Abstract

Analog optical array processors are limited in dynamic range and accuracy to about 1000:1. To circumvent this limitation residue, binary, or other radix encoding schemes can be employed. Invariably, nonlinear logical operations must be performed on the input data streams in order to perform operations such as additions and multiplications on the encoded data samples. The standard approach used in electronic systems is to perform pairwise locally connected bit operations sequentially in a tree or array configuration of adder or multiplier primitives, each formed out of boolean switching devices. This approach is dictated because of the local wiring constraints inherent to planar technologies, and the availibility of only low dimensionality switching primitives (i.e. logic circuits which have very few inputs).

© 1985 Optical Society of America