Monday, April 2, 2007

Using Calibrator Models

Wisdom from Stefanie Muehle:

BTW, I now got a nice calibration of my spectal line data using the models as recommended in the AIPS cookbook. The thing you have to get your head around is that when you apply a model, you don't _expect_ the phases to be zero, because you are dropping the default assumption that the calibrator is a point source. The clean components and the visibility plots are not much help in checking whether your calibration is ok, either. So, here's the tip from Michael Rupen, one of the gurus at NRAO:
The structure of the phases will depend in detail on the model.  To check whether you're getting something reasonable, try the following:

* If you restrict the uv-range in UVPLT to match the original (no model)  recommendations, you should still get flat (zero) phases.  Do you?

* You can check the phases predicted by the model by using UVSUB with opcode 'MODL' (this may require splitting off 3C xxx to make single source  files).  Try that and see whether the predicted phases match what you observe.

If the post-calibration phases do NOT match the model, there is indeed something wrong, and we'll work with you to track it down.  I hope and expect this is not the case though... :}

So, you basically first split off the calibrators into single-source files (SPLIT), then you substitute in each file the observed phases with those derived from the model images provided by AIPS (UVSUB with optype = 'modl'). When you then plot the phases of these files, they should look like the UVPLT of your calibrated data, only without the noise. At U-band (2cm) the model phases of 0137+331 and 0542+498 are so close to zero that the difference between the model and a point source is lost in the noise. But for 1331+305, the effect was quite spectacular. The model predicts phases up to |\phi| = 60\degr! That doesn't look like one should "force" the phases down to around zero by assuming a point source.

The catch is that I don't know whether the calibration with a model affects the polarization calibration. PCAL should be ok, since we are using the phase calibrator, which is assumed to be a point source. But RLDIF requires the polarization calibrators 0521+166 and 1331+305, for which the phases (RCP and LCP) will be substantially different from zero.

No comments: