Spice-compatible Verilog-ams Model For Inferior Olive Neuron

Abstract

The field of brain modeling is a very promising step towards the understanding of human brain functions and the discovery of new treatments for brain diseases. Appearing research challenges are being tackled by several scientific approaches including fMRI, Neuroimaging and High Performance Computing. Depsite the numerous scientific advances towards the demystification of the human brain, there is always a constant need for higher optimization and standardization of the simulation procedures. Neuromorphic computing is a very interesting concept and a major step towards the better understanding of the brain, as it establishes the connection between designed neural systems and actual biological nervous systems. This analogy between electrical circuits and neural functionality is explored using traditional EDA tools that are prominent in the IC design domain.The purpose of this thesis is to provide a detailed transient response of a inferior olivary nuclei (InfOli) model as a single neuron and as part of multi-neuron interconnection network, through a standard, multi-physics simulator for integrated circuits. The software of our choice is Spectre by Cadence. Starting from a Matlab implementation a compatible Verilog-A model is created, which is run in the Cadence Spectre simulator. The results from Spectre and the MATLAB simulator are compared to derive figures of merit as different input parameters and inter-connection schemes are explored and finally the accuracy limitations of the SPICE-like commercial tool are quantified.