System and method for dynamic scaling of LDPC decoder in a solid state drive
In some embodiments of the present invention, a data storage device includes a controller and a memory. The data storage device further includes an LDPC encoder and decoder, with the decoder implementing a dynamic precision-rescaling technique for improving performance. In one embodiment, the technique works by rescaling the binary representations of the input log-likelihood ratios (LLRs) and messages upon activation of decoder-state-based triggers. Various triggering functions are introduced, e.g., checking if the number of output LLRs smaller than a certain limit crosses a threshold, checking if the weight of a syndrome crosses a threshold, etc. This technique offers an improvement in the performance of the decoder.
1. A decoder, the decoder configured to:
exchange, during each of a plurality of iterations, a plurality of messages between a plurality of variable nodes and a plurality of check nodes, each message indicating a degree of reliability in an observed outcome of data;
determine a current operating condition of the decoder;
for each of the iterations:
scale the respective message when the current operating condition of the decoder satisfies a predetermined condition, and
pass the message without scaling when the current operating condition does not satisfy the predetermined condition; and
determine a valid codeword based on the exchanging.
2. The decoder of claim 1 , wherein the decoder being configured to scale the respective message comprises the decoder being configured to reduce a number of bits representing the message and the degree of reliability indicated by the message.
3. The decoder of claim 1 , wherein the plurality of messages comprise a plurality of log-likelihood ratios (LLRs) for a unit of data.
4. The decoder of claim 3 , wherein scaling the respective message comprises reducing a numerical range for a respective one of the LLRs.
5. The decoder of claim 4 , wherein scaling the respective message comprises right shifting a plurality of bits indicative of the respective one of the LLRs and truncating or rounding the respective one of the LLRs.
6. The decoder of claim 3 , wherein the decoder is further configured to:
perform, during each iteration, a mathematical operation on each of the LLRs; and
determine whether a respective one of the LLRs should be scaled at the end of the iteration,
wherein, when the respective one of the LLRs is scaled, a next mathematical operation is performed on the scaled LLR.
7. The decoder of claim 3 , wherein the decoder being configured to scale the respective message when the current operating condition of the decoder satisfies a predetermined condition comprises the decoder being configured to determine that absolute values of a predetermined number of the plurality of LLRs are greater than a predetermined limit.
8. The decoder of claim 3 , wherein the decoder being configured to scale the respective message when the current operating condition of the decoder satisfies a predetermined condition comprises the decoder being configured to determine that a predetermined number of the plurality of iterations has been reached.
9. The decoder of claim 8 , wherein the predetermined number of the plurality of iterations is a predetermined number of the plurality of iterations since a prior scaling operation.
10. The decoder of claim 3 , wherein the decoder being configured to scale the respective message when the current operating condition of the decoder satisfies a predetermined condition comprises the decoder being configured to determine that a current number of unsatisfied checks on at least a portion of the plurality of messages exchanged between the plurality of variable nodes and the plurality of check nodes does not satisfy a predetermined threshold and a predetermined number of iterations have been performed since a prior scaling operation.
11. A machine-implemented method comprising:
exchanging, in a decoder during each of a plurality of iterations, a plurality of messages between a plurality of variable nodes and a plurality of check nodes, each message indicating a degree of reliability in an observed outcome of data;
determining a current operating condition of the decoder;
for each of a plurality of iterations:
scaling the respective message when the current operating condition of the decoder satisfies a predetermined condition, and
passing the message without scaling when the current operating condition does not satisfy the predetermined condition; and
determining a valid codeword based on the exchanging.
12. The machine-implemented method of claim 11 , wherein scaling the respective message comprises reducing a number of bits representing the message and the degree of reliability indicated by the message.
13. The machine-implemented method of claim 11 , wherein the plurality of messages comprise a plurality of log-likelihood ratios (LLRs) for a unit of data.
14. The machine-implemented method of claim 13 , wherein scaling the respective message comprises reducing a numerical range for a respective one of the LLRs.
15. The machine-implemented method of claim 14 , wherein scaling the respective message comprises right shifting a plurality of bits indicative of the respective one of the LLRs and truncating or rounding the respective one of the LLRs.
16. The machine-implemented method of claim 13 , further comprising:
performing, during each iteration, a mathematical operation on each of the LLRs; and
determining whether a respective one of the LLRs should be scaled at the end of the iteration,
wherein, when the respective one of the LLRs is scaled, a next mathematical operation is performed on the scaled LLR.
17. The machine-implemented method of claim 13 , wherein determining whether the respective one of the messages should be scaled based on a current operating condition of the decoder comprises:
determining that absolute values of a predetermined number of the plurality of LLRs are greater than a predetermined limit.
18. The machine-implemented method of claim 13 , wherein determining whether the respective one of the messages should be scaled based on a current operating condition of the decoder comprises:
determining that a predetermined number of the plurality of iterations has been reached.
19. The machine-implemented method of claim 18 , wherein the predetermined number of the plurality of iterations is a predetermined number of the plurality of iterations since a prior scaling operation.
20. The machine-implemented method of claim 13 , wherein determining whether the respective one of the messages should be scaled based on a current operating condition of the decoder comprises:
determining that a current number of unsatisfied checks on at least a portion of the plurality of messages exchanged between the plurality of variable nodes and the plurality of check nodes does not satisfy a predetermined threshold and a predetermined number of iterations have been performed since a prior scaling operation.