A Comparative Analysis of Neuromorphic Neuron Circuits for Enhanced Power Efficiency and Spiking Frequency in 22nm CMOS Technology

Neuromorphic computing aims to replicate the brain's function and relies on the design of circuits inspired by biological processes. The neuron, a fundamental component of these architectures, must be optimized for operation speed and energy efficiency in order to approach the scale and complexity of the brain. Numerous designs have focused on optimizing performance metrics across a range of CMOS technologies, highlighting the necessity for a framework focusing on neuron circuit designs. Addressing this gap, our study presents the implementation of neuron models, inspired by the Leaky-Integrate-and-Fire, Axon-Hillock and Morris-Lecar mechanisms, specifically within the 22nm technology node. With this approach and through parameter sweeping, we provide a more nuanced comparison of power efficiency and spiking frequency among these neurons and insight into further design optimizations. This methodology yielded significant improvements achieving a peak spiking frequency of 80 MHz and minimizing energy consumption to 241 aJ per spike in the Axon-Hillock inspired design, marking the highest efficiency among the models tested.