CAREER: Multitrack Read Channel Designs for Modern Two-Dimensional Magnetic Recording

Project: Research project

Project Details

Description

This project develops efficient noise prediction, synchronization, and symbol detection algorithms that support modern and future generations of ultra-high density two-dimensional magnetic recording. Two-dimensional magnetic recording is a novel technology in hard disk drives that allows a drastic increase in the data density to up to 10 Terabits per square inch, on already-existing head and media designs. This is achieved through shingled writing, where the adjacent data tracks are written with partial overlap on top of each other, like roof shingles, in order to squeeze many more tracks on the disk. Powerful signal processing algorithms enable efficient data recovery from highly interference-laden and noisy readback signals. This project investigates powerful signal processing algorithms, from purely communication-theoretic to machine learning models, and their combinations, for data recovery in such channels. The results of this project are expected to increase the data density of two-dimensional magnetic recording well beyond the current state-of-the-art. The project integrates an educational component in signal processing and communication theory in the form of 1) graduate student training, 2) research-oriented undergraduate student education, and 3) collaborative participation with industrial research.The novel elements of this project are that: 1) it adopts a multiple-input multiple-output (MIMO) model in accordance with the industry expectation for low-latency implementation; 2) it develops multitrack detection strategies as opposed to the current industry standard of single-track detection, in order to reach higher areal densities as well as throughput; and 3) in addition to considering all the channel impediments that are often considered separately in earlier works, this project also addresses the problem of timing asynchrony between the adjacent tracks and the analog-to-digital converter sampling rate. To achieve the above goals, four research thrusts will be pursued. In the first thrust, reduced-state redesigns of the trellis-based multitrack symbol detectors will be investigated. The second thrust develops media noise mitigation techniques for asynchronous multitrack detection. Here, both the MIMO extension of the pattern-dependent noise prediction algorithm and neural network noise predictor models will be studied. In the third thrust, deep neural network symbol detectors will be developed, both as a stand-alone joint symbol detector and synchronizer, as well as a joint symbol detector and synchronizer coupled with a low-density parity check decoder following the turbo detection mechanism. The fourth thrust develops novel read channels built entirely of neural networks, both as a separate network equalizer followed by a network symbol detector, and as a holistic deep neural network read channel.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
StatusActive
Effective start/end date1/07/2330/06/28

Funding

  • National Science Foundation

Fingerprint

Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.