A extendable COTS multicomputer/CPU design for a MCAO control system

Many reconstructors, or Real Time Controllers (RTC), for mono-conjugate AO systems are currently operating(6,7,8), with many more about to be commissioned. The advent of faster and more efficient CPUs has permitted this task to be accomplished on a single processing element, for all but the highest order systems. However, the demands on the RTC increase by an order of magnitude or so in the case of a Multi-Conjugate AO (MCAO) system. Multiple Wavefront Sensors (WFS) and multiple deformable mirrors increase the complexity, processing load and data flow rates that the RTC must deal with. No currently available single processing unit is capable of meeting this demand and retain the advantages of a cost-effective, flexible system. Multiple processing units must be employed. We present in this paper a general architecture that addresses these issues. We present an analysis of the requirements of the Gemini South MCAO system on the RTC. This is followed by an algorithmic decomposition that simplifies the problem, lending itself to the use of commercially available multi-CPU single board computers. This is supported by the results of benchmark tests aimed at verifying the capabilities of one sample SBC. We conclude by presenting a description of the extendability of this architectural approach in the face of yet higher demands such as more mirrors, WFSs or complexity.