Linear parallel interference cancellation in long-code CDMAmultiuser detection

摘要

Parallel interference cancellation (PIC) is a promising detectionntechnique for code division multiple access (CDMA) systems. It hasnpreviously been shown that the weighted multistage PIC can be seen as annimplementation of the steepest descent algorithm used to minimize thenmean squared error (MSE). Following this interpretation, a unique set ofnweights, based on the eigenvalues of the correlation matrix, was foundnto lead to the minimum achievable MSE for a given number of stages in anshort-code system. In this paper, we introduce a method for finding annappropriate set of time-invariant weights for systems using long codes.nThe weights are dependent on moments of the eigenvalues of thencorrelation matrix, exact expressions of which can be derived. This setnof weights is optimal in the sense that it minimizes the ensemblenaveraged MSE over all code-sets. The loss incurred by averaging rathernthan using the optimal, time-varying weights is practically negligible,nsince the eigenvalues of sample correlation matrices are tightlynclustered in most cases of interest. The complexity required forncomputing the weights increases linearly with the number of users but isnindependent of the processing gain, hence on-line weight updating isnpossible in a dynamic system. Simulation results show that a few stagesnis usually sufficient for near-MMSE performance