Content area
Abstract
In this dissertation, the issues of coding and signal detection for digital magnetic recording systems with partial response channels are studied. Partial response channels are channels with controlled intersymbol interference. The topics researched in this dissertation are methods for combating Thermal Asperities, methods of improving signal detection for colored and media noise, DC offset invariance for partial response channels, and a postprocessing approach utilizing single-parity interleaved block codes.
Chapter 1 serves as an introduction and summarizes some of the previous literature on the subjects covered in this dissertation.
Chapter 2 describes Thermal Asperities and the existing methods of countering them. A new method is proposed and its performance is found by computer simulation. The new method utilizes two channels that are running in parallel. The output of the unmodified channel (EPR4) is selected when no Thermal Asperities are detected. The output of the second channel (PR5) with a (1-D) filter is selected when a Thermal Asperity is detected. The (1-D) filter improves the bit detection for the bits impacted by a Thermal Asperity.
In Chapter 3 a trellis expansion is proposed and studied which serves to close the SNR performance gap between the AWGN performance of the EPR4 channel and the PR5 channel (proposed in Chapter 2). The proposed method also improves the detrimental effect of the noise coloration caused by an equalizer without the undesirable effect of error propagation which occurs in noise prediction methods.
In Chapter 4 the performance of EPR4 and PR5 channels is studied under the conditions of AWGN noise, media noise, and the mix of these two noises.
Chapter 5 concentrates on the impact of DC offset on the detection process. For a particular class of the partial response channels it is proved that an unknown DC offset does not impact the performance of a maximum-likelihood sequence detector.
Chapter 6 contains a study of postprocessing for a partial response channel when the data is encoded using an interleaved, single-parity check code. A matched filter is used to locate the error patterns.
Chapter 7 contains a summary of the dissertation.
