Recreating neural oscillators in VLSI

Abstract

As a prelude to a VLSI implementation of a biologically-based neuronal locomotor network, the phenomena of reciprocal inhibition and recurrent cyclic inhibition were recreated in VLSI-based artificial neurons for parametric analysis of oscillatory range and stability. The artificial neurons used in this study are behaviorally comprehensive and highly configurable, allowing for a variety of transient and steady characteristics to be precisely and continuously adjustable. Circuit tests on both oscillatory phenomena indicate that reciprocal inhibition between two neurons requires a measure of synaptic dynamics, while the recurrent cyclic inhibitory prototypes did not. In addition, the cyclic prototypes demonstrated remarkable stability, even when the frequency of self-excitation of each component neuron varied over two orders of magnitude.<>