Memory system with adaptive information propagation and method of operating such memory
Systems, memory controllers, decoders and methods perform decoding by exploiting differences among word lines for which soft decoding fails (failed word lines). Such decoding generates extrinsic information for codewords of failed word lines based on the soft decoding. The soft information obtained during the soft decoding is updated based on the extrinsic information, and the updated soft information is propagated across failed word lines. Low-density parity-check (LDPC) decoding of codewords of failed word lines is performed with the updated soft information.
1. A memory system, comprising:
a memory device including a super block; and
a memory controller configured to:
perform soft decoding of all word lines in the super block to generate respective codewords and generate soft information as a result of the soft decoding, the super block including multiple word lines and a last word line, which stores a result of an XOR operation on codewords of the multiple word lines;
generate extrinsic information for codewords of failed word lines by scaling respective soft information for each of the failed word lines using a scaling factor specific for each of the failed word lines, wherein the scaling factor is determined based on iteration information and unsatisfied check information for each failed word line;
propagate the extrinsic information across a Tanner graph associated with the XOR operation, wherein the extrinsic information is propagated to check nodes from variable nodes corresponding to the failed word lines and the last word line on the Tanner graph;
update check information based on the propagated extrinsic information to generate updated check information by updating the check nodes based on the propagated extrinsic information; and
perform low-density parity-check (LDPC) decoding of the codewords of the failed word lines with the extrinsic information and the updated check information.
2. The memory system of claim 1 , wherein the memory controller generates the scaled extrinsic information for each word line for which soft decoding failed at each iteration.
3. The memory system of claim 2 , wherein the scaling factor is determined as a function of a number of iterations performed with respect to the corresponding word line for which soft decoding failed and the number of unsatisfied checks for the corresponding word line for which soft decoding failed a current iteration.
4. The memory system of claim 3 , wherein the function has an inverse relationship with the number of unsatisfied checks for the corresponding word line at the current iteration.
5. The memory system of claim 1 , wherein the scaling factor for each word line for which soft decoding failed is determined in real time.
6. The memory system of claim 1 , wherein the scaling factor for each word line for soft decoding failed is determined heuristically based on simulation results.
7. A memory controller comprising:
a low-density parity-check (LDPC) decoder configured to:
perform soft decoding of all word lines in a super block to generate respective codewords and generate soft information as a result of the soft decoding, the super block including multiple word lines and a last word line, which stores a result of an XOR operation on codewords of the multiple word lines;
generate extrinsic information for codewords of failed word lines by scaling respective soft information for each the failed word lines using a scaling factor specific for each of the failed word lines, wherein the scaling factor is determined based on iteration information and unsatisfied check information for each failed word line;
propagate the extrinsic information across a Tanner graph associated with the XOR operation, wherein the extrinsic information is propagated to check nodes from variable nodes corresponding to the failed word lines and the last word line on the Tanner graph;
update check information based on the propagated extrinsic information to generate updated check information by updating the check nodes based on the propagated extrinsic information; and
perform low-density parity-check (LDPC) decoding of the codewords of the failed word lines with the extrinsic information and the updated check information.
8. The memory controller of claim 7 , wherein the LDPC decoder generates the scaled extrinsic information for each word line for which soft decoding failed at each iteration.
9. The memory controller of claim 8 , wherein the scaling factor is determined as a function of a number of iterations performed with respect to the corresponding word line for which soft decoding failed and the number of unsatisfied checks for the corresponding word line for which soft decoding failed at a current iteration.
10. A method of decoding, comprising:
performing soft decoding of all word lines in a super block of a memory device to generate respective codewords and generating soft information as a result of the soft decoding, the super block including multiple word lines and a last word line, which stores a result of an XOR operation on codewords of the multiple word lines;
generating extrinsic information for codewords of failed word lines by scaling respective soft information for each the failed word lines using a scaling factor specific for each of the failed word lines, wherein the scaling factor is determined based on iteration information and unsatisfied check information for each failed word line;
propagate the extrinsic information across a Tanner graph associated with the XOR operation, wherein the extrinsic information is propagated to check nodes from variable nodes corresponding to the failed word lines and the last word line on the Tanner graph;
updating check information based on the propagated extrinsic information to generate updated check information by updating the check nodes based on the propagated extrinsic information; and
performing low-density parity-check (LDPC) decoding of the codewords of the failed word lines with the extrinsic information and the updated check information.
11. The method of claim 10 , wherein the scaled extrinsic information is generated for each word line for which soft decoding failed at each iteration.
12. The method of claim 11 , wherein the scaling factor is determined as a function of a number of iterations performed with respect to the corresponding word line for which soft decoding failed and the number of unsatisfied checks for the corresponding word line for which soft decoding failed at a current iteration.
13. The method of claim 12 , wherein the function has an inverse relationship with the number of unsatisfied checks for the corresponding word line at the current iteration.
14. The method of claim 10 , wherein the scaling factor for each word line for which soft decoding failed is determined in real time or heuristically based on simulation results.