Teaching and Designing Antenna Arrays

Using Signal and Image-Processing

Toolboxes of MATLAB
Joseph Sahaya Kulandal Raj and Arunachalam Ponnusamy Kabilan
Department of Electronics and Communication Engineering
Ban nani Amman Institute of Technology
Sathyamangalam-638 401 ,Tamilnadu, India
Tel: 91 4295 221289; Fax: 91 4295 223775; E-mail: sahajose@rediffmail.com

Abstract
The goal of this paper is to apply some of the similarities between signal-processing theory and antenna-array theory to
analyze and synthesize antenna arrays by using the signal and image-processing toolboxes of MATLAB T. A detailed
description and examples are given. The synthesis of one-dimensional and two-dimensional rectangular arrays is
demonstrated using this simple method. These tools can be used for teaching antenna design.
Keywords: Antenna arrays; signal processing; analysis; synthesis; FIR digital filters

1. Introduction MA TLAB was chosen in part because it is available world-

wide and relatively affordable, even in developing countries. While
programs can be written in MATLAB itself to analyze and synthe-
A ntenna arrays are important components of present-day wire-
less communication systems. It has been proven that adaptive size arrays, students can gain additional knowledge and insight by
comparing signal-processing theory with antenna theory, using the
arrays have significant performance advantages over conventional
signal and image-processing toolboxes. These toolboxes have
antennas. It has also been shown that having antenna arrays at both
the base station and the wireless terminal will improve the spectral many built-in functions for filter synthesis that are ideal for array
efficiency and quality of service (QoS). Therefore, it is imperative synthesis.
for communication-engineering students to understand antenna
theory. Antenna theory at the undergraduate level is typically an
advanced subject, with EM theory as a prerequisite. Visualizations
of the radiated field and the radiation patterns dramatically increase 2. Analysis of an Antenna Array
student understanding and interest in this subject. [1]
The array factor of N isotropic point sources with different
Generally, the learning process involves augmenting one's excitations placed along the x axis on the x-y plane is given by [2]
existing knowledge with new concepts. Therefore, while teaching a
new concept to students their already acquired knowledge can be
exploited to assimilate the concept. Explaining a new concept to AF =-W + e+ We + we +. + wn 1e- ),
the student is made effective if we compare the new concepts to
other concepts with which the student is already familiar. The
strong similarity between antenna theory and signal-processing where
theory can be used in this way. Students typically learn signal-
processing theory earlier than antenna theory, and antenna educa- W'=,gd cos0+ a,
tors can make use of the former to illustrate the later. This paper
presents some of the analogies between signal-processing theory the %, are the current excitations, and a is the phase shift
and antenna-array theory. A detailed account of usage of the signal between adjacent elements. The frequency response of an FIR filter
and image-processing toolboxes of MA TLAB is given with regard is given by
to the analysis and synthesis of certain antenna arrays. The synthe-
sis of one-dimensional and two-dimensional rectangular arrays is + hn-le j(fll)w
H(jw) = ho+ hle io + he-o,+ h3 e-j3o, +..
also demonstrated using the toolboxes. In addition, Section 4 gives
the similarity between the sampling rate in FIR filters and the (2)
spacing between the elements in antenna arrays. Section 5 illus-
trates the use of convolution to explain the concept of pattern mul- where the hn are the inmpulse-response coefficients of the FIR fil-
tiplication. ter.

