Professional Documents
Culture Documents
An Ten A
An Ten A
31
DSP Journal, Volume 8, Issue 1,December, 2008
2.1.1 Linear Prediction [LP] Method 2.2.1 Least Mean Squares Algorithm
This method estimates the output of one sensor using The LMS algorithm uses instantaneous gradient
linear combination of the remaining sensor outputs vector ∇(ξ) to update the weight vector. If w(n)
and minimizes the mean square prediction error i.e., denotes the estimate of the weight vector at the nth
the error between the estimate and the actual output. iteration and ξ(n) is the mean square error, the next
^
estimation of the weight vector for the (n+1)th
Expression for the array weights W and power
iteration, w(n+1) is estimated according to the
spectrum PLP (θ) are given following simple recursion
W (n + 1) = W (n) − μ ∇ (ξ ) (6)
^ R −1u1 (1)
W = where μ is the Step size that controls the convergence
u1T R −1u1 characteristic of the algorithm.
and 2.2.2 RLS Algorithm
u1T R −1u1 (2) Recursive Least squares Algorithm uses inverse of
PLP (θ ) = 2
u R A(θ )
T
1
−1 correlation matrix in place of step size (μ). Recursive
converges fast compared to the Least squares
where u1 is the column vector such that one of its Algorithm.
elements is unity and other elements are zero. A(θ) The algorithm is initialized by first setting
denoting steering vector that corresponds to direction
θ and R is correlation matrix. R −1 (0) =1 / δ * I (7)
2.1.2 MUSIC Method Where δ>0 and I- Identity matrix
In this method once the noise subspace has been The weights are updated using
estimated, a search for M directions is made by W (n + 1) = W (n) − R −1 (n + 1) x(n + 1)ε (8)
looking for steering vectors that are as orthogonal to
noise subspace. This is accomplished by searching
2.2.3 Sample Matrix Inversion Algorithm
for peaks in the spectrum as shown below Weights are chosen to minimize the MSE (Mean
Square Error) between the beam former output and
1 (3)
PMU (θ ) = 2
the reference signal as
A(θ )U N
w = R −1 r (9)
Where UN denotes an L by L-M dimensional matrix,
SMI has faster convergence since it employs direct
with L-M columns being the eigenvectors
inversion of the covariance matrix R. The initial
corresponding to the L-M smallest eigen values of the
optimum weight can be calculated by using
array of correlation matrix and A(θ) denoting steering
Equation.(9) where covariance matrix is
vector that corresponds to direction θ.
R(n)=x(n)xH(n) and cross-correlation matrix is r(n)=
2.1.3 Minimum Norm Method d*(n)x(n).
This method is applicable for Uniform linear array 3. Features and Implementation of
and finds the DOA estimate by searching for peak
locations in the spectrum. The power spectrum is Dielectric Lens
given by 3.1 Features of Dielectric Lens
The demand for higher data rate and capacity in
1 wireless communication has drastically increased in
PMN (θ ) = (4)
2 the recent years. One way allowing better spectrum
w A(θ )
T
usage is incorporating the concept of smart antennas.
Where w denotes an array weight such that it is of the Adaptive Beam forming used in smart antenna is a
minimum norm, has first element equal to unity and technique in which an array of antennas is exploited
is contained in the noise subspace. The solution to to achieve maximum reception in a specified
this problem leads to the following expression for the direction by estimating the signal arrival from a
spectrum desired direction (in the presence of noise) while
1 (5) signals of the same frequency from other directions
PMN (θ ) = 2 are rejected. This spatial separation is exploited to
A(θ ) U N U NT e1
T
separate the desired signal from the interfering
Where the vector e1 contains all zeros except the first signals. By transmitting only in the directions in
element equal to unity. which active users are present, unneeded wastage of
2.2 Beam forming Algorithms power is avoided. To improve the performance of this
Adaptive beam forming is a technique in which an system, a dielectric lens can be used along with the
array of antennas are exploited to achieve maximum antenna elements, to further collimate the rays in the
reception in the direction of the desired signal while specified direction, improving directivity and
rejecting the interference signals. Beamforming reducing interference. This enables implementation
methods discussed in this paper are LMS, RLS and of a smart antenna system in a densely populated user
SMI area, while also reducing cost and complexity of the
antenna array [6].
