Nonlinear equalization for data storage channels

Abstract

Nonadditive and non-Gaussian noises and nonlinear signal distortions are significant sources of error in high density digital magnetic recording. Transition noise, transition broadening and partial erasure are a few such impediments. We propose a multilayer perceptron (MLP) based nonlinear equalizer to counter such distortions. It is shown that for realistic channel models the performance of such a nonlinear equalizer is superior to that of the conventional linear partial response equalizer, when used with a Viterbi detector.<>