[ATM] CZ

["James Mulherin" <jcmulherin@opticalmechanics.com>]

Hi Scott:

 

I just read Carl Zambuto's most recent post. I'm a little offended at being
called a fool. Since I'm not allowed to respond directly to Carl I'll
respond here.

 

Here's what I said that makes me a fool:

A Strehl ratio .999 based on Foucault measurements is over optimistic.

Foucault ignores asymmetric errors, so even the P-V reading can be over
optimistic.

 

In the very post where Carl calls me and others fools he agrees with
everything I have said in the past but with his usual spin in favor of
Foucault testing. I've pasted his message below. Compare what he says with
what I've said in the past and you'll see that he's parroting others in the
know now because his old arguments are indefensible after all that we've
learned recently in the open discussions on the other groups. The only point
Carl and I still disagree on is his claim that you can measure astigmatism
by measuring across different diameters with Foucault. As you mention in
your response to Carl, your astig term changes drastically with time due to
air. You can sit there and watch it change real time on the interferometer
monitor. The larger the mirror the more dramatic the effect because you are
testing through a larger volume of air. The only way to get to the bottom of
the astigmatism question for any particular mirror is to average a large
number of data sets to reduce the effect of random astigmatism and other
asymmetrical errors in the air column. 

 

There's nothing magical in the Foucault test that eliminates this practical
problem in testing large mirrors.

 

I agree with you that avoiding astigmatism is a matter of good process and
that most asymmetric errors seen in interferometry are due to test stand and
air. As you know we take steps to reduce these errors in our analysis but we
stop short of turning off all asymmetric terms. We rely partly on averaging
data to reduce their effects on the test results. If we did turn off all
asymmetric terms I'm sure interferometry might match Foucault pretty well.
Some of Steve Koehler's and your work has shown a strong indication that
this would be the case. But then, I think the results may end up being over
optimistic. In other words, every mirror might end up with a Strehl of .99+.
I choose to leave in as much of the data as practical to give giving a fair
representation of the mirror's wavefront.

 

We've talked about some simple methods of checking for astigmatism with the
Ronchi grating (fringe rotating or swirling through focus) on the atm_free
group. This is a test that we perform constantly as we figure and test
mirrors. This is a good qualitative test that will detect significant
problems with astigmatism. John Abrahamian's thin 16" plate glass mirror was
a very good example of how this test can be used effectively. In John's case
the astigmatism was severe. Based on my experience with auto-collimation
testing the threshold of astigmatism delectability is low enough that, if
the mirror looks good in the test stand it will perform well on the sky. For
lack of similar experience with ROC testing I can't say if this test has the
sensitivity required to make the same assessment in the Foucault and ROC
Ronchi test. Apparently it doesn't because Carl has shipped several
astigmatic large aperture mirrors to his customers (although he won't admit
it.) 

 

I would like to reiterate a point that I have made often. Skilled people
routinely make very fine mirrors using only the zonal Foucault test. All
along I have only argued that the results of this test tend to be
overoptimistic. I'm not attacking the test as invalid or useless. I'm only
saying that it has practical limitations. The test can not tell you
conclusively that your mirror has a Strehl ratio of .999. Neither can
interferometry. I have frequently pointed out the limitations of our own
interferometric test results and will continue to do so to avoid
misunderstanding. Every test has it's limitations. Fortunately, in the right
hands the tests are more than accurate enough to make very fine mirrors.

 

Best,

 

James

 

James Mulherin

Optical Mechanics, Inc.

jcmulherin@opticalmechanics.com

www.opticalmechanics.com

Tel: (319) 351-3960

Fax: (319) 351-3943

 

 

Carl Zambuto writes:

Scott,

 

I absolutely count on you to bring up the tough subjects. If I ever decide I
don't want to deal with tough questions anymore, I'll know who to send my
cousins to visit (okay, its not as funny as chuckling in your coffee, but I
tried).

 

The subject before us now is comparing peak to valley measurements between
the two tests. I know you understand this already, so I'm writing to the
readership. This is a subject I probably never got to in our test method
agreement thread of long ago, and that subject was never finished.  But here
we are again.

 

The peak to valley surface or wavefront result of an interferometric test is
based on a two dimensional measurement. It is measuring the entire surface,
with each respective data point representing that one data point locally.
However the peak to valley wavefront result of a knifedge test is based on
single, or one-dimensional measurement, along a straight line where each
data point represents a zone which extends around the entire mirror. They
are two different measurements which represent two different things.

 

One spot on a mirror will affect the PV measurement of the interferometry
output. If we have one small hill that is say 1/4" in diameter  existing on
only one place on the mirror, it will affect the total PV result by its
amplitude. But it will not significally affect the performance of the
mirror, not in the ratio of its PV rating, because that one spot is
representative of only that one spot. The PV rating of an interferometer
consequently has little to do with the mirror's performance, because it is
not an indicator of the mirror's overall, averaged spherical aberration.

 

But each knifedge plot or data point in the knifedge test represents a zone
which extends around the entire mirror, as in a ring. Yes, we also measure
multiple axes to check for figure of revolution and will average the
measurements in cases, for the most accurate representation of the overall
surface. That is why a good figure of revolution is paramount for a knifedge
test to have validity.

 

So the bottom line PV numbers of the two tests are two different animals.
They cannot be compared in any meaningful way. This is why I have said in
the past on this forum that a typical difference between the PV measurements
on a good figure of revolution between interferometry and Foucault may be as
much as 2.5x. That is a typical, although not consistent, ratio.

 

I recall one of my public crucifixions on the Obsession mirror group some
time ago. OMI did an interferometric test on one of our mirrors that had a
knifedge PV that was very accurate. It was measured at about 1/50 wavefront
for averaged spherical aberration, close to the limit of the knifedge's
capability in the best of circumstances. The OMI interferometric report put
the mirror at 1/8 wave. So they really went to town on that one. They ranted
and made a big deal of how important it was to have "real" data on your
mirror. Sorry, folks, they don't know what they're talking about. Truth is
they made fools of themselves because they don't understand that basic
aspect of optics testing as I've just explained above. If they had looked at
the comparable Strehl ratios they would have had a more realistic picture of
meaningful comparisions.

 

There is a lot more to this, and we will get to it over time. The value to
compare is the Strehl ratio. And we will deal with that to some degree I'm
sure, during the course of the subject.

 

Now here's a little side-note to keep things interesting- The PV as stated
by the interferometer is *not* an indication of what you will see in the
star test. And even more interesting, the PV as put out by the knifedge *is*
a very good indicator. The reason is, the PV of the knifedge is the
*averaged spherical aberration as measured over the entire optic*. If you
have an interferometrically tested mirror that has a PV of 1/4 wave, that
tells you little of what you will see in the star test. It is probably a
good mirror, but you don't know where that 1/4 wave error is. It could (and
probably will) be one single bump anywhere on an x-y picture of the surface
of that mirror, and it will likely have little effect on the overall
performance. But if you have a 1/4 wave mirror as tested by the knifedge you
will see 1/4 wave of spherical aberration in the star test. It could be
anywhere along the radius in any zone or combination of zones, but you
*will* see 1/4 wave of spherical aberration.

 

Scott has mentioned that if we could have the unwrapped phase data and
Zernike coefficients from the interferogram we could make more realisitic
comparisons. Indeed we could. What I typically do is look at the 3-d
isometric surface profile of the mirror, to see where the hills and holes
are that go all the way around the mirror (spherical

aberration) and what their amplitude is on a graph. This one looks pretty
rough, and as was pointed out by Mark S might be more about air currents
than mirror surface. So we might be out of luck on this particular instance.

 

I'll stop for now. Much more to come later, I'm sure.

 

CZ

 

Scotts response:

Very well put, Carl, on all points.  Your description of what the knife edge
sees as being "averaged spherical" aberration, and the better correlation
between this and the star test vs, the overall P-V error is, I think very
concsiely stated, and true when the fabricator's process has evolved to
indepdendently control astigmatism (as yours has, from the examples that you
have shown me in the past).  

 

I would only add that the comparison between 1D and 2D methods could perhaps
be made more meaningful if the 2D method (i.e. the

interferogram) were to have been analyzed with all of the non- rotationally
symmetric aberration terms removed from the data.  

Perhaps if this had been done, OMI's answer might have more closely matched
what you got by using the knife. 

 

Just considering the possibility of astigamtism in the data, for example,
complicates the 2D analysis quite a bit, because astigmatism can have
multiple sources, only one of which is actual astigmatism "in the glass"  As
you already know, this is a separate thread itself, so I won't go there,
except to say that it has not been uncommon in looking at live video images
of interferometry done in my own shop to see variations between
interferograms separated only in time (i.e. 

same mirror, same set-up, same orientation), show changes in the astigmatism
of 1/4 wave P-V or more.  Under these conditions, even the RMS, and hence
the calculated Strehl ratio, will be negatively affected if only 1 frame is
used.  Averaging multiple frames to mitigate atmospheric effects is
therefore mandatory for obtaining the most accurate results.  Bob Royce has
a nice discussion of this topic on his website, for anyone folowing this who
wants a more in depth discussion of how averaging reduces random noise in
interferometric testing.

 

Calibration of the interfeometer for spherical aberration, however, is a
relatively straighforward thing, so long as the interferometer optical
layout creates a real (as oposed to a virtual) focal point.  

To keep everyone with me, what I'm talking about here is that the beam that
the interferometer sends into the optic under test should converge to a
point focus in air and then diverge to fill the optic.  

If such is the case, then placing a small, plane (flat) mirror at this focus
point will effectively "self test" the interferometer optics for all orders
of spherical aberration, as such a mirror functions as an effectively
perfect reference over the few microns aperture that the beam sees.  I
routinely perform this test on my own interferometer when evaluating new
diverger objectives before using them in critical optical tests.  Hence, I
feel I have a very solid understanding of the "error bars" for spherical
aberration that my interferometer introduces, and in fact, can easily
subtract the small residual errors from the test data, if required.

 

I'm sure you would agree that it all comes back to the same thing; you can't
get the best result from any method by just "flipping switches"; you have to
think about what you are doing, and how you will overcome the weaknesses of
whatever test method you are using.

 

Am I really having this conversation at 10:30 P.M?  Time to snooze.

 

Scott

 

 

 

_______________________________________________
ATM mailing list http://www.atmlist.net/