Matched Feedback Integrate-and-Fire Neuron Circuit for PVT Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing integrate-and-fire neuron circuits suffer from high variability and reduced accuracy due to process, voltage, and temperature variations, leading to significant quantization errors and loss of information during reset periods.
Innovation Solution
The proposed neuron circuit incorporates a feedback subcircuit that generates charge packets of opposite signs based on spike output signals, maintaining charge accumulation and reducing reset periods, while employing structural matching of functional blocks to minimize variations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a reset operation is performed to clear charge on the capacitor when an output spike is created, then the circuit can generate spike output signals, but continuing inputs are lost during the blind period and quantization error is not stored
Solution Approach 1:
The patent introduces an intermediary storage mechanism (additional capacitor or memory element) that holds quantization error and blind period input signals separately from the main integration capacitor. This intermediary storage allows the reset operation to clear the main capacitor for new integration while preserving quantization error and missed inputs for later processing, thus resolving the contradiction between spike generation and information retention
Solution Approach 2:
The patent implements a mechanism where quantization error and blind period inputs are temporarily discarded from the main integration path during reset, but are recovered and stored in separate storage elements. These recovered signals are then reintegrated in subsequent operations, eliminating permanent information loss while maintaining spike generation functionality
2Reliability
If the capacitor is reset after each spike output, then the circuit can maintain stable operation, but quantization error is lost and accuracy decreases
Solution Approach 1:
The patent segments the capacitor system into multiple functional components: a main integration capacitor for spike generation and separate storage elements (additional capacitors or memory) for quantization error and blind period inputs. This segmentation allows the main capacitor to be reset for stable operation while preserving precision-critical information in separate storage regions that are not cleared during reset
Solution Approach 2:
The patent implements feedback mechanisms where stored quantization error and recovered blind period inputs are fed back into the integration process. This feedback loop continuously corrects accumulated errors and improves measurement precision over time, while the main reset operation maintains stable spike generation behavior
3Measurement precision
If structural matching is used to reduce PVT variations, then accuracy improves, but device complexity increases due to additional feedback subcircuit and storage elements
Solution Approach 1:
The patent designs the additional storage elements and feedback subcircuit to serve multiple functions: storing quantization error, preserving blind period inputs, enabling accuracy improvement through feedback, and providing PVT variation compensation. This multi-functionality reduces the need for separate dedicated components, thereby limiting the increase in device complexity while achieving improved precision and robustness
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 approach enhances accuracy and reduces noise by allowing quantization errors to accumulate over time, providing higher precision and less sensitivity to PVT variations, with improved performance in cascaded neuron circuits.
Implementation Method 1
A capacitance between the common line and a reference potential stores the charges from the positive and negative charge packets and converts the sum of the charges into a voltage
Implementation Method 2
A first comparator compares this voltage against an upper reference voltage for quantization and is configured to generate the positive spike output signals at the positive output, while said voltage is larger than the upper reference voltage
Implementation Method 3
A second comparator compares said voltage against a lower reference voltage for quantization and is configured to generate the negative spike output signals at the negative output, while said voltage is smaller than the lower reference voltage
Data Source
AI summary
The present invention relates to an integrate-and-fire neuron circuit for signed processing which is characterized by a feedback subcircuit connected to the positive and negative outputs of the circuit. This feedback subcircuit is configured to generate and output positive charge packets to a common line of the neuron circuit on each negative spike output signal and stored weights and to generate and output negative charge packets to the common line based on each positive spike output signal and stored weights. Due to this feedback circuit that is build in the same way as the input weighting circuit, structural matching and therefore higher accuracy and less variation of the behavior over PVT variations is achieved.