Granted Patent B1

US 8,760,782 · App. 13/404,380 · Granted Jun 24, 2014

Abstract

Claims (55)

1. A disk drive comprising:

a processor configured to decode data written over a plurality of tracks of a disk using joint self-iterating soft equalization and an iterative turbo two-dimensional (2-D) maximum a-posteriori (MAP)-based detection, the processor being coupled to a memory and being configured to:

receive input samples from a data channel;

perform soft self-iterating linear 2-D minimum mean square error (MMSE) equalization on the received input samples;

provide a 2-D equalization filter having a predetermined order that defines a number of rows and columns;

compute MMSE estimates for coefficients of the provided 2-D equalization filter based on a-priori values, mean of the a-priori values, the 2-D equalization filter and the received input samples;

update the coefficients of the 2-D equalization filter using a variance based on the a-priori values when carrying out a first iteration and based on an extrinsic Log Likelihood Ratio (LLR) when carrying out second and subsequent iterations;

determine whether individual bits in the input samples are 0s or 1s using the updated coefficients;

compute an output LLR of each individual received sample based on the computed MMSE estimate, the mean and the variance, and

compute the extrinsic LLR by subtracting a priori LLR from the output LLR.

2. The disk drive of claim 1 , wherein a-priori values are initialized to zero during the first iteration.

3. The disk drive of claim 1 , wherein the processor is further configured to apply a scale factor that is less than 1 to the extrinsic LLR.

4. The disk drive of claim 1 , wherein the processor is further configured to update the coefficients by at least one of:

adaptively using least mean squares algorithms; and

non-adaptively using a Weiner-Hopf equation.

5. The disk drive of claim 3 , wherein the processor is further configured to feed the scaled extrinsic LLR to a MAP detector and scale an output of the MAP detector to generate the a-priori values.

6. The disk drive of claim 3 , wherein the processor is further configured to feed the scaled extrinsic LLR to one of a 2-D detector, a soft output Viterbi algorithm and a Max-log MAP detector and scale an output thereof to generate the a-priori values.

7. The disk drive of claim 5 , wherein for the MAP detector, the processor is further configured to:

select a mask configured to account for Inter-Symbol-Interference (ISI), the mask comprising selected channel coefficients and the mask defining a trellis defining a plurality of states;

generate equalized samples from the input samples, each of the equalized samples corresponding to an input pixel;

compute branch probabilities for each received input pixel based on a Gaussian noise distribution using the equalized samples and ideal samples;

compute forward and backward probabilities via recursions using the computed branch probabilities for each of the plurality of states of the trellis;

combine the computed forward probabilities, backward probabilities and the branch probabilities to generate a-posteriori probabilities;

marginalize the a-posteriori probabilities over values of neighboring pixels to generate an a-posteriori probability for a pixel in a given state; and

decode the pixel as a first logical state or a second logical state from the marginalized a-posteriori probabilities.

8. The disk drive of claim 7 , wherein the processor is further configured to decode the pixel by:

summing each of the plurality of states of the trellis corresponding to the pixel assuming a value of +1 and summing each of the plurality of states of the trellis corresponding to the pixel assuming a value of −1;

calculating a ratio of a sum over all the states corresponding to pixels that are +1 and a sum of all states corresponding to pixels that are −1 to obtain an output pixel log likelihood ratio (LLR), and

decoding the current pixels as a 1 if the output pixel LLR is greater than 0 and decoding the current pixels as a 0 if the output pixel LLR is less than or equal to 0.

9. A computer-implemented method of decoding data written over a plurality of tracks of a disk of a disk drive using joint self-iterating soft equalization and an iterative turbo two-dimensional (2-D) maximum a-posteriori (MAP)-based detection, comprising:

receiving input samples from a data channel;

performing soft self-iterating linear 2-D minimum mean square error (MMSE) equalization on the received input samples;

providing a 2-D equalization filter having a predetermined order that defines a number of rows and columns;

computing MMSE estimates for coefficients of the provided 2-D equalization filter based on a-priori values, mean of the a-priori values, the 2-D equalization filter and the received input samples;

updating the coefficients of the 2-D equalization filter using a variance based on the a-priori values when carrying out a first iteration and based on an extrinsic Log Likelihood Ratio (LLR) when carrying out second and subsequent iterations;

determining whether individual bits in the input samples are 0s or 1s using the updated coefficients;

computing an output LLR of each individual received sample based on the computed MMSE estimate, the mean and the variance, and

computing the extrinsic LLR by subtracting a priori LLR from the output LLR.

10. The computer-implemented method of claim 9 , further comprising initializing the a-priori values to zero during the first iteration.

11. The computer-implemented method of claim 9 , further comprising applying a scale factor that is less than 1 to the extrinsic LLR.

12. The computer-implemented method of claim 9 , wherein the coefficient updating is carried out at least one of adaptively using least mean squares algorithms and non-adaptively using a Weiner-Hopf equation.

13. The computer-implemented method of claim 9 , further comprising feeding the scaled extrinsic LLR to a MAP detector, scaling an output of the MAP detector to generate the a-priori information.

14. The computer-implemented method of claim 9 , further comprising feeding the scaled extrinsic LLR to one of a 2-D detector, a soft output Viterbi algorithm and a Max-log MAP detector and scaling an output thereof to generate the a-priori values.

15. The computer-implemented method of claim 13 , wherein for the MAP detector, the method further comprises:

selecting a mask configured to account for Inter-Symbol-Interference (ISI), the mask comprising selected channel coefficients and the mask defining a trellis that defines a plurality of states;

generating equalized samples from the input samples, each of the equalized samples corresponding to an input pixel;

computing branch probabilities for each received input pixel based on a Gaussian noise distribution using the equalized samples and ideal samples;

computing forward and backward probabilities via recursions using the computed branch probabilities for each of the plurality of states of the trellis;

combining the computed forward probabilities, backward probabilities and the branch probabilities to generate a-posteriori probabilities;

marginalizing the a-posteriori probabilities over values of neighboring pixels to generate an a-posteriori probability for a pixel in a given state; and

decoding the pixel as a first logical state or a second logical state from the marginalized a-posteriori probabilities.

16. The computer-implemented method of claim 15 , wherein decoding further comprises:

summing each of the plurality of states of the trellis corresponding to the pixel assuming a value of +1 and summing each of the plurality of states of the trellis corresponding to the pixel assuming a value of −1;

calculating a ratio of a sum over all the states corresponding to pixels that are +1 and a sum of all states corresponding to pixels that are −1 to obtain an output pixel log likelihood ratio (LLR), and

decoding the pixels as a 1 if the output pixel LLR is greater than 0 and decoding the pixels as a 0 if the output pixel LLR is less than or equal to 0.

Assignments (8)