Simplification of Neural Network Equalization for PRML Channel with Nonlinear Distortion

Abstract

A simplification of a neural network equalization for a partial response maximum-likelihood (PRML) channel with a partial erasure and a nonlinear distortion in magneto-resistive (MR) head is studied. First, a model of the recording/reproducing channel with nonlinear distortions and the neural network equalizer are described. Then, the bit-error rates are obtained by computer simulation and the performances are compared with those of the conventional transversal filter and more complex neural network equlizer. The results show that the error rate performance of this simplified neural network equalizer is still better than that of the transversal filter.