184 IEEE Antennas and Propagation Magazine, Vol. 49, No. 4, August 2007
 Equations (1) and (2) are of similar format, except for a
minus sign in the power of the complex exponentials. Therefore,
the tool that is used for finding the frequency response can be used
to find the array response as a function of direction, q0.The built-in
function f reqz in the signal-processing toolbox of MATLAB
evaluates the frequency response given the impulse-response coef-
ficients [4]. The same function can be used to evaluate the radia-
tion pattern of the array given the current excitations. The follow-
ing example demonstrates the idea. A four-element array has [1,
1.2, 1.2, 1] as its current excitations. Figure 1 shows the pattern
obtained by using f reqz command. The code in Figure 2 gener-
ates the radiation pattern of the four-element array.
The array factor for an M x N rectangular array on the x-y
plane [3] is given by
v'Y =kdYsin 0osin0,
d,, and dyare the spacings between the adjacent elements along
the x axis and the y axis, respectively, and the Wrnn are the current
excitations. The frequency response of a two-dimensional filter is
given by
QOý iO2, y 2J-(4)
rn-I n=1
where the hrnn are the impulse-response coefficients.
Equations (3) and (4) are of similar format, except for a
minus sign in the power of the complex exponentials. Therefore,
the tool that is used for finding the frequency response can be used
to find the array response as a function of variables 0 and gi. The
image-processing toolbox of MfATLAB has the function f reqz2
for finding the frequency response of two-dimensional filters. The
same could be effectively used to evaluate the radiation pattern of
the two-dimensional arrays [5]. For example, the code in Figure 3
generates the radiation pattern of a 5 x 5 uniform rectangular array.
The radiation pattern of the array on the x-y plane is shown in Fig-
ure 4.
3. Synthesis of an Antenna Array
One of the problems of array synthesis consists of finding the
current excitations of the array elements, given the desired radia-
tion pattern or some data about the radiation pattern. This synthesis
problem is equivalent to the problem of finding the impulse
response of a digital FIR filter for the desired frequency response,
as follows from the explanation in the previous section. The
MA7hAB Signal-Processing Toolbox has the f iris function for
designing a one-dimensional FIR filter that is optimal in a least-
squares sense [4]. The arguments of the function are
N, one less than the number of impulse response coefficients;
IEEE Antennas and Propagation Magazine, Vol. 49, No. 4, August 2007
F, a vector of frequency points, in ascending order between 0
and 1; 1 corresponds to half the sampling frequency;
A, a real vector of the same size as F, which specifies the
desired amplitude of the frequency response of the
resultant filter.
If the function is given as firls (N, F, A), it returns an FIR filter
with N + 1 impulse-response coefficients that has the best approxi-
mation to the desired frequency response described by F and A in
the least-squares sense. This function can be used to design an
antenna array if a proper mapping is executed between the fre-
quency variable, F, and the variable Vf, which in turn is related to
gi. For a broadside array, if the spacing between the elements is
A/2, then V/=ir cos 0. In this case, when gi varies from 0 to Yr,
V/will vary from ir to -7r. Thus, for the F-vector values in the
f iris function to vary from 0 to 1 (the normalized range of w),
the corresponding V/ values will range from 0 to -;r. The normal-
ized E-field values of the desired radiation pattern at the V/' vector
values is given by A. N will now be equal to the number of ele-
ments in the array minus one. The values in the vector F should be
in ascending order and normalized to unity. For instance, let us
desigu an antenna array that will give an approximately bidirec-
tional pattern as in Figure 5 (indicated by the blue line). The pat-
tern is defined by
A(g9=0, O<gi<7r/3,
A(gi)=l, ;r/3<gi<2,r/3,
Using the above approach for a 16-element linear array, the fol-
lowing relative current excitations were obtained: [-0.0284,
-0.0375, 0.0382, 0.0517, -0.0627, -0.0929, 0.1473, 0.4518,
0.4518, 0.1473, -0.0929, -0.0627, 0.0517, 0.0382, -0.0375,
90
240 300
270
0urn
Fue1 . Th foreeetarywt
ttern of.......
....
a.....
.....
excitations of [1, 1.2, 1.2, 11
185
 phi=0:0.01 :2*pi;
shi=pi*cos(phi);
Current excitations-[1, 1.2, 1.2, 1];
E=freqz(Current excitations, 1,shi);
polar(phi,abs(E)/max(abs(E)))
% 0<0<27c
" for broad side array with X/2 spacing
" Relative currents to the elements
" E for different yi values
" Generating the radiation pattern

Figure 2. The code for generating the radiation pattern of the four-element array.

