Cmos Resonate-and-Fire Neuron Emulates Biological Oscillations for Low-Power Edge Processing

The demand for energy-efficient computing is driving innovation in bio-inspired hardware, and researchers are increasingly turning to the brain for solutions. Giuseppe Leo, Paolo Gibertini, and Irem Ilter, working with colleagues at the University of Groningen, Bilkent University, the Technical University of Denmark, and elsewhere, have created a novel electronic neuron that mimics the behaviour of biological cells controlling oscillations in the nervous system. This team successfully fabricated a mixed-signal “Resonate-and-Fire” neuron circuit using standard semiconductor technology, offering a pathway towards real-time, low-power processing at the edge of computing networks. The resulting circuit, which selectively responds to specific frequencies, demonstrates the potential for large-scale integration and represents a significant step forward in exploiting bio-inspired designs for efficient temporal signal processing.

Low-Power Resonate-and-Fire Neuron Circuit Design

This research presents the design and implementation of a Resonate-and-Fire (R and F) neuron circuit using analog Very Large Scale Integration (VLSI) technology, aiming to create a low-power, biologically plausible neuron model for neuromorphic computing systems. R and F neurons selectively respond to specific frequencies of input signals, making them valuable for signal processing, pattern recognition, and sensory encoding. This work improves upon existing designs by focusing on power efficiency and addressing variations inherent in analog circuits. The design achieves significantly lower power consumption, measuring 571 picowatts, crucial for large-scale neuromorphic systems where power is a major constraint.

The team also analyzed how manufacturing imperfections affect the circuit’s performance, demonstrating its resilience to these variations. The circuit consumes 571 picowatts of power and exhibits an energy consumption of 196 femojoules per spike, demonstrating robustness to variations in transistor parameters through simulation results. The R and F neuron is implemented using subthreshold Complementary Metal-Oxide-Semiconductor (CMOS) technology, allowing for very low power consumption but requiring careful design to overcome sensitivity to variations. The circuit achieves resonance through a combination of capacitive and resistive elements, creating a frequency-selective response.

Analog VLSI enables a compact and efficient implementation, incorporating techniques to mitigate process variations through careful layout and component matching. The R and F neuron is well-suited for processing sensory signals, such as audio, video, and tactile data, and can be incorporated into event-based computing systems. The design is intended for integration into larger neuromorphic systems, and future work will explore integrating the neuron with other neuromorphic components and developing more sophisticated learning algorithms.

CMOS Neuron Circuit Emulates Biological Oscillation

Researchers engineered a novel Complementary Metal-Oxide-Semiconductor (CMOS) mixed-signal Resonate-and-Fire (R and F) neuron circuit, meticulously designed to emulate the oscillatory behavior of biological neurons. The fabricated circuit, measuring 35μm × 155μm, relies on a modified Lotka, Volterra (LV) oscillator to generate the neuron’s intrinsic oscillatory dynamics, with the resonant frequency precisely controlled by specific currents. Mathematical modeling, based on exponential relationships in sub-threshold circuits, allowed the team to define the circuit’s behavior using differential equations approximating the R and F dynamics. To detect and transmit signals, the circuit incorporates a spiking mechanism employing a comparator and cascade of inverters to generate a request (REQ) signal when the membrane voltage exceeds a defined threshold.

This REQ signal is central to an asynchronous handshake protocol, enabling event-based communication and minimizing energy consumption. The team integrated this handshake logic directly into the circuit, ensuring reliable signal transmission. The circuit’s synaptic connections, implemented as transistor-based voltage-controlled current generators, allow the neuron to process both spiking and continuous inputs. Comprehensive characterization across one hundred dies involved meticulous measurement of system parameters and overall power consumption, revealing the circuit’s suitability for large-scale integration. Experimental results demonstrate the circuit’s frequency selectivity and operational range, confirming its potential for real-time, low-power processing of signals like audio, electromyography (EMG), and electrocardiography (ECG).

CMOS Neuron Emulates Oscillatory Control Mechanisms

Scientists have achieved a breakthrough in bio-inspired computing with the fabrication and characterization of a Complementary Metal-Oxide-Semiconductor (CMOS) Resonate-and-Fire (R and F) neuron circuit, emulating the behavior of neurons responsible for oscillatory control within the central nervous system. This work demonstrates a pathway towards low-power, real-time edge processing by translating biological neural properties into hardware. The fabricated circuit, measuring 35μm × 155μm, integrates sub-threshold analog membrane potential dynamics with digital output spikes and asynchronous handshaking capabilities for compatibility with neuromorphic hardware platforms. Extensive testing across one hundred dies revealed the feasibility of large-scale integration within neuromorphic systems.

The core of the circuit, a modified Lotka, Volterra oscillator, exhibits dynamics modeled by differential equations describing the neuron’s resonant frequency and response to external input. Measurements confirm the circuit’s ability to selectively respond to specific frequencies, a crucial property for sensory processing and temporal pattern recognition. The team characterized the circuit in silico, measuring variability in system parameters and overall power consumption, demonstrating the circuit’s operational range and energy efficiency.