Putting some thoughts down here on the 'proper' approach to common resolution.
Right now, in the imager stage, we convolve an entire field to a common resolution across beams and frequency.
In principle, we actually only need to convolve each beam along the frequency axis. Then, when we go to LINMOS each cubelet together, we can make a last minute check and convolution before images are co-added. This way, each cubelet would be kept at the best possible resolution.
For top-level purposes, one could then decide to convolve all (or some subset) of the cubelets to whatever resolution and reproduce a catalogue from those convolved data. We just need to ensure the right check happens in the LINMOS stage, and that the pipeline picks up the correct PSF information when making a catalogue.
Putting some thoughts down here on the 'proper' approach to common resolution.
Right now, in the imager stage, we convolve an entire field to a common resolution across beams and frequency.
In principle, we actually only need to convolve each beam along the frequency axis. Then, when we go to LINMOS each cubelet together, we can make a last minute check and convolution before images are co-added. This way, each cubelet would be kept at the best possible resolution.
For top-level purposes, one could then decide to convolve all (or some subset) of the cubelets to whatever resolution and reproduce a catalogue from those convolved data. We just need to ensure the right check happens in the LINMOS stage, and that the pipeline picks up the correct PSF information when making a catalogue.