Adaptive desaturation in min-sum decoding of LDPC codes
A system implements adaptive desaturation for the min-sum decoding of LDPC codes. Specifically, when an-above threshold proportion of messages from check nodes to variable nodes (CN-to-VN messages) are saturated to a maximum fixed-precision value, all CN-to-VN messages are halved. This facilitates the saturation of correct messages and boosts error correction over small trapping sets. The adaptive desaturation approach reduces the error floor by orders of magnitudes with negligible add-on circuits.
1. A method comprising:
receiving, by an electronic device, input data; and
performing, by the electronic device, layered min-sum decoding of the input data using a Tanner graph to generate decoded output data, the layered min-sum decoding including passing messages from variable nodes (VNs) to check nodes (CNs) in VN-to-CN messages and from CNs to VNs in CN-to-VN messages, each VN and CN having a value associated therewith, the values of the VNs being updated in response to CN-to-VN messages and the values of the CNs being updated in response to VN-to-CN messages; and
during performing layered min-sum decoding, in response to detecting that a proportion of the values of the CNs that exceed a maximum value meets a threshold condition, halving all of the VN-to-CN messages and all first a-posterior probability (APP) messages of a current iteration of the layered min-sum decoding.
2. The method of claim 1 , wherein performing layered min-sum decoding of the input data to generate decoded output data comprises performing algorithm 3 of Table 3.
3. The method of claim 1 , wherein performing layered min-sum decoding of the input data to generate decoded output data comprises performing algorithm 4 of Table 4.
4. The method of claim 1 , wherein performing layered min-sum decoding of the input data to generate decoded output data comprises iteratively performing a plurality of iterations, each iteration including passing of VN-to-CN messages followed by CN-to-VN messages, wherein halving at least one of all of the CN-to-VN messages and all of the VN-to-CN messages is performed between consecutive iterations of the plurality of iterations.
5. The method of claim 1 , wherein performing layered min-sum decoding comprises
halving all of the CN-to-VN messages and all of the VN-to-CN messages.
6. The method of claim 1 , wherein the electronic device is a dedicated hardware device.
7. The method of claim 1 , wherein the Tanner graph implements a quasi-cyclic low density parity code matrix.
8. The method of claim 1 , wherein the maximum value is a maximum fixed point value representable by the CNs.
9. An electronic device including circuits configured to:
receive input data; and
perform layered min-sum decoding of the input data using a Tanner graph to generate decoded output data, the min-sum layered decoding including passing messages from variable nodes (VNs) to check nodes (CNs) in VN-to-CN messages and from CNs to VNs in CN-to-VN messages, each VN and CN having a value associated therewith, the values of the VNs being updated in response to CN-to-VN messages and the values of the CNs being updated in response to VN-to-CN messages; and
during performing layered min-sum decoding, if a number of the values of the CNs that exceed a maximum value meets a threshold condition, halving all of the VN-to-CN messages.
10. The electronic device of claim 9 , further comprising circuits configured to perform layered min-sum decoding of the input data to generate decoded output data according to algorithm 3 of Table 3.
11. The electronic device of claim 9 , further comprising circuits configured to perform layered min-sum decoding of the input data to generate decoded output data according to algorithm 4 of Table 4.
12. The electronic device of claim 9 , further comprising circuits configured to perform layered min-sum decoding by iteratively performing a plurality of iterations, each iteration including passing of VN-to-CN messages followed by CN-to-VN messages, wherein halving at least one of all of the CN-to-VN messages and all of the VN-to-CN messages is performed between consecutive iterations of the plurality of iterations if the number of the values of the CNs that exceed the maximum value meets the threshold condition.
13. The electronic device of claim 9 , further comprising circuits configured to perform layered min-sum decoding, wherein halving at least one of all of the CN-to-VN messages and all of the VN-to-CN messages comprises halving all of the CN-to-VN messages and all of the VN-to-CN messages.
14. The electronic device of claim 9 , further comprising circuits configured to, if the number of the values of the CNs that exceed the maximum value meets the threshold condition, halving all first a-posterior probability (APP) messages of a current iteration of the layered min-sum decoding.
15. The electronic device of claim 9 , wherein the electronic device comprises a dedicated hardware device including specialized circuits implementing the layered min-sum decoding of the input data.
16. The electronic device of claim 9 , wherein the Tanner graph implements a quasi-cyclic low density parity code matrix.
17. The electronic device of claim 9 , further comprising one or more storage registers;
wherein the maximum value is a maximum fixed-point value representable by the one or more registers.