92 replies to this topic

[MKV]

The point I disagree with is about how seeing overwhelms any of these minor defects....

The atmospheric effects include scintillation (sparking), turbulence (dancing), and changes in air density (due to temperature) in small pockets traversing the aperture.

 

Texereau talks about in his book, and I think I've read somewhere (can't recall where) that after about 1/10 wave PV atmospheric effects dominate over optical errors in telescopes on earth making it pointless to pursue correction beyond that parameter. Thus, pursuing correction for astronomical optics past that marker has diminishing returns for telescope on earth.