Equalization for sample value estimation and sequence detection in a sampled amplitude read channel

Abstract

In a sampled amplitude magnetic read channel, pulses in an analog signal corresponding to flux transitions on a magnetic medium are sampled and equalized into a first equalization for estimating sample values and into a second equalization for sequence detection of digital data. A gain and phase error detector generate respective error signals corresponding to the difference between estimated and actual sample values. Gain control and timing recovery use the error signals to adjust the amplitude and sampling frequency/phase of the analog read signal. A pair of programmable discrete time filters equalize the signal samples into the desired equalization. In a first embodiment, the signal samples are equalized to PR4 for estimating sample values and to EPR4 for sequence detection. A slicer processes the PR4 equalized sample values to generate the estimated sample values. The gain and phase error detectors generate the corresponding error signals according to a minimum mean squared error stochastic gradient algorithm. In a second embodiment, the signal samples are equalized to EPR4 for estimating sample values and to EEPR4 for sequence detection. A pulse detector processes the EPR4 equalized sample values in order to detect pulses in the analog signal and generate a corresponding pulse detect signal. The gain and phase error detectors generate the corresponding error signals in response to the pulse detect signal. To minimize circuitry and associated cost, the discrete time equalizing filters are implemented in series. For d=1 recording, the second equalizing filter is a (1+D) notch filter that attenuates the noise caused by clocking the discrete time circuitry at half the sampling rate, thereby increasing the accuracy of the sequence detector.

Dependencies are inferred from the claim text.