Calibration is a process of characterising response of the telescope. It is typically performed by observing stable and well characterised (described by a model) astronomical sources (calibrators). This process leads to visibilities calibrated in the units of Jansky (Jy) and ultimately to sky images, which also have correct flux density in units of Jy. Based on observations of a calibrator source (or target field) complex gains of each MWA tile are calculated.
The two typical calibration scenarios are :
Both ways lead to calibration solutions of each antenna represented by complex gains (amplitude and phase) as a function of frequency. These solutions are subsequently applied to uncalibrated visibilities in order to calculate calibrated visibilities, which enable further data processing ultimately leading to sky images (for details on how calibration solutions are applied to the uncalibrated data see: Calibration and Stokes Imaging with Full Embedded Element Primary Beam Model for the Murchison Widefield Array, Sokolowski, M., et.al. 2017. Based on the known spectrum of the calibrator it calculates band-passes of each antenna (MWA tile). Hence, application of these calibration solutions to raw visibilities converts them into calibrated visibilities with a correct flux scale. Consequently, leading to correct flux scales of sources in the final sky images. It also calculates additional phase corrections which were not accounted for in the pre-calibration steps (wrong cable lengths, ionosphere, etc - see above). Hence, leads to phase calibrated visibilities (coherent across the tiles), which can then be inverse Fourier transform to produce a dirty image.
Calibration solutions change in time as they depend on ambient temperature, ionospheric conditions etc. Our assumption, based on the MWA experience, is that the telescope is stable over many hours. Hence, every morning and evening a set of six calibrator observations covering the standard MWA observing band (~ 70 - 230 MHz) are performed, which enables calibration of any MWA data collected during this day/night. These calibrator scans plus additional project-specific calibrator observations have been used to calculate calibration solutions and populate the calibration database (CALDB) used for the MWA ASVO.
When an MWA ASVO user requests a conversion job and for the observation to be calibrated, the following steps are performed by the system:
For a given target observation, the system will try to find an optimal calibration solution as the closest-in-time to the target observation and no more than 12 hours apart. Calibration solutions for the entire MWA Archive are being calculated and stored in the CALDB, however this process is very time and compute intensive and is expected to continue into early 2019. If no suitable calibration solution is found in the CALDB for the requested observation, the conversion job is failed and a request will be queued on our servers to produce a calibration solution for that observation to store in the CALDB. This normally takes 24-48 hours, so the user may then resubmit the conversion job later on and the calibration solution should be available.
When the system generates calibration solutions to populate the CALDB, if there are no suitable calibration solutions in the database, the system will find a close-in time calibrator observation (or other field suitable for calculating calibration solutions) in the main MWA archive, calculate calibration solutions using this observation (as described in the earlier section) and also save them in the calibration database for future use. Therefore, any potential gaps in the calibration solutions database will be filled as new user requests arrive. This may also be necessary on request for the nights with significant ionospheric activity when calibration solutions can change on timescales even shorter than 1 hour.
Once a suitable calibration solution is found in the calibration database (or calculated using close-in-time calibrator observation) its quality is verified based on the quality fields in the database (x_phase_fit_quality, y_phase_fit_quality, x_gains_fit_quality and y_gains_fit_quality). All these quality indicators are value between 0 and 1 and are calculated as a ratio of number of channels close to the fitted curve (less than 5 standard deviations away) to the total number of channels. If the value of quality indicator is below a certain threshold (currently 0.6) the data from such a tile gets automatically flagged and files with list of flagged tiles is attached to the output zip file (obsid_flagged_tiles.txt).
As part of the conversion job steps, the metafits file for the observation (containing metadata for the observation) is updated to flag any tiles that failed the calibration quality control (tiles listed in obsid_flagged_tiles.txt). The calibration solutions (in cal_obs.bin) and metadata (obsid.metafits) are applied to the raw visibilities by the cotter processing software and the calibrated visibilities are returned to the MWA ASVO user in one of the standard radio-astronomy formats (CASA measurements set or UVFITS files).