[ATM] How many zones

[jonathan_bishop at sbcglobal.net (Jonathan Bishop)]

Hi,

Thanks again for the details. I gave sixests a try and your explanation was
right on.

Another question. Have you tried the program from Andreas Reifke called
obviously "Foucault Test Analysis". It seems quite nice as it allows you to
log successive readings and has an easy way to plot. One downside is that it
is implemented using VB, but it is pretty simple to download VB runtime from
his site at the same time you down load the program.

I am curious what you think of it.

Thanks again,

Jon Bishop

----- Original Message -----
From: "Mark Holm" <mdholm@telerama.com>
To: "Jonathan Bishop" <jonathan_bishop@sbcglobal.net>
Sent: Thursday, February 19, 2004 6:57 PM
Subject: Re: [ATM] How many zones


> Jon,
>
> You might be surprised how quickly the paraboloid figure can develop.
> Limit yourself to certainly no more than 10 minute figuring sessions
> till you see how it is progressing.  Let the mirror stand on the test
> stand at least 1/2 hour before testing.  You need that, and perhaps
> more, to let it settle down thermally so you will get a true measurement
> of the surface.   Having a fan blowing on the mirror could certainly
> help the temperature equilibration, though not to many atm's do that.
>
> I think Tex (the program) uses a method very similar to that in
> Texerau's book.  If you use Sixtests, you will have the benefit of more
> sophisticated math working for you, but it isn't any more difficult for
> you to use.  Sixtests, is available in an older DOS version and a
> Windows version that is a simple .exe file with no auxiliary files to
> download (my VB based programs certainly are not like that!).  Sixtests
> presents three graphs.  The first one, which presents deviations from
> the best fit parabola is the one that is closest to Texerau's.  It also
> shows the deviations from the best fit conic section and deviation from
> the best fit sphere (or any other conic section you wish to enter).  The
> best fit conic section plot is useful, because it gives you the "b" or
> conic constant value that best fits your mirror in its current state.
> The conic constant for a sphere is zero and for a parabola it is -1.
> Negative values between zero and -1 are "under corrected" ellipses.
> Values lower (more negative than) -1 are hyperbolas (over corrected).
> It helps you know where you are at.
>
> Sixtests fits a curve to your error numbers.  It is pretty good, and I
> have found that the error curve usually seems to match what I am seeing
> in the shadow patterns pretty well.   If there is a narrow zone though,
> that isn't sufficiently sampled by your mask zones, Sixtests may miss
> it. (True of any Foucault analysis program.)  With five zones in your
> mask, a zonal ditch or raised ring would have to be pretty narrow for
> Sixtests to miss it entirely.  The only exception is a narrow turned (or
> raised) edge.  Your masks can only sample on the inside of the edge, and
> that makes it more likely that Sixtests will not quite get the curve
> right near the edge. (This is true of any Foucault analysis program, not
> just Sixtests.)  When I saw that I had a turned edge, I made a special
> mask with two zones half the normal width, one right up to the edge and
> the other immediately inside that.  These extra zones straddled the
> outer zone openings in my main mask and gave better data on what was
> happening right up to the edge.  The extra narrow zones were rather
> difficult to read.
>
> It can be useful to use the third plot in Sixtests in a slightly non
> obvious way.  First set it to a fixed conic constant of -1 and let it
> find the best fit ROC.  The result should be the same as on the first
> plot.  Then, you set both the conic constant and ROC to manual control.
> Leaving the conic constant at -1, you make small changes in the ROC and
> observe how the error plot changes.  It will always get worse!, but
> sometimes the worse curve will actually be better from the point of view
> of having something you can correct with known polishing techniques.
> Texerau suggests essentially the same method in his section on
> correcting figuring errors.
>
> One minor quirk about Sixtests: it presents the errors in nanometers
> instead of waves.  You can convert using the formula
> error in nM / 550 = error in waves
>
> Most of us are used to thinking in 1 / waves  The formula for that is 1/
> ( 550 / error in nanometers).  In other words, if your error is 110
> nanometers, it is 0.2 waves or 1/5 waves.
>
> The other thing about Sixtests is that (for very good reasons) Jim
> Burrows pushes the use of RMS rather than P-V error.  To get the P-V
> value that many atm's still use, and as described in Texerau, you just
> have to scale it directly off of the graph.  The graph vertical axis is
> always in nanometers.  You can do the scaling easily to plenty good
> enough accuracy just by eyeballing it off your computer screen.
>
> The quarter wave (at the wavefront) Raleigh criterion translates into
> about 1/28 wave RMS error at the surface, about  20 nM.  1/8 wave at the
> wavefront (a quite good mirror) translates into about 1/56 wave RMS
> error, about 10 nM.
>
> The Strehl ratio, calculated from the RMS value, is even an easier
> measure of quality.  The Raleigh criterion (1/4 wave at the wavefront)
> is about equivalent to a Strehl of 0.8   Strehl = 0.95 is about
> equivalent to 1/8 P-V wave at the wave front.  Sixtests displays the
> Strehl for you, the one you want is  on the first plot, the one with
> conic constant, b, = -1    Don't worry about the Z coefficients.  That
> is getting way beyond what a beginner, or most more advanced atm's,
> needs to worry about.
>
> Mark Holm
>
> Jonathan Bishop wrote:
>
> >Hi,
> >
> >Thanks for the detailed response! Since I am just starting I will
probably
> >keep it simple and use five zones. My last measurements show the maximum
> >zonal error was 1/.8 wave! So I have a long way to go before I need more
> >precise measurements.
> >
> >By the way, I am having trouble printing the graphs created by TEX. The
> >method described in the README does not work for me (I am running windows
> >XP). Do you know of a data reduction program which runs directly in
windows
> >and is able to print teh type of graphs I see in TEX?
> >
> >Thanks,
> >
> >Jon Bishop
> >
> >
> >
>