Nonvolatile Synapse Circuit for Dense Multiply-Accumulate Arrays

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing neural network devices require a large mounting area per synapse due to the need for increased load capacity of field-effect transistors to maintain channel resistance over time and to suppress leak currents, making high-density integration difficult.

Innovation Solution

A semiconductor device and multiply-accumulate operation device are designed with synapses comprising a nonvolatile variable resistance element and a fixed resistance element connected in series, allowing for reduced area usage by maintaining resistance values without power supply and integrating at higher densities.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If field-effect transistors are used to form synapses in neural network devices, then the channel resistance can be controlled to implement neural network operations, but the mounting area per synapse increases due to the need for increased load capacity and larger transistor area to suppress leak current

Engineering Contradiction:
Improvechannel resistance maintenanceVSAvoidmounting area per synapse
Core Design Contradiction:
ReliabilityVSArea of stationary object

Solution Approach 1:

The patent changes the fundamental parameter of resistance element type from volatile (field-effect transistor requiring power) to nonvolatile (variable resistance element). This parameter change allows the synapse to maintain its resistance value without power supply, eliminating the need for large load capacity transistors and reducing the mounting area per synapse while maintaining reliable channel resistance.

Inventive Principle:
Principle #35Parameter changes

2Stability of the object's composition

If field-effect transistors are used with increased load capacity to maintain channel resistance, then the resistance stability improves, but the device complexity and manufacturing difficulty increase

Engineering Contradiction:
Improvechannel resistance stabilityVSAvoidtransistor configuration complexity
Core Design Contradiction:
Stability of the object's compositionVSDevice complexity

Solution Approach 1:

The patent extracts the voltage control function from the synapse structure itself and places it in a separate control electrode, while the synapse core uses a simple variable resistance element without complex transistor configurations. This separation simplifies the synapse structure while maintaining resistance stability through the nonvolatile nature of the variable resistance element.

Inventive Principle:
Principle #2Taking out (Extraction)

3Reliability

If larger field-effect transistors are used to suppress leak current, then the leak current suppression improves, but the mounting area per synapse increases

Engineering Contradiction:
Improveleak current suppressionVSAvoidmounting area per synapse
Core Design Contradiction:
ReliabilityVSArea of stationary object

Solution Approach 1:

The patent changes the resistance element from volatile to nonvolatile, which fundamentally alters the leak current characteristics. The nonvolatile variable resistance element maintains its resistance state without power, effectively suppressing leak current without requiring larger transistor dimensions, thus reducing the mounting area per synapse.

Inventive Principle:
Principle #35Parameter changes

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This configuration reduces the mounting area per synapse, enabling higher-density integration and improved accuracy in neural network operations by stabilizing resistance values and minimizing current leakage.

Implementation Method 1

a nonvolatile variable resistance element taking a first resistance value and a second resistance value lower than the first resistance value

Methodology Applied
Scientific EffectNonvolatile resistance memory effect:

Implementation Method 2

a nonvolatile variable resistance element taking a first resistance value and a second resistance value lower than the first resistance value and a fixed resistance element having a resistance value higher than the second resistance value are connected in series

Methodology Applied
Scientific EffectElectrical resistance: Electrical Resistance

Implementation Method 3

an output line that outputs a sum of currents flowing through the plurality of synapses

Methodology Applied
Scientific EffectKirchhoff's current law:

Data Source

PatentUS11972229B2Semiconductor device and multiply-accumulate operation device
Publication Date: 2024.04.30 SONY SEMICON SOLUTIONS CORP
  • US11972229B2 patent drawing
  • US11972229B2 patent drawing
  • US11972229B2 patent drawing

AI summary

Semiconductor devices and multiply-accumulate operation devices are disclosed. In one example, a semiconductor device includes synapses in which a nonvolatile variable resistance element taking a first resistance value and a second resistance value lower than the first resistance value and a fixed resistance element having a resistance value higher than the second resistance value are connected in series. An output line outputs a sum of currents flowing through the plurality of synapses.