Systems and methods for adaptation of a two-dimensional magnetic recording channel

Patent metadata

Field	Value
Publication number	US-11145331-B1
Application number	US-202017084370-A
Country	US
Kind code	B1
Filing date	Oct 29, 2020
Priority date	Oct 29, 2019
Publication date	Oct 12, 2021
Grant date	Oct 12, 2021

Abstract

Systems and methods for adaptation of a two-dimensional magnetic recording (TDMR) channel are provided. Read-back signals from respective read sensors of a TDMR channel are received at an equalizer, the read-back signals corresponding to a digital signal value. A log-likelihood ratio (LLR) signal is generated based at least in part on the read-back signals. A cross-entropy value is generated indicative of a mismatch between a probability of detected bit and a probability of the true recorded bit. The equalizer is adapted by setting an equalizer parameter to a value that corresponds to a minimum cross-entropy value from among the computed cross-entropy value and one or more previously computed cross-entropy values, to decrease a read-back bit error rate for the TDMR channel.

First claim

What is claimed is:

1. A method for adaptation of a two-dimensional magnetic recording (TDMR) channel, comprising:
receiving, at an equalizer, read-back signals from respective read sensors of a TDMR channel, the read-back signals corresponding to a digital signal value;
generating a log-likelihood ratio (LLR) signal based at least in part on the read-back signals;
computing a cross-entropy value indicative of a mismatch between a probability of detected bit and a probability of the true recorded bit; and
adapting the equalizer by setting an equalizer parameter to a value that corresponds to a minimum cross-entropy value from among the computed cross-entropy value and one or more previously computed cross-entropy values, to decrease a read-back bit error rate for the TDMR channel.

2. The method for adaptation of a TDMR channel claimed in claim 1 , wherein the equalizer comprises a plurality of filter taps having a plurality of coefficients, respectively, and wherein the adapting the equalizer based on the cross-entropy value comprises setting one or more of the plurality of coefficients to one or more respective values that correspond to a minimum cross-entropy value from among the computed cross-entropy value and one or more previously computed cross-entropy values.

3. The method for adaptation of a TDMR channel claimed in claim 1 , wherein the read-back signals contain inter-symbol interference (ISI) from data recorded on the media, and wherein the LLR signal is generated by feeding an output of the equalizer through a a soft sequence detector configured to remove inter-symbol interference.

4. The method for adaptation of a TDMR channel claimed in claim 1 , wherein generating the LLR signal comprises:
generating branch metrics at a Viterbi detector based on an output of the equalizer; and
generating the LLR signal at a soft output Viterbi algorithm module based on the branch metrics, the LLR signal being a soft output indicative of both a detected digital bit value and a reliability of a likelihood that the detected digital bit value is accurate.

5. The method for adaptation of a TDMR channel claimed in claim 1 , wherein computing the cross-entropy value comprises:
receiving, at the equalizer via the read sensors of the TDMR channel, a plurality of read-back signals corresponding to a plurality of digital bit values stored on a recording medium, the plurality of digital bit values representing a set of training data;
computing a plurality of cross-entropy values for the plurality of digital bit values, respectively;
computing an average cross-entropy function based on the plurality of cross-entropy values; and
utilizing the average cross-entropy function as a cost function for the adapting of the equalizer.

6. The method for adaptation of a TDMR channel claimed in claim 1 , wherein computing the cross-entropy value comprises computing a gradient of between cross-entropy values and values of coefficients of filter taps of the equalizer.

7. The method for adaptation of a TDMR channel claimed in claim 6 , wherein the TDMR channel comprises a branch metric unit (BMU) and a path metric unit (PMU), and wherein computing the gradient between the cross-entropy values and the values of the coefficients of the filter taps of the equalizer comprises computing a gradient between LLR values and values of a path metric difference (PMD) of the PMU, a gradient between the values of the PMD and values of a branch metric (BM) of the BMU, and a gradient between the values of the BM and the values of the coefficients of the filter taps.

8. The method for adaptation of a TDMR channel claimed in claim 7 , wherein the BMU comprises a Viterbi detector configured to generate branch metrics based on an output of the equalizer.

9. The method for adaptation of a TDMR channel claimed in claim 8 , wherein the PMU is configured to execute a soft output Viterbi algorithm (SOVA) based on the branch metrics to generate the LLR signal, the LLR signal being a soft output indicative of both a decoded digital bit value and a reliability of a likelihood that the decoded digital bit value is accurate.

10. The method for adaptation of a TDMR channel claimed in claim 6 , wherein adapting the equalizer based on the cross-entropy value comprises setting one or more of the coefficients to a value that corresponds to a minimum cross-entropy according to the computed gradient.

11. The method for adaptation of a TDMR channel claimed in claim 1 , wherein the read-back signals comprise error-correcting codes and wherein the method further comprises computing, at an error-correcting decoder, the digital signal value based on the error-correcting codes of the read-back signals.

12. The method for adaptation of a TDMR channel claimed in claim 1 , wherein the TDMR channel comprises a channel estimation filter having a plurality of filter tap coefficients, and wherein the method further comprises adapting the filter tap coefficients of the channel estimation filter based on the computed cross-entropy.

13. The method for adaptation of a TDMR channel claimed in claim 1 , wherein the equalizer is a nonlinear equalizer configured to perform nonlinear equalization upon the read-back signals to reduce non-linear noise originating from non-linear noise sources.

14. The method for adaptation of a TDMR channel claimed in claim 13 , wherein the nonlinear equalizer comprises a neural network including a plurality of hidden node layers, with each hidden node layer comprising a plurality of hidden nodes, and wherein the method further comprises executing hyperbolic tangent function activation functions (tanh) at the plurality of hidden nodes.

15. The method for adaptation of a TDMR channel claimed in claim 13 , further comprising:
estimating a plurality of bit error rates for a plurality of configurations of the nonlinear equalizer, respectively, each of the plurality of configurations corresponding to a number of hidden node layers having respective numbers of hidden nodes; and
configuring the nonlinear equalizer to have one of the plurality of configurations corresponds to a minimum bit error rate value from among the plurality of bit error rates.

16. The method for adaptation of a TDMR channel claimed in claim 13 , further comprising:
generating, based on a probability distribution function (PDF) of noise detected at an output of the nonlinear equalizer, a curve fitting model comprising a plurality of branch metric parameters; and

configuring the curve fitting model to be utilized as a modified branch metric of the BMU.

17. The method for adaptation of a TDMR channel claimed in claim 16 , further comprising adapting the values of one or more of the plurality of branch metric parameters to minimize the cross-entropy signal by setting values of the branch metric parameters to values that correspond to a minimum cross-entropy value from among the computed cross-entropy value and one or more previously computed cross-entropy values.

18. The method for adaptation of a TDMR channel claimed in claim 1 , wherein the equalizer is configured to perform linear equalization upon the read-back signals to reduce linear noise originating from linear noise sources.

19. A two-dimensional magnetic recording (TDMR) read-back channel, comprising:
equalizer circuitry configured to execute an equalization algorithm upon read-back signals received from respective read sensors, the read-back signals corresponding to a digital signal value;
soft sequence detector for an inter-symbol interference (ISI) channel

Assignees

• Marvell Asia Pte Ltd

Inventors

• Nangare Nitin
• Shen Jinlu

Classifications

• G11B20/10046 Primary

  filtering or equalising, e.g. setting the tap weights of an FIR filter · CPC title