Hybrid Digital-Optical Processors: A Performance Assessment

Hybrid digital-optical processors are designed to use the best features of their components; the speed of an analog optical processor and the accuracy of the digital processor; to overcome the accuracy limitations of analog systems and the relatively slower speed of digital systems. The application of hybrid processors to the solution of systems of linear algebraic equations using the bimodal algorithm is shown by analysis to be potentially more rapid than exclusively digital processing only when the range of significant system eigenvalues is at least two times less than the linear dynamic range of the analog optical processor provided the proper preconditioning is employed. Thus, while performance is significantly improved by preconditioning and by maximizing the accuracy of the analog optical processor, only a limited number of problems - those with a small range of eigenvalues - can be efficiently solved to high accuracy using the hybrid processor. This precludes its use as a replacement for a digital processor. The hybrid processor may find applications as an adjunct to analog optical processor. However, by design the hybrid processor has increased complexity and reduced speed with respect to the simple analog optical processor. These sacrifices may not be justified in real-time applications to which an optically-based processor would be uniquely suited.