Circular Arch Isolators as Static Supports
Warren Davison, Steward Observatory
John Hill, Steward Observatory
Bruce Hille, Steward Observatory
Large Binocular Telescope Project
Technical Memo
UA-96-01
April 11, 1996
http://medusa.as.arizona.edu/lbtwww/tech/ua9601.htm
Abstract
The static supports for the telescope mirror are there to rest the mirror on when it is not
operating. They have to take the full weight of the mirror at any telescope angle including horizon
pointing. They also take upward push, and random earthquake forces. They must have a clear band
where they do not touch the mirror during operations. They must have a low spring constant so they
provide equal forces on the mirror, and a high spring constant to limit the travel so the mirror or its
components do not touch solid objects when excess force is applied. This paper describes circular
arch isolators which we have selected as static supports for 6.5 and 8.4 m borosilicate honeycomb
mirrors. The circular arch isolators we tested met all of our requirements but did have an annoying
feature, that is a dead band in the loading that requires you to set them carefully at installation. Their
best feature however is that it is very hard to overload them in a manner that hurts the mirror.
Description
The circular arch isolator is made from a base plate with eight 1/4" stainless aircraft cables
that arch to a top plate (figure 1).

It has several desirable characteristics for static supports and a couple undesirable ones. The
desirable properties are: it is made of materials that will last the life of the telescope, it is very
nonlinear so it uses up a small portion of our travel budget and is nearly impossible to overload to a
dangerous stress level, the axial and lateral spring constants are in the correct range, it has a lot of
damping, it is low cost, and it fits in the space available. Its faults are: it has an initial hysteresis
curve with lost motion of 0.5 mm axial and 1 mm lateral, it work hardens when subjected to the same
loading (it however partially resets this hardening when a different load is applied). It meets all of
our requirements for a static support and the faults can be overcome.
The top of the basic unit is usually bolted down so in our application the stiffness properties
are different from the catalog values. The axial stiffness is lower because the top can twist as it is
compressed. For the 100 lb gravity load we anticipate the axial deflection for the first cycle is 1.6 mm and for repetitive cycles is 0.9 mm. In the free state the lateral stiffness is a function of where you press on it. If you press at the height of the plate it pivots up as it translates and the spring constant
is too low. If you press at 20 mm higher the plate is nearly flat as it translates and the spring constant
is appropriate. The height we require to match our cell is this elevated height. The hysteresis is a
classic loop and is a result of the friction in the cables, for repeated loads it is very repeatable and
predictable. There does not appear to be any jumps or discontinuities in any of the data. When we
subjected the unit to repetitive cycles there was a distinctive hardening, where the maximum
deflection was reduced and the zero load position was increased. The slope at load was increased
by almost a factor of two. Figures 2 and 3 show some test results.
Requirements
Based on Parodi's analyses of local stresses in the glass (BCV Report #160), the maximum
allowable lateral force from a static support is 850 N and the maximum axial force is 1130 N. The
mean gravity load on each static support is 422 N for the 6.5m. The difference between the mean
load and the maximum allowable force is what gives us tolerances on positions, spring constants, etc..
We have reconstructed the historical division of force tolerances in the following tables:
(these apply to all types of static supports)
Lateral force errors:
- from CTE mismatch between mirror and cell (x1.2)
- from position and spring constant variations (x1.3)
- from over-travel of dropping the mirror (x1.3)
Total lateral force errors (1.2x1.3x1.3=2) and (2x422=850)
Axial force errors:
- from CTE mismatch between mirror and cell (x1.01)
- from position and spring constant variations (x2.0)
- from over-travel of dropping the mirror (x1.3)
Total axial force errors (1.01x2.0x1.3=2.6) and (2.6x422=1100)
We can now combine these force tolerances with the maximum allowable travel in the
actuators to generate position tolerances for the static supports and load spreaders. (Or, we are
going to calculate the spring constant of the static supports given certain assumptions about the
position tolerances.) If the deflection of the static supports under a 1g load is 3 mm, and we want
no more than 30% variation in the force, then the tolerance on the positioning stack-up from one
support to the next is 1 mm. Given the allocation of travel in Table 4 of "Mirror Support II/R"
(UA-95-02) 3 mm is the largest 1G deflection that the actuators can handle. If the deflection of the
static supports under a 1G load is 1 mm, and we want no more than 30% variation in the force, then
the tolerance on the positioning stack-up is 0.3 mm. This is why you want soft static supports --- to
avoid very tight positioning tolerances in the lateral direction. Even if you had perfect positioning,
the CTE mismatches drive you to soft supports. We also note that since the allowable axial forces
are higher, we are allowed an x2 variation in the axial force; that is from zero to twice the nominal
load (provided that the variations are random across the mirror so as not to make global stresses).
