Content area
Abstract
In this dissertation we investigate coding for, and the performance of, detectors on recording systems in which controlled amounts of intersymbol interference (ISI) is allowed. New modulation codes for a peak detector that allows ISI to occur are developed in Chapter 2. In the next several chapters, we estimate the performance of a Viterbi detector, based upon a partial response (PR) polynomial h(D), in a recording system with Lorentzian step response. Due to noise correlation, the nonstationary nature of media noise, and misequalization, the Viterbi detector is not a maximum likelihood detector for this channel. In Chapter 3 we examine the optimal choice of h(D) assuming that no modulation code is used, the recording channel is linear, and the playback signal is embedded in additive white Gaussian noise. In Chapter 4, we compare run length limited modulation codes with spectrally constrained modulation codes. The spectrally constrained codes are shown to have the larger minimum distance error events for a fixed information density on the media. In Chapter 5 we reconsider the choice of PR polynomial under the assumption that a DC/Nyquist free modulation code is used. The best PR polynomials from Chapter 3 also are the best polynomials in the presence of a DC/Nyquist free modulation code. In addition, the class I, (1 + D), PR polynomial is shown to be a low complexity alternative to polynomials of the form (1 + D)$\sp{n}$(1 $-$ D). In Chapter 6, we present a sophisticated model of media noise and describe how it has been integrated into a software package for evaluating the choice of modulation code in light of the specific head and media of a system. Finally, in Chapter 7 we examine the architecture of a maximum likelihood (ML) detector for nonlinear magnetic recording channels. It is shown that the ML detector for the nonlinearities in magnetic recording is considerably simpler than the ML detector for general nonlinear channels.
