are: becomes an important tool irrespective of the large
¨ Telegraphy using copper wire line growth in terrestrial communication networks
particularly when the technology is to be taken to rural
¨ Telephony using copper wire line
India.
¨ Television
¨ News papers, books, etc. Space Communication is provided by satellites in
¨ Means of Transport various orbits. Satellites have been put in to all these
orbits and successful communication links have been
All these have undergone tremendous changes and
established. It will be a matter of large discussion if
the new technologies emerged are
one starts explaining the pros & cons of the orbits &
¨ Mobile Satellite Telephone - Personal their satellites. Satellite communication started with
Communication Systems fixed services and expanded to greatest potential
¨ INFOSAT (information satellite) applications in the mobile and broad cast services.
Satellites have innate advantage that make them an
¨ A major shift from analogue to digital domain
attractive alternative or complement to terrestrial
¨ An increase in the value of software as broadband circuits. Satellite communication provides
opposed to hardware content reliable means of providing information or monitoring
¨ Extension of optical technology towards the inaccessible areas even in the case of severe cyclones
local loop coupled with increasing use of optical and disasters and when the entire terrestrial network
switches and optical processing fails. The conventional services provided by satellites
¨ Use of wireless in access technology - New are: Telephony/TV Broadcasting/Data reception and
access & Modulation codes distribution/Direct Television broadcasting/Disaster
warning/Continuous weather monitoring/Spacecraft
¨ Ultra broad band services
Vehicle Tracking and Commanding/ Inter satellite links/
¨ Extensive use of spread spectrum and code Mail /Internet/Data mining, Position (GPS) and time
division multiple access techniques. determination / Moving motor vehicle tracking etc.
Services planned / in existence are:
Mobile Personal Communications
¨ Virtual Private Networks
¨ PCS - Personal Communication Services Most of the time technological advancements have
¨ Call Collect Services taken place due to defense requirements. However in
¨ Desk top Video conferencing the area of communications where more than the
military requirements, it is the business requirements
¨ Card Phone Services
which have given a big flip to the overall scenario of
¨ Tele-shopping communication and ushered us from fixed copper
¨ Video Telephone wire communication era to the era of mobile
¨ Tele-health communication - meaning an individual (stationary or
PAL & RAO : SATELLITE TECHNOLOGY UTILIZATION 279
Television Internet TV
Cable TV
Video on
AUDIO SYSTEM
Demand
DVD Space
Telephone Internet
Cellular
Cellular
Pager Internet
Telephony World
Fax Wide
Home Banking Web
Network
Printer Computing
Video Camera Terrestrial
Video
Desktop PC Conferencing
PDA
Credit Card Smart Card
on move) carrying small inexpensive hand held rather than marketing led. In spite of cellular telephones
communicator and being reached by voice, fax or data and other mobile services, the two systems have shown
with a single telephone number independent of location. their usefulness over others in Afghanistan operations
Basically it means to provide communication to and and also communications to inaccessible places.
from users located anywhere on the globe and
In the foreseeable future new equipments and
possessing a portable light weight hand held mobile
techniques may be used in mobile “satellite technology
telephone. This could be termed as Personal
to bring down the costs and make the system affordable.
Communication Services (PCS).
The day is not far off for providing the benefits of
Today, a hand held mobile phone has become a technological advancement being enjoyed by urban
personal hold and has become affordable by every people to the people in rural inaccessible India.
citizen across societies. This service is available only
in densely populated cities and towns. Even today, the Internet over satellite
cellular mobile service is a dream for people in remote
It will be quite important and relevant to talk about
areas. This can be fulfilled with mobile Satellite
latest arrival on the communication technology scene:
Communication though it is yet to take off commercially.
the internet and the web. Today, the internet provides
The advantage of space communications is that one
a vast array of services with high bandwidth links that
does not have to be at a particular place to derive its
can simultaneously carry telephone, video (Television)
benefits. The Communication through space is
and data and is accessible from anywhere through a
independent of place and time and geo political limit.
variety of information appliances ranging from personal
Though the Irridium and Global Star mobile satellite computers and hand-held digital assistants to screen
communication projects have failed due to the phones and televisions, not to mention the computers
astronomically high terminal & service cost as embedded in everything from automobiles to vending
compared to terrestrial based systems, on engineering machines. The Net is a phenomenon that cannot be
account the two projects cannot be considered as ignored. It is an agent of change in all sectors of
failures since the projects were engineering led - society. In turn, this will lead to truly profound
280 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
changes in society & the present technological communication network using the Ku-band capacity
paradigm. of INSAT system. The Ku-band capacity of INSAT
will be used for promoting satellite based communication,
To help all these & to spread the net at a faster
specifically, in the areas of distance education, tele-
pace even to inaccessible & remote places satellite
medicine, agricultural extension, e-governance, self-
communication plays a major role, besides the
help groups, marketing and HRD, community internet
conventional terrestrial links, optical links etc., which
centres, etc. While ISRO will provide the transponder
cater to cities and larger population bases owning to
capacity on board INSAT and the technical support of
the economics.
the state governments will operate the satellite
It is expected that the space based high speed communication network for the various objectives afore-
internet service will reflect triple digit growth rate. As mentioned.
of today internet is the greatest thing to happen in the
SRO has already helped the Orissa State
satellite industry. For many space segment operators
Government in the implementation of Vidya Vahini
INTERNET is the biggest earner. As of today internet
Yojana. Besides, it has established the Jhabua
is going towards zero tolerance for failure. The goal is
Developmental Communication project in Madhya
more & more critical services to NET. The availability
Pradesh that is in progress since 1996. ISRO has also
of internet services through space is a big boon to rural
started the INSAT training and developmental
folk to get all the information required for increasing
communication channel in 1995 that is being used by
their knowledge including agricultural productivity.
several private and Government agencies for industrial
training, training of panchayat workers, agricultural
Services Provided by ISRO
extension workers, etc. The MOU signed with Andhra
Pradesh marks yet another milestone in the
ISRO had mainly three types of satellites viz.,
implementation of a countrywide Gramsat network
Communication satellites mainly for communication
for grassroots level development.
operating in GEO, Remote sensing satellites deployed
in LEO and scientific satellites; for various applications
Video-conferencing facility
driving the growth and development of entire India.
Many applications are mainly focused towards
Provided by INSATs is a great achievement in
upliftment of rural areas by bringing education, health
bringing various segments of administration together.
etc. to the door steps of rural homes/community centres.
To give an example, the facility enabled then Prime
Minister, Mr Atal Behari Vajpayee, on August 01,
INSAT Utilization
2000 to have a virtual “tete-a-tete” with the chiefs of
Panchayat Raj Institutions from 14 video/studio link
Satellite Television broadcast was first introduced
centres spread across the states of Uttar Pradesh,
by ISRO in India 32 years back when the Satellite
Gujarat and Karnataka. In an hour-long conversation,
Instructional Television Experiment (SITE), targeted
the Prime Minister had a first hand assessment of the
for rural areas, was commenced on August 1, 1975.
nitty-gritty of the local problems being faced at the
The video conference involved the use of INSAT-2C
grassroots level in the country. The video conferencing
and INSAT-3B satellites. Over the years, ISRO has
centres had been established at Bakshi ka Talab,
also conducted interactive training and distance
Gaurabaug, Sarojini Nagar, Nishatganj and Cantonment
education programmes through the INSAT satellite
Park Road in Lucknow district (UP), Himmat Nagar,
system.
Mehsana, Nadiad, Palanpur and Gandhinagar in Gujarat
INSAT series satellites deployed in GEO orbit are and at Mangalore, Dharwar, Raichur and Tumkur in
mainly providing Karnataka. Lauding the role of science in the rapid
strides being made by the nation, the Prime Minister
• Telecommunication
congratulated the scientists and engineers for their
• Television endeavor to make the video conference a reality. Mr
• Search and Rescue Vajpayee said that this would bridge gaps and open up
• Meteorology new vistas of growth and communication.
• Radio Networking.
Video conference provided a platform for the
The Indian Space Research Organisation (ISRO) Prime Minister to give a patient hearing to the
signed Memorandum of Understanding with various Panchayat representatives and, at the same time,
state governments for establishing a satellite-based apprised them of the limitations and practical problems
PAL & RAO : SATELLITE TECHNOLOGY UTILIZATION 281
faced by the government at times and to discuss vast provide a common platform to all concerned agencies
array of issues viz., specific problems faced by like communication systems/software and medical
panchayat institutions like education, water shortage, equipment providers, super-specialty hospitals,
and irrigation. healthcare administrators, various departments of
Government, trust and private hospitals, NGOs and
Tele-medicine corporate hospitals dealing with healthcare both in
India and abroad. It will also help to exchange ideas,
The advances and convergence of IT and practices and methodology to effectively implement
telecommunication can bring the entire health care and use the emerging telemedicine technology
services to the patient’s doorstep. Tele-medicine is
delivery of health care information across distances In the field of tele-medicine, the first pilot project
using telecom technology. This includes transfer of was started in Andhra Pradesh in 2000 connecting
images like X-rays, CT, MRI, ECG, etc from patient Apollo Hospitals at Chennai and Aragonda village and
to expert doctors seamlessly, apart from the live video ISRO’s hospital at Satish Dhawan Space Centre,
conferencing between the patient at remote hospital SHAR, Sriharikota, in Nellore District. The tele-
with the specialists at the super speciality hospital for medicine network is extended to more hospitals.
tele-consultation and treatment.
DTH-TV Broadcasting
ISRO has been spearheading satellite based tele-
medicine programme in the country with remote district Common entertainment today is through Television.
hospitals connected to Super Specialty Hospitals in Television broadcasted by satellites enabled every
major cities using INSAT satellite (Fig 2). citizen in all parts of the country to have the benefits.
DTH - TV broadcasting comes under BSS
The advantage of tele-medicine in reaching out to (Broadcasting services) as well as under FSS (Fixed
the rural and remote population has been well Satellite Services). Under FSS services it is from point
established through the experience of this present to multipoint but whose location is known while under
telemedicine network. However, there is a need to BSS it is always in universal broadcast mode. With
digital TV in 36 MHz standard transponder, one can observation to disseminate a variety of services
transmit almost 12 TV channels using latest emanating from the space systems and other
compression, error correcting codes etc, where even Information Technology tools to address the changing
at low SNR better TV quality can be received. This and critical needs of rural communities. The VRC is a
will give rise to the possibility of reducing orbital totally interactive VSAT (Very Small Aperture
separations between DTH Satellites. This is only the Terminal) based network.
means to provide entertainment communication to rural
Users located at one node of this network can
inaccessible areas.
fully interact with others located at another node through
video and audio links. Each of the nodes can be
Village Resource Centres (VRC)
further extended using other technologies like Wi-Fi,
Communication Satellites are effectively used to Wireless and Optical Fibre. These extensions may
provide all most all the useful information needed for serve as local clusters around the areas where the
the village folks by establishing village resource centres/ VRC is located.
information kiosks (Fig 3). This is another highly The information provided will be in the form of
successful project to bring technology to the benefit of geo-referenced land record, natural resources, suitable
rural areas. sites for drinking water as well as sites for recharging
VRC project strives to promote a need based to replenish ground water, water harvesting, wastelands
single window delivery system for providing services that can be reclaimed, rural employment creation,
in the areas of education, health, nutrition, weather, watershed, environment, infrastructure, alternate
environment, agriculture and livelihoods to the rural cropping pattern, and so on. By suitably blending the
population to empower them to face challenges. This information derived from earth observation satellites
satellite based project, aims for digital connectivity to with ground derived and weather related information,
remote villages for providing multiple services such as locale-specific community advisory services can be
tele-medicine, tele-education and remote sensing provided. Community based vulnerability and risk
applications through a single window. The setting up related information, provision of timely early warning
of VRC is yet another saga of adventure that has been and dissemination of severe weather related information
taken up by ISRO to benefit the rural society. can lead to reliable disaster management support at
the village level.
The VRC concept has been evolved and
implemented by integrating ISRO’s capabilities in VRCs can also provide a variety of services like
satellite communications and satellite based earth tele-education, tele-medicine, online decision support,
interactive farmers’ advisory services, tele-fishery, e- used along with the VHRR imageries for weather
governance services, weather services and water forecasting.
management. By providing tele-education services,
Satellite derived products data are increasingly
the VRCs act as virtual community centric learning
used in conjunction with conventional meteorological
centres. At the same time, VRCs will provide
observations in the synoptic analysis and conventional
connectivity to speciality hospitals thus bringing the
weather forecasts to extract information of relevance
services of expert doctors closer to the villages.
to various sectors in India. The impact of satellite data
In addition, VRCs will facilitate access to spatial is phenomenal in certain areas of meteorological
information on important subjects like land use/land applications such as thunderstorm forecasts, Tropical
cover, soil and ground water prospects which can Cyclone monitoring, aviation forecasts etc. The major
enable the farmers to get support in taking important application of satellite data has been the monitoring of
decisions based on their query. Besides, VRCs will Synoptic weather systems ranging from thunderstorms,
enable online interaction between the local farmers fog detection to cyclones and planetary scale
and agricultural scientists. Fishermen can obtain phenomena such as monsoon.
information on sea state and wave heights. Provision Synoptic applications of satellite imagery are in
of information on many governmental schemes, use at Indian Meteorological Department. Following
location and farming system specific action plans satellite data products are derived from Kalpana-1
based on weather, community specific advice on soil and the same are also archived and displayed on IMD
and water conservation are the other services rendered website daily.
by VRCs.
• Earth cloud imagery in visible, IR bands and
Search & Rescue water vapor every hour on black & white and
color.
