Turbo-Detection for Multilayer Magnetic Recording Using Deep Neural Network-Based Equalizer and Media Noise Predictor
Turbo-Detection for Multilayer Magnetic Recording Using Deep Neural Network-Based Equalizer and Media Noise Predictor
复制标题
使用基于深度神经网络的均衡器和介质噪声预测器进行多层磁记录的 Turbo 检测
DOI：
10.1109/tmag.2021.3122136
复制
发表时间：
2022
影响因子：
2.1
通讯作者：
Chan, Kheong Sann
中科院分区：
文献类型：
--
作者：
Sayyafan, Amirhossein;Aboutaleb, Ahmed;Belzer, Benjamin J.;Sivakumar, Krishnamoorthy;Greaves, Simon;Chan, Kheong Sann
本文考虑基于深度神经网络（DNN）的多层磁记录（MLMR）涡轮检测，这是一种新兴的硬盘驱动器（HDD）技术，使用垂直堆叠的磁介质层，读取器位于最顶层之上。所提出的系统使用具有两个上层轨道和一个下层轨道的两个层。阅读器信号由卷积神经网络（CNN）处理，以分离上层和下层信号，并分别将其均衡为2-D和1-D部分响应（PR）目标。上层和下层信号分别馈入2-D和1-D Bahl-Cocke-Jelinek-Raviv（BCJR）检测器。检测器的软输出馈送多层基于CNN的媒体噪声预测器，其预测的噪声输出被反馈到BCJR均衡器以降低其误码率（BER）。BCJR均衡器还与低密度奇偶校验（LDPC）解码器接口。通过从上层BCJR向下层BCJR发送软信息来实现额外的BER降低。这种涡轮检测系统上的一个双层MLMR信号产生的颗粒切换概率（GSP）介质模型的模拟显示，密度增益11.32%的可比系统，没有较低的层，并实现2.6551兆兆比特每平方英寸（Tb/in2）的整体密度。
This article considers deep neural network (DNN)-based turbo-detection for multilayer magnetic recording (MLMR), an emerging hard disk drive (HDD) technology that uses vertically stacked magnetic media layers with readers above the top-most layer. The proposed system uses two layers with two upper layer tracks and one lower layer track. The reader signals are processed by convolutional neural networks (CNNs) to separate the upper and lower layer signals and equalize them to 2-D and 1-D partial response (PR) targets, respectively. The upper and lower layer signals feed 2-D and 1-D Bahl–Cocke–Jelinek–Raviv (BCJR) detectors, respectively. The detectors’ soft outputs feed a multilayer CNN-based media noise predictor whose predicted noise outputs are fed back to the BCJR equalizers to reduce their bit error rates (BERs). The BCJR equalizers also interface with low-density parity-check (LDPC) decoders. Additional BER reductions are achieved by sending soft-information from the upper layer BCJR to the lower layer BCJR. Simulations of this turbo-detection system on a two-layer MLMR signal generated by a grain-switching-probabilistic (GSP) media model show density gains of 11.32% over a comparable system with no lower layer and achieve an overall density of 2.6551 terabits per square inch (Tb/in2).