[evlatests] Referenced Pointing Successes ... and Failures!

[Rick Perley]

    The elevation tracking tests done last Saturday morning offers much 
insight into the referenced pointing effectiveness, and the tracking 
stability of our antennas.  In general, very good news.  As always, 
there are exceptions too. 

    The test started (at midnight) at an elevation of 8 degrees and an 
azimuth of ~62 degrees (to the ENE).  Most of the antenna motion for the 
next 5 hours is in elevation.  For the final two hours, until source 
transit, the azimuth moves much more, while the elevation changes 
little.   It is important to note that the weather for this run was 
ideal, with no wind, clear skies, and small temperature changes. 

    Referenced pointing was done every ten minutes throughout the 7 hour 
run.  For the great majority of the antennas, and throughout most of the 
run, the solutions are very consistent, with changes nearly always less 
than 5 arcseconds, and usually less than 2".   Plots of the elevation 
and azimuth solutions (attached) show gradual changes reflecting small 
residual pointing errors.  Some antennas (notably ea13) have significant 
offsets -- since a pointing run was done the previous night, and the 
results applied, the origin of such offsets remains a mystery. 

    Examination of the gain solutions at all frequencies (including 48 
GHz!) show excellent continuity in amplitudes over *most* of the run.   
These results indicate that our general recommendation of doing 
referenced pointing approximately hourly, using a source within 15 
degrees of the target, should suffice to remove residual pointing errors 
to a level of 5" or better.  (Note:  a 5" offset at 48 GHz results in a 
power loss of about 3%). 

    As always, the exceptions provide the most useful information.

    1) The pointing solutions for many antennas 'go wild', generally 
within 30 minutes of transit.  And the amplitude gains -- even with 
these wild solutions applied, are poor, indicating the solutions found 
don't actually reflect the offsets when the observations were taken 
(generally, 5 minutes after the referenced pointing was established).  
The origin of this problem is well known -- the antennas were above 80 
degrees in elevation during this period, and many of our antennas simply 
don't point well when we are so close to the zenith.  Some antennas are 
notably worse than most, and some (remarkably!) point perfectly well.  A 
short list:
    Notably bad:  5, 7, 8, 11, 12, 14, 15, 17, 23, 25, 26
    Notably good:  2, 3, 6, 9, 16. 

     Based on this one source, even the worst antennas still point well 
at 80 degrees elevation -- it is only above this that things go bad. 

    2) On occasion, a solution is found which is 10 arcseconds or so 
different than the preceding, and following, solutions.  There is 
excellent evidence from the Q-band gain solutions that the solution is 
bad.  A good example is with ea11, which shows two unusual solutions in 
elevation -- both are matched by measurable drops in forward gain.

    3) But by far the most peculiar issue is with ea02.  Examination of 
the pointing results shows one major and one minor deviation, in 
elevation, from the smooth curve.  The larger deviation comprises 6 
consecutive solutions (covering more than an hour) suggesting a 20 
arcsecond offset in elevation.    If these offsets are real, there 
should be no observable effect in the gain solutions at any band.  But 
there is -- and it's not small ...
    There is no measureable effect at C or X bands -- but I wouldn't 
expect it in any event, as 20 arcseconds at X-band only provides a 1% 
gain drop.
    There are very large effects in gain seen at K, Ka, and Q bands.  
The gain drops, however, don't match the size of the pointing error in 
any way.  Below is a table of the expected power loss (from the 20 
arcsecond offset), the measured power loss, and the pointing offset 
which would cause the measured loss:

     Freq.         Expected         Measured         Real Offset
---------------------------------------------------------------------------
       19.5 GHz        0.97             0.75                  45 arcsec
       25.5                0.90             0.66                   47
       29.5                0.87             0.80                    25
       36.5                0.81             0.65                   28
       42.0                0.75             0.25                    45
       48.0                0.70             0.14                    46
----------------------------------------------------------------------

    These numbers are rough estimates -- an accuracy of 5 to 10 
arcseconds can be assumed. 
    The power loss at Ka-band is close to that expected from the actual 
measured offset.  The losses at K and Q bands indicate a much larger 
offset was in place -- more than twice as large. 
    The only rational explanation that I can think off is that the 
'subreflector rotation' application failed.  But it had to have failed 
in a band-dependent way, and it didn't 'fail' in an on-off manner -- the 
plots of the gain loss mirror exactly the pointing solution offsets. 
    No other antenna showed this effect.  And on the test done the 
preceding week (this one of a source setting), ea02 behaved perfectly 
well.