# US 2021/0241081 A1 · Spiking neural networks circuit and operation method thereof

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

- Number: US 2021/0241081 A1
- Document ID: US20210241081A1
- Status: Published application
- Kind: A1
- Application number: 16/872,404
- Filed: 2020-05-12
- Published: 2021-08-05
- Earliest claimed: 2020-02-04
- Assignees: Macronix International Co., Ltd.
- Applicants: Macronix International Co., Ltd.
- Inventors: Cheng-Lin SUNG; Teng-Hao YEH
- Web page: https://patentgrep.com/patent/US20210241081A1

## CPC

- G06N 3/063: using electronic means
- G06N 3/049: Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
- G11C 11/54: using elements simulating biological cells, e.g. neuron
- G11C 16/08: Address circuits; Decoders; Word-line control circuits
- G11C 16/10: Programming or data input circuits
- G11C 16/16: for erasing blocks, e.g. arrays, words, groups
- G11C 16/24: Bit-line control circuits
- G11C 16/26: Sensing or reading circuits; Data output circuits
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G11C: STATIC STORES (semiconductor memory devices H10B)

## Abstract

A spiking neural networks circuit and an operation method thereof are provided. The spiking neural networks circuit includes a bit-line input synapse array and a neuron circuit. The bit-line input synapse array includes a plurality of page buffers, a plurality of bit line transistors, a plurality of bit lines, a plurality of memory cells, one word line, a plurality of source lines and a plurality of source line transistors. The page buffers provides a plurality of data signals. Each of the bit line transistors is electrically connected to one of the page buffers. Each of the bit lines receives one of the data signals. The source line transistors are connected together. The neuron circuit is for outputting a feedback pulse.

## Claims

### Claim 1

A spiking neural networks (SNN) circuit, comprising:
a bit-line input synapse array, including:
a plurality of page buffers, providing a plurality of data signals;
a plurality of bit line transistors, each of which is electrically connected to one of the page buffers;
a plurality of bit lines, each of which is electrically connected to one of the bit line transistor for receiving one of the data signals;
a plurality of memory cells, each of which is electrically connected to one of the bit lines;
one word line, electrically connected to the memory cells;
a plurality of source lines, each of which is electrically connected to one of the memory cells; and
a plurality of source line transistors, each of which is electrically connected to one of the source lines, wherein the source line transistors are connected together; and
a neuron circuit, for outputting a feedback pulse.

### Claim 2 (depends on 1)

The spiking neural networks circuit according to claim 1, wherein the bit line transistors and the bit lines form a presynaptic zone, the memory cells and the word line form a synaptic zone, and the source line transistors and the source lines form a postsynaptic zone.

### Claim 3 (depends on 1)

The spiking neural networks circuit according to claim 1, wherein the memory cells storing a plurality of weights are connected to the only one word line.

### Claim 4 (depends on 1)

The spiking neural networks circuit according to claim 1, wherein the bit line transistors are triple well devices.

### Claim 5 (depends on 1)

The spiking neural networks circuit according to claim 1, wherein the source line transistors are triple well devices.

### Claim 6 (depends on 1)

The spiking neural networks circuit according to claim 1, wherein the neuron circuit includes:
a current mirror circuit, electrically connected to the bit-line input synapse array;
a membrane capacitor, electrically connected to the current mirror circuit;
an integrated-and-fired circuit, electrically connected to the membrane capacitor; and
a pulse generator, electrically connected to the integrated-and-fired circuit and the bit line transistors, wherein the integrated-and-fired circuit triggers the pulse generator to generate the feedback pulse or an output pulse when a potential of the membrane capacitor exceeds a threshold.

### Claim 7 (depends on 6)

The spiking neural networks circuit according to claim 6, wherein the feedback pulse is used to turn on the bit line transistors.

### Claim 8 (depends on 6)

The spiking neural networks circuit according to claim 6, wherein the neuron circuit further includes:
an inhibitory field-effect transistor (FET), connected to the membrane capacitor, wherein the inhibitory field-effect transistor suppresses an firing action of another neuron circuit when the potential of the membrane capacitor exceeds the threshold.

### Claim 9 (depends on 6)

The spiking neural networks circuit according to claim 6, wherein the pulse generator is connected to the source line transistors.

### Claim 10 (depends on 6)

The spiking neural networks circuit according to claim 6, wherein the current mirror circuit includes:
a first current mirror, including two N type Metal-Oxide-Semiconductor (NMOS) transistors; and
a second current mirror, connected to the first current mirror, wherein the second current mirror includes two P type Metal-Oxide-Semiconductor (PMOS) transistors.

### Claim 11 (depends on 10)

The neuron circuit of the spiking neural networks circuit according to claim 10, wherein the NMOS transistors are triple well devices.

### Claim 12 (depends on 1)

The spiking neural networks circuit according to claim 1, wherein each of the page buffers includes:
a level shifter for outputting a plurality of shifted data signals, wherein a variety of the shifted data signals is larger than a variety of the data signal.

### Claim 13

An operation method of a spiking neural networks (SNN) circuit, wherein the spiking neural networks circuit comprises a bit-line input synapse array and a neuron circuit, the bit-line input synapse array includes a plurality of bit lines, a plurality of memory cells, one word line and a plurality of source lines, each of the memory cells is electrically connected to one of the bit lines, the word line is electrically connected to the memory cells, each of the source lines is electrically connected to one of the memory cells, and the operation method comprises:
inputting a plurality of data signals to the bit lines during a read phase; and
collecting, by the source lines, currents reflecting the data signals during the read phase.

### Claim 14 (depends on 13)

The operation method according to claim 13, wherein each of the data signals is a 0V signal or a 0.2V signal.

### Claim 15 (depends on 13)

The operation method according to claim 13, wherein during the read phase, a plurality of bit line transistors, each of which is connected to one of the bit lines, are turned on, and a plurality of source line transistor, each of which is connected to one of the source lines, are tuned on.

### Claim 16 (depends on 13)

The operation method according to claim 13, wherein if the neuron circuit is a winner, the operation method further comprises:
inputting a plurality of shifted data signals to the bit lines during an update phase, wherein the shifted data signals are obtained by a level shifter and a variety of the shifted data signals is larger than a variety of the data signal;
applying a long-term potentiation signal to the word line during an erasing step of the update phase, such that an erasing pulse is formed between the word line and the bit lines; and
applying a long-term depression signal to the word line during a programming step of the update phase, such that a programming pulse is formed between the word line and the bit lines.

### Claim 17 (depends on 16)

The operation method according to claim 16, wherein each of the shifted data signals is −6V signal or a 6V signal, the long-term potentiation signal is a −6V signal, the erasing pulse is a −12V signal, the long-term depression signal is a 6V signal, and the programming pulse is a 12V signal.

### Claim 18 (depends on 16)

The operation method according to claim 16, wherein during the update phase, if the neuron circuit is the winner, a plurality of bit line transistors, each of which is connected to one of the bit lines, are turned on, and a plurality of source line transistors, each of which is connected to one of the source lines, are turned off.

### Claim 19 (depends on 18)

The operation method according to claim 18, wherein during the update phase, if the neuron circuit is a loser, the bit line transistors and the source line transistors are turned off.

### Claim 20 (depends on 13)

The operation method according to claim 13, wherein the memory cells storing a plurality of weights are connected to the only one word line.