You are on page 1of 17

Computer Physics Communications 182 (2011) 2638–2654

Contents lists available at ScienceDirect

Computer Physics Communications


www.elsevier.com/locate/cpc

PETOOL: MATLAB-based one-way and two-way split-step parabolic equation tool


for radiowave propagation over variable terrain ✩
Ozlem Ozgun a,∗ , Gökhan Apaydin b , Mustafa Kuzuoglu c , Levent Sevgi d
a
Department of Electrical and Electronics Engineering, Middle East Technical University, Northern Cyprus Campus, Guzelyurt, Mersin 10, Turkey
b
Department of Electrical and Electronics Engineering, Zirve University, Gaziantep, Turkey
c
Department of Electrical and Electronics Engineering, Middle East Technical University, 06531 Ankara, Turkey
d
Department of Electronics and Communications Engineering, Dogus University, Istanbul, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: A MATLAB-based one-way and two-way split-step parabolic equation software tool (PETOOL) has been
Received 1 April 2011 developed with a user-friendly graphical user interface (GUI) for the analysis and visualization of radio-
Received in revised form 5 July 2011 wave propagation over variable terrain and through homogeneous and inhomogeneous atmosphere.
Accepted 9 July 2011
The tool has a unique feature over existing one-way parabolic equation (PE)-based codes, because it
Available online 27 July 2011
utilizes the two-way split-step parabolic equation (SSPE) approach with wide-angle propagator, which is a
Keywords: recursive forward–backward algorithm to incorporate both forward and backward waves into the solution
PETOOL in the presence of variable terrain. First, the formulation of the classical one-way SSPE and the relatively-
Electromagnetic propagation novel two-way SSPE is presented, with particular emphasis on their capabilities and the limitations. Next,
Refractivity the structure and the GUI capabilities of the PETOOL software tool are discussed in detail. The calibration
Ducting of PETOOL is performed and demonstrated via analytical comparisons and/or representative canonical
Split-step parabolic equation tests performed against the Geometric Optic (GO) + Uniform Theory of Diffraction (UTD). The tool can
Terrain factors be used for research and/or educational purposes to investigate the effects of a variety of user-defined
Multipath effects
terrain and range-dependent refractivity profiles in electromagnetic wave propagation.
Validation, verification and calibration
MATLAB program
Program summary

Program title: PETOOL (Parabolic Equation Toolbox)


Catalogue identifier: AEJS_v1_0
Program summary URL: http://cpc.cs.qub.ac.uk/summaries/AEJS_v1_0.html
Program obtainable from: CPC Program Library, Queen’s University, Belfast, N. Ireland
Licensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.html
No. of lines in distributed program, including test data, etc.: 143 349
No. of bytes in distributed program, including test data, etc.: 23 280 251
Distribution format: tar.gz
Programming language: MATLAB (MathWorks Inc.) 2010a. Partial Differential Toolbox and Curve Fitting
Toolbox required
Computer: PC
Operating system: Windows XP and Vista
Classification: 10
Nature of problem: Simulation of radio-wave propagation over variable terrain on the Earth’s surface, and
through homogeneous and inhomogeneous atmosphere.
Solution method: The program implements one-way and two-way Split-Step Parabolic Equation (SSPE)
algorithm, with wide-angle propagator. The SSPE is, in general, an initial-value problem starting from a
reference range (typically from an antenna), and marching out in range by obtaining the field along the
vertical direction at each range step, through the use of step-by-step Fourier transformations. The two-
way algorithm incorporates the backward-propagating waves into the standard one-way SSPE by utilizing
an iterative forward–backward scheme for modeling multipath effects over a staircase-approximated
terrain.


This paper and its associated computer program are available via the Computer Physics Communications homepage on ScienceDirect (http://www.sciencedirect.com/
science/journal/00104655).
*
Corresponding author.
E-mail address: ozozgun@metu.edu.tr (O. Ozgun).

0010-4655/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cpc.2011.07.017
O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2639

Unusual features: This is the first software package implementing a recursive forward–backward SSPE
algorithm to account for the multipath effects during radio-wave propagation, and enabling the user to
easily analyze and visualize the results of the two-way propagation with GUI capabilities.
Running time: Problem dependent. Typically, it is about 1.5 ms (for conducting ground) and 4 ms (for
lossy ground) per range step for a vertical field profile of vector length 1500, on Intel Core 2 Duo 1.6 GHz
with 2 GB RAM under Windows Vista.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction Apart from the Fourier split-step algorithm, the solution of the
PE has also been achieved by the finite difference (FD)-based [15–
Radio-wave propagation over the Earth’s surface and in an in- 17] and finite element (FE)-based [18–22] algorithms. The Fourier
homogeneous atmosphere is affected by several scattering phe- split-step algorithm is more robust, since it provides the use of
nomena, such as reflection, refraction, and diffraction. Understand- larger range increments and a faster solution for long-range propa-
ing the effects of varying conditions on radio-wave propagation is, gation scenarios. Apart from these studies, there exist several com-
therefore, essential for designing reliable radar and communication puter software programs, most of which have been developed for
systems. Especially, tropospheric waves may play a dominant role military purposes, for predicting radar coverage under the effect of
in communications because they can propagate over the horizon environmental factors that influence refractivity. These are IREPS
and increase the coverage area, and hence, they may disrupt the (Integrated Refraction Effects Prediction System), EREPS (Engineer’s
communication links due to the interference that is not normally Refractive Effects Prediction System), TESS (Tactical Electronic Sup-
there. Such waves are propagated by bending or refraction due to port System), AREPS (Advanced Refractive Effects Prediction Sys-
the abrupt change in the refractive index in the troposphere, and tem), TEMPER (Tropospheric Electromagnetic Parabolic Equation
cause so called ‘anomalous’ propagation. If the refractive gradient Routine), and TPEM (Terrain Parabolic Equation Model). These pro-
exceeds some certain limits, the radio waves may be trapped in a grams implement the ray optics or one-way PE techniques, or a
‘duct’ and guided over distances far greater than the normal range. hybrid model combining these methods.
In addition to tropospheric effects, irregular terrain surfaces have In this paper, a novel software tool (PETOOL), which is devel-
considerable influence on radio-wave propagation because they re- oped in MATLAB® with graphical user interface (GUI), is intro-
flect and diffract the electromagnetic waves in a complex way. duced for the analysis and visualization of radio-wave propagation
Hence, the design of an effective radar or communication system through homogeneous and inhomogeneous atmosphere, by incor-
can be achieved by using a model that can properly incorporate porating variable terrain effects with the aid of the two-way split-
the refractivity and terrain factors. step algorithm employing wide-angle propagator. Why yet develop
The rigorous analytical and numerical modeling of radio-wave another PE-based program for radio-wave propagation? The reason
propagation in such environments is a challenging task and has is twofold: The first is that PETOOL is a free/open-source pro-
attracted the attention of researchers for many decades. The dif- gram, and has been designed with a user-friendly GUI, to serve
ficulty stems from the vast variability of the properties of the as a research/educational tool for propagation engineers/instructors
medium and also the surfaces and obstacles that re-direct the to investigate the phenomenon in an illustrative manner, and/or
propagating energy, making the radio propagation somewhat un- to achieve the analysis/design/planning of reliable communication
predictable. Initially, analytical techniques (such as ray tracing systems. It displays the propagation factor/loss on a range/altitude
methods, diffraction methods, and waveguide mode theory) have scale, and enables the user to easily visualize, enter and save all
been employed to predict the radio propagation [1–6]. However, related input/output parameters. More importantly, the user can
they require the geometry to be represented as a member of a easily create arbitrary terrain and refractivity profiles. The sec-
set of some canonical geometries and suffer from the presence of ond reason is that PETOOL is indeed the first software package
the vertically-varying refractivity profile in the troposphere. With implementing both one- and two-way SSPE algorithms, the lat-
the advances in computers, some numerical techniques have been ter of which incorporates the multipath effects into the PE solu-
devised to easily handle the above-mentioned difficulties. Parabolic tion of the radio-wave propagation through a recursive forward–
Equation (PE) model has been widely used in propagation modeling backward algorithm. The standard one-way PE method, in spite of
to predict the wave behavior between a transmitter and a receiver its wide-spread usage, suffers from two major drawbacks: (i) The
over the two-dimensional (2D) Earth’s surface, because of its high PE method handles only the forward-propagating waves, and ne-
capability in modeling both horizontally- and vertically-varying glects the backscattered waves. The forward waves provide almost
atmospheric refraction (especially ducting) effects. The standard accurate results for typical long-range propagation problems, only
PE is derived from Helmholtz’s equation in such a way that the if there does not exist obstacles that redirect the incoming wave
rapidly-varying phase term is discarded to obtain a reduced func- in the form of reflections and diffractions. However, the accurate
tion having slow variation in range for propagating angles close estimation of the multipath effects, occurring during propagation
to the paraxial direction. Helmholtz’s equation is approximated by over terrain, requires the correct treatment of backward waves as
two differential equations, corresponding to forward and backward well. Moreover, the PE method takes the diffraction effects into
propagating waves, each of which is in the form of a parabolic par- account within the paraxial approximation, degrading the accuracy
tial differential equation. The standard PE method takes only the of the approach in deep-shadow regions where the diffracted fields
forward part into account, namely, it is a one-way, forward scatter dominate. There are a number of studies in the literature to over-
model, valid in the paraxial region. Although the initial introduc- come such difficulties [23–30]. Recently, a two-way SSPE algorithm
tion of the PE method has been credited to [7], its wide-spread has been proposed by the authors to incorporate the backward-
usage has become possible after the development of Fourier split- propagating waves into the standard one-way SSPE by utilizing
step algorithm by [8]. The split-step based PE method (SSPE) is, in an iterative forward–backward scheme for modeling multipath ef-
general, an initial-value problem starting from a reference range fects over a staircase-approximated terrain [31–33]. (ii) The second
(typically from an antenna), and marching out in range by ob- drawback is that the standard PE is a narrow-angle approximation,
taining the field along the vertical direction at each range step, which consequently restricts the accuracy to propagation angles
through the use of step-by-step Fourier transformations [9–14]. up to 10◦ –15◦ from the paraxial direction. A typical long-range
2640 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

