Norman L. Koren

Matched filter limits and code performance in digital magnetic recording

Presents a method for evaluating the performance of recording channels and codes for systems with additive spectral Gaussian noise. Starting with a noise-whitened isolated pulse, matched filters are calculated for the amplitude and timing channels which have the maximum signal-to-noise ratio (SNR) and minimum jitter, respectively. In high-density recording systems, the pulse slimming equalization required to control intersymbol interference degrades the SNR and jitter. For the timing channel, a detector SNR is defined whose error statistics are nearly identical to the amplitude channel SNR. Procedures for calculating losses relative to the matched filter SNR are developed that facilitate performance comparisons of both channels. A series of runs for the Lorentzian input pulse and cos/sup n/ equalized output pulses compares the performance of several modulation codes and detection techniques. Write equalization and partial response type four equalization are examined. >

Signal processing in recording channels utilizing unshieldede magnetoresistive (UMR) heads

Because unshielded magnetoresistive (UMR) heads offer excellent short-wavelength sensitivity, immunity from Barkhausen noise, a velocity-independent readback signal, and simple construction, they were chosen for use in a new high-performance, multichannel tape system that operates at 80 kbit/in with a 1.5-mil track width. UMR heads have two properties that present a challenge to system designers: (1) a tendency to saturate with large applied magnetic fields, and (2) a high sensitivity to long-wavelength signals and noise which corresponds to an isolated pulse that extends far beyond the systems minimum transition time. These problems were solved through a combination of write and read equalization. System performance, measured by amplitude and timing margins, is shown to be excellent. In selecting write equalization parameters for amplitude-qualified peak-detecting systems, both amplitude and timing margins are significant. Write-equalized pulses that superficially appear to be highly irregular can frequently be effectively read equalized. The only caution in using write equalization is that write current risetime must be short enough to fully record the added pulses. This should not be a problem in systems with separately optimized write and read heads. >

Signal processing for single and dual element unshielded magnetoresistive heads

Single and dual element unshielded magnetoresistive (UMR and DMR) heads have potential for extremely high density magnetic recording, but saturation and unusual readback waveforms must be dealt with. Two new peak detector channels take advantage of the unusual waveforms: (1) The quasi-integrating channel for the UMR, whose output is similar to that of integrating channels with inductive read heads. Performance may be enhanced by adding short pulses to the write current at distances greater than T/sub min/ from code 1 transitions. (2) Controlled polarity recording with decision feedback, which is a special case of write equalization where successive pulses may have the same polarity. The detection window is twice that of conventional recording. Simulated performance is comparable to PRML channels. >

(1+D)/sup n/ partial response channels for UMR heads

Most partial response channels described in the magnetic recording literature are characterized by the (1-D)(1+D)/sup n/ pulse response polynomial, where n=1 for PR4 and 2 for EPR4. Partial response channels with (1+D)/sup n/ pulse response polynomials, which include partial response classes 1 and 2 (PR1(1+D) and PR2(1+D)/sup 2/), have received scant attention because magnetic recording systems lack the DC component implicit in these polynomials. A new channel developed for the unshielded magnetoresistive (UMR) head that closely approximates a (1+D)/sup n/ channel at all but the very lowest frequencies and works well when used with DC-free modulation codes is discussed. This channel is an excellent match to the UMR head response, which has greater long wavelength response than the more familiar inductive and shielded MR heads. >