Crosstalk cancellation and memory truncation in transmultiplexer filter banks — Transmission over non-ideal channels

Abstract

When transmultiplexers with overlapping frequency bands are used for the transmission of data over non-ideal channels, intersymbol interference and crosstalk between different data channels will arise. This paper addresses the design of optimal linear networks that reduce the effects mentioned above. A receiver structure based on a combination of crosstalk reduction, memory truncation and Viterbi detection will be proposed. The filter design method presented here is based on the maximization of a signal-to-noise ratio (SNR) at the detector input. The SNR will be defined for channel memories being truncated to arbitrary lengths. Thus, low-complexity Viterbi detectors working independently for all data channels can be used. The design of minimum mean squares error (MMSE) equalizer networks is included in the framework.