propagation encounters propagation angles that are usually less The differential operator in (2) can be factored in terms of two
than a few degrees, whereas the short range propagation prob- pseudo-differential operators and put into forward and backward
lems, as well as the problems involving multiple reflections and propagating waves, as follows:
diffractions because of hills and valleys with steep slopes, can only 
∂u −ik(1 − Q )u forward,
be solved by a PE model that is effective for larger propagation = (3)
angles. To handle propagation angles up to 40◦ –45◦ , wide-angle ∂x −ik(1 + Q )u backward,
propagators have been introduced [34–39]. PETOOL implements where
the two-way algorithm proposed in [31], with the exception that  1 ∂2  
narrow-angle propagators are replaced by wide-angle propagators Q = 1+q and q = + n2 − 1 . (4)
to handle larger propagation angles. k2 ∂ z2
The validation, verification, and calibration (VV&C) process of The formal solution of the forward propagation part in (3) can be
PETOOL is also demonstrated in this paper through several numer- expressed as
ical tests and comparisons realized in some canonical and more 
complex scenarios. It is known that the VV&C process is one of u (x + x, z) = exp −ikx(1 − Q ) u (x, z), (5)
the key issues for electromagnetic simulation codes because it is which is amenable to numerical solution by marching-type algo-
necessary for the user to specify and master the validity domain rithms along range. Note that backward propagating waves in (3)
of the simulation. Validation process is the process of determin- are omitted in the standard PE. The operator Q
ing whether the right model is built (or solving the right equations).
√ is approximated
by using the first-order Taylor expansion (i.e., 1 + q ≈ 1 + q/2),
Verification assessment examines if the model is built right (or solv- yielding the standard PE as follows [12]:
ing the equations right). The verification of a code usually involves  
the error analysis, which intends to search for bugs, incorrect im- ∂2 ∂ 
2 2

+ 2ik + k n − 1 u (x, z) = 0. (6)
plementations, errors in the inputs or in other parts of the code, ∂ z2 ∂x
accuracy in the calculations, etc. Finally, the calibration process
The accuracy of the standard PE is limited to propagation angles
is “the process of adjusting numerical or physical modeling parame-
less than 10◦ –15◦ , and the error in the approximation increases
ters in the computational model for the purpose of improving agreement
with sin4 θ , where θ is the propagation angle from the horizon-
with experimental data (AIAA G-077-1998)”. The calibration is usu-
tal, due to the first neglected term in Taylor’s expansion. Hence,
ally done against exact solutions and/or other numerical methods,
the standard PE is known as the “narrow-angle” approximation to
and can occur as a part of either validation or verification. The cal-
the wave equation. As the propagation angles encountered in long-
ibration results of PETOOL are checked against the analytical data,
range propagation problems are usually less than a few degrees,
the results of existing software (AREPS), as well as the Geometric
the accuracy of the standard PE is adequate for numerical model-
Optic (GO) + Uniform Theory of Diffraction (UTD). The 3D image
ing.
visualization of the GO + UTD fields in a propagation scenario in-
In problems involving large propagation angles (such as short
volving multiple knife-edges is introduced and compared with the
range propagation problems or the problems exhibiting strong
PE method.
multipath effects), a more accurate expansion of the operator Q
This paper is organized as follows: Section 2 presents the brief
is needed. In such cases, the use of higher-order polynomials for
theoretical formulation of the one- and two-way SSPE methods, for
the operator causes instabilities in the numerical results. There are
the sake of completeness. Section 3 discusses the structure and the
various convenient methods proposed in the literature to handle
GUI capabilities of PETOOL. Section 4 illustrates the VV&C tests via
large propagation angles, such as Claerbout equation [39] and split-
several comparisons. Finally, Section 5 draws some conclusions.
step Padé parabolic equation method [35], both of which depend
Throughout the paper, the suppressed time-dependence of the
on the first-order Padé approximation. A very efficient approach,
form exp(−i ωt ) is assumed.
providing a good approximation to the wave equation for large
angles, has been proposed in [36], and re-visited√ in [37]. In this
2. Formulation of one-way and two-way SSPE methods
approach, the operator √Q is re-written
√ as Q = 1 + A + B, and is
approximated as Q ≈ 1 + A + 1 + B − 1 where A =√∂ 2 /k2 ∂ z2
2.1. One-way SSPE
and B =√ n − 1. By −making
2
use of the operator identity 1 + A =
The problem of our interest is the 2D electromagnetic problem 1 + A [ 1 + A + 1] 1 , the wide-angle parabolic equation is given
bounded by the ground/sea at the bottom, and unbounded at the by
top, i.e. extending to infinity. The problem is governed by the scalar ∂ u  √ − 1
Helmholtz equation expressed as follows: = ik A 1 + A + 1 + ik(n − 1) u . (7)
∂x
∂ 2ϕ ∂ 2ϕ In this paper, the wide-angle PE in (7) is preferred and imple-
+ 2 + k 2 n 2 ϕ = 0, (1) mented because it can be solved by the Fourier split-step algo-
∂ x2 ∂z
rithm as if it is a narrow-angle PE, which will be clear in the
where k = 2π /λ is the free-space wavenumber (λ is the wave-
sequel.
length), n(x, z) is the refractive index, and ϕ denotes the electric
The numerical solution of the PE is achieved by the Fourier
or magnetic field in horizontal or vertical polarization, respectively.
split-step method, which is a widely-used and robust algorithm.
Here, x and z represent range and altitude coordinates, respec-
The algorithm starts at a reference range (usually at an antenna),
tively.
and marches the solution in range in a way such that it obtains the
The PE is derived from Helmholtz equation by separating the
vertical field profile at a given range by using the field at the pre-
rapidly varying phase term to obtain a reduced (amplitude) func-
vious range, with appropriate boundary conditions at the top and
tion that varies slowly in range for propagating angles close to
bottom boundaries of the domain (see Fig. 1). The split-step solu-
the paraxial direction (horizontal x-direction in this paper). By in-
tion of the narrow-angle parabolic equation in (6) is given by [12]
troducing the reduced function as u (x, z) = exp(−ikx)ϕ (x, z), the
wave equation in terms of u is given as follows [12]: u (x + x, z)
      
