ATM lies, damn lies, and mirror tests

[Bratislav Curcic <epabcc@epa.ericsson.se>]

I too noticed the increased traffic on 'revolutionary' mirror tests.
It is easy to sit back and let the imagination fly, but that won't make
a successful testing procedure. Just take for example 'laser pointer
running on the track perpendicular to mirror's axis'. There are quite
few FUNDAMENTAL shortcomings of such test, not to mention difficulty in
execution (how do you assure that laser pointer won't tilt say 1/10
wave on its track - we're talking NANOMETERS here ! Just breathing on
the jig will induce temperature differences more than enough to twist
it much more). Designer proposes to use frosted glass to see
aberrations by looking at displacements of in-focus image! An ideal f/5
mirror is capable of creating a spot which is less than 7 microns
across. A 1/2 wave junk one will do just the same. How do you
distinguish between these ? Don't forget that focused laser spot will
be :
a) large (you'll be using effective f/1000 or longer, so 'in focus spot' 
   will in fact be some MILLIMETERS across; the smaller laser beam you
   make, the larger efective "Airy disc" will be)
b) very bright; a large area of the glass will be so brightly lit in 
   fact by scatter that you will see just about nothing

Designer laconically writes :

> 7) Now move the laser so it strikes the mirror at various predetermined
> zones. If the spot on the screen does not move, the mirror is a paraboloid.
> If the spot does move, then that particular zone deviates from the desired
> curve. This can be measured and and the slope of the mirror corrected in the
> usual way.

Did you actually try to 'measure' slope in such way ? A lateral 'dot
movement' of IN FOCUS image representing say 1/4 wave error is going to
be VERY small (don't forget how large in focus spot will be). How do
you propose we measure this displacement, let alone 'eyeball' it on the
focus screen ?

And on and on. The method is so full of fundamental flaws that it is
not worth listing them all here.

Foucault test is an ingenious method of amplifyng mirror errors
millions of times. It uses measurement of longitudinal aberrations at
center of curvature (easily measured in fractions of a millimeter) to
measure errors at mirror surface whose dimensions are fraction of
wavelength of light (measured by NANOMETERS). To date this test firmly
stays alone as single most important tool for evaluating optics during
manufacture in BOTH amateur and professional hands. It has been
elaborated upon gazillions of times in French, German, Russian, 
Japanese, Chinese ... you name it. It has been used to produce and 
verify many fine optics in last few centuries. It is not perfect, but
its genial simplicity and natural presentation makes it choice test 
for most TNs world wide. There aren't many mirrors around that are 
finished without being peeked upon with good ol' KE. And many, many
optics documented with flashy graphs and impressive 3-D plots have 
in fact been finished using Foucault (don't forget that tests using
auxiliary optics to achieve null are basically still variant of 
Foucault).

I'm not trying to rain on anyone's parade, but if you write something
here, be prepared to defend it. I'd in fact suggest to prospective
inventors that there is a very appropriate venue for such ideas. A host
of magazines and proceedings dedicated to optics (not to forget
venerable ATMJ) will be quite happy to be first to publish about new
and revolutionary testing method. But you better back it up with
something a bit more substantial than armchair optician's commentry;
namely a proper mathematical apparatus (with extensive error analysys) 
and preferably 'real life' experiences, backed up with cross testing
by other, already established method(s). Some call it a scientific 
method. 

- just to preventively address any flames of 'this is ATM list, not 
a new committee for new KECK' nature -

Amateur telescope making it may be, but testing of optics MUST be
scientifically based. There may be some black magic involved in
_making_ ultrasmooth, close to perfect optics, but there's NOTHING
magic in testing them. It is very boring, very scientific and very
exact. And as such, it _has_ to comply with certain minimal requirements
in order to be accepted and used as 'new testing method'.

off the 'box,
Bratislav