Neuromorphic Memory Readout With Reference ADC Compensation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neuromorphic computing devices using resistive memory cells face accuracy reductions due to temperature and time dependency, affecting the reliability of inference operations.
Innovation Solution
A neuromorphic computing device is designed with a first memory cell array and a second reference memory cell array, utilizing a current-voltage converting circuit and an analog-digital converting circuit to convert read currents into digital signals using reference voltages, thereby mitigating the effects of temperature and time dependency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If resistive memory cells are used for neuromorphic computing operations, then the device can perform MAC operations efficiently, but the accuracy of inference is reduced due to temperature and time dependency
Solution Approach 1:
The patent implements a feedback mechanism where read currents from resistive memory cells are converted to voltages, compared against reference voltages, and the comparison results are used to determine digital output values. This feedback loop compensates for temperature and time dependency by continuously adjusting the inference based on actual measured values rather than relying solely on predetermined weights.
Solution Approach 2:
The patent changes the operational parameters by converting from direct current-based computation to voltage-based computation with digital output. By using current-voltage conversion and analog-digital conversion, the system transforms the physical parameters (current, voltage) and data representation (analog to digital) to achieve temperature and time independence while maintaining computational efficiency.
2Speed
If analog signals are used directly from resistive memory cells, then the computation speed is high, but temperature and time dependency reduces inference accuracy
Solution Approach 1:
The patent substitutes the mechanical/electrical analog system with a digital system. By converting analog voltages to digital signals through analog-digital conversion, the system replaces continuous analog computation with discrete digital computation, thereby eliminating the sensitivity to temperature and time variations while preserving computation speed.
Solution Approach 2:
The patent introduces intermediary conversion circuits (current-voltage conversion circuit and analog-digital conversion circuit) between the resistive memory cells and the output. These intermediaries transform the temperature-sensitive analog currents into temperature-insensitive digital signals, acting as a buffer that preserves speed while improving precision.
3Reliability
If reference voltages are used for analog-digital conversion, then temperature and time dependency is compensated, but the device complexity increases
Solution Approach 1:
The patent segments the conversion process into distinct functional modules: current-voltage conversion circuit, reference voltage generation circuit, and analog-digital conversion circuit. Each module performs a specific function, making the overall complex system manageable and maintainable while achieving temperature and time compensation through coordinated operation of these segments.
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
The device maintains accuracy and reliability of inference operations by compensating for temperature and time-dependent variations in resistive memory cells, enhancing the overall performance of neuromorphic computing.
Implementation Method 1
a current-voltage converting circuit configured to output a plurality of signal voltages respectively corresponding to the plurality of read currents
Implementation Method 2
an analog-digital converting circuit configured to convert the plurality of signal voltages to a plurality of digital signals using the at least one reference voltage
Data Source
AI summary
A neuromorphic computing device includes a first memory cell array comprising a plurality of resistive memory cells and configured to output a plurality of read currents through a plurality of bit lines or source lines; a second memory cell array comprising a plurality of reference resistive memory cells and configured to output at least one reference current through at least one reference bit line or at least one reference source line; a current-voltage converting circuit configured to output a plurality of signal voltages respectively corresponding to the plurality of read currents and output at least one reference voltage corresponding to the at least one reference current; and an analog-digital converting circuit configured to convert the plurality of signal voltages to a plurality of digital signals using the at least one reference voltage and output the plurality of digital signals.