Proceedings of the 7th European Radar Conference

Side-Lobe Suppression Techniques for a Uniform Circular Array

Francesco Belori #1 , Stefania Monni 2 , Wim Van Rossum 3 , Peter Hoogeboom #4
#

IRCTR Department, Delft University of Technology Mekelweg 4, 2628CD Delft, The Netherlands
1

f.belfiori@tudelft.nl

TNO - Defence, Security and Safety P.O. Box 96864, 2509JG The Hague, The Netherlands
2 3 4

stefania.monni@tno.nl wim.vanrossum@tno.nl peter.hoogeboom@tno.nl

Abstract This paper provides a comparison between two different techniques for the array pattern side-lobe reduction in a passive radar system based on a circular array. The rst method retrieves the phase modes pattern for the reference array and then applies a conventional tapering for the side lobe suppression of the obtained virtual uniform linear array. The second approach foresees instead an optimization algorithm in order to obtain the desired level of the side lobes. Both techniques have been adapted for being used with an 8 elements FM-based circular array for passive radar applications developed at TNO - Defence, Security and Safety in The Hague. The effect of the directional elements has been also considered for the array pattern synthesis. The different advantages of the algorithms are described and a distinction about the possible elds of application is retrieved.

being r the array radius, and the elevation and the azimuth angle respectively, n = 2 n is the angular position N of element number n around the circle, an is the complex coefcient which performs the beam shaping and pointing and fn is the radiation pattern of the n th element of the array. For a specic pointing direction (0 , 0 ), the selection of a phase-only coefcient an as: an (0 , 0 ) = ejkr sin(0 ) cos(0 n ) (2)

I. I NTRODUCTION The main advantage of using circular arrays, instead of the linear ones, is the ability of these symmetrical systems to steer the beam electronically through 360 ([1]). The pattern behavior is also characterised by very small dependance of the azimuth resolution and of the side-lobe level on the steering angle. The array conguration, as is well known, allows the application of digital beam forming (DBF) algorithms for the formation of multiple beams, for the insertion of angular nulls inside the antenna pattern (AP) in order to minimize jamming, interferences and, for the passive radar applications, the direct path signal of the reference transmitter. On the other hand, the utilization of a system of multiple antenna elements introduces the mutual coupling (MC) effect that has to be compensated. Several techniques for the MC parameter estimation have been presented in literature for the different classes of antenna congurations ([3],[4]) and also for the compensation of the aforementioned system ([2]). As this paper it is not focused on that topic, in the rest of the article the MC compensation will be assumed already performed. Under this assumption and referring to a N elements circular array, the AP can be expressed as:
N

provides the simplest synthesis of the AP for an uniform circular array (UCA). In several applications the requirement on the side-lobe level can be a very stringent constraint and the pattern obtained with (2) could not be able to satisfy it. In order to get a better behavior of the side-lobes a different selection of the excitation coefcients is needed. This paper analyzes two different approaches for the side-lobe reduction; the rst one is based on the phase modes decomposition of the circular array pattern while the second algorithm is based on an optimization procedure. In Sec. II the phase mode technique is shown and the nal form of the coefcient an is estimated both for the omnidirectional and for the directional element pattern case. In Sec. III the latter approach is depicted while results and conclusions are reported in Sec. IV and V. II. P HASE M ODE T ECHNIQUE The array pattern of a circular array is a periodic function in the interval [0, 2] and this characteristic allows its representation in terms of a complex Fourier series. Referring to (1) and rst considering the case of omnidirectional radiating elements (fn (, n ) = 1), we can write:

AP (, ) =
p=

Cp ()ejp

(3)

being: Cp () = 1 2

AP (, )ejp d

(4)

AP (, ) =
n=1

an fn (, n )ejkr sin() cos(n )

(1)

Each term of the sum in (3) is normally referred as phase mode of the radiation pattern ([1]) and it has a 2p phase

978-2-87487-019-4 2010 EuMA

113

30 September - 1 October 2010, Paris, France

variation as varies from 0 to 2. From (3) it follows that, in order to synthesize a directional far eld pattern, a pth order phase mode ejp must be produced in the far eld. The Cp coefcients can be generally evaluated as a sum of the an excitations, but that sum cannot be solved in a closed form for all the an . However, a specic set of symmetrical excitation currents can be identied and computed ([9]) as :

pth phase mode pattern for a uniform circular array with ideal directive elements is given by:
N

APp (, ) =
n=1

Ap ejpn fn (, n )ejkr sin() cos(n )

(10) Thanks to the element pattern periodicity, it can also be expressed by means of a Fourier series:
L

an =
p=

ap = n
p=

Ap ejpn

(5) fn (, n ) =

Bl ()ejl(n )
l=L

(11)

