ABSTRACT
This chapter presents incoherent fiber-optic systolic array processors (FOSAPs), which employ a digital-multiplication-by-analog-convolution (DMAC) algorithm and the extension of the DMAC algorithm, are proposed for real-valued digital matrix computations. It describes the important role of optics in optical computing and a variety of existing optical architectures using the DMAC algorithm. The chapter also presents mathematical formulations of the DMAC algorithm and the two’s complement binary arithmetic. It describes three elemental optical signal processors, namely, an optical splitter, an optical combiner, and a binary programmable incoherent fiber-optic transversal filter. The chapter analyses the incoherent FOSAP architectures for computation of positive-valued digital matrix–vector, matrix–matrix and triple-matrix products by using the DMAC, HO-DMAC and S-DMAC algorithms. It focuses on the performance of the FOSAP architecture using non-binary data. The chapter considers the performance of the positive-valued high-order FOSAP matrix multiplier is compared with that of its digital electronic counterpart.
