Bad column handling in flash memory

1. A method of operating a memory array that includes at least one bad column, comprising:

determining location information for the at least one bad column;

compressing the location information;

providing the compressed location information to a soft-input ECC decoder as a first soft data indicating low likelihood for hard data obtained from the at least one bad column; and

decoding hard data in combination with the soft data indicating low likelihood for the hard data obtained from the at least one bad column.

2. The method of claim 1 wherein the providing step comprises:

providing the compressed location information to a soft-input ECC decoder as soft data indicating low likelihood for hard data obtained from any column which shares the compressed location information of the at least one bad column.

3. The method of claim 1 further comprising performing a high resolution read to obtain a second soft read data, the second soft read data combined with the first soft data.

4. The method of claim 1 wherein compressing the location information includes dividing all columns into a plurality of sections and, in each section, treating a particular column as a bad column if the location information indicates that there is a bad column in any section of the plurality of sections at a location corresponding to the particular column.

5. The method of claim 1 wherein the memory array consists of a plurality of columns, each of the plurality of columns containing a plurality of bit lines, and wherein a column is considered as a bad column if it contains at least one defective bit line.

6. The method of claim 5 further comprising determining location information for at least one defective bit line in the at least one bad column, and wherein the compressed location information includes compressed bit line location information.

7. The method of claim 1 wherein the at least one bad column consists of at least one low-ranked bad column, and further comprising:

identifying at least one high-ranked bad column; and

replacing the at least one high-ranked bad column with at least one redundant column.

8. The method of claim 7 further comprising identifying one or more redundant columns that are not used for replacement of high-ranked bad columns and using the one or more redundant columns to provide additional redundancy for stored data.

9. The method of claim 7 further comprising identifying low-ranked bad columns and high-ranked bad columns by ranking according to the number of bad bit lines per column.

10. The method of claim 1 wherein the memory array is a NAND flash memory array.

11. A method of operating a memory array comprising:

identifying a number of columns of the memory array as bad columns;

ranking the identified bad columns;

replacing high-ranked bad columns with redundant columns;

recording low-ranked bad columns in a record;

subsequently, storing data in the memory array including in the low-ranked bad columns;

subsequently, reading the data from the memory array including from the low-ranked bad columns;

identifying data from the low-ranked bad columns from the record; and

performing soft-input Error Correction Code (ECC) decoding of the data, where data from the low-ranked bad columns is treated as having low likelihood in response to identifying the data as coming from the low-ranked bad columns.

12. The method of claim 11 further comprising identifying one or more redundant columns that are not used for replacement of high-ranked bad columns and using the one or more redundant columns to provide additional redundancy for stored data.

13. The method of claim 11 wherein the identified bad columns are ranked according to the number of bad bit lines per column.

14. The method of claim 13 wherein high-ranked bad columns have a number of bad bit lines that is greater than a predetermined number and low-ranked bad columns have a number of bad bit lines that is less than or equal to the predetermined number.

15. The method of claim 11 wherein the identified bad columns are ranked according to ECC results from data stored in the bad columns.