Model based error avoidance

Patent metadata

Field	Value
Publication number	US-12517816-B2
Application number	US-202318383712-A
Country	US
Kind code	B2
Filing date	Oct 25, 2023
Priority date	Nov 3, 2022
Publication date	Jan 6, 2026
Grant date	Jan 6, 2026

Abstract

Aspects of the present disclosure configure a system component, such as a memory sub-system controller, to perform adaptive read level threshold voltage operations. The controller determines a first read level offset associated with reading a first set of data from a first level using a first read level of a plurality of read levels. The controller applies the first read level offset to a machine learning model to estimate a second read level offset, associated with reading a second set of data from a second level of the plurality of levels, using a second read level of the plurality of read levels. The controller updates, based on the first read level offset and the estimated second read level offset, a look-up table that includes a set of read level offsets used to read data from the plurality of levels of the individual component.

First claim

What is claimed is:

1 . A system comprising:
a memory sub-system comprising a set of memory components; and a processing device, operatively coupled to the set of memory components and programmed to perform operations comprising: maintaining a plurality of tables comprising different sets of read level offsets for a same set of read levels, each of the plurality of tables being associated with a different features of the set of memory components; determining a first read level offset, associated with reading a first set of data from a first level of a plurality of read levels of an individual component of the set of memory components, using a first read level of a plurality of read levels stored in a first table of the plurality of tables; determining an environmental factor associated with the individual component, the environmental factor comprising a temperature or program-erase count (PEC) of the individual component; applying the first read level offset and the environmental factor to a machine learning model to estimate a second read level offset, associated with reading a second set of data from a second level of the plurality of read levels, using a second read level of the plurality of read levels; updating, based on the first read level offset and the estimated second read level offset, the first table by replacing the set of read level offsets stored in the first table using the estimated second read level offset; applying an individual read level offset of a second table of the plurality of tables to the machine learning model to estimate an additional read level offset; and updating, based on the individual read level offset and the estimated additional read level offset, the second table by replacing the set of read level offsets stored in the second table using the estimated additional read level offset.

2 . The system of claim 1 , the operations comprising:
identifying a bin of the first table corresponding to the first read level offset; and replacing an individual read level offset value of the set of read level offsets associated with the identified bin and corresponding to the second read level with a value of the estimated second read level offset.

3 . The system of claim 1 , the operations comprising:
estimating, by the machine learning model, read level offsets for each of the plurality of read levels using the first read level offset.

4 . The system of claim 3 , the operations comprising:
identifying a bin of the first table corresponding to the first read level offset; and replacing each read level offset value of the set of read level offsets associated with the identified bin and corresponding to a respective one of the plurality of read levels with a respective value of the estimated read level offsets.

5 . The system of claim 1 , wherein the first read level offset is determined by:
performing a scan operation on the individual component to determine the first read level offset.

6 . The system of claim 5 , the operations comprising:
gradually applying different read levels to read the first set of data; and determining that the first read level offset has been reached in response to determining that an individual one of the different read levels reaches a center of valley (CoV).

7 . The system of claim 1 , the operations comprising:
determining a bin of the first table that is currently associated with the individual component; accessing the set of read level offsets of the determined bin stored in the first table; and reading one or more data sets from an individual level of the plurality of read levels of the individual component using a read level that is computed using an individual read level offset of the set of read level offsets corresponding to the individual level.

8 . The system of claim 1 , wherein the machine learning model is trained during manufacture of the memory sub-system.

9 . The system of claim 1 , wherein the machine learning model comprises at least one of an artificial neural network or a linear regression model.

10 . The system of claim 1 , wherein the machine learning model is trained based on training data to establish a relationship between a plurality of features comprising one or more first read level offsets and ground truth read level offsets of one or more other levels of the plurality of read levels.

11 . The system of claim 10 , wherein the plurality of features include at least one of: an operating temperature, a current bin of a plurality of bins, an elapsed programming time of data, or the PEC of the individual component.

12 . The system of claim 1 , wherein the first level corresponds to level 7 of the individual component, and wherein the second level corresponds to any one of levels 1-6.

13 . The system of claim 1 , wherein the machine learning model comprises a plurality of machine learning models, each of the plurality of machine learning models being trained to estimate a read level offset for a respective one of the plurality of read levels based on the first read level offset.

14 . The system of claim 1 , the operations comprising:
storing the first table of the plurality of tables, the first table associated with a first feature of the different features associated with the set of memory components, the first table comprising a first set of read level offsets of the plurality of read levels; and storing the second table of the plurality of tables, the second table associated with a second feature of the different features associated with the set of memory components, the second table comprising a second set of read level offsets of the plurality of read levels.

15 . The system of claim 14 , the operations comprising:
identifying a current feature of the individual component; selecting an individual table from the plurality of tables comprising the first table and the second table based on the identified current feature; and accessing data from the individual component using the selected individual table.

16 . The system of claim 14 , wherein the plurality of features include at least one of: an operating temperature, a current bin of a plurality of bins, an elapsed programming time of data, the PEC of the individual component.

17 . The system of claim 1 , wherein the memory sub-system comprises a three-dimensional NAND storage device, and wherein the individual component includes a superblock comprising a plurality of memory blocks across a plurality of memory dies.

18 . A method comprising:
maintaining a plurality of tables comprising different sets of read level offsets for a same set of read levels, each of the plurality of tables being associated with a different features of a set of memory components; determining a first read level offset associated with reading a first set of data from a first level of a plurality of read levels of an individual component of the set of memory components using a first read level of the plurality of read levels, the individual component configured to store data in the plurality of read levels each associated with a read level of the plurality of read levels stored in a first table of the plurality of tables; determining an environmental factor associated with the individual component, the environmental factor comprising a temperature or program-erase count (PEC) of the individual component; applying the first read level offset and the environmental factor to a machine learning model to estimate a second read level offset, associated with reading a second set of data from a

Assignees

• Micron Technology Inc

Inventors

• Chang Li-Te
• Kwong Charles S
• Wang Wei
• Lang Murong
• Zhou Zhenming

Classifications

• G11C29/021

  in voltage or current generators · CPC title

• G06F12/023 Primary

  Free address space management · CPC title

• G11C29/028 Primary

  with adaption or trimming of parameters · CPC title