Orthogonal multichannel ds-cdma system with low-complexity multiuser detection
EDUARDO ANTONIO DA SILVA ESTEVES · 1997
We introduce a general model to represent transmission of DS/CDMA signals in frequency-selective fading channels. We derive a vector tapped-delay-line model that includes the effects of signal spreading, time dispersion due to the channel, and receive filtering operation. A suboptimum reduced set of statistics is introduced, resulting in a vector discrete-path equivalent channel model. In addition, a multivariable whitening filter is derived using an iterative algorithm which solves a set of algebraic equations. We use the framework developed in this dissertation to derive some multiuser detectors presented in the literature. We propose an orthogonal multichannel CDMA system that combines random channel access and multiuser detection schemes to achieve high channel utilization. Because of the multichannel random access scheme, low-complexity multiuser detectors can be practically implemented at the receiver, providing good performance results when no power control is available. Error probability and throughput analyses of the proposed system are presented for both the additive white Gaussian and Rayleigh fading channels. Numerical results show that our system can achieve great performance advantage compared to a conventional detection DS-CDMA system with the same total processing gain. We present a bit error probability analysis of linear multiuser detectors when there exists some remaining MAI at the input of the detector. That is, the receiver does not have full knowledge of the active users' signature sequences. Results show that linear multiuser detectors are quite sensitive to unknown MAI. In fact, just one unselected transmission is sufficient to significantly affect performance of these detectors. We propose two suboptimum maximum likelihood sequence detectors for DS/CDMA signals where the likelihood computations are based on reduced-state trellises. In the first algorithm, which uses a reduced-state single trellis, we truncate states in a way similar to the one proposed for intersymbol interference channels. In the second algorithm, we introduce the idea of multiple reduced-state trellises, where survivor information is exchanged among different trellises. Simulation results show that significant complexity reduction can be achieved with minimal performance loss.