QPSO for Synthesis of Linear Array of Isotropic

Antennas to Generate Flat-top Beam Including

Multiple Null Placement

Hemant Patidar G. K. Mahanti

Department of Electronics and Communication Engineering, Department of Electronics and Communication Engineering,
National Institute of Technology National Institute of Technology
Durgapur 713209.India Durgapur 713209.India
[email protected] [email protected]

Abstract— The authors proposed an evolutionary algorithm using particle swarm optimization has been described in [8]
based on quantum particle swarm optimization to generate flat- and antenna array pattern synthesis with wide null control
top beam from a linear array of isotropic antennas. An effort is using enhanced particle swarm optimization has been
also made to reduce the side lobe level including multiple null described in [9]. For reducing the computational time to a
placements with minimum variation in ripple of flat-top beam. great extent utilization of iterative fast-Fourier transform for
This is generally done by changing excitation current amplitude null synthesis of scanned linear antenna array with minimum
of the elements. Phases of some of the elements are kept either at side lobe level was also described in article [10].
zero degree or at 180 degree. The generated pattern is broadside
in the vertical plane. One example has been presented with In this paper, the authors have applied quantum particle
twenty-six isotropic antennas. Array factor is calculated by using swarm optimization (QPSO) [11-13] for synthesis of linear
inverse fast Fourier transform to reduce the computational time array of isotropic antennas to generate flat-top beam with
significantly. The results obtained from the given example desired side lobe level including multiple null placement.
demonstrate the effectiveness of the proposed method. Although,
the proposed method is developed and applied to a linear array II. THEORY
of isotropic antennas; however, the principle can easily be
A linear array of N isotropic antennas [1, 10, 11] is
extended to other type of arrays.
considered along the y-axis. It is assumed that elements in the
Index Terms— Flat-top beam; inverse fast Fourier transform; array are uncoupled and equally spaced at a distance d apart as
linear array; quantum particle swarm optimization; ripple; side shown in Fig.1. Total generated field of the antenna array is
lobe level; single null depth. given by the vector summation of the field radiated by
individual element. The free space [1, 10, 11] far-field pattern
I. INTRODUCTION FFP (u) in the vertical plane (YZ-plane) is given by (1):
Antenna arrays are used to generate highly directive
N
∑ Bnei(n−1)kdu
radiation pattern and this property makes its huge application
in wireless communication, radar etc. Flat-top beam finds FF P(u ) =
application in broadcasting from satellite for uniformly
illuminating more than one sector of region. It is also used in n =1 (1)
wireless communication systems, under-water acoustics etc.
By controlling the property of array factor like changing the Where n= the element number, λ= the wavelength, Bn= the
geometrical configuration and changing the antenna excitations (current) of the individual antenna elements, i= the
parameters, we can change the shape of antenna array pattern imaginary unit, k=2π/λ= the wave number, d= inter element
[1-2]. Several methods for generating the shaped beam pattern spacing and u=sinθ, where θ is the angle of the far-field point
for antenna arrays have been described in [3-7].Woodward [3] measured from z-axis as shown in Fig.1.
familiarized the analytical method to generate the shaped
beam pattern. Steyskal [4] and Stutzman [5] also have Normalized absolute far-field in dB is written in (2) as
proposed several different methods to shape the antenna array follows:
pattern. A simple method for linear array to provide solution
for pattern is described in [6]. Main parameters of flat-top ⎡ F P (u) ⎤
beam are Side lobe level, ripple and transition width which is FFn (u) = 20 log10
⎢ F ⎥
contradictory to each other is detailed in [7]. To place null ⎢ ⎥ (2)
along the direction of interference [8, 9], null steering method ⎢⎣ FF P (u) max ⎥⎦
finds applications in radar and communication systems. Linear
array synthesis with minimum side lobe level and null control

