UC: DetermineAntennaPolarization -------------------------------- S.T.Myers revision 2000-10-18 This pipeline task uses data taken in full polarization mode (with cross-hand products) in quasi-real time, in an automatic way, to determine the instrumental antenna polarizations for a particular band setup. Data is taken from the raw data archive and an antenna polarization "object" is written to the calibration archive. Role(s)/Actor(s): Primary: AOS, ALMA Archive System Secondary: Operator, Staff Astronomer, Observer Priority: Major Performance: seconds Frequency: minutes to hours Preconditions: 1. correlator data for a potential calibration run is available I am assuming that the calibration observations themselves were scheduled and carried out as a full-polarization standard observation of an appropriate source with sufficient parallactic angle coverage (see UC: ObserveFullPolarization) 2. either programme dependent or standard system scripts are available for pipeline reduction - in particular system standard scripts might be preferred to ensure conformity of results, however for alternate methods of polarization calibration user-supplied scripts might be preferred Basic Course: calibrator of known or unknown polarization is observed for either a single snapshot or over a range of parallactic angle 1. the ALMA Archive system or the AOS initiates pipeline when the observations for a potential polarization calibration are available 2. the script produces a solution for the cross-polarization leakage terms (D-terms) and estimates of robustness, perhaps in comparison to previous solutions in addition to internal error estimates Basic Primary Course: calibrator of known or unknown polarization observed sufficiently over range of parallactic angle to separate instrumental and source polarization 3.1.1 instrumental D-terms (Jones matrix) can be determined from data with known polarization, up to an arbitrary single D-term over the whole array. 3.1.2 instrumental D-terms (Jones matrix) can be determined from data with unknown polarization, up to a single D-term and arbitrary phase (eg. R-L phase difference for circular, orientation on-sky for parallel linear). This phase can be determined by single scan on source of known polarization (Basic Secondary Course below). Basic Secondary Course: calibrator of known polarization observed for single snapshot at arbitrary parallactic angle 3.2.1 instrumental D-terms can be determined up to an overall distortion corresponding to an overall single unknown matrix, plus and arbitrary phase (eg. R-L phase difference for circular, orientation on-sky for parallel linear). Basic Tertiary Course: unpolarized calibrator observed for single snapshot at arbitrary parallactic angle 3.3.1 instrumental D-terms can be determined up to an overall distortion corresponding to an overall single unknown matrix, plus and arbitrary phase (eg. R-L phase difference for circular, orientation on-sky for parallel linear). A snapshot of a source of known polarization (Basic Secondary course) needed to establish this unknown phase. Exception Course: the script fails to obtain a valid solution (eg. due to insufficent parallactic angle coverage or non-convergence) 1. notification is given to the operator 2. the calibration schedule is resubmitted to the dynamic queue for execution Alternate Course: on-line execution 1. operator or observer instigates calibration run and expects pipeline to take care of the analysis 2. the pipeline normally executes script commands when complete calibration data is available in archive These commands retrieve data for the raw data archive, manipulate data, optionally display results, store results in calibration archive. Exception Course: the operator is notified if the script fails 1. the pipeline makes results available to the AOS. 2. the pipeline makes quality check results available to the scheduler. Alternate Course: off-line execution 1. the operator, observer, or staff scientist initiates the pipeline off line, perhaps with a modified script 2. this behaves in all other ways as the basic course Postconditions: 1. Parameters are fed back to either the AOS or the Scheduler, and results are written to the Calibration Archive 2. Science data (images) produced as a by-product of the calibration (or perhaps the calibration a by-product of the science observing) are written into the Science Archive 3. Data reduction scripts and logs are written into the Science Archive Issues to be Determined or Resolved: 1. I am assuming that the physical receiver hardware consists of some sort of ortho-mode transducer to separate out the orthogonal polarizations plus a fixed quarter-wave plate if circular polarization mode is desired (unnecessary if linear), or a quasi-optical device such as a grid diplexer which will send one linear polarization to a different path. Some other scheme with a single rotating grid would complicate things. 2. The possible use of a calibration tone to determine the receiver plus receiver optics polarization would be a separate UC, as it would then be similar to the use of a secondary load or signal for amplitude calibration. This would presumably be applied in quasi-real time (at least as frequently as once per run). 