[daip] STUFFR

[Andy Biggs]

Hi Eric. Okay, that explains why the maps look so good. Interestingly, I 
ran MATCH on the five of my six datasets with identical configurations, 
then ran STUFFR on them and attached the final dataset (which has a 
slightly different configuration) with DBCON (with DOARRAY=-1). I maybe 
made a mistake, but the resultant map actually has a higher rms than doing 
the 'wrong thing' without MATCH.

Another thing has occured to me though. As expected, if I plot the u,v 
coordinates of a particular baseline in the STUFFR dataset, it corresponds 
to multiple actual u,v tracks. In one case I get 3 different tracks on the 
one baseline. Therefore, attempting to self-calibrate a STUFFR dataset 
would appear to be incorrect. However, I could see that CALIB might do 
sort of the right thing as the model subtraction is presumably done using 
the u,v coordinates of each visibility, regardless of its baseline 
identifier.

I've been attempting self-calibration at this stage mainly because 
Frazer's low frequency reduction guide suggests it. I think in the past 
it's made very little difference due to the each dataset being 
self-calibrated to the same model before STUFFR.

Andy

On Mon, 27 Aug 2007, Eric Greisen wrote:

> Andy Biggs writes:
> > Hi again Eric. This was what I was afraid of and would appear to be
> > disastrous for various datasets that myself and others have mapped in the
> > past. Just looking back at another wide-field VLA dataset that I still
> > have on disk, the data were taken in two sessions about a year apart and
> > the antenna numbers for each are, unsurprisingly, completely different,
> > pad-wise. However, the final maps look excellent, without any hint of a
> > problem with the data. I'm surprised that anything recognisable appears in
> > the maps at all.
>
> Actually, UBAVG compares adjacent uv data samples in BT and averages
> them only if the u,v's are within a certain distance set by the
> desired field of view.  This almost certainly helps prevent wrong
> averaging although it also prevents right averaging thereby leaving
> you with too many samples in the imaging.  I suspect that this means
> that your images are ok but in future you might wish to use MATCH to
> align the data sets before STUFFR.
>
> Eric Greisen
>