ALMA Polarization data
Hiroshi's polarization calibration guide (from SCIREQ-580): Polarization_Calibration_Guide.pdf
Initial overview and calibration
Before starting, you may wish to run au.parallacticAngleForField. It has two arguments, the MS and the field ID. Run this on each session (i.e. ASDMs taken consequectively, in the same Array), and sum the parallactic angle coverage. Anything >70deg is probably fine. If the session has <70deg please consult with a DRM.
Create and run the regular calibration script, but add refantmode='strict' to the calls to gaincal to prevent gaincal flipping the refant.
Flag bad data and check that the refant is the same in each execution and is good throughout each execution. If the phase cal is faint and the phase_int solutions on the polarization calibrator are noisy, use the phase_inf instead of the phase_int table for applying the calibrations in the last step. Note that because the data are taken in sessions you may need to use the bandpass table from the first session to calibrate the others. Also the flux calibration will only be done in the first ASDM, so use setjy fix the flux of the phase cal in subsequent ASDMs.
Remember that plotms will plot all correlations by default, the XY and YX correlations usually have very low signal even for bright calibrators and should have very low amplitudes.
Look at the project in the OT to figure out the calibration scheme for polarization. Typically there is one leakage calibrator observed multiple times over the session (to build up a parallactic angle coverage).
Once you have the calibrated measurement sets, you need to run the scriptForPolCalibration.py (example attached, from SCIREQ-580). Make sure you use the version with the FIX_PRTSPR-20069.
This first concatenates the individual executions.
The polarization calibrator scans should span a wide range of parallactic angle (at least 50deg). To check the parallactic angle coverage run plotms with xaxis='parang' on the ms from the concatenated session (see below).
The script should work out of the box after chaning the input ASDMs, refants and intents at the top, though may need some editing if e.g. the polarization and phase calibrators are the same. If there are problems I find the parallel hands gain ratio plots to be very useful, e.g.:
These provide an independent check on the polarisation calibration derived from the cross-hands. If the gain table for the calibrated data does not show a parallel hands ratio that is constant very near unity there is probably something wrong (e.g. one dataset from June 2015 needed 2 spw and 8 antennas flagged before these plots looked right!).
The standard NA imaging script does not accommodate polarization, so instead I use one first written by the Italian ARC node. Both it, and the scriptForImagingPrep that should be run first are attached below. Note that the new imaging script ( scriptForImagingPol_NA_20180326.py.txt
) should be used.
In step 3, the target is imaged. The default is set to non-interactive cleaning. Change this to interactive cleaning if you want to change the mask or threshold.
You may also need to modify the polithresh parameter in when creating the POLA image.
QA2 guidelines are still being developed, here's what I use:
1) check the image of the polarization calibrator, it should correspond to the model derived during the calibration process. In particular the V signal should be consistent with noise.
2) Check the parallel hands RMS plot concat.ms.GainRatiosPol.png. This should show a flat RMS as a function of scan number with PolCal
(blue points), significantly lower than the No Polcal points (in red)
3) Check that the V polarization of the phase calibrator (if different from the polarization calibrator) is consistent with noise (or is less than a few tenths of a percent).
Spectral line and circular polarization
Spectral line and circular polarization have no special additional requirements on calibration or imaging (though the data reducer may need to make a cube for spectral line reduction if the QA2 criteria assume spectral line imaging). Circular polarization products are already produced by the default imaging script.
The QA criteria for spectral line (FDM) and circular (V) polarization are less strict than for linear. From the Cycle 6 technical handbook, Section 8.7:
"The expected minimum detectable degree of linear polarization, defined as three times the systematic calibration
uncertainty, is 0.1% (1%) for compact sources (i.e., within the inner 1/3 of the primary beam FWHM) and
0.3% (3%) for extended sources for TDM (FDM) observations, respectively. The minimum detectable degree of
circular polarization is 1.8% of the peak flux for both TDM and FDM observations. Note that the systematic
calibration uncertainty can degrade by a factor of ⇡2 depending on the choice of calibrator, parallactic angle
coverage, etc. With the current calibration scheme, linear polarization imaging of a compact source on-axis in
TDM mode is feasible at the level of 0.1% (3 sigma) fractional polarization for the very brightest calibrators,
and 0.2% (3 sigma) level for a typical observation,"
In principle FDM should be no less accurate than TDM, but the signal-to-noise ratio of the calibration solutions is lower. The limitation on the accuracy of circular polarization is less clear, however the primary beam has a significant beam squint, so off-axis imaging is particularly problematic.