# US 2025/0095764 A1 · Conductance modulation in computational memory

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

- Number: US 2025/0095764 A1
- Document ID: US20250095764A1
- Status: Published application
- Kind: A1
- Application number: 18/368,158
- Filed: 2023-09-14
- Published: 2025-03-20
- Earliest claimed: 2023-09-14
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Ghazi Sarwat Syed; Benedikt Kersting; Abu Sebastian
- Web page: https://patentgrep.com/patent/US20250095764A1

## CPC

- G11C 11/54: using elements simulating biological cells, e.g. neuron
- G11C 29/52: Protection of memory contents; Detection of errors in memory contents
- G11C 11/1659: {Cell access}
- G11C 11/1673: {Reading or sensing circuits or methods}
- G11C 11/5628: {Programming or writing circuits; Data input circuits}
- G11C 11/5642: {Sensing or reading circuits; Data output circuits}
- G11C 11/5678: {using amorphous/crystalline phase transition storage elements}
- G11C 13/0004: {comprising amorphous/crystalline phase transition cells}
- G11C 13/003: {Cell access}
- G11C 13/004: {Reading or sensing circuits or methods}
- G11C 2207/2254: Calibration
- G11C 2213/79: Array wherein the access device being a transistor
- G11C: STATIC STORES (semiconductor memory devices H10B)

## Abstract

Systems and methods for modulating conductance of a plurality of unit cells are described. A controller can determine a plurality of conductance values for a plurality of unit cells arranged in a crossbar arrangement. The plurality of unit cells can include non-volatile memory devices. The plurality of conductance values can be less than maximum conductance values of the plurality of unit cells. The controller can determine a read gate voltage for driving a plurality of access transistors to read from connected to the plurality of unit cells under a read operation. The controller can map a plurality of weights to the plurality of conductance values and to the read gate voltage.

## Claims

### Claim 1

A computer-implemented method comprising:
determining a plurality of conductance values for a plurality of unit cells arranged in a crossbar arrangement, wherein the plurality of unit cells comprises non-volatile memory (NVM) devices, and the plurality of conductance values are less than maximum conductance values of the plurality of unit cells;
determining a read gate voltage for driving a plurality of access transistors to read from the plurality of unit cells; and
mapping a plurality of weights to the plurality of conductance values and to the read gate voltage.

### Claim 2 (depends on 1)

The method of claim 1, wherein a plurality of calibration weights are mapped to the plurality of unit cells, and determining the plurality of conductance values comprises:
setting the read gate voltage to an initial value;
using the initial value of the read gate voltage to perform a read operation on the plurality of unit cells;
determining a conductance distribution of the plurality of unit cells based on a result of the read operation; and
determining the plurality of conductance values using the conductance distribution.

### Claim 3 (depends on 2)

The method of claim 2, wherein determining the plurality of conductance values using the conductance distribution comprises:
specifying a conductance value that is greater than conductance values of a portion of the plurality of unit cells; and
setting the specified conductance value as one of the plurality of conductance values.

### Claim 4 (depends on 1)

The method of claim 1, wherein determining the read gate voltage comprises:
using the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
using an initial value of the read gate voltage to perform a read operation on the plurality of unit cells;
determining an error of a result of the read operation; and
using the error to adjust the read gate voltage.

### Claim 5 (depends on 1)

The method of claim 1, wherein determining the read gate voltage comprises:
using the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
using a specific value of the read gate voltage to perform a read operation on the plurality of unit cells;
determining an error of a compute result of the read operation;
determining a difference between the error and a previous error, wherein the previous error is from a previous iteration of read operation on the plurality of unit cells mapped to the calibration weights using previously determined values of the plurality of conductance values; and
using the difference to adjust the read gate voltage.

### Claim 6 (depends on 5)

The method of claim 5, wherein using the difference to adjust the read gate voltage comprises:
in response to the difference indicating the error is equal to or greater than the previous error, setting the specific value of the read gate voltage as the read gate voltage; and
in response to the difference indicating the error is less than the previous error, reducing the read gate voltage to a new value.

### Claim 7 (depends on 1)

The method of claim 1, further comprising:
using the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
using a plurality of candidate values of the read gate voltage to perform a plurality of read operations on the plurality of unit cells;
identifying a specific value among the candidate values of the read gate voltage that causes a result of the read operation to have the lowest compute error; and
determining the specific value as an optimal value of the read gate voltage.

