Caulfield Hj

Optical computing and the Fredkin gates

The use of optics to implement the Boolean logic functions traditionally used in conventional electronic computing is an active area of optical computing research. Many proposed optical implementations duplicate the configuration of electronic logic gates and hence may not optimally utilize the full benefits of optical techniques. We present here a new optical gate, the Fredkin gate, which may, in principle, be minimally dissipative (i.e., exhibit reversible logic) and whose response time may be limited in some implementations only by the duration of optical pulses (i.e., in the picosecond range). Such gates, which consist of three input and three output lines, can be programmed to produce a standard set of Boolean functions and appear well matched to the parallelism of optics. We present here a number of optical implementations of Fredkin gates and suggest ways of composing their interconnections to achieve combinatorial logic, circulating memories and generalized interconnects.

Dynamics of hologram recording in DuPont photopolymer

Several dynamical aspects of the DuPont photopolymer film HRF-150-38 for holographic storage are described. We study temporal aspects of exposure, exposure time, processing situations, and storage effect. The quantities studied are diffraction efficiencies, thickness changes, and Bragg angle. The experimental results are performed with an argon-ion laser at 514.5 nm.

Parallel N 4 weighted optical interconnections

While full optical interconnects of an NxN input signal array to an N×N output signal array through N weighted interconnects is an important goal for optical artificial neural systems (ANSs), methods for doing this are rare. Goodman et αl. fully connected an N × 1 array to a 1 × N array. Sawchuk has suggested a fixed N interconnection method using replicated holograms for optical cellular logic. This works in principle but has extreme space-bandwidth requirements for large N. Sawchuk has described a 3-D dynamic interconnection network for interconnecting 2-D N × N arrays in parallel computing, but this network does not have arbitrarily variable weights. I hope to show a simple optical N interconnection method which uses only one noncritical lens, an N × N reflective spatial light modulator and a beam splitter as components. It is convenient to think of the N × N input array as a matrix A with components αkl· Likewise the output is an N × N array B with components bij. These are interconnected by a 4-D tensor T, i.e.,

Massive holographic interconnection networks and their limitations

Fundamental and practical limitations to be encountered in the implementation of massive free space optical interconnects are discussed in detail, and some improved architectures are proposed. The long term optimum design uses currently unavailable large arrays of laser diodes. An interim solution, using available spatial light modulators, is shown to be capable of storing ~10(10) bits of information and performing ~10(11) interconnections/s.

Wave particle duality considerations in optical computing

The wave particle duality inherent in the propagation of light or particles can be exploited for energy efficient computing leading to energy requirement per calculation below kT. Although several reversible computers with similar characteristics were proposed in the past, only optical implementations can be made with the present technology.

Optical learning (inference) machines

Two different approaches to decision aids are explored. In one, forced binary decisions are made. In the other, a set of a posteriori probabilities is calculated. In both cases fast inferences can be calculated by optical matrix operations.

High-efficiency rapidly programmable optical interconnections.

An array of optical Fredkin gates implemented by optically controlled waveguide couplers is shown to constitute a very efficient and versatile optical interconnection network with parallel addressing capabilities. The characteristics of the array are analyzed using linear algebra to indicate interconnect programming procedures. In terms of SNR this network is estimated to be comparable with previously proposed architectures. However, from many other aspects (light transmission efficiency, number of switching elements, speed, and fault tolerance) it has significant advantages.

Design for a massive all-optical bidirectional associative memory: the big BAM.

An optical bidirectional associative memory offering a potential of operating up to 10(6) neurons with 10(12) interconnections is described. Except possibly for input and output all operations are optical and parallel.

Flexible two-way optical interconnections in layered computers

Realisation de communications bidirectionnelles a grande vitesse adaptative flexible entre des couches optoelectroniques et une region de commande centrale a travers une colonne de lumiere. Analogie avec le systeme nerveux vertebral

Residue arithmetic processing utilizing optical Fredkin gate arrays

A cascadable residue arithmetic processor based on optical Fredkin gate arrays and page-oriented holographic memories is introduced. The implementations of residue functions and operations by this processor are described. Analytic expressions are derived for the number of holograms and waveguide channels required for the implementation of residue addition and multiplication. The practical cases of 16-bit addition and multiplication are analyzed as specific examples. It is shown that, using the proposed architecture, these operations can be implemented with state-of-the-art technologies in holography and integrated optics.