A programmable multi-function optoelectronic cellular neural network

Made by simply forming two Schottky contacts on an undoped semiconductor layer, the metal-semiconductor-metal photodiode (MSMPD) has all the desirable attributes of a practical photodetector which include low dark current, high-speed response and the need for little internal gain. The output current of the MSMPD is nearly linearly dependent on the bias voltage applied on the detector´s two contacts and on the light power detected. We propose a novel cellular neural net (CNN) architecture that combines the MSMPDs and the current-mode electronic CNN to form a programmable multifunction optoelectronic cellular neural net (OECNN). Applying the bi-symmetry attribute of the MSMPD, we can map the control matrix onto a fixed size MSMPD array that requires few control voltages. By adjusting these control voltages, we can easily and quickly obtain a new control matrix in the OECNN thus providing the programmability and the multifunctionality. Designed using the current-mode method, the electronic neurons are directly matched to the MSMPD array such that the interface between them is very easy to implement and the components used in the electronic circuits are very economical