Beyond...Details on FEA modelling.
The Active Surface Control System
This page is intended to give a few details and to provide some background information on the
Active Surface Control System (ASCS) being worked on at the JACH.
Status 29 July 2002 - In the past, changes have been made in the hardware
of the temperature acquisition system and now the software is being
updated (among others it is not known which sensor is where...), hopefully
To improve the image quality of optical and near-infrared telescopes, a number of adaptive optics schemes
are being used to compensate for mirror distorsions and for atmospheric disturbancies. In contrast, radio
and (sub)millimeter telescopes do not normally suffer much from image degradation due to the atmosphere.
Here, loss of image quality and loss of beam efficiency resulting from thermal deformations in the (large)
telescope structures are the problem. Since the mechanical structure of a radio telescope is rather simple,
it should be possible (at least in principle) to calculate the thermally and gravitationally induced deformation
of the main dish once the thermal expansion and gravitational deflection of each structural element is known.
Such calculations can easily be carried out using standard FEA software packages. For this particular application,
we are only concerned with linear expansion and much less with stress. Consequently, a relatively simple FEA model
of the telescope is all that is needed to obtain acceptable results.
The JCMT telescope was build with steppermotors attached to each panel of the surface, making it possible
to remotely adjust the surface. This feature of the telescope makes it possible to correct for thermal distorsions
once they have been calculated, e.g. to "freeze" the surface into its optimal shape. This is the concept of our
A prototype of an Active Surface Control System was tested at the telescope during August 1997.
This was done by measuring
across the telescope structure and calculating the displacements of the surface nodes, and by comparing the
calculated surface with the shape obtained directly from simultaneous holography. This early implementation
however failed to predict the surface shape accurately, partly because of insufficient resolution of the
temperature acquisition system, and partly because the system relied on lookup tables (see below) for FEA
modelling which had their own problems. These problems were overcome early 1998.
The 2 pictures below show the results from measurements made in March 1998.
The figure on the left is the calculated dish profile from FEA (no lookup
tables were used in this case) and the
figure on the right is the profile measured with holography. The scale is the same for both pictures.
Considering that some critical parts of the telescope were still rather poorly modelled
at that time, these early results proved that the concept really worked.
Temperature acquisition system
A new temperature acquisition was installed at the telescope in November 1998 after it was found that the
resolution and stability of the old system was totally inadequate for active surface control. Temperature
gradients on certain sensitive elements of the structure, notably the supporting elements for the backing structure,
as low as 0.1 degrees could be shown to result in significant amounts of distorsion. The new system is
a commercial system obtained from Keithley and performs extremely well. Presently, it consists of 3 switch main
frames (Model 7002) holding a total of 24 Model 7067 4-Wire scanner cards. A Model 2010 DMM is used for measuring
the switched signals from Platinum Resistor Thermometers (PRTs). A total of 200 sensors have been hooked up, but
the number can be expanded with the present hardware to a total of 240. The 2010 DMM converts the measured
signals directly into
calibrated temperatures. The system is interfaced over an IEEE bus.
Resolution and longterm stability of the system is checked during each readout with zero-temperature coefficient
precision calibration resistors that have been substituted for a number of PRTs. Repeatabilily between
subsequent measurements is better than 0.003 C, and longterm drift has been less than 0.01 C over a period of
one year. Sampling of all 200 sensors takes about 3 minutes, but can be done much faster if needed.
The pictures below show the locations of the PRT sensors on the telescope structure. At several locations, a
single PRT sensor has been assigned to multiple nodes, and they appear in the plots as strings of sensors.
With 200 PRT sensors installed, the telescope structure is still greatly undersampled, and a thermal conduction
FEA program is used to obtain the temperatures of the "missing" nodes.
Side view of the telescope. Each red square, or row of red squares, indicates the location of a PRT sensor.
You should be able to count a total of 200.
Front view of the telescope. Each red square, or row of red squares, again indicates the location of a PRT sensor.
Analysis of the temperature data files and FEA modelling is done on a SPARC-5 workstation running scripts.
First step in the reduction process is a calculation of the temperatures of ALL nodes in the FEA model.
