A new generation of opto-electronic signal processors has been developing during the past several years. These processors are designed to perform algebraic operations like matrix-vector and matrix-matrix multiplication. In this paper key architectural developments are reviewed and major algorithmic methods and problems are discussed.
Methods of performing elementary computational operations using integrated optics technology are reviewed. It is then shown how these operations can be combined to form a number of matrix-vector and matrix-matrix multipliers. These include a compact modification of the engagement architecture which allows a more efficient multiplication of two N x N matrices than possible with previous engagement processors.
The parallelism and non-interfering propagation of optics offers a way around some of the communication problems which limit the computational throughput of current computers such as interconnection bandwidth, clock skew, and the Von Neumann bottleneck. Optical bistability offers the possibility of optical logic gates with speed and power requirements comparable with traditional electronic approaches. A pipelined architecture is used to illustrate how the parallelism and communication abilities of optics can be exploited as well as how the optical requirements can be simplified.
Optical technique for data storage has been under intense development for the past twenty years. This technology is finally reaching maturity, and its unique features are appreciated by the video, audio, image and digital data recording industry. This paper reviews the status of the key components and the performance of representative devices, and forecasts the contributions of optical memory technology at present and in the future.
This paper briefly reviews analog and discrete optical computing systems; it then concentrates on binary digital optical computers in which photons are the primary information medium. A critical review of various techniques for optically implementing combinatorial and sequential logic by individual devices and arrays of devices are given. The problems of communications, interconnections and input-output among logic devices (gates) , among arrays of devices (chips) and among processors are discussed. A particular optical system architecture that uses a computer-generated hologram to interconnect a planar array of logic gates in a third dimension is presented. Some of these architectures offer the potential of parallel processing, including non von Neumann digital computers. Some existing limitations and needs of optical logic devices and computing systems are discussed.
As the complexity of microelectronic circuits increases, performance becomes more and more limited by interconnections. Continued scaling and packing lead to a dominance of interconnect delays over gate delays. This paper explores the potential of optical inter-connections as a mean for alleviating such limitations. Various optical approaches to the problem are discussed, including the use of guided waves (integrated optics and fiber optics) and free space propagation (simple broadcast and imaging interconnections). The utility of optics is influenced by the nature of the algorithms that are being carried out in computations. Certain algorithms make far greater demands on interconnections than do others. Clock distribution is a specific application where optics can make an immediate contribution. Data interconnections are more demanding, and require the development of hybrid Si/GaAs devices and/or heteroepitaxial structures containing both Si and GaAs layers. The possibilities for future developments in this area are discussed.
Vision systems are becoming increasingly important sources of information for automated inspection and assembly tasks. Spatial information along directions transverse to the viewing axis is easily obtainable with imaging systems and one-or two-dimensional detectors. A challenge for optics is to develop fast, accurate range sensors. This paper reviews range sensing techniques. Four basic categories of range sensing techniques are discussed: geometric techniques, time-of-flight techniques, interferometric techniques and diffraction techniques. The basic principles are elucidated and general comparisons are made between the groups. Representative examples are given of many different approaches.
Several promising applications of optical signal processing are reviewed: radar warning receivers/ESM, synthetic aperture radar, large time-bandwidth radar, and sensor-array processing. For these applications an attempt is made to identify the need for optical signal processing and to indicate how optical processors can be configured to meet these needs. The ESM application is also used to indicate the types of developments needed for practical implementations.
Optical pattern recognition has always offered the advantages of high speed and parallel processing with the basic linear system operations of Fourier transformation and correlation being its hallmark. Recent years have considerably broadened this repertoire of operations to include optically-generated features, distortion-invariant correlators, space-variant processors, optically-produced local operators, increased use of digital post-processing al-gorithms, more extensive tests on large multi-class image databases, various real-time architectures, etc.
At present, digital electronics are the most often used method for the solution of partial differential equations (PDE's) and integral equations (IE's). This method has the advantages of high accuracy and programmability. The disadvantages are that total time to compute an entire solution may be long, iterative methods are sometimes needed (further reducing speed), and the solutions are computed only at discrete points within the domain.
Devices for optical processing and computing systems are discussed with emphasis on the materials requirements imposed by functional constraints. Generalized optical processing and computing systems are described in order to identify principal categories of requisite components for complete system implementation. Three principal device categories are selected for analysis in some detail: spatial light modulators, volume holographic optical elements, and bistable optical devices. The implications for optical processing and computing systems of the identified materials requirements for these device categories are described, and directions for future research are proposed.