Memristor Neural Network Circuit With Periodic Offset Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural network circuits face challenges in reducing power consumption and circuit scale while maintaining calculation accuracy and signal processing speed, particularly due to the inefficiencies in applying bias voltages and correcting offsets in memristor-based systems.
Innovation Solution
The neural network circuit incorporates a lattice-shaped memristor storage portion, D/A converters, drive amplifiers, I/V conversion amplifiers, and A/D converters, with an offset correction mechanism that controls bias and reference voltages to optimize memristor operations, reducing bias current duration and correcting offsets efficiently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If bias voltage is continuously applied to memristors for offset correction, then calculation accuracy is improved, but power consumption increases
Solution Approach 1:
The patent implements periodic offset correction by controlling the bias application amplifier to apply bias voltage only at specific timing intervals rather than continuously. The controller activates the bias application amplifier during offset correction periods and deactivates it during normal operation, achieving both high calculation accuracy through regular offset correction and low power consumption by minimizing bias current duration.
2Measurement precision
If multiple separate circuits are used for bias application and offset correction, then correction accuracy is improved, but circuit scale increases
Solution Approach 1:
The patent merges the bias application amplifier and offset correction amplifier into a single integrated circuit. The bias application amplifier is configured to output both bias voltages for normal operation and correction voltages for offset correction, eliminating the need for separate amplifier circuits while maintaining correction accuracy through controller-managed voltage output switching.
Solution Approach 2:
The bias application amplifier is designed with multi-functionality to serve dual purposes: applying bias voltages during normal operation and applying correction voltages during offset correction. The amplifier responds to control signals from the controller to switch between these functions, reducing overall circuit scale while preserving correction accuracy.
3Measurement precision
If bias current duration is extended for thorough offset correction, then offset correction accuracy is improved, but power consumption increases
Solution Approach 1:
The controller implements periodic offset correction by activating the bias application amplifier only during specific correction intervals rather than continuously. This periodic activation achieves thorough offset correction accuracy while minimizing bias current duration and associated power consumption by keeping the amplifier inactive during normal operation periods.
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 effectively reduces power consumption and circuit scale while maintaining high calculation accuracy and signal processing speed by optimizing memristor operations through controlled bias voltage application and offset correction.
Implementation Method 1
an element having two terminals as a synapse, the element being nonvolatile and capable of varying a conductance value and being referred to as a memristor
Implementation Method 2
a D/A converter that receives the data to apply a voltage signal
Implementation Method 3
a drive amplifier that receives the output signal of the D/A converter to apply the voltage signal to the voltage input terminal
Implementation Method 4
an I/V conversion amplifier connected with a current output terminal of the crossbar circuit, and converting a current flowing in the current output terminal into a voltage signal
Implementation Method 5
an A/D converter that A/D-converts the voltage signal converted by the I/V conversion amplifier
Implementation Method 6
At least one of the drive amplifiers may be a bias application amplifier that may apply the bias voltage having an opposite polarity and a reference voltage
Data Source
AI summary
A neural network circuit includes: a storage portion that includes memristors; D/A converters; drive amplifiers; I/V conversion amplifiers; A/D converters; and offset correctors. The offset corrector includes a first latch circuit, a second latch circuit, a subtractor that subtracts latch data, and a controller. In performing a bias setting operation, the controller controls a bias application amplifier to output the bias voltage, controls each of the D/A converters to cause the drive amplifier other than the bias application amplifier to output a reference voltage, and also cause the first latch circuit to latch the output data. In performing a normal operation, the controller controls the bias application amplifier to output the reference voltage, controls each of the D/A converters to cause the drive amplifier other than the bias application amplifier to output the signal voltage, and also cause the second latch circuit to latch the output data.