Вы находитесь на странице: 1из 4

Design of Discrete Phase-only Dual-beam Array Antennas with Minimum Dynamic Range Ratio

T.K.Sinhamahapatra l , A.Ahmed l , G.K. Mahanti l , N.Pathak2 and A.ChakrabartY


IDepartment of Electronics and Communication Engineering National Institute of Technology Durgapur-713209, India (E-mail: gautammahanti@yahoo.com) 2Department of Electronics and Communication Engineering Bankura Unnayani Institute of Engineering Pohabagan, Bankura-722146, India (E-mail: narendra.pathak@rediffmail.com) 3Department of Electronics and Electrical Communication Engineering Indian Institute of Technology Kharagpur-721302, India (E-mail: bassein@ece.iitkgp.ernet.in)
Abstract-In this paper, we describe the use of real-coded Genetic Algorithm (GA) for design of reconfigurable dual-beam linear isotropic antenna arrays with phase-only control of discrete phase shifters. The problem is to find a fixed amplitude distribution that will generate two broadside symmetrical beams in vertical plane: a pencil beam with zero phases and a flat-top beam with discrete phases of a five-bit discrete phase shifter, with or without minimizing dynamic range ratio (amaxlamin) of excitation amplitude distribution.

I.

INTRODUCTION

Phase-only reconfigurable array antennas capable of radiating multiple radiation patterns with a fixed amplitude distribution are used in many applications. In general, the design and implementation of the circuitry feeding the array are simpler if the element excitations corresponding to different patterns differ only in phase than if they also differ in amplitude. Several methods of synthesizing phase-only multiple pattern antenna arrays have been described [1-7]. The synthesis of phase-only multiple radiation patterns with pre-fixed amplitude distributions [1] is reported with modified Woodward-Lawson technique. Bucci et al. [2] proposed the method of projection to synthesize reconfigurable array antennas with asymmetrical pencil and flat-top beam patterns using common amplitude and varying phase distributions. The design of a phase-differentiated reconfigurable array has been described [3] using particle swarm optimization in angle domain. Baskar et al. [4] synthesized reconfigurable array antennas with phase-only control of 6-bit discrete phase shifter and continuous amplitude distribution using generalized generation gap model genetic algorithm and better synthesis results were obtained, as compared to continuous phase excitations with subsequent quantization. Synthesis of 978-1-4244-1864-0/07/$25.00 2007 IEEE

continuous phase-only reconfigurable array is described in [5]. Design of phase-differentiated multiple pattern antenna arrays [6] have been described based on simulated annealing optimization technique. Beam reconfiguration of linear arrays of parallel dipoles has been discussed [7] with the help of mechanical displacement of a parasitic array in front of an active one. On the other hand, an evolutionary algorithm such as genetic algorithm (GA) is a global iterative optimizer that performs population-based probabilistic searches with an ability to escape from local optima. In this paper, optimization technique using real-coded GA [8] with some variation in its implementation is applied for the synthesis of a reconfigurable symmetrical dual-beam linear isotropic antenna array in sine space (sine of far-field angle) with a 5-bit discrete phase shifter with or without minimizing dynamic range ratio (DRR) of excitation amplitude distribution. However, minimizing DRR to a lower value makes little compromise on the design specifications. Patterns are optimized in sine space (sine of far-field angle) instead of angle space [3, 4]. The optimized phase excitations obtained by this method can be directly implemented without further quantization. II. PROBLEM FORMULATION

The design of a reconfigurable dual-beam antenna array leads to finding a common amplitude distribution that can generate either a pencil beam or a flat-top beam, when the phase distribution of the array is modified appropriately. All excitation phases are set at 00 to generate a pencil beam, and are varied in the range -180 0 ~ <p ~ 1800 in steps of 3600 /2 5 or

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on March 24, 2009 at 01:48 from IEEE Xplore. Restrictions apply.

11.25 of a 5-bit discrete phase shifter to form a flat-top beam pattern [3]. We consider a linear array of N isotropic antennas that are assumed uncoupled and equally spaced a distance d apart along the Y-axis. This is shown in Fig.I. The free space [9] far-field pattern F (u) in the principal plane (YZ-plane) is given by eqn. (1):
F(u) = Lane1qJne
n=l
N .

