Granted Patent B2

US 10,607,710 · App. 16/414,714 · Granted Mar 31, 2020

Abstract

Claims (21)

1. A method of detecting data drift in an array of non-volatile memory cells in an analog neuromorphic memory system, the method comprising:

selecting a sample of non-volatile memory cells within the array; and

detecting data drift in the selected cells for data drift, the detecting comprising:

programming each of the selected cells at one of N levels;

determining a difference between each pair of adjacent levels; and

comparing each difference against a target value.

2. The method of claim 1 , further comprising: when any difference exceeds the target value, performing a data restore.

3. The method of claim 2 , further comprising:

when any current of a restored cell does not meet a target, identifying the restored cell as a bad cell.

4. The method of claim 3 , further comprising:

replacing the bad cell with a redundant cell.

5. The method of claim 2 , wherein the data restore comprises cell tuning.

6. The method of claim 5 , wherein the cell tuning comprises iterative programming and verifying.

7. The method of claim 2 , wherein the data restore comprises cell tuning with a target amount corresponding to one of the N levels.

8. The method of claim 7 , wherein the target amount is stored in a look-up table.

9. The method of claim 2 , wherein the data restore comprises re-transferring training weights to the memory cells.

10. The method of claim 2 , wherein the memory cells are split gate memory cells.

11. The method of claim 2 , wherein the memory cells are stacked gate memory cells.

12. The method of claim 1 , further comprising: when any difference exceeds the target value, setting a data restore flag.

13. The method of claim 1 , wherein the memory cells are split gate memory cells.

14. The method of claim 1 , wherein the memory cells are stacked gate memory cells.

Assignments (12)