The search and rescue facility provided by satellites • Clouds Motion vectors over Bay of Bengal,
is a breakthrough in technology in ensuring safety and Arbian Sea and Indian Ocean at 00, 07, 12
providing support in case of disasters/accidents to the and 1 8 UTC and being disseminated over
tourists, fishermen etc. GTS.
• Outgoing long wave radiation (OLR) on daily/
Meteorology weekly/monthly basis at 0600UTC.
• Sea surface Temperature (SSTs) at 0600 UTC
All the first generation INSATs and INSAT-2A,
is being derived over Bay of Bengal, Arabian
2B and 2E in the second generation in the operational
Sea and Indian Ocean.
INSAT system were configured for multi purpose
mission with transponders for telecommunication & • Quantitative Precipitation Estimates (QPE)
broadcasting and Meteorological payloads. To meet on daily/weekly/monthly basis.
the demanding requirements of Indian Meteorological
Satellite bulletins based on 3 hourly / 1 hourly
Department’s, ISRO launched a satellite with dedicated
Kalpana-1 cloud imageries are prepared and transmitted
meteorological payloads - Very High Resolution
to all the forecasting offices on Global Telecom Service
Radiometer (VHRR) and Data Relay Transponder
(GTS) through RTH, New Delhi. Satellite imageries
(DRT), and is named as Kalpana-1 after launch in the
and animations are also put on the IMD website on
memory of Late (Ms) Kalpana Chawla, Indian born
regular basis for the users and public. Heavy rainfall
American Astronaut.
advisory, bulletins are also transmitted regularly for
Very High Resolution Radiometer (VHRR) is a IMD forecasting offices and other centres daily. During
three band VHRR capable of imaging the earth in the tropical cyclones in Arabian Sea and Bay of
three spectral bands namely Visible (VIS), Water Bengal hourly special satellite bulletins with intensity
Vapor (WV) and Thermal Infra Red (TIR), to provide and position are also issued for all the users by the
both day and night coverage. satellite Div., IMD.
The weather data relay transponder (DRT) receives During winter season, fog formation is one of the
402.75 MHz signals from unattended weather data most important weather events over northern parts of
collection platforms, translates them to 4506.05 MHz India which affects the aviation badly. The Kalpana-1
and retransmits it to a central facility - Meteorological imageries are regularly utilized for capturing such
Data Utilization Centre, IMD, Delhi. This data is being events.
284 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Remote Sensing satellites launched and operated The 1050 kg satellite placed in a Polar
by ISRO are highly successful in providing vital sunsynchronous orbit of 720 km height. IRS-P4 carried
information regarding earth resources and also in on board an Ocean Colour Monitor (OCM) and a
providing cartographic maps for town planning and Multifrequency Scanning Microwave Radiometer
development. (MSMR).
Data from Indian Remote Sensing Satellite is used OCM is a solid state camera operating in eight
for various applications of resources survey and narrow spectral bands. The camera is used to collect
management under the National Natural Resources data on chlorophyll concentration, detect and monitor
Management System (NNRMS). phytoplankton blooms and obtain data on atmospheric
aerosols and suspended sediments in the water.
• Preharvest crop acreage and production
estimation of major crops. The Microwave Scanning Radiometer (MSMR)
• Drought monitoring and assessment based on which operates in four microwave frequencies both in
vegetation condition. vertical and horizontal polarisation with its all weather
capaility is useful for measuring sea surface temperature
• Flood risk zone mapping and flood damage
and meteorological parameters like Sea Surface
assessment.
Temperature (SST), Atmospheric water vapor and
• Hydro-geomorphological maps for locating Sea surface winds.
underground water resources for drilling well.
• Irrigation command area status monitoring The Oceansat data helped fisherman to a great
extent in locating fish concentration in the sea and in
• Snow-melt run-off estimates for planning water increasing the fish produce. The satellite data also
use in down stream projects helped to study the ocean behaviour and to provide
• Land use and land cover mapping safety guidelines and precautions to fisherman, divers
• Urban planning and navy to a great extent.
• Forest survey
INFOSAT
• Wetland mapping
• Environmental impact analysis The future Telecommunication spacecrafts will
• Mineral Prospecting be developed from transmission in to Information
• Coastal studies Satellites (INFOSAT). They will be given many of the
properties of terrestrial telephone exchanges and signal
• Integrated Mission for Sustainable
processing equipments and it will be possible to integrate
Development for generating locale-specific
them directly into future global networks. They will
prescriptions for integrated land and water
thus permit immediate applications of many existing
resources development in 174 districts.
and future services. Because of their inherent built up
Many state governments are utilizing the remote flexibility, these satellites will be able to support and
sensing data for various applications including town speed up the initial experimental phase of many new
planning. For example, Andhra Pradesh has been services before their trial on terrestrial networks. These
extensively using ISRO satellites for societal type of satellites will enable new services to be tried
applications. Andhra Pradesh is one of the foremost in out over a large area before being put in to the market
the utilisation of remote sensing technology for and optimally adapted to suit the most appropriate
management of land and water resources and disaster transmission medium.
management. Remote sensing technology is used for
The future INFOSATS will be of three types:
watershed management in drought prone districts of
National / Regional, International & Relay.
Adilabad, Ananthapur, Kurnool, Mahaboobnagar,
Nizamabad and Ranga Reddy. Management plans for Future technologies will enable the construction of
major irrigation projects like Srirama Sagara and an INFOSAT network in which the above three will
Nagarjuna Sagara have been planned using remote be connected to each other. The onboard processors
sensing data. More than 35,000 bore wells have been will ensure that the signals to be exchanged between
drilled using remote sensing data with better than 90 the satellite and the terrestrial subscribers are combined
percent success rate. using Time Division and Space Division multiplexing
PAL & RAO : SATELLITE TECHNOLOGY UTILIZATION 285
techniques and distributed in accordance with the user necessary for planning of the satellite network and the
requirements. Special coding techniques will ensure terrestrial network to be closely coordinated. Until
the security of transmissions. now satellites have connected the terrestrial networks
of various countries and organizations.
Configuration of INFOSAT
Services to be provided by INFOSAT
The satellite platform will have multiple
Radio and data distribution services to many users
reconfigurable antennas/transmitters with dynamic
which are spread over a large area. Dara collection
power sharing / Receivers in X various frequency
services for large areas with many data transmitter
ranges / large reconfigurable switching matrices at
stations (multi point-to-point operation : weather, oil,
baseband and at RF level / intersatellite links permitting
electricity & water meter reading etc.).
signals to be exchanged between satellites according
to changing requirements / complex and efficient analog • Telecommunication services for thin routes
/ optical / digital signal processors / New modulation • Telecommunication services with ships, aeroplanes,
technique and multiple access techniques. It will be space vehicles, etc).
sufficiently broadband systems & may have even
• In the area of business communications demand is
optical space communication components.
growing for broad band internet & multimedia
The satellites besides their autonomous control communication facilities which can be applied
and power generation equipment may have sensors to flexibly using satellites.
observe the earth’s atmosphere and pass the data to • Worldwide Radio paging
an appropriate station after processing. This will help • Video conferencing and high resolution TV
to deal with the situation like Orissa cyclone. Such a broadcast trials.
system will obviously have Geo synchronous
spacecrafts with some orbiting satellites to take care • Mobile radio services can be combined with location
of North/South pole regions. finding services and be used for automatically
locating subscribers.
The INFOSAT Network (Fig 4) is to be an integral • Earth observation with special warning mechanism
component of the planned worldwide broadband and environmental protection services in close
telecommunication network. Therefore it will be coordination with terrestrial sensors.
By integrating INFOSAT into terrestrial network, Satellite technology connects total country irrespective
it will be possible to supply all the subscribers of a of location. This paper briefed about the services
future global network with all essential information. provided by satellites to bring the cutting edge
This will make the location of the user quite irrelevant. technology and benefits to the rural and inaccessible
area.
CONCLUSION
REFERENCES
There is a great influence of present day fast
evolving technologies particularly the digital techniques 1. Surendra Pal, Advanced Satellite Communication
Technology-A Perspective, Journal of Spacecraft
on the development of the society. Satellite technology
Technology, vol 12, no 2, July 2002.
bridges the gap between urban and rural areas in
utilizing the technologies. Inspite of advancement of 2. Lectures by the author Surendra Pal on Satellite
technology to a great extent, many parts of India are Communication and various studies carried out in
still backward and inaccessible by terrestrial means. the area.
Authors
S Pal an alumnus of Birla Institute Millennium Medal for various developments, inventions and
of Technology & Science (BITS), Pilani overall contributions towards the growth of communication and
and Indian Institute of Science (IISc), space technology in India. Besides these he holds Indian,
Bangalore is presently Outstanding European & International patents for his various inventions.
Scientist & Deputy Director (Digital &
Communication Area) and also Program Dr Pal has published more than 170 papers in international
Director, Satellite Navigation Program and national journals of repute , one book on communications &
at ISRO Satellite Centre, Bangalore. guided PhD students. He has been on the editorial board of
various technical journals. He is a Distinguished Visiting Professor
He joined Indian Space Research of Indian National Academy of Engineering. His fields of interest
Organization (ISRO) in 1971, after a brief tenure at TIFR. Dr are Space Communications, Information Technology,
Pal is responsible for starting Antenna & Microwave activities Electromagnetics, Antennas and Satellite Navigation.
in the ISRO Satellite Centre at the inception. He has been
responsible for the development & fabrication of all spacecraft Dr Pal has been closely associated with IETE and served
related telecommunication systems for India’s all satellites starting IETE in various capacities. He was also a Council Member for
with Aryabhata to present day IRS & INSAT series of spacecraft. seven years. He was instrumental in starting IETE research
He has been consultant to: (i) INMARSAT/ICO (UK) for board activities. He has also been guest editor of many IETE
LEOSAT definitions, responsible for the development of Satellite Technical Review issues on Antenna and Microwaves.
Hand Held Telephone Antenna System for ICO (UK) * * *
(ii) International Telecommunication Union (ITU) for defining
Regional African Satellite Communication System and
V Sambasiva Rao obtained BE
(iii) Nanyang Technolgical University, Singapore on X-Sat. He
(Electronics and Communication
was also Chairman of Information Infrastructure Working Group Engineering) from College of Engineering,
of Ministry of Information Technology. Presently he is spear Kakinada, (Andhra University) in 1973.
heading the Indian Satellite Based Wide Area Augmentation
He joined ISRO Satellite Centre,
System – GAGAN (GPS Aided Geo Augmented Navigation
Bangalore in 1974 and presently, heading
System) and Indian Satellite Navigation activities for civil aviation
Communication Systems Group. He is
applications.
responsible for the development of high
He is also on the United Nations panel of experts on Global bit rate data transmitters for all IRS
Navigation Satellite Systems (GNSS). satellites and various RF and microwave
systems for IRS and INSAT missions.
Dr Pal a Distinguished Fellow of IETE, Fellow of Indian
National Academy of Engineering, Indian National Academy of Shri Rao is a fellow of IETE and received Distinguished
Sciences, Fellow, IEEE(USA), MAMTA(USA), MASI and has Achievement Award for launching Aryabhatta, NRDC award
received more than a dozen awards, including, Performance 1994 for the development of X-band high bit rate QPSK Modulator
Excellence Award by ISRO, Distinguished Achievement Award and IETE-IRSI (83) award 2006. He has published more than 30
for launching Aryabhatta, Hari Om Ashram-Vikram Sarabhai technical papers in national and international journals.
Award, IETE Ramlal Wadhawa & Hari Ramji Toshniwal Awards,
Om Prakash Bhasin Award, four NRDC awards and IEEE Third * * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 287-311
BHARGAB B BHATTACHARYA
Advanced Computing and Microelectronics Unit, Indian Statistical Institute,
203, BT Road, Kolkata 700 108, India.
email: shibaji@vlsi.iitkgp.ernet.in; drc@cse.iitkgp.ernet.in; bhargab@isical.ac.in
A memory test algorithm for detecting neighborhood pattern sensitive faults (NPSFs),
including static NPSF (SNPSF), passive NPSF (PNPSF) and active NPSF (ANPSF), is proposed
in this paper. The patterns can also detect all the traditional faults present in the memory
array such as stuck-at faults (SAFs), transition faults (TFs), coupling faults (CFs) and address
decoder faults. Next, a programmable BIST architecture is designed. The BIST circuit allows
the users to select a vast variety of test algorithms based on their choice. The single BIST
circuit is capable of testing different types of memory cores embedded in SOC. The proposed
BIST circuit is shared among the different memory cores in an SOC. For this purpose, test
wrappers for the shared BIST circuits and the memory cores are designed. Finally, a test
scheduling algorithm is developed to reduce the overall test time.
Today, an SOC may contain more than 100 BIST module and shared BIST module. The dedicated
memory cores. In order to reduce test cost and shorten BIST module does not reuse the exiting on-chip
time-to-market, testing time for these cores should be resources for testing of embedded memories. The
minimized by adopting an appropriate test access dedicated BIST can be classified into categories: FSM-
architecture. However, power dissipation is becoming based and micro-code based. The FSM based BIST
a key challenge for the SOC. Therefore, power circuit [8-12] can generate a single simple pattern or
consumption of the embedded memory cores has to complex suite of patterns. These types of BISTs are
be considered during SOC testing. In SOC test mainly used to generate a single pattern by adopting a
scheduling, multiple cores have been tested concurrently suitable March algorithm. However, a series of patterns
while satisfying the power constraints during test. is required to obtain a better fault coverage. This
makes the design of FSM complex. The FSM based
To lower the BIST area overhead by implementing
approach requires large number of flip-flops to
an efficient architecture, the main challenge lies in
implement the FSM. So, modifying the test algorithms
finding a solution, which also reduces the overall testing
or memory architecture requires changing of the BIST
time. The main contributions of this paper can be
design. Micro-code based BIST is a programmable
summarized as follows.
