On the improvements of channel unwrapping for conventional-adaptive MLSD

Per-survivor processing (PSP) and conventional adaptive MLSD (CA-MLSD) are well-known approaches to performing approximate adaptive maximum likelihood sequence detection over time-varying intersymbol interference (ISI) channels. PSP-based algorithms outperform CA-MLSD because the use of parallel decision feedback eliminates the need for a delay in the decision-directed channel estimation loop. Another method suggested to eliminate this delay is to use the sequence of previous delayed channel estimates to predict the current channel. We compare the performance of this modification of CA-MLSD with that of PSP and CA-MLSD. It is found that the addition of a predictor yields a performance improvement relative to CA-MLSD, but that there is still a significant degradation relative to the performance of PSP-based receivers.