[evlatests] Delays and Closure

[Rick Perley]

    Spurred by a question from Frazer and Sanjay, I've investigated the 
closure issues a little more closely.

    The BLCAL that I ran on the 4 Jy source was applied to the SPLITed 
data -- the only deviant points remaining were those from the 'wobble' 
baselines. 

    I thus ran SPLIT *without* the BLCAL -- and got a *very* different 
picture ...

    Both the RCP and LCP visibilities now show significant offsets in 
the visibilities of some baselines.  These are easy to separate from the 
'wobble' baselines -- wobbleing baselines always oscillate about the 
mean amplitude of phase.  The true closure baselines are always offset 
and steady -- and are only seen in amplitude.  The offsets in LCP are 
much larger -- up to 10%! 

    It was immediately found that *all* of the visible offset (closure) 
baselines are identified with those antennas with large delays.  This is 
a one-to-one relation -- there are no baselines with large delays 
without offsets.  Running BLCAL on this database perfectly removes the 
visible closure offsets.

    The trouble is:  It's my understanding that AIPS corrects for the 
decorrelation from these large delays.  I followed the recommended 
procedure:  FRING and CLCAL, which created a CL table which was then 
applied for further processing. 

    Furthermore, comparing the amplitude closure offsets with the 
measured delays shows a curious relation:  All the negative offsets 
arise from baselines whose delays are positive.  All the positive 
offsets arise from baselines where the delays are of opposite sign.  Yet 
these latter offsets are much smaller than the negative ones, when they 
should be larger (the two contributing antennas are 'further apart' in 
delay space). 

    This indicates that the delay correction is not being applied 
correctly (or there is a pilot error by your faithful scribe). 

    All of this will be eliminated once we get these pesky delays out of 
the system!  We might then be able to measure the *true* closure offsets 
...