The terms of the sum in (5) are referred as the phase mode excitation currents. By using (1), with the omnidirectional case assumption, and (5) the pth phase mode pattern takes the form:
N

Where the Bl () coefcients can be computed as in (4). By replacing (11) in (10) and rearranging the exponential term we obtain:
L

APp (, ) =
n=1

Ap ejpn ejkr sin() cos(n )

(6)

APp (, )

= Ap ejp
l=L N

Bl ()

Introducing the Bessel function approximation for the array factor exponential term and neglecting a constant term of no interest, (6) can be rewritten as: APp (, ) Ap j p Jp (kr sin())ejp (7)

n=1

ej[(pl)(n )+kr sin() cos(n )](12)

where Jp () is the rst kind Bessel function of order p. The behavior of the Bessel functions also introduces a limitation on the maximum number of phase modes that can be excited on the array. Decreasing rapidly to zero when the order of the function exceeds its argument, the maximum phase mode order is estimated as the maximum value of the argument which is: P = kr . The Ap coefcients can now be selected in order to synthesize the far eld pattern of the array. By choosing the coefcients as: ejp0 (8) Ap = p j Jp (kr) the array pattern resembles the pattern of a linear array [5]. The only difference is that, while in the linear case the azimuth scanning angle is related to a sin() function, providing therefore a possible steering sector of 180 , in this case the scanning capability has no angular restrictions over the 360 . Referring to the N excitations that must be applied to the circular array, they can be evaluated by combining (5) and (8):
P

That can be also approximated by using a Bessel function representation, providing the following nal form for the pth phase mode component of a circular array based on directive radiating elements:
L

APp (, ) = Ap ejp
l=L

Bl ()j pl Jpl (kr sin())

(13)

Following this result, the an excitations that must be provided to the reference N elements circular array, can be derived as it has been done in(9):
P

an =
p=P

ejp(n 0 )
L l=L

Bl ()j pl Jpl (kr)

(14)

III. O PTIMIZATION T ECHNIQUE The other algorithm that has been considered in order to reduce the side-lobe level of the reference AP is based on an optimization approach. By using a matrix representation, it can be written as: AP = sH x (15) where sH is the hermitian array steering vector of the weights in (2) and x is the array manifold elements. As suggested in [6], the amplitude and phase weights of the steering vector can be computed by directly applying an optimization procedure on the AP behavior. A simple objective function can be written as:
NS NM

an =
p=P

ejp(n 0 ) j p Jp (kr)

(9)

By using the an coefcients computed with (9), a conventional window tapering may be applied in order to obtain the desired level of the side lobes. 1) Directive Elements Case: The expression of the excitation coefcients evaluated in the previous section can be extended to the directive elements case. Referring to (6), the

O(sH ) = c
s=1

|AP (s )|2
m=1

|AP (m )|2

(16)

being the s values a set of points of the array pattern selected in the side lobe region and the m values a set of points inside the 3dB main beam region. The coefcient c has the functionality of emphasising the contribution of the side lobe region in respect to the one coming from the main beam. The steering

114

vector is computed by minimising the objective function and this minimisation is based on the Broyden-Fletcher-GoldfarbShanno (BFGS) Quasi Newton method ([7],[8]) implemented in the Matlab environment. IV. N UMERICAL R ESULTS A. Theoretical Performances In this section the results of the theoretical array patterns are shown. The reference system is an 8-element circular array, working in the FM bandwidth (88-108 MHz) at the main frequency of 96.8 MHz. The radius of the array is 0.48 leading to an element spacing of 0.38. This circular array has been built at TNO - Defence, Security and Safety in The Hague for passive radar applications. Further characteristics about the system and the receiver chain can be found in [2]. Fig. 1 shows the ideal pattern for the array with the mentioned
Fig. 2. Comparison between different Chebishev windows applied to the phase mode array pattern

shows a side-lobe reduction of 17dB, so 9dB better than the conventional case, and at the same time it is able to keep almost the same beamwidth of the main lobe (43 ). A comparison between these two methods can be also carried

Fig. 1.

