See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261355137

Synthesis of cosecant squared pattern in linear antenna arrays using differential

evolution

Conference Paper · April 2013

DOI: 10.1109/CICT.2013.6558248

CITATIONS READS
7 1,100

5 authors, including:

Sujit K Mandal
National Institute of Technology, Durgapur
52 PUBLICATIONS   153 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

On-chip Antenna for telemetry application View project

Development of Time-modulated Antenna Arrays for Present Communication Systems View project

All content following this page was uploaded by Sujit K Mandal on 14 December 2017.

The user has requested enhancement of the downloaded file.

 Proceedings of 2013 IEEE Conference on Information and Communication Technologies (ICT 2013)

Synthesis of Cosecant Squared Pattern in Linear

Antenna Arrays using Differential Evolution
Ananda Kumar Behera, Aamir Ahmad, S. K. Mandal, G. K. Mahanti, Rowdra Ghatak
Dept. of Electronics and Communication Engg.
National Institute of Technology Durgapur
West Bengal, India
[email protected], [email protected], [email protected].

cosecant squared radiation pattern are used for costal and

Abstract— In this paper, a cosecant squared shaped beam air surveillance radars because of its efficient scanning
pattern is synthesized from a uniformly spaced but non- ability in the space. A cosecant squared radiation pattern
uniformly excited linear antenna array where the uniform can be achieved by using multiple feeders and a reflector
spacing of the array elements is considered as half of the antenna.
desired signal wavelength. The shaped beam pattern is
synthesized for the angular beam width of 320 by reducing In [7], Haddadi and Ghorbani applied a weighted
the side lobe level (SLL) to -20.06 dB and almost without alternating reverse projection (WARP) method to
any ripple in the shaped region. To get the desired pattern, synthesize a 12 element linear array with a spacing of
excitation coefficients and the static phase shift of the array 0.5 λ and they obtain the cosecant squared pattern in the
elements are taken as the optimizing parameters and the range about 350 with the ripple level of about 0.5 dB.
differential evolution (DE) algorithm is used as the Although, Orchard et al. [8] introduced a method to
optimizing algorithm. It is further observed that the desired synthesize the power pattern from linear array but the
array pattern is remained unaltered for a band of frequency solution is not the optimum one because, the reduced
on either side of the operating frequency. ripple results in shrinkage of the shaped region and
increases the complexity of the array excitation. A
Index Terms— Linear Array (LA), Side lobe level (SLL), number of optimization methods such as particle swarm
Cosecant Squared Pattern (CSP), Differential Evolution optimization [9], genetic algorithm [10, 11], simulate
(DE).
annealing [12] etc. have been applied for the synthesis of
I. INTRODUCTION antenna array. Hwang synthesized a CSP for a 13 element
non-uniformly slotted rectangular waveguide array [9].
The synthesis of antenna arrays with the desired
radiation pattern is a nonlinear based optimization In this paper Differential Evolution (DE) algorithm is
problem and for its solution many analytical techniques used for optimizing the non uniform excitation
such as Taylor series method and Dolph–Chebyshev coefficients and static phase shifts of a linear array
methods are introduced [1]. But the analytical techniques elements to generate a cosecant squared pattern (CSP) in
in synthesizing shaped beam pattern such as flat-top the far field power pattern of the array. The obtained CSP
pattern, cosecant squared pattern (CSP), are not efficient has the much fewer ripples in the shaped region and also
enough to provide optimum pattern as it produce have a lower side lobe level (SLL). DE algorithm
unwanted ripple in the shaped region. So in order to get introduced by Storn and Price [13] is preferred because it
the best result the stochastic methods [2, 3] are necessary. is a population based, stochastic evolutionary algorithm.
For a symmetrical array synthesis a Genetic Algorithm The algorithm uses a fewer control parameters and also
(GA) method is presented in [4] for the optimization of has the ability to find the true minima not depending
the weights of the array elements with element spacing of
Ȝ/2; where Ȝ is the wavelength of the operating signal. Also a
binary genetic algorithm is used in [5] for thinning
isotropic array. The Cosecant Squared Pattern (CSP) is
preferred due to the fact that, this pattern can provides
high gain near the horizon [6] where the range is higher
but low gain at high-elevation angles where range is
lower. Thus, this pattern is useful for an aircraft system
because when the aircraft is near the horizon its range is
higher whereas at high elevation angle its range is lower.
As a result when it moves from the horizon to towards
the higher elevation, the corresponding aircraft receiver
gets almost constant power at different elevation angles
although its range is changing. Also antennas with

