Maximum SNR prefiltering for MIMO systems

In this paper, we derive a class of FIR prefilters for MIMO systems that maximize the signal-to-noise ratio (SNR) defined as the sum of SNRs for the output branches with uncorrelated noise. This definition is motivated by an analysis of the pairwise error probability. The maximum-SNR (MSNR) class of prefilters includes decision feedback equalization (DFE) and channel shortening equalization (CSE) prefilters. The feedback taps are completely suppressed by the CSE prefilter and not suppressed at all by the DFE prefilter. The MSNR class includes hybrid prefilters which are a combination of CSE and DFE prefilters. Hybrid prefilters allow middle ground for suppressing feedback taps which is desirable for equalization using decision feedback sequence estimation (DFSE). We show simulation results for joint DFSE demodulation of two 8-PSK modulated cochannel EDGE users using a single antenna receiver. We find that a hybrid prefilter performs best for this example.