From: "Kreig McBride" 
To: "Jeff Anderson-Lee" 
Cc: 
Reply-To: "Kreig McBride" 


Question:

Why are we using points???

Why not use "large" points such as 1/4" diameter soft nylon or teflon?

In my 54 point cell, i used the caps found on 35mm film containers. The
contact points are a ring instead of a point. (The inside of the cap.)

The 54 point cell is not in use yet so dont know how it will perform. (experimental)

This is a question, not a suggestion.

Kreig  McBride
Telescope Optics Workshop
March 22-23, 2003
Bellingham, WA


-----Original Message-----
From: Jeff Anderson-Lee 
To: ATM Mailing List 
Date: Thursday, February 27, 2003 6:16 AM
Subject: ATM supporting thin mirrors


>
>With all of the "too big, too thin" talk, it brings me to a matter of
>concern for me: how best to support a thin mirror.
>
>I have a 16x7/8 plate glass blank that I had Dan Cassaro pre-generate to
>f/5.  I went with f/5 because Plop seemed to indicate that I could support
>it with 18 points.  However later analysis shows that may not be so...
>
>First, by the time I get a smooth/polished surface I figure [no pun
>intended] I will more likely be looking at 3/4in of glass, not 7/8in.  That
>still seems to be doable with 18 points -- at first glance.
>
>However being a skeptic of sorts I did a study using Monte Carlo variations
>to look at the effects of mirror cell fabrication errors.  The results were
>not very
>encouraging.
>
>You can see the whole study at
>http://http.cs.berkeley.edu/~jonah/18plus/p18.html but I will summarize the
>results here.
>
>I studied 5 vatiants of an 18-point cell using even angles, varying angles,
>Plop optimized varying force, 0.8 varying force, and refocusing with
varying
>force. Simple Monte Carlo testing on 2mm error showed at least a 3.5x loss
>in average versus design performance with no refocusing and 8.5x loss in
>maximum versus design performance on 100 Monte Carlo runs.  When allowing
>refocusing, the results looked more promising, but that is because simple
>Monte Carlo testing introduces systemic errors that can be refocused out.
>
>Changing the model to reflect more realistic fabrication errors lead to
>larger changes in performance; I had to reduce the tolerances to 1mm before
>I began to see "acceptable" levels of performance once again.
>
>Using varying force ultimately did not seem to help, in that any seeming
>performance increase was lost back in increased sensitivity to errors.
This
>was especially true with the model optomized with refocusing on.  (I
>generally optomize with refocusing off.)  The best case still saw a roughly
>3x average (8x maximum) perfomance loss in implementation (for 1mm errors!)
>Allowing refocusing after the fact (i.e. design without it, Monte Carlo
test
>with it) reduced the spread to about 2x average loss (4x maximum loss).
>
>Variation in balance contributed the most to performance loss.  A friend
has
>suggested using laser-cut stainless steel parts for the mirror cell for
more
>precise fabrication.  However I'm still puzzled at how to join the parts so
>that the balance points do not shift by more than 1mm!
>
>I've heard of using ball-bearing supports for the triangles, but have not
>found a manufacturer/ part number/ distributor that would seem to supply
>workable parts at a reasonable price.  (I would need 6 for an 18-point
>cell).
>
>The alternative would seem to be an astatic cell design.  I have a drill
>press, but no complex machine shop, so any pointers/suggestions for simple
>astatic cell units would be appreciated.
>
>The alternative would be to go for more points of support.  27 or 36 points
>might give more support, but I'm probably still going to be stuck with
>fine-tolerances in cell manufacturing--more than a drill-press alone may
>suffice.
>
>For those who are interested, I have a collection of Plop cell designs of
18
>or more points as well at http://www.cs.berkeley.edu/~jonah/18plus/ which
>includes sixteen designs ranging from 18 to 108 point cells.  The overview
>there describes the collection and notes on the design process used, as
well
>as discussions of why I chose to optimize with "Refocus Error
Calculation"
>turned off and the like.
>
>There are two downloads there for the package: a slim 38KB which includes
>the .gr mesh files, an Excel spreadsheet, and a README file, or a 700KB
>distribution package with the "full deal" for anyone wanting
all the gifs
>and html files for private use or redistribution.  The report on 18-point
>cell optomization is also available there in both .html and .doc format.
>
>Phew.  That's a lot!  for those of you who made it this far, thanks for
>reading.  Any help with cell manufacturing suggestions would be
appreciated.
>
>Jeff Anderson-Lee
>Sacramento, California; 38.5566N 121.4525W
>
>

--- BBBS/NT v4.00 MP