Calibration strategies
MeerKAT follows standard interferometric calibration procedures. Here we list a high-level overview of the recommended observational set-up for calibration, with more details given in the sections below. These strategies are applicable to both UHF and L-band.
In general, imaging observations are preceded by a delay calibration on the subarray build, done by the telescope operators. This is applied at the correlator and does not need to be solved for again, though it can be a useful sanity check to solve for delays on the bandpass calibrator during reductions.
A typical duration and cadence for bandpass calibrator visits is for 10 minutes every 3 hours.
It is recommended to visit a phase calibrator every 30 minutes for a duration of 2 minutes.
An absolute polarisation calibrator such 3C286 or 3C138 should be visited at least once during the observation if the user wishes to have polarisation angle information. J1939-6342 is effectively unpolarised and can be used to solve for leakage when the source is observed. The gain calibrator can generally be used to obtain more extensive parallactic angle coverage. Please find further details on other polarisation calibration scenarios here: Polarisation calibration.
Delay calibration
Some a priori corrections are generally applied to the correlator before imaging observations are done: key among them is a delay calibration. While the cable delays of the array are stable, small timing offsets can be introduced during synchronisation of the digitizers. After initialising the array the standard operating procedure is to run a short delay calibration observation. During the observation, the noise diode, as well as a bright, well-known source, are used to calculate and apply time-variable solutions for the antenna-based delays.
The delay calibration observation consists of multiple stages: initially, predefined complex gains are applied across the band in the correlator for each antenna; subsequently, a suitable calibrator is observed and simple antenna-based delays are calculated; next, the noise diodes are activated and cross-polarisation delay, as well as phase, is measured for the entire array. The delays are derived and combined by the real-time calibration pipeline. The delays are applied to the data with the exception of the cross-polarisation phases which are stored in the observation metadata and can be applied at a later stage.
Flux and bandpass calibration
Flux calibration requires that the gains for a given gain calibrator which were scaled to a flux density of unity be re-scaled to the true flux density of said gain calibrator. This scaling requires measuring the flux density of the gain calibrator by computing the ratio of the (flux calibrated) gains of the flux calibrators to the uncalibrated gains for a whole observation on the gain calibrator. The scaling factor will be different for each gain calibrator used in the observation.
J1939-6342 is the recommended and preferred MeerKAT flux calibrator at L-band and can be used without a multi-component model. Please see this page for more information on source structure and models to be used, particularly at UHF. J0408-6545 is also used but has more contamination from secondary sources. Due to MeerKAT's wide field of view and sensitivity, there is some structure introduced into the bandpass response by secondary sources in the field. The scale of these effects on the system bandpass ranges from 0.3% to 1.0% at L band (and are far more severe at UHF). The effects can be mitigated to a minor degree through fringe washing with long solution intervals. However, if an accurate bandpass correction is required, it is recommended that a source model be used for the calibrators. Another source that is listed as MeerKAT flux calibrator in L-band is 3C 286 (J1331+3030) which is mainly used as a polarisation calibrator (please see here), owing to its multi-decade stability. In general, choosing 3C 286 (J1331+3030) as both flux and polarisation calibrators in an observing block shall be avoided. For more details, we refer the readers to this page.
The structure of the MeerKAT bandpass response is stable to within 3% over 3 hours. A typical duration and cadence for bandpass calibrator visit is 10 minutes every 3 hours. However, for spectral line observations, the required time on the bandpass calibrator will depend on the continuum flux density of your target; if the target has a similar flux density to that of the calibrator, the per-channel bandpass solutions will transfer noise to the target spectrum.
Extremely short tracks, typical for monitoring of bright transients, can incur over 100% overhead due to the time required on the bandpass/flux calibrator. We recommend that observations with total duration less than ~3 hours be bracketted with two scans of 5 minutes each on the bandpass/flux calibrator (which can be combined if no significant change is noticed in individual scans)(see Observation overheads).
Gain and phase stability
The MeerKAT receivers are temperature stabilised, ensuring that the typical amplitude response remains stable to levels well below 2-3% over the span of a horizon to horizon observation. Under typical ionospheric conditions, the L-band system phase is stable to within 1% level over time-spans exceeding 1 hour (Figure 1). It is recommended to visit a phase calibrator every 30 minutes for a duration of 2 minutes. Two minutes is sufficient for any of the calibrators listed in the Observation Planning Tool (OPT).
Given MeerKAT’s large field of view and sensitivity, contributions from secondary sources in the calibrator field can be significant (see discussion above under Flux and bandpass calibration). Work is currently underway to develop models of our standard calibrator fields. Further information will be added to our pages on MeerKAT L-band and UHF calibrators when available. Note that fits images for L-band calibrators are available, UHF still to follow.
