Content area
Abstract
In this dissertation, detection and coding techniques for channels with intersymbol interference are studied and applied to the magnetic recording channel.
In Chapter Two, an optimization criterion for decision feedback equalization that minimizes the bit error probability is studied. It is shown that the optimization has the simple geometric interpretation of determining a hyperplane that maximally separates two groups of points.
In Chapter Three, a search algorithm that determines all the open and closed error events on partial response systems is used to tabulate low Euclidean distance error events for certain partial response polynomials. The lists of error events are used to calculate the first few terms of the generating function for the error event and bit error probability. The effect of noise correlation on the effective distance of error events is studied.
In Chapter Four, a detection technique called partial local feedback noise prediction is studied. This technique incorporates a noise predictor into the Viterbi detector branch metric calculation in order to improve its performance in correlated noise.
In Chapter Five, a different method for improving the performance of a Viterbi detector in correlated noise is developed. In this method, the samples going into the Viterbi detector are processed in blocks. A correlated noise branch metric is used within each block.
In Chapter Six, trellis codes that double the free squared Euclidean distance of the dicode channel are determined. These codes have finite truncation depth which allows the use of a Viterbi detector with a small amount of path memory.
In Chapter Seven, some modulo N trellis codes for binary partial response channels with an input constraint that prohibits the string '101' are developed.