Un-weighted ideal array patterns

parameters; it is characterized by a side-lobe level of 8dB for a 3dB beamwidth of 42 . The conventional pattern is compared with the one obtained by applying the phase mode excitations in the form of (14), so without a tapering window. It can be noticed that, as expected, even the simple phase mode excitations provide a small reduction of the side lobes; this reduction is due to the nal form of the array pattern, similar to the one of a linear array, which the excitations realise. On the other hand an enlargement of the main beamwidth, of the order of 7 , occurs with this method. The application of a further tapering function, therefore, can ensure the desired level of the sidelobes but it consequently introduce an additional degradation of the azimuth resolution. In Fig. 2 the behavior of different array pattern shapes, evaluated by applying three different orders of Chebishev windows, are plotted. For the side lobe reductions that are presented (18dB, 22dB and 27dB) the related beamwidth apertures are 56 , 60 and 62 respectively. The result obtained by applying the optimization algorithm is presented in Fig. 3. The behavior of the optimised pattern

Fig. 3.

Result of the optimization approach for the side-lobe reduction

out by evaluating the efciency ensured by the different set of weights. Being the efciency dened as: = ( N
N 2 n=1 |an |) N 2 n=1 |an |

(17)

and considering similar levels of side-lobe suppression ( 20dB), the efciency values are almost the same for both the approaches ( 0.81). However, the phase mode technique shows better efciency values at higher levels of sidelobe reductions; we have indeed P mode (25dB) 0.83 and P mode (35dB) 0.86.

115

B. Real Performances The described algorithms have been applied to the real circular array presented in [2]. By using a reference transmitter, several data acquisitions have been done from different locations around the array. These measurements have been rst used to perform the mutual coupling compensation and then to compute the array patterns with both techniques. In Fig.4 the real phase mode pattern is evaluated and compared with the ideal one. The pattern is characterised by a main beam 3dB aperture of 55 and by a side-lobe level of 22dB. The results of the optimization method are shown in Fig.5;

Fig. 5.

Optimised pattern computed on real data

The second method used an optimization procedure to compute the complex excitations of the array. This approach has shown better general result, as it was able to provide a good suppression of the side lobes while maintaining the same 3dB main beam aperture. On the other hand, due to the optimization procedure the computational cost can be very demanding when also a kind of operational adaptivity is requested. ACKNOWLEDGMENT
Fig. 4. Phase mode pattern computed on real data

they have been computed by setting c = 20 in (16), by dening an area of 40 degrees around the main beam direction and by using a transition sector, between the main beam area and the side-lobe region, of 10 degrees. The input steering vector for the optimization has been chosen equal to the phase mode steering vector computed with the previous method. In this case the main beam keeps its original azimuth aperture ( 43 ) and a side-lobe level of -17dB is also achieved. V. C ONCLUSIONS In this paper, two different approaches for the side-lobe reduction in circular arrays have been presented and compared. The rst technique is based on the possibility to rewrite the circular array pattern by means of the phase mode pattern components. In such form, the circular array behavior is similar to the one of a linear array, so the desired side-lobe reduction can be achieved by simply applying a conventional window tapering. The main advantages of this method reside in the less computational requirement for the weights evaluation and in an higher adaptivity; however, the worse behavior for the higher levels of side-lobe suppression has to be taken into account when also the azimuth resolution is a stringent parameter for the system design.

This work was supported by TNO (Dutch Organization for Applied Scientic Research) and by NL Ministry of Defense. The authors acknowledge the excellent work of TNOs internal workshop. R EFERENCES
[1] A.W. Rudge, K. Milne, A.D. Olver and P. Knight, The Handbook of Antenna Design, IEE Electromagnetic Waves Series 16, Chap. 12, 1983. [2] F. Belori, S. Monni, W. Van Rossum and P. Hoogeboom, Mutual Coupling Compensation Applied to a Uniform Circular Array, International Radar Symposium 2010. [3] H. Steyskal and J.S. Herd, Mutual Coupling Compensation in Small Array Antennas, IEEE Trans. on Antennas and Propagation, 1990. [4] G.V. Borgiotti and Q. Balzano, Mutual Coupling Analysis of a Conformal Array of Elements on a Cylindrical Surface, IEEE Trans. on Antennas and Propagation, 1970. [5] T. Rahim and D.E.N. Davies, Effect of Directional Elements on the Directional Response of Circular Antenna Arrays, IEE Proc., Vol. 129, Pt. H, No. 1, February 1982. [6] M. Malanowski and K. Kulpa, Digital Beamforming for Passive Coherent Location Radar, IEEE Radar Conference, 2008. [7] D.F. Shanno, Conditioning of Quasi-Newton Methods for Function Minimization, Mathematics of Computation, Vol. 24, No. 111, pp. 647-656, 1970. [8] C.G. Broyden, The Convergence of a Class of Double-rank Minimization Algorithms, IMA Journal of the Institute for Mathematics and Applications, Vol. 6, pp. 222-231, 1970. [9] R.J. Mailloux, Phased Array Antenna Handbook Artech House, 2nd edition, 2005.

116