CN-122019249-A - Flash memory decoding method, flash memory decoding device and computer storage medium

CN122019249ACN 122019249 ACN122019249 ACN 122019249ACN-122019249-A

Abstract

The application provides a flash memory decoding method, a flash memory decoding device and a computer storage medium. The flash memory decoding method comprises the steps of responding to hard decision failure, obtaining the current check state of check nodes, wherein the check nodes are connected with a plurality of variable nodes, obtaining the variable state of the variable nodes at the previous stage, traversing and updating each variable node in the variable nodes, obtaining an updating factor of the current stage, determining a target variable node and remaining variable nodes in the variable nodes, obtaining the remaining log likelihood ratio of the remaining variable nodes sent to the check nodes, and calculating the target remaining log likelihood ratio of the check nodes sent to the target variable nodes according to the remaining log likelihood ratio and the updating factor of the current stage. By the flash memory decoding method, the updating factors of iteration change are introduced, so that the error correction performance of soft decoding is improved, and soft decoding convergence is accelerated.

Inventors

DONG GUIQIANG
WANG TAO
ZHANG YIFAN

Assignees

慧忆微(上海)科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. A method of flash memory decoding, the method comprising: Responding to the failure of the hard decision, and acquiring the current check state of a check node, wherein the check node is connected with a plurality of variable nodes; Acquiring variable states of the variable nodes at the last stage; Traversing updating each variable node of the plurality of variable nodes: acquiring an update factor of the current stage, wherein the update factor is subjected to iterative change according to the update stage; Determining target variable nodes and residual variable nodes in the variable nodes; Acquiring a residual log likelihood ratio value of the residual variable node and the check node; And calculating a target residual log likelihood ratio sent to the target variable node by the check node according to the residual log likelihood ratio and the update factor of the current stage.
2. The method of claim 1, wherein, The step of obtaining the update factor of the current stage comprises the following steps: and calculating the update factor of the current stage according to the log likelihood ratio value of the check node sent by the variable node of the previous stage.
3. The method for decoding a flash memory according to claim 2, wherein, The step of calculating the update factor of the current stage according to the log likelihood ratio value sent by the variable node of the previous stage to the check node comprises the following steps: acquiring a log likelihood ratio transmitted by a variable node received by each check node; And calculating the update factor of each check node in the current stage according to the log likelihood ratio value received by each check node.
4. The method for decoding a flash memory according to claim 3, wherein, And calculating an update factor of each check node in the current stage according to the log likelihood ratio received by each check node, wherein the update factor comprises the following steps: Traversing a plurality of variable nodes connected with each check node to obtain the log likelihood ratio of other variable nodes corresponding to each variable node; And calculating the update factors of the check nodes in the current stage for each variable node according to the log likelihood ratio values of the other variable nodes.
5. The method of claim 4, wherein, And calculating the update factor of the check node in the current stage for each variable node according to the log likelihood ratio values of the other variable nodes, wherein the update factor comprises the following steps: And determining the update factors of the check nodes in the current stage for each variable node according to the average value of the log likelihood ratio values of the other variable nodes.
6. The method of claim 1, wherein, The step of obtaining the update factor of the current stage comprises the following steps: Acquiring the iterated times of the current stage; and determining the update factor of the current stage according to the iterated times.
7. The method of claim 6, wherein, The step of determining the update factor of the current stage according to the iterated times comprises the following steps: calculating an iteration factor according to the iterated times and preset updating parameters; determining an update factor of the current stage by comparing the iteration factor with a larger value in a preset upper limit value; or determining the update factor of the current stage by comparing the iteration factor with a larger value in a preset lower limit value.
8. The method of claim 1, wherein, The step of obtaining the update factor of the current stage comprises the following steps: acquiring a first updating factor of the current stage according to a first soft decoding algorithm; acquiring a second updating factor of the current stage according to a second soft decoding algorithm; determining an update factor of the current stage according to the first update factor and the second update factor; wherein the first soft coding algorithm and the second soft coding algorithm are different, at least one of the first update factor and the second update factor iteratively changing according to an update phase; And calculating a target residual log likelihood ratio sent to the target variable node by the check node according to the residual log likelihood ratio and the update factor of the current stage, wherein the method comprises the following steps: determining an update function according to the first soft decoding algorithm and the second soft decoding algorithm; Substituting the update factor of the current stage and the residual log likelihood ratio into the update function, and calculating a target residual log likelihood ratio sent to the target variable node by the check node.
9. A flash memory decoding device, wherein the flash memory decoding device comprises a memory and a processor coupled with the memory; Wherein the memory is configured to store program data and the processor is configured to execute the program data to implement the flash decoding method according to any one of claims 1 to 8.
10. A computer storage medium for storing program data which, when executed by a computer, is adapted to carry out a flash memory decoding method according to any one of claims 1 to 8.

