In late 2020 we became aware of some problems with the accuracy of the delays applied to the antennas. The causes included the limitations of a single-precision astrometry library, incorrect use of the UT1 correction, and a faulty pressure sensor on site which caused large changes to the tropospheric component of delay.
In many cases the nearest suitable gain calibrator is > 10 degrees from the target field. The worst case errors occur when the phase calibrator is > 45 degrees from the target. Currently the source position errors can be up to 2 arcsec. The errors get smaller when the calibrator is closer.
Potential solutions are under investigation. For now users should be aware of the possibility of small position errors, especially when the target is far from the calibrator. For more information, please see this memo.
Time and frequency errors
The earliest data had a 2 second of time and a half channel in frequency errors resulting in incorrectly calculated 𝑢, 𝑣, 𝑤 baseline coordinates. These result in a rotated and miss-scaled image producing apparent position errors up to 2” at the edge of each pointing image. This issue was corrected in March 2019. Katdal will automatically correct these errors during the measurement set conversion process.
Poor calibrator positions
The initial MeerKAT L-band calibrator list included sources with errors up to several arcseconds. These will result in pointing images with constant position errors if the calibrator positions are not corrected (e.g. use fixvis in CASA).
The calibrator database underwent revisions between 8-18 November 2019, with coordinates updated to ICRF positions, if relevant. If you use data from observations prior to 18 November 2019, please check the listed coordinates against the current calibrator database.