A NOVEL SEMI-BLIND EQUALIZATION SCHEME FOR ROBUSTNESS by ryg11839

VIEWS: 5 PAGES: 3

									                A NOVEL SEMI-BLIND EQUALIZATION SCHEME FOR
                ROBUSTNESS ENHANCEMENT IN ADAPTIVE ARRAYS

                                            Shweta Sharma
                   Birla Institute of Technology & Science, Pilani-333031 INDIA
                                  e-mail: f2003854@bits-pilani.ac.in
                                   Hema Singh, Rakesh Mohan Jha
          Computational Electromagnetics Lab., Aerospace Electronics and Systems Division
                    National Aerospace Laboratories, Bangalore-560017, INDIA
                        e-mail: hemasingh@css.nal.res.in, jha@css.nal.res.in

Abstract: A novel approach called the Semi-blind DF-GSC is proposed for active RCS reduction. Simulation
results of the Blind and the Semi-blind DF-GSC shows that two-staged Semi-blind approach has better performance
than the Blind DF-GSC. The proposed scheme achieves higher output SINR with enhanced robustness even with a
DOA mismatch. In this paper, interference suppression capabilities of both the schemes are compared using
structured and improved LMS algorithms. Better results are obtained in the case of improved LMS algorithm used in
conjunction with the Semi-blind DF-GSC scheme.

                                      I. INTRODUCTION
Blind channel identification and equalization have been widely studied from various perspectives
in order to combat multipath effects in the absence of training samples. Blind methods are those
methods in which no initial training is required. The performance of the adaptive array is
enhanced while simultaneously maintaining the robustness against any mismatch errors. A major
advantage of this approach is that it is not affected by erroneously detected data symbols due to
noise. This provides better robustness as compared to normal decision feedback generalized
sidelobe canceller (DF-GSC) scheme of adaptive arrays.

The problem with blind adaptation techniques is their poor convergence rate as compared to the
traditional techniques using training sequences. Generally a gradient descent based algorithm is
used with the blind adaptation schemes. Earlier analysis shows that the improved least mean
square (LMS) proves to be the best among various forms of LMS algorithm (Sharma et al.[1]).
The performance of an array can be further enhanced using a Semi-blind method. The simulation
studies show better results as compared to the well known blind adaptation technique (Singh et
al.[2]). In the present paper, computations are performed using improved LMS algorithm for the
Blind and Semi-blind DF-GSC. A comparative study is carried out for the interference
suppression capabilities of both these sidelobe cancellers. Standard and structured LMS
algorithms are employed for weight estimation.

                                        II. BACKGROUND
A digital system, transmitting data over a linearly distorting channel, usually contains an
equalizer to compensate for the channel distortion. The quantitative characteristics of the channel
distortion (i.e. channel response) are often not known a priori. There are three basic approaches
for the equalization of the unknown channel:
        a) Transmit and analyze a known training sequence. But in actual practice, these initial
            training snapshots are difficult to acquire.
        b) Try to detect the reliable data and adjust equalizer accordingly.
        c) Use statistical property of the transmitted signal for adjusting the equalizer.


EMTS 2007 International URSI Commission B - Electromagnetic Theory Symposium • July 26-28, 2007 • Ottawa, ON, Canada
Since the initial training snapshots are difficult to acquire, it becomes necessary to find a way,
which can result in sufficient robustness without the use of initial training snapshots. Inclusion of
a blind equalizer to DF-GSC was proposed by Lee and Wu[3]. In this approach, three stages are
included for the weight estimation process, illustrated in Fig. 1. The proposed design gives better
robustness than normal DF-GSC and same signal-to-interference-noise ratio (SINR) target.
However results obtained reveal that convergence of this scheme is not so good and it takes
longer time for simulations. Thus, this approach was modified to a two-staged Semi-blind
approach where only two stages namely, initialization and transition stage are used for the
computations.

                                                                                Iterative wa
                                                                                Iterative wb


                                y(k)
                                                                                                                       e(k)

                 Conventional                    wq Fixed                Iterative wq            wq Fixed
                                                                         Iterative B             B Fixed
                    GSC                          B Fixed




                                                            Fig. 1 Blind adaptation scheme


                                  III. SIMULATION STUDIES
Structured LMS algorithm is considered for the weight estimation. For comparison of the
nullifying capabilities of the two schemes, beam pattern with the Blind and Semi-blind DF-GSC
is presented in Fig. 2. A case of three hostile sources, probing a uniformly spaced linear antenna
array mounted on an aircraft, is considered. The direction of arrival (DOA) of three sources is
arbitrarily taken to be -35°, 20° and 50°. Radar sources are assumed to have an equal power level
of 100 each. Nulls up to 70 dB down are obtained in case of the Semi-blind DF-GSC against -60
dB nulls of the Blind DF-GSC. This verifies that the Semi-blind DF-GSC has superior
interference cancellation properties with almost 10 dB deeper nulls than the three-staged Blind
DF-GSC. A comparative analysis has been carried out for Blind DF-GSC where all the three
stages are used for the computations, subject to two forms of LMS algorithm viz. structured LMS
and improved LMS algorithm. Figure 3 represents the synthesized adapted beam pattern of Blind
DF-GSC. A difference of almost 12 dB is observed in the nulls of two algorithms showing
superior performance of improved LMS as compared to structured LMS algorithm.

