Knowledge-based parameter estimation for identification and equalization of storage channels

The emerging sampling detectors in digital magnetic recording require precise filtering of the readback signal. In this paper, we propose a technique for fast estimation of the channel response from which the filter coefficients are calculated. Utilizing the a priori knowledge about the general shape of the transition response in magnetic recording, the channel identification problem is reduced to estimation of one or more parameters. Specifically, the pulse width at half of the transition response peak magnitude is first estimated. The algorithm is then extended to include more practical cases. It is shown that this simplified approach has much faster convergence rate than the direct least-mean-square channel identification method. Once the algorithm converges, the estimates can be used to select or compute an appropriate set of equalizer coefficients or to modify the decoder parameters. We will describe methods for recursive filter design based on the estimated channel for partial response as well as decision feedback systems.