The present implementation uses a thermal conduction model for this purpose. Once all the temperatures
are known, a consistency check is performed to flag unrealistic values. The second step is a calculation
of the thermal expansion of the telescope structure using a full-blown FEA analysis. Third step is to fit an
ideal parabolid through the displaced surface nodes. This fit gives us values for the focus change, pointing
offsets in Azimuth and Elevation and values for how much each panel adjuster has to move to compensate for
the residuals (is: difference between the displaced nodes and the fitted paraboloid). This is an important
step in the calculation process since it minimizes the adjuster moves by eliminating "trivial" moves caused
by focus and pointing drifts. This is dramatically illustrated by the 2 figures below for a temperature
change in one of the
elevation drive rings. The figure on the left is the raw FEA result, requesting a large correction of the
surface. The figure on the right shows the residual adjustments which are needed to tweek the telescope back
into a perfect shape
after a simple pointing correction in Elevation is made.
Scale left picture: +/-100 microns.
Scale right picture: +/- 5 micons.
Surface Control System
After fitting the claclulated surface node displacements, the number of steps for each one of the
828 steppermotors for moving the panels is determined. This information is send to a VME microcomputer
which runs the motors in sets of 6 and in 4 sectors (out of 12) simultaneously. Step rate is 40 Hz.
A full dish adjustment takes less than 1 minute. The motor-driven adjuster system at the JCMT is an
open-loop system with no provisions for feedback from the steppermotors. Absolute positions of the
adjusters are obtained from end-of-travel datums and kept in a position file which is being updated
after each adjustment.
Although this system is far from fail-safe, the surface can be recovered from the position files and
from the datums if something goes horribly wrong.
The new VME adjuster control system is essentially finished and will be installed at the telescope in
With the present design, a typical looptime (temperature measurement, FEA analysis, dish adjustment) is
5 minutes, as has been verified by our prototype system which has been place and been used intermittently
for over a year. With this loop time, most adjusters will not have drifted enough for even a single
step, with 10-15% of the remaining adjusters requiring only one step up or down. During periods of rapid
drift, typically one hour after sunrise, and during sunset, a small fraction (5%) may require 2 steps.
The Active Surface Control System can therefore operate continuously without disturbing ongoing
observations. The looptime is, in fact, variable and is adjusted to keep the demand for single steps
within the 10-15% bracket. During nighttime, the system effectively puts itself asleep.
An open-loop surface control system as described above needs periodic checks. This will be done with our
new holography system (See RxH3 news on the Home Page menu) as often as needed. It is not known how fast
errors will accumulate with the system in continuous operation. We expect to make a holography measurement
and to "tweek" the surface back to its optimum shape perhaps once a week, or perhaps once every two weeks.
It must be possible, in priciple, to skip the step where we do the FEA analysis completely, and to
calculate the surface shape from lookup tables. We have made such a table which contains surface node
displacements for 95 individual elements, or related groups of elements. Of the 95 table entries, only
60 or so have relevant contributions, with the remaining entries contributing not more than a few microns
per degree C. The lookup table thus makes it possible to characterize the thermal behaviour of the dish
with only 60 coefficients. The results obtained from the lookup tables are not yet as accurate as those from
a full FEA analysis, but they are being improved. When this on-the-side project is finished, we will have
accompliced two more things: A more fundamental understanding of the thermal properties of the telescope, and
the capability to calculate the surface from temperature measurements in less than a millisecond, not 3 minutes!
The picture below shows a "page" of the lookup table with the stored surface node displacements displayed as small
Lookup table data. Each table entry has the displacement of the 300 surface nodes for a given element
for 1 C temperature difference.
FEA model results and the resulting images for the dish shapes and
beam profiles during a typical day were made into MPEG movies. The quality of these movies can vary from very good to downright poor depending
on the PC or workstation used. Nevertheless, it may be illustrative to see how the surface
shape of the dish deforms during a typical day, and to see the 450 micron beam profile "explode" at mid-day.
Note that the surface rms reaches a plateau value during most of the day, and also note the low noise level of the system
which is apparent from the repeatability of subsequent images. You can choose one out of four animations from the
menu on the Home Page. The shortened version of the 450 micron beam in false colors works best when the quality of the
the movies turns out to be poor.
Time scale for final implementation
Although the software and several prototype versions have been in place for
quite some time now,
delays in the fabrication of a more reliable control system for the stepper
motors have prevented us
from implementing and using active surface control up to this point. This
control system is working in the mean time, and with RxH3 also in use we
can now concentrate on the temperature measurements. Hopefully ASCS tests
can be made again in 2003.
F. Baas, updated 11 December 2002 by: