8
BER Performance of Adaptive RAKE Receiver Using Tap Weights
Obtained by poes Deconvolution Technique
Shehzad Hussain" Zoran Kostic' B. Gopinath"
Abstract
Th e RAKE receiver is used in the mult ipath channels to make use of the inherent time-diversity
effect achieved by the use of a direct sequence spread spectrum (DSSS) communication system.
Goherent combining of mult ipath signals by a RAKE receiver improves system performance. The
improvement in performance is based on selecting the stronger paths and rejecting the noisy or
weak paths. This criterion requires good estimates of the multipath channel parameters . We obtain
multipath parameter estimates by the set theoretic deconvolution technique . The method is based on
the constrained iterative deconvolution using the method of Projection Onto Gonvex Sets (P OGS).
In this paper we analyze the bit error rate (BER) performance of an adaptive RAKE receiver
by computer sim ulations for a BPSK/DSSS system in an indoor environment. The performance
is analy zed for indoor multipath channel models, with/without Inter Symbol Interference (lSI) . In
order to have a realistic effect, we generate the impulse response of the channels by software package
SIRGIM.
For comparison purposes, performance of five receivers was analyzed , i.e., S trongest Path (SP) ,
RAKE with ideal estimates (Ideal) , Post Detection Integrator (PDI), RAKE with matched filter
estimates (MF), and RAKE with POGS estimates. The simulation results show that the RAKE
receiver with tap weights being adjusted by POGS estimates performs close to an ideal RAKE for
a sound ing signal transmitted at Signal to Noise Ratio (SNR) of greater than 25 dB. It performs
about 1-2 dB better than the RAKE with matched filter estimates at BER 0/10- 3 • The performa nce
being superior at low sounding SNR. It was also observed that performance of a PDI receiver was
extremely poor, i.e. a simple implementation of RAKE receiver in an indoor mult ipath environment,
having large num ber of taps and small number of useful paths is not feasible.
1 Introduction
In mobile radio environment the signals arrive at the receiver via various pa ths. In order to make
use of this path diversity. RAKE [1-3J receiver is employed to coherently combine the energies in
the multipaths to improve system performance. To resolve multipaths , wideband DSSS technique
·WINLAB , Dept. of Electrical and Computer Engr., Rutgers Un iver sity, PO Box 909, Pi scataway, NJ 08855-0909
I AT &T Bell Labora to ries , Holmd el, New J ersey 07733
B. D. Woerner et al. (eds.), Wireless Personal Communications
© Springer Science+Business Media Dordrecht 1995
120
is used . The paths ar e resol vable if the adjacent paths are separated by at least t he spreading
sequence bit (chip) duration, T e • RAKE receiver is implemented as a tapped delay line filter .
Optimum performance from RAKE is achie ved if the tap weights ofthe RAKE represent the impulse
response of the channel. When impulse response of the channel is time-varying , adaptive scheme is
needed to keep track of these variations. Various schemes have been proposed to adaptively track
the tap weights, some of these techniques are discussed in [9-12].
In thi~ paper we evaluate the tap weights of the RAKE combiner by set theoretic decon volut ion
technique , poes [4-7]. This technique has been extensively used in image processing . poes
application for estimating impulse response parameters of a multipath channel for sonar systems
was first studied by Kostic [5]. Its usefulness for DSSS system was studied in [6-7J. A synthetic
view of set theoretic estimation, along with an extensive list of references is available in [13J.
In section 2, we will discuss the channel/receiver model, section 3 gives a brief description about
parameter estimation by poes. Section 4 discusses the computer simulation and results .
2 'Transmitter, Channel and Receiver Model
A block diagram of the simulated system is shown in Fig . 1. Input is the sampled information
symbols (±1), multiplied by the spreading sequence, s . Impulse response of the rnultipath channel
is represented by h, n is the additive white Gaussian noise sequence, h m is the impul se response of
the matched filter, matched to the spreading sequence, s. The sampled estimate of the multipath
channel is ob t ained by poes algorithm and matched filter ing are represented by hp oCl and hAf F
respectively. The processor processes the output of the matched filter depending on five type of
receivers discussed latter. The processor output is sent to a detector, which makes a hard decision
on the received symbol to be +1 or -1.
Inlcrmatlcn
Symbo ls h
b (·~® ~E9
: tc3 1-ehlp
AWGN~
I m·sequence n ~
(.1.-1) r
C, .'Mr
L
,
HARD
ERROR DECISION
COUNT
Figure 1: Block diagram of the simulated system