Now, to allow for devices with non-linear spring constants, let us define a pseudo spring
constant which tells us the deflection under a 1G load. We'll adopt the units of millimeters/100# of
load. Let's call this value D --- remember, this is the deflection at 100# regardless of how you got
there. The maximum value of D is 3 mm since that's all the travel we have allocated for the
compression of the static supports. In the lateral case, the minimum value of D is also 3 mm since less compression gets into stress problems from CTE mismatch (20% variations around the nominal 3 mm value are permitted). In the axial case, smaller values of D are permitted, where the minimum is set by the range of distance over which all the static supports contact the mirror as the mirror is raised or lowered. The historical contact allocation is 1.5 mm, thus Dmin is 1.5 mm.
Let's also define a more traditional spring constant K, which is the slope of the force vs.
deflection curves at a loading of 100 pounds. (We'll be tricky and express it as the linear deflection
caused by a 100# load so it has the same units as D.)
Properties of the Circular Arch Isolators
Now let us look at the measured properties of the circular arch isolators based on the
measurements Bruce has done.
| Condition | D | K |
| Axial First cycle | 2.1 mm |
| Axial New after 4 cycles | 1.7 mm | 3.5 mm |
| Axial after 20K cycles | 0.95 mm | 1.3 mm |
| Exercised 20000 lateral cycles |
| Axial first cycle | 1.3 mm |
| Axial after 4 cycles | 0.9 mm | 1.6 mm |
| Axial then 100 more cycles | 0.8 mm | 1.3 mm |
| Axial after 40K total cycles |
| then 200# load | 7.1 mm | 30 mm |
| Lateral New | 2.8 mm |
| Lateral after 20K cycles | 2.4 mm |
The lateral deflection matches our 3 mm ± 20% even after 20000 cycles, so no problem there.
The tricky part is that the axial deflection is low so there is potentially a problem relative to the
contact distance allocation. The nice thing is that the isolators apparently continue to soften so that
a deflection or order 10 mm is needed to get the force up to 200# (love those mixed units!). To
control global stresses, we just want to be sure that the deflection of the mirror cell is not more than
half of D_axial (< 0.5 mm), and the discussion is easy so long as the total contact stack-up is less than D_axial (< 1 mm). The decreasing D with lifetime seems to be due to lack of rebound caused by increasing friction rather than a continued sagging of the device. This means that the devices
compensate for initial position errors as the age --- quite a convenient property.
Mounting
The mounting holes in the baseplate (as delivered from the factory) do not allow enough
translation to meet our requirements. The desire to adjust from the bottom of the cell would be
greatly facilitated if we had captive nuts in the unit. To solve both problems we have designed a large captive washer shaped piece with six .25-28UNF-2b tapped holes on a 2" bolt circle. This will clamp down the base of the static support with an adjustment range of plus or minus 3 mm. The top of the Isolator will have a 1.06" hole to receive a specially machined "chase nipple". This will allow vertical adjustment. We will need to accommodate the vertical hysteresis by compressing the unit and then adjusting the height. The horizontal hysteresis can be taken care of in the cell by shoving it over with a bar at the time of alignment.
View Figure 2 here
View Figure 3 here