Re: ATM Rumak Fabrication Tolerances

[Aplanatic@aol.com]

Bratislav expounds:

> Aplanatic Dave wrote :

Well, there are 14 million AOL accounts and every one of us poor souls has to
have a unique 10 character alphanumeric "name."  So, I could have been
dave5438234 or some such nonsense.  Actually, aplanatic fits me well, my eyes
require no spherical correction and I have never been in a coma.  I do,
however, have 2 diopters of astigmatism....

>  I can't offer you a generic proof, but think this way : a menisk
>  introduces certain amount of (negative) spherical aberration. There
>  will ALWAYS be a spherical mirror (or set of primary-secondary) that
>  has exact amount of SA to cancel. Yes, you may need to change some
>  parameters (you may end up with f/14 or f/16 instead of targeted f/15,
>  but so what ?).

I tried this on the Rumak and even with small changes in R1, could not find a
solution that was reasonable.  Don't know, but if one choses the approach of
"tuning" the mirrors, one better be dead sure that a reasonable solutions
exists.

> The important thing is achromatization; and for that
>  R1-R2 MUST be spot on. You can fiddle with residual spherical by
>  retouching one of the surfaces (even secondary, if you REALLY insist
>  :-). Once the color creeps in, it can only be fixed with #120 carbo.

I find that the corrector can be substantially off from designed values of R1
and R2 (0.2%) without introducing a lot of chromatic aberration.  The severe
problem is spherical aberration.
 
>  If you use a precise spherometer, steep radii on the meniscus actually
>  HELP us to reduce the error. Unlike on slow mirrors, saggita error on a
>  Mak is almost directly proportional to a radius change. If your
>  spherometer isn't good enough, use one of the proposed methods in books
>  I mentioned (use microscope on the Foucault apparatus to observe image
>  at ROC and compare that with a set of spacers/rods). You can measure
>  BOTH sides this way. Concave is measured directly, and convex is
>  tested THROUGH concave. Displacement is easily calculated and taken
>  into account, the only tricky thing is to learn to ignore reflections
>  from first surface (takes practice but it is no problem at all). This
>  way you can easily measure radii to within 20-30 microns, or even
>  better if you're careful.

Yes, I now understand.  This method is superb for this particular problem
because, as I mentioned before, it is the differential error in R1, R2 that
really causes great grief.  Since your proposed measurement can be fashioned
as a very accurate differential measurement, the severe problem of spherical
aberration can be overcome by this technique.  Using this method relaxes the
absolute requirements on RoC accuracy when executed correctly.
  
>  Mind you, this is an opinion of a twisted minded, frustrated TN who's only
>  pleasure in life comes from finding minute wavefront errors :-)

Whatever makes you happy. <g>

-------------
Peter John Smith wrote:

>A possible approach is to back test when figuring the corrector and measure
>the DIFFERENCE of the image positions from both surfaces.  

>One formed only in air and the other through the glass reflected back from
>the convex side.

Yes, this is exactly the right approach, as I see it, as Bratislav noted
above.  Great minds think alike (yours, not mine!).

>If the tolerance is 0.03 % this means RC chances about 0.1 mm.

>The spacing of the KE positions change from 12.249 mm to 12.397

>The convex side takes an aspheric of -.000308 which degrades Rumak image
>very slightly but it is still very good if it is back figured.

I'm not sure I understand this point.  Are you saying that when figured
through the concave side, there is an aspheric component impressed on the
convex side?

>This gives a RELATIVE test keeping the meniscus surfaces linked.

Yes, absolutely the key point, making the fabrication possible.

>If a small tube is turned and faced to exactly the correct length and a
>thread stretched over each end
>this could possibly be used like a wire test.

A very nice idea, requires an accuracy of maybe 0.2% if the two surfaces are
linked, as above.  Certainly doable.

Now that the council of the wise (not me!) has found a way to make the Mak,
everyone should be running out to their local BK7 store, stocking up on
supplies.

Dave Rowe
Torrance, CA