Create Standard Datasets for SDP Pipeline Testing

Develop "standard" datasets that can be used for testing imaging solve and predict cycles, that might be considered representative for SKA1 Mid or Low. These datasets would be used to test performance and especially accuracy of pipeline implementations.

As an initial suggestion it would be convenient to have three different sized images:

• Small: 10s of sources in an 2048 by 2048 image.
• Medium: 100s of sources in an 8192 by 8192 image.
• Large: 500+ sources in an 16384 by 16384 image.

Having the images a power of two is handy but not essential, and its good if its taken from a typical Low or Mid array configuration. The number of sources is not critical and potentially the small/medium images could be taken from the same field of view of the sky as the large image.

For each it would be good to have:

• The list of visibilities (up to around 30-50 million for the large image) and their u,v,w coordinates in a similar order to that produced by a correlator (so frequencies separated)
• The field of view or cellsize, frequency or frequencies and any other configurations used
• The corresponding image ideally produced by an inverse DFT, including w-term, for the large image the inverse DFT will require a fair amount of compute so may need to be restricted to a central region of the image and produce the outer regions by gridding (Note: NZ_APP has a GPU-accelerated iDFT on the SKAO github which can be used).
• The sky catalogue (list of sources) that was used to produce the visibilities

As an initial specification things could be kept simple so not worrying about eg sources outside the primary beam.