Suboptimal detection for intersymbol interference inflicted cable channels

This paper presents some features of combining the design of a finite impulse response linear equalizer, which minimizes the noise variance (anchoring the first non zero sample), with detection techniques, which are frequently used in severe intersymbol interference inflicted channels. These include symbol by symbol detection using zero forcing decision feedback equalizer and Tomlinson filtering, and sequence detection invoking maximum likelihood sequence estimation and decision feedback sequence estimation (DFSE). Two trellis coded modulation schemes for this channel are also presented. The first uses Tomlinson precoder, for which we examine a DFSE with an unfolded branch metric, and the second employs DFSE (with states that combine the code states with part of the channel states). It is demonstrated that for the twisted pair and coax cable channels DFSE with 2-4 states achieves a very good performance when compared to other examined detection techniques in the uncoded transmission case. For the coded transmission case, it is shown that the performance of the coding scheme with Tomlinson precoding is close to the performance of the idealized coding scheme that invokes a sample whitened matched filter and an ideal tail canceller, and it improves (/spl sim/2 dB for bit error rate of 10/sup -4/) on the performance of a Viterbi algorithm with modulo folded metric (the commonly used metric in this case). A simple analytical upper bound for the error probability of a DFSE that accounts for the error propagation is also investigated.