# US 2022/0068430 A1 · Apparatuses and methods for repairing defective memory cells based on a specified error rate for certain memory cells

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0068430 A1
- Document ID: US20220068430A1
- Status: Published application
- Kind: A1
- Application number: 17/454,443
- Filed: 2021-11-10
- Published: 2022-03-03
- Earliest claimed: 2019-06-26
- Assignees: Micron Technology, Inc.
- Applicants: Micron Technology, Inc.
- Inventors: DAVID HULTON; TAMARA SCHMITZ; JONATHAN D. HARMS; JEREMY CHRITZ; KEVIN MAJERUS
- Web page: https://patentgrep.com/patent/US20220068430A1

## CPC

- G06F 3/0616: {in relation to life time, e.g. increasing Mean Time Between Failures [MTBF]}
- G11C 29/76: {using address translation or modifications}
- G06F 3/0619: {in relation to data integrity, e.g. data losses, bit errors}
- G06F 3/064: {Management of blocks}
- G06F 3/0659: {Command handling arrangements, e.g. command buffers, queues, command scheduling}
- G06F 3/0679: {Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP]}
- G06F 12/0246: {in block erasable memory, e.g. flash memory}
- G06F 12/126: with special data handling, e.g. priority of data or instructions, handling errors or pinning
- G11C 29/883: {using a single defective memory device with reduced capacity, e.g. half capacity}
- G06F 11/1048: {using arrangements adapted for a specific error detection or correction feature}
- G06F 12/0813: with a network or matrix configuration
- G11C 7/10: Input/output [I/O] data interface arrangements, e.g. I/O data control circuits, I/O data buffers
- G11C 11/408: Address circuits
- G11C 11/418: Address circuits
- G11C 29/04: Detection or location of defective memory elements {, e.g. cell constructio details, timing of test signals}
- G11C 29/44: Indication or identification of errors, e.g. for repair
- G11C 29/4401: {for self repair}
- G11C 29/787: {using a fuse hierarchy}
- G11C 2029/4402: {Internal storage of test result, quality data, chip identification, repair information}
- H04L 61/2575: using address mapping retrieval, e.g. simple traversal of user datagram protocol through session traversal utilities for NAT [STUN]
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G11C: STATIC STORES (semiconductor memory devices H10B)
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Methods, apparatuses, and systems for repairing defective memory cells in regions of a memory array associated with high or low priority levels are disclosed. A repair address generator may be configured to generate a memory address map for repair (e.g., blowing fuses at a fuse circuit), depending on whether certain applications may operate at a high priority level indicative of a low bit error rate or a low priority level indicative of a higher bit error rate. For example, a specified error rate associated with a low priority level may correspond to a threshold error rate for certain applications, such as a neural network application that stores trained weights. Such neural network applications may access trained weights being partially stored in defective memory cells, with the least significant bits of such trained weights being stored in defective memory cells that are not repaired according to the memory address map.

## Claims

### Claim 1

A method comprising:
obtaining a memory access pattern associated with a plurality of memory access operations, the memory access pattern indicative of first regions of a memory that are accessed for data more frequently than second regions of the memory associated with a specified error rate;
generating a memory address map for the plurality of memory access operations based on the first regions of the memory and the second regions of the memory identified associated with the specified error rate; and
providing the memory address map to a memory data register of the memory, the memory data register accessible by the memory for repairing at least one region of the second regions of the memory.

### Claim 2 (depends on 1)

The method of claim 1, wherein the second regions are configured to store weights of a neural network.

### Claim 3 (depends on 1)

The method of claim 1, wherein the specified error rate comprises a specified error rate for storing weights of a neural network.

### Claim 4 (depends on 1)

The method of claim 1, wherein the memory data register corresponds to a register or local cache of the memory.

### Claim 5 (depends on 1)

The method of claim 1, wherein the specified error rate corresponds to a threshold error rate for storing weights of a neural network.

### Claim 6 (depends on 5)

The method of claim 5, wherein the threshold error rate for storing the weights of the neural network is based on a bit flip tolerance of the neural network.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
identifying a first portion of memory addresses associated with the first regions of the memory being accessed more frequently than the second regions of the memory, as having a high priority; and
identifying a second portion of memory addresses associated with the second regions of the memory, as having a lower priority than the high priority.

### Claim 8 (depends on 7)

The method of claim 7, further comprising:
identifying, in the first portion of memory addresses, most significant bits of each memory address of the first portion of the memory addresses; and
aligning each memory address of the first portion of memory addresses with a memory identifier of the memory address.

### Claim 9 (depends on 8)

The method of claim 8, wherein the memory identifier corresponds to a page of the memory or a block of the memory.

### Claim 10 (depends on 1)

The method of claim 1, wherein the memory corresponds to at least one of a DRAM, an SRAM, a NAND, resistive memory, or phase change memory, or any combination thereof.

### Claim 11 (depends on 1)

The method of claim 1, further comprising repairing the second regions corresponding to most certain significant bits of a neural network weight.

### Claim 12

An apparatus comprising:
a memory array;
a register, the register configured to store a memory address map including a plurality of memory addresses and indicators that certain of the plurality of memory addresses are associated with a high priority level and certain are associated with a lower priority level than the high priority level, wherein the high priority level is associated with a first bit error rate and wherein the lower priority level is associated with a second bit error rate, the second bit error rate different than the first error rate; and
a memory controller coupled to the memory array and configured to:
based on a comparison of defective memory addresses of defective memory cells of the memory array with the memory address map, prioritize repair of the first memory addresses associated with the high priority level before repair of the second memory addresses associated with the lower priority level.

### Claim 13 (depends on 12)

The apparatus of claim 12, wherein the memory controller is further configured to identify the defective memory addresses.

### Claim 14 (depends on 12)

The apparatus of claim 12, wherein the memory controller is further configured to:
determine that a portion of repairs of the first memory addresses associated with the high priority level are complete in a first region of the memory array; and
repair at least one memory address of the memory addresses associated with the lower priority level.

### Claim 15 (depends on 12)

The apparatus of claim 12, wherein the memory controller is further configured to:
determine that a portion of repairs of the memory addresses associated with the high priority level are not complete in a first region of the memory array; and
fail the first region of memory array.

### Claim 16 (depends on 12)

The apparatus of claim 12 further comprising:
a processor or host device coupled to the memory controller, wherein the memory controller is configured to receive memory commands from the processor or the host device.

### Claim 17 (depends on 12)

The apparatus of claim 12, further comprising:
a memory interface coupled to the memory controller and configured to communicate with the memory array, wherein the memory controller is configured to provide the memory map to the memory array via the memory interface.

### Claim 18 (depends on 12)

The apparatus of claim 12, wherein the certain of the plurality of memory addresses associated with the high priority level are accessed at a higher frequency than the memory addresses associated with the lower priority level.

### Claim 19 (depends on 12)

The apparatus of claim 12, wherein the plurality of memory addresses associated with the high priority level are configured to store weights for a neural network.

### Claim 20 (depends on 12)

The apparatus of claim 12, further comprising at least one fuse circuit configured to perform at least one of the repairs.
