PARALLEL SUBSPACE SUBCODES OF REED-SOLOMON CODES FOR MAGNETIC RECORDING CHANNELS

Abstract

Read channel architectures based on a single low-density parity-check (LDPC) code are being considered for the next generation of hard disk drives. However, LDPC-only solutions suffer from the error floor problem, which may compromise reliability, if not handled properly. Concatenated architectures using an LDPC code plus a Reed-Solomon (RS) code lower the error-floor at high signal-to-noise ratio (SNR) at the price of a reduced coding gain and a less sharp waterfall region at lower SNR. This architecture fails to deal with the error floor problem when the number of errors caused by multiple dominant trapping sets is beyond the error correction capability of the outer RS code. The ultimate goal of a sharper waterfall at the low SNR region and a lower error floor at high SNR can be approached by introducing a parallel subspace subcode RS (SSRS) code (PSSRS) to replace the conventional RS code. In this new LDPC+PSSRS system, the PSSRS code can help localize and partially destroy the most dominant trapping sets. With the proposed iterative parallel local decoding algorithm, the LDPC decoder can correct the remaining errors by itself. The contributions of this work are: 1) We propose a PSSRS code with parallel local SSRS structure and a three-level decoding architecture, which enables a trade off between performance and complexity; 2) We propose a new LDPC+PSSRS system with a new iterative parallel local decoding algorithm with a 0.5dB+ gain over the conventional two-level system. Its performance for 4K-byte sectors is close to the multiple LDPC-only architectures for perpendicular magnetic

xviii

recording channels; 3) We develop a new decoding concept that changes the major role of the RS code from error correcting to a "partial" trapping set destroyer.