Note that MeerKAT currently does not make use of noise diodes to calibrate system gain or Tsys during standard interferometric imaging observations, though the capability exists to fire the noise diodes in a wide range of timing patterns.
S-band calibration considerations
The general calibration strategy outlined above (bandpass/flux calibrator visits of 10 minutes every 3 hours; gain calibrator visits of 2 minutes every 30 minutes) is also applicable to S-band continuum observations. The following additional considerations apply, with polarisation calibration considerations documented on the page Polarisation calibration and references therein.
Gain and phase stability
S-band gain stability measurements from long tracks on J1939-6342 show a fairly linear drift that is well-sampled at 30-minute intervals, consistent with the L-band cadence. However, users should be aware that:
Pointing-related gain variability is more pronounced at S-band than at L-band due to the smaller beam size. Mean blind pointing is approximately 40″ at night and 50″ during the day. This can introduce significant gain scatter, particularly in the S3 and S4 sub-bands. See the page S-band capability and status for further details.
Weather sensitivity is increased — gain instability has been observed to coincide with rainstorms. Users are advised to inspect SDP calibration reports for high scatter in gain solutions before selecting a reference antenna.
Flux scale errors are generally < 2% (based on comparison between the Reynolds (1994)/Partridge (2016) and Perley-Butler (2017) scales). However, large inter-observation gain variations have been noted in the S3 and S4 sub-bands, likely due to pointing errors. In one S3 commissioning observation, residual pointing problems produced average flux errors of ~5% at the highest frequencies. Users should therefore treat S3/S4 flux measurements with caution and inspect their gain solutions for pointing-related scatter (Hugo, 2022).
Flux and bandpass calibration
J1939-6342 and PKS B0407-65 remain the recommended flux and bandpass calibrators at S-band. A point source model is adequate at S-band frequencies, as field source contamination is reduced by the smaller far-field beam relative to L-band (Ranchod [2024]; Hugo [2022]). Bandpass solutions have been shown to remain fairly stable over the course of an observation.
Self-calibration
Self-calibration is strongly recommended for S-band continuum imaging. Reference calibration alone may not be sufficient due to small-scale gain fluctuations on some baselines. Commissioning observations have demonstrated that artefacts from reference calibration are significantly improved by at least one round of phase self-calibration. Readers are referred to ThunderKAT team 2023 report, Cotton (2021), and the page Dynamic range considerations for further details.
Direction-dependent calibration
For high dynamic range imaging (fields containing bright sources), direction-independent calibration alone produces unacceptably high artefact levels. Direction-dependent gain calibration (peeling) is recommended and has been shown to greatly reduce artefacts, consistent with antenna pointing errors being the dominant cause at S-band frequencies. Readers are referred to Cotton (2021), and the page Dynamic range considerations for further details. Please also see pages Imaging simulations, External pipelines, Ionosphere , The MeerKAT primary beam and Sensitivity calculators for more context.
Gain calibrators
A dedicated list of S-band gain calibrators is available at S-band gain calibrators. The selection criteria are documented there. Note that this list is evolving as the calibrator survey continues. Table 1 sources were observed at S4 (2625–3500 MHz); Table 2 sources at S3 (2406–3281 MHz).
RFI environment
Globalstar RFI at 2483.5–2495.0 MHz affects the S3 sub-band, mainly on short baselines. Baseline-dependent flagging is recommended. Please refer to the page Radio Frequency Interference (RFI) for further details.
Overall, S3 observations show approximately 10% data loss due to flagging. Please see this 2023 report from the ThunderKAT team for further details.
Sub-band selection
S-band is observed in sub-bands (S0–S4). Switching sub-bands takes 15–20 minutes, including delay calibration and telescope operations QA checks. The S3 sub-band is generally considered better quality than S4 for calibration purposes. SDP calibration intervals are centred at 2010–2096 MHz, 2517–2603 MHz, or 2876–2962 MHz, depending on sub-band selection. Please see the SDP pipelines overview page for more details.
Antenna availability
Users should assume a lower limit of 56 antennas when estimating sensitivity for S-band proposals. Please see the page MeerKAT specifications for further details.
Example calibration strategies
Here are some example calibration strategies for different time on target. These are illustrated using the OPT.
In all of these examples the time on target is 1800s. Each target scan is sandwiched between two gaincal scans (each 120s).
For each observation a sample json file can be downloaded and used in the OPT.
Observation 1: 9 hours on target
Observation 2: 6 hours on target
Observation 3: 4 hours on target
Observation 4: 3 hours on target
Observation 5: 1 hour on target
For observations with total duration less than ~ 3 hours the observation should be bracketted with two scans of 5 minutes each.
Observation 6: snapshot observation
This observation observes three targets for 15 min each. Each target is bracketted by two gaincals.