Multistage Widely-Linear DF Equalizers for MIMO Channels

The paper addresses the low-complexity design of minimum mean square error (MMSE) multistage decision-feedback (DF) equalizers for multiple input/multiple output (MIMO) channels. More specifically, the multistage DF equalization strategy is combined with the widely-linear (WL) processing in order to exploit the properties of the rotationally variant transmitted signals. The optimization (in the MMSE sense) of WL-DF equalizer is re-examined in terms of the classical least-square approach and it is generalized to thecase of the MIMO multistage DF receivers, which require at least one equalization stage to initialize the input of the feedback filter. The performance analysis is aimed at comparing the performances of the linear multistage DF equalizers and their WL counterparts in terms of the symbol error rate (SER), the number of equalization stages required to minimize the SER and the number of floating point operations required to optimize the MMSE feedforward and feedback filters.