Figure 4 shows the learning curve of the Blind and Semi-blind DF-GSC with improved LMS
algorithm. It is emphasized that the performance of sidelobe cancellers is better with Semi-blind
approach as compared to Blind DF-GSC. Semi-blind DF-GSC achieves an output SINR of
almost 19 dB whereas Blind DF-GSC converges around 15 dB only. The corresponding beam
pattern of the two approaches is presented in Fig. 5. In this case too, the Semi-blind DF-GSC
with improved LMS algorithm proves to be the best combination placing much deeper nulls
accurately than in the case of three-staged Blind DF-GSC. Nulls of up to -70 dB are achieved
which shows that the Semi-blind is more efficient than the Blind DF-GSC. It is apparent that
Semi-blind DF-GSC scheme is not only efficient but also takes lesser time to achieve the
optimum value.


EMTS 2007 International URSI Commission B - Electromagnetic Theory Symposium • July 26-28, 2007 • Ottawa, ON, Canada
                                    IV. CONCLUSIONS
Simulations results verify that the novel Semi-blind approach proposed here, for DF-GSC
scheme, converge faster and result in better performance. It is quite robust even in the case of
DOA mismatch and has better suppression capabilities. Nulls of up to 70 dB down are achieved
with higher output SINR. Furthermore, a comparative analysis is carried out for two forms of
LMS algorithm. It is inferred from the simulations that improved LMS algorithm is better than
structured LMS algorithm for the DF-GSC scheme. It is capable of yielding higher output SINR
with faster convergence rate.

                                       REFERENCES
[1] Sharma, S., H. Singh, and R. M. Jha, “Active RCS reduction: Studies on adaptive algorithms
    in sidelobe cancellers,” Project Document PD-AL-0614, National Aerospace Laboratories,
    Bangalore, Sep. 2006.
[2] Singh, H., S. Sharma, and R. M. Jha, “Scheme for suppression of narrowband probing
    sources for active RCS reduction,” Project Document PD-AL-0627, National Aerospace
    Laboratories, Bangalore, Nov. 2006.
[3] Lee, Y., and W. -R. Wu, “A robust adaptive generalized sidelobe canceller with decision
    feedback,” IEEE Transactions on Antennas and Propagation, vol. AP-53, Nov. 2005, pp.
    3822-3832.

        0                                                                                       0

      -10                                                                                   -10
                  Blind DF-GSC                                                                            Structured LMS
      -20                                                                                   -20

      -30                                                                                   -30              Improv ed LMS

      -40                                                                                   -40

      -50        Semi-blind DF-GSC                                                          -50

      -60                                                                                   -60

      -70                                          Hostile Sources                          -70                                          Hostile Sources

      -80                                                                                   -80
                -80    -60       -40       -20        0         20    40     60   80                  -80        -60       -40    -20         0     20     40   60   80



 Fig. 2 Effect of scheme on beam pattern: Comparison                                    Fig. 3 Effect of LMS algorithm on beam pattern:
 of the Blind DF-GSC and the Semi-blind DF-GSC                                          The Blind DF-GSC is studied using the structured
 using the structured LMS algorithm                                                     and the improved LMS algorithm

      20                                                                                   0

                                            Semi-blind DF-GSC                             -10
      15
                                                                                          -20             Blind DF-GSC
      10                                         Blind DF-GSC
                                                                                          -30
       5                                                                                  -40             Semi-blind
                                                                                                          DF-GSC
                                                                                          -50
       0
                                                                                          -60
       -5
                                                                                          -70                                           Hostile Sources

      -10                                                                                 -80
            0          50            100           150          200        250    300               -80        -60       -40     -20      0        20      40   60   80



  Fig. 4 Learning curves using the improved LMS                                          Fig. 5 Beam patterns corresponding to Fig. 4
  algorithm for the Blind and Semi-blind DF-GSC
  scheme


EMTS 2007 International URSI Commission B - Electromagnetic Theory Symposium • July 26-28, 2007 • Ottawa, ON, Canada

								
To top