[daip] CALIB with a model

[Eric Greisen]

Rupal Mittal wrote:
> Hi,
> 
> I am trying to calibrate P-band VLA data with 3C286 as the phase calibrator 
> and Virgo A as the band pass and flux calibrator. Even though 3C286 is as 
> good as being a point source at this frequency, the primary beam contains a 
> few contaminating sources. So I did the bandpass and flux calibration using 
> Virgo A and made hybrid maps of 3C286 (from a single-source file) comprising 
> 123 facets. As the next step, I'd like to calibrate the phases and amplitudes 
> of 3C286 (in the primary multi-source file) from this derived model 
> containing 123 subfields in CALIB, so that I can use 3C286 as the 
> phase-calibrator in order to map the target source. I tried two different 
> things (the input parameters for CALIB and the output messages created in 
> both the cases are pasted below):
> 
> One was to let SETJY calculate the flux-density (optype 'CALC') from the known 
> spectrum. Please note the FACSET messages. It finds only 17 mJy in the model. 
> I also tried inserting in SETJY the sum of all CC's in the model (~48 Jy) 
> using the adverb ZEROSP but the FACSET results remain the same. On changing 
> NMAPS from 123 to 50 results in "FACSET: 0.000000 Jy found from 971 
> components". Why is this the case? The same exercise with Virgo A (inserting 
> in SETJY the sum of the CCs from the model) and using the FITS file from 
> http://lwa.nrl.navy.mil/tutorial/omodels.html works just fine. FACSET finds 
> the right flux density.

      I found this disturbing and took a look at the code.  Indeed 
FACSET was looking in a radius of 3 arc-sec in EACH facset - so I am 
fixing it to look for standard sources only in facet 1.  It is clear 
that you have no flux in your model at radius < 3 arcsec in facet 1 
(through 50) of the reference pixel which should be accurately the 
coordinate of 3C286.

At P band this is not so surprising - the ionosphere can move sources 
around quite a bit so 3C286 might not appear where it is supposed to.
Of course, if you self-cal with a shifted model you freeze in that shift 
and cannot correct for it.

SETJY 'CALC' computes the flux of 3C286, not the flux of the full P-band 
field which should be rather more due to all the other sources visible 
at low frequency in the wide field.  CALIB then tries to adjust the flux 
scaling so that the model flux, when scaled, equals the 'CALC' flux. 
Since your model contains no flux at the 3C286 position, this is not 
possible.
> 
> The second was to set the flux-density of 3C286 in SETJY to zero. As far as I 
> know, the scaling is meaningless in the case the source flux is found to be 
> zero in the SU table. But FACSET reports the correct flux density in this 
> case. Also the intensity vs. uv distance plot after applying the calibration 
> looks reasonably good and similar to the model visibilities. Is this 
> technique correct then (as long as one doesn't use GETJY after this)?

It works better than the first method for reasons explained.  But, if 
the model is shifted so that 3C286 does not appear at its correct 
position, then all later images will show this shift as well.  Not too 
that the phase error may vary across your 123 facets so that the correct 
phase cal at the center may not be correct toward the edges due to the 
size of the isoplanatic patch in the ionosphere.
> 
> An easier method is to copy the last SN table produced in the course of 
> self-calibration and use Virgo A as the primary calibrator to bootstrap the 
> flux-density of 3C286 in GETJY (flux in the SU table before GETJY is anything 
> but zero). But I have datasets where 3C286 is the only calibrator in which 
> case there doesn't seem to be any other choice but to use the maps in CALIB.

Have you looked at the images in the 123-facet?  Is there a source of 
~26 Jy somewhere near the center?

Eric Greisen