Soft-decision decoding of Reed-Solomon codes using pattern information over partial response channels.

摘要

Reed-Solomon (RS) codes have been used for a long time in many data storage systems for error correction. Popularity of RS codes is due to their powerful burst error correction capability and well-established decoding algorithm with efficient hardware implementation providing high throughput. Conventional hard-decision decoding algorithms are able to correct up to half the minimum distance.Many soft-decision decoding algorithms for RS codes have been proposed and have been investigated for application to magnetic recording channels. However, even with increased computing power, the complexity of soft-decision decoding algorithms is too high for them to be employed in real applications which require high data rate, small hardware area and low power.Some soft-decision decoding algorithms adopt bit-based decoding schemes using bit reliability. Performance gains in magnetic recording channels provided by the bit-based soft-decision decoding algorithms are not as good as those in memoryless additive white Gaussian Noise (AWGN) channels, because of the burst error characteristics of magnetic recording channels. On the other hand, symbol-based soft-decision decoding algorithms can extend burst, error correction capability, but require excessively high complexity.In this thesis, we propose a series of algorithms for soft-decision decoding of RS codes: successive error-and-erasure correction decoding, pattern-flipping Chase-type algorithms, reliability matrix construction method and multiplicity assignment algorithm for Koetter-Vardy (KV) algorithm. All proposed algorithms target partial response (PR) channels of use in magnetic recording channels where burst errors are dominant.To achieve higher areal density needed these days, enhanced error correction capability is required. Soft information, which is probability information about binary decision, can be used to correct more errors than half the minimum distance. Rapid increase of computing power enables the systems to execute complicated decoding algorithms using soft information.While most soft-decision decoding algorithms use soft information for bits or symbols, the proposed algorithms take advantage of soft information for patterns (An error pattern can extend to multiple bits and straddle symbol boundaries). In order to provide probable error patterns and corresponding reliabilities for the proposed algorithms, we also propose a modified Viterbi algorithm, named pattern-output Viterbi algorithm. Each proposed algorithm is supported by simulations and analysis, to show good competitive performance and complexity.