[evlatests] Modcomp-Free data.

[Rick Perley]

    Perfect, it ain't.  But I think there's hope...

    I ran a 30 minute test, with the usual modes, with Modcomp-free, 
then the same file, with Modcomps in charge, 30 minutes later. 

    *** Special note to Ken and Walter:  The data we were looking at 
were in fact the Modcomp-on data!!! *** 

    I reviewed the fast-dump continuum data this evening for L, C, and X 
bands, for the Modcomp-free data.   There is a wealth of different 
phenomena.  But some commonality appears:

    1) The system is slow to get going.  At C-band, the first 15 seconds 
for all antennas was dead.  At X-band, it was 28 seconds.  At L-band, 
it's hard to judge, as other curious phenomena (seen also with Modcomps 
on) dominate. 

    2) For all three bands, all gains and phases discontinuously jump 
10.0 seconds before the end of the scan (as judged by the presence of 
data).  EVLA antenna amplitudes jump a lot more than VLA antennas.  
There is no commonality between antennas or even IFs.  Every antenna-IF 
has its own 'personality'. 

    3) There is a curious pattern of dead antennas, which only partially 
(and weakly) reflects which antennas were actually working.  
    At L-band, 14 and 16 were dead on all IFs.  Antenna 4 had low 
amplitudes. 
    At C-band, the pattern was really bizarre.  All IFs for antennas 1, 
16, 20, 22, 23, 25, 27 and 28 were dead.  We know 23 has no receiver, 
but the others?  Antenna 21 has no receiver but gave (weak) fringes on 
IF A.  In addition to these, 10D, and 26 A, C, and D were dead. 
    At X-band, 16C, and 14 B and D were dead.  Amplitudes are low on 
some other antenna-IFs. 

    4) Some antenna-IFs gave ramp-like gain changes!  (Looks like the 
delay 'clunk', but in fact these are amplitudes).  I have plots to show 
the experts. 

    5) There are large step-changes in phase on some antennas.  These 
are tens of degrees, and of short (a second of two) in duration.  The 
phase returns to the original value in each case.  (This is not a common 
phenomenon, but is certainly easily seen). 

    6) After removing the first ~20, and last 10 seconds, and flagging 
the dead and weak antennas, and ignoring the odd amplitude ramp and 
phase hop, the remainder of the data (by far the majority) look pretty 
decent. 

    Despite the litany of ills, this is much better than what I've seen 
before.  But not prime-time yet ...