NAND flash memory with reconfigurable neighbor assisted LLR correction with downsampling and pipelining

1. A non-volatile data storage device, comprising:

memory cells, each memory cell being a 4-bit quad-level cell (QLC) associated with a word line and a bit line;

a memory controller coupled to the memory cells for controlling operations of the memory cells;

wherein the memory controller comprises:

an LLR (log likelihood ratio) generation block for generating current LLR values;

a down-sampling block for grouping 4-bit cell values of neighboring memory cells and a target memory cell to respective 3-bit indices based on effect of neighboring memory cells on the LLR of the target memory cell;

an LLR compensation LUT (look-up table) for providing LLR compensation values;

an LLR correction block for receiving the current LLR values and the LLR compensation values to produce compensated LLR values;

wherein the memory controller is configured to:

perform a soft read operation of a target memory cell in response to a read command from a host;

determine a current LLR value based on result from the soft read operation using the LLR generation block;

use the down-sampling block to convert 4-bit cell values of neighboring memory cells and the target memory cell to respective 3-bit indices;

determine an LLR compensation value based on 3-bit indices using the LLR compensation LUT;

use LLR correction block to determine a compensated LLR value based on the current LLR value and the LLR compensation value;

perform soft decoding using the compensated LLR value;

correct the cell value based on the soft decoding; and

store corrected cell value in the memory controller.

2. The non-volatile data storage device of claim 1 , wherein the down-sampling block comprises:

a first look-up table (LUT) for converting a 4-bit cell value of a first neighboring memory cell to a first 3-bit index;

a second look-up table (LUT) for converting a 4-bit cell value of a second neighboring memory cell to a second 3-bit index; and

a third look-up table (LUT) for converting convert a 4-bit cell value of the target memory cell to a third 3-bit index.

3. The non-volatile data storage device of claim 2 , wherein the first, second, and third look-up tables are formed by grouping two or more 4-bit cell values of a neighboring cell that have similar effect on the target cell to a same 3-bit index.

4. The non-volatile data storage device of claim 3 , wherein the first, second, and third look-up tables are formed using offline characterization to determine effects of neighboring memory cells on the LLR values of the target memory cell.

5. The non-volatile data storage device of claim 3 , wherein the first, second, and third look-up tables are formed using on-line learning to determine effects of neighboring memory cells on the LLR values of the target memory cell.

6. The non-volatile data storage device of claim 5 , wherein the down-sampling block is reconfigurable by varying entries in the look-up tables in the down-sampling block in response to changing memory technologies, wherein new entries are determined based on effects of neighboring memory cells on the LLR values of the target memory cell by offline characterization or on-line training.

7. The non-volatile data storage device of claim 1 , wherein the LLR compensation values in the LLR compensation LUT are determined based on effect of neighboring memory cells on the LLR values of the target memory cell by offline characterization or on-line training.

8. The non-volatile data storage device of claim 7 , wherein the LLR compensation LUT is reconfigurable by varying entries in look-up tables in the LLR compensation LUT in response to changing memory technology, wherein new entries are determined based on effect of neighboring memory cells on the LLR values of the target memory cell by offline characterization or on-line training.

9. A non-volatile data storage device, comprising:

memory cells, wherein each memory cell is an m-bit multi-level cell (MLC) associated with a word line and a bit line, where m is an integer;

a memory controller coupled to the memory cells for controlling operations of the memory cells;

wherein the memory controller is configured to:

perform a soft read operation of a target memory cell in response to a read command from a host;

determine a current LLR (log likelihood ratio) value based on result from the soft read operation;

group m-bit cell values of neighboring memory cells and the target memory cell to respective n-bit indices based on effect of neighboring memory cells on the LLR of the target memory cell, where n is an integer and n<m;

determine an LLR compensation value based on the n-bit indices;

determine a compensated LLR value based on the current LLR value and the LLR compensation value; and

perform soft decoding using the compensated LLR value.

10. The non-volatile data storage device of claim 9 , wherein the memory controller is further configured to:

correct the cell value based on the soft decoding; and

store corrected cell value in the memory controller.

11. The non-volatile data storage device of claim 9 , wherein the memory controller comprises:

an LLR (log likelihood ratio) generation block for generating current LLR values;

a down-sampling block for converting m-bit cell values of neighboring memory cells and the target memory cell to respective n-bit indices;

an LLR compensation block for providing LLR compensation values; and

an LLR correction block for receiving the current LLR values and the LLR compensation values to produce compensated LLR values.

12. The non-volatile data storage device of claim 11 , wherein the down-sampling block comprises a look-up table (LUT) mapping m-bit cell values to n-bit indices for each of the neighboring memory cells and the target memory cell.

13. The non-volatile data storage device of claim 12 , wherein the look-up tables in the down-sampling block is formed by offline characterization or on-line training to determine effect of neighboring memory cells on the LLR of the target memory cell.

14. The non-volatile data storage device of claim 11 , wherein the LLR compensation block comprises a look-up table (LUT) for associating LLR compensation values with the n-bit indices.

15. The non-volatile data storage device of claim 14 , wherein the look-up table in the LLR compensation block is formed by offline characterization or on-line training to determine effect of neighboring memory cells on the LLR of the target memory cell.

16. A method of operating a storage system, the storage system including memory cells and a memory controller coupled to the memory cells for controlling operations of the memory cells, wherein each memory cell is an m-bit multi-level cell (MLC) associated with a word line and a bit line, where m is an integer, the method comprising:

performing a soft read operation of a target memory cell in response to a read command from a host;

determining a current LLR (log likelihood ratio) value based on result from the soft read operation;

grouping m-bit cell values of neighboring memory cells and the target memory cell to respective n-bit indices, based on effect of neighboring memory cells on the LLR of the target memory cell, wherein n is an integer and n<m;

determining an LLR compensation value based on the n-bit indices;

determining a compensated LLR value based on the current LLR value and the LLR compensation value; and

performing soft decoding using the compensated LLR value.

17. The method of claim 16 , wherein the method further comprises:

correcting the cell value based on the soft decoding; and

storing corrected cell value in the memory controller.

18. The method of claim 17 , further comprising forming look-up tables (LUTs) for grouping m-bit cell values of neighboring memory cells and the target memory cell to respective n-bit indices, based on effect of neighboring memory cells on the LLR of the target memory cell.

19. The method of claim 18 , further comprising grouping m-bit cell values to respective n-bit indices by offline characterization or on-line training to determine effect of neighboring memory cells on the LLR of the target memory cell.

20. The method of claim 16 , further comprising forming a look-up table (LUT) for associating LLR compensation values with the n-bit indices.