Multitrack Detection with 2D Pattern-Dependent Noise Prediction
Shanwei Shi, John R. Barry · 2018
The advent of multiple readers in magnetic recording opens the door to multitrack detection, in which multiple tracks are detected jointly. Multitrack detection is a key enabler for both coding across tracks (including modulation and error-control codes) and crosstrack noise prediction, neither of which can be fully exploited using single-track detectors. In this paper, we propose the two-dimensional pattern- dependent noise-prediction (2D-PDNP) algorithm as a solution to the joint maximum-likelihood multitrack detection problem in the face of pattern-dependent autoregressive Gaussian noise. The solution takes the form of the Viterbi algorithm over a trellis that models the combined memory of the channel and noise, with a branch metric that can be interpreted as 2D pattern- dependent noise prediction, where noise is predicted in both downtrack and crosstrack directions, taking into account transitions occurring in both downtrack and crosstrack directions. Numerical results show that, on a set of quasi-micromagnetic simulated channel waveforms with a pair of readers and a multitrack detector detecting two tracks simultaneously, the 2D-PDNP algorithm provides a 4% increase in areal density.