Antenna Array Geometries to Reduce the Compute Load in Radio Telescopes

The proposed next-generation radio-telescope, the Square Kilometre Array (SKA), has ~3000 parabolic dish antennas and operates over many gigahertz of bandwidth. This results in a telescope with correlator requirements ~10 000 times more than any current telescope. This will be a challenge using traditional approaches. A method is proposed that will reduce the computational requirements for that part of the SKA where the antenna separation is not too great. This is done by deliberately introducing redundancy so that many of the outputs of the correlator measure the same data. We then show that the sum of these redundant correlations can be obtained efficiently by the use of fast convolution. The degree to which computational requirements are reduced is selectable. This translates directly into a power saving in the correlator and image processing.