∂2
∂ ∂ 2      x −1 x

+ 2ik + 2 + k n2 − 1 u (x, z) = 0.
2
(2) = exp ik n2 − 1 F exp −ip 2 F u (x, z) , (8)
∂ x2 ∂x ∂z 2 2k
O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2641

Fig. 1. One-way SSPE framework with forward propagating waves. (White nodes represent zero fields.)

where F indicates the Fourier transform, and p = k sin θ is the where α1 and α2 are constants. The cases where α1 = 0 and
transform variable (θ is the propagation angle from the horizon- α2 = 0 refer to Dirichlet (horizontal polarization) and Neumann
tal). The wide-angle split-step solution of (7) is computed by [37] (vertical polarization) BCs, respectively, over the perfectly conduct-
 ing (PEC) surface. For lossy ground surface, the Cauchy-type BC
u (x + x, z) = exp ik(n − 1)x is defined by setting α1 = 1, α2 = ik(εr + i60σ λ)1/2 and α1 = 1,
  − 1  α2 = ik(εr + i60σ λ)−1/2 for the horizontal and vertical polariza-
−1 ip 2 x p2 tions, respectively. Here, σ and εr are the conductivity and the
×F exp − 1− +1
k k2 relative permittivity of the Earth’s surface, respectively. The stan-
 dard split-step PE method cannot handle the BCs automatically.

The ground losses can be incorporated into the standard split-step
× F u (x, z) . (9)
PE approach through the use of mixed Fourier transform [11]. To
satisfy the BCs over PEC ground, the upper boundary is extended
In radio-wave applications two parameters are of interest: prop- from [0, zmax ] to [− zmax , zmax ], and then, in accordance with the
agation factor and propagation loss (usually called path loss). After image theory, the odd and even symmetric field profiles of u are
calculating the field inside the entire computational domain via constructed for Dirichlet and Neumann BCs, respectively, to be able
(8) or (9), the propagation factor (field strength relative to its free- to apply the FFT. Another option, to avoid the height extension,
space value in dB) is computed by is to reduce the Fourier transform to one-sided discrete sine or
cosine transforms (DST or DCT), for Dirichlet and Neumann BCs,
PF = 20 log |u | + 10 log x + 10 log λ. (10)
respectively.
The path loss, which is the ratio between the power radiated by (ii) As obvious from (8) and (9), the split-step algorithm oper-
the transmitter antenna and the power available at a point in ates between z and p domains (namely, Fourier transform pairs)
space, can be determined by in a consecutive manner. In the numerical implementation, the
domains are truncated at zmax and p max . The altitude and range
PL = −20 log |u | + 20 log(4π ) + 10 log x − 30 log λ. (11) step sizes  z and x, respectively, and the maximum altitude
zmax are determined according to the source/observation require-
In the numerical realization of the split-step algorithm, there
ments, as well as the sampling criterion to avoid aliasing effects.
are certain important issues that require special and careful treat-
Once zmax is decided, p max is obtained from the Nyquist criterion
ment. These issues can be classified as (i) imposition of the bound-
zmax × p max = π N where N is the Fourier transform size. Note that
ary conditions on the upper and lower terminations, (ii) discretiza-
p max = k sin θmax , where θmax is the maximum allowable propa-
tion (the determination of the altitude and range increments,
gation angle. Since  z = zmax / N, the altitude increment should
 z and x, respectively), (iii) definition of the initial vertical pro-
satisfy  z  λ/(2 sin θmax ). Although the choice of  z is quite crit-
file (source imposition), (iv) handling terrain factors, and (v) han-
ical in simulations, the selection of the range increment x (taking
dling atmospheric refractivity variations. The subsequent para-
the refractivity gradients into account) is chosen by the user, and
graphs address these issues, which are the essential components
can be much larger than the wavelength.
of the validation, verification and calibration process.
(iii) The initial vertical field at the starting range position (usu-
(i) The propagation problem under consideration involves a ver-
ally x = 0) must be properly determined in accordance with the
tically open region, and therefore, the condition u (x, z)|z→∞ = 0
parameters of the antenna pattern being modeled. The initial field
must be satisfied. Since the Fourier transform is realized by dis-
can be computed by means of near-field/far-field transformation
cretization (such as by Fast Fourier Transform, FFT), the field is
that relates the aperture field and beam pattern, along with the
truncated abruptly along z, yielding non-physical reflections from
utilization of fast Fourier transform. The height and elevation an-
the upper boundary. Such artificial reflections can be removed
gle of the antenna can be included by using Fourier shift theorems.
by extending the maximum height, and by decaying the field
The antenna pattern is specified by three parameters: height za ,
smoothly in this extended region. This can be achieved by using
the 3 dB beamwidth θ B W , and the tilt (or elevation) angle θtilt . The
absorbing regions or by applying windowing functions (such as
first step is to specify the initial field in the p domain via
Hanning, Hamming), or by adding a small imaginary part to the
refractive index (by making n complex).
The boundary condition (BC) that must be satisfied over the U (0, p ) = f ( p ) exp(−ipza ) − f ∗ (− p ) exp(ipza ). (13)
Earth’s surface is expressed as
The initial field in spatial z-domain is found by taking the In-
 
∂ verse Fast Fourier Transform (IFFT) of (13). The Gaussian antenna
α1 + α2 u (x, z) = 0, (12) pattern is often used in applications since it represents vari-
∂z
2642 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

ous antenna types (such as parabolic antennas). The horizontally- 2.2. Two-way SSPE
polarized Gaussian antenna pattern can be defined as
√ As mentioned in Section 2.1, the standard PE method is a one-
 2 ln 2 way, forward propagation model, because of ignoring the backward
f ( p ) = exp − p 2 w 2 /4 where w = . (14) propagation term in (3). The standard PE model cannot reflect the
k sin(θ B W /2)
effect of the interaction between the forward and backward waves,
The tilt angle is introduced by shifting the antenna pattern, i.e., especially if there are valleys or hills with steep slopes along the
f ( p ) → f ( p − k sin θtilt ). propagation path. In the presence of such obstacles, the effects
(iv) Terrain factors can be incorporated into the split-step PE of not only forward but also backward reflected, refracted, and
algorithm with different approaches. It is worthwhile mention- diffracted waves must be very well predicted to be able to get re-
ing that, the majority of the PE approaches in the literature in liable results.
modeling radio-wave propagation over a variable terrain, charac- The two-way SSPE algorithm is basically the iterative imple-
terizes only forward-propagating waves, and neglects backward mentation of the one-way SSPE by simply switching the direction
ones. Although there are algorithms considering the terrain diffrac- of propagation back-and-forth to estimate the multiple-reflection
tion, they still omit the backward waves. Recently, the two-way effects [31–33]. As illustrated in Fig. 2, the algorithm can be ap-
algorithm has been proposed to model multipath effects by mod- plied to a variable terrain by using staircase approximations. If the
eling an arbitrary terrain by staircase approximations [31] (see vertical field meets an obstacle, it is split into two components
Section 2.2). In the context of forward propagation, the simplest propagating in forward and backward directions. The forward field
and the effective approach is the staircase terrain modeling. In this profile continues in the usual way after setting it to zero on the
approach, on each segment of the constant altitude, the vertical vertical terrain facet. During forward propagation, the wide-angle
field profile is calculated in the usual way, applying the desired split-step solution in (9) is employed to march the solution. How-
boundary conditions on the ground surface, and then, is simply set ever, it is evident that, with regard to the physics of the problem,
to zero on the vertical terrain facets. This approach provides sat- the field must be partially-reflected from the terrain facet. This is
isfactory results in the approximate sense because the boundary achieved in the two-way algorithm in such a way that, first the ini-
conditions on sloping facets are not properly taken into account, tial field of the backward field is obtained by imposing the bound-
and also, the corner diffraction is ignored. There are more accurate ary conditions at the facet (e.g., the total tangential field must be
approaches in the literature to modeling of sloping facets by repre- zero on the PEC facet), and then this initial field is marched back in
senting the terrain as a sequence of piecewise linear functions, or the −x-direction by reversing the signs of k and x in (9). Hence,
by employing coordinate transformations [23–25], which achieve the backward vertical field profile is found at range x − x as fol-
conformal mappings over terrain curvature. lows:
(v) The lower troposphere affects radio-wave propagation in 
u b (x − x, z) = exp ik(n − 1)x
numerous ways. Especially, its non-uniform nature causes electro-
  − 1 
magnetic waves to be bent or refracted. Refraction of electromag- ip 2 x p2
−1
netic waves is due to the variation of the velocity of propagation ×F exp − 1− +1
k k2
with altitude. It is known that index of refraction (n) is the ra-