Current -excitations-ones(5); 0/
0 Relative currents to the elements
theta-0:.04:pi;phi=0:0.04:2*pi; o < 0 < 7r, 0 < qI<27t
[the,ph]-ndgrid(theta,phi); Making grid of points
shix.~cos(ph).*sin(the); % Normalized yix andyy
shiy-sin(ph).*sin(the);
E =freqz2(Current-excitations,shix,shiy); % E for different yix and y' y
En--abs(E)Imax(max(abs(E)));
x En.*sin(the).*cos(ph); % Converting the spherical coordinates
y--En.*sin(the).*sin(ph); % into rectangular coordinates
z En.*cos(the);
mesh(x,y,z) % 3-D plotting

Figure 3. The code for generating the radiation pattern of a 5 x 5 uniform rectangular array.

a7-[0:0.0l :pi!3];
b-4pi/3 :0.01 :2*pi/3];c4-2*piI3:0.01 :4*pi/3];
d=[4*piI3 :0.01: 10*pi/6] ;e--[10*piI6:0.01:2*pi];
T=[a b c d el; % vector of direction variable
Spec- jzeros(1 ,length(a)),ones(l ,length(b)),zeros(l1,length(c)),ones(1 ,length(d)),zeros(1 ,length(e))]
polar(T,Spec) %plot of the specifications
shi~pi*cos(T);
F shi((length(T)/4-1):-1 : 1)/pi; % Sorting y in ascending order and normalizing to 1
A Spec((length(T)/4- 1):-1::1); % Sorting the spec. also in the order of y
Cur-rent -excitations--flrls(1 5,F,A); % Synthesis
E-freqz(Current excitations, 1,shi); % Analysis
hold on
polar(T,abs(E),Yr)

Figure 6. The MA TLAB code for the 16-element linear array example.

phi=0:0.01 :2*pi; % 0<qI<27c

shi=rpi*cos(phi); % y for broad side array with k/2 spacing
Current -excitations -chebwin(6, 15); % synthesis
E freqz(Current excitations,1,shi);
polar(phi,abs(E)/max(abs(E)));

Figure 7. An example of the use of the chebwin function.

-0.02841]. The pattern of the synthesized array is indicated by the
red line in Figure 5. The MATLAB code for the above example is
given in Figure 6.
Another usefuil function related to synthesis is chebwin [4].
For instance, if a six-element Chebychev array is to be designed
with a -15 dB sidelobe level (SLL), the chebwiri function, avail-
able in the signal-processing toolbox, can be used (see Figure 7).
The fuanction chebwin (n, r) returns the n-point Chebychev win-
dow with a sidelobe level of r dB. The window values are the cur-
rent excitations of the array elements. Figure 8 shows the radiation
pattern of the synthesized Chebychev array.
Similarly, two-dimensional rectangular arrays can be synthe-
sized using tools available in the image-processing toolbox. The
tools are f windl, f sanp2, f trans2, and f wind2 [5]. For exam-
ple, H = FWIND2 (FR, WIN) designs a two-dimensional FIR filter
H with frequency response FR. FWIND2 uses the two-dimensional
window WIN to truncate the infinitely large impulse response
required to meet the given frequency response, FR. WIN can be the
boxcar (rectangular), hamming, hanning, bartlett,
blackman, kaiser, or chebwin window functions of the signal-
processing toolbox. To synthesize the rectangular array, the
f wind2 function's argument FR is given by the desired radiation-
pattern samples in the 0 and q5 directions, and the dimension of

186 IEEE Antennas and Propagation Magazine, Vol. 49, No. 4, August 2007
 0.8-~

0.61.

0.4-

&2-~1IIjj/(l
0-

A21

-1A

5
Figure 4. The radiation pattern of a 5 x 5 uniform rectangular
06
array.
44

90

Figure 5. The desired (blue line) and synthesized patterns. 0

270

-rwe- -at

as6

Figure 10b. The synthesized pattern for the 10 x 10 two-dimen-

x
sional array.

