Quantization issues in blind equalization: a case study

A simple case study of blind equalization on a first-order channel (one pole and one zero) is presented. Quantization issues are addressed. Estimates of the transient time to error recovery are obtained. The authors focus on analytical techniques. First, they establish necessary and sufficient conditions for error recovery regardless of the input sequence. These conditions are severe, and not met by many practical channels. Using averaging techniques the class of channels allowing error recovering for most, but not all, input sequences is enlarged. Using a Markov model, the authors estimate a conservative overbound for the mean time to error recovery, valid for almost all channels and almost all input sequences