Modeling, detection, and adaptive signal processing for magnetic disk recording.

摘要

At the current rate, the storage capacity of magnetic disk drives doubles every 18 months. An important factor in maintaining this trend is improvements in signal processing. A relative newcomer to the disk drive industry, read channels with signal processing other than the traditional method of peak detection first appeared in IBM products in 1990. Today, nearly all hard disk drives under design include advanced signal processing as a means of increasing the capacity and yield.; The reason for including such advanced channels was simple. For a relatively small additional cost, the first alternative to peak detection, known as partial response maximum likelihood (PRML), boosted the storage density by 30% and was able to adapt to variations in manufacturing, pushing the yield higher as well. As the feature size and power of integrated circuits continue to fall, more complex signal processing will remain a cost-effective way of meeting the increasing demands of storage density.; This dissertation begins by introducing current state-of-the-art signal processing technology with a look to future technology. The second chapter characterizes a new model of the magnetic recording channel known as the microtrack model. This new model is used to help analyze a new nonlinear signal processing algorithm presented in the third chapter. Known as the Random Access Memory, Reduced-state Sequence Estimator, the RAM-RSE will help meet the areal density goal of 10 Gbits/in{dollar}sp2.{dollar} For a relatively modest rise in complexity, the RAM-RSE yields another 20% to 25% increase in density to that of PRML. The presentation of a new structure for decision feedback-style detectors follows in Chapter Four. The new structure improves the conditioning of the adaptive process for decision feedback detectors like the RAM-RSE. The dissertation concludes with a brief discussion on topics of future research.