Manuscript received February 15, 2013. Ananda Kumar

Behera1, Aamir Ahmad2, S. K. Mandal3, G. K. Mahanti4, Rowdra Fig.1. Linear array having N-element
Ghatak5. Dept. of Electronics and Communication Engg. National
Institute of Technology Durgapur, West Bengal, India
upon the initial parameters values [13]. In the DE

978-1-4673-5758-6/13/$31.00 © 2013 IEEE 1025

 Proceedings of 2013 IEEE Conference on Information and Communication Technologies (ICT 2013)

algorithm the vector differences of floating-point vector is called mutation. Now this parameter vector is
parameters are found out instead of the uniform or point further mixed with the predefined parameters to produce
crossovers on binary strings which are followed in the the trial vector and this process is called crossover. Then
genetic algorithm (GA). The most significant advantage in the last step, called selection in which trial vector is
of the DE compared to the GA is that it helps us in replaced by the target vector if the trial vector reduces the
optimizing large antenna arrays with more computational values of the cost function than that obtained due to the
efficiency compared to the GA. target vector. To realize the algorithm let the problem is a
function of D number of independent parameters. In this
II. LINEAR ARRAY work for N element array, the number of independent
parameters is 2N in which the first N parameters are the
A configuration of multiple antenna elements normalized amplitude coefficient of the N array elements
arranged along a straight line is known as linear array. A and remaining N parameters are to represent the static
linear array having N-isotropic elements placed along y- phase of each element. Hence, if NP be the population
axis, as shown in figure.1. Generally suitable values of size then the parameter vectors are represented as -
three parameters as amplitude, phase and the inter
element spacing are considered to design an antenna X k ,G = [ x1, k ,G , x2, k ,G ,......xD , k ,G ]k … (2)
array for the desired radiation pattern. The radiation
pattern of an antenna array is the sum of the contribution Where k = 1, 2, 3 ...NP and G is the generation number.
of all the signals radiated by the individual array elements The four basic steps of DE algorithm are as follows -
and can be obtained by the array factor expression as
given in (1), [1]. Step-1 Initialisation:
If the upper and lower bounds for each parameter
 A F = N A e i ( k y n s i n θ + φ n ) ǤǤǤǤǤǤሺͳሻ l u
are defined as xi ≤ xi ,k ,1 ≤ xi for i = 1,2,….D, then
¦ n
n =1 the NP number of D dimensional vectors are randomly
Where, An is the normalized excitation coefficient
generated from the interval [ xil , xiu ] to initiate the
of the n-th element, k = 2π λ , the wave number; θ , the searching process.
angle made by the line joining the observing point and
the origin with the z-axis; yn and φn be the position Step-2 Mutation:
and static phase of the n-th element respectively. Hence In this step the search space is expanded from its
from (1), it is observed that the desired power pattern initial locations. Two vectors X r1,G and X r 2,G are
from a linear array can be synthesized with the suitable
values of the normalized amplitude coefficients, A n and selected randomly for a kth target vector X k ,G . Let

