Semi-Analytical Performance Prediction Methods for Iterative MMSE-IC Multiuser MIMO Joint Decoding

In this paper, two semi-analytical performance prediction methods are proposed and compared for multiuser MIMO transmission over block-fading multipath channels and iterative MMSE successive interference cancellation based joint decoding. In the first method, we develop a completely analytical parameterisation of the whole turbo receiver up to the users' channel decoders. The method which follows the classical framework of extrinsic information transfer charts tracks the evolution of the average mutual information defined at coded bit level and circulating between the multiuser detector and the bank of channel decoders. The modelling of the demapping output in the multiuser detector is based on an explicit calculation of the (conditional) mutual information for each symbol digit in a symbol label under arbitrary (possibly non-uniform) a priori probability distributions on others. Then, shifting in viewpoint and considering user demapping and decoding as a joint process, we present an alternative method which tracks the evolution of the average mutual information defined at symbol level and circulating between the MMSE successive interference based interface and the bank of joint demappers and channel decoders. This allows to avoid the critical issue of parameterising the demapping. Numerical results demonstrate the superiority of the second method for high-order non-linear mapping.