Suprathreshold stochastic resonance and signal-to-noise ratio improvement in arrays of comparators

We consider parallel arrays of threshold comparators similar to those modelling flash analog-to-digital converters, or digital sonar arrays, or networks of sensory neurons. Such arrays are used here for the transmission of a noisy periodic input signal, with a performance assessed by a signal-to-noise ratio in the frequency domain. We show that independent noises injected on the comparators can improve the performance of the array, with a nonzero optimal amount of noise that maximizes the output signalto-noise ratio. This represents, for periodic signals, a new instance of the recently introduced phenomenon of suprathreshold stochastic resonance. We also study the capability of the arrays of comparators to realize an input–output enhancement of the signal-to-noise ratio, and show the existence of conditions where the arrays act as signal-to-noise ratio amplifiers, with a gain maximized by a nonzero level of the injected noises.  2004 Elsevier B.V. All rights reserved.