Content area
Abstract
In this dissertation, medium noise in thin film magnetic media is examined from the perspective of a communications systems engineer. In general, coding and detection schemes for magnetic recording channels are designed assuming the presence of only white Gaussian noise. However, with data being written at higher densities and more sensitive read heads being introduced, medium noise is becoming a dominant factor in system performance. Unlike white noise, medium noise is correlated, nonstationary, and signal dependent and, therefore, can affect system performance very differently.
Chapter Two describes transition noise in thin film media and introduces a model for this noise known as the microtrack model. A model for partial erasure is incorporated into the microtrack model. In Chapter Three, the noise produced by the model is characterized.
In Chapter Four, one of the most important applications of the microtrack model, bit-by-bit simulation, is examined. Methods for experimentally setting the parameters for the microtrack model and for using the model to simulate the performance of detection schemes in the presence of transition noise with partial erasure are discussed. Results showing the effect of medium noise on the performance of PRML systems are presented.
The microtrack model is expanded to include some nonlinear effects of the medium and their interactions with transition noise and partial erasure in Chapter Five. The resulting effects are examined and compared with those seen in experiment.
In Chapter Six, the effective distance $(d\sb{eff})$ of error events for PRML systems in magnetic recording channels with medium noise are found using the microtrack model. These results are compared to the case of no medium noise. The presence of medium noise is shown to have a strong negative impact on the minimum effective distances. The effective distances and the hierarchy of error events are shown to depend on the characteristics of the medium noise.
In Chapter Seven, two detection techniques are employed with PRML systems to achieve a performance improvement in the presence of medium noise. The first technique is noise prediction which takes advantage of the correlated nature of medium noise. Next, data dependent noise prediction is introduced to also take advantage of the fact that medium noise is data dependent.
