Beyond PRML: linear-complexity turbo equalization using the soft-feedback equalizer

The combination of partial response and trellis-based maximum-likelihood sequence detection that dominates magnetic recording is known as PRML. The receiver front-end includes an analog filter whose role is to essentially shorten the impulse response of the underlying channel; this filter transforms the channel response into a target partial response that has very little memory, so that a trellis-based equalizer will have a manageable number of states. There are several drawbacks of the PRML approach that only grow worse as areal densities increase which includes penalties arising from noise enhancement and correlation in the noise. We propose an alternative equalization architecture for magnetic recording that addresses all of the shortcomings of the PRML approach. Specifically, we propose to abandon the PR strategy altogether; we abandon trellis-based equalizers in favor of simple equalization strategies based on nonlinear filters whose complexity grows only linearly in their length; and we propose an integration of the proposed structure into a turbo equalization framework. This paper has proposed the use of a linear-complexity algorithm as a promising alternative to partial-response for magnetic recording. The SFE algorithm can provide up to 20% more capacity than EPR4-based detectors while maintaining low complexity costs.