The Artificial Neuron: Built From Nanosheet Transistors to Achieve Ultra Low Power Consumption

Abstract

The performance of semiconductors has greatly improved due to the miniaturization of the transistor. To shrink the size of a transistor, the channel length must be decreased. Short-channel effects become noticeable as the gate length is reduced. Short-channel effects are observable at dimensions below 7 nm as a consequence of the shorter gate length. However, nanosheet structures have been proposed to replace FinFET and nanowire transistors at the 7 nm and 3 nm technological nodes. The purpose of this research is to construct an artificial neuron based on nanosheet transistors and evaluate its spiking behaviour and power consumption. We are building a regular Axon Hillock and our proposed neuron model using GPDK 45 nm and nanosheet transistor 20 nm technology. The simulation results clearly show that the nanosheet transistor consumes less power than the GPDK 45 nm transistors. Furthermore, we verified the validity of our suggested neuron model by performing Monte Carlo simulation, PVT analysis, AC response, noise response and layout. Additionally, the proposed neuron was tested with a variety of input currents, including pulse current, ramp current, sinusoidal current, and arbitrary current, and their associated spike patterns were recorded. According to our research, nanosheet-based Axon Hillock consumes 156 fW, and our proposed neuron consumes 1.9 pW of power.