Memristor Neural Network Circuit for Temperature-Stable I-V Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural network circuits using memristors face challenges in correcting temperature characteristics, which are time-dependent and input-voltage-polarity-dependent, making it difficult to maintain consistent retention and output polarity.
Innovation Solution
A neural network circuit design that includes memristors connected in a matrix with a voltage application device and current-voltage conversion amplification circuits, where the feedback resistor is also a memristor with aligned polarity direction, allowing for uniform temperature characteristics and correction of temperature-dependent issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If memristors are used as memory elements in a neural network circuit, then the circuit can achieve non-volatile storage and analog computing capabilities, but the temperature characteristics become time-dependent and input-voltage-polarity-dependent, making it difficult to maintain consistent retention and output polarity
Solution Approach 1:
The patent applies feedback by using a sense amplifier that reads the voltage on the bit line and feeds back a corrected voltage to compensate for temperature drift. The sense amplifier detects the actual voltage level affected by temperature and polarity variations, then adjusts the read operation to output a consistent logic level, thereby resolving the temperature characteristic inconsistency without requiring complex circuit reconfiguration
Solution Approach 2:
The patent changes the operating parameters of the memristor circuit by applying different read voltages depending on the detected temperature and polarity conditions. The control circuit adjusts the magnitude and polarity of read voltages dynamically to compensate for temperature drift, allowing the same memristor array to maintain reliable operation across varying temperature conditions without increasing device complexity
2Reliability
If the polarity direction of feedback resistor memristors is aligned with memory element memristors, then temperature characteristics can be corrected uniformly, but the circuit design becomes more constrained in terms of polarity alignment
Solution Approach 1:
The patent deliberately introduces asymmetry in the circuit design by aligning the polarity direction of feedback resistor memristors with memory element memristors. This asymmetric polarity configuration allows the feedback mechanism to effectively counteract temperature drift in one polarity direction, and the circuit is designed to handle the other polarity case separately, achieving temperature correction while managing manufacturing constraints through deliberate asymmetric design
3Measurement precision
If multiple I-V conversion amplification circuits are used to convert currents to voltages, then the neural network computation can be performed accurately, but the number of components and circuit complexity increases
Solution Approach 1:
The patent merges the I-V conversion amplification function with the sense amplifier that already exists in the memory read circuitry. By combining the current-to-voltage conversion with the existing sensing and feedback mechanisms, the patent achieves accurate neural network computation without adding separate dedicated amplification circuits for each function, thereby maintaining measurement precision while controlling device complexity through functional integration
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 enables appropriate correction of temperature characteristics and uniformity of electric current polarity, effectively addressing time-dependent and input-voltage-polarity-dependent variations in memristor resistance, ensuring stable output.
Implementation Method 1
A feedback resistor of each I-V conversion amplification circuit includes a memristor. A polarity direction of the memristor of the feedback resistor of the I-V conversion amplification circuit is aligned with polarity directions of the memristors of the memory elements acting as an input resistor of the I-V conversion amplification circuit
Implementation Method 2
a plurality of current-voltage (I-V) conversion amplification circuits arranged to convert currents flowing via the memory elements into voltages and output the voltages
Implementation Method 3
a voltage application device arranged to apply a bias voltage to the memory device
Data Source
AI summary
A neural network circuit includes a memory device in which memristors being variable resistance elements are connected in a matrix and serve as memory elements of the memory device. The neural network circuit further includes a voltage application device arranged to apply a bias voltage to the memory device and current-voltage (I-V) conversion amplification circuits arranged to convert currents flowing via the memory elements into voltages and output the voltage. A feedback resistor of a respective I-V conversion amplification circuit includes a memristor. The feedback resistor of a respective I-V conversion amplification circuit and the memory elements acting as an input resistor of the I-V conversion amplification circuit are connected to align a polarity direction of the memristor of the feedback resistor and polarity directions of the memristors of the memory elements acting as the input resistor.