Changes/invert
invert Item Complexity Research Milestone Time Theoretical Noise easy moderate 6 5d Visibility Weights easy moderate 6 5d Primary Beam moderate some 10 2-5d MFS moderate some 11 15d
Theoretical Noise
Use Tsys spectrum for computing theoretical noise in image.
Visibility Weights
Use Tsys spectrum for determining weights in image formation.
Primary Beam
Check and, if necessary, correct usage of primary beam models in invert. Make sure variation with frequency is taken into account for mosaic case. Amount of time needed will increase if current code is not handling this correctly. Should check linmos, mossdi and mosmem as well, hopefully they share the primary beam code...
Multi Frequency Synthesis
Add a spectral curvature term (2nd order term) to MFS imaging. At present the average flux and a spectral index are handled. For 2-4 GHz bandwidth we may require higher order terms to model the sources in a typical field, especially at the lower frequencies. Some research is required to determine whether this is indeed a limiting factor in high dynamic range imaging and how much improvement can be expected.
Polarimetry
Improving polarimetry requires a synthesised beam that scales with frequency; having a beam cube with 1000 channels in it is impractical, according to the practicioners. This change would be required in restor as well.
Questions
- If we split the UV data into dozens of files all separately calibrated, what is the max. number of UV data files invert will accept?
- We could just glue all the pieces together for spectral line data, but what about the continuum multifrequency synthesis case?
- Look at the code to find out: does invert process channels independently with regard to multi frequency synthesis?
- A useful test seems to be: generate some data with strong spectral curvature. Image with MFS and compare to an image where the data are inverted channel by channel and summed in the image plane.