3. In the current optical design, there is significant polarization aberration (squint of ~4%), and the polarization properties will likely vary over the primary beam. I do not know how to deal with this easily, though perhaps the holography measurements of the beam will produce the necessary numbers. See also (9) below. 4. I presume that there is some sort of schedule flag or tag that indicates that a particular observation is intended as or is suitable for determining polarization calibration. This will alert the Archive control to instigate this process. It may be that the operator or staff scientist flags a particular science observation as suitable even though the observer did not indicate it thus. For longer observing programs the observer will likely supply their own calibrator observations, while snapshots will likely rely on cal-transfer for other (possibly system) programs. 5. How do we envision the calibration processes working together (eg. baselines, flux, bandpass, polarization)? In particular the phase calibration and polarization calibration are related by the XY or RL phase difference. In spectral line mode there are probably polarization considerations in the bandpass calibration. 6. The must be a mechanism to notify the Archive Control that a calibration parameter has been updated and needs to be propagated through the Science Archive 7. Where in the data structure does the antenna polarization go? In AIPS this is stored in the Antenna (AN) Table. In aips++ this is stored in a table as part of the measurement set in a Jones matrix form. 8. Is the Archive System the controller of calibration operations? For example, on failure should it resubmit the schedule to the queue, or just notify a human? 9. Just what are the limitations of the various polarization calibration schemes? I guess the papers by Sault, Hamaker et al. are a good place to start (see Notes below). Another question is what parts of polarization calibration must be deferred to the imaging (pipeline) itself, such as the application of wide-field polarization calibration of the primary beam? 10. So far I have only really considered continuum polarization issues. Are there modifications needed for spectral line (eg. Zeeman) polarization observations? Notes: 1. I am assuming that the calibration observations themselves are scheduled and carried out as a full-polarization standard observation of an appropriate source. Suitable parallactic angle coverage must be ensured. I do not see that there need to be a special UC for these observations - in fact for most non-snapshot programs the phase calibrator will serve as a reasonable polarization calibrator. 2. Polarization calibration is an example of a "state" calibration where some parameter describing the system is relatively time-independent (assuming no physical changes are made to the array) and thus a global storage of these parameters is effective and can be shared among many observations using the same band setup 3. I am assuming that there is a Calibration Archive where the results of the various calibrations (baseline,polarization, bandpass,flux) are stored for access by the on-line system and the pipeline 4. The papers by Sault, Hamaker and others are good references for the subtleties of polarization observations and calibration. In particular: Sault, Hamaker & Bregman (1996) A&ASup 117, 149-159; Hamaker (2000) A&ASup 143, 515-534. There seem to be three main optical geometries: a. circular - two orthogonal circualr poln. R and L each antenna b. parallel linear - two orthogonal linear poln. V and H oriented the same on all antennas c. crossed linear - two orthogonal linear poln. V and H oriented at 45 deg w.r.t. half of the antennas The differences between these cases are mostly in implementation, and we dont really know what ALMA will have yet! There is a claim that a hetergeneous configuration like (c) breaks the degeneracies and thus a simple intensity alignment will determine all parameters. I think WSRT uses some such scheme and it might be useful to look into what they do. 5. There seem to be three modes for calibration using celestial sources: a. a single scan of an unpolarized source b. a single scan of a source of known polarization c. several scans spread in parallactic angle of a source of known or unknown polarization plus obvious combinations, eg. (a+b), (c+b), and multiple sources (n x b). Hamaker (2000) splits the polarization errors into a "polrotation" and "poldistortion" (matrices), with the main difference being that the poldistortion couples I into the QUV while the polrotation is a rotation of QUV. It seems that an intensity calibration (a) removes the poldistortion, some assumptions about the instrument can fix 2 of the 3 DOF of the polrotation, and an observation of a polarized source of known polarization can fix the remaining XY or RL phase difference. Thus the combination (a+b) seems to be a good way to calibrate. If there is no suitable unpolarized calibrator available (as in high-frequency VLBI), then option (n x b) or (c + b) is the way to go. Created: 2000/10/10 smyers Revised: 2000/10/17 smyers Revised: 2000/10/18 smyers