tio of the velocity in free-space to the velocity in the medium of

interest (atmosphere in our case); and is caused by pressure, tem- × F u (x, z) . (16)
perature and water vapor variations in both space and time. It is
conventional to define a new quantity, so called modified refractiv- Note that the same form of the equation is derived for the reduced
ity, which takes the Earth’s curvature into account, as follows: function in backward propagation (as expected), but the original
  field is expressed as ϕb (x, z) = u b (x, z) exp(−ikx).
M = n2 − 1 + 2z/ae × 106 (M-units), (15) In this way, both forward and backward fields continue to
march out in their own paraxial directions. At each time the wave
where ae is the Earth’s radius and z is the height above sur- hits a terrain facet, the field is again separated into forward and
face. Note that 2z/ae corresponds to the Earth’s curvature. The backward components. The total field is then obtained by super-
variations of the vertical gradient dM /dz of the modified refrac- posing the backward- and forward-fields at each range step. Bear-
tivity determine four types of atmospheric conditions: subrefrac- ing in mind that the field contributions of the multiple reflections
tion (dM /dz > 118 M-units/km), standard (dM /dz = 118 M-units/ √
decrease as the iterative calculations are performed (due to 1/ r
km), superrefraction (dM /dz < 118 M-units/km), and ducting term in 2D Green’s function), the convergence of the algorithm is
(dM /dz < 0). Non-standard atmospheric conditions cause anoma- achieved and checked against a certain threshold criterion com-
lous propagation because rays bend upwards in subrefraction, and paring the total fields at every iteration (i.e., un − un−1 /un−1 
downwards to the Earth’s surface in superrefraction and ducting where un and un−1 are the superposed fields at the nth and
conditions, in a way different from the standard atmosphere. Espe- (n − 1)st steps). The default value of the threshold is set to 0.025 in
cially, atmospheric ducts (i.e., wave trapping layers) are of special PETOOL, but is automatically decreased unless the maximum range
interest, because the negative vertical gradient leads to the capture is arrived. It is useful to note that, in the implementation of the it-
of energy within the duct, and the trapped energy can propagate erative two-way algorithm, the computation time and the memory
to ranges beyond the normal horizon what would be expected requirements may increase depending on the number of forward–
with a standard atmosphere. Such conditions significantly affect backward propagating waves that must be handled at once. SSPE is
the radio communication links and radar performance. There are ideally capable of handling infinite number of propagating waves,
basically four types of ducting conditions: surface duct, surface- of course limited by the available memory, and therefore, it can
based duct, elevated duct and evaporation duct (which is indeed be used for any type of terrain in a flexible way. However, the
a type of surface duct occurring over water due to water vapor bookkeeping process might place burden on the computational
evaporated from the sea) (see Fig. 5 in Section 3). These are the resources, but can be eliminated by utilizing parallel processing
consequences of several meteorological conditions, which are be- techniques. The parallel version of the software is currently being
yond the scope of this study. For detailed discussion, the readers developed, and will be available as a freeware program in the fu-
are referred to [40]. ture.
O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2643

Fig. 2. Recursive implementation of two-way SSPE with forward (F)–backward (B) propagating waves for staircase-approximated terrain.

3. PETOOL software The main m-file to run the program is petool.m. An overview of
the program structure, as a flowchart, is shown in Fig. 3. The main
window of the program is depicted in Fig. 4. The window is basi-
PETOOL software package has been developed in MATLAB® ver-
cally divided into four panels. The left and right panels are located
sion R2010a with a user-friendly GUI for the analysis and visual- on a large gray background, whereas the top and bottom panels
ization of radio-wave propagation. The GUI has been designed so are located on thin blue backgrounds. These panels are defined in
as to meet the following goals: detail below.
The top blue panel is reserved for five operational pushbuttons
• The user should be able to easily visualize, load and save the (load, save, exit, run, about). The load and save buttons are used for
propagation factor/loss on a range/altitude scale in radio-wave all input parameters of the simulation. Once clicked, a modal dialog
propagation problems over variable terrain and through homo- box is opened to select or specify a file the user wants to create
geneous and inhomogeneous atmosphere. or save. While exiting PETOOL, the user is also warned whether or
• The user should be able to define her/his own input param- not s/he wants to save the parameters.
eters, and to load/save them if desired. The user should be The bottom blue panel is used to show warning text messages
warned if s/he enters inappropriate input values. whenever needed, especially in case of inappropriate input en-
• The user should be able to easily define an arbitrarily-shaped trance.
terrain profile by just locating a number of points on the On the left panel, there are six sub-panels (domain, analysis, an-
graph through left-clicking the mouse. The user should be tenna, surface, atmosphere, terrain), where the input parameters
able to load/save the terrain parameters from/in a user-defined are defined by the user. The input parameters are summarized in
file. Table 1.
• The user should be able to easily specify range-dependent or
independent refractivity profiles by just selecting from a list of • In specifying a refractivity profile, the user selects an atmo-
various types of atmosphere profiles. sphere type from a menu list, and a new window is opened
2644 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

Fig. 3. Flowchart of the program structure.


O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2645

Fig. 4. PETOOL main window and 2D graphics windows obtained by right-clicking the mouse on the 3D map. (For interpretation of the references to color in this figure, the
reader is referred to the web version of this article.)

accordingly, enabling the user to enter the modified refractiv- • In specifying a terrain profile, the user has three options:
ity ( M ) values for the specified atmosphere type. The avail- (i) s/he can locate a number of points on the top graphics
able atmosphere profiles are standard atmosphere, surface of the right panel by clicking “Locate Points” button, (ii) s/he
duct, surface-based duct, elevated duct, evaporation duct, and can define the terrain points manually by entering the val-
user-defined duct, whose windows are illustrated in Fig. 5. If ues into the range-height list boxes, or (iii) s/he can load a
the range-dependent refractivity profile is to be defined, the user-generated text file including the terrain parameters. In
above-mentioned selection is performed for each range value. all cases, the user can save/load/clear/plot the terrain profile.
The profiles lying between two consecutive range values are If the user prefers to create her/his own terrain by locat-
computed automatically through linear interpolation. It is use- ing points on the graphics, the values of the selected points
ful to note that the user can load/save the atmosphere pa- are automatically placed into the range and height list boxes.
rameters separately. In addition, the user can easily modify or Hence, it is possible to store the graphically-generated terrain
delete the parameters in the profile and range lists, by means profiles in files, as well as to modify or delete the parameters
of a special dialog box that is opened when the user clicks an in the list boxes. Once the terrain points are specified by using
item from the list. one of the above-mentioned ways, the overall terrain profile is
2646 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

Fig. 5. Windows for specifying atmosphere types: (a) standard atmosphere, (b) surface duct, (c) surface-based duct, (d) elevated duct, (e) evaporation duct, (f) user-defined
duct.

