Modern radio astronomy telescopes the world over require digital back-ends. The complexity of these systems depends on many site-specific factors, including the number of antennas, beams and frequency channels and the bandwidth to be processed. With the increasing popularity for ever larger interferometric arrays, the processing requirements for these back-ends have increased significantly. While the techniques for building these back-ends are well understood, every installation typically still takes many years to develop as the instruments use highly specialised, custom hardware in order to cope with the demanding engineering requirements.
Modern technology has enabled reprogrammable FPGA-based processing boards, together with packet-based switching techniques, to perform all the digital signal processing requirements of a modern radio telescope array. The various instruments used by radio telescopes are functionally very different, but the component operations remain remarkably similar and many share core functionalities. Generic processing platforms are thus able to share signal processing libraries and can acquire different personalities to perform different functions simply by reprogramming them and rerouting the data appropriately. Furthermore, Ethernet-based packet-switched networks are highly exible and scalable, enabling the same instrument design to be scaled to larger installations simply by adding additional processing nodes and larger network switches. The ability of a packetised network to transfer data to arbitrary processing nodes, along with these nodes’ reconfigurability, allows for unrestrained partitioning of designs and resource allocation.
This thesis describes the design and construction of the first working radio astronomy imaging instrument hosted on Ethernet- interconnected reprogrammable FPGA hardware. I attempt to establish an optimal packetised architecture for the most popular instruments with particular attention to the core array functions of correlation and beamforming. Emphasis is placed on requirements for South Africa’s MeerKAT array. A demonstration system is constructed and deployed on the KAT-7 array, MeerKAT’s prototype.
This research promises reduced instrument development time, lower costs, improved reliability and closer collaboration between telescope design teams.