IEEE Antennas and Propagation Magazine, Vol. 49, No. 4, August 200718 187
WIN is given by N x N, the dimension of the array. The Output, H, cation of two frequency responses, corresponding to two time
is the array excitations. The code in Figure 9 synthesizes a 10 x 10 sequences. The resultant frequency response obtained by multipli-
array. Figure IlOa shows the desired pattern, and Figure l Ob shows cation is actually the frequency response of the resultant sequence
the synthesized pattern, of the two-dimensional array. obtained by convolving the two sequences. In the same way, the
array for which the pattern is needed can be thought of as a result
of convolution of two arrays, one comprising mnindividual ele-
4. Spacing and Sampling Rate ments of the array and the other comprising n elements. Each of the
n elements is a combination of m elements. For instance, to dem-
The effect on the array pattern of changing the spacing onstrate the pattern-multiplication concept, a four-element iso-
between the adjacent elements of the array is same as the effect on tropic array was used in [61, as shown in Figure 13a. The four-
the frequency response of changing the sampling period of the FIR element array is treated as a combination of two arrays of two ele-
filter. To illustrate this point, the frequency responses of a 15-tap ments each. The two-element isotropic array can be regarded as a
FIR filter with uniform impulse response coefficients for different directional antenna. Then, the four-element array is equivalent to
sampling rates were compared with the array responses of a 15- an array of two directional antennas separated by a distance 2d.
element uniform linear array with different spacing. The frequency Now, by multiplying the pattern of the 2d -spaced two-element
responses are plotted from -1 Hz to 1 Hz for the Nyquist rate, one- array with the pattern of the d-spaced two-element array, the pat-
half the Nyquist rate, and one-quarter of the Nyquist rate in Fig- tern of the four-element array can be obtained. The four-element
ures 1Ila, 11 b, and 11 c, respectively. It can be observed that the array is comparable to a discrete time sequence having four unit
frequency response began to shrink as the sampling frequency samples with sampling period d. The two-element array separated
decreased (the sampling period increased). As a result, more main
lobes crept into the fundamental range, which otherwise had only
one main lobe.
90
In Figures 12a, 12b, and 12c, the array responses for a 15-
element uniform linear array are plotted to show the analogy. As
the spacing between the elements was increased, more main lobes
occurred within the range of 9, just as in the frequency response of
the FIR filter. This analogy can be used to explain the occurrence
120 60
of "grating lobes" to the students, who have already come across
the phenomenon of "aliasing." Aliasing occurs when the sigual is
sparsely sampled. Aliasing is reflected in the frequency response of
the filter, because it should respond in the same way for two differ-
ent frequencies having the same identity. Similarly, aliasing of the
pattern in the spatial domain occurs when the array elements are
sparsely arranged. If the sampling rate is sufficiently high, the fre-
quency response expands, and only the main lobe may occupy the
entire fundamental range of -1 Hz to 1 Hz. In the same way, if the
array elements are arranged densely enough, all the sidelobes will
disappear from the range of 9.

240 300
5. Convolution and Pattern Multiplication
270
Pattern multiplication is used to find the array pattern from Figure 8. The synthesized pattern of a six-element Chehychev
the patterns of individual elements. This is analogous to multipli- array.

phi=~2*pi*((O: 1: 100)/100); % 0 < 0<27i

theta~pi*((0: 1: 100)/100); % 0<0 <7
[shix,shiy] = freqspace(l 01 ,'meshgrid');
r-sqrt(shix.A2+shiy.A2);
FR~ones(101); % Spec. formation
FR(r>sin(piI6))=0;
Currents =fwvind2(FR~boxcar(l 0)*boxcar(l 0)') % Synthesis
[the,ph]-ndgrid(theta,phi);
% Normalized yjx and y,
shiy~csi(ph).*sin(the);
E = freqz2(Currents,shix,shiy); " Normalized E for different y.,and y
x~abs(E).*sin(the).*cos(ph); " Converting the spherical system
y-abs(E).*sin(the).*sin(ph); " into rectangular coordinates
z-abs(E).*cos(the);
mesh(x,y,z) " 3-D plotting

Figure 9. The code for synthesizing a 10 x 10 array.