created by performing a kind of interpolation (linear or cubic- values (range, height, PF/PL) automatically appear in the ‘current
spline) between two consecutive terrain points along range. If point’ panel. The user can plot the 2D graphics (PF/PL versus range
the interpolation method is chosen to be ‘none’, the terrain for fixed altitude, or PF/PL versus altitude for fixed range) either
profile appears as a collection of knife-edges. After defining by entering the values into the boxes in ‘2D graphics’ panel, or by
the terrain profile, the user must click on the ‘run’ button to right-clicking the mouse on the desired point of the 3D map (see
see the analysis results of the new geometry. It is useful to Fig. 4). The ‘colorbar’ panel is used to adjust the colorbar scale of
note that the program does not allow the terrain to extend the 3D map for better visualization. Finally, the ‘save result’ panel
below zero level. is used to store the PF/PL maps in the form of a MATLAB file (.mat)
or a picture file (.tif). We also note that 2D graphics in Fig. 4 can
be saved in text files by using the ‘save’ button.
On the right panel, there are two graphics (top graphics where
the terrain is specified, and the bottom graphics showing the col-
4. Validation, verification and calibration tests
ored 3D map of the propagation factor/loss), together with five
sub-panels (plot type, current point, 2D graphics, colorbar, save re- VV&C starts with the validation process, which is related to
sult) related to the visualization or storage of output parameters. whether the model is correct. Referring to the above-mentioned
After the user clicks on the ‘run’ button, the code performs the discussions, the limitations of the standard split-step based para-
one- or two-way split-step algorithm, and then, plots the 3D prop- bolic equation (SSPE) model are two-fold: (i) it is a one-way, for-
agation factor map on the bottom graphics. Although the default ward propagation model, which neglects backward waves; (ii) both
is propagation factor (PF), the user can switch to path loss (PL) narrow- and wide-angle PE models are valid within the parax-
map by clicking the appropriate button in the ‘plot type’ panel. ial region. For long-range propagation problems, these limitations
Whenever the user moves the mouse over the bottom graphics, the may not be serious. However, they must be treated with special
O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2647

 
Table 1 1 −( z − za )2
g ( z) = √ exp . (18)
Input parameters of PETOOL.
2πσ 2 2σ 2
Domain parameters
Maximum range (km)
Here, za and σ are the height and the spatial width of the Gaus-
Maximum altitude (m) sian source, respectively.
In the second canonical problem, the 2D surface duct over the
Antenna parameters
Polarization (horizontal or vertical)
PEC flat Earth is caused by a linearly decreasing vertical refractivity
3 dB beamwidth (degree) profile (i.e., n2 ( z) = 1 − a0 z, where a0 is a positive constant con-
Elevation angle (degree) trolling the duct strength) and the exact solution is available in
Antenna height (m) terms of Airy functions for the range-independent vertical refrac-
Frequency (MHz) (must be > 30 MHz)
tive index. The analytical modal solution using N modes is given
Analysis parameters by [41,42]
One-way or two-way (which algorithm is to be performed)
Range step (m): horizontal step size 
N
Altitude step (m): vertical step size u (x, z) = cq ψq ( z) exp(i βq x) (19)
Surface parameters q =1
Perfectly conducting or impedance surface
If impedance surface is chosen: where cq is the normalization constant, βq is the longitudinal
Type: sea, fresh water, wet ground, medium dry ground, very dry ground, propagation constant for the mode q, ψq ( z) = Ai[(a0 k2 )1/3 z − σq ] is
user-defined ground the mode function satisfying the 1D wave equation in along z, Ai is
If user-defined ground is chosen:
Dielectric constant
Airy function of the first kind, and σq is the root of the equation
Conductivity (S/m) satisfying the boundary condition Ai (−σq ) + α Ai(−σq ) = 0. The
cases where Ai(−σq ) = 0 and Ai (−σq ) = 0 denote the Dirichlet
Atmosphere parameters
Range-independent or range-dependent refractivity
and Neumann boundary conditions, respectively. Here, the prime
If range-independent refractivity is chosen: refers to the derivative with respect to the vertical coordinate. In
Type: standard atmosphere, surface duct, surface-based duct, elevated the numerical calculation, first the antenna pattern is determined.
duct, evaporation duct, and user-defined duct Then, the modal superposition in (17) or (19) is used together with
If range-dependent refractivity is chosen:
Refractivity type for each range value defined in list-boxes the orthonormality condition, and the number of modes and their
excitation coefficients are computed for a given error boundary.
Terrain parameters
Assuming a horizontally polarized Gaussian antenna at 3 GHz
None (flat surface) or terrain
If terrain is chosen: with 3 dB beamwidth of 30◦ , located at 0.5 m height and tilted
Interpolation type (none, linear, cubicspline) 45◦ downwards, inside a 1 m wide parallel-plate PEC waveguide,
Number of points: number of points to be placed on top graphics to define the performances of narrow- and wide-angle SSPE tools with re-
the terrain profile spect to the analytical result are illustrated in Fig. 6 by means
Range and height values for terrain points defined in list-boxes
of 3D field maps. Fig. 6(a) alone seems to be logical and phys-
ical. A down-tilted Gaussian beam hits the lower boundary and
attention for waves propagating upwards/downwards with larger reflects up causing interference, reaches the upper boundary and
tilts and/or for propagation paths having steep irregular terrain reflects down with the pattern shown there. The true 3D wave
profiles. The discussion of the former issue is left to Section 4.2 patterns are shown in Figs. 6(b) and 6(c). Fig. 6(b) (the analyti-
in the context of the two-way SSPE. The latter issue in conjunction cal result) is produced with the superposition of 53 modes with a
with the one-way SSPE is the main concern of Section 4.1. maximum error of 4 × 10−10 . With reference to the exact result,
the maximum and mean errors for the narrow-angle SSPE are 73%
4.1. Numerical simulations with one-way SSPE and 13% respectively, whereas for the wide-angle SSPE, the errors
are only 2.3 × 10−5 and 4.8 × 10−6 respectively. Hence, these re-
A critical issue in the VV&C procedure is the construction of the sults validate the accuracy of the wide-angle SSPE, and show that
reference (analytical or exact) solution. Two canonical problems the narrow-angle SSPE cannot handle large tilt angles. To validate
are taken into account to produce the reference solution: (i) 2D the narrow-angle SSPE, the angle must be constrained to be less
propagation inside a parallel-plate waveguide with PEC boundaries, than 10◦ while considering both beamwidth and tilt angle. Finally,
(ii) 2D groundwave propagation through a surface duct. The wave Fig. 6(d) shows field strength vs. range computed from analyti-
function inside the PEC parallel-plate waveguide with Dirichlet BC cal exact, narrow and wide angle SSPE models. As shown here,
having the width of d and located longitudinally along x may be narrow-angle SSPE is far away from handling propagation charac-
represented in terms of modal superposition given as: teristics of this scenario.
The performance of the narrow-angle SSPE in horizontal polar-
  
2
N ization can be visualized in Fig. 7, for the propagation scenario

u (x, z) = cq sin z exp(i βq x), involving two Gaussian antennas at 250 m and 500 m, with 3 dB
d d beamwidth of 1◦ , and tilt angle of 0.5◦ downwards and upwards,
q =1
 respectively. The frequency is 300 MHz; and a vertically decreas-
βq = k2 − (qπ /d)2 . (17) ing refractivity profile with a slope of dM /dz = −600 M-units/km,
corresponding to a strong anti-guiding atmosphere, is taken into
Here, βq is the longitudinal propagation constant for the mode q consideration. The range and altitude increments chosen in the
and cq is the modal excitation coefficient, numerically computed simulation are x = 25 m and  z = 0.5 m, respectively. In the an-
from the given Gaussian source function g ( z), using the orthonor- alytical result, 350 modes are used with a maximum field error of
mality property as: 10−8 . With reference to the exact result, the maximum error for
the narrow-angle SSPE is obtained as 1.4%.
d   Note that mathematically exact representations do not neces-

cq = g ( z) sin z dz, sarily yield reference solutions; unless numerical data is generated
d with a desired accuracy. Eqs. (17) and (19) are exact solutions in
0
2648 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

Fig. 6. [Calibration of one-way SSPE.] Field maps for a Gaussian antenna at 0.5 m height with 3 dB beamwidth of 30◦ , tilt angle of 45◦ inside a parallel-plate waveguide at
3 GHz: (a) narrow-angle SSPE, (b) analytical result, (c) wide-angle SSPE, (d) horizontal field profiles at z = 0.25 m and z = 0.5 m [x = 5 mm,  z = 2.5 mm].

