Research on Architecture of Optical Computer

本文研究了一种具有大容量信息处理能力的光计算机的结构，并阐明了它的研制方案。在我们的研究中，取得了以下结果.建立了一个光-数混合并行阵列逻辑系统实验系统。在该系统中，液晶显示器和电荷耦合器件电视摄像机被用作光学和电子系统之间的接口设备。用个人计算机系统实现反馈回路。界面显示的对比度是影响系统处理速度的重要因素。采用振镜结构，可以提高系统的灵活性和处理能力.研究了OPALS的并行编程技术。在这项研究中，已经开发了一种方法，用于空间变化的操作使用光学阵列逻辑。光学阵列逻辑是OPALS的处理基础。在光阵列逻辑程序设计中，研究了适用于各种数据处理范围的模式逻辑的概念，并验证了它的能力。本文提出了一种用于数组逻辑的程序设计语言，并对运算核的乘、反、簇等概念进行了研究。利用这些技术和概念，开发了各种并行处理程序，并通过OPALS实验系统进行了验证.提出了一种基于神经网络的光学处理系统结构。研究了一种具有正交变换的联想记忆系统。还考虑了利用神经网络的恢复能力的高可靠性系统。在这项研究中，多层系统的有用性是明确的，即使是光学系统。在OPALS上实现神经网络仍是未来的工作。

An architecture of an optical computer capable of large capacity information processing has been investigated and a scheme of its decelopment has been clarified. In our research the following results have been obtained.1. An experimental system of the opto-digital hybrid OPALS (optical parallel array logic system) has been constructed. In the system a liquid crystal display and a charge coupled device TV camera are used as interface devices between optical and electronic systems. Feed back loop is implemented with a personal computer system. Contrast of the display for interface is an important factor in processsing speed of the system. With a galvanometer mirror, flexibility and capability of processing of the system can be increased.2. Techniques of parallel programming for OPALS have been investigated. In this study, a method for space-variant operation has been developed using optical array logic. Optical array logic is the processing basis of the OPALS. In programming of optical array logic, a concept of pattern logic has been studied for various range of data processing and its capability has been verified. A programming language for potical array logic has been developed and some useful concepts such as multiplication and negation of operation kernel and cluster have been satudied. Using these techniques and concepts various programs of parallel processing are developed and verified by the experimental system of the OPALS.3. An architecture based on neural network has been investigated as a a new type of optical processing system. an associative memory system with orthogonal transform has been studied. A high reliable system utilizing restoration capability of neural network also has been considered. In this study usefulness of a multi-layer system is clarifid even for optical systems. Implementation of neural network on the OPALS remains as a future work.