UVdistances in blflag a factor of 20 too large

[PKamphuis]

Hi All,

I've been trying to flag some data based on uv distances in blflag and the uvdistances that blflag returns are about a factor 20 too large. I'm not sure wether this is a bug or wether I am misunderstanding something.

The data I use is a 500 channel cut out from a concatenated zoom band on 1934-638 and can be found here http://www.atnf.csiro.au/people/Peter.K ... xample.tar
The data is merely loaded with atlod, subsequently the 500 channels are cut out of the concatenated zoom band and the data is split with uvsplit. No other processing has been done.

If I use uv plot to look at the data I get the following:

>uvplt vis=1934-638.1284 select="time(14APR11:01:58,14APR11:02:00)" stokes=xx,yy axis=uvdist,amp device=/XWINDOW options=nofqav,nobase
uvplt: Revision 1.18, 2014/05/21 04:34:59 UTC
Will plot unflagged data


File # 1 = 1934-638.1284
1934-638.1284 1.2842 GHz
Read 360 visibilities from all files
Plot 179980 points

uvplot of data

1934-a.png (44.05 KiB) Viewed 5549 times

Which is what I would expect. However, when I use blflag xx,yy get averaged (Which they should) and the uvdistances become a factor 20 larger (which I don't understand) as you can see below.

serpens-HI>blflag vis=1934-638.1284 select="time(14APR11:01:58,14APR11:02:00)" stokes=xx,yy axis=uvdist,amp device=/XWINDOW options=nofqav,nobase
blflag: Revision 1.20, 2013/05/23 03:43:33 UTC
Reading the data...
Number of points to edit: 89990
Entering interactive mode...

blflag plot of the data

1934-b.png (37.52 KiB) Viewed 5549 times

It seems to be related to not averaging over frequency as when I remove the nofqav command uvdistances return to normal (see below). however I do not understand why that should make such a difference.

>blflag vis=1934-638.1284 select="time(14APR11:01:58,14APR11:02:00)" stokes=xx,yy axis=uvdist,amp device=/XWINDOW options=nobase

blflag: Revision 1.20, 2013/05/23 03:43:33 UTC

Reading the data...
Number of points to edit: 180
Entering interactive mode...

blflag averaged over frequency

1934-c.png (32.99 KiB) Viewed 5549 times

I checked this behaviour against the phase calibrator and it is similar. I have no idea why the uvdistances should be so much larger when the frequency averaging is off.

[Mark.Wieringa]

Hi Peter,

thanks for your report. This bug has been around for quite a while (2010), so clearly the uvdist option is either not used much or you inspect axis labels a lot closer than most people.
The numbers were out by a factor sqrt(numchan). I've now fixed this and also made the uvdistance scale with the frequency in the spectrum. It was using a single point before.
This means the results without 'nofqaver' differ from uvplt, which still uses the first point in the spectrum for the uvdistance calculation in that case.

Cheers,

Mark