978-1-4799-6761-2/15/$31.00 ©2015 IEEE 46

 In this paper our objective is to find thee set of excitation III. QUANTUM M PARTICLE SWARM
current amplitude of the elements using quantum particle OPTIMIZATION
N ALGORITHM
swarm optimization (QPSO) to obtain the desired values. The
QPSO algorithm [11-13] based on the fundamental theory
proposed algorithm will minimize the following fitness
of particle swarm and properrty of quantum mechanics is a
function to generate the flat-top beam patttern with desired
novel optimization algorithm m. All particles show the
value of SLL, ripple and single null at two poositions.
characteristics of quantum behavior. However, QPSO
algorithm’s global optimizatioon performance is better than
standard PSO algorithm’s becaause of QPSO is stated only by
the position vector; there is noo velocity vector in QPSO. The
dynamic behavior of the particlle in QPSO is different from that
of the particle in standard PS SO. Position and velocity of a
particle cannot be found simultaneously according to
uncertainty principle of quanttum world, each particle finds
many position in search space with
w certain probability.
Several steps, texts and eqquations of QPSO algorithm are
taken from paper [11] and reprooduced below:
“Step1: Initialize a populaation of particles with random
positions between the maximuum and the minimum operating
limits of the generators in the D-dimensional
D space.
Step2: Evaluate the fitness value of each particle.
Step3: Compare the personnal best (pbest) of every particle
with its current fitness value. If the current fitness value is
Fig.1. Geometry of an N-element linear array of isotroopic antennas along the better, then assign the current fitness
f value to pbest and assign
Y-axis. the current coordinates to pbestt coordinates.

[ ]
Step4: Calculate the meann best position (mbest) of all K
Fitness = rr1 × F12 + rr2 × F2 + rr3 × F3 + rr4 × F4 particles using the following eqquation given below:
K
1
(3)
mbest =
K
∑ pbest
i =1
i
Where
Step5: Determine the currennt best fitness value in the whole
⎧SLLo − SLLd , if → SLL
Lo > SLLd population and its coordinattes. Compare the global best
F1 = ⎨ (4)
⎩0, if → SLLo ≤ SLLd position (gbest) of the currennt particle population with the
previous one. If the current best
b fitness value is better than
⎧⎪ SN max
o d
− SN max o
, if → SN max
m
d
> SN max global best (gbest), then assignn the current best fitness value to
F2 = F3 = ⎨ gbest and assign the current cooordinates to gbest coordinates.
⎪⎩0, if → SN max
o d
≤ SN max Step6: Determine the vecttor local focus of the particle
(5) using the following equation givven below:
u =0.4 2 (
Vf idt = rand1tid ∗ ( pbestt id ) + 1 − rand1tid ∗ ( gbest ) )
F4 = ∑ (OP − DP)
u =0
(6)
Step7: Update position (Yid) of the d-th dimension of the i-
th particle using the following equations:
e
The coefficients rr1, rr2, rr3 and rr4 deteermine the relative t

weight (importance) given to each term in (3). The value of Yidt = Vf idt + (−1) ceil ( 0.5+ rand 2id ) × δ
rr1=250, rr2=1, rr3=1 and rr4 =1 has been asssumed by trial and
× mbest − Yidt −1 × log e (1 / rand 3tid )
error for generation of power pattern. SLL Ld and SLL o are
desired and obtained values of side lobe leevel (SLL) for F1
o d
If Yidt < Ymin
d

from u =0.4 to 1. SN max and SN max are obbtained and desired

value of maximum null depth at u= 0..4720 for F2 and Then
u=0.9209 for F3. In F4, OP and DP are the amplitude of
obtained and desired pattern from u =0 to t 0.4 for flat-top Yidt = Ymin
d
+ 0.25 ∗ raand 4 idt ∗ Ymax
d d
(
− Ymin ) (7)
region.

