[daip] AIPS problem

[Eric Greisen]

Gary A. Welch writes:
 > Dear AIPS helper,
 > 
 > Recent comments by Eric Greisen have motivated me to revisit an unsolved problem 
 > in hopes of at least finding the cause, if not a fix.  First, let me explain 
 > that I am far from being an AIPS expert, so part of my difficulty might result 
 > from inexperience.
 > 
 > I have D-array data distributed over 4 days in 1998.   The configuration was the 
 > same on all days.  The IFs were tuned separately to increase the baseline, and 
 > then UJOIN-ed at the end of the calibration process.  
 > 
 > The problem shows up in IF 2 data from 2 of the 4 days.  Symptom: Each 
 > single-channel image in the velocity range covered by IF 2 shows a bull's-eye 
 > pattern.  The rms on those images is 2-4 times larger than on the images from IF 
 > 1.  Curiously, the pattern seems to reverse from channel to channel, so that 
 > 2-channel averaging significantly lowers its amplitude - in most cases.  After 
 > averaging, though, I'm still left with a few objectionable images, some of which 
 > occupy interesting velocity ranges!
 > 
 > Flagging has been ineffective.  I've used TVFLG, and also UVFLG to get rid of 
 > 1-2 occasionally flaky antennas and a few very aberrant phases.  I've looked 
 > closely at UVPLT plots of amplitude vs baseline.  I've spent a lot of time on 
 > the flagging issue, although I might have missed something obvious to a more 
 > experienced AIPS user.  Plots of the BP tables seem to show more "jitter" in the 
 > phases for IF 2 compared to IF 1, but I cannot judge whether that is 
 > significant.
 > 
 > What do you advise?

Try looking at the suspect data with SPFLG.  Set DPARM(2) to 1.

A bulls-eye at the origin suggests a constant disk with zero phase -
perhaps simply something at zero spacing or very short spacings.  Have
auto-correlation data snuck in?  IMAGR has no adverb to exclude them
but they are likely to be badly calibrated if at all.  Many other
display tasks do have adverbs requiring you to include the
autocorrelation data explicitly or they are excluded.

Eric Greisen