[daip] visibility weight problem in output of DIFUV

[Eric Greisen]

Huib Intema wrote:
> Dear DAIP,
> 
> I use DIFUV to subtract a model UV database from a real UV database. The 
> model UV database is created with UVSUB with OPCODE = 'MODL', using the 
> real UV database as input, which results in model visibilities having 
> the same weights as in the input. Before (using a mid-2008 copy of 
> AIPS), I needed to run WTMOD on the output of DIFUV with APARM(1)=0.5 to 
> compensate for the unwanted summation of (quadratic?) weights by DIFUV.
> 
> I found that DIFUV in the current 2009 AIPS causes the output visibility 
> (quadratic?) weights to be halved rather than doubled. This I could fix 
> by using WTMOD with APARM(1)=2.0. However, when the input visibility 
> weights are both zero the output weight is NaN. This looks like a zero 
> division error while adding the reciprokes of the input visibility 
> weights rather than the weights themselves.

Sorry - this slipped through the cracks of EVLA needs.  I have fixed the 
computation to be more careful (it would abort on Suns and perhaps 
others).  Tomorrow's MNJ should fix it.

Eric Greisen