On BCJR state metric quantization for turbo equalization

Vector quantization of the BCJR and Viterbi algorithms´ state metrics for detection of finite-state channels is considered. An estimate is given for the gain associated with vector quantization, over conventional implementations. This is expressed using the volume of the recurrent region, and the maximum state metric difference, which are both intrinsic properties of the channel detector. One application of this gain is the complexity evaluation of a previously proposed lookup-table BCJR implementation. The BCJR algorithm is of interest in turbo equalization used for communication over intersymbol-interference and finite-state Markov channels