dsp motion controller needed for telescope

Dan McKenna wrote:

Go to the Matlab (i.e. Mathworks) website and see what hardware they
support for real time control. I'm sure you'll get just what you need.
(I've seen demos but can't remember all the details). The DSpace
solutions are good but very expensive and probably way overkill for
what you want. Look at some of the cheaper solutions.

Fred.

Fred,

Thanks for your input.

Yes, I have and am using the matlab support site to find supported
hardware. I never considered Dspace as I prefer a US based company
for support.

On the other hand, Dspace has what I need and may be what I use as the
I/O is modular and in a rack format.

In the past I have used ARCS, and Bitware for my control systems, both
found by using the Matlab site. The Bitware system is up and running
and was developed with Matlab. The ARCS system is now a portable system
used for system ID and as a controller playground.

My goal here is to see if I missed something.

Thanks again

Dan




Fred Stevens wrote:

Dan Mckenna wrote:

   ...

See what Danville Signal Processing, offers at
http://www.danvillesignal.com . They make modular I/O also.

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Jerry Avins wrote:

Maybe also http://www.bittware.com     (603) 226-0404 x536

Jerry
--
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯

Yes,

I used Bittware for one system and the SDL product for Matlab.
These a both fine companies with excellent support.

I had to use a DSP programmer to get all working with a LINUX system
and am looking for a solution that can be totally done in Matlab including
our encoder IO. The Bittware system used an FPGA for the encoder I/O
which I will do again if necessary but trying to avoid. This way I can do
it all
as I am not a real programmer.

Thanks again

Dan

Jerry Avins wrote:

Thanks Jerry,

Went to Danville and they do not have the digital quad encoder interface
that I need.


Jerry Avins wrote:

Dear Dan,

last summer in Glasgow me and my friends Toomas Erm (Caltech, formerly ESO)
and Bertrand Bauvier (ESO) have presented a paper at SPIE Conference on
Astronomical Telescopes and Instrumentation, in which we showed a slight
improvement in control of elevation position of a VLT telescope using
H_infinity based loopshaping:

@InProceedings{erm_2004,
author = {T.~Erm and Z.~Hur'k and B.~Bauvir},
title = {Time to go $\mathcal_$?},
booktitle = {Proceedings of SPIE International Symposium Astronomical
Telescopes},
year = ,
address = {Glasgow, Scotland, UK},
month = ,
organization = ,
}


If you are interested, and have no access to the proceedings, I can send you
a copy.

Best regards,
Zdenek Hurak



Dan McKenna wrote:


--
======================================================
Zdenek Hurak
Center for Applied Cybernetics, Czech Technical University in Prague
Karlovo namesti 13/E, 12135 Prague, Czech Republic
Tel.: +420 22435 7683, Fax: +420 224916648
Email: z.hurak@c-a-k.cz, Web: http://ar.c-a-k.cz/hurak
======================================================

Zdenek,

Thanks for the info and I will look up the paper.

Did you test the disturbance rejection ?

Is the increased performance offset by stability margin ?

Was tuning done by hand ?

Did you use friction compensation ?

Thank you for your time

Dan McKenna



Zdenek Hurak wrote:

Dear Dan,

Dan Mckenna wrote:


In fact, our major motivation for this adventure of using the H_infinity
stuff was to make the bandwidth higher, to attenuate the effect of the
disturbing torque induced by the wind buffets over larger frequency range.
Finally we succeeded in stretching the bandwith a bit (but the physical
limits did not let us go too far). I don't remember exactly, but it was
something like one or two Hz, with the spectrum of the wind buffets being
significant up to 1 Hz or so.


No. With H_infinity-optimization based loopshaping, you have a direct
control over the shape of the frequency response and robustness
requirements can easily be included. In fact, with the existing cascaded PI
control, the magnitude frequency response must start roll off pretty soon
(I mean, at low enough frequency), to guarantee, that at about 8Hz, where
the slowest structural resonant modes appear, the amplification of the
input signal by the system is low enough. As all these new optimization
methods like H_inf usually give controller of higher order, the magnitude
frequency response can stay high quite long and then start roll off
steeply.


Well, in fact, the whole computation was done in Matlab and Simulink, we
could only compare the performance by simulation, using a fairly precise
model. As Toomas Erm (control engineer, formerly with ESO) said during his
presentation in Glasgow: "we have demonstrated that further improvement in
disturbance rejection for positioning the VLT is still possible. The
question is, whether this slight improvement is sufficient motivation for
the bosses to give us go-ahead with this new controller".


No, the two guys from ESO claimed that friction is no considerable problem
with these huge and expensive VLT machines.

--
zh

Zdenck,

Thanks again for your information and time.

Depending on who I talk to, the performance gain of higher order controllers
as applied to telescope control systems is marginal to significant. The VLTs
are very impressive, well engineered systems that might represent the peak
of 8 meter class technology. I met Toomas Erm once and he seemed to be
very competent and knowledgeable and so I conclude that at least in the case of
the VLT,
application of H inf results in only a small gain in performance.

At the moment it seems that I can use simulink/matlab to interface and design
with a few
high performance DSP systems that will allow me to explore any number of
control strategies.
In addition, using the computing power of these systems, I can program the
complete control system including
the tracking vector generation and system telemetry all done in simulink. For
safety we will have
a small PLC as an observer that will monitor limits (position velocity and
acceleration). The independent
saftey observer was used in a large telescope control system here and has
proven it worth.

Dan


Zdenek Hurak wrote:

Dan,

thanks for letting me know about your work. I am ready to share some
"advanced control-computational" skills exactly as I did with Toomas. Keep
in touch if you are interested in such a collaboration.

Zdenek


Dan McKenna wrote:


--
======================================================
Zdenek Hurak
Center for Applied Cybernetics, Czech Technical University in Prague
Karlovo namesti 13/E, 12135 Prague, Czech Republic
Tel.: +420 22435 7683, Fax: +420 224916648
Email: z.hurak@c-a-k.cz, Web: http://ar.c-a-k.cz/hurak
======================================================

(my apologies - since I work here you can consider this first part a bit
of a plug ...)

here is an alternative to a dSpace system at a fraction of the price:

http://www.quanser.com/industrial/html/solutions/fs_power1.asp

This thread is interesting due to our discussions with the NRC (and
related parties in North America) regarding the VLOT (now TMT) project.
Jennifer Dunn had their controller simulations take into effect wind
loading on the flexible telescope structure, but they were only assuming
PI control I believe.

Anyway - that's my random two cents.

(P.S. - Hello Dr. Hurak!)


Dan McKenna wrote: