You are on page 1of 4

Compensation of Mutual Coupling in Dual

Polarized Arrays
Bjom Lindmark
Department of Signals, Sensors, and Systems,
Royal Institute of Techtlology
SE-100 44 Stockholm Sweden
Email: bjom.lindmark@s3.kth.se

Abstract
Mutual coupling in dual polarized antenna arrays is compensated by matrix multiplication. Using far-
field measurement data, a least squares estimation of the coupling matrix is made including CO- and
cross-polar coupling. A compensation for measurements off the phase center is important as well as a
proper assumption of the ideal element pattern. We study the performance of a dual polarized patch may
with respect to far-field phase error and cross-polarization level and the performance of the compensation
method exceeds or equals the use of dummy elements.
I. INTRODUCTION
Base station antennas in mobile communications typically consist of linear arrays of vertically or dual
polarized antenna elements. In an adaptive antenna system we increase the antenna aperture by placing
several linear arrays side-by-side. We instead get the possibility to focus the reception and transmission
of radio waves to the direction of the desired user. In array signal processing, it is normally assumed
that we have a Uniform Linear Array (ULA) of isotropic elements. Real antenna arrays will differ from
the ULA model for several reasons: un-isotropic elements, edge diffraction from finite-sized arrays, and
mutual coupling. In this paper we will focus on the effects of mutual coupling.
There are different ways to mitigate mutual coupling. The most straight-forward is to try to minimize
the inter-element coupling itself as in [l]. However, this is problematic and costly, especially for dual
polarized arrays where CO- and cross-polar coupling must b: compensated separately. Another solution
is the use of dummy columns on each side of the array which improve the performance [2] at the price
of increased array size. An altemative to dummy columns is to compensate for the coupling through
matrix multiplication [3] derived from measured scattering ]mameters. This requires knowledge of the
reference plane corresponding to the center of radiation for each radiating element which makes the
method unpractical for real-life antennas with complex feed networks and non-ideal elements. It is also
possible to calculate decoupling matrices from the complex far-field radiation pattem by integration [3]
or a direct Least-Squares estimate [4], [SI. In this paper we extend these methods by using non-isotropic
amplitude pattems and compensation for measurements wheie the array has been rotated around a point
off its phase center. We also present de-coupling for dual polarized arrays to reduce cross-polarization as
well. We then compare the de-coupling method for a 4-element array to the performance of an array with
2 dummy columns on each side. The de-coupling method provides superior array performance compared
to the use of dummy columns, even though the array size is one-half.
11. THEORY
In array signal processing one typically assumes a model for the antenna array, the most common being
a ULA with its corresponding Vandermonde structure of the array response matrix:

where M is the number of elements in the array, d the nuniber of signals, k = 2x/X the wave number,
A the element spacing, and 0 the angle of incidence, or DOA, measured from the normal to the array
axis. Now, in order to include the effect of mutual coupling: as well as other errors we follow [5] and

02003 IEEE
0-7803-7846-6/03/$17.00 896
express the measured far-field of the array as multiplication of a coupling matrix C with an ideal array
response A(@):
A(Q) CA(8) (2)
Now, at the manufacture of our antenna array, we may have calibration data A(Q) available from far-field
measurements where 0 typically includes f180" in steps of 1'. We use the least-squares estimate of C:

C =A(Q~)A(QO)~[A(~O))A(Q~)H)~-' (3)
where 00is a vector of the calibration data used in the estimate. We need to limit the data used since
the deviations from ideal behavior tend to be very large as we approach 0 = f90° and if we try to fulfill
(2) for all angles we could get a poor result. Also, if the spacing in the array is slightly above X/2 we
need to limit us to the angles where there is no ambiguity (i.e. grating lobes) so that A(&) has full rank.
Note the difference to the integration method of [3] where the spacing must be greater than X/2. In this
+
paper eo= [-60°, -59"... 59", +60"]. This seems a logical choice if we consider cellular systems
where the sector of coverage is 120'. but is quite possible that another choice of angles or a weighting
of the data could provide even better performance.
Finally, we have to decide which A(@) we should use in (3). We could use the ULA of (l), but it
turns out that it is not a good idea to match measured data of a directive element to an ideal behavior
with constant amplitude over incidence angle. It is then impossible to match the amplitude over a wide
sector, and this deteriorates the phase performance as well. Instead, we assume a more realistic ideal
behavior, namely:
a(@) COS"^ auLA(b), A(Q) = [a(b,),. . . ,a(Od)] (4)
where n is an exponent chosen to best fit the particular elements used. Note that it is important that
the calibration measurement is performed so that the array is rotated around a common phase center,
precisely as assumed in (1). If not, we would try to match array responses with different phase centers
which would lead to a poor result. We therefore calculate A(0)from the measured data A,l(Q) as:
A(Q) = ACQl ( ~ ) ~ j k ( &
cosO-1)+6, sins)
(5)
where 6, and hY are displacements during the calibration normal to and along the array, respectively. In
this paper we choose 6, and 6, so that the far-field phase becomes as symmetrical as possible.
If the antenna array is dual polarized, e.g. *45"-polarized, the mutual coupling is more complicated
because of CO- and co-polar coupling. However, it is possible to extend the model (2) to consider such
coupling as well. If we denote the measured CO- and cross-polar response of the -45"- and +45'-polarized
+450
channels with Ai4' (Qo), ..., A,, ( 8 0 ) . the equation system is:
~

The equations for the different parts of the coupling matrix c d p are conveniently decoupled and may be
calculated as in (3). The compensated output for the two set of polarization channels x * (1)~ is ~then:~

where k*450 denotes the received signals from the non-ideal array. However, since any element which is
not a Huygens' source will have inherent cross-polarization as we scan off boresight, we cannot expect
the same performance for the cross-polar coupling as for the CO-polarcase.

