On neural networks for analog to digital conversion

Abstract

In this paper we compare analog to digital conversion (ADC) delay in Hopfield ADC and asymmetrical (lower triangular) neural network-based ADC due to Avitabile et al. (1991). It is shown that, although Hopfield ADC has extensive feedback, its behavior in asynchronous mode is similar to that of lower triangular ADC. It is also shown that any constant delay n-bit feedforward ADC must have an exponential number of neurons.