A note on naming convention: bandpass vs. passband. The MIRIAD item is called bandpass, but you will often also find the name passband uses in source code and script names.
First we measured the variations of the antenna based solutions of typical 2" integrations of the noise source. There are two types of data available in the archive: - fringetest: usually 6 samples of 2" integrations preceding the calibrator, 1 minute of 2" integrations now?? used to be 10" The correlator is usually set up with 8 500MHz bands spread out over about 4-5 GHz in the IF. - science data: usually 1 sample of 2" each time before a calibrator The correlator setup varies, depending on science track. mfcal bandpass solutions are computed, and a Ant-Chan-Time cube for phase and amplitude is created. For each Ant-Chan point a time averaged value is subtracted, and the variations in this cube are given as an indication of the stability. - 6 sample fringetests consistently show amp variations 2e-4 and phase variations 0.1 degrees - a 2-3 hour science track show different variations: * amp 4e-3 phase 0.7 deg for noise,c0566 2" over 2-3 hrs * amp 0.04 phase 3 for 3c454.3,c0566 30" over 15 mins * amp 4e-4 phase 0.2 for noise,c0645 2" over 2-3 hrs * amp 0.03 phase 3 for 3c273,c0645 30" over 15 mins Notables: - the noise source sometimes does go wacko - after subtracting mean banding phasetest.py - array health - can be used.PJT 2-sep-2010