Binary versus symbolic performance prediction methods for iterative MMSE-IC multiuser MIMO joint decoding

This paper presents a dual mode fast performance prediction method for iterative minimum-mean square error interference cancellation based joint decoding in a very general multiuser MIMO setting, considering a transmission over block-fading frequency-selective channels. The symbolic mode relies on a signal to interference-plus-noise ratio compression at symbol level, by means of the mutual information effective signal-to-noise metric (MIESM), bridging the gap between the extrinsic information transfer (EXIT) charts and the compression methods developed for link-to-system interface. The binary mode follows the classical EXIT chart approach relying on a new analytical formula of the bit interleaved coded modulation average mutual information in presence of non-uniform a priori on input bits. This paper demonstrates the superiority of the symbolic mode for high order modulation.