[daip] signal 11 abort during CALIB

[Eric Greisen]

Andy Biggs wrote:
> Hi Eric. I did some more testing and it seems that CALIB does not abort if NMAP=1, but otherwise, it always aborts if gridded model subtraction is used. Does this help?
> 
> By the way, these data are also giving what is probably a small problem when I run IMAGR. This task seems to think that a single visibility that corresponds to a ridiculously long baseline that doesn't exist. No other task thinks this and plotting the data against u,v distance shows nothing that far out. More worrying is the message about a guard being applied in IMAGR. I don't know too much about what this does, but is there a way to avoid it? GUARD = -1?
> 
> Thanks,
> 
> Andy
> 
> IMAGR message:
> 
> IMAGR1 10:13:49 IMACPY: Only   341085 samples of   341086 fell on UV grid
> IMAGR1 10:13:49 IMACPY: Copied   341085 visibilities to be imaged
> IMAGR1 10:13:49 **** WARNING data discarded outside usable part of UV plane ****
> IMAGR1 10:13:49 **** Actual Umax 8.3519E+09 exceeds limit 7.7707E+04        ****
> IMAGR1 10:13:49 **** Use cellsize <   0.00001 not   0.90000 to get all data ****
> IMAGR1 10:13:49 **** using a U guard band of  0.300 of a radius             ****
> IMAGR1 10:13:49 **** Actual Vmax 1.0420E+10 exceeds limit 7.7707E+04        ****
> IMAGR1 10:13:49 **** Use cellsize <   0.00001 not   0.90000 to get all data ****
> IMAGR1 10:13:49 **** using a V guard band of  0.300 of a radius             ****

I will try this on my machine - but what kind of computer are you using 
and what $ARCH?  Mac or Linux, LINUX or LNX64, flavor of Linux?  Do you 
compile locally or use the binary distribution?

It is worrisome that 1 vis appears to be so insanely larger than the 
rest in U and V.  But the code correctly ignores that sample and all the 
rest of the messages are solely about having discarded that one.  The 
"guard" region is an area toward the outside of the gridded UV plane - 
it was found that samples falling in that region were inclined to put 
high spatial frequency information in the image that could not be 
recovered in Cleaning.  Thus we issue warnings about too large cell 
sizes.  Not to worry in your case except to see why there is this 1 sample.

The actual Umax is determined after a rotation of the coordinate to the 
numerous facet locations.  That is a simple rotation and should not make 
U and V blow up unless they started that way.

Eric Greisen