891
111. N U M E R I C A LRESULTS
The compensation for coupling was tried on a measure data from two different dual polarized arrays
for UMTS. A n array of 4 columns was compared to an array of 8 columns where the 2 outermost
columns on each side served as dummy elements. The 8-column array is shown in Fig. 1; the 4-
column array is of the same design except for the number of columns. Each column has 5 dual po-
larized aperture coupled patches of the design in [6] and a spacing between the columns of 80 mm.
The un-compensated radiation pattems in of the two ar-
rays are shown in Figs. 2-5. The effect of a finite ground-
plane causes the edge columns to radiate somewhat
more towards the nearby edge. The cross-polarization
is also rather high, especially at 160”. In order to make
a fair comparison between the compensated and un-
compensated arrays, the far-field phase has been set to
zero in boresight for both arrays. For the compensated
array, this is not necessary since any phase bias will be
compensated for, but the un-compensated array could
show a poor performance simply due to a single cable
of incorrect length. The measured data was compensated
with a phase displacement of 6, = 11 mm and 6, =
5 mm according to (5). This reduced the average r.m.s.
phase error after compensation with a factor of 1.6. The Fig. 1. An 8-column dual polarized aperture coupled
coupling matrices was then calculated according to (6) pach array for UMTS with an inter-column spacing of
above with 7~ = 1 and the resulting pattems and phase 80 nun. Courtesy Allgon Systems AB.
error is seen in Figs. 6; both cross-polarization and phase
error has been drastically reduced.
Iv. CONCLUSIONS AND FUTURE WORK
The potential for mutual coupling compensation based on measured far-field pattems of 4-element aperture
coupled patch arrays has been demonstrated. It is important to compensate for measurements where the
array has not been rotated around its phase center and to use a suitable ideal element factor. The method
leads to superior performance compared to the use of 2 dummy elements on each edge of the array,
also for dual polarized arrays. We have also compared the SIR performance of the different patch arrays
with an array with stochastic errors [7]. These results will be shown at the conference and confirm the
performance of the coupling compensation.

ACKNOWLEDGEMENTS - The author wishes to express his gratitude to Allgon Systems AB for the use
of far-field measurement data of the dual polarized UMTS mays.
REPERENCES
[I] B. Lindmark, S.Lundgren. J. Sanford, and C. Beckman, “Dualpolarized array for signal prccessing applications in wireless
communications:’ IEEE Trans. Antennas Propagat.. vol. 46, no. 6. pp. 758-763, June 1998.
[21 S. Lundgren, “A study of mutual coupling effects on the direction finding performance of esprit with a linear microstrip
patch array using the method of moments:’ in IEEE Antennas Propatat. Soc. Int. Symp. Dig.. Baltimore, MD. July 19%.
pp. 1372-1375.
[31 H. Steyskal and 1. S. Herd, “Mutual coupling compensation in small amay antennas,” IEEE Trans. Antennas and Propagat..
vol. 38, no. 12, pp. 1971-1975. Dec. 1990.
141 P.N. Fletcher and M. Dean, “Least squares pattern synthesis for confcmnal arrays:‘ Elertmnics Lerrers, vol. 34. no. 25, pp.
2363-2365, Dec 1998.
[ 5 ] T. Su, K. Dandekar, and H. Ling, “Simulation of mutual coupling effect in circular arrays for direction-finding applications:’
Micmwave and Optical Technology Letters, vol. 26, no. 5, pp. 331-336, sep 2000.
[6] B. Lindmark, S. Lundgren, and C. Beckman, ‘IJual polarized multibeam antenna,” Elecrmnics Lerrers, no. 25, Dec. 1999.
[7] B. Friedlander and A. J. Weiss, “Effects of modeling errors on waveform estimation using the MUSIC algorithm:’ IEEE
Trans. Signal. Processing, vol. 42, no. I , pp. 147-155, Jan. 1994.

898
.Pot
&-15
4 0 40 -20 0
,

20
,

40
1
BO
e [degrees]

Fig. 5. Measured far-field phase error in vertical polarization


-90 -75 -Bo -45 -30 -15 0 15 30 45 Bo 75 90 at 1900 MHz of the 4-column array. The r.m.s. error is 2.1'
e [degrees] to 4.9".

Fig. 2. Measured co- and cross-radiation pattern at 1900 MHz


for the I-column array in Fig. 1. The thick solid and dashed
lines are the CO- and cross-polar patterns, respectively, of the
center 4 columns. The thin dashed-dotted and dotted lines are
the CO-and cross-polar pattems of the outer dummy columns.

unmmpensafad phase emr. 19wMHZ

10
8 5
g o

4
I -5
-10
- EO1 Y EO-MI

P I I
4 -40 -20 0 20 40 Bo .(..,,m1.4m-po1
e 10ep-I -- m1.1x-pol ,"..,..'.
m1.2 x-po1
- - m1.3 x-pol
Fig. 3. Measured far-field phase error in vertical polarization rm.4x- I
at 1900 MHz of the 8-column array in Fig. 1. The thin dotted
lines correspond to the 2 dummy elements on each side. The
r.m.s. enor of the 4 center columns is 1' to 1.3'.

l ' " " ' " " ' 1


e [degrees]
Campenpaledphase error. 19W MHz

Fig. 6. Compensated co- and CO-polarpattems C-'D(S) and


compensated far-field phase error in vertical po&zation at
1900 MHz of the 4-column array. The r.m.s. error is 0.8" to
1.3'.

Fig. 4. Measured CO-and co-polar pattems at 1900 MHz of


the 4-column array.

899

You might also like