Common Counter ADC for Temperature-Stable Neural Compute
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neural network computing techniques are energy-intensive and temperature-dependent, making them inefficient for various applications that require high-performance compute operations.
Innovation Solution
A low-power, temperature-independent analog-to-digital converter (ADC) method using a common counter that adjusts source and sink currents based on temperature changes, allowing for accurate measurement of average neuron current through charge accumulation in non-volatile cells.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional neural network computing techniques are used, then compute operations can be performed, but energy consumption is high and performance is temperature-dependent
Solution Approach 1:
The patent changes the operating parameters of the ADC circuit by dynamically adjusting the counter clock frequency based on temperature conditions. At higher temperatures, the clock frequency is reduced to compensate for increased thermal noise and leakage currents, thereby maintaining reliable operation across temperature variations while enabling energy-efficient neural network computations
2Productivity
If high-performance neural network compute operations are implemented, then processing capability is improved, but energy consumption increases
Solution Approach 1:
The patent implements periodic action through the use of a counter-based ADC that operates in discrete time intervals. The counter increments periodically based on clock cycles, converting analog neuron currents to digital values through time-discretized sampling. This periodic operation enables high-performance neural network computations while consuming significantly less energy than continuous-operation conventional ADCs
Solution Approach 2:
The patent replaces traditional high-power ADC conversion mechanisms with a simplified counter-based digital accumulation approach. Instead of using complex analog-to-digital conversion circuits that consume high power, the system uses a digital counter to accumulate and measure current over time, achieving the same conversion function with much lower energy consumption suitable for energy-efficient neural network computing
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 solution enables efficient neural network computations by reducing energy consumption and maintaining performance across temperature variations, enhancing the applicability of neural networks in diverse applications.
Implementation Method 1
The method uses charge accumulation for detecting the average neuron current
Data Source
AI summary
The method provides for a low power and a temperature independent analog to digital convertor for systems which use non-volatile cells for forming neurons to be used for neural network applications. The method uses a common counter which can be an up-counter or a down-counter depending on implementation, but in which the source and sink currents to a comparator are changed with temperature by the same percentage as the average bit line current for specific weight distributions programmed in the non-volatile cells forming the neurons. The method uses charge accumulation for detecting the average neuron current.