To date, much of the "back-end" hardware used in radio astronomy, such as formatters, receivers, samplers and correlators, has been custom designed and built. It is clear that future developments will increasingly make use of commercial off-the-shelf (COTS) components. For example, the massive computing power required for LOFAR, albeit with limited bandwidth per telescope, is provided by standard computer chips in an overwhelmingly large (Blue-Gene) supercomputer architecture. While purely software-based correlators are increasingly being deployed, it is far from clear whether the increase in computing power predicted by Moore's law will be sufficient for future radio-astronomical instruments, such as the SKA. Power consumption and cooling issues (costs) are becoming increasingly important, and may to a high degree determine the direction of future developments, especially for very large arrays.
In this work-package a generic digital board (UniBoard) will be developed. This board will provide as much computing power as will fit on a reasonably sized PCB, in the form of a large number of state-of-the-art FPGAs, yielding up to 4 Tops (Tera-operations per second). The board will communicate with the external world using a large number of high-speed links. FPGAs are extensively used nowadays (e.g. for ALMA, EVLA, Korean VLBI Network correlators), but the attractive, and unique, aspect of the UniBoard is that it will intrinsically be a multi-purpose instrument, re-programmable for a wide variety of radio-astronomical applications, as well as be usable as a building block for larger systems. Combining unprecedented computing power and I/O capacity, it will enable new capabilities for a next generation of instrumentation, crucial for the future of the RadioNet facilities, as well as serve as a Peta-operations test bench for SKA development.
The UniBoard PI is A. Szomoru (JIVE)
szomoru [at] jive [dot] nl
More information can be found on the wiki page
