Influence of capacitor variability on the electronic spiking neurons

Currently, the neuromorphic hardware receives a growing interest in modelling the brain behaviour due to the parallel operation of the neurons and very low power consumption. However, the main difficulties in implementation of electronic synapses that mimic rigorously the physiology of the natural neurons represents the non-volatile weights storage and their adjustment in real time following the biological plasticity rules. Among the IC implementable methods for the weights storage, the capacitors represent the cheapest solution due to the simplicity of the circuits for weight adjustment. Considering that component variability represents a disadvantage of nanoelectronics, in this work we evaluate the tolerance to the parameter changes of an electronic neuron that uses capacitors for the weight storage. The simulation results show that despite the fact that the neuron operation stability is not affected by the typical variability in ±10% range of the capacitors, the regulatory performance of small SNN is significantly changed.