2024/12/17 by Seyedmohamadjavad Motaman, Motaman, Seyedmohamadjavad, Safura Sharifi +3
Engineering · Computer Science · #Advanced Memory and Neural Computing #Neural Networks and Reservoir Computing #Advancements in Semiconductor Devices and Circuit Design
paper · pdf · doi:10.48550/arxiv.2412.12443
Enhancing power efficiency and performance in neuromorphic computing systems is critical for next-generation artificial intelligence applications. We propose the Nanoscale Side-contacted Field Effect Diode (S-FED), a novel solution that significantly lowers power usage and improves circuit speed, facilitating efficient neuron circuit design. Our innovative integrate-and-fire (IF) neuron model demonstrates exceptional performance metrics: 44 nW power consumption (85% lower than current designs), 0.964 fJ energy per spike (36% improvement over state-of-the-art), and 20 MHz spiking frequency. The architecture exhibits robust stability across process-voltage-temperature (PVT) variations, maintaining consistent performance with less than 7% spike amplitude variation for channel lengths from 7.5nm to 15nm, supply voltages from 0.8V to 1.2V, and temperatures from -40°C to 120°C. The model features tunable thresholds from 0.8V to 1.4V and reliable operation across input spike pulse widths from 0.5 ns to 2 ns. This significant advancement in neuromorphic hardware paves the way for more efficient brain-inspired computing systems.