static phase shift φn of the array elements. In this paper, X best ,G is the best vector of the current population then
the effectiveness of DE algorithm is applied to select the the donor vector Vk ,G is formed as in (3).
suitable values of A n and φn for n=1,2,….N to generate
the desired pattern. Vk ,G = X best ,G + F ( X r1,G − X r 2,G ) …… (3)
Where F is the mutation factor and the range of F is [0,
III. DE ALGORITHM 2].
In 1995, Price and Storn commenced the Differential
Evolution (DE) algorithm which is based upon Step-3 Crossover:
differential mutation operator. Practically there are many
problems with different types of objective functions such In this step the solutions from the previous
as non-linear, noisy, flat, non-differentiable, non- generation is obtained successfully. A trial vector U k , G is
continuous, and multi-dimensional or have many local obtained by exchanging the elements of the target vector
minima which are difficult to solve analytically. DE is a X k , G and the donor vector V k , G with a crossover
robust statistical method for cost function minimization
which does not make use of a single nominal parameter probability CR [0, 1].
vector but instead it uses a population of equally
important vectors and is very advantageous to find the Step-4 Selection:
most approximate solution to any of this type of The best value for the next generation is selected as
problems. In DE algorithm the objective function is
sampled by a set of initial points which are chosen ­U k ,G , if f (U k ,G ) ≤ f ( X k ,G )
X k ,G +1 = ® ......... (4)
randomly from the entire search space. Then in the next
¯ X k ,G , if f (U k ,G ) > f ( X k ,G )
step the algorithm adds the weighted difference between
the two randomly selected population vectors to the third For k=1, 2, 3 ... NP
random population vector to generate a new parameter
vector. This process of generating the new parameter

978-1-4673-5758-6/13/$31.00 © 2013 IEEE 1026

 Proceedings of 2013 IEEE Conference on Information and Communication Technologies (ICT 2013)

The target vector compared with the trial vector and the
minimum value is admitted to the next generation. The
above steps are continued until the predefined number of
generation is reached or the desired value of the cost
function is obtained.
Now for this work the DE parameters are as follows.
The array is considered as a 30 element array. Hence, it is
a 60 dimensional problem. The initial population size is
taken as 50 and these are randomly generated by a [50 ×
60] matrix whereas the mutation factor and the crossover
probability are taken as 0.5 and 0.85 respectively.

IV. RESULTS AND DISCUSSIONS

To demonostrat the proposed work, linear antenna
array of 30 eIement is considered with equal element
spacing of 0.5 λ . It is desired that the radiated power
pattern from the array follow the cosecant square function
for at least 300 angular widths and the maximum value of Fig.2. The optimized pattern obtained using DE (solid line) and the actual
SLL is ” -20dB. For the desired pattern, the normalized cosec square plot (dotted line) for the desired range.
amplitude coefficients and the static phase shifts of the
array elements are taken as the optimizing parameters of
Table.1. Normalized amplitude and phase of the individual the DE algorithm. The cost function used for the
antenna array element optimizing algorithm is as given in (5).
Normalized Static phase shift,
Element number
amplitude f cos t = W1 H1δ12 + W2 H 2δ 22 + ∆ sum (5)
(n)
coefficients, (An)
φ n (degree)
Where, δ1 = SLLmax − SLLd , δ 2 = Rmax − Rd ; SLLmax
1 0.0226 122.24
and Rmax are the maximum value of SLL and ripple
2 0.0237 159.94
3 0.1251 155.61
measured at each iteration of the searching process
whereas SLLd and Rd are their corresponding desired
4 0.0556 159.65
5 0.2130 182.66
values. H1 and H2 are the Heaviside step function of the
corresponding terms and ∆ sum is the sum of the
6 0.0058 201.90
7 0.0966 242.32
difference between the sampled values of the array factor
and actual cosecant square values at each point within the
8 0.0999 268.10
desired angular width. The MATLAB code is used to
9 0.1306 285.95 write the algorithm. With the DE parameters as
10 0.0750 313.40 mentioned in section III, the DE algorithm takes 1000
11 0.0119 347.09 iterations to get the desired result. The DE optimized
12 0.2027 019.88 pattern is shown in Fig. 2 and the corresponding optimum
13 0.2079 051.89 values of An and φn are shown in table-1. In Fig. 2, the
14 0.2425 098.95 actual cosecant square plot (dotted line) for the desired
15 0.3240 054.16 range is also shown. From Fig. 2, it is observed that the
16 0.1774 053.32 DE optimized CSP approximately follows the actual plot
17 0.1874 056.92 for the angular width of 330 with the range (98-131)
degrees. The maximum value of the side lobe level is also
18 0.7759 099.41
reduced to ņ20.06 dB and almost without any ripple in
19 0.9655 114.33 the desired pattern. It is further observed that with the
20 0.9804 130.86 variation of the uniform element spacing from 0.3 λ to
21 0.9956 288.43 0.7 λ the obtained pattern is remain unaltered with SLL
22 0.9852 076.73 less than -20dB. Hence this pattern is applicable for the
23 0.9527 090.77 frequency range of 1.67f0 to 0.71f0. However as seen
24 0.8825 092.86 from table-1, the dynamic range ratio (DRR) of
25 0.6870 326.21
amplitude is quite higher (170.4). But when the elements
with lowest amplitude coefficient (6th element) is
26 0.6397 091.09
discarded the corresponding values of DRR becomes
27 0.4839 343.88 almost half (83.66) and the resultant pattern is not
28 0.3902 094.29 affected for that.
29 0.2212 079.21
30 0.1358 356.45 V. CONCLUSION
The differential evolution algorithm is used
successfully for synthesizing a cosecant square pattern in
the radiation pattern of a uniformly spaced but non-