For the dual-beam array optimization, the fitness function must cover the entire array radiation pattern. The fitness function to be minimized for dual-beam array optimization problem is expressed as follows:

i(n-l) 21l' du A

(1)

Where n the element number, A the wavelength, qJn the excitation current phases of the elements, an the excitation current amplitudes of the elements, i the imaginary unit, d is the inter-element spacing, and u=sinB, e being the polar angle of far-field measured from broadside (-90 to +90). The amplitude and phase distribution are assumed symmetric with respect to the center of the array. Normalized power pattern in dB can be expressed as follows:
!F(u)! P(u) = 1OioglO [ F(u) max

Where the superscript p is meant for the design specification of pencil beam and the superscript f is meant for the design specification of flat-top beam pattern. Pj , d and Pj represent respectively the applicable desired and calculated value of each design specification, as given in Table1 and Table2. The fourth term in the second summation in eqn. (3) is a ripple parameter for the flat-top beam pattern. The third term in eqn. (3) is dynamic range ratio of excitation amplitude, defined as:

I I

]2 = 201oglO[I !F(u)1 I

DRR=(an)max/(an)min VnE [l,N]

(4)

F(u) max

(2)

The lower the fitness, the more fit the array to the desired specifications. The desired maximum ripple level (RL) in the entire coverage region (-0.19~ u ~ 0.19) is not to exceed 0.5 dB from the peak value of 0 dB. The difference terms in connection to side lobe level and ripple level in fitness function eqn. (3) are made zero when their respective calculated values are less than their desired values by multiplying appropriate Heaviside step function with these terms whose value is unity when the calculated value is greater than or equal to the desired value and zero elsewhere. The weighting factors associated with all the terms in eqn. (3) are made equal to unity. III. OVERVIEW OF REAL-CODED GA Genetic Algorithm [8] is an iterative stochastic optimizer that works on the concept of survival of the fittest motivated by Darwin, using methods based on the mechanics of natural genetics and natural selection to construct search and optimization procedures that best satisfies a predefined goal. Real-coded GA uses floating-point number representation for the real variables and thus is free from binary encoding and decoding. It takes less memory space and works faster than binary GA. The real-coded GA is summarized as follows:
Step 1: Randomly generate an initial population of P individuals within the variable constraint range. Step 2: Evaluate the fitness of the individuals from the fitness function. Step 3: Select the superior individuals using nonlinear ranking [8] and place them in the mating pool. Numbers of individuals in the mating pool are same as P in order to accommodate more and more copies of superior individuals in the new population. Step 4: Individuals so called parents placed in the mating pool are now allowed to mate followed by mutate using arithmetic crossover and non-uniform mutation [8] respectively. In the

FAR-FIELD

VPS

PDN

Fig. 1. Geometry of a dual-beam linear antenna array, VPS is variable phase shifter and PDN is power divider network

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on March 24, 2009 at 01:48 from IEEE Xplore. Restrictions apply.

crossover process, two parents produce two children. Subsequent mutations of the parents add diversity to the population and explore new areas of parameter search space. Select C pairs of parents at random to participate in crossover to produce C pairs of offspring without replacement of the parents. Select M number of parents at random from the mating pool to participate in mutation to produce M number of offspring without replacement of the parents.

