Refresh, run, aggregate decoder recovery
View Patent ↗A data processing system includes a likelihood input operable to receive encoded data, a decoder operable to apply a decoding algorithm to likelihood values for the received encoded data and to yield a decoded output, and a decoder input initialization circuit operable to generate new decoder input values based in part on the likelihood values for the received encoded data after the likelihood values for the received encoded data have failed to converge in the decoder.
1. A data processing system comprising:
a memory;
a data input of the memory to receive encoded data, the encoded data comprising a log likelihood ratio input;
a decoder operable to apply a decoding algorithm to likelihood values for the received encoded data and to yield a decoded output when the likelihood values for the received encoded data fail to converge in the decoder; and
when one or more likelihood values for the received encoded data fail to converge in the decoder, a decoder input initialization circuit operable to:
generate new decoder input values based at least in part on the likelihood values for the received encoded data, wherein sign values used for parity bits in the new decoder input values are derived from sign values of the log likelihood ratio input of the received encoded data, and sign values used for data bits in the new decoder input values are derived from a log likelihood ratio value resulting from the decoder applying the decoding algorithm to the likelihood values for the received encoded data;
replace magnitudes of at least some of the likelihood values for the received encoded data with fixed values based on one of the one or more generated new decoder input values; and
reduce the fixed values each time a decoding run fails to converge.
2. The data processing system of claim 1 , wherein the decoder comprises a message-passing decoder.
3. The data processing system of claim 1 , wherein the decoder comprises a low density parity check decoder.
4. The data processing system of claim 1 , wherein the magnitudes of all of the likelihood values for the received encoded data are replaced with the fixed value.
5. The data processing system of claim 1 , wherein the decoder input initialization circuit is operable to scale magnitudes of at least some of the likelihood values for the received encoded data by a scaling factor to generate the new decoder input values.
6. The data processing system of claim 1 , wherein the decoder input initialization circuit is operable to generate the new decoder input values based in part on the decoded output from a previous decoding run.
7. A method for processing data, comprising:
receiving log likelihood ratio values for encoded data at a data input of a memory of a decoder of a storage device or a data transmission device;
applying, by the decoder, a decoding algorithm to the log likelihood ratio values;
when the decoding algorithm fails to converge, generating, by the decoder, new log likelihood ratio values based at least in part on the log likelihood ratio values for the encoded data, wherein signs used for parity bits in the new log likelihood ratio values are derived from signs of the log likelihood ratio values for the encoded data, and sign values used for data bits in the new log likelihood ratio values are derived from a log likelihood ratio value resulting from applying the decoding algorithm to the log likelihood ratio values for the encoded data, and wherein magnitudes of the new log likelihood ratio values comprise fixed values that are reduced with each successive decoding run; and
reapplying, by the decoder, the decoding algorithm to the new log likelihood ratio values.
8. The method of claim 7 , further comprising performing dampening when reapplying the decoding algorithm to the new log likelihood ratio values fails to converge, wherein the dampening comprises replacing the magnitudes of the new log likelihood ratio values based on a number of unsatisfied checks for each variable node in a previous decoding run.