[ATM] Re: Second Thoughts on Mirror Support

[atmpob at yahoo.com (Dale Eason)]

--- Jeff Anderson-Lee <jonah@eecs.berkeley.edu> wrote:
> Jeff Anderson-Lee wrote:
> 
> Suppose instead of a THICK steel plate you use a
> very THIN plate--even 
> sheet metal.  You sandwhich three layers: the metal,
> some plush 
> carpeting, and an underlying support (e.g 1.5 inches
> of cured/dried 
> plywood sealed against humidity changes).  Use a
> torque driver and 
> screws to make sure all of the connecting bolts have
> the same tension.  
> Put a central hole in the sandwhich so that air can
> be forced between 
> the mirror and the back plate using a small muffin
> fan. Then, RTV the 
> mirror to the metal using multiple, small area pads.
> 
> 
> The hope is that the sandwhich would give similar
> all-over support as 
> the carpet used to grind/polish the mirror on, yet
> have airspace for 
> forced ventilation, while the RTV avoids the need
> for a sling.
> 

You are mixing to different processes here.  The
carpet used to grind/polish the mirror works because
we repeatedly pick up the mirror and put it down in a
different orientation.  Thus using a semi random
process to even out the high areas as we work the
mirror.  This process will not work for a stationary 
mirror in the telescope.

This discussion has forgotten first principals.  That
is when you place two relatively flat surfaces
together they will only touch at the three highest
points between them. Because three points define a
plane.  You can try make more points but it will still
be only three most of the time.  Thus your nine point
cell made on a flat plate will only be using three of
the points.  

Try this experiment.  Take three ball bearings the
same size.  Place them on a slightly rough surface
like a weathered board or a brick (this will simulate
the metal plate at the scale of roughness we need to
control to.) Place them in a triangle pattern.   Now
lay a flat piece of metal or wood on top of them to
simulate the mirror back. This will be a very stable
platform.  The top piece will not rock in relation to
the bottom.  
Now add a fourth ball bearing between the two
surfaces.  Doe the top plate rock up and down now?  It
probably will.  It will rock between two different
groups of three ball bearings.

 Add more ball bearings of the same size.  Does it
touch more than three at a time?  Probably not.

You can try adding some RTV to the ones that don't
quite touch.  Can you still make it rock with the RTV?
Probably but not as fast.

Now this was all done at a different scale than what
we need for our mirror cell.  It needs to not rock on
the order of less than 1/10 the wave length of light.

A whiffle tree arrangement can do this, has been doing
this for years and is easy to make.

Two flat plates (a mirror back and the cell) and more
than three RTV globs can not do it.  You can't get the
plates optically flat enough to know what three of N
points are going carry the load.  So if you need more
than three support points a whiffle tree is your
better bet.


Why do people spend so much time polishing a mirror to
exacting standards and then expect to hold it with
devices that are not up to the same standards?

Why does it work for the scope in the previous email
example that started this discussion?  Maybe because 3
points are enough for that thick of mirror.  Or
perhaps it is not good enough.  The poster said the
mirror was tested at 1/5 wave on the wave front. 
Perhaps the errors in the mirror swamps out any error
caused by the mirror cell.  Perhaps no one who looked
through the scope actually did a star test where it
would show up.  



Dale Eason




__________________________________
Do you Yahoo!?
Yahoo! Finance: Get your refund fast by filing online.
http://taxes.yahoo.com/filing.html