[evlatests] Separating Temperature and Elevation dependencies

[Rick Perley]

    Some previous tests have shown there is a strong elevation or 
temperature dependence of the observed visibilities.  The 'or' is 
present because these past tests were taken when both were changing 
simultaneously, making a clear separation difficult. 

    The Thursday night test was designed to allow a clear separation.  
We observed a far northern source (J0019+7327) through meridian transit, 
at a time when the temperatures dropped dramatically.  Specifically:

    At the start (4:20 PM MST), the source elevation was 41.7 degrees, 
and the external temperature was 11C. 
    At meridian transit (8:17 MST), source elevation was 50.4 degrees, 
and the external temperature was  -2C.
    At the end (11:19 MST), the source elevation was 45.0 degrees, and 
the external temperature was -5C. 

    Over a 3 hour period centered on meridian transit, the source 
elevation changed by less than 1.5 degrees, while the outside  
temperature dropped monotonically by 5C. 

    Referenced pointing (at X-band) was done every 10 minutes.  We 
cycled through the eight observing bands, spending 1 minute at each. 

    Results:

    As expected, the demodulated Switched Power monitor (PDif) rose 
throughout the observing period -- for most antennas at all bands, the 
rise in switched power reflected the outside temperature (in reverse -- 
as the temperature dropped, the monitored switched power rose).  
However, the rise in switched power is a *dramatic* function of 
observing band.  Below is a table, showing the fractional rise in 
swithced power:   Value = [PDif(end) - PDif(start))/PDif(start)].

    Band        Value              Comments
----------------------------------------------------
    L              2 -- 3%           all antennas the same
    S              2 -- 3%           all antennas the same
    C                7%               all antennas the same
    X              5 -- 7%              some variation present
    Ku           12 -- 22%        median is about 15%
    K             15 -- 30%        median is about 20%
    Ka            10 -- 22%       median is about 15%
    Q              8 -- 25%        median is about 10%
----------------------------------------------------------------------------
   
    It is notable that the most affected bands are Ku and K -- not Ka 
and Q.  I'm speculating that since the former two have their horns 
exposed to the night air, while the latter two are protected by weather 
windows not attached to the horn, the larger variation at Ku and K bands 
is due to a 'cold pipe' effect.  This argument is especially true for 
Q-band, where the horn itself is well down from the roof. 

    These PDif variations are most certainly due to gain changes, as the 
visibilities faithfully replicate these changes.  Application of the 
PDif values to the visibilities does a very nice job in *completely 
removing* the gain changes.  The maximum remaining change in visibility 
amplitude over this time (and temperature) range is no more than 2%, and 
in most cases, no slope at all can be discerned. 
   
    Despite this success in PDif removing the changes, the remarkably 
large magnitude of the gain changes is discomfiting.  Although the 
temperature range on this night was fairly large, it was by no means as 
large as many others we see -- changes of up to 30 degrees C have been 
seen, especially in the spring.  We might expect gain changes well in 
excess of 50% due to this -- which could induce changes in bandpass and 
polarization from the long lengths of cable between the front-ends and 
the IF/sampler racks. 

    Keith was monitoring the temperatures, and other variables, within 
the antennas during this period.   It will be interesting to see what 
the internal temperatures were doing during this period.