MMSE-based linear parallel interference cancellation in long-code CDMA

Parallel interference cancellation (PIC) is a promising detection technique for multiuser code-division multiple access (CDMA) systems. It has previously been shown that weighted multistage PIC can be seen as a realisation of the steepest descent method for minimising the mean squared error (MSE). Following this interpretation, a unique set of weights, based on the eigenvalues of the correlation matrix, leads to the minimum achievable MSE for a short-code CDMA system. In this paper, we develop a method for finding a set of fixed weights, minimising the ensemble average of the MSE over all code-sets. Exact expressions of the ensemble averaged moments of the eigenvalues are derived. Simulation results show that a few stages can be sufficient for near-MMSE performance.