Decomposition of noise sources in recording applications using symbol-rate readback samples

Noise in recording applications is often modeled as a combination of additive noise and medium noise. In addition to the total noise power, the nature of the different noise processes and their relative contributions to the total noise affect detection performance. Two methods for noise decomposition that are based on symbol-rate readback samples of arbitrary channel input sequences are presented. They rely on the estimation of the variance of an ensemble of noise samples as a function of sample index, and its cross correlation with the input data sequence. Performance is assessed by using well-established models of the magnetic recording channel.