A Study on Iterative Decoding by Neural Network Detector in SMR System

We study the low-density parity-check (LDPC) coding and iterative decoding system, as a signal processing system for the shingled magnetic recording (SMR) in two-dimensional magnetic recording (TDMR). Previously we reported that a waveform equalization using a two-dimensional finite impulse response (TD-FIR) filter or an inter-track interference (ITI) canceler reduced the influence of ITI. We also proposed a neural network detector (NND), and evaluated the performance of the first decoding by the NND. In this paper, we propose the NND which iteratively calculates the log-likelihood ratio (LLR) as the decoding reliability using the returned sum-product (SP) decoder output sequence in addition to the TD-FIR filter output sequence. Furthermore, we evaluate the performance of the iterative decoding system using an NND, and compare it to that of the system using a soft-output Viterbi algorithm (SOVA) detector with the signal-dependent noise predictor.