Spike Neural Network Circuit Dynamic Membrane Capacitor Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The dynamic range of the membrane in spike neural network circuits implemented with semiconductor circuits is limited, leading to charge calculation errors, especially when large weights are stored in synapses, as current sources often operate outside their saturation region.
Innovation Solution
A spike neural network circuit design that includes a neuron with a firing unit, membrane capacitors, a switch controller, and a spike output generator, which compares membrane node potentials to a reference potential, switches membrane capacitors between power supply and ground, and generates spike outputs based on firing signals, thereby expanding the dynamic range and ensuring current source operation within the saturation region.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the membrane voltage is constrained to ensure current source saturation region operation, then the current source operates reliably in saturation region, but the dynamic range of the membrane is limited
Solution Approach 1:
The membrane capacitor is divided into multiple segments (first membrane capacitor and second membrane capacitor) that can be independently controlled. This segmentation allows the circuit to expand its effective dynamic range by switching between different capacitor configurations while maintaining proper current source operation in each segment.
Solution Approach 2:
The patent implements dynamic switching of membrane capacitors based on the accumulation state of membrane signals. The switch controller dynamically connects or disconnects membrane capacitors from the membrane node, allowing the system to adapt its capacitance value in real-time to maintain current source saturation while handling varying signal accumulation requirements.
2Measurement precision
If the membrane dynamic range is expanded to improve charge calculation accuracy, then charge calculation accuracy improves, but current sources may operate outside saturation region
Solution Approach 1:
The switch controller monitors the accumulation state of membrane signals and provides feedback control to dynamically adjust the configuration of membrane capacitors. This feedback mechanism ensures that the system expands its dynamic range only when needed and maintains current source operation within the saturation region by adjusting capacitor connectivity based on real-time signal conditions.
Solution Approach 2:
The system dynamically reconfigures the membrane capacitor network in response to signal accumulation levels. By switching capacitors in and out of the circuit based on operational needs, the system achieves expanded dynamic range for accurate charge calculation while ensuring current sources remain in saturation region through controlled capacitance adjustment.
3Adaptability or versatility
If multiple membrane capacitors are used to expand dynamic range, then membrane dynamic range is expanded, but device complexity increases
Solution Approach 1:
The membrane capacitor is segmented into multiple independent units that can be selectively activated. This segmentation approach expands the functional dynamic range of the neuron circuit while managing complexity by using simple switching mechanisms to control each segment, rather than requiring a completely redesigned complex circuit architecture.
Solution Approach 2:
The multiple membrane capacitors serve universal functions within the neuron circuit - they can be individually or collectively connected to the membrane node depending on operational requirements. This multi-functionality allows the same capacitor components to provide both standard capacitance operation and expanded dynamic range operation, reducing the need for entirely separate circuit paths.
Data Source
AI summary
Disclosed is a spike neural network circuit, which includes an axon generating a spike input, a synapse performing a weight calculation and generating a membrane signal based on the weight calculation, and a neuron accumulating the membrane signal to generate a spike output, and the neuron includes a firing unit that compares a potential of a membrane node where the membrane signal is accumulated with a reference potential and fires based on the comparison result, membrane capacitors connected to the membrane node, a switch controller that outputs switching signals based on the firing of the firing unit, switches that connects each of membrane capacitors to one of a power supply voltage and a ground voltage in response to the switching signals, and a spike output generator that generates the spike output based on the plurality of switching signals and the firing of the firing unit.