Soft-Information Flipper based on Long-Short Term Memory Networks for Ultra-High Density Magnetic Recording

Abstract

An inter-track interference (ITI) effect can severely degrade the performance of ultra-high density magnetic recording systems such as bit-patterned media recording (BPMR) [1]. The two-dimensional (2D) modulation code can protect the severe ITI [2] which can efficiently improve the overall system performance: however, it cannot offer performance gain over the conventional system when the recording system must encounter with severe ITI from its sidetracks. Then, the flipping technique [3] was proposed to deal with that ITI sidetrack effect using the normalized soft-information values to be as bit-flipping criteria, which can efficiently improve the overall system performance. However, we found that those normalized values are difficult to confirm where is the optimal threshold value in the bit-flipping process. Therefore, this paper proposes a soft-information flipper based on long-short term memory (LSTM) networks [4] to deal with the mentioned shortcoming, which leads to improving the bit-error rate (BER) performance of three-track/three-head (3T/3H) BPMR system. Here, we directly map the soft-information, e.g., the datasets that were obtained from the 2D soft-output Viterbi algorithm (SOVA), sl,k with the hard-information that are achieved from encoding process, xl,k as shown in Fig. 1. Simulation results indicate that, at the same user density (UD), the proposed system (areal density (AD) = 3 Tb/in2) provides BER performance gain over the conventional uncoded (AD = 2.5 Tb/in2), coded [2], and normalization threshold systems (AD = 3 Tb/in2) [3] as shown in Fig. 2. Moreover, the results also reveal that the proposed system is more robust to the media noise than the other systems.