Antenna Position Correction Bug

A bug has been found in CASA and the CASA pipeline (present in CASA 6.5.3 and earlier) that ignores VLA baseline corrections for an antenna that has zero values for both the x (i.e., Bx) and the y (i.e., By) parameters and a non-zero value for the z parameter (i.e., Bz), e.g.,

;MOVED OBSDATE   Put_In_ MC(IAT)  ANT  PAD    Bx      By      Bz
 FEB13  FEB14     FEB14   20:00   18   W32  0.0000  0.0000  0.0133 

This bug goes back to the first implementation of this functionality within CASA's gencal task and the Scripted Pipeline in 2012, and was carried over into the CASA-integrated Pipeline which is the main calibration pipeline run on all science observations at the VLA currently.

While this bug has been present for a long time, the impact is minimal and isolated to a very small number of observations, for several reasons:

1. Only those datasets that were observed between the time when an antenna was moved and when the correction was 'Put in' to the observing system will have missing antenna position correction(s).
2. The time between moving an antenna and its corrections being put in to the system is generally short, a few days at most.
3. Antenna position corrections where Bx=By=0 and Bz≠0 were infrequent, so typically even if this effect is present, it is present only on one or two antennas.
4. Most of the effect is taken out in the normal course of calibration.

The main effect of this bug is a small delay offset, which manifests itself mostly as a phase error on baselines involving any affected antennas (one or two at most in any given time). The magnitude of the error is proportional to the inverse of the wavelength being observed, the magnitude of the antenna position offset in the z direction, and the difference of the sines of the declinations of the target source and complex gain calibrator source. The average magnitude of the offsets in the z direction is a few mm, making the error unimportant for all but high-frequency observations (Ka- and Q-bands). For target-calibrator declination differences of a few degrees, even at those high frequency bands the error is negligible. Furthermore, if it is possible to self-calibrate your data, the error is corrected. Because of all of this we do not think that this should be cause for concern, nor for a re-calibration of any data taken in the past. However, if your particular observations are during an affected time (you may submit a helpdesk ticket to determine this), are high declination, are high frequency, and have a target-calibrator declination difference of more than 10 degrees, you may want to consider re-calibration, but we do not expect this to be the case except in extremely unusual circumstances.

To mitigate this problem until the bug is fixed in CASA and the pipeline, on February 8, 2023, we have modified the entries in the baseline corrections database that trigger this bug by introducing a very small (insignificant) value in Bx. Observers who wish to reprocess their data may do so using the current version of the pipeline in order to check whether their results are significantly impacted by this bug.

This bug will be fixed in the next version of the CASA+pipeline package which will be based on CASA 6.5.4+ that is expected to be released in late-September 2023. Standalone versions of CASA will also have the fix starting with CASA 6.5.4 and later; the upcoming CASA 6.5.3 will still have the bug.