Fig. 7. [Calibration of one-way narrow-angle SSPE.] Two Gaussian antennas at 250 m and 500 m with 3 dB beamwidth of 1◦ , tilt angles of −0.5◦ and 0.5◦ at 300 MHz above
PEC ground, and inside a vertically decreasing refractivity with a slope of 600 M-units/km: (a) field map for analytical result, (b) field map for narrow-angle SSPE, (c) vertical
field profiles at two different ranges [x = 25 m,  z = 0.5 m, 350 modes are used for the analytical result].

Table 2 as a function of antenna tilt. Table 2(a) belongs to the parallel-


Number of modes as a function of tilt angle, for maximum initial field er- plate waveguide with PEC boundaries, showing that the number
ror (< 10−8 ). (a) Parallel plate PEC waveguide ( f = 3 GHz, za = 0.5 m, θ B W =
of modes (for the given parameters) would be much less than
30◦ ). (b) Surface duct ( f = 300 MHz, za = 250 m, θ B W = 65◦ , dM /dz =
−600 M-units/km). 100 for tilts up to 45◦ (note that, although only trapped modes
with real propagation constants contribute fields beyond several
(a) (b)
wavelength distances, all trapped and non-trapped modes are still
Tilt angle Number Tilt angle Number
(degree) of modes (degree) of modes
required to precisely reconstruct near fields). The number of modes
increases as the frequency increases and the beamwidth gets nar-
0 37 0 19
10 40 2 191 rower, but all are numerically manageable. On the other hand,
20 43 4 795 Table 2(b) shows the number of modes for the surface duct prob-
30 46 6 2099 lem. Tens of thousands of modes would not be sufficient to take
40 49 8 4380
the tilt angles more than 10◦ into account (even for the twice wide
45 50 10 7926
beamwidth). Also the computations become a real challenge! First
of all, modes are confined between the ground and their caustics
terms of modal summation, therefore may serve as reference if nu- [13] which go higher and higher as the mode number increases.
merically computed with a specified accuracy. Although an exact The modal excitation coefficients are calculated numerically using
mathematical solution is available at hand, it might be inefficient the orthonormality principle, therefore a numerical integration is
to perform numerical computation of the modal summation for essential in the vertical domain for each mode. This means that
larger tilts, because the number of modes increases as the antenna both the integration step and the upper integral boundary should
tilt increases. This is clarified in Table 2, which lists the number be changed dynamically. Finally, all the modes are required in field
of modes required to establish a given Gaussian antenna pattern computations at every range.
O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2649

Hence, it might be difficult to produce numerical reference data polarization is horizontal, the range and the altitude increments
for the VV&C tests. A convenient way of testing narrow- and wide- are x = 10 m and  z = 0.13 m, respectively. The 3D field maps,
angle SSPE models is to tilt up or down the pattern up to 40◦ – corresponding to narrow- and wide-angle SSPE are illustrated in
45◦ . In the third scenario, five Gaussian antennas located at the Fig. 8. The specular reflection points (r i ) on the ground surface
same place (1000 m height) with 3 dB beamwidth of 0.5◦ each, are tabulated in Table 3 for the tilt angles used in Fig. 8, together
and 10◦ –20◦ –30◦ –40◦ –50◦ tilt angles illuminate the PEC ground in with various tilt angles. Although there is an almost exact match
a standard atmosphere, assuming that the frequency is 1 GHz, the
between the desired specular reflection points and those found by
the wide-angle SSPE, the results of the narrow-angle SSPE start to
deteriorate as the tilt angle increases.

Table 3
Specular reflection points and errors as a function of tilt angle.

Tilt angle Exact and wide- Narrow-angle Percentage


(degree) angle reflection reflection point difference
point (km) (km) (%)
8 7 .1 7.2 1.41
9 6 .3 6.4 1.58
10 5 .6 5.77 1.76
11 5.14 5.25 2.14
12 4 .7 4.82 2.55
13 4.33 4.45 2.77
14 4.01 4.14 3.24
15 3.73 3.87 3.75
20 2.74 2.92 6.57
25 2.14 2.36 10.28
30 1.73 2 15.61
35 1.43 1.74 21.67
40 1.56 1.19 31.09
Fig. 8. [Calibration of one-way SSPE.] Propagation Factor (PF) maps for five Gaussian
45 1 1.42 42.00
antennas with different tilt-angles illuminating PEC ground: (Top) narrow-angle,
50 0.84 1.3 54.76
(Bottom) wide-angle [specular reflection points are shown in Table 3].

Fig. 9. [Calibration of one-way SSPE with AREPS.] A Gaussian antenna at 50 m height with 3 dB beamwidth of 3◦ , tilt angle of 0◦ inside a surface duct: (Upper-left) main
window with PF map, (Upper-right) modified refractivity profile for surface duct, (Lower-left) PF vs. range at 50 m altitude, (Lower-right) PF vs. altitude at 120 km range
[ f = 3 GHz, x = 200 m and  z = 0.29 m].
2650 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

Fig. 10. [Calibration of two-way SSPE via GO + UTD.] A finite-height wall (50 m
height, 50 km range) in front of an infinite-height wall at 60 km illuminated by a
line-source at 250 m: (a) PF maps, (b) PF vs. altitude at 55 km range, (c) PF vs.
range at 50 m altitude [ f = 3 GHz, x = 200 m and  z = 0.29 m].

Finally, the performance of the one-way SSPE is demonstrated


against AREPS1 ver. 3.0, which simulates only the one-way prop-
agation. The comparison is based on the narrow-angle SSPE be-
cause this version of AREPS implements so. As shown in Fig. 9,
a Gaussian antenna at an altitude of 50 m illuminates a variable
terrain inside a surface duct, which is modeled by the modified
refractivity profile M. As illustrated by the propagation factor (PF)
Fig. 11. [Calibration of two-way SSPE via GO + UTD.] Two finite-height walls (100 m
maps, a good agreement is observed between PETOOL and AREPS high wall at 20 km range and 150 m high wall at 40 km range) illuminated by a
in narrow-angle case. line-source at 5 m: (Top) PF maps, (Bottom-left) PF vs. altitude at 15 km (in the in-
The good agreement among the results illustrates the success terference region), (Bottom-right) PF vs. altitude at 35 km (in the shadow/diffraction
region) [ f = 3 GHz, x = 200 m and  z = 0.29 m].
and completeness of the VV&C process for the one-way SSPE.

4.2. Numerical simulations with two-way SSPE In Fig. 10, the finite-height (50 m) wall is located at 50 km; the
infinite-height wall is at 60 km; the source is at 250 m; and the
This section demonstrates the test results of the VV&C pro- polarization is vertical. This is one of the classical structures in
cess of the two-way SSPE tool over different propagation scenar- the field of diffraction theory, and its approximate solution can
ios. In the first two scenarios, the calibration is done against the be computed by using ray-optic techniques, combined with special
GO + UTD results, assuming that the frequency is 3 GHz. It is use- diffraction methods. In the implementation of the GO method, the
ful to note that the calibration via GO + UTD is feasible only for reflected waves from the ground and the wall can simply be cal-
sufficiently high frequencies such that the ray-optics interpretation culated by employing the principles of image theory that replaces
is valid. the original problem with the equivalent problem represented by
The first scenario involves two walls, one of which is finite (also image sources with respect to the boundary conditions that must
known as a knife-edge) and the other one is infinite in height. be satisfied on the boundaries depending on the polarization. The
total field is obtained by the sum of the direct ray, reflected rays
emanating from image sources, and the diffracted rays from the tip
1
http://areps.spawar.navy.mil/. of the wall, by also checking the line-of-sight (LOS) conditions be-
O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2651

Fig. 12. [Calibration of two-way SSPE.] PF maps for two Gaussian antennas above ground with different tilt angles illuminating two opposite infinite-walls: (Left) narrow-
angle, (Right) wide-angle [specular reflection points: z1 = 176 m, z2 = 577 m (wide-angle and exact), z1 = 174 m, z2 = 500 m (narrow-angle)] [ f = 3 GHz, x = 5 m and
 z = 0.06 m].

