Stokes V calibration for slow visibilities

[binchensun]

Commissioning for Stokes V calibration of slow visibilities.

[surajit4444mondal]

I am attaching my slides. Please note that these are extremely fluid and these slides just tracks the work done and how it reached the current state, whatever that is. work_notes_march12_2024.pdf ovro-lwa_polarisation_calibration_feb19_2024.pdf ovro-lwa_polarisation_calibration_jan16_2024.pdf

[surajit4444mondal]

Multiple functionalities implemented in #58

[surajit4444mondal]

A summary of what has been done and needs to be done is written in the attached presentation. The red lines shows parts which is not implemented and blue ones shows parts which has been implemented. The code only works for unpolarised sources. Also I think running quartical after the final image plane leakage correction, can have issues due to incorrecting modeliing of the sky. Care should be taken when using the quartical features.

work_notes_march18_2024.pdf

[surajit4444mondal]

crosshand_phase_Ruby_Bryne_derivation.pdf

[surajit4444mondal]

There has been a host of improvements in this regard, which I go over here:

1. Most of the steps done in my previous efforts were concerned with image based corrections which does not take into account any antenna dependent effects. Hence these methods can also be tested in the beam data, where testing is much easier and less time consuming.
2. The steps I am currently following for the beam calibration are:

• First do the crosshand phase calibration by minimising the absolute value of Stokes U. The only free parameter is a rotation angle between U and V. I have to also take into account some leakage, as the observed data also have leakage from Stokes I, which cannot be minimised in this way, and some approximate leakage subtraction is sufficient at this step.
• Next calculate the fractional polarisation of the data for all Stokes (Q,U,V), over the primary beam contribution, and subtract a low order polynomial fitted to it.
• The residual gives the final calibrated data
  3. The leakage correction steps assumes that `true' polarisation fraction should be zero for all Stokes, which is valid, strictly speaking, only in case of quiet Sun. Thus the data for determining the beam parameters should be a predominantly quiet Sun, although some bursts are fine, as they will not contribute significantly to the parameters.
  4. I expect the fractional polarisation due to leakages in the signal chain to be constant over days and hence interpret that the linear term in the fit is arising primarily because of primary beam effects. If this is true, the same parameters should also hold for nearby days at similar time. I have also verified that this expectation is true.

I am attaching a zip file which 2 files: a jupyter notebook and a html of the same. These have these steps and a bit of explanations. beam_polarisation_calibration.zip

It is also quite comforting to see that the crosshand theta calculate also shows a linear relationship with frequency. I am attaching the crosshand theta variation with frequency here. Note that the crosshand theta calculate here is just theta, which is equal to 2\pi \nu \Delta t

[surajit4444mondal]

The previous steps were done using the Woody beam. I implemented the code for reading Nivedita's beam in our package and did the same analysis of beam calibration with both the beams. I find that Nivedita's beams also leave the linear trend shown earlier, but the mean leakages after primary beam has become smaller. So my conclusion is while Nivedita's beam is better, they still need some improvement for taking care of all the systematic issues present. I am attaching 2 plots, one with Nivedita's beam and one with Woody beam, where the crosshand phase has been calibrated out, following the method mentioned earlier.

[surajit4444mondal]

Final test is on the imaging data. I apply the polcal parameters on the imaging data, after bandpass corrections. I show 2 images with 4 panels. Beginning from top left and in clock wise direction, the panels are I,Q/I, V/I and U/I respectively. The line plot shows a cut along a line. The solid line is the original and the dashed line is after the polcal. The improvement is evident.

[surajit4444mondal]

Same for another time. See that in these plots, Stokes V and Stokes U both had a strong compact source. After polcal, the strong source in source U is gone.

[surajit4444mondal]

New developments and tutorials summarised in these documents. Contains Jupyter notebooks about running using the software, showcases the issue of primary beam errors etc. beam_polcal_tutorial.zip