32
DSP Journal, Volume 8, Issue 1,December, 2008
The shape of the lens is then determined based on the of the material used in the lens. The DOA and Beam
determined composite radiation and the desired forming algorithm use this data to form the steering
collimation characteristics. Shaping of lens based on vector, and are found to yield the correct direction of
such considerations potentially results in accurate arrival estimate and Beam forming radiation pattern
beam shaping in the desired pre-specified directions so that the beams are formed at the required correct
[3]. directions, corresponding to the sources and it can
The main advantage of using a hemispherical lens even place nulls for interferers and jammers.
antenna is its non dependency on its focal length The signal processing algorithms for beam forming
which spelt in other terms mean non-dependency on can be chosen according to the resolution required.
frequency. Therefore this lens is an inherently Thus an algorithm is developed which enable us to
broadband which can work up to several GHz .In nullify the aberration brought about by the
other words we can place a S-band, L–band, X-band introduction of the dielectric lens. Thus the cons of
and K-band radiators in front of this lens using the dielectric lens are countered while enjoying
simultaneously and expect collimation. Figure 2 the advantages of higher gain and better collimation
shows the collimation provided by hemispherical afforded by the dielectric lens
dielectric Lens. As shown in figure 3 practically the
lens will provide better directivity.
33
DSP Journal, Volume 8, Issue 1,December, 2008
Figure 6,7 and 8 indicates the DOA of the signals Figure 8: Plot of Minimum Norm Method with
after nullifying the aberration brought about by the directions [ 100 200 300 400]
introduction of the dielectric lens.
Figure 6: Plot of MUSIC Method with directions Figure 9: Backward ray tracing after finding the
[ 350 200 280 400 ] changed DOA
Figure 7: Plot of Linear Prediction Method with Figure 10: Polar Plot for LMS method for signal
directions [100 200 300 400] direction 350
34
DSP Journal, Volume 8, Issue 1,December, 2008
6 Acknowledgements
The authors sincerely acknowledge the financial
support by DRDO (Defence Research &
Development Organization) New Delhi, India.
6. References.
[1] “Application Of Antenna Arrays To Mobile
Communications", IEEE Antennas Propagation
Mag , vol 85 , Aug 1997
[2] Carlos A Fernandes, “Shaped dielectric lenses
for Wireless Millimeter-Wave
Communications,” IEEE Antennas and
Propagation Mugazirze, vol41, No 5, pp 141-
150, October 1999
[3] Geogre V. Tsoulos , "Smart antennas for mobile
communication systems: Benefits and
Figure 11: Polar plot for RLS method with signal Challenges" , Electronics and Communication
direction 300 Journal , Apr 1999.
st
[4] Lal Chand Godara, Smart Antennas, 1 Edn.,
5. Conclusions: CRC Press, 2003
To adapt a smart antenna system for use of lens [5] Proakis J. G and D. G. Manolakis, Digital
antenna, we successfully used a ray tracing approach Signal Processing: Principles, Algorithms and
to generate the steering vector needed to implement Applications, Prentice-Hall, 1996.
the direction of arrival estimate algorithms and beam [6] Ravishankar S “Analysis of shaped beam
forming algorithms to generate the complex weights. dielectric lens antennas for mobile broadband
We used a hemispherical lens of Teflon with applications” IEEE international workshop on
dielectric constant 2.08. antenna technology 2005 proceedings.
We found that a rays incident on the lens at about 45º [7] Tiong sieh kiong, Mahamed Ismail,etc
and above are totally internally reflected. This limits “Dynamic characterized genetic algorithm for
the use of the antenna in a close circle radius. power usage reduction in WCDMA adaptive
Amongst the three direction of arrival estimate Beamforming. AIML Journal volume (6)
algorithm we implemented, the MUSIC algorithm Issue(3), September 2006.
works best in a multiple signal environment with a
considerable noise level as well. Linear prediction
tends to give an average direction estimate in the case Biographies
of two closely spaced signals. S. Ravishankar received his
Each beam forming algorithm has its own merits and Ph.D. degree in Electro-
de-merits depending on the application. Minimum magnetic Scattering from IIT
Mean Square Error (MMSE), Minimum Variance, Madras. Presently he is working
Maximum SIR are the optimization criteria used in in R V College of Engineering,
adaptive beam forming. Since they require intense Bangalore India, as Professor.
computational load, it’s difficult to implement in real His research interests are in the
time. Hence, we go for adaptive beam forming fields of Wire line Signal
algorithms like LMS, RLS. Processing Technologies, Electromagnetic Scattering
LMS is simple to implement but slow convergent and Antennas. He has published more than 25 papers
whereas, RLS is the fast convergent. CMA has the in journals and international conference proceedings.
ability to restore constant envelope property which
can be used in digital modulation techniques. SMI H.V.Kumaraswamy is current-
converges faster than LMS but its error prone due to ly working as Assistant
estimation of the correlation matrix. Professor in the Department of
There is no published literature on lens and its effect Telecommunication Engineer-
in smart antenna steering scheme, hence the ing, at R V College of
extension of this work depends on type of Engineering, Bangalore India.
application. He is currently pursuing his Ph
D Degree in Smart Antennas.
These algorithms were implemented on MATLAB His research interests are Digital Signal Processing,
7.1 and few algorithms on DSP TMS320C6713 to get Adaptive Signal Processing and Communication.
optimal code.
35