Detection and error correction for partial response channels
In the first part of the dissertation, we explore the maximum-likelihood sequence estimation (MLSE) detector for binary input intersymbol interference channels corrupted by independent and identically distributed Gaussian noise. The standard method of implementing the MLSE detector is commonly referred to as the state metric implementation. Here, we present a new alternative implementation referred to as the difference metric implementation. The difference metric approach is a generalization of a technique currently used to implement the MLSE detector for the (1-D) channel.
In the second part of the dissertation, we focus on error-correction coding for the partial response channels commonly used in high-density recording systems, which are the (1 $\pm$ D), (1 $-$ D)(1 + D), (1 + D)$\sp2$, and (1 $-$ D)(1 + D)$\sp2$ channels. We review the current two-stage decoding process where the channel and the code are decoded separately, using an MLSE detector for the channel and a soft-decision decoder for the code. Since the soft-decision decoder is designed to correct random errors, the code is typically time interleaved to randomize the correlated errors at the output of the MLSE detector.
Here, we introduce a new two-stage decoding process where the soft-decision decoder corrects error-events, rather than random errors. At high signal-to-noise ratios, minimum distance error-events are the dominant error-events. Therefore, we present a low-complexity error-event decoder which corrects only minimum distance error-events. In this case, an error-correction code is used to detect minimum distance error-events, and an error-event decoder is used to correct the most-likely sequence of minimum distance error-events satisfying the code. We apply this technique to matched spectral-null (MSN) codes and linear block codes and show that the new two-stage decoder is a viable alternative to conventional techniques.