Fig. 13. [Calibration of two-way SSPE.] PF maps for a variable terrain illuminated by a Gaussian antenna at 50 m in an elevated-duct environment: (Left-top) one-way
narrow-angle, (Left-bottom) one-way wide-angle, (Right-top) two-way narrow-angle, (Right-bottom) two-way wide-angle [ f = 3 GHz, x = 200 m and  z = 0.29 m].

tween the source(s) and the observation point. As shown in Fig. 10, observed in the GO + UTD model, which are inherently not ob-
multiple reflections (almost resonance behavior) occur especially served in the SSPE map. The two plots on the bottom belong to PF
in the region between the walls. In comparing the two-way SSPE vs. height at two different ranges. As observed from two-way SSPE
with the GO + UTD approach, the contribution of the waves hitting and GO + UTD results, a good agreement is obtained in the first
the walls up to three times is superposed. To achieve fair compar- region (at 15 km that is in the interference region), but a slight dis-
isons up to third degree of reflections, the GO + UTD code accounts crepancy is observed in the shadow region (at 35 km that is in the
for 35 types of rays bouncing from the walls and the ground. In shadow/diffraction region). This discrepancy can be eliminated by
addition to reflected waves, the diffracted waves from the finite- taking into account the slope diffraction coefficient in the GO+UTD
height wall are also computed. However, the multiple bouncing of model, whose accuracy decreases in deep-shadow regions. Further-
the diffracted fields from the walls and the ground is ignored due more, this discrepancy might be due to ignoring some of the less
their negligible effects compared to strong reflections. The good contributing components (such as double-diffractions), and due to
agreements among the results illustrate the success of the two- the limitation of the SSPE within the paraxial regions; therefore
way SSPE with respect to GO + UTD approach. further investigation is required in order to speculate about this
The scenario presented in Fig. 11 involves two finite-height discrepancy.
walls along the propagation path (100 m high wall at 20 km range The forthcoming scenario is designed to better grasp the dif-
and 150 m high wall at 40 km range) illuminated by a line-source ferences between narrow- and wide-angle propagators in an en-
at 5 m (i.e., multiple-wedge problem). The figures on the top show vironment exhibiting strong reflections. Two Gaussian antennas,
the PF maps obtained with the one-way SSPE, two-way SSPE and located just above ground, with 3 dB beamwidth of 0.3◦ each, and
GO + UTD models. The boundaries of both incident (i.e., LOS) and −10◦ , −30◦ tilt angles illuminate two opposite infinite-walls, as
reflected fields are observed in both two-way SSPE and GO + UTD shown in Fig. 12. It is assumed that the polarization is horizon-
maps. The artificial effects around these boundaries are also clearly tal, and the medium is free-space. The expected specular reflection
2652 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

Fig. 14. [Miscellaneous PETOOL simulations.] PF maps for a variable perfectly-conducting terrain illuminated by a Gaussian antenna at 10 m assuming different refractivity
profiles: (a) surface duct (one-way), (b) surface duct (two-way), (c) surface-based duct (one-way), (d) surface-based duct (two-way) [ f = 3 GHz, x = 200 m and  z =
0.29 m].

Fig. 15. [Miscellaneous PETOOL simulations.] The same scenario in Fig. 14, except that the antenna is at 100 m height.
O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654 2653

Fig. 16. [Miscellaneous PETOOL simulations.] PF maps for a variable terrain illuminated by a Gaussian antenna at 5 m assuming different ground surface parameters in
standard atmosphere: (a) perfectly-conducting surface (one-way), (b) perfectly-conducting surface (two-way), (c) very dry ground (εr = 3, σ = 1e–4) (one-way), (d) very dry
ground (two-way) [ f = 100 MHz, x = 100 m and  z = 10 m].

points of the beams on the walls are as follows: z1 = 176 m and Core 2 Duo 1.6 GHz with 2 GB RAM PC. Note that the computations
z2 = 577 m. The wide-angle version of the two-way SSPE provides times are dependent on the number of iterations in the two-way
accurate results for all tilt angles and specular reflection points. SSPE, which in turn depend on the amount of the wave interac-
However, the narrow-angle SSPE is accurate for only −10◦ tilt an- tions (multiple reflections) between the hills. In Fig. 15, the same
gle (Narrow: z1 = 174 m and z2 = 500 m). The error increases scenarios in Fig. 14 are performed, except that the antenna is lo-
as the tilt angle increases in the narrow-angle SSPE, as expected. cated at 100 m. Computations times are: 4 s (one-way SSPE), 672 s
The new scenario in Fig. 13 illustrates the comparison between (Fig. 15(b)), 963 s (Fig. 15(d)). In Fig. 16, different ground parame-
one-way and two-way SSPE models with narrow and wide angle ters are simulated assuming that a Gaussian antenna is located at
propagators over a variable terrain and in an elevated duct envi- 5 m height and radiates into a standard atmosphere. In Fig. 16(a)
ronment. A Gaussian antenna with 2◦ tilt angle is located at 50 m and 16(b), the surface is perfectly conducting. In Fig. 16(c) and
height. The frequency is 3 GHz, and the polarization is horizontal. 16(d), the surface is “very dry ground” with dielectric constant
Referring to the descriptions in Fig. 5, the elevated duct is defined εr = 3 and conductivity σ = 1e–4 siemens/m. Polarization is ver-
as follows: M 0 = 300, M 1 = 330, M 2 = 310, M 3 = 350, z1 = 100 m, tical, and the frequency is 100 MHz. Computations times are: 8 s
z2 = 150 m, z2 = 300 m. These examples help us to visualize the (Fig. 16(a)), 112 s (Fig. 16(b)), 8 s (Fig. 16(c)), 129 s (Fig. 16(d)).
importance of the correct representation of large propagation an-
gles, as well as the backward propagating waves. 5. Conclusions
As emphasized in the Introduction part, PETOOL has been de-
signed for research/educational purposes. Hence, propagation en- A MATLAB-based one-way and two-way split-step based para-
gineers/instructors can simulate different propagation scenarios bolic equation program — called PETOOL — has been introduced
to investigate the radio-wave propagation and/or to design re- and discussed systematically for a radio-wave propagation problem
liable communication links. In order to demonstrate what en- over an arbitrary ground profile through homogeneous and inho-
gineer/instructors can do with PETOOL, we have shown various mogeneous atmosphere. It has been concluded that the standard
scenarios in Figs. 14, 15 and 16. In Fig. 14, a Gaussian antenna, PE model has certain drawbacks in terms of handling backward
which is located at 10 m height, radiates into surface duct and propagation, as well as large propagation angles. These issues may
surface-based duct environments. Referring to the descriptions in be predominantly observed in problems involving arbitrary obsta-
Fig. 5, the surface duct is defined as follows: M 0 = 350, M 1 = 300, cles on the propagation path or in short-range problems, or combi-
M 2 = 350, z1 = 200 m, z2 = 300 m. The surface-based duct is nation of two (for example, in urban propagation scenarios). They
defined as follows: M 0 = 340, M 1 = 356, M 2 = 340, M 3 = 358, may also be observed in propagation scenarios where abrupt, se-
z1 = 135 m, z2 = 150 m, z2 = 300 m. Polarization is horizontal, vere changes in atmospheric conditions occur. The wide-angle ver-
and the ground is perfectly conducting. Computations times are: sion of the two-way SSPE overcomes these limitations up to a great
4 s (one-way SSPE), 820 s (Fig. 14(b)), 643 s (Fig. 14(d)), on Intel extent, as demonstrated by several numerical calibration tests per-
2654 O. Ozgun et al. / Computer Physics Communications 182 (2011) 2638–2654

