[evlatests] EVLA C-band Sensitivity

[Rick Perley]

    I've looked carefully at the continuum data taken by Ken and me
yesterday afternoon.  These data were taken with 400 ms averaging,
and 50 MHz bandwidth on a 2.1 Jy calibrator (quite strong).  We also
looked at blank sky to see if the noise characteristics there were the
same as on the calibrator. 

    a) Pointing was done to determine the offsets.  Antenna 14 was
mispointed by 1.9 arcminutes, (loss of 12% in power), antenna 16
by 0.4 arcminutes (negligible loss).  I have adjusted the results below
to account for this.

    b) The loss of coherence due to the mismatched bandpasses was
easily seen.  For the VLA-EVLA baselines, the loss in gain is
by a factor of 0.79.  Because the AIPS calibration of antennas
14 and 16 is (almost) entirely based on these baselines, the
14 x 16 baseline amplitude, after calibration is a factor
of (1/.79)^2 too high = 1.52.  (It varies slightly between the
IF pairs).  I have adjusted the results to account for this. 

    c) The overall amplitude and phase stability (that is, on
timescales of minutes) is the same on 14 and 16 as on
nearby antennas.  All seems well here.  There may be small
'clunks' here and there, but they don't stand out. 

    d) Antenna gains were determined on the 400 ms data.  There
is nothing to distinguish antennas 14 and 16, other than the
higher scatter.  (See below). 

    e) Channel 14B has the 'stepped' amplitude problem.  As this
is certain to be a T304 issue, which is being addressed elsewhere,
I'll not comment on it further. 

    f) The most interesting data are the amplitude and phase
fluctuations.  After appropriate calibration, and correction for the
gain bias of 14 and 16 (as noted in (b), above), I find the rms
amplitudes and phases to be, at a minimum, 50% higher on
baseline 14 X 16 than on a good VLA baseline. 
    For example:

    IF 'D' is by far the best of the four parallel-hand combinations
between 14 and 16.  (14B has the gain problem, 16C has an
oscillating gain of 15% amplitude and 40 second period, 16A is
about 20% noisier than 16D).  The rms noise on 14Dx16D is
0.100 mJy.  Presuming both antennas are identical in sensitivity,
this translates into an SEFD = 607 JY, which is reduced to
570 Jy after accounting for pointing.  For 16A x 14A, it's about
20% worse (i.e., 685 Jy). 
    A good VLA baseline gives an rms of 62 mJy, or an SEFD of
375 Jy.

    What *should* we see?  For the VLA, with an efficiency of
60%, and a Tsys = 40, we should get an SEFD = 375.  (These
are the best values we can expect, and I selected the best baseline
I could find for the measurement).

    For the EVLA, we expect efficiency of 60%, and a Tsys of
24K or so.  Hence, SEFD should be 225 Jy. 

    Bottom line:  The EVLA antennas, operating as well as we
can make them, have a sensitivity 2.5 times worse than
they should be, and 1.5 times worse than the current best
VLA antennas. 

    The rms in the phase is in line with what we would predict
from the amplitude rms.  So we are not 'resolving out' any
hypothetical fluctuations in phase at 400 msec sampling. 

    The reported system temperatures for antenna 14 was
40 K (backend Tsys), a little better than most VLA antennas,
but far from the 23K we hope for (and which was being reported
a few days ago).   The change from 23 to 40 has been seen in
the data, and occured in less than 10 seconds.  No explanation
has been offered...

    Perhaps we should check the efficiency, via observations of
a strong source.  This is easy to do. 

    Rick

PS  I will try review the L and X band data over the weekend.