BIST that also does not reuse the on-chip resources
• Introduction of a memory test algorithm, which but, it overcomes the previous limitations. These types
can detect all NPSFs, SFs, TFs, and CFs of BISTs are highly flexible for different test structures.
present in the memory array. They use a storage element to store the test algorithms.
• Implementation of a new programmable The micro-based BIST architecture can be found in
memory BIST. The proposed BIST module [10,13,14]. In [13], authors used two operations per
allows testing of memories for various fault March element. However, this algorithm was not able
models. A simple test pattern generator for to detect NPSFs and CFs. In [14], authors proposed a
NPSFs is implemented by employing the regular diagnosis process in a much more efficient way.
structure of Cellular Automata (CA). Hence, However, implementations of these architectures
the proposed BIST architecture provides a require a certain amount of storage elements, which
very simple and cost effective memory test leads to increase in the area overhead. In summary,
solution for handling various fault models. The an FSM-based programmable BIST circuit requires
area overhead due to proposed BIST is similar less area overhead but provides relatively less flex-
to that of the existing memory BISTs. ibility, while a micro-code based BIST circuit provides
more flexibility but requires high area overhead.
• Design of wrapper cells for the embedded
memories, so that they can share a common On the other hand, the shared BIST approach
BIST circuit. Sharing of BIST by several reuses the on-chip resources, which result low area
memory cores reduces area overhead. overhead. Since, a complex SOC usually contains one
• Power-constrained test scheduling to reduce or more processors (or microprocessors), reusing these
the overall testing time. existing on-chip processors for testing the embedded
memories can lead to lower area overhead. But, reusing
The rest of the paper is organized as follows. In a processor as a BIST needs much more sophisticated
section 2, we discuss the related work on memory wrapper cells for both the embedded memories and
BIST, different March algorithms, and their processors. In [15], an on-chip microprocessor was
implementation. Section 3 presents the fault model reused for testing the other cores in an SOC. It
used to develop the proposed test algorithm. The eliminated the area overhead caused by dedicated
memory test algorithm is proposed in section 4. The BIST hardware. In [16], authors developed an SOC
proposed programmable BIST circuit design and the test solution, which includes both software and
shared BIST approach are described in section 5 and hardware based approaches. They introduced a new
section 6 respectively. Section 7 presents the BIST approach, called hardware/software co-testing
experimental results and finally section 8 concludes for embedded memories. An embedded processor-
the paper. based built-in self-repair (BISR) design for embedded
memories was proposed in [17]. In [17], authors
2. RELATED WORK removed the controller a typical BISR by reusing the
embedded processors. They also developed a memory
Recently, several approaches have been proposed wrapper for this purpose. In [18], a processor-based
for designing of the BIST circuits. We can classify BIST design for embedded DRAMs was proposed. In
these approaches into two categories, the dedicated [18], the authors used two separate instruction storage
SHIBAJI BANERJEE et al : A PROGRAMMABLE BUILT-IN SELF-TEST FOR EMBEDDED MEMORY CORES 289
memories, combined with flexible address, data, and scheduling problem. While in [27], the authors used
clock generators to perform the test using a minimum network transportation approach to solve the above
of dedicated test pins. The processor-based BIST problem, in [29,33], the authors used a bin packing
design can also be found in [19,20]. However, shared approach, which provides an efficient and effective
BIST techniques require significantly longer testing solution for test scheduling. In [33], the authors pro-
time, which increases the test cost. Again, reusing a posed a restricted 3-D bin packing approach to optimize
processor for BIST may also introduce delay in the the test time under pin and power constraints. A
functional path of the chip. This is not desirable for power-constrained test compatibility graph (P - TCG)
high speed SOC. was created in [24]. Using the P - TCG, the authors
had constructed a set of power-constrained concurrent
Most of the existing BIST techniques have used
test set (PCTSs) to facilitate concurrent testing. They
the March test algorithms. However, these BIST
scheduled the test sets by dynamically partitioning and
techniques are not capable of detecting NPSFs. In
allocating the tests. The memory test scheduling
[21], the authors proposed a modified March algorithm
problem can be found in [16,34]. In [34], it is assumed
that can also detect NPSFs along with other
that the March elements of a March algorithm can be
conventional faults. They used complex data
performed by distinct test resources. The disadvantage
background to detect the NPSFs. The BIST
of this method is that it needs a special control environ-
implementation of this algorithm requires complex
ment, which adds complexity to the test architecture.
control circuits, which increase the area overhead. In
In the present work, we have designed a programmable
[22], the authors have used auxiliary memory to create
BIST. This BIST is capable of testing different memory
the periodic data backgrounds in the memory cell
cores with various sizes and types. The proposed
array for NPSFs. They loaded the desired data
BIST can be shared among the different memory
background by means of a scan path. Thus, gener-
cores in the SOC to reduce the area overhead. The
ation of the periodic data background needs an auxiliary
test wrappers for these shared memory cores are
memory. In addition, complex control circuits are
presented next. Finally, a test scheduling algorithm is
required to deliver the data from the auxiliary memory
proposed to reduce the overall testing time.
to the memory under test. In the present work, we
have proposed a memory test algorithm which removes
3. FAULT MODEL
the requirement of the periodic data background. The
proposed test algorithm can detect the NPSFs, SFs,
TFs, and CFs present in the memory array. A
programmable BIST circuit is then designed for the
embedded memories. The BIST technique supports B E N
the proposed algorithm along with the other existing S W B
March algorithms and allows the users to select a
N S
specific algorithm based on their choice.
W B E
Embedded memories are usually much smaller
than stand-alone memories [23]. This leads to high S
BIST overhead. So, sharing a BIST with different
memory cores in an SOC leads to a cost effective N
solution. However an SOC contains several memories
W B
of different sizes, with different access protocol and
timing. The shared BIST circuit should be able to test
them. Thus, the BIST circuit should be independent of
the type and size of the memory cores and should be Fig 1 The five-cell NPSF
programmed for different types of memory cores. In
other words, a single BIST circuit may test different
types and size of memory cores. Next, an appropriate A pattern sensitive fault (PSF) is a conditional
scheduling algorithm is required to schedule the memory coupling fault in which the content of a memory cell or
cores for the BIST. Several approaches of test the ability to change its content is influenced by a
scheduling for the digital cores in the SOC can be certain bit pattern in other cells in the memory. Here
found in [24-32]. In [26], the authors presented an the data retention and transition of the victim cell are
integer linear programming (ILP) model to formulate affected by a set of aggressor cells. A neighborhood
the integrated wrapper/TAM co-optimization and test pattern sensitive fault (NPSF) is a special case of
290 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
pattern sensitive faults, wherein the influencing affect the value of another (coupled) memory
(coupling) cells are in the neighborhood of the influenced cell. The coupled cell can have a lower address
(coupled) cell. The coupled cell is called the base (or than the coupling one (it is indicated as Ù), or
victim) cell and the coupling cells are called the deleted a higher address (Ú). There are three types of
neighborhood cell. The neighborhood includes all the CFs:
cells in the deleted neighborhood as well as the base a. Idempotent CF (CFid): A transition in
cell. The five-cell NPSF as shown in Fig 1 is considered coupling cell forces the coupled cell to
in the paper. The neighborhood includes the base cell have a certain value. There are eight
(B) as well as the deleted neighborhood cells (including subtypes of CFid: Ù(-,0), Ù(-,1), Ù(¯,0),
N, E, W and S cells that are physically adjacent to B). Ù(¯,1) Ú(-,0), Ú(-,1), Ú(¯,0) and Ú(¯,1).
The boundary cells on the memory cell array are
b. Inversion CF (CFin): A - or ¯ transition
special cases that have only 2 or 3 deleted neighborhood
in coupling cell inverts the value of coupled
cells. This fault model can further be categorized into
cell. The following four CFin sub-types
three classes as follows.
may occur: Ù(-, ), Ù(¯, ), Ú(-, ) and
«
«
«
Ú(¯, ).
«
• A static NPSF (SNPSF) occurs if the base
cell is forced to a certain state because of the
appearance of a certain pattern in the deleted
neighborhood. To detect SNPSF, all the 32
TABLE 1: Neighbourhood pattern
static neighborhood patterns shown in Table
1a should be applied. B 00000000000000001111111111111111
• A passive NPSF (PNPSF) occurs if the base N 00000000111111110000000011111111
E 00001111000011110000111100001111
cell can not change its state from 0 to 1, or
W 00110011001100110011001100110011
from 1 to 0 because of the appearance of a S 01010101010101010101010101010101
certain pattern in the deleted neighborhood.
To detect PNPSF, all 32 passive neighbor- (a) Static neighborhood patterns (SNPs)
hood patterns shown in Table 1b should be
applied. B ----------------¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
N 00000000111111110000000011111111
• An active NPSF (ANPSF) occurs if the base E 00001111000011110000111100001111
cell is forced to a certain state when a transi- W 00110011001100110011001100110011
S 01010101010101010101010101010101
tion occurs in a deleted neighborhood cells,
while other deleted neighborhood cells as- (b) Passive neighborhood patterns (PNPs)
sume a certain pattern. To detect ANPSF, all
the 128 active neighborhood patterns shown B 00000000000000001111111111111111
in Table 1c must be applied. N --------¯¯¯¯¯¯¯¯--------¯¯¯¯¯¯¯¯
E 00001111000011110000111100001111
Traditional NPSF test patterns (as shown in Table W 00110011001100110011001100110011
1 are used to detect all the NPSFs. In addition to S 01010101010101010101010101010101
NPSFs, other faults can also be present in the memory
B 00000000000000001111111111111111
array. These faults are stuck-at faults, transition faults, N 00000000111111110000000011111111
coupling faults, decoder faults and Read/Write logic E ----¯¯¯¯----¯¯¯¯----¯¯¯¯----¯¯¯¯
faults as described below [35]: W 00110011001100110011001100110011
S 01010101010101010101010101010101
• Stuck-at fault (SAF): A permanent stuck-at-0 B 00000000000000001111111111111111
fault (SAF0) or stuck-at-1 fault (SAF1) that N 00000000111111110000000011111111
E 00001111000011110000111100001111
may occur in any memory cell.
W --¯¯--¯¯--¯¯--¯¯--¯¯--¯¯--¯¯--¯¯
• Transition fault (TF): A memory cell in state s S 01010101010101010101010101010101
(s Î {0,1}) fails to undergo an s to s transition B 00000000000000001111111111111111
when s is to be written in the cell. A TF can N 00000000111111110000000011111111
be either a rising-transition fault (TF1 or TF-) E 00001111000011110000111100001111
or a falling-transition fault (TF0 or TF¯). W 00110011001100110011001100110011
S -¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯-¯
• Coupling Fault (CF): During write operation a
transition in one (coupling) memory cell may (c) Active neighborhood patterns (ANPs)
SHIBAJI BANERJEE et al : A PROGRAMMABLE BUILT-IN SELF-TEST FOR EMBEDDED MEMORY CORES 291
Fig 2 Divide 8´8 memory into 9 blocks Ý{RB0, WB1}; ß {RB1, WB0}; ß {RB0, WB1};
292 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
ß{WNp1}, {WE p2, WS p3, WW p4, WB´ p1}; Option D: To detect only SAFs, TFs and CFs but
not NPSFs, the following test may be used:
ß{RB1, WB0}; Ý {RB0}.
Ý {WB0}; Ý {RB0, WB1}; RB1, WB0}; ß{RB0,
Here WN p1 means that a write operation using p1 WB1}; ß {RB1, WB0}; Ý {RB0}.
value as data is performed in the north neighborhood
of the selected base cell. Similarly WB0 (RB0) means 5. THE BIST ARCHITECTURE
that a write (read) operation using the 0 value as the
data is performed in the selected base cell. The notation The present programmable BIST circuit is ca-
Ý indicates that all the base cells are processed in up pable of testing different types of memories using
address order, while ß. indicates that all the base cells various test algorithms. For this purpose, BIST module
are processed in down address order. The total test uses an instruction register (IR)- The IR can be loaded
operation is divided into 9 sub-operations. During each serially from the primary input of the chip. Thus one
sub-operation, one base cell from each block is selected, primary test input pin is required to load the data into
and the patterns according to the proposed test algorithm the IR. If a primary test input pin is not available, one
are applied. can multiplex a functional primary input pin to load the
IR. If we have more than one BIST circuits in the
Option B: If the user wants to detect only NPSFs
SOC, they can share one primary test input pin. So, no
the test algorithm can be described as
additional primary test input pins are required for these
Ý{WB0}; Ý {WN p1, WE p2, WS p3, WW p4}; Ý{RB0, BISTs. Once a pattern is loaded into the IR, the
control circuit of the BIST decodes and generates the
WB1}; ß {WN p1, WE p2, WS p3, WW p4}; ß {RBl}.
different control signals based on the pattern. The
Option C: For detecting all NPSFs, SAFs and pattern loaded into the IR will determine a particular
TFs but not CFs, one may use: March algorithm, size, and the type of the memory
under test. The block diagram of the proposed BIST
Ý {WB0}; Ý {WN p1, WE p2, WS p3, WW p4}; Ý{RB0, circuit is shown in Fig 4. It consists of following four
WB1}; ß {WNp1, WEp2, WSp3, WWp4}; ß {RBl}. blocks.
