[evlatests] EVLA Performance

[Rick Perley]

    Following the resynchronization, and corrections of various problems 
yesterday, I ran a pair of test to establish current performance. 

    I ran two one-hour tests, one with Modcomps-on, and one with 
Modcomps-off. 

    The Modcomps-off test failed completely -- no data made it to the 
archive.  There is reason to believe the correlator produced data (as 
there were not notes in the operator's log to signify any problems), so 
I suppose the pipeline between it and the archive was broken.  Experts 
will have to peer in. 

    The Modcomps-on test produced excellent data from all antennas.  All 
10 EVLA antennas produced stable data, at least at some bands and IFs.  
There are numerous issues (mostly of a minor nature) which I summarize 
below. 
    Tests were done at L, C, X, and K bands, both line and continuum.  
The continuum data provides the following information.  (Delays and 
sensitivity are derived from line data, which I will reduce later). 

    I apologize for the length of this report.  The problems noted below 
have very clear signatures, which I hope can lead to speedy 
resolutions.  I have plots to illustrate these, for those who (like me) 
prefer pictures to words. 

    A) Global Issues  (seen at all bands). 

    1) Antenna 23 produces good data *only* on IF B.  The others give 
only noise. 

    2) Antenna 21 gives no data in LCP (IFs C and D). 

    3) Antenna 19 shows a perfect 10-second square wave in phase on IFs 
C and B only.  The phase jumps by ~105 degrees -- at both L and X-band 
(no data was present at K or C bands) -- every 5 seconds.  IF 'B' and 
'C' are in perfect anti-phase -- the jumps are simultaneous, and in 
opposite directions.  There is no intermediate phase state, and no 
corresponding amplitude effect.  Whatever is doing this is in perfectly 
synchronism with the 416 ms integration, or is doing it after the 
integration is completed. 

    4) Antenna 24 also has square-wave phase jumps, on IFs B and D.  The 
character of these jumps is quite different than for 19:
          * The jumps are not regular, and are infrequent:  6 are seen 
at L-band (in 7 minutes), 11 at C-band (in 9 minutes), 4 at X-band (in 7 
minutes), and 7 at K-band (in 7 minutes).  The durations are always 10 
seconds, except that two of the jumps at C-band lasted 20 seconds. 
          *  The jumps are of the same magnitude at all bands:  70 
degrees. 
          *  The jumps are of opposite sign at X-band -- decreasing by 
70 degrees, while at all the other bands, the jumps were positive. 
         *  There is always an intermediate phase in the middle of the 
jumps -- the duration of the jump is longer than the sample time (0.417 
seconds).  The expected drop in amplitude for the intermediate phase is 
seen.  However, at  C-band, the 'intermediate' phase point is very close 
to one of the two 'stable' phase values.  Hence, the apparently 
excellent synchronism of the mechanism with the integration cycle works 
differently at this band. 

    5) Antenna 21, IFs A and B show a remarkable sinusoidal phase, with 
period about 1 minute, and amplitude which scales with frequency:  100 
degrees at K-band, and 30 degrees at X-band.  It is not visible at 
L-band (but I would predict an amplitude of only 4 degrees there, which 
makes it too small to see with the 'delay clunk' going on).  Nothing can 
be said about IFs C and D, as they were dead at all bands. 

    6) The 'phase parabolas' seen at long spacings are present again.  
The characteristics are:
          *  The parabolas have a 10-second period.  All four IFs behave 
identically. 
          *  The amplitude is proportional to frequency:  (40 degees at 
K-band on antennas 17 and 18, and 15 degrees at X-band, 7 degrees at 
C-band, and not visible at L-band).
          *  The amplitude is proportional to baseline length (from the 
array center).
          *  The amplitude and sign of the error (i.e., whether the 
parabolas point 'up' or 'down') are given by the derivitive of the 
U-component of the baseline.  The parabolas look like upward-facing cups 
when the derivitive is positive. 
          Is it possible that the 2nd derivative of the phase is not 
being used in the phase calculation?  The residual phase looks like only 
the phase and phase rate are being applied over a ten-second period. 

     7)  There are the usual 'short shallow' amplitude drops, always 
seen in opposite polarizations together, but independently between the 
two IFs.  They are very infrequent (typically less than 1% of the 
integrations), and small (a few percent in amplitude), with no 
discernible effect in phase.  They are always single -- 0.416 second -- 
drops.  There is never any discernible pattern or frequency.  Some 
antennas (17, 24, 26) are much more likely to show these.  Two antennas 
(14, 16) never show them. 

    B)  Antenna-Specific Problems. 

    1) Antenna 26, IFs A, C, and D, is misbehaving at C-band.  
Amplitudes are far too high, and very unsteady.  IF B looks normal. 

    2) Antenna 19, IF B is dead at L-band. 

    3) Antenna 19 gave no fringes at K-band, on any IF.