[evlatests] Test Run, Saturday evening

[Rick Perley]

    I used a few hours of dynamic time to try a deep, high-fidelity 
imaging experiment on a well-known strong source, 3C273. 

    Many interesting things were learned, I'll summarize only a few 
major ones here (after all, it's Sunday afternoon ...).

    I started the experiment using two different modes:

    1) Continuum, 50 MHz BW.
    2) Mode 'PA', which provides full polarization, in a single IF pair 
(A and C), with 12.5 MHz BW.  (64 channels total). 

    It was immediately apparent that no data from the 'PA' mode were 
getting to the archive.   
    It was initial felt that the integration time might be to blame (3.3 
second), so a 6.6 second integration version was tried.  No change.  I 
then switched to mode '4', with 3.3 seconds averaging -- everything 
worked fine.  Hence, conclusion #1:

    ******* Mode 'PA' is not working  ***********

    Beyond this, from the continuum data:

    *  IF 'D' suffers greatly from the correlator problems, already 
noted last week.  Although easy to spot (kilo-Jy visibilities can't be 
hidden!), it's a darn nuisance.  I trust this will be addressed on Monday. 

    *  On two occasions, all visibilities went to low values:  00:18:14 
-- 00:18:29, and 02:51:51 -- 02:52:09.  Both are short, and were easy to 
see when on a strong source.  But such events won't be easy to find when 
on a noise background... 

    *  The long-term phases are very strange, looking neither like a 
baseline problem, nor a short-term problem.  More work will be needed to 
extract something useful from these. 

    And, from the spectral line data:

    *  In the *** middle of a scan***, the EVLA and VLA went incoherent 
w.r.t each other for IF 'B' only -- but each array remains coherent with 
itself.  The classic problem.  This affected only the spectral line 
data, not the continuum data.  And the problem, once it started ( at 
0:27:48 on Sunday morning, IAT), never went away through the end of the 
run (despite numerous change to continuum and back to spectral line).    
Weird!  How can this happen?????

    *  The IF 'D' correlator problem seems to be absent in the spectral 
line data, despite being present in the continuum...

    Lots of other more minor issues will be reported on later, as time 
permits.