The testing and validation process of the archive-new restore versus the output of the production calibration pipeline revealed minor differences between the results. The differences were tracked down to the production calibration pipeline using a standalone installation of CASA that did not have the continually updating data repository, thus Earth Orientation Parameters were becoming out of date relative to the regular CASA installation. It was decided that the production pipeline should be running on the current and updating data repository to provide the most accurate calibration.
Approved: John Tobin/Jeff Kern
Date:Jan 25, 2019
Consulted: John Tobin
Claire Chandler
Juergen Ott
Jeff Kern
Drew Medlin
The Earth Orientation Parameters (EOP) define the uvw-coordinates of a measurement set and are generated during the importasdm procedure. When the data repositories do not match up, especially for recently obtained data, the uvw coordinates will be different. The production pipeline runs on data shortly after coming off the telescope and will be calibrated using the EOP model. However, data that is recalibrated later, after the EOP model is replaced by EOP data, will have slightly different uvw values. Thus, a restoration of a measurement set to a calibrated state will not exactly match a recalibration or a current calibration, unless they are using identical EOP data.
The production VLA pipeline CASA was using EOP Data from Dec. 8th, when it was installed in the pipeline directory at NM. However, the archive is utilizing the normal CASA installation that uses continuously updated EOP data, thus the results were non-identical. However, the differences were generally small in the calibrator images. I conducted some additional tests using A-config data in Ka-band, where resolution is the highest and position shifts might be introducted by using outdata EOP data. I reduced data with old EOP data, current EOP data and restored the original calibration. The results produced position matches within less than 10% of the beam and flux densities were within a few percent. Thus, restoring with different EOP data does not seem to alter the scientific accuracy of the data at the levels of accuracy needed for the VLA.
Risks
Users may obtain non-identical results as compared to the pipeline calibration run and/or a recalibration.
Risk is mitigated by using the best available EOP data, and we can warn users that if they absolutely need the best possible EOP to wait until the data are updated and submit for recalibration.