Multilayer perceptron-based array reader optimization for ultra-high density magnetic recording

Abstract

We have previously presented a utilization of a long-short-term memory (LSTM) network to improve the reliability of the log-likelihood ratios (LLRs) in a coded three-head/three-track bit-patterned magnetic recording (BPMR) system. To continuously enhance a bit error rate (BER) performance; therefore, this article introduces a novel LLR flipper utilizing a multilayer perceptron (MLP), which operates under the optimization of input numbers of array reader positions. We first optimize an input number of MLP-based soft-information flippers with varying the number of recording bits in various data patterns. Then, the right positions of an array reader under the operation of our proposed MPL flipper are also optimized to obtain better BER performance. The simulation results demonstrate that the proposed MLP flipper significantly improves BER performance compared to the previous LSTM flipper using the same input number. Moreover, our proposed system still provides superior performance when it is operated under an optimal input number and can be higher when working with the right array reader positions.