Efficient implementation of iterative multiuser decoding

Iterative multiuser decoding based on the sum-product algorithm can give excellent performance. To reduce complexity, inherent maximum a posteriori detection is often replaced by linear MMSE detection combined with parallel successive cancellation. The powers of the users are often unbalanced, either by design to improve performance or due to independent fading in the reverse link. During iterations, the power distribution of the residual interference evolves. This can lead to a severely mismatched MMSE detector after few iterations unless it is re-calculated for every iteration. Here, a computationally less costly approach is presented.