1. Instruction Register
the memory. Similarly, during each read binary sequence of length (n + 0.5)2n using an n-bit
operation it provides the data value expected shift register, an (n – l)-bit counter, (n – 1) two-input
to be read from memory. This expected value XOR gates, a FF (flip flop) and a two-input AND
is supplied to the comparator of the output gate. In [43] a recursive technique of generating all
response verifier. For bit-oriented memories, SIC pairs is proposed using a structure called
the LSB of the data register is used to supply complementation sequence. An iterative design of the
the data during read and write sequences. TPG is proposed in [43], which produces exactly n.2n
SIC pairs in optimal time (i.e., n.2n + 1) with less area
3. Test Pattern Generator (TPG) overhead.
A TPG is used to generate the patterns for NPSFs. In SIC test scheme of NPSF testing, a two-
The total number of 2-pattern tests needed to sensitize pattern test consists of a vector X1(xn , xn–1, ..., xi,...
all NPSFs of a neighborhood of size k is ,x2,x1) applied to the CUT, followed by another vector
(k – l).2 k–1 . A sequence of all such tests for X2(xn,xn–1,...,xi,... ,x2,x1) that differs from X1 by unit
k = 5 is needed to sensitize all NPSFs of size five hamming distance in a single bit xi, 1 £ i £ n. To
(Table 4). It can be observed that the test patterns include all SIC pairs, X must appear at least n times in
form a unit distance code, where the two successive S, and exactly n times if S has to be of minimum
test vectors differ only in one bit position. If the length. Furthermore, if (X1, X2) is a SIC test pair, (X2,
vectors of a two-pattern test differ in single (multiple) X1) must also be a SIC test pair. The sequence S can
bit (s), such a pair is known as single (multiple)-input- thus be described in terms of graph traversal in a
change, i.e., SIC (MIC) pair. directed graph.
For an n-input CUT, there are n.2n SIC two For an n input CUT, all the n.2n SIC pairs can be
pattern tests. In pseudo-exhaustive adjacency testing represented by a directed graph G(V, E). The graph G
(PEAT), all such test pairs are generated in (n + l).2n can be obtained by an undirected graph G´(V,E/2),
clock cycles [39]. The TPG proposed in [40] generates
whose 2n nodes are labeled with n-bit binary vectors
all SIC test pairs in n.2n+1 clock cycles using a
and two nodes v i and v j are adjacent if their
(2n + l)-bit scan shift register, an n-bit LFSR and n
corresponding binary labels lie at unit hamming
2-input XOR gates. Design-for-testability techniques
suited to BIST applications for robust testing of delay distance. The directed graph G can be obtained by
faults with SIC pairs, appeared in [41], where the replacing each undirected edge (vi, vj) in G´ with two
TPG is driven by a Johnson counter. The TPG directed edges (vi,vj) and (vj,vi). Such a graph for 3
proposed in [42], generates all SIC test pairs in a input CUT (i.e., n=3) is shown in Fig 5.
0 0 0 0 0 0 0 0 1 0 1 0 1 0 1 0
0 0 0 1 0 1 0 0 1 0 1 1 1 1 1 0
0 0 1 1 1 1 0 0 1 0 0 1 0 1 1 0
0 1 1 1 1 1 0 1 1 1 0 1 0 1 1 1
1 1 1 1 1 1 1 1 0 1 0 1 0 1 0 1
1 1 1 0 1 0 1 1 0 1 0 0 1 0 0 1
1 1 0 0 0 0 1 1 0 1 1 0 1 0 0 1
1 0 0 0 0 0 1 0 0 0 1 0 1 0 0 0
0 0 0 0 0 0 0 0 1 0 1 0 1 0 1 0
0 0 1 0 1 0 0 0 1 0 0 0 0 0 1 0
0 1 1 0 1 0 0 1 1 1 0 0 0 0 1 1
1 1 1 0 1 0 1 1 0 1 0 0 0 0 0 1
1 1 1 1 1 1 1 1 0 1 0 1 0 1 0 1
1 1 0 1 0 1 1 1 0 1 1 1 1 1 0 1
- - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - -
296 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
1111
25 0000
32
8 116 9 17
24
1000 0001 0010 0100
7
18 10 31
26 15 23
2
D Q D Q D Q D Q
1000 0001 0010 0100
7 Cleck
26
10 27 5
23 9 6 CL CL CL CL
24 8
21 12
11 25 28 CL-Combinational Logic
1001 1100 1010 0010 0101 0110
31 20
16 13
2 1 Fig 9 An one-dimensional CA
14
30 19
3 17 32
18 cells are arranged in a two-dimensional plane.
4 15
1100 1011 1110 0111 Additionally, to achieve dynamic behavior (discussed
later) the connection order of the top row depends on
a signal named MODE.
CL CL
D Q a7 D Q a6
D Q a5 D Q a4
clk clk clk
clk
clk clk clk
a3 a2 a1 a0
D Q D Q D Q D Q
clk clk clk
It can be shown that T28 = 1, i.e., it is a group CA Table 5 indicates that when MODE = 0, the top
of length 8. row implements a single left shift function and when
MODE = 1, it implements a double left shift function.
Here, the CA is made to change dynamically with Initially MODE signal is made to 0. After 7 clock
time by changing its characteristic matrix. The cycles it becomes 1 and in the next clock cycle it
composite characteristic matrix is chosen as T17T2, so again resets to 0. The MODE signal can be
that the CA is run for 4 subcycles, where in each generated by AND-ing the output bits of a 3 bit up-
subcycles the characteristic matrix is T1 for 7 time counter. The technique of generating MODE signal
steps and T2 for 1 time step. So the CA forms a group is shown in Fig 11. The counter is initially set to 000
CA of length (7 + 1) 4 = 32 and generates 32 different state and after 8 clock cycles it again returns to 000
patterns. state. The counter is fed by the same clock as used
Thus the CA generates 32 patterns in a single by the CA.
cycle. If the cycle is run with a new initial state then Moreover, for the loading of the next seed 00011010
another set of 32 patterns can be generated. By we see from the Table 5 that the end of the first cycle
selecting suitable starting seed, the required 64 patterns of length (7 + 1) 4 = 32 can be detected by observing
indicated in Table 4 can be generated. An Euler line the pattern < a7,a6,a5,a4,a3,a2,a1,a0 > = < 00010000 >.
can be obtained in G1 with < a7,a6,a5,a4,a3,a2,a1,a0 > Thus for 4 bit, the present technique requires 24.2 = 64
= < 00010000 > is chosen as initial seed, while for G2 clock cycles to generate all the SIC pairs.
SHIBAJI BANERJEE et al : A PROGRAMMABLE BUILT-IN SELF-TEST FOR EMBEDDED MEMORY CORES 299
Mode a 7 a6 a5 a4 a3 a2 a1 a0 Mode a7 a6 a5 a4 a3 a2 a1 a0
0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 1 0 0
0 0 0 1 0 0 0 0 1 0 0 0 0 1 1 1 0 0
0 0 1 0 0 0 0 1 1 0 0 0 1 0 1 1 0 1
0 1 0 0 0 0 1 1 1 0 0 1 0 0 1 1 1 1
0 0 0 0 1 1 1 1 1 0 1 0 0 0 1 0 1 1
0 0 0 1 0 1 1 1 0 0 0 0 0 1 0 0 1 1
0 0 1 0 0 1 1 0 0 1 0 0 1 0 0 0 1 0
1 1 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0
0 0 1 0 0 0 0 1 0 0 0 0 1 0 1 0 0 1
0 1 0 0 0 0 1 1 0 0 0 1 0 0 1 0 1 1
0 0 0 0 1 1 1 1 0 0 1 0 0 0 1 1 1 1
0 0 0 1 0 1 1 1 1 0 0 0 0 1 0 1 1 1
0 0 1 0 0 1 1 0 1 0 0 0 1 0 0 1 1 0
0 1 0 0 0 1 0 0 1 1 0 1 0 0 0 1 0 0
1 0 0 0 1 0 0 0 1 0 0 0 0 1 0 0 0 0
0 0 1 0 0 0 0 0 0
Mode a 7 a6 a5 a4 a3 a2 a1 a0 Mode a7 a6 a5 a4 a3 a2 a1 a0
0 0 0 0 1 1 0 1 0 0 1 0 0 0 1 1 1 0
0 0 0 1 0 0 0 1 1 0 0 0 0 1 0 1 1 0
0 0 1 0 0 1 0 0 1 0 0 0 1 0 0 1 1 1
- - - - - - - - - - - - - - - - - -
- - - - - - - - - - - - - - - - - -
data independent of the size of the data line coming logic circuit used to generate the content of the control
(read) out from the memory i.e., the comparator is register is shown in Fig 13 where the DT1 field is
common for both bit-oriented and word-oriented assumed to be 2-bit long.
memories. Figure 12 shows the block diagram of the
(ii) Fault Indicator (FI): On getting an error
comparator. The data control register indicates the
signal from the comparator, the FI sends the controller
number of bits of the data (read out from the memory
and the other units into the wait state by issuing a
under test) to be compared. The content of the data
BIST error signal (BE). It also sets the ERROR port
control register is determined by the DT1field of the
of the BIST. At the same time, it sets the status of the
IR. The size of the data control register is equal to the
current operation in the EOP (Error Operation) register.
DT2 field of the IR. For bit-oriented memories, the
The error operation includes the failed operation i.e.,
content of the data control register is 000... 01, which
the position of the failed read operation in the read-
indicates that only the last bit of the data register is to
write sequence during which the fault is detected and
be considered during comparison. For a bit-oriented
the address of the faulty cell. For example, the proposed
memories, as shown in Fig 12, the outputs of the AND
test algorithm has 5 read operations (maximum) in the
gates An–1 to A1 and A´n–1 to A´1 zero while the outputs
read-write sequence; so 3 bits (indicated as “session”)
of the AND gates A´0 and A´0 are depend on the data
in Fig 14 are needed to indicate the failure. The FI
read out from the memory and D0 (LSB of the data
then sends the content of EOP register to the user
register) respectively. Thus, the outputs of the all XOR
through ERROR_ADDR port of the BIST. When the
gates are zero except X0 gate whose output will be
content of EOP is completely shifted out, the FI waits
determined by D0 and the data read out from the
for the user input. When the user sends a pulse
memory. For a word-oriented memory, the number of
through CONT_U port, it issues a continue signal
bits from the LSB side of the data control register set
(CONT) and resets the BE signal to allow other units
to one indicates the size of the word line. Thus, for a
to continue the execution. This has an advantage that
word-oriented memory of word size n1 (n1 < n), the
the user can stop the test on every fail.
n1 bits of the data register from the LSB will be
considered during comparison.
6. SHARED BIST APPROACHES
The present comparator is independent of the size
of the word line of the memory under test. The content In current scenario, an SOC may contain more
of the data control register is determined from the than 100 memory cores. In order to reduce the test
DT1 field of the IR by using some logic circuit. The cost and shorten time-to-market, the testing time for
Data coming
from the
Memory under
A0 A n–1 test
A n–1 A1 A0
A1
X n–1
Output X1
X0
Comparator
Fig 12 Logic circuit to generate the content of the data control register
SHIBAJI BANERJEE et al : A PROGRAMMABLE BUILT-IN SELF-TEST FOR EMBEDDED MEMORY CORES 301
DT1 field of IR
1 DT1 DT0 DC3 DC2 DC1 DC0 Decision
DT1 DT0
0 0 0 0 0 1 Bit-oriented Memory
0 1 0 0 1 1 Word-oriented Memory
(2 bits word line)
1 0 0 1 1 1 Word-oriented Memory
(3 bits word line)
DC3 DC2 DC1 DC0 1 1 1 1 1 1 Word-oriented Memory
(3 bits word line)
Data Control Register
Fig 13 Logic circuit to generate the content of the data control register
ATPG) for a scan-based core, or can be pseudo- The TAM switch has two useful modes of operation
random vectors generated by a BIST or can be namely, cross and pass modes as shown in the Fig 16
functional test vectors. and Fig 17. In the cross configuration, the switch is
configured to send the test patterns to the CUT. In the
pass configuration, the switch simply passes the test
patterns to the next switch.
responses are sent to BIST circuit through the Dout pin that, memory cores nearer to each other will fall in the
and the TAM switch. The BIST then compares the same partition. Next, we assign a certain number of
observed responses with the expected responses. When BIST circuits to each partition based on the total
a fault is detected, the BIST sends the faulty cell number of available BIST circuits. In the present
address and the order of the read operation in the work, we have allotted two BIST circuits to each
read-write sequence of the March algorithm. The partition. For example, if the total number of available
BIST core is also encapsulated by a wrapper cell, BIST is 10, then we divide the memory cores into 5
which is similar to the memory wrapper cell. The size partitions and assign two BIST circuits to each partition.
of TAM switch for the BIST wrapper cell, depends on The test structure thus obtained is guaranteed to offer
the number of input and output pins of the BIST. a good trade-off between the minimum test routing
overhead and minimum test time. The scheduling
B. Proposed test scheduling algorithm algorithm is then used to schedule the memory cores
for the BIST circuits available in a partition. The
It is observed that sharing a BIST with large scheduling algorithm is described below.