formed with respect to reference results. The program structure [21] G. Apaydin, L. Sevgi, Numerical investigations of and path loss predictions for
and the GUI capabilities of PETOOL have been discussed. It is be- surface wave propagation over sea paths including hilly island transitions, IEEE
Trans. Antenn. Propag. 58 (2010) 1302–1314.
lieved that PETOOL serves as a user-friendly research/educational
[22] G. Apaydin, O. Ozgun, M. Kuzuoglu, L. Sevgi, A novel two-way finite-element
tool for propagation engineers/instructors.
parabolic equation groundwave propagation tool: tests with canonical struc-
tures and calibration, IEEE Trans. Geoscience and Remote Sensing 49 (2011)
References 2887–2899.
[23] M.F. Levy, Parabolic equation modeling of propagation over irregular terrain,
[1] D.E. Kerr, Propagation of Short Radio Waves, Peninsula Publishing, Los Altos, Electron. Lett. 26 (1990) 1153–1156.
CA, 1988. [24] A.E. Barrios, A terrain parabolic equation model for propagation in the tropo-
[2] K.G. Budden, The Waveguide Mode Theory of Wave Propagation, Logos Press, sphere, IEEE Trans. Antenn. Propag. 42 (1994) 90–98.
London, 1961.
[25] R. Janaswamy, A curvilinear coordinate-based split-step parabolic equation
[3] J.B. Keller, Geometrical theory of diffraction, J. Opt. Soc. Amer. 52 (1962) 116–
method for propagation predictions over terrain, IEEE Trans. Antenn. Propag. 46
130.
(1998) 1089–1097.
[4] R.G. Kouyoumjian, P.H. Pathak, A uniform geometrical theory of diffraction for
[26] D.J. Donohue, J.R. Kuttler, Propagation modeling over terrain using the parabolic
an edge in a perfectly conducting surface, Proc. IEEE 62 (1974) 1448–1461.
[5] P.Ya. Ufimtsev, Theory of Edge Diffraction in Electromagnetics, Tech. Science wave equation, IEEE Trans. Antenn. Propag. 48 (2000) 260–277.
Press, Encino, CA, USA, 2003. [27] T.K. Tsay, B.A. Ebersole, P.L.F. Liu, Numerical modelling of wave propagation
[6] L.B. Felsen, F. Akleman, L. Sevgi, Wave propagation inside a two-dimensional using parabolic approximation with a boundary-fitted co-ordinate system, Int.
perfectly conducting parallel plate waveguide: hybrid ray-mode techniques and J. Numer. Methods Engin. 27 (1989) 37–55.
their visualizations, IEEE Antenn. Propag. Mag. 46 (2004) 69–89. [28] M.F. Levy, A.A. Zaporozhets, Parabolic equation techniques for scattering, Wave
[7] M.A. Leontovich, V.A. Fock, Solution of propagation of electromagnetic waves Motion 31 (2000) 147–156.
along the Earth’s surface by the method of parabolic equations, J. Phys. USSR 10 [29] M.D. Collins, R.B. Evans, A two-way parabolic equation for acoustic backscat-
(1946) 13–23. tering in the ocean, J. Acoust. Soc. Am. 91 (1992) 1357–1368.
[8] R.H. Hardin, F.D. Tappert, Applications of the split-step Fourier method to the [30] M.J. Mills, M.D. Collins, J.F. Lingevitch, Two-way parabolic equation techniques
numerical solution of nonlinear and variable coefficient wave equations, SIAM for diffraction and scattering problems, Wave Motion 31 (2000) 173–180.
Rev. 15 (1973) 423. [31] O. Ozgun, Recursive two-way parabolic equation approach for modeling terrain
[9] K.H. Craig, Propagation modelling in the troposphere: parabolic equation
effects in tropospheric propagation, IEEE Trans. Antenn. Propag. 57 (9) (2009)
method, Electron. Lett. 24 (1988) 1136–1139.
2706–2714.
[10] G.D. Dockery, Modeling electromagnetic wave propagation in the troposphere
[32] O. Ozgun, G. Apaydin, M. Kuzuoglu, L. Sevgi, Two-way Fourier split step algo-
using the parabolic equation, IEEE Trans. Antenn. Propag. 36 (1988) 1464–1470.
rithm over variable terrain with narrow and wide angle propagators, in: IEEE
[11] J.R. Kuttler, G.D. Dockery, Theoretical description of the parabolic approxima-
Int. Symp. on Antenn. and Propag., ON, Canada, 2010.
tion/Fourier split-step method of representing electromagnetic propagation in
the troposphere, Radio Sci. 26 (1991) 381–393. [33] O. Ozgun, G. Apaydin, M. Kuzuoglu, L. Sevgi, Two-way split-step parabolic
[12] M.F. Levy, Parabolic Equation Methods for Electromagnetic Wave Propagation, equation algorithm for tropospheric propagation: tests and comparisons, in:
IEEE Electromagnetic Wave Series, vol. 45, Institution of Electrical Engineers Mediterranean Microwave Symposium MMS’2010, Guzelyurt, Turkey, 2010.
(IEE), London, 2000. [34] M.F. Levy, Diffraction studies in urban environment with wide-angle parabolic
[13] L. Sevgi, Complex Electromagnetic Problems and Numerical Simulation Ap- equation method, Electron. Lett. 28 (1992) 1491–1492.
proaches, IEEE Press/John Wiley, New York, 2003. [35] M.D. Collins, A split-step Padé solution for the parabolic wave equation,
[14] L. Sevgi, C. Uluisik, F. Akleman, MATLAB-based two-dimensional parabolic J. Acoust. Soc. Am. 94 (1993) 1736–1742.
equation radiowave propagation package, IEEE Antenn. Propag. Mag. 47 (2005) [36] M.D. Feit, J.A. Fleck, Light propagation in graded-index fibers, Appl. Opt. 17
164–175. (1978) 3990–3998.
[15] S.W. Marcus, A hybrid (finite difference-surface Green’s function) method for [37] D.J. Thomson, N.R. Chapman, A wide-angle split-step algorithm for the
computing transmission losses in an inhomogeneous atmosphere over irregular parabolic equation, J. Acoust. Soc. Am. 74 (1983) 1848–1854.
terrain, IEEE Trans. Antenn. Propag. 40 (1992) 1451–1458.
[38] J.R. Kuttler, Differences between the narrow-angle and wide-angle propagators
[16] P.D. Holm, Wide-angle shift-map PE for a piecewise linear terrain — a finite-
in the split-step Fourier solution of the parabolic wave equation, IEEE Trans.
difference approach, IEEE Trans. Antenn. Propag. 55 (2007) 2773–2789.
Antenn. Propag. 47 (1999) 1131–1140.
[17] C. Mias, Fast computation of the nonlocal boundary condition in finite differ-
[39] J.F. Claerbout, Fundamentals of Geophysical Data Processing with Application
ence parabolic equation radiowave propagation simulations, IEEE Trans. An-
to Petroleum Prospect, McGraw–Hill, New York, 1976.
tenn. Propag. 56 (2008) 1699–1705.
[18] D. Huang, Finite element solution to the parabolic wave equation, J. Acoust. [40] M.I. Skolnik, Introduction to Radar Systems, McGraw–Hill, New York, 2001.
Soc. Am. 84 (1988) 1405–1413. [41] J.R. Wait, Electromagnetic Waves in the Stratified Media, Pergamon Press, Ox-
[19] G. Apaydin, L. Sevgi, FEM-based surface wave multi-mixed-path propagator and ford, 1962.
path loss predictions, IEEE Antenn. Wireless Propag. Lett. 8 (2009) 1010–1013. [42] L. Sevgi, Modeling and simulation strategies for electromagnetic wave propaga-
[20] G. Apaydin, L. Sevgi, The split step Fourier and finite element based parabolic tion in complex environments: groundwave path loss prediction virtual tools,
equation propagation prediction tools: canonical tests, systematic comparisons, in: Special Issue on Electromagnetic Wave Propagation in Complex Environ-
and calibration, IEEE Antenn. Propag. Mag. 52 (2010) 66–79. ments: A Tribute L.B. Felsen, IEEE Trans. Antenn. Propag. 55 (2007) 1591–1598.

You might also like