Simplified Neural Network Equalizer With Noise Whitening Function for GPRML System

A new design method of the simplified neural network equalizer (NNE) with the noise whitening function for a generalized partial response (GPR) channel is proposed. The long-term bit error rate performance of GPR class-I maximum likelihood (GPR1ML) system employing NNE in perpendicular magnetic recording (PMR) channel with thermal decay is also obtained by computer simulation using our thermal decay model for a CoPtCr-SiO2PMR medium withKuV/kT=60. The performance of GPR1ML system with NNE (GPR1ML-NNE) is superior to that of GPR1ML with a transversal filter equalizer. The PR1ML system using a noise whitening NNE and a two-state Viterbi detector provides almost equivalent performance to that of GPR1ML-NNE having a 16-state detector.