number of memories, located at different regions in
the chip results in a certain routing overhead. Sharing We first explain the scheduling algorithm with two
a BIST with two memory cores separated by a large BIST circuits. Later, we will extend the same algorithm
distance leads to high routing overhead. So, in this with some modification when more than two BIST
case sharing is not desirable. The long routing wire circuits are available. In order to illustrate the proposed
length gives rise to many unwanted effects in deep test scheduling algorithm, we consider an SOC with
sub-micron technology which degrades the six memory cores as an example. Each core is provided
performance of the chip. In the present work, we with total test cycles needed to test it and maximum
divide the memory cores into different partitions based power dissipation during testing. These parameters
on their location in the chip. This location is determined are listed in Table 6. In addition, in the present case,
by the functional proximity between the memory cores we assume that the Pmax is 100 mw.
and the other cores in the SOC. Thus an approximate
location of each memory core can be found before The power-constraint test-scheduling problem can
layout [47]. Based on these approximate locations, we be solved by finding compatible memory cores based
divide the memory cores into different partitions so on test power dissipation. To determine the compatible
Fig 19
to B1. M 5 is connected to M 4 by an edge in G3. So, incident on M 6 are deleted from G´ to obtain G´1 (Fig
M4 is assigned to B2. The node M 5 and edges 21b). In G1´ the new weight of the M 3 is obtained by
incident on it are deleted from G3 to obtain subgraph subtracting its weight from the weight of M6. In G´1,
G4. In G4 the weight of M5 is changed by subtracting M 1 is connected to M 3. So M 1 is assigned to B2.
its weight from the weight of M4. The subgraph G4 is But, in G1´ the weight of the M 1 is greater than the
shown in Fig 19e. The subgraph G4 contains only one weight of the M 3. So, in this case M 3 will be deleted
node and so scheduling process is ended here. The from G´1. The node M 3 and the edges incident on it
final allotment of the cores to B1 and B2 is shown in are deleted from G´1 to obtain G´2. In G´2, the weight
Fig 20. of M 1 is obtained by subtracting the weights of M 1
and M 3. The subgraph G1´ is shown in Fig 21c. The
We consider another SOC as an example whose
remaining scheduling processes are same as in the
P – TCG (G´ (V´,E´)) is shown in Fig 21a. The number
previous one and are shown in Fig 21d, and Fig 21e.
of cores and the parameters of each core are kept the
Finally, the allotment of the cores to B1 and B2 is
same as in the previous one except the test time of the
shown in Fig 22. The outline of the Algorithm_l is
core M 1 and M 6 are now changed to 1000 and 1200
presented in Fig 23.
respectively. The change in scheduling process due to
these is described below.
Now we generalize the algorithm for any number
As in the previous case, the node M 3 has the of available BIST circuits. Here instead of one weight,
maximum test time in G´ and between M 1 and M 6, two sets of weights W and P are assigned to each
M 6 has larger test time. So, M 3 and M 6 are assigned node in P – TCG. The sets of weights are determined
to B1 and B2 respectively. The node M6 and edges from the test time and the power dissipation of each
node. Let n and m be the number of nodes in P – TCG
and the number of available BIST circuits respec-
tively. The weights assign to ith node will be {w1i,w2i
,...,wmi} and {p1i,p2i,....,pmi}, where w1i and p1i are
4300
the test time and power dissipation of node i
B2 M6 M1 M5 respectively. Initially, other elements of Wi and Pi are
B1 M3 M2 M4
set to zero (i.e., w2i = 0, w3i = 0,..., wmi = 0, p2i = 0, p3i
= 0,..., pmi = 0).
1000 2000 3000 4000
In P – TCG, consider the node having maximum
Time w1 value. Let Mi be the node having maximum w1.
Next, find an adjacent node of Mi with maximum w1.
We assume it is Mj. After getting Mj, modify Wi =
{w1i, w2i ,..., wmi} of Mi by replacing w2i with w1j and
Fig 20 Allotment of the memory cores of Fig 19a to Pi = {p 1i ,p 2i,...,pmi } by replacing p2i with p 1j
B1 and B2 respectively. After assigning Mj to Mi (i.e., to B2),
306 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Fig 21
and column 3 represent the number of gates and area To verify the correctness of the proposed test
overhead due to BIST for each of these memory scheduling algorithm, several experiments have been
cores respectively. The results indicate that the area carried out. The scheduling algorithm is applied to 50,
overhead due to BIST is low and it decreases when 100, and 150 embedded memory cores. Table 8 to
the size of the memory core increases. Table 10 show the results of testing time due to the
scheduling algorithm with the number of available
BIST circuits for different number of memory cores.
Each table corresponds to a particular maximum power
TABLE 7: Experimental results for different constraint. For Table 8 maximum power limit is
BIST circuits
considered as 250 mW and the power consumption of
memory BIST gate BIST area each core is randomly assigned from 50 mW to 150
size count overhead mW. Similarly, for Table 9 (Table 10) maximum
64K ´ 1 5890 8.24% power limit is taken as 500 mW (750 mW) and the
power consumption of each core is randomly assigned
128K ´ 1 6958 5.04%
from 100 mW to 300 mW (200 mW to 500 mW). The
256K ´ 1 8235 3.05%
memory cores included in this experiment are of
64K ´ 4 7894 2.92% different sizes with the number of address lines randomly
128K ´ 4 9885 1.85% assigned from 7 to 20 and bits per word are also taken
randomly between 1, 2, 4, 8, 16, and 32.
308 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
TABLE 8:
TABLE 9:
TABLE 10:
Before applying the scheduling algorithm, we first 3. K J Lin & C W Wu, PMBC: a programmable BIST
divide memory cores into different partitions based on compiler for memory cores, In Proc of IEEE Int,
Workshop on Testing Embedded Core-based
the number of available BIST circuits. Then we allot System Chips, April 1999, pp P2.1-P2.6.
two BIST circuits to each partition. Memory cores are
grouped randomly such that the difference in number 4. Kuo-Liang Cheng, Chia-Ming Hsueh, Jing-Reng
Huang, Jen-Chieh Yeh, Chih-Tsun Huang & Cheng-
of cores between any two partitions can be at most Wen Wu, Automatic generation of memory built-in
one. For example, when the number of available BIST self-test cores for system-on-chip, In Proc of Asian
circuits and total number of memory cores are 6 and Test Symposium, 19-21 Nov 2001, pp 91-96.
100 respectively, the memory cores will be divided 5 A J van de Goor, Testing Semiconductor Memories:
into 3 partitions. The number of cores in each partition Theory and Practice, John Wiley and Sons,
will be 33, 33, and 34. The partitions thus created are CHicEester, England, 1998.
balanced with respect to each other. Among the 6 6. J P Hayes, Testing memories for single-cell pattern
available BIST circuits, we assign 2 of them to each sensitive fault, IEEE Transactions on Computers,
partition and schedule them by using the proposed 1980, vol 29, pp 249-254.
algorithm. 7. K K Saluja & K Kinoshita, Test pattern generation
for API faults in RAM, IEEE Transactions on
When the number of available BIST circuit is one, Computers. March 1985, vol 34, p 284-287.
cores will be tested sequentially one after another and
8. Y Zorian, S Dey & M J Rodgers, Test of future
the test time can be represented by Tmax. We can system-on-chips, In Proc of IETE/ACM
define the test time reduction factor (R) as, International Conference on Computer Aided
Design, 5-9 Nov 2000, pp 392-398.
Tmax – test-time-with m BIST circuits
R = ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ´ 100% 9. P Camurati, P Prinetto, M S Reorda, S Barbagallo, A
Tmax Burri & D Medina, Industrial BIST of embedded
RAMs, IEEE Design and Test of Computers, 1995,
vol 12, Issue 3, p. 86.
8. CONCLUSIONS
10. K Zarrineh & S J Upadhyaya, On programmable
The paper demonstrates that for SOCs compris- memory built-in self test architectures, In Proc of
Design, Automation and Test in Europe Conference
ing of SRAMs and DRAMs the proposed, algorithm and Exhibition, 9-12 March 1990, pp 708-713.
can be used to test them. For DRAMs it can detect all
NPSFs, SAFs, TFs, and CFs. For SRAMs it can 11. Po-Chang Tsai, Sying-Jyan Wang & Feng-Ming
Charig, TSM-based programmable memory BIST
detect all SAFs, TFs, and CFs. A programmable BIST with macro command, In Proc of IEEE International
circuit is proposed, which provides good flexibility on Workshop on Memory Technology, Design, and
both test algorithms and memory types. It supports Testing, 3-5 Aug. 2005, pp. 72-77.
different types of test algorithms and memory cores 12. Chih-Tsun Huang, Jing-Reng Huang, Chi-Feng Wu,
according to the user choice. The user needs to program Cheng-Wen Wu & Tsin-Yuan Chang, A
the BIST before testing of a particular memory core programmable BIST core for embedded DRAM,
with a particular test algorithm. Today, an SOC may IEEE Design and Test of Computers, Jan-March
consist of tens to hundreds of embedded memories. 1999, vol 16, Issue 1, pp 59-70.
Using a dedicated BIST for each of them is not a 13. H Koike, T Takeshima & M Takada, A BIST scheme
feasible solution due to large area overhead. So, sharing using microprogram ROM for large capacity
of BIST is needed for these types of memories. A memories, In Proc of International Test Conference,
10-14 Sept 1990, pp 815-822.
power-constraint test scheduling algorithm is also
proposed to reduce overall testing time. The 14. I Schanstra, D Lukita, A J van de Goor, K Veelenturf,
experimental results verify the above claims. & P J van Wijnen, Semiconductor manufacturing
process monitoring using built-in self-test for
embedded memories, In Proc of International Test
REFERENCES Conference, 18-23 Oct 1998, pp 872-881.
15. R Rajsuman, Testing a system-on-a-chip with em-
1. Semiconductor Industry Association. Int.
bedded microprocessor, In Proc of International
Technology Roadmap for Semicconductors, http://
Test Conference, 28-30 Sept 1997, pp 499-508.
public.itrs.net/files/1999_51A. Roadmap/Home.htm.
16. Bai Hong Fang, Qiang Xu, & N Nicolici, Hardware/
2. Shankar Hemmady, Tom Anderson, & Yervant software co-testing of embedded memories in
Zorian, Verification and Testing of Embedded Cores, complex SOCs, In Proc of International Conference
In Proc of Design SuperCon, January 1997, pp s122- on Computer Aided Design, 9-13 Nov. 2003, pp 599-
1-19. 605.
310 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
17. Chin-Lung Su, Rei-Fu Huang, & Cheng-Wen Wu, A 30. S K Goel & E J Marinissen, Effective and efficient
processor-based built-in self-repair design for test architecture design for SOCs, In Proc of
embedded memories, In Proc of Asian Test International Test Conference, 7-10 Oct. 2002, pp
Symposium, 16-19 Nov 2003, pp 366-371. 529 538.
18. J Dreibelbis, J Barth, H Kalter & R Kho, Processor- 31. K Chakrabarty, Design of Systein-on-a-Chip Test
based built-in self-test for embedded DRAM, IEEE Access Architectures using Integer Linear
Journal of Solid-State Circuits, Nov 1998, vol Programming, In Proceedings IEEE VLSI Test
33(11), p 1731. Symposium (VTS), 2000, pp 127-134.
19. Ching-Hong Tsai & Cheng-Wen Wu, Processor- 32. K Chakrabarty, Test Scheduling for Core-Based
programmable memory BIST for bus-connected Systems Using Mixed-Integer Linear Programming,
embedded memories, In Proc of Asia and South In Transactions on Computer-Aided Design, Oct
Pacific Design Automation Conference, 30 Jan-2 2000, vol 19, pp 1163-1174.
Feb 2001, pp 325-330.
33. Yu Huang, Sudhakar M. Reddy, Wu-Tung Cheng,
20. C A Papachristou, F. Martin & M Nourani,
Paul Reuter, Nilanjan Mukherjee, Chein-Chung Tsai,
Microprocessor based testing for core-based
Omer Samman & Yahya Zaidan, Optimal Core
system on chip, In Proc of Design Automation
Wrapper Width Selection and SOC Test Scheduling
Conference, 21-25 June 1999, pp 586-591.
Based on 3-D Bin Packaging Algorithm, In Proc of
21. Kuo-Liang Chen, Ming-Fu Tsai & Cheng-Wen Wu, International Test Conference, Oct 2002, pp 74-82.
Efficient Neighborhood Pattern-Sensitive Fault Test
Algorithm for Semiconductor Memories, In Proc of 34. Wei-Lun Wang, March based memory core test
IEEE VLSI Test Symgosium, 29 May 2001, pp 225- scheduling for SOC, In Proc of Asian Test
230. Symposium, 15-17 Nov 2004, pp. 248-253.
22. S Boutobza, M Nicolaidis, K M Lamara & A Costa, 35. M L Bushnell & V D Agrawal, Essentials of
Programmable memory BIST, In Proc of Electronic Testing for Digital, Memory, and Mixed-
International Test Conference, 2005, pp 1155-1164. Signal VLSI Circuits, Kluwer Academic Publishers,
Boston/Dordrecht/London, 2000.
23. Said Hamdioui & Georgi Gaydadjiev, Future
challenges in memory testing, citeseer.ist.psu.edu/ 36. S Hamdioui, Z Al-Ars & A J van de Goor, Testing
649203.html. static and dynamic faults in random access
memories, In Proc of IEEE VLSI Test Symposium, 28
24. D Zhao & S Upadhyaya, Dynamically partitioned
April-2 May 2002 pp 395-400.
test scheduling with adaptive TAM configuration
for power-constrained SoC testing, IEEE 37. M L Bushnell & V D Agrawal, Essentials of
Transactions on Computer-Aided Design of Electronic Testing for Digital, Memory and Mixed-
Integrated Circuits and Systems, June 2005, vol 24, Signal VLSI Circuits, Springer, Berline, Germany,
Issue 6, pp 956-965. 2000.
