Adaptive data-dependent noise-predictive maximum likelihood detection for channels with non-linear transition shifts

A method comprising:

• receiving a signal representing data stored on a medium;
• applying one or more noise whitening filters to the signal using a bank of noise whitening filters to produce a filtered signal;
• estimating a data-dependent noise mean from the filtered signal using a data-dependent noise mean calculator;
• performing one or more branch metric calculations on the filtered signal using a branch metric calculator to produce one or more branch metrics, the one or more branch metric calculations accounting for the data-dependent noise mean;
• and
• generating an output stream from the one or more branch metric calculations using an adaptive data-dependent noise-predictive maximum likelihood (D3-NPML) detector.

A data storage system, comprising:

• a magnetic head configured to produce a signal representing data stored to a magnetic tape;
• a least mean squared (LMS) engine configured to produce an array of predictor coefficients based on the signal;
• a bank of noise whitening filters configured to apply one or more noise whitening filters to the signal to produce a filtered signal using one or more predictor coefficients from the array of predictor coefficients;
• a data-dependent noise mean calculator configured to estimate a data-dependent noise mean from the filtered signal using exponential averaging;
• a branch metric calculator configured to perform one or more branch metric calculations on the filtered signal to produce one or more branch metrics, the one or more branch metric calculations accounting for the data-dependent noise mean;
• and
• an adaptive data-dependent noise-predictive maximum likelihood (D3-NPML) detector configured to generate an output stream from the one or more branch metric calculations.