MIRIAD 's Calibration Model

MIRIAD 's suite of calibration tasks generally produce or use items which contain the relevant calibration tables and parameters. The tasks to manipulate these calibration tables generally have names start with gp, though there are a few exceptions such as selfcal and mfcal. The calibration parameters are antenna-based ones, with the calibration data being used to derive a model of the response of each antenna. The response of an antenna to radiation is modelled by four sets of parameters:

• Antenna gains. These are complex-valued gains which vary with time, but not frequency. These are predominantly atmospheric in origin, though there is a significant instrumental component (the two cannot be distinguished readily). MIRIAD stores these in the gains item of a visibility dataset.

• Delay factors. These are antenna-based parameters which give a time delay, which will probably be a sum of atmospheric and instrumental components. This term is assumed to vary with time, but not frequency. It is stored in the gains item.

• Antenna bandpass functions. These are complex-valued gains modelling the instrumental bandpass. These vary with frequency but not with time. These are largely instrumental in origin. These are stored in the bandpass item (and some associated items).

• Antenna leakages. These are complex-valued terms (which do not vary with time or frequency) which model the leakage of one polarisation into another. These are entirely instrumental in origin. Current experience suggests that leakages may vary moderately with frequency. These are stored in the leakage item.

Ignoring leakage (which is discussed in too much detail in Chapter 11), we model the composite gain function of an antenna (representing both atmospheric and instrumental terms) as

Here g(t) is the frequency-independent part of the antenna gain, is the bandpass function, is a delay term. The delay is calculated with respect to a reference frequency, . For dual polarisation systems the two polarisation bands are assumed to have independent gains and bandpasses, although the delay is common.

Note that MIRIAD datasets cannot contain more than one set of calibration tables -- you cannot have multiple versions of calibration tables. Running a calibration task twice will result in the first calibration table being overwritten. This may be inconvenient if the dataset has two sets of data that need to be calibrated separately.