### Claim 8

A system comprising:
a memory device including:
a plurality of unit cells arranged in a crossbar arrangement, wherein the plurality of unit cells comprises non-volatile memory (NVM) devices; and
a plurality of access transistors connected to the plurality of unit cells;
a controller configured to:
determine a plurality of conductance values for the plurality of unit cells, wherein the plurality of conductance values are less than maximum conductance values of the plurality of unit cells;
determine a read gate voltage for driving the plurality of access transistors to read from the plurality of unit cells; and
map a plurality of weights to the plurality of conductance values and to the read gate voltage.

### Claim 9 (depends on 8)

The system of claim 8, wherein a plurality of calibration weights are mapped to the plurality of unit cells, and the controller is configured to:
set the read gate voltage to an initial value;
use the initial value of the read gate voltage to perform a read operation on the plurality of unit cells;
determine a conductance distribution of the plurality of unit cells based on a result of the read operation;
determine the plurality of conductance values using the conductance distribution.

### Claim 10 (depends on 9)

The system of claim 9, wherein the controller is configured to:
specify a conductance value that is greater than conductance values of a portion of the plurality of unit cells; and
set the specified conductance value as one of the plurality of conductance values.

### Claim 11 (depends on 8)

The system of claim 8, wherein the controller is configured to:
use the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
use an initial value of the read gate voltage to perform a read operation on the plurality of unit cells;
determine an error of a result of the read operation; and
use the error to adjust the read gate voltage.

### Claim 12 (depends on 8)

The system of claim 8, wherein the controller is configured to:
use the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
use a specific value of the read gate voltage to perform a read operation on the plurality of unit cells;
determine an error of a compute result of the read operation;
determine a difference between the error and a previous error, wherein the previous error is from a previous iteration of read operation on the plurality of unit cells mapped to the calibration weights using previously determined values of the plurality of conductance values; and
use the difference to adjust the read gate voltage.

### Claim 13 (depends on 12)

The system of claim 12, wherein the controller is configured to:
in response to the difference indicating the error is equal to or greater than the previous error, set the specific value of the read gate voltage as the read gate voltage; and
in response to the difference indicating the error is less than the previous error, reduce the read gate voltage to a new value.

### Claim 14 (depends on 8)

The system of claim 8, wherein the controller is configured to:
use the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
use a plurality of candidate values of the read gate voltage to perform a plurality of read operations on the plurality of unit cells;
identify a specific value among the candidate values of the read gate voltage that causes a result of the read operation to have the lowest compute error; and
determine the specific value as an optimal value of the read gate voltage.

### Claim 15 (depends on 8)

The system of claim 8, wherein each one of the plurality of unit cells is one of an unprojected phase change memory (PCM) device and a projected PCM device.

### Claim 16

A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by a device to cause the device to:
determining a plurality of conductance values for a plurality of unit cells arranged in a crossbar arrangement, wherein the plurality of unit cells comprises non-volatile memory (NVM) devices, and the plurality of conductance values are less than maximum conductance values of the plurality of unit cells;
determining a read gate voltage for driving a plurality of access transistors to read from the plurality of unit cells; and
mapping a plurality of weights to the plurality of conductance values and to the read gate voltage.

### Claim 17 (depends on 16)

The computer program product of claim 16, wherein a plurality of calibration weights are mapped to the plurality of unit cells, and the device is further caused to perform:
set the read gate voltage to an initial value;
use the initial value of the read gate voltage to perform a read operation on the plurality of unit cells;
determine a conductance distribution of the plurality of unit cells based on a result of the read operation;
determine the plurality of conductance values using the conductance distribution;
specify a conductance value that is greater than conductance values of a portion of the plurality of unit cells; and
set the specified conductance value as one of the plurality of conductance values . . .

### Claim 18 (depends on 16)

The computer program product of claim 16, wherein the device is further caused to perform:
use the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
use an initial value of the read gate voltage to perform the read operation on the plurality of unit cells;
determine an error of a result of the read operation; and
use the error to adjust the read gate voltage.

### Claim 19 (depends on 16)

The computer program product of claim 16, wherein the device is further caused to perform:
use the plurality of conductance values to perform a programming operation to map a plurality of calibration weights to the plurality unit cells;
use a specific value of the read gate voltage to perform a read operation on the plurality of unit cells;
determine an error of a compute result of the read operation;
determine a difference between the error and a previous error, wherein the previous error is from a previous iteration of read operation on the plurality of unit cells mapped to the calibration weights using previously determined values of the plurality of conductance values; and
use the difference to adjust the read gate voltage.

### Claim 20 (depends on 16)

The computer program product of claim 16, wherein the device is further caused to perform:
in response to the difference indicating the error is equal to or greater than the previous error, set the specific value of the read gate voltage as the read gate voltage; and
in response to the difference indicating the error is less than the previous error, reduce the read gate voltage to a new value.
