[evlatests] Results of ea21 Pointing tests (part 1) Corrected

[Barry Clark]

There are two structural resonances, a torsional one and a rocking
one (the antenna is shaking his head or nodding it).  Both are at
slightly above 2 Hz (I think they were made about the same to
minimize weight).  Rick's results suggest that it is slightly easier
to excite the torsional oscillation than the nodding one.  That is,
the +Az -> -Az step produced more 2Hz than the +El -> -El step.
I have no guess why this might be.  The size of the oscillation
(30" peak-peak corresponding to about 1.5 mm motion at the edge
of the dish) doesn't appear serious to me, though it would be nice
if it could be made to die out a little more quickly than the two
or three seconds it seems to take it.  At a guess, we are likely
to be at a point that fiddling the parameters to minimize the
excitation of the oscillation is likely to cost more in arrival
time than it's worth.

It might be interesting to see what the tracking looks like in a
moderate wind, say 10-12 mps.

For a long time I was an advocate for the GBT memo 203 and related
methods - kick it in the butt as hard as you can to get it moving
and deal with the consequences later - but as Greg pointed out in
his initial presentation here, we want to be kind to the machinery
as well as to get on source quickly.  I don't think they are worth
exploring in this context.


>
> On 01/04/2014 10:03 PM, Gregory Maglathlin wrote:
>> Rick,
>> Thank you for the quantitative information.  I will talk to this data
>> (referenced pointing mode)  first as it is
>> the easiest for me to assimilate.
>>
>> First, the reduced overshoot in the new ACU is due to position
>> pre-processing that is performed on the
>> commands provided by the MIB.   The position (and velocity) loops have
>> limiting integrators which will
>> saturate if allowed.  The position pre-processor of the new ACU is
>> designed to turn any step function (which
>> of course cannot be realized) into an S-curve which does not tax the
>> integrator in the system preventing
>> saturation and hence overshoot.  It is good to see that this is working.
>>
>> Second, I am assuming that the 2 Hz oscillation is the fundamental
>> resonance of the antenna structure.
>> Controlling the acceleration allowed in the position pre-processor has
>> reduced the excitement of this
>> frequency.  There are a few methods to further reduce this oscillation.
>> Two come to mind immediately.
>>
>> First, reduce the allowed acceleration in the pre-processor.  This is
>> very simple and can be done by changing
>>     a parameter in the configuration display at the ACU.  This will cause
>> each axis to get to the desired position a little more slowly
>> but should reduce the oscillation at the end.  We should be able to find
>> a sweet spot here, which may be antenna
>> dependent.  Antenna dependency is not a problem as the ACU is designed
>> to have an antenna specific configuration.
>>
>> Second, I have always been intrigued by a paper that Ken introduced me
>> to (GBT Memo 203) when we first met years ago
>> regarding impulses at the beginning of a move counteracted by impulses
>> through the move to
>> counteract any oscillations.  This would be tricky to  implement
>> (especially considering the ACU/MIB interaction)
>> but warrants further thought.  The fact that the ACU is running both
>> velocity and position loops at 1 KHz may make
>> this a realizable option.
>>
>> Of course, I will study this data some more and move on to Part 2.
>>
>> Best regards and Happy New Year,
>> Greg
>>
>> _______________________________________________
>> evlatests mailing list
>> evlatests at listmgr.cv.nrao.edu
>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests