Signal Processing and Coding for Two-Dimensional Magnetic Recording
Many novel approaches have been proposed to improve storage densities of magnetic recording beyond the existing 1 Tb/in2. These include heat-assisted magnetic recording (HAMR), bit-patterned media (BPM) and two-dimensional magnetic recording (TDMR). TDMR is a promising technology that relies on sophisticated signal processing algorithms driven within a systems framework to achieve high areal densities. In TDMR, the tracks are closely packed together to reduce the bit-size, thereby increasing areal densities. This allows for 2D inter-symbol interference (ISI) and other artifacts that need to be mitigated using the signal processing algorithms. Additional areal density gains can be achieve by deploying HAMR/BPM along with TDMR. Other challenges in TDMR include 2D timing recovery and handling 2D burst erasures. The timing errors in TDMR include jitter and frequency components in both the directions that are interdependent on each other. Burst erasures occur in magnetic recording due to thermal asperities, inherent defects/scratches on the medium, as well as, due to mechanical vibrations. The effect of reducing the bit-size in TDMR results in (1) increased ISI in both crosstrack and down-track directions resulting in 2D-ISI, and 2) increased jitter noise due to irregularities on the medium. 2D-ISI MAP detection is known to be NP-hard problem. Therefore, we need detection algorithms endowed with noise prediction capability that come close to near maximum likelihood (ML) performance with tractable computational complexity. In this dissertation, we discuss our novel contributions to TDMR and physical storage technologies that include 1) Design of 2D separable and non-separable partial response (PR) targets for partial-response maximum likelihood detection; 2) 2D softoutput Viterbi algorithm (2D-SOVA): A tractable 2D-ISI detector that operates on 2D surfaces that inherently handles the jitter noise; 3) Joint timing recovery and signal detection using 2D-SOVA; 4) A 2D defect detection algorithm that accurately identifies the region of 2D burst erasures for subsequent decoding using LDPC codes; 5) Construction of a native 2D LDPC codes that can effectively correct 2D burst erasures 6) Exact performance analysis of 1D sequential detectors towards obtaining a look-up table based most optimal sequential detector, that performs better than the Viterbi algorithm; 7) A low complexity approximation of the GBP algorithm. Our work is directly extendable to other multi-dimensional storage technologies such as 3D NAND flash and holographic storage.