25. M Nourani & C Papachristou, An ILP formulation to
38. A J van de Goor & I B S Tlili, A systematic method
optimize test access mechanism in system-on-chip
for modifying inarch tests for bit-oriented memories
testing, In Proc of International Test Conference, 3-
into tests for word-oriented memories, IEEE
5 Oct 2000, pp 902-910.
Transactions on Computers, Oct 2003, vol 52(10), p
26. Vikram Iyengar, Krishnendhu Chakrabarty, E J 1320.
Marinissen, Test Wrapper and Test Access
Mechanism Co-optimization for System-on-chip, in 39. G L Craig & C R Kime, Pseudo-exhaustive adjacency
Journal of Electronic Testing: Theory and test: A BIST approach for stuck-open faults, In Proc
Applications, April 2002, vol 18, pp 213-230. of International Test Conference (ITC), 1985, pp
126-137.
27. Sandeep Koranne & Vishal Suhas Choudhary,
Formulation of SOC Test Scheduling as a Network 40. W Wang & S K Gupta, Weighted random robust
Transportation Problem, In Proc of Design, path delay testing of synthesized multilevel circuits,
Automation and Test in Europe Conference and In Proc of VLSI Test Symposium (VTS), 1994, pp 291-
Exhibition, 4-8 March 2002, p 1125. 297.
28. Yu Huang, Wu-Tung Cheng, Chien-Chung Tsai, 41. I P Shaik & M L Bushnell, A graph approach to DFT
Omer Samman, Yahya Zaidan, & Sudhakar M Reddy, hardware placement for robust delay fault BIST, In
Resource Allocation and Test Scheduling for Proc of International Conference on VLSI Design,
Concurrent Test of Core-Based SOC Design, Nov 1995, pp 177-182.
2001, pp 361-366.
42. I Voyiatzis, A Paschalis, D Nikolos & C Halatsis,
29. V Iyengar, K Chakrabarty, & E J Marinissen, On An efficient built-in self test method for robust
using rectangle packing for SOC wrapper/TAM co- path delay fault testing, Journal of Electronic
optimization, In Proc of IEEE VLSI Test Symposium, Testing : Theory and Applications (JETTA), 1996
28 April-2 May 2002, pp 253-258. pp 219-222.
SHIBAJI BANERJEE et al : A PROGRAMMABLE BUILT-IN SELF-TEST FOR EMBEDDED MEMORY CORES 311
43. D K Das, I Chaudhuri & B B Bhattacharya, Design of 46. Subhayu Basu, I Sengupta, D Roy Chowdhury & S
an optimal test pattern generator for built-in self Bhowmik, An Integrated Approach to Testing
testing of path delay faults, In Proc of International Embedded Core and Interconnects Using Test
Conference on VLSI Design, January 1998, vol 139, Access Mechanism (TAM) Switch, in Journal of
pp 205-210, IEEE CS Press, USA. Electronic Testing: Theory and Applications
(JETTA), Special Issue on Plug-and-Play Test
44. S Wolfram, Statistical mechanics of cellular Automation for System-on-a-Chip, Aug 2002, vol
automata, Rev Mod Phys, vol 55, no 3, pp 601-644, 18, pp 475 485.
July 1983.
45. D Roy Chowdhury, I Sengupta & P Pal Chaudhuri, 47. A Sehgal, F Liu, S Ozev & K Chakrabarty, Test
Cellular automata based pattern generator for planning for mixed-signal SOCs with wrapped analog
testing RAM, IEE Proceeding – Part E, November cores, In Proc of Design, Automation and Test in
1992, vol 139, pp 469-476. Europe, 2005, pp 50-55.
Authors
Shibaji Banerjee received BTech. Bhargab B Bhattacharya received
and MTech degrees from the Institute BSc degree in physics from the
of Radio Physics and Electronics, Presidency College, Calcutta, the BTech.
Kolkata, in 2001 and 2003, respectively. and MTech. degrees in radiophysics
He is currently working toward the PhD and electronics, and the PhD degree in
degree in Computer Science and computer science all from the University
Engineering at the IIT, Kharagpur. His of Calcutta, India. Since 1982, he has
research interests are in the area of scan been on the faculty of the Indian
architecture, system-on-chip testing, Statistical Institute, Calcutta, where he
and memory testing. is full professor. He held visiting
positions at the Department of Computer Science and Engineering,
University of Nebraska-Lincoln, USA, during 1985-1987, and
* * * 2001-2002, and at the Fault-Tolerant Computing Group, Institute
of Informatics, the University of Potsdam, Germany during
Dipanwita Roy Chowdhury is 1998-2000. In 2005, he visited the Indian Institute of Technology
Professor of Computer Science and Kharagpur as VSNL Chair Professor. His research interest includes
Engineering, Indian Institute of logic synthesis and testing of VLSI circuits, nanotechnology,
Technology, Kharagpur, India. Her digital geometry, and image processing architecture. He has
current research interests are in the field published more than 200 papers in archival journals and refereed
of VLSI design and testing, cryptography conference proceedings, and holds 9 United States Patents.
and cellular automata. Dr Roy Dr Bhattacharya is a Fellow of the Indian National Academy
Chowdhury received her BTech. and of Engineering, a Fellow of the National Academy of Sciences,
MTech degrees in Computer Science India, and a Fellow of the IEEE. He is on the editorial board of
from the University of Calcutta in 1987 the Journal of Circuits, Systems, and Computers (World Scientific,
and in 1989 respectively, and obtained the PhD degree in Singapore), and the Journal of Electronic Testing - Theory and
Computer Science and Engineering from the Indian Institute of Applications (Springer).
Technology, Kharagpur, in 1994. She has been teaching for more
than 13 years.
* * * * * *
INVITED PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 313-321
V JAYARAMAN, FIETE
Earth Observation (EO) inputs on various natural resources parameters are effectively
used for integrated land and water resources development at grass roots helping in
enhancing the productivity levels in dryland areas. EO inputs and a judicious mix
Geographical Information System (GIS) and Management Information System (MIS) has
helped in characterization, prioritization, developmental plan preparation, generation of
implementation strategies (with peoples’ participation), concurrent project monitoring and
impact assessment at various stages of implementation. The innovativeness of this study
lies in the infusion of modern technologies in the form of providing EO-based inputs on
natural resources for development, simple-to-use information system with MIS/GIS in local
language, reporting and monitoring through simple client-server solutions, participation of
people, experts, Local Government, NGOs, project monitoring group and other stakeholders
with one clear objective of achieving sustainable development. This innovative method of
project implementation and monitoring has brought about significant impacts on the natural
resources conservation and positive trends in livelihood condition of the people.
313
314 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
image analysis, new software tools and geospatial This is a community driven program wherein village
databases for participatory watershed development people involve in participatory planning, implementation
program is well adapted in the ‘Sujala’ project being and maintenance of assets.
implemented in Karnataka state. Attempt has been
made to use an optimal combination of earth observation Geospatial inputs and IT solutions
inputs, field observations and Information Technology
for planning in implementation, monitoring and impact Satellite Remote sensing inputs, GIS, a near real-
assessment. time MIS solution and GIS solution for participatory
planning are the core areas of technology utilisation
COMMUNITY BASED WATERSHED for resource mapping, database generation, analysis
DEVELOPMENT and information extraction for watershed planning,
implementation and monitoring. In Sujala project, high
Organization and societal issues have recently resolution satellite data with a spatial resolution of 6
become important subjects of GIS research meter has been utilised to generate maps on 1:12,500
(Obermeyer, 1998). Watershed development is a well scale. Typical areas of technology utilization under the
accepted method of treating land through scientific project are
means for sustainable development of natural resources.
• Watershed prioritization
Watershed itself is a natural boundary connecting
similar elevation points in the form of ridge tops which • Resources Inventory and Mapping
is self contained polygon of natural resources which is • Land and Water resources action plan
amenable for conservation and protection. Different • Site selection for implementation
methods, suitable to locale specific conditions, have
• Web based Management information system
been practiced for ages across the world, that involves
(MIS) solution
treating land on a ridge-to-valley basis to restore
equitable use of soil and water resources in a given • GIS based action plan development
terrain. It is a well known fact that any watershed • Implementation of concurrent Monitoring
development program poses varieties of challenges as • Impact Assessment
there are many stakeholders and players who are • Post project Evaluation
important for achieving success in implementation. A
few typical parameters considered while watershed • Run off estimation by imperial methods.
development are: large extent of private land (belonging
to farming community), common land (sharable by the EO DATA FOR DEVELOPMENTAL
village community), livestock, fuel and fodder areas, PLANNING
tanks and waterbodies, waterholding structures,
EO data from Indian Remote Sensing satellite has
farming community with land, landless labourers, infants
proved to be very effective in similar such programs
and so on. The newer techniques of image processing,
earlier. However, with the availability of high resolution
extraction of natural resources information, using GIS
images from space platforms, RESOURCESAT and
for data integration and modeling, providing extract of
CARTOSAT, there has been a significant shift in the
such processing in the form of land and water resources
way the technology is used for monitoring. LISS 4
plans, IT tools for monitoring and impact assessment
multispectral images (5.8 m) and high spatial resolution
have made significant impacts in successful
CARTOSAT 1 PAN (2.5 m) could be effectively
implementation of watershed development.
used to carry out an on-line tracking of the various
Watershed Development program developmental works effectively and accurately. It is
rare that such high resolution image data is used, not
Sujala is a participatory watershed development only for facilitating the farming community to take up
programme, which is being implemented by Government technically sound decisions for watershed development,
of Karnataka with World Bank assistance in 5 districts, but also for monitoring such implementations. The
viz., Kolar, Tumkur, Chitradurga, Dharwad and Haveri usage of EO data has been facilitated by adopting
of Karnataka State. The major objective of the project special product generation techniques, like, image fusion,
is to improve the productive potential of degraded natural color composites, landuse classification, change
watersheds in dry land areas and poverty alleviation of detection and other thematic layers creation to aid in
rural community. The project is spread across 77 sub- locale specific action plan preparation. Geospatial layers
watersheds covering an area of about 0.51 million at high resolution is created and integrated with
hectares and benefiting about 400,000 households. cadastral/parcel data to facilitate proper planning of
DIWAKAR & JAYARAMAN : SPACE ENABLED ICT APPLICATIONS 315
Randomization, Area based sampling with probability resources but also provide critical inputs for on-line
proportion to size) and use of the same for such a big monitoring of the implementation works. This type of
project and delivering the same on the field effectively. possibility has given a big boost in the area of project
The sub-watersheds are randomly selected based on monitoring by bringing about transparency in project
their agroclimatic conditions, status of land use and implementation and peoples’ participation at grass roots.
soil. Within each sub watershed, 3 micro watersheds
High-resolution satellite images have been
are selected at random, representing Ridge, Middle
processed and provided at large scales to the village
and Valley portions. Households are selected and
communities involved in watershed development to
randomly sampled based on land holding class, i.e.,
help them understand the locale specific nature of the
Marginal, Small, Big and Landless using Probability
terrain for better planning and management of natural
Proportion to Size (PPS) criteria. To achieve acceptable
resources. Image fusion and natural color composite
accuracies of estimates and as per the project
based product generation techniques were successfully
guidelines, 10% sampling intensity is considered for
employed to make the basic satellite images community-
data collection and analysis (Fig 2).
friendly. Image fusion is extensively used to take full
advantage of high-resolution geometry of the pixel and
Geospatial databases on Natural
the optimum spectral mixing of the multispectral
Resources
component to produce best possible output. It was
Very few community based developmental projects discovered that for preserving spectral characteristics,
adopt newer technologies for resource mapping. The one needs a high level of similarity between the
technique based on EO inputs has tremendous potential panchromatic image and the respective multispectral
to not only establish a strong baseline on natural intensity [1]. Varieties of image fusion methods have
been used in the present study amongst some of the
popularly used techniques, such as, IHS, PCA,
Broovey’s method, Discrete Wavelet Transforms,
adaptive linear band combination and so on. However,
these techniques do depend on the inter-band correlation
existing amongst the multispectral channels to
appreaciate the results.Multispectral transformation
techniques are adopted to generate natural color
composite images by establishing a relationship between
the spectral bands producing false color composite
with that of true color composite (Fig 3a).
These images and other ancillary data have been
analysed and integrated under GIS to generate various
types of resource maps, viz., land use / land cover,
soils, slope, hydrogeomorphology, drainage, transport
network, settlements, land parcels, etc., at the micro-
Fig 2 Geostatistical approach for baseline establishment
watershed level. These maps play an important role in
understanding the spatial nature and interrelationship
that exists between different resources. From a practical
point of view, a high-resolution satellite image depicting
terrain in true color could be the most comfortable one
for conventional interpretation and visualization [2].
Locale specific action plans for sustainable development
of land and water resources are generated on micro
watershed basis by integrating thematic information
from the resource maps, peoples’ aspirations and
socioeconomic inputs with special emphasis on
community needs (Fig 3b).
Such action plans, prepared through community
participation, basically address private lands (the land
belonging to individuals in the villages) and common
lands (the Government land utilized by the village
DIWAKAR & JAYARAMAN : SPACE ENABLED ICT APPLICATIONS 317
undertaken in the project and also provides varieties of intervals to establish the net contribution of the project
graphics and analysis functions as value added tool. to poverty alleviation and natural resource regeneration
Other unique feature of the package is the possibility (Fig 7). Impact is evaluated using a variety of qualitative
to have access to GIS maps through WebGIS and quantitative indicators with respect to baseline,
technology with MIS reports and graphs to provide a midterm and post-project status. Impacts are also
spatial dimension for monitoring and management of analysed based on observations made in the project
the project more efficiently (Fig 6). and control areas. A comprehensive benchmark data
has been established through judicious combination of
conventional and remote sensing data to facilitate
monitoring and impact assessment. Participatory
Observations, Focus Group Discussions, Transect
Walks are some of the other methods of data collection
at community level. The data available from MIS/GIS
system, thematic reviews and specific case studies
are also utilised for the impact assessment. One of the
most crucial points to be noted in effectively carrying
out such impact studies is the establishment of a
strong baseline database, both from satellite remote
sensing, GIS and field based observations.