In case of dual-beam pattern with DRR minimization, GA is run for 2000 generations. Excitation amplitude dynamic range ratio is found to be 5.18 and the final fitness value is found to be 5.287. Results are shown in Table 2. In this case, ripple level is 0.825dB, little higher than the desired one. Amplitude and 5-bit discrete phase distributions in degree are shown in Table 3. Fig.3 shows normalized absolute power patterns in dB for dual-beam array with DRR minimization. In our design, phases are not required to generate pencil Step 5: Score all individuals again. These include P parents, C beam compared to [1, 2] where it is required. Moreover, pencil pairs of crossover children and M candidate mutation children. and flat-top beams are also asymmetrical in [2] resulting The best scoring P individuals survive to the next generation. asymmetrical amplitude and phase distributions that complicate This step has been introduced to prevent loosing the best-found the feed network further. Coverage region near zero dB for calculating ripple of flat-top beam is not mentioned in [3,4]. In individuals by chance because of crossover and mutation. our case, they are all clearly mentioned. Our method of Step 6: Repeat steps 2-5 until a stopping criterion, such as a calculating ripple of flat-top beam is quite different from [3, 4]. sufficiently good solution being discovered or a maximum Moreover, the authors [3] did not optimize dynamic range ratio number of generations being completed, is satisfied. The best (DRR) of excitation amplitude distribution. We not only scoring individual in the population is taken as the final minimize DRR but also synthesize the dual-beam array with a five bit discrete phase shifter instead of a six bit discrete phase answer. shifter as used in [4]. Equally spacing pattern points in sine IV. RESULTS AND DISCUSSIONS space provide a more uniform sampling and less number of sampling points than angle space [3,4], which in tum reduces We consider a uniformly spaced d=0.5A apart linear array of the complexity of optimization. The optimization technique 20 isotropic elements in order to generate a symmetric pencil used here is very simple and different from [3,4]. beam and a flat-top beam. Because of symmetry, only ten TABLE 1 amplitudes and ten phases are to be optimized. For DESIRED AND OBTAINED RESULTS WITHOUT DRR MINIMIZATION convenience, the genetic algorithm is designed to provide vectors of 20 real values between zero and one. Each 20Pencil beam Flat-top beam Design parameters element vector is mapped to 10 amplitude and 10 phase Desired Obtained Desired Obtained weights, where the first 10 values from the vector are scaled to Side lobe level -25.00 -24.99 -25.00 -25.00 desired amplitude weight range and the second 10 values are (SLL, in dB) scaled to desired phase weight range. All phases are restricted Half-power beamwidth 0.498 0.120 0.124 0.500 (HPBW, in u-space) to lie between -180 and 180 degrees, as mentioned before, and Beamwidth at SLL amplitudes between 0 and 1. 0.720 0.300 0.306 0.700 (in u-space) For design specifications as given in Tablel and Table2, GA Ripple N/A N/A 0.500 0.459 is run independently ten times, each time with different random (in dB, -0.19:5 u :50.19) number but fixed number of generations with an initial TABLE 2 population of 200 and nonlinear ranking with probability of DESIRED AND OBTAINED RESULTS WITH DRR MINIMIZATION 0.35 for selecting the best individual. The shape parameter in non-uniform mutation is taken to be 3.0. Crossover and Pencil beam Flat-top beam mutation operators are called sixth times every generation Design parameters Desired Obtained Desired Obtained respectively in order to ensure that only six pairs of parents Side lobe level -25.00 -25.00 -24.98 -25.00 take part in crossover and six numbers of parents take part in (SLL, in dB) mutation in stead of all. This will reduce the overall Half-power beamwidth 0.120 0.464 0.116 0.500 computational time in optimization considerably. Best run (HPBW, in u-space) Beamwidth at SLL having the lowest fitness value is taken as final. 0.300 0.272 0.700 0.692 (in u-space) In case of dual-beam pattern without DRR minimization, GA Ripple is run for 1500 generations. Excitation amplitude dynamic N/A N/A 0.500 0.825 (in dB, -0.19:5 u :50.19) range ratio (DRR) is found to be 9.17 and the final fitness value is 0.00061. Results are shown in Table 1. There is a very good agreement between desired and obtained results using GA. Corresponding common amplitude and 5-bit discrete phase distributions in degree are shown in Table 3. Fig.2 shows normalized absolute power patterns in dB for dual-beam array without DRR minimization.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on March 24, 2009 at 01:48 from IEEE Xplore. Restrictions apply.

