Dr. Dan Werthimer, University of California, Berkeley
High Performance Radio Astronomy Signal Processing
The upcoming Arecibo Focal Plane Array and large radio telescope arrays such as the Allen Telescope Array (ATA), and the square kilometer array (SKA) require peta-ops of real time signal processing. Such computational requirements are far beyond the capabilities of general purpose computing clusters (eg: Beowolf clusters) or supercomputers. Traditionally instrumentation for radio telescope arrays has been built from highly specialized custom chips, many different boards and racks; these instruments take five to ten years to design and debug; they are very expensive to develop, inflexible, hard to maintain, and are usually out out of date before they are working well.
We have developed a modular scalable high performance FPGA-based computing platform and software design methodology that target a range of real-time signal processing applications, including beamforming, wide band spectroscopy, imaging correlation, VLBI, and RFI mitigation. The modular system architechture provides orders of magnitude reduction in cost and design time, and closely tracks state-of-the-art fabrication by FPGA vendors. Users have the freedom to choose the appropriate number of computational modules needed to tackle the applications at hand, and rapidly reconfigure the system for different applications.