Description

Flash memory decoding method, flash memory decoding device and computer storage medium Technical Field The present application relates to the field of communications technologies, and in particular, to a flash memory decoding method, a flash memory decoding device, and a computer storage medium. Background As flash memory density increases, raw bit error rate (rber, raw bit error rate) gets higher and higher. LDPC (Low-DENSITY PARITY-check) codes, which provide powerful soft decoding error correction capability, have been widely used in flash memory systems. In order to maintain data reliability, various measures are taken in flash memory systems to perform read recovery. In conventional practice, when decoding failure occurs, a retry process is performed first, and read voltage offset is adjusted according to a retry table to re-read, i.e., hard read, and then hard decoding, i.e., hard decoding, is performed. When the retry still cannot correct errors, soft information is acquired through a soft reading process, and then the soft information is mapped into LLR (Log-Likelihood Ratio) values, and the LLR values are input to a soft decoder, namely soft decoding. The error correction performance of the LDPC code determines the reliability of NAND data, the error correction performance of the LDPC soft decoding determines the upper limit of the data reliability, and the soft decoding algorithm corrects the error code word by repeatedly and iteratively updating the LLR value of the code word bit between the variable node and the check node. The soft decoding algorithm is classified into BP (Belief Propagation ), NMS (Normalized Min-Sum) and OMS (Offset Min-Sum), MS (Min-Sum) and the like according to functions used by CN (check nodes, c-nodes) update. In practice, LDPC soft decoder generally needs to provide higher throughput, the calculation amount of the decoding process of BP decoding algorithm is larger, and the error correction performance of MS is poorer, so NMS and OMS algorithms are often used in a storage system to correct error code words, in the existing soft decoder, NMS algorithm only uses single normalization factor alpha to optimize performance in the decoding process, OMS algorithm also only uses one offset factor beta to optimize performance, and the improvement of the error correction performance of soft decoding is limited. ‌ A Disclosure of Invention In order to solve the above technical problems, the present application provides a flash memory decoding method, a flash memory decoding device and a computer storage medium. In order to solve the above technical problems, the present application provides a flash memory decoding method, which includes: Responding to the failure of the hard decision, and acquiring the current check state of a check node, wherein the check node is connected with a plurality of variable nodes; Acquiring variable states of the variable nodes at the last stage; Traversing updating each variable node of the plurality of variable nodes: acquiring an update factor of the current stage, wherein the update factor is subjected to iterative change according to the update stage; Determining target variable nodes and residual variable nodes in the variable nodes; Acquiring a residual log likelihood ratio value of the residual variable node and the check node; And calculating a target residual log likelihood ratio sent to the target variable node by the check node according to the residual log likelihood ratio and the update factor of the current stage. Wherein, the obtaining the update factor of the current stage includes: and calculating the update factor of the current stage according to the log likelihood ratio value of the check node sent by the variable node of the previous stage. Wherein, the calculating the update factor of the current stage according to the log likelihood ratio value sent by the variable node of the previous stage to the check node comprises: acquiring a log likelihood ratio transmitted by a variable node received by each check node; And calculating the update factor of each check node in the current stage according to the log likelihood ratio value received by each check node. The calculating the update factor of each check node in the current stage according to the log likelihood ratio received by each check node comprises the following steps: Traversing a plurality of variable nodes connected with each check node to obtain the log likelihood ratio of other variable nodes corresponding to each variable node; And calculating the update factors of the check nodes in the current stage for each variable node according to the log likelihood ratio values of the other variable nodes. The calculating the update factor of the check node for each variable node in the current stage according to the log likelihood ratio of the other variable nodes comprises the following steps: And determining the update factors of the check nodes in the current stage for each v