Thoughts on polcal in crosscal and selfcal

[SpheMakh]

[bennahugo]

Ok @PeterKamphuis @paoloserra as well as meerkathi admins you should have read access to https://github.com/ska-sa/fleeting-pol - our boresight polarization calibration pipeline for the tied array beamformer

The relevant reports detailing MeerKAT leakage are the following:

• https://docs.google.com/document/d/1Goa7jnDFR94hnnboM47d9p0-O0At5cNts5vj5nvO2fk/edit?usp=sharing
• https://docs.google.com/document/d/19lSzq5Y1qYtNwgq3rWHfbxH0dHLGPiPwGxtXIDvOLpE/edit?usp=sharing
• https://www.overleaf.com/read/yqyvmznvytcf
• https://docs.google.com/document/d/1PCcp4I_QxtVnEWr0Siv3Zt3wev3k9C3BzbE0CUyYfjc/edit?usp=sharing

The following needs to be edited in the pipeline:

• crosscal: gain and bandpass calibration must apply P Jones on MODEL_DATA
• crosscal: gain and bandpass calibration must use the same reference antenna as X, D, K and Kcross Jones
• polarization worker needs to be added which:
  • apply 90 degree FEED rotation using python-casacore to FEEDS table
  • first averages the data to improve SNR on X(chan) and D(chan) solutions
  • then compute crosshand delay (dominated by digitizer timestamp offsets between X and Y, on the Fengine data). This is the slope of X Jones on strongly linearly polarized source (when source crosshands are not 0 due to parallactic angle rotation)
  • then compute crosshand phase, X Jones. DC first, then per channel X(chan)
  • then compute delays, DC first, then per channel D(chan)
• reports has to be migrated from fleetingpol which requires python-casacore as dependency
• additional html report on crosshand equality needs to be compiled as described in Polarization calibration testing procedures (company and stakeholder confidential) which I'm currently working on. I have the jupyter notebook ready - just needs to be incorporated in this pipeline.
• then switch fleeting-pol to have meerkathi as dependency and act as wrapper for computing the jones corrections for the tied array beamformer

[gigjozsa]

@SpheMakh Can this be closed? It seems that polcal is now being implemented in #1099 , which is not this?

[Athanaseus]

Adding @francescaLoi to the discussion.