Energy-efficient Hybrid CMOS-NEMS LIF Neuron Circuit in 28 nm CMOS Process

摘要

Designing analog sub-threshold neuromorphic circuits in deep sub-microntechnologies e.g. 28 nm can be a daunting task due to the problem of excessiveleakage current. We propose novel energy-efficient hybrid CMOS-nanoelectro-mechanical switches (NEMS) Leaky Integrate and Fire (LIF) neuron andsynapse circuits and investigate the impact of NEM switches on the leakagepower and overall energy consumption. We analyze the performance ofbiologically-inspired neuron circuit in terms of leakage power consumption andpresent new energy-efficient neural circuits that operate with biologicallyplausible firing rates. Our results show the proposed CMOS-NEMS neuron circuitis, on average, 35% more energy-efficient than its CMOS counterpart with samecomplexity in 28 nm process. Moreover, we discuss how NEM switches can beutilized to further improve the scalability of mixed-signal neuromorphiccircuits.