188 IEEE Antennas and Propagation Magazine, Vol. 49, No. 4, August 2007
188
 Desired pattern

0.5
0
Y .0.5 -0.5
x

Figure 10a. The desired pattern for the 10 x 10 two-dimen-

sional array.

-Nyqulat Rate1 -d=O.5L-arbda

10 1 10-

.1 -0.8 -0.6 -0.4 -0.2 9_, 0.2 0.4 0.6 0.8 1 0 0.5 1 1.5 2 2.5 3
(a)
015
__Halt
th C -d=Lambda
lo- Nyquist Re 0
C. lo~-
in

10J

-1 -0.8 -0.6

10
-0.4 -0.2 ?b) 0.2

Iis
0.4 0.6 0.8

Quarter thet
Nyquiat Rat
1

15-
00 0.5 1 1.5 () 2 2.5 3

01 08 -0.6 0.4 -.02 0 0.2 0.4 0.6 0.8 1 1 5

:0 0.5 1 . (c) 2 -- 3=Lmd
(c) frequency (Hz)

Figure 11. (Top to bottom: a, b, and c) The frequency Figure 12. (Top to bottom: a, b, and c) The array responses of
responses of a 15-tap FIR filter for different sampling rates. a 15-point uniform linear array for different spacings between
the elements.

* .-.

2d
Figure 13a. The convolution of two arrays of two elements with
spacings of 2d and d, respectively.

Sequence 1 Sequence 2 Sequence 3

* I I

<
I
>
ý I>
r
2dr
Figure 13b. The convolution of two sequences with sampling
rates of spacings of 2d and d, respectively.

IEEE Antennas and Propagation Magazine, Vol. 49, No. 4, August 200719 189
by distance 2d is comparable to a sequence having two unit sam-
ples with sampling period 2d, and so on. Just as the time sequence
3 can be considered as a result of a convolution of time sequences
1 and 2, as shown in Figure 13b, the four-element array can be
regarded as the result of a convolution of two arrays of two ele-
ments with spacing 2d and d distances. Since the convolution of
two sequences in the time domain corresponds to multiplication of
frequency responses in the frequency domain, the convolution of
two arrays corresponds to multiplication of tbe radiation patterns of
the respective two-element arrays.
6. Conclusion
Exploiting the similarity between signal-processing theory
and antenna-array theory, it has been demonstrated that with little
effort analysis of antenna arrays can be done using signal- and
image-processing tools. The analogy between FIR filters and arrays
can be effectively used to explain the concepts of the arrays to stu-
dents. Also, synthesis of arrays, to a first approximation, can be
carried out using signal-processing tools in the case of one-dimen-
sional arrays, and image-processing tools in the case of two-dimen-
sional arrays, ignoring the mutual coupling between the elements.
190
7. References
1. C. Bachiller, H. Esteban, A. Belenguer, J. V. Morro, A. Vidal,
and V. E. Boriaet, "Teaching of Advanced Wave-Propagation Phe-
nomena in Open-Space problems and Waveguide Devices Using
MATLAB GUls," IEEE Antennas and PropagationMagazine, 48,
2, April 2006, pp. 128-13 1.
2. C. A. Balanis, Antenna Theory: Analysis and Design, Second
Edition, Singapore, John Wiley, 2001.
3. Panayiotis Ioannides and Constantine A. Balanis, "Uniform Cir-
cular and Rectangular Arrays for Adaptive Beaniforming Applica-
tions," IEEE Antennas and Wireless PropagationLetters, 4, 2005,
pp. 35 1-354.
4. The MATLAB Signal Processing Toolbox Manual, Mathworks
Inc.
5. The MATLAB Image Processing Toolbox Manual, Mathworks
Inc.
6. Edward C. Jordan and Keith G. Balmain, Electromagnetic
Waves and Radiating systems, Second Edition, Prentice-Hall of
India, 2000. 'I
IEEE Antennas and Propagation Magazine, Vol. 49, No. 4, August 2007