[daip] Combining two sets of UVDATA increases noise?

[Eric Greisen]

landon at inaoep.mx writes:

 > I took one track of 13CO data from OVRO and flagged all the visibilities that
 > occured in the same positions in the UV plane as the visibilities in both of 2
 > other tracks. I then did the opposite and flagged all the visibilities that
 > weren't common to all four tracks. I made cleaned maps of both of these (with
 > the same adverb settings) and obtained noise levels of ~1.5mJy/beam and
 > ~2.5mJy/beam for a certain channel of interest.
 > 
 > The strange thing is that although these two sets of flagged UV data are
 > completely complimentary, the noise level of the cleaned map of the original
 > data (ie: equivalent to these two sets combined) is ~4.5mJy/beam. From what I've
 > learned, more data (especially complimentary data) should lower the noise level,
 > not raise it by a factor of ~2.
 > 
 > The cleans were all done with the same adverb settings and the beam sizes are
 > all the same. 

I agree that it is counter-intuitive.  You did not say whether the
source parts of the images agree or not.  You did not say what the
robust weighting noise degradation parameters were for the various
images.  With different sampling one would get different weighting
unless you use natural weights.

I would guess that each data set is calibrated internally moderately
well but that the two are not calibrated the same.  In that case,
combining them would make an apparent noise greater than the
individual noises.

Eric Greisen