47
 If Yidt > Ymax
d Through this mapping procedure, the sampling in u domain
is transformed into v domain. Note that Equation (9) is very
Then identical to standard definition of one dimensional inverse fast
Fourier transform (IFFT), which indicates that the array pattern
Yidt = Ymax
d
− 0.25 ∗ rand 5 tid ∗ (Ymax
d d
− Ymin ) (8) can be directly computed through an IFFT operation on the
excitations Bn. .It reduces computational complexity
Where t is the current generation, rand1, rand2, rand3, significantly. IFFT used in the program is 4096-point IFFT
rand4 and rand5 are uniform random numbers between 0 and filled with zeros if excitation current has less than 4096 points.
1. Equations (7) and (8) have been introduced to limit the
(d
position along each dimension within Ymax , Ymin
d
)
if they try
V. SIMULATED RESULTS
To synthesize flat-top beam pattern in the principal vertical
to cross the desired domain of interest. To stop the particles
from explosion sometimes this clipping techniques are plane (YZ-plane) a linear array of 26 isotropic antennas with
necessary. equal spacing of 0.5λ between any two consecutive elements
has been considered along the y-axis with specified side lobe
Step8: The steps 2-7 above are uninterrupted until a stop level of -20dB, specified ripple of 0 dB from u=0 to 0.4 of
criterion is satisfied; ordinarily it is stopped when there is no desired pattern and specified null depth of -42 dB or less at
further update of best fitness value or maximum number of u=0.4720 and u=0.9209. Desired results are obtained by
iteration reached. taking asymmetric current excitation amplitude of the
The parameter Vf is the local attractor of each particle. δ is elements to generate flat-top shaped pattern including null
the contraction and expansion coefficient that control the placement.
convergence speed and performance of the particle. In order to
avoid premature convergence, mean best position (mbest) is QPSO is run for 3000 iterations with a particle size of 60
regarded as the barycenter of all particles. Ceil (A) rounds the to generate the flat-top beam pattern in vertical plane. A set of
elements of A to the nearest integers greater than or equal to A. 26 excitation current amplitude values is obtained.
The value of δ is changed with iteration by the relation given Matlab has been used here for writing the program for
below: synthesis of the linear array antenna. The computational time is
measured with a PC with Intel core2 duo processor of clock
(max gen − t ) frequency 2.93GHz and 4 GB of RAM. Table I shows the
δ = (δ max − δ min ) × + δ min desired and obtained results using QPSO. Table II shows the
(max gen − 1)
excitation current amplitude distribution from -0.2 to 1. Fig.2
Where δmax =1 and δmin =0.7 are the maximum and shows the desired pattern of flat-top beam in u domain. Fig. 3
minimum value of δ, maxgen is the maximum number of shows the best fitness value versus iteration number obtained
iteration (generation) and t is the current generation”. by QPSO. Fig. 4 shows the current amplitude distribution to
generate the flat-top beam pattern with asymmetric excitations
IV. ARRAY FACTOR AND INVERSE FAST between -0.2 to 1. Fig. 5 shows the normalized power pattern
FOURIER TRANSFORM in dB (flat-top beam pattern) including multiple null
An inverse Fourier transform relationship exists between placements for a linear antenna array.
the array factor (FFP) and the element excitations for an array
TABLE I. DESIRED AND OBTAINED RESULTS
having an equal spacing between the elements. This property
used in the IFFT [10].
Array factor (FFP) in Y-Z plane is given by Design Parameters Desired Value Obtained Value
Side lobe level in dB
(u=0.4 to 1)
N -20 -18.16

FF P(u) = ∑ Bnei(n −1)kdu

Ripple in dB
(u=0 to 0.4)
Single null in dB
0 1.16

n =1 (u=0.4720) -42 -42.79

Let N Single null in dB
v =1+ kdu , (u=0.9209)
2π -42 -43.02
Computation time
Then (Seconds) …………… 606.98

N 2π
(n−1)(v −1)
i
FF P(v) = ∑ Bne N (9)
n=1
FFP (v) =IFFT (Bn, no of sample points)

48
 TABLE II. EXCITATION CURRENT AMPLITUDES

C u rren t A m p litu d e D istrib u tio n

1
Element Current Element Current
Number Amplitude Number Amplitude
0.9
1 -0.1986 14 -0.1973 0.8
2 -0.1988 15 0.0533 0.7
3 -0.1940 16 0.3593 0.6
4 -0.0313 17 0.3287 0.5
5 0.1284 18 0.0922
0.4
6 0.2188 19 -0.1795
7 0.3041 20 -0.1822 0.3
8 0.4914 21 -0.1009 0.2
9 0.5466 22 0.1010 0.1
10 0.5732 23 0.2303 0
11 0.3403 24 0.1014 -0.1
12 -0.0387 25 -0.0387
-0.2
13 -0.1984 26 -0.1313 0 5 10 15 20 25 30
Element Number
0
-1 Fig.4. Excitation Current amplitude distribution versus element number
-2 obtained by QPSO.
-4
Desired Pattern in dB

N o rm alized P o w er P attern in d B
-6 0
-8
-10 -10
-12
-20
-14
-16
-30
-18
-20 -40
-22
-1 -0.4 0 0.4 1
-50
u
Fig.2. Desired flat-top beam pattern. -60
-1 -0.5 0 0.5 1
x 10
4
u
14
Fig.5. Normalized power pattern in dB using QPSO with two nulls of
12 prescribed depth -42 dB at u=0.4720 and 0.9209.
Best Fitness Value

10 VI. CONCLUSIONS
The paper applied quantum particle swarm optimization
8 algorithm for synthesis of a linear array of isotropic antennas
to generate flat-top beam with desired side lobe level (SLL),
6 multiple nulls at two locations and fixed ripple. The results
obtained using simulation depicts a great deal between the
4 desired and obtained specifications. The advantage of using
this technique is that it finds the suitable excitation current
2 amplitude distribution using QPSO to obtain the desired
values with minimum variation. There would be a large scope
0 of work by doing comprehensive study on the parameter i.e.,
0 500 1000 1500 2000 2500 3000
spacing between the element, length of antenna and
Iteration Number considering mutual coupling effects to generate the flat-top
beam pattern. Results obtained by synthesis of a linear array
Fig.3. Best fitness value versus iteration number obtained by QPSO. of isotropic antennas have illustrated the performance of this
proposed technique.

49
 REFERENCES
[1] Balanis, C. A., Antenna Theory: Analysis and Design, 2nd ed.
Singapore: John Wiley and Sons (Asia), 2003.
[2] R. L. Haupt, and D. H. Werner, “Genetic algorithms in
Electromagnetics”, IEEE Press, 2007.
[3] Woodward P. M., “A method of calculating the field over a plane
aperture required to produce a given polar diagram”, J.Inst. Elect. Eng.
vol. 93, pp. 1554-1558, 1947.
[4] Steyskal H., “On antenna power pattern synthesis’’, IEEE Trans
Antennas Propagat. vol. Ap-18, pp. 123-124; Jan. 1970.
[5] Stutzman W. L.,“Synthesis of shaped-beam radiation patterns using the
iterative sampling method”, IEEE Trans. Antennas Propgat., vol. Ap-19,
pp. 36-41, Jan. 1971.
[6] Woodward P. M., Lawson J. D.,“The theoretical precision with which an
arbitrary radiation pattern may be obtained from a source of finite size”,
J.IEE (London), vol. 95, pp. 363-370, Sept. 1948.
[7] Stutzman W. L., Thiele G. A.,“Antenna Synthesis in Antenna Theory
and Design”, Wiley Publication, Second Ed., 1997.
[8] M. M. Khodier and C. G. Christodoulou, “Linear array geometry
synthesis with minimum sidelobe level and null control using particle
swarm optimization”, IEEE Trans. Antennas Propag., vol. 53, no. 8, pp.
2674–2679, Aug. 2005.
[9] M.A. Mangoud and H.M. Elragal, “Antenna array pattern synthesis and
wide null control using enhanced particle swarm optimization”,
Progress in Electromagnetics Research B, vol. 17, pp. 1-14, 2009.
[10] B.C. Mahato, G.K. Mahanti and A.K. Panda, “Null synthesis of scanned
linear array antenna with minimum side lobe level and fixed dynamic
range ratio using iterative fast Fourier transform”, ICTACT Journal on
communication technology, vol. 2, no. 3, pp.377-381, 2011.
[11] R. Muralidaran, A. Vallavaraj, Hemant Patidar and G. K. Mahanti,
“QPSO for failure correction of linear array antenna including wide Null
placement”, ICTACT Journal on Communication Technology, vol. 05,
issue: 03, pp.959-962, September 2014.
[12] J. Sun, W.B. Xu, B. Feng, “A global search strategy of quantum
Behaved particle swarm optimization,” in Proc. IEEE Conference On
Cybernetics and Intelligent Systems, Singapore, pp.111–116, 2004.
[13] Said M. Mikki and Ahmed A. Kishk, “Quantum particle swarm
optimization for Electromagnetics,” IEEE Transactions on Antennas and
Propagation, vol. 54, no. 10, pp. 2764-2775, October 2006.

50