[evlatests] EVLA X-Band Antenna Gain Stability

[Rick Perley]

    One of the benefits of the flux density run is the ability to judge the
long-term amplitude (power) gain stability of the antennas.  In this
context, 'amplitude gain stability' means the effectiveness of  electronics
to remove the variations in gain and system temperature, so that the
derived flux density of a known point source is the same, hour after
hour, day after day, to some typical accuracy (which is band dependent --
see below). 

    I judge this stability after the following actions:

    a) Clearly dead data ('dead antennas') are removed. 

    b) Data corresponding to scans were referenced pointing has
failed, or some other failure clearly associated with any antenna within
any scan, are removed.

    c) Short-term variations, typically at the beginning of the scan, are
removed. 

    d) Elevation-dependent gain changes are removed (through L. Kogan's
program ELINT).

    e) Opacity variations are removed through application of the known
opacity (measured via tip scans) via the program 'CLCOR'. 

    For X-band, the judgement is made using four standard calibrators,
observed 37 times over the assigned 30 hours.  This year, we were given
an extra ~12  hours (and 12 more calibrator observations) to make up for
losses due to wind.  These extra data were added in. 

    The amplitude stability of the VLA antennas is (to me, at least)
astonishingly good -- and has been since I began this program in 1983.
At X-band, the typical gain variations, following the editing and correcting
effects listed above, are less than 0.5%.   (The same level applies at L and
C bands -- the scatter worses at higher frequencies in a manner consistent
with residual pointing errors -- more on this in a later report). 

    This standard was easily matched in the 2006 data by the VLA 
antennas.  It
has not been met by any of the EVLA antennas -- but some are close. 

    Antenna 13:

    Quite well behaved, but the gain scatter is ~6% (pk - pk).  No temporal
trend (over 5 days) is seen, and the expected elevation gain change is
detected.   IFs A and C behave similarly in the statistical sense, but there
is little correlation in the gain changes between A and C. 

    Antenna 14:

    IFs A and C gains vary by ~5%, in a random way. 

    IF B is terrible -- gain changes of over 200% are seen.

    IF D varies by ~3%.  A fit to the gain elevation dependence gives an
inverted result, but this is based on rather little data. 

    Antenna 16:

    IF A is terrible -- random gain variations of 50% throughout the 5 
days. 

    IFs B and C are very good -- variations are only ~1 - 2 % -- getting 
close
to VLA standards, and are probably due to uncorrected pointing errors --
no referenced pointing was available, while the VLA antennas benefit by
this.   Residual gains are highly correlated between B and C (and D too) --
supporting the residual pointing hypothesis. 

    IF D is a little worse the B or C --- very good but not quite to the 
required
standard. 

    I would guess that the Tsys correction machinery is not yet accurate
enough for antennas 13 or 14, but may be close to VLA standards for 16. 

    I'll show a couple of plots at the Monday meeting, which should help
explain the above.