The influence of photo-detector non-linearity on the multi-functionality of an opto-electronic cellular neural network

The multi-functionality of an opto-electronic cellular neural network (OECNN) is investigated in this paper. Investigations are carried out to see if the network can still perform as an image edge detector, noise remover, or shadow creator when its metal-semiconductor-metal photodiodes (MSMPDs) have a non-linear response and device variations (i.e. variation in sensitivity from one device to another). From HSPICE simulations, it was found that the edge detector is not affected by the MSMPD's non-linearity and tolerates 20% device variation. It was also found that by adjusting the cell's behaviour, the impact of the non-linearity on the noise remover is alleviated to an acceptable level and that the network can perform as a shadow creator and tolerates 40% device variation. This discovery means that the restriction on the fabrication method of the MSMPD may be significantly reduced, if the programmability of the OECNN can be limited.