LIFR CMOS Cell Layout With Self-Reset for Low-Power Neuromorphic ICs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neuromorphic computing implementations rely on bulk, power-hungry designs such as SRAM-based and comparator-based designs, which significantly increase the size and power consumption of integrated circuits, and lack an ideal artificial neuron with leaky-integrate-fire and self-reset functionalities.
Innovation Solution
A leaky integrate-fire and reset (LIFR) integrated complementary metal oxide semiconductor (CMOS) cell is developed, utilizing a partially depleted semiconductor-on-insulator (PDSOI) wafer to reduce footprint and power consumption, incorporating a reset field effect transistor to detect and reset the LIF CMOS cell.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If SRAM-based or comparator-based designs are used for neuromorphic neural hardware activation functions, then the computational functionality is achieved, but the footprint significantly increases the size of the integrated circuit
Solution Approach 1:
The patent merges the leaky integrate-fire neuron core with reset functionality into a single integrated LIFR CMOS cell structure. By combining the integration capacitor, fire threshold detection, and reset mechanisms within one cell, the design achieves compact footprint while maintaining full neuromorphic activation function capability.
Solution Approach 2:
The LIFR CMOS cell serves multiple functions: it performs leaky integration of input spikes, detects fire threshold conditions, and enables self-reset functionality. This multi-functional design eliminates the need for separate SRAM blocks or comparator circuits, significantly reducing the overall IC footprint.
2Adaptability or versatility
If SRAM-based or comparator-based designs are used for neuromorphic neural hardware activation functions, then the computational functionality is achieved, but power consumption increases
Solution Approach 1:
The LIFR CMOS cell implements self-reset functionality where the cell automatically resets itself after firing without requiring external control signals. The reset mechanism is triggered by the fire event itself, eliminating the need for continuous external power management and reducing overall power consumption compared to SRAM-based designs.
Solution Approach 2:
The neuron operates in periodic cycles of integration, fire detection, and reset. The leaky integration continuously maintains the membrane potential while consuming minimal power, and the reset occurs periodically only when needed after fire detection, optimizing power consumption during idle periods.
3Adaptability or versatility
If conventional SRAM-based designs are used, then neural network functionality is implemented, but the complexity of the integrated circuit increases
Solution Approach 1:
The patent segments the neuron functionality into distinct functional blocks within the LIFR CMOS cell: the integration capacitor for leaky integration, the fire threshold detection mechanism, and the reset switch. This segmentation allows each component to perform its specific function efficiently while keeping the overall design simpler and more modular compared to monolithic SRAM-based approaches.
Data Source
AI summary
An integrated circuit (IC) is described. The IC includes a leaky integrate-fire (LIF) complementary metal oxide semiconductor (CMOS) cell. The LIF CMOS cell is configured to receive an input spike train and having an output to fire in response to the input spike train. The IC also includes a reset field effect transistor (FET) switch coupled to the output of the LIF CMOS cell. The reset FET switch is configured to detect the fire at the output of the LIF CMOS cell in response to the input spike train and to reset the LIF CMOS cell.