James Fitzpatrick

A generalized frequency domain nonlinearity measurement method

In this paper, we discuss an accurate and effective nonlinearity measurement method that is generalized from the frequency domain method proposed by Y.S. Tang and C. Tsang(1991). The new method discussed here offers the following features. It can test any recording density by observing only one fixed frequency. This method can be used to study nonlinearities due to interactions between two transition groups such as a tribit interacting with a single transition. By following a general rule given in this paper, one can construct a recording pattern that simultaneously satisfies code constraints and allows the measurement of nonlinearities. Therefore, it allows a nonlinearity measurement on a HDA. >

The dependence of overwrite on proximity effects

The widely held theories of overwrite (OW) on thin film media predict very little sensitivity to the low frequency used in the measurement. Here experimental data is shown with a variation of more than 15 dB as the low frequency is changed from 1/2 to 1/14 of the high frequency pattern. Tang and Tsangs work (1993) on interpattern interactions is used to explain the sensitivity of OW to the specific frequencies used in the measurement.

The relationship of medium noise to system error rate in a PRML channel

High density disk drive systems utilize fine grain polycrystalline thin films as storage medium. These films exhibit noise that is pattern dependent and has appreciable magnitude near the center of each recorded transition. A model for error rates in a PRML channel is presented to explore how variations in medium noise parameters effect system performance. The channel is assumed to be characterized by both pattern dependent nonstationary medium noise and white Gaussian electronic and/or head noise. Results are presented in terms of both error rate and sensitivity of error rate to changes in medium parameters. >

An evaluation of partial response polynomials for magnetic recording systems

The class of partial response polynomials of the form (1-D)(1+D)/sup n/ are evaluated in recording systems with both electronic noise and nonstationary media noise. The detector performance for each polynomial is evaluated versus normalized recording density for systems with 100% electronic noise, and then for systems with 100% media noise. Simple mathematical relations are developed that show how to convert these results into plots of detector performance for contours of constant areal density. In the case of 100% electronic noise, it is shown that with the proper choice of bit length and width EPR4 yields the lowest probability of error at any areal density. Finally, a technique for combining the above results is presented so that systems with arbitrary proportions of electronic and media noise can be evaluated. >