Self-calibration

       

The most reliable method for removing atmospheric fluctuations from the data is self-calibration, if the target is strong enough. Self-calibration should be used if the signal-to-noise on the target is high enough to determine the gains in the typical coherence time of the atmospheric phase screen, using baselines involving all antennas. For an array of N elements, the r.m.s. noise in the determination of an individual antenna phase gain by self-calibration on a point source of flux density Jy will be

(4) 

where is the sensitivity in Jy per second per MHz of bandwidth, and is the averaging time used for the gain solutions. Unless when , self-calibration will not improve the data.

There are several reasons why external calibration is useful even if you know you will be able to self-calibrate your final images.

• External calibrators will provide flux-density and position scales for self-calibrated images (on which this information will otherwise generally be lost).
• Observations of the time scale of the phase fluctuations on a phase calibrator near your target can also be used to estimate the coherence time of the atmosphere during your observations. This will enable you to select a suitable averaging time for the self-calibration: it should be .
• External calibration may also tell you that some parts of your data were obtained under more stable atmospheric conditions than others; the ``good'' parts may yield a better initial model of your target that helps self-calibration of the whole data set to converge quickly.

Fortunately, the class of source for which images of high dynamic range are most important is also the class for which self-calibration is most likely to work well--namely, those with weak extended structures around bright small-diameter components. There is however a range of flux densities and structural complexities over which self-calibration cannot be guaranteed to work in typical atmospheric coherence times , and for which external calibration therefore remains essential.