978-1-4673-5758-6/13/$31.00 © 2013 IEEE 1027

 Proceedings of 2013 IEEE Conference on Information and Communication Technologies (ICT 2013)

uniformly excited linear antenna arrays. The obtained [5] Haupt, R. L., “Thinned arrays using genetic algorithms,” IEEE
pattern is also compared to the cosec square plot in the Trans. Antennas and Propagation, Vol. 42, No. 7, 993–999, 1994.
desired region. The side lobe level is reduced to ņ20.06 [6] Hui, K.Y., Luk, K.M.: ‘Design of wideband base station antenna
arrays for CDMA 800 and GSM 900 systems’, Microw. Opt.
dB with almost zero ripples at the desired region. The echnol. Lett., 2003, 39, (5), pp. 406–409
synthesized pattern of the antenna is much more helpful [7] Haddadi, A. Ghorbani, A.:‘Cosecant-squared pattern synthesis
at the sea bank for effective communication. using a weighted alternating reverse projection method’ IET
Microw. Antennas Propag., 2011, Vol. 5, Iss. 15, pp. 1789–1795
REFERENCES [8] Orchard, H.J., Elliott, R.S., Stern, G.J.: ‘Optimizing the synthesis
of shaped antenna patterns’, IEEE Proc. (London) H, 1984, 132,
[1] Balanis, C., Antenna Theory Analysis and Design, second edition, (1), pp. 63–68
John Wiley and Sons, Inc., 1997.
[9] Hwang, K.C.: ‘Optimization of a non-uniformly slotted
[2] Ares-Pena, F., “Application of genetic algorithms and simulated rectangular waveguide antenna’, Microw. Opt. Technol. Lett.,
annealing to some antenna problems,” Electromagnetic 2008, 50, (7), pp. 1935–1937
Optimization by Genetic Algorithms, Y. Rahmat-Samii and E.
Michielssen (Eds.), Wiley & Sons, New York, 1999. [10] Marcano, D., Duran, F.: ‘Synthesis of antenna arrays using genetic
algorithms’, IEEE Antenna Propag. Mag., 2000, 42, (3), pp. 328–
[3] Murino, V., A. Trucco, and C. S. Regazzoni, “Synthesis of 332
unequally spaced arrays by simulated annealing,” IEEE Trans.
Signal Processing, Vol. 44, 119–123, 1996. [11] Donelli, M., Coarsi, S., Natale, F.D., Pastorino, M., Massa, A.:
‘Linear antenna synthesis with hybrid genetic algorithm’, Prog.
[4] Yan, K. and Y. Lu, “Sidelobe reduction in array-pattern synthesis Electromagn. Res., 2004, 49, pp. 1–22
using genetic algorithms,” IEEE Trans. Antennas and Propagation,
Vol. 45, No. 7, 1117–1122, 1997. [12] Farhat, N.H., Bai, B.: ‘Phased-array antenna pattern synthesis by
simulated annealing’, IEEE Proc., 1987, 75, (6), pp. 842–844
[13] Storn, R. and K. Price, “Differential evolution—A simple and
efficient heuristic for global optimization over continuous spaces,”
Journal of Global Optimization, Vol. 11, No. 4, 341-359, 1997.

978-1-4673-5758-6/13/$31.00 © 2013 IEEE 1028

View publication stats