Neural Network Equalization of Nonlinear Distortion in MR Head

Abstract

The neural network equalization for the read/write data using a spin-stand is studied. As a recording code, the (1,7) RLL code is used and the readback waveform influenced by a nonlinear distortion in MR head is equalized to EPR4 characteristic by a three-layered neural network. For the read/write data at a recording density of 165 kFCI, the equalization characteristics and the bit-error rate of the EPR4ML sytem are obtained by computer simulation and are compared with those for a conventional Nyquist equalization. The results show that the neural network equalizer can effectively mitigate the effect of a nonlinear distortion in MR head and provide a significant performance improvement.