SCIOS - Scottish Collaborative Initiative in Optical Sciences

Project Objectives

(i) Demonstration of parallel connections to Si-IC's based on assembly with III-V semiconductor optoelectronic chips containing arrays of detectors, and modulators or emitters. The target specification is 4k optical channels (64×64) each operating at 250Mbits^-1. This corresponds to an aggregate I/O of 1Tbits^-1. Both InGaAs detector/modulator arrays and VCSELs will be investigated as the optical interface devices.

(ii) Assessment of a number of free-space optical interconnect architectures, suitable for linking high performance processor/memory modules or other electronic nodes with particular attention to scalability.

(iii) Construction of an experimental demonstrator system suitable for showing the potential of such free-space links and acting as the basis for advancing the required optical and packaging hardware. This will take the form of a single-stage optoelectronic 64 x 64 crossbar switch, capable of fully non-blocking, self-routing of >200Mbits^-1 signals.

Project Status: Completed 1995.

Rationale

As feature sizes on silicon IC's become smaller and the performance of individual chips becomes more powerful, so the problem of providing a proportional increase in external input/output capacity becomes harder. Physical limits on the numbers, and bandwidth of, electrical pin connections create a substantial bottleneck. This project aims to develop and exploit three-dimensional optical interconnects based on multiple free-space connections distributed across the full area of the chip in array format.

The proposed 3-D free-space approach is particularly attractive as, by permitting the construction of high-throughput reconfigurable interconnects (such as the crossbar switch), it could rapidly open up the construction of powerful information processing machines based on highly interconnected processor and memory blocks. This would include massively parallel systems and advanced shared/distributed memory machines. One could therefore anticipate significant impact on all those applications currently looking to exploit highly-parallel computing systems – an increasing domain, encompassing tasks such as engineering simulations, computational science applications, real-time image processing and game/entertainment systems. Distributed sensing/control systems in which large amounts of data need to be communicated from node to node with minimal delay will be another important application.

The project will explore the long-term potential of this approach, which offers the prospect of creating wide-band links to advanced Si-IC's with no limitation on connection distance and with aggregate bandwidths in excess of 10Tbits^-1.

Last Modified 04/04/03 11:04:16.