Neural Network Equalizer for GPRML System with Post-Processor

Abstract The neural network equalizer (NNE) for the generalized partial response maximum likelihood (GPRML) system with a postprocessor in perpendicular magnetic recording is studied. First, a new designing method of NNE is proposed for suppressing frequency of a high-level noise at the discrimination point which degrades the bit error rate (BER) performance of the GPRML system with a post-processor. Then, the BER performance of GPR class-1 ML system with a post-processor using the NNE is obtained and compared with that using a conventional transversal filter as an equalizer. The result shows that the gain of the former over the latter is about 0.6 dB at a BER of 10-5.