The adaptive MSINR algorithm to improve error rate for channel equalization

Abstract

Based on the expression of the error probability for the infinite length equalizer and the error bound for the finite length equalizer for M-ary PAM channels using the unbiased decision rule, we are inspired to develop an adaptive algorithm called the maximum signal-to-interference-plus-noise ratio algorithm (MSINR) to improve error rate performance. The MSINR algorithm is found to outperform the LMS both in convergence behaviors and error rate performance. When comparing with a true minimum error rate algorithm called approximate minimum bit error rate (AMBER), it is found that AMBER yields a lower error probability as expected, but as the equalizer length is sufficiently long, the error probabilities obtained from MSINR very closely approximate the minimum error probabilities given by AMBER. Moreover, the MSINR convergence performance, both in speed and smoothness, is far superior to the AMBER.