TABLE 3 AMPLITUDE AND PHASE DISTRffiUTIONS IN DEGREE

v. CONCLUSIONS
An optimization technique based on real-coded Genetic Algorithm for the design of a reconfigurable dual-beam array antenna with or without DRR minimization has been proposed in this paper. The effect of coupling between antenna array elements is reduced by minimizing the excitation amplitude dynamic range ratio, with little compromising on the design specifications. The method presented here takes discrete phases directly into account during synthesis. This leads to better synthesis results, compared to conventional methods where phases are subsequently quantized [4]. Results clearly show a very good agreement between the desired and GA synthesized one with or without minimizing DRR, even with a 5-bit discrete phase shifter instead of a continuous phase shifter [2-3] and a 6-bit discrete phase shifter [4]. This demonstrates the effectiveness of the proposed procedure. The patterns, amplitude and phase distributions are all symmetric in nature, and this symmetry greatly simplifies the feed network. Moreover, for practical applications, the design of reconfigurable antenna arrays with fixed amplitude and quantized phases is preferred in order to keep costs low, control simple and maintain accuracy. This design method can be used directly in practice to synthesize reconfigurable linear isotropic antenna arrays with phase-only control of discrete phase shifters. Results for a linear isotropic antenna array have illustrated the performance of this proposed technique. This method is very simple and can be used in practice to synthesize multiple beam patterns with continuous as well as discrete phase shifter.

Element Number
1&20 2&19 3&18 4&17 5&16 6&15 7&14 8&13 9&12 10&11

Unoptimized DRR Phase (5-bit) Amp. (De2ree) 146.25 0.0967 101.25 0.1366 56.25 0.2097 11.25 0.3175 -22.50 0.4377 -45.00 0.6135 0.5720 -67.50 -101.25 0.6623 -135.00 0.7277 -146.25 0.8874

Optimized DRR Phase (5-bit) Amp. (Degree) -157.50 0.1510 -146.25 0.1545 -90.00 0.2340 -67.50 0.3828 -45.00 0.5268 -22.50 0.5684 0.00 0.5842 0.5899 33.75 0.6912 67.50 78.75 0.7833

(/
I
J

-1.1

a=doI

.tl

Fig. 2. Normalized absolute power patterns in dB for dual-beam array without DRR minimization. Dashed line, pencil beam pattern; solid line, flat-top beam pattern.

ACNOWLEDGMENT The authors are grateful to ISRQ-Kalpana Chawla Space Technology Cell, Indian Institute of Technology, Kharagpur for supporting this work. REFERENCES

,
1

n .~;
,
1.2

1\ (

..

c~
I \

,1'
'.'.

[1] M. Durr, A. Trastoy, and F.Ares, "Multiple-pattern linear antenna arrays with single prefixed amplitude distributions: modified Woodward-Lawson synthesis", Electronics Letters, Vol.36, No.16, pp.1345-1346, 2000. [2] O.M. Bucci, G. Mazzarella, and G.Panariello, "Reconfigurable arrays by phase-only control", IEEE Trans. Antennas and Propag., Vol.39, No.7,pp. 919-925, 1991. [3] D. Gies, and Y.Rahmat-samii, "Particle swarm optimization for reconfigurable phase differentiated array design," Microwave Opt. Tee/lIlol. Lett, Vol.38, pp.168-175, 2003. [4] S. Bhaskar, A. Alphones, and P.N. Suganthan, "Genetic Algorithm based design of a reconfigurable antenna array with discrete phase shifter", Microwave Opt. Technol. Lett., Vol. 45, pp.461-465, 2005. [5] G.K. Mahanti, A. Chakraborty, and S. Das, "Design of Phase-Differentiated Reconfigurable Array Antennas with Minimum Dynamic Range Ratio," IEEE Antennas and Wireless Propagation Letters, vol.5, pp.262-264, 2006. [6] X. Diaz, 1. A. Rodriguez, F. Ares, and E. Moreno, "Design of Phasedifferentiated multiple-pattern antenna arrays", Microwave Opt. Technol. Lett., Vol.26, pp.52-53, 2000. [7] I.A. Rodriguez, A. Trastoy, 1. C. Bregains, F. Ares, and G. Franceschetti, "Beam reconfiguration of linear arrays using parasitic elements," Electronics Letters, Vo1.42, pp.l31-133, 2006. [8] Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, Springer-Verlag, Berlin, 1999. [9] C.A. Balanis, Antenna Theory: Analysis and Design, Second Edition, John Wiley and Sons (Asia), Singapore, 2003.

l ,..r . . f . . ~ ~ 11 II \
0..5

1'

fr\.J. r J

-1.1

0
a~dDI

Fig. 3. Normalized absolute power patterns in dB for dual-beam array with

DRR minimization. Dashed line, pencil beam pattern; solid line, flat-top beam
pattern.

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY DURGAPUR. Downloaded on March 24, 2009 at 01:48 from IEEE Xplore. Restrictions apply.

Вам также может понравиться