A method for improving the encoding rate of flash memory and flash memory

Abstract

The invention discloses a method for improving the encoding rate of flash memory and flash memory, which is characterized in that flash memory of the same batch or similar performance is extracted for durability test, and the data distribution data of BER of each page of the flash memory is obtained statistically, and the page is divided according to BER It is divided into K categories, K>2, and the maximum BER value of each category is used as the typical BER value of each category, and the recommended value of the page verification data size of each category is obtained according to the typical BER value. By using the machine learning algorithm to classify the BER values ​​of all pages, and formulating a BER value for each type of page, it is expected to reduce the BER value of the flash memory as a whole, that is, to increase the coding rate of the flash memory as a whole. At the same time, the K-means clustering algorithm can flexibly select the K (category number) value. The firmware of the solid-state hard disk only needs to record the typical BER value of K categories or the size of the verification data, and hardly occupies RAM space.

Description

technical field [0001] The invention relates to solid-state hard disk control technology, in particular to a method for improving the coding rate of flash memory and flash memory. Background technique [0002] A typical NAND is composed of multiple blocks, and each Block is composed of multiple pages. Die is the unit of independent concurrent operation supported by the NANDFLASH; the block can be erased independently, and the data of each physical location in it must be erased before the next write before writing a new one. Data; page Page is the unit of reading and writing, that is, writing or reading data to FLASH must be performed with the page as the unit of operation. [0003] The data on the flash memory (NAND) may be wrong, that is, the read data is inconsistent with the written data, so we need to generate some check data in addition to the user data, and use it to correct the error when the data is wrong; a Coding rate of a page=user data size / (user data size+chec...

AI Technical Summary

Problems solved by technology

[0005] Before the flash memory leaves the factory, the manufacturer conducts an endurance test for the batch of flash memory, measures the BER value of each page and the size of the verification data, selects the largest BER value as the typical BER value of the batch, and sets the corresponding verification data size Tell the customer that it is recommended that the customer generate such a large verification data at the same time when writing a page of data, figure 1 It is an example of the size of common flash memory data and valid data. It can be seen from the figure that the BER and parity data size of each page is actually different. Reasonable, but a DIE of flash memory contains several million pages. If the information of the checksum data size of all pages is recorded, a large additional RAM space is required.

Therefore, the manufacturer directly provides the maximum BER value and the corresponding check data size, and all pages use this typical value. The disadvantage of this is that the overall BER and check data of this batch of flash memory are enlarged, and the encoding rate becomes smaller. up

Images