Adaptive Cancellation of Intersymbol Interference for Data Transmission

For the past twenty years, engineers have been seeking new techniques to combat the intersymbol interference (ISI) in data transmission over band-limited channels. Adaptive equalization with the meansquare algorithm has been the major technique that allowed a substantial increase in attainable transmission rate. 1 If the channel has 1997 only phase distortion, then the linear fractionally spaced equalizer can eliminate virtually all of the ISI without enhancing the noise level.2,3 However, when amplitude distortion is present in the channel, any adaptive linear equalizer (LE) must compromise between inverting the channel transfer function and avoiding excessive noise enhancement. Inevitably, some noise enhancement occurs. Decision-feedback equalization can offer somewhat improved performance when amplitude distortion is present.4,5 By using the Viterbi algorithm,6 maximumlikelihood receivers, in principle, offer the best performance possible but depend on adaptive estimators of the channel and require an unpractically high complexity when the channel impulse response is long, as in the case of the typical telephone channel. If there were no isi, the probability of error in detecting the transmitted data level (i.e. ±1 or ±3) would be the same as if only one such pulse were transmitted in isolation. In that case, the optimal receiver (for Gaussian noise) would be a matched filter and would yield a certain error probability, Po. When pulses are sent sequentially, the effect of ISI cannot be totally eliminated.