[evlatests] Modcomp-Free Tests, 18 May

[Rick Perley]

    I ran my standard 'torture test' of the system Friday afternoon.    
I've carefully reviewed the L-Band data from this, and the top-level 
summary is that, once a few rather gross problems are identified and 
removed, the situation doesn't look hopeless at all!  Indeed, there may 
be reason for optimism ...

    Read on for some details. 

    The test was in two blocks:

    A)  23 minutes in 'Modcomp-ON' mode, observing first in continuum 
(std. 50 Mhz, 0.418 sec averaging), each band from L through Q, then in 
spectral line (mode 4, 12.5 MHz BW, 3.3 second avg). 

    The Modcomp-free system was then installed (takes about 5 minutes)

    B) The same 23-minute file as above. 

    Major results (from L-Band only):

    I filled the data with all flags turned off, so I could see 
everything, no matter how bad.  I haven't yet filled with the flags on 
-- that effort may have to be delayed a bit, due to time pressures from 
other activities. 
    At first review, the Modcomp-free data looked just dreadful.  But it 
was fairly quickly found that once two major issues were identified and 
removed, data quality looked pretty good:

    1) Antennas 3 and 7 appeared to be near-dead on all IFs, all 
baselines, both in line and continuum -- so the problem is not a delay.  
In review the baseline-spectra, it was quickly seen that the phases are 
oscillating back and forth by 180 degrees on alternative channels.  
(This will very neatly kill the coherent signal!) 

    2) In spectral line mode only, all the X-baselines between EVLA and 
VLA showed nonsense spectra, particular in phase, where channel-channel 
deviations are tens to hundreds of degrees, with no recognizable pattern. 

    After removing these effects, the standard bandpass and calibration 
procedures worked well.  The following items were noted:

    1) In continuum, several antennas required very high gain 
corrections for calibration.  In all cases, these were due to very large 
delay errors in the Modcomp-free system.  Antenna 18 is particularly bad 
(~170 nsec), but many others need better delays. 
    2) A high amplitude spike is seen in channels 2 and 3, in IF B -- 
this is the frequency of one of the GPS signals (which, back in NVSS 
days, came on for a few minutes every few hours).  The odd thing is that 
it doesn't appear to be in IF 'D' (LCP). 
    3) There was 20 seconds of bad data, affecting all antennas, in the 
middle of the spectral line data.  No suspected cause. 

    Interestingly, the Modcomp-controlled L-band data showed 
considerably worse amplitude stability than the Modcomp-free.  I doubt 
this has anything to do with the presence or absence of the Modcomps, as 
the effect was strongest at the center of the array, and the two 
antennas at the end of the west arm (17 and 22, at W64 and W72 
respectively) are not affected at all.  I'm speculating this is a 
weather related issue, as the four IFs on any antenna wander or jump 
about identically.  This problem was definitely not an EVLA issue, as 
both types of antennas behaved similarly.   Whatever the cause was -- it 
was gone by the time the Modcomp-free observations were taken. 

    By calibrating the bandpasses (to remove the delay errors), I 
checked the antenna sensitivities, via ANBPL.  The Modcomp-free 
sensitivities were much better than the Modcomp-on ones by a large 
margin, for nearly all antennas.  Notable exceptions were antennas 5 and 
17.  I doubt the overall difference is relevent -- the poor performance 
of the Modcomp-on sensitivities is very probably related to the poor 
gain stability already noted above. 

    Peculiar behaviors were noted on some antennas and IFs -- these are 
in all cases independent of which computer was in charge:

    We are seeing many 'short drops' on ALL antennas.  In all cases, 
these affect all four IFs simultaneously and identically.  These were 
much worse during the 'Modcomp-on' data. 
    Some antennas behave very badly, over and above the small drops 
noted above:
       Antenna 4 is very bad -- with very large, short-duration drops in 
amplitude.
       EVLA antenna 18D is showing multiple amplitude levels, which have 
large phase changes as well.  The effect is unlike any of the 
multiple-level problems I've reported before.
       EVLA antenna 26C also has multiple amplitudes -- but here there 
is no phase effect.  The structure of the amplitude effect is unique. 
    The preceding two problems are seen both with Modcomps on and off. 

    Antenna 12B and D has rapid 5 degree oscillations in phase -- seen 
only with Modcomps on. 

    I'll return this evening to repeat these checks on the X-band data, 
and perhaps the C-band data if time permits.