Optimization of Soft Interference Cancellation in DS-CDMA Receivers
Parallel interference cancellation for DS-CDMA has been shown to suffer from biased amplitude estimates if a matched-filter estimator is used. The bias magnitude is proportional to the number of interfering users. For heavy system loads, the bias has been shown to adversely effect the accuracy of the interference cancellation process, thereby impairing BER after cancellation. Empirical simulation work has demonstrated that weighting down interference estimates can improve BER performance.
This thesis substantiates these BER improvements by modelling and analyzing a soft interference cancellation technique which mitigates the effects of the bias by minimizing BER after cancellation in a bit-synchronous parallel interference cancellation CDMA receiver. We analyze system decision metrics with down-scaled interference estimates and determine both the mean and variance of the biased decision statistics. From these two metric moments, system BER is evaluated, and the optimal interference scaling function which minimizes BER is derived. We demonstrate BER performance enhancements by simulating this soft interference cancellation technique in systems under perfect power control and in the near-far situation. We further discuss the applicability of the results to asynchronous systems.