[evlatests] Temperature-Sensitive Gains at Ku-band

[Rick Perley]

    A major justification for the 'Flux Redux' run (taken last 
Thursday/Friday) was to judge whether thermal isolation of the 
calibration noise diodes at Ku-band is sufficient to remove the strong 
temperature dependence of the post-calibration gains. 

    I've attached an example of just how strong this is, taken from the 
May 02/03 flux densities run.  Shown are the residual gains over 30 
hours for two antennas.  The 24-hour periodicity is very easy to see, 
with pk-pk amplitude of about 4% -- an 8% variation in power.  This is 
far above the EVLA spec. 

    I emphasize here that this is the variation after application of the 
switched power.  Were the switched power monitor not applied, the 
variation would be very much larger -- about 20%, pk-pk! 

    It is believed that the residual gain variation is due to 
temperature-induced change in output noise power from the diode.  the 
primary driver for the temperature change is believed to be a 'heat 
pipe' effect -- the diodes are mounted with good thermal conduct to the 
Ku-band horn, which extends well outside the vertex room.  One antenna 
-- ea15, had its noise diode thermally isolated in the May 02/03 run, 
and indeed its residual variation is much less -- barely notable at 
all.   This is suggestive, but not proof, as there are a couple of other 
antennas which have much smaller variations as well. 

    To better judge whether thermal isolation is sufficient, three other 
Ku-band receivers had their noise diodes thermally isolated -- ea05, 
ea12, and ea16.  We then ran the 'Flux Redux' program, which extended 
over about 17 hours (from a bit after 5PM to a bit after 10AM), to see 
if there was a notable reduction in time (temperature) gain sensitivity.

    Unfortunately, the results are inconclusive.  ea15 seriously 
misbehaved on this later run (with a very strong linear gain change of 
some other origin).  Of the other three, two seemed to show no change in 
gain vs. time, while one seemed to show a temperature-like variation 
remaining. 

    I use 'seemed' in the description, since the results are hard to 
judge, for two main reasons:

    1) The night chosen for this had a much reduced temperature swing, 
compared to the May 02/03 run.  This was largely because we had the 
misfortune (or bad judgment) of observing on a very windy night.   The 
high winds also cause a larger scatter in the amplitude solutions (but I 
don't think this is a major problem for these Ku-band data). 
    2) Identifying a weak 24-hour period in a 17-hour span is difficult, 
even with stable conditions.  A longer run would have been better. 

    I don't recommend trying this experiment again -- (I have too much 
data already,  and temperature swings in summer are much reduced 
compared to winter.)  It is probably best now that lab tests be done on 
these diodes, to better understand their temperature sensitivity.  


-------------- next part --------------
A non-text attachment was scrubbed...
Name: UGainVar.png
Type: image/png
Size: 63868 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20130605/1d4cd3f8/attachment.png>