An autocontrol system based on neurons in a self-routing analogue converter

The paper is concerned with a self-routing neural network analog-to-digital converter. A neural network is a two-layer structure that has input and output layers, as well as two latent layers. The input and output layers play supplementary roles, with the first latent layer (multiplexers) providing input signal passage to the free portion of neurons in the second latent layer and the second latent layer (basic neurons) being responsible for analog-to-digital conversion. A basic neuron is a very simple single-digit analog-to-digital converter. Combining the necessary number of basic neurons provides the formation of an individual analog-to-digital converter of a given width for a specific input signal. Analog-to-digital conversion is performed by the bit-by-bit method. After completion of conversion, an individual analog-to-digital converter is deactivated and the basic neurons are returned to their original states. An autocontrol system of basic neurons is designed that provides a basis for determining their operability. The autocontrol system is a part of a common system of diagnostics of neurons in a self-routing analog-to-digital converter. This system generates a “readiness” flag of a basic neuron, which indicates to the system of signal self-routing that this neuron is in good condition and may be used to construct an individual analog-to-digital converter. © 2014, Allerton Press, Inc.