EO data processing and GIS modeling for The application of cutting edge technology including
Impact Assessment: remote sensing, GIS and Computer based monitoring
system in conjunction with ground observations has
The potential of space technology in generating provided robust baseline and change data and wealth
the base line information on land and water resources of information for in-depth analysis (Grant, 2006). The
and in monitoring the progress and success of watershed mid-term assessment of the project has indicated that
development programme has been substantiated from project investments are having a measurable impact
various studies carried out so far [3]. Impacts due to on the indicators representing the project development
various interventions are monitored through a objectives. Some of the visible changes observed are:
combination of remote sensing data, GIS, MIS data, increase in average crop yields, crop diversity has
process monitoring data and farmers / household increased from an average of 2-5 crops in the baseline
surveys. Through a scientifically designed mechanism, to 4-9 crops, increase in annual household income due
impact assessments are done at pre-determined time to employment, income generating activities and
320 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Authors
K V V S REDDY
Department of ECE, Andhra University College of Engineering,
Vishakhapatnam 530 003, India.
email: anu_praise2004@yahoo.co.in
This paper deals with simulation of Nakagami-m fading in wireless channels using
generalized SC DS-CDMA and MC DS-CDMA schemes using binary phase shift keying
modulation scheme. The paper proposes for deriving Bit Error Rate (BER) in BPSK-SC DS-
CDMA and MC DS-CDMA systems over Nakagami-m fading channels. The numerical results
are plotted as BER vs SNR for various values of Nakagami factor m, number of users K and
jamming interference (JSR) using MAT LAB software. The performance of Nakagami-m
channel using BPSK-SC DS-CDMA and MC DS-CDMA techniques are compared. It is
observed that there is a decrease in Bit Error Rate for increase in m for SC DS-CDMA and
MC DS-CDMA. The number of users k, JSR increases as Bit Error Rate increases for MC DS-
CDMA. For the same values of m BER for MC DS-CDMA is smaller when compared with SC
DS-CDMA using Maximal Ratio Combining (MRC) diversity technique.
SYSTEM MODELS
F ADING is observed in wireless communication
channels [1] due to multi path propagation. The
Nakagami-m distribution accurately models the fading
(a) SC DS-CDMA
effect for short distance communications. This paper
Introduction
deals with the derivation of BER in BPSK- SC DS-
CDMA and MC DS-CDMA systems over Nakagami-
In this section we apply the MGF-based approach
m channel. In the generic form, DS-CDMA access is
presented to derive the average BER performance of
a spread spectrum technique for simultaneously
binary DS-CDMA systems over these channels [3].
transmitting a number of signals representing
The results presented in this section are applicable to
information messages from a multitude of users over a
systems that employ RAKE reception with coherent
channel employing a common carrier. The method by
Maximum Ratio Combining (MRC).
which the various users share the channel is the
assignment of a unique pseudo noise (PN) type code
1. Transmitter
to each user (which accompanies the transmission of
information) with orthogonal like properties that allows
We consider a binary DS-CDMA system with
the composite received signal to be separated into its
Ku independent users sharing a channel simultaneously,
individual user components, each of which can then be
each transmitting with power P at a common carrier
demodulated and decoded. A complete discussion of
frequency fc=wc/2P .using a data rate Rb=1/Tb and a
techniques for accomplishing these functions and their
chip rate Rc=1/Tc. The kth user, k=1, 2…Ku is assigned
a unique code sequence {ak, j} of chip elements
Paper No 126-A; Copyright © 2007 by the IETE. (–1, +1). So that its chip waveform is given by
323
324 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Lp
L
hk (t) = S a k,l e–jqk,l d(t–tk,l) rk = S a k,l rk,l k=1,2...Ku
l=1 l=1
Ne 1. Transmitter
¾ ¾ º s2N + s2S + s2M
2
(2Ku +1) W T – 3 N0
= ¾ ¾ ¾ ¾ ¾ ¾ ¾ W lEb + ¾ ¾ s0(i)
6PG 2
Channel r(t)
With gl = W 1Eb / N0 the average received SNR per bit s1(i) Trans- Rake s (n)
corresponding to the first path. mitter h(l) Receiver k
Pilot
Since we are assuming BPSK modulation, the Channel
n(t)
average SNR per bit of lth path Channel
sN–1(i)
Estimation
–1
W l Eb é (2Ku + 1) W T – 3 ù
gl,e = ¾ ¾ ¾ = gl ê 1 + ¾ ¾ ¾ ¾ ¾¾ gl ú Multi-user Interference
Ne ë 3PG û
The expression for average BER is given by: Fig 1 System model for the MC DS-CDMA system
326 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
The block diagram for the transmitter for the kth 2. Channel Model
user is shown in Fig 2.
Following the system design and modeling
Consider a BPSK multi carrier coherent DS- assumptions of kando and well as their bandwidths are
CDMA system with Ku independent users each chosen so that the separate subbands fade slowly and
transmitting with power P. The users are simultaneously nonselectively. Under these assumptions, the channel
sharing an available bandwidth BW=(1+a )/Tc, where transfer function of the lth subband for the kth user is
T c is the chip duration of a corresponding
a ´k,l exp (jqk,l) where the {(a ´k,l}l=1Mf are the fading
single-carrier wideband DS-CDMA system, and
a (0 £ a £ 1) is the roll off factor of the chip wave- amplitude RV’s and {q´k,l}l=1Mf are independent
shaping Nyquist filter. The available spectrum BW is uniformly distributed RV’s over [0,2P ].The average
divided into (not necessarily contiguous) Mf equal fading power of lth subband is denoted by
bandwidth subbands each of width BWMf approxi- W l´ = (a ´k,l)2 and is assumed to be independent of k.
mately equal to the coherence bandwidth of the channel.
Each subband is assigned a carrier which is DS-
CDMA modulated with the same user information at 3. Receiver
the bit rate 1/Tb and chip rate 1/(MfTc) (see Fig 2).
Each user is effectively assigned a specific periodic The receiver consists of a bank of Mf matched
code sequence of chip elements (+1, –1) and of filters followed by MRC (see Fig 4). Each of the
processing gain per subband PG´ = PG/Mf. We assume received modulated sub band carriers is first passed
deterministic subband PN codes with ideal auto through a band pass chip-modulated filter H*(f), then
correlation function. The use of band limited (Nyquist coherently demodulated, sampled, despread and
shaped spreading) waveforms with wave-shaping filter summed, all these operations assume that the receiver
transfer function denoted by H(f) guarantees that the is correctly phase and time synchronized at every
DS waveforms do not overlap. branch. We denote by X(f) = H(f)H*(f) = |H(f)|2 the
overall frequency response of the chip waveshaping
The spectrum of the MC DS-CDMA signal is Nyquist filter and assume that X(f) is a root-cosine
shown in Fig 3. frequency response given by
Chip
pulse
Shaping
2cos(2P f1t+qk,l)
Data Transmitted
Sequence Multi carrier
Dn{k} DS signal
Impulse
modul H(f)
ator
Cn{k}
PN code
sequence
¥
S dh{k} Cn{k} d(t–MfTc)
–¥
¥
S dh{k} Cn{k} d(t–MfTc)
–¥
2cos(2P f1t+qk,l)
4. Performance Analysis
f1 f2 f3 fU
U
¾ Conditional SNR
D Te
Mf Mf
Fig 3 Spectrum of the MC DS-CDMA signal Eb Mf
rk = S wk,l rk,l = ± S wk,l a ´k,l ¾ + S wk,l (IMl+N)
l=1 l=1 M f l=1
1 W
¾ , 0 £ | f | £ ¾ (1–a ) Gaussian RV with Conditional mean and conditional
W 2
ì1 é
2W ë
æ 1 æ2p|f|
è 2a è W
ö öù
¾ ê1– sin ç ¾ ç¾ ¾ – p ÷ ÷ú ,
ø øû
variance are given by
Mf æ Mf ö Eb
X(f) =
í W W
¾ (1–a ) £ | f | £ ¾ (1+a )
2 2
E[rk| {a ´k,l}l=1 = ± ç S wk,l a´k,l ÷
è l=1 ø
¾ ¾
Mf
î W
0, ¾ (1+a ) £ | f |
2
Mf Mf
N0
var (rk | (a ´k,l}l=1 = S (wk,l)2 ¾
l=1 2
Sample at
H*(f–f1) T=nMfTc
+H*(f+f1) 1 PG´–1
LP ¾ ¾ S() Wk,l
F PG´ n=0
{Cn {k}}
Removes R{k}
second harmonic
of carrier
Sample at
T=nMfTc
rk
Wk,Mf
1 PG´–1
H*(f–fMf)
+H*(f+fMf)
LP ¾ ¾ S()
F PG´ n=0
{Cn{k}}
2cos(2P f1t+qk,l Mf)
ì ¾h2 , f – ¾W2
j
j
j W
£ | f | £ fj + ¾ j
2 SC DS-CDMA (for different values of m)
Snj (f) = í 100
î 0, elsewhere
10-1
Where fj denotes jammer carrier frequency. Average BER (dB)
10-2
The decision variable of Kth user may now be
written as sum of desired signal component and 3 10-3
intereference /noise components [9].
10-4
Mf Mf Mf
æ ö Eb
rk = S wk,l rk,l = ± ç S wk,la ´k,l ÷ ¾ + S wk,l
l=1 èl=1 ø Mf l=1 10-5
0 5 10 15 20 25 80
(IJl + IMl+N) Average SNR (dB)
where Ijl is the Gaussian PBI present in lth subband Fig 5 Ber versus SNR of SC DS-CDMA for different m
with variance values
NJ 1 ¥
s2Jl º ¾ l = ¾ ò [Snj (f–fl) + Snj (f+fl)] X (f) df
2 2 –¥ MC DS-CDMA (for different m values)
100
The Kth user maximum conditional SNR is given by
–1 10-5
é K –1 a Nj ù
bl = ê1 + ¾ u¾ ¾ (1 + ¾ ) gl´ + ¾ i¾ ú
ë Mf PG 4 N0 û
Average BER (dB)
10-10
Nj JSRvgv
Where ¾ = ¾ ¾ ¾ ¾ ¾ v = 1,2,…..or Mf
N0 PG´ (1+a )
10-15
The Pb(E) of MC DS-CDMA over the MRC Fig 6 Ber versus SNR of MC DS-CDMA for different m
combiner is given by values
S ANURADHA et al : SC DS-CDMA AND MC DS-CDMA SYSTEMS 329
1984.
10-10 3. M S Alouini, M K Simon, & A J Goldsmith, A unified
performance analysis of DS-CDMA systems over
generalized frequency-selective fading channels,
Proc IEEE Int Symp Inf Theory (ISIT '98),
10-15
Cambridge, MA, August 1998, p 8.
4. G L Turin, Communication through noisy, random-
multipath channels, IRE Natl Conv Rec, March 1956,
10-20
0 5 10 15 20 25 80 pp 154-166.
Average SNR (dB) 5. H Suzuki, A statistical model for urban multipath
propagation, IEEE Trans Commun, vol COM-25,
Fig 9 Comparisions B/W SC & MC DS-CDMA for
July 1977, pp 673-680.
different m values
330 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
6. H Hashemi, Impulse response modeling of indoor 8. T Eng & L B Milstein, Coherent DS-CDMA
radio propagation channels, IEEE J Sel Areas performance in Nakagami multipath finding, IEEE
Commun, vol SAC-11, September 1993, pp 967-978. Trans Commun, vol COM-43, February-March-April
1995, pp 1134-1143.
7. S A Abbas & A U Sheikh, A geometric theory of
Nakagami mobile radio channel with physical 9. S Kondo & L B Milstein, Performance of Multi carrier
interpretations, Proc IEEE Veh Technol Conf DS CDMA Systems, IEEE Transactions on
(VTC'96), Atlanta, GA, April 1996, pp 637-641. Communications, vol 44, pp 238-246, Feb 1996.
Authors
* * * * * *
SHORT PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 331-334
The paper presents high speed realization of transposed direct form FIR filter core by
using sampler switch with its low resistive and capacitive components value that is to be
done in the core itself. It enables computationally efficient implementation of some of the
type of FIR filters by saving the time of the individual processing blocks in the path of the
final response due to these smaller value of components and also the computational time
can be saved to enhance speed by placing few blocks in parallel. The significant reduction of
resistance and capacitance of the sampler switch with their typical values enable to increase
speed by a large multiplying factor. The designed core will be suitable for the real time digital
signal processing applications.
x(n)
b0 b1 b2 bM–1 bM
y(n)
wM(n) w2(n) w1(n)
wM (n) = bM x (n) (2) in the form of the high speed core are connected by
the same timing and control unit as shown in Fig 2.
wk (n) = wk+1 (n–1) + bk x(n) k = M–1, M–2,....1 (3) First of all, the input data and various coefficients
corresponding to the particular input are stored in the
input data and constant coefficient memories, as shown
y (n) = w1 (n–1) + b0 x(n) (4)
in the Fig 2 along with the other individual components.
The timing and control unit will provide the simultaneous
3. ARCHITECTURE clock to these memories and both the data’s will be
available to the arithmetic logic unit, which will perform
The system architecture of FIR filter is shown in the multiply-shift-add operation as well as accumulator
Fig 2 for the designing of high speed core. Most of the to store the intermediate result for the further processing
individual blocks in the hardware of the designed filter through the next set of data’s.
x(t) X
Sample- x(n) Register x
and-hold ADC Input Data
circuit Memory Arithmatic Y y(n)
and Logic DAC LPF
Unit
Register b
h(t)
Sample-
b B Constant
and-hold ADC Coefficient Accumulator
circuit Memory
4. SIMULATION RESULTS Compared with the direct form structure, the direct
form structure will also obtain the same value of the
The final response of the system is expressed output as obtained for transposed form structure. As
mathematically in the following form. far as, the analysis contains the similar steps with the
same equation and difference observed only in the
M structure.
y (n) = S bk x (n–k) (5)
k=0
5. CONCLUSION
= b0 x(n) + b1 x(n–1) + ... + bM x (n–M)
In this paper, a high speed FIR filtering core for
First of all, all the data’s are stored in the two transposed form structure is presented. This design
memories i.e. input data memory contains the binary approach can also be applied for the implementation
information after conversion by sample-and-hold circuit of other configuration of digital FIR filter and that will
and ADC corresponding to the input signal in analog equally work well as far as high speed implementation
form represented by x(t), and, in the similar fashion, is concerned. The digital portion of design can be
the constant coefficient memory also contains the easily simulated by VHDL software and also it can be
same type of the information corresponding to the downloaded on the FPGA package to allocate the
coefficient values that are present in the h(t) [9]. various functioning of the pins for making core in the
Now, the various data’s are suitable for arithmetic and IC industry where micro chips are being fabricated.
logic circuit in the binary form. This circuit will perform
REFERENCES
all the computations i.e. multiplications, additions, and
shiftings operation. The intermediate results are stored 1. Ahmet Teyfik Erdogan & Tughrul Arslan, On the
in the accumulator, which will be required to obtain the Low-Power Implementation of FIR Filtering
final digital output values. After getting it, this calculated Structures on Single Multiplier DSPs, IEEE
value will pass through a DAC to convert into discrete Transactions on Circuits and Systems-II: Analog
analog form, which will be the desired output value and Digital Signal Processing, vol 49, no 3, March
2002.
[10].
2. A T Erdogan & T Arslan, Low power multiplication
Let the input data x(n) has the following discrete scheme for FIR filter implementation on single
values after converting x(t) by sample-and-hold circuit. multiplier CMOS DSP processors, Electronics
Letters, vol 32, no 21, 10th October, 1996.
x(n) = {0.1, 0.2, 0.3, 0.4} 3. V A Bartlett & E Grass, A Low power Asynchronous
VLSI FIR Filter, IEEE 2001.
and the coefficient corresponding to impulse response 4. A Chandrakasan, S Shang & R W Broderson , Low-
has the following values as per the input values Power CMOS Digital Design, IEEE Journal of Solid-
accordingly. State Circuits, vol SC-27, pp 473-484, 1992.
5. J Bhasker, VHDL Primer, Third Edition, Pearson
b = {0.01, 0.02, 0.03, 0.04} Education Asia, India.
These above data’s are taken arbitrarily for getting 6. E P Zwyssig, A T Erdogan & T Arslan, Low power
system on chip implementation scheme of digital
time output signal and the input data may be taken for
filtering cores, Low Power IC Design Seminar, 19
any real time signal processing application i.e. speech January 2001, London, UK.
signal, audio signal, vedio signal, any other 7. Artur Krukowski & Izzet Kale, Simulink/Matlab-to-
communication signal, and any biomedical signal etc. VHDL Route for Full-Custom/FPGA Rapid
Prototyping of DSP Algorithms, Matlab DSP
From equation (5) for M = 3, the output equation will Conference (DSP'99), Tampere, Finland, 16-17
be as below. November 1999.
8. John G Proakis & Dimitris G, Manolakis., Digital
y(n) = b0 x(n) + b1x (n–1) + b2 x (n–2)+b3x(n–3) (6)
Signal Processing Principles, Algorithms, and
The final discrete output obtained after the Applications, Third Edition, Prentice Hall of India
Private Limited, New Delhi- 110001, India, 2004.
processing is {0.001, 0.004, 0.010, 0.020}, which is
9. Willis J Tompkins, Biomedical Digital Signal
determined by the MATLAB Software using
Processing, Prentice-Hall of India Private Limited,
convolution to the maximum length of the one of the New Delhi, India, 2006.
signal between input signal and impulse response in 10. Simon Haykin, Communication System, Third
the discrete domain and the analog output y(t) will be Edition, Printed by Thomson Press (India) Ltd., John
reconstructed by the LPF. Wiley & Sons.
334 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
Author
M Arif born on 11th January, 1965, College Kurukshetra, Haryana, India as Lecturer in 1988, and
at Ghazipur, UP, India. He received BE presently serving in Electronics and Communication Engineering
in (Electronics Engineering) from Department. National Institute of Technology (Formerly Regional
Gorakhpur University and MTech in Engineering College) Kurukshetra, Haryana, as Lecturer in Senior
(Electronics and Communication Scale. His research interest is Digital Signal Processing. He is life
Engineering) from Kurukshetra member of ‘The Institution of Society for Technical Education’.
University, India in 1987 and 1997
respectively. He joined in Electronics
and Communication Engineering * * *
Department, Regional Engineering
SHORT PAPER
IETE Technical Review
Vol 24, No 4, July-August 2007, pp 335-337
Science and technology education have enjoyed a meaningful partnership across most
of this century. The work of scientists embraces an array of technologies, and major
happenings in science are often accompanied by complicated applications of technology. As
a result, a complete science education has involved a commitment to the inclusion of
technology, both as a tool for learning science content and processes and as a topic of
instruction in itself. Due to the fast development of information highway there is a remarkable
raise in the quality of education throughout the world. With the invention of new technologies
like high speed computer networks and multimedia computers, there is an increasing
awareness that direct face to face teaching is not the only possible mode of teaching in a
school system. There is a demand for preparing high quality multimedia course materials
across all faculties, which can be used by learners of the course who either cannot attend
the live lectures or prefer to study in an off-line mode. This paper discusses and suggests
some ideas and necessary changes in implementation of independent learning of Science
and Technology education through e-learning concepts in education which leads to improve
the education culture in rural area.
335
336 IETE TECHNICAL REVIEW, Vol 24, No 4, July-August 2007
needed in developing course structure. Before study. The web based support system helps for
preparing course structure it must be discussed with interaction and evaluation. The interaction is achieved
subject experts for standards. Good courseware must in the web by chatting or by mailing. Some times
give guidance to candidate and also provide a rich advanced technology of video conferencing may also
variety of background material when ever such is be utilized.
needed. The advent of digital libraries provides not
Evaluation and certification can be done by
only the most significant chances for computer-based
evaluating individual students by giving assignments.
training but also the extensive background reservoir of
Assignments can be submitted from home and discussed
material needed in many situations.
with the subject experts or with peers in an interactive
When designing the course structure in teaching distant mode.
science and Technology, the following three stages of
teaching structure must be adopted for better Cultural change
understanding of the subject
Though the students learn a subject through
• Presentation e-education the culture of a country should never be
• Practice affected. Every country esteems its own culture and
• Application so the fast development in education should go along
in a balanced way without neglecting its own culture.
Presentation is a one-way traffic. The main aim A country will look more gorgeous if it only shines
of presentation is to make clear-cut meaning of the with blooming flowers of cultural values mingled with
structures to the students. The presentation involves developed education. To succeed in modern society,
oral teaching and situation teaching. Structures are you have to know a lot. And you have to have this
first presented orally and then through reading and knowledge instantly accessible at your mental fingertips.
writing. Structures are taught in situations so that they
become real and meaningful for the students. Showing In the eyes of sociologists culture is the sum total
objects, pictures and charts, matchstick figures, actions of human experiences. It comprises of all human
and gestures and verbal situations creates different traits, determined by a cluster of values. These cultural
situations. values are manifested in our traditions, customs,
ideologies, and even in day-to-day life. Each group of
The importance of the second stage practicing is people beholds different but certain values and attitudes.
to reinforce the structure in the minds of the students. A change in attitude alone can bring about a change in
The importance of third stage application is to ask the acquisition of cultural values. Languages on its
the students to apply the structure in new situations. part are more helpful for the development of culture.
Every language is a product of its own culture. All the
The CD-ROM materials are prepared, with the chunks and expressions we come across help us to
help of multi-media technology by subject experts so understand varied concepts. It is none other than a
that it will motivate their learning interest. The materials language, which assists each and every one to think
are prepared by combining short phases of presence and speak in terms of concepts. It is the language that
with interleaved media-based self-study blocks; they has a great control over the thought process of an
are well suited for web-based education. An innovative individual. The younger generation is blessed with the
approach is taken by integrating all course elements- cultural elements through the decorative pipes of the
acoustic and written information, diagrams and figures, language.
animations, simulations, video clips, and laboratory
exercises by presenting in a CDROM materials. The The disadvantages of class room teaching are
material should include animation system. Animation • The classroom pace is moving too slowly for
is a best tool for teaching dynamic phenomena. the gifted child.
Animations are used so that they will promote interest • The classroom pace is moving too fast—the
in further studying. child needs 55 minutes to master a concept
but the bell always rings at 50.
The web based education support system and • The child feels left out or bullied.
CD-ROM materials are very useful to improve
candidate’s knowledge and skills, regardless of prior • The child can’t sit still in class or has a learning
knowledge and skills. CD-ROM materials significantly challenge.
enhance the effectiveness of teachers creating materials, • The child falls outside of the middle range in
especially when the CD-ROM was used for self- the classroom.
S A PERUMAL : INTERACTIVE EDUCATION THROUGH ECT 337
• The class is overcrowded. First, digital technologies are changing the ways
• The child is homebound or traveling for sports teachers interact with students in the classroom.
Second, teacher education courses are not only
Factors to be concentrated on to improve influenced by new K-12 curricula, they are also
the quality of e-learning influenced by instructional approaches that
incorporate a variety of digital technologies.
1. Technology should be introduced in the context Technological applications go beyond K-12 curriculum
of science content. to the delivery of college level content. For instance,
2. Technology should address worthwhile science faculty and students explore web resources for
with appropriate pedagogy. educational statistics or education-related reports and
3. Technology instruction in science should take course resources.
advantage of the unique features of technology.
Third, faculty and students alike are interacting in
4. Technology should make scientific views more new ways afforded by digital technologies. Faculty
accessible. and students have virtual discussions related to course
5. Technology instruction should develop content, advice, and counseling in a wide variety of
students’ understanding of the relationship times and places via email, cell phones, pagers, and
between technology and science. features of the web. Faculty and students now produce
documents with more information and in far more
CONCLUSION diverse formats as a result of desktop publishing,
online libraries and databases, and file transfer
The impact of digital technologies on science capabilities. The pervasiveness of digital technologies
teacher education is more pervasive than any curricular motivates a thorough review of technological impacts
or instructional innovation in the past. The impact can on curriculum and instruction in science teacher
be felt on three fronts. education.
Author
Microwave technology has rich history dating back to the 19th Century and fundamental discoveries in
electromagnetic with a very strong and elegant foundation established by Maxwell with his famous equations. The
thrust of the research supporting microwave technology has been in establishing the basic elements to support the
growth and development of the application and also to extend the usable spectrum to higher and higher frequencies.
Although major thrust and applications of Microwave Circuits are for the Defense field, the growth of Satellite and
Wireless Communication in last few decades has proved the utility of this band for civilian applications also. The
purpose of this special issue on ‘Microwave Circuits and Systems’ is to exhibit the state-of-the-art research in our
country in this area and also to give future direction for research to young scientists and engineers.
• RF Circuit Design
• CAD of Microwave Circuits
• EMI/EMC
• Microwave Integrated Circuits
• Satellite Communication Systems
• Passive Microwave Devices
• Radar Components
• Super Components for EW systems
• Feed Networks for Phased Arrays
• TR/ATR Modules
• Air Borne/Space Borne Microwave Circuits and Systems
• MEMS
• Circuit Synthesis and Simulation
• Microwave Measurement techniques
Conceptual review papers covering the theme with less mathematics are invited. Research papers with original contribution
in any of the above area with more stress on ‘Results and Discussion’ are most welcome. Articles of about 4000 to 5000
words including figures, graphs, tables and references typed neatly in A4 size paper in double space as per IETE
guidelines to author (given in website www.iete.org under the heading publication) should be sent to the Guest Editor at
the address given below. Two hard copies along with a soft copy in MS word should be sent to:
Prof V M Pandharipande,
Guest Editor, IETE Journal of Research,
Director,
Centre for Excellence in Microwave Engineering,
Department of Electronics & Communication Engineering,
Osmania University, Hyderabad - 500 007 (AP).
Email: ceme_uceou@yahoo.com;
Ph:040-27071273,27682234
Last Date for receiving the papers: 30th October 2007; Papers to be reviewed: 31st December 2007;
Issue to be published: March 2008.
Printed and Published by Brig (Retd) V K Panday for the Institution of Electronics and Telecommunication Engineers,
2, Institutional Area, Lodi Road, New Delhi 110 003 (India) and printed at the Shivam Offset Press, A-12/1, Naraina Indl. Area,
Phase I, New Delhi-110 028 (India). Dy Managing Editor : CDR A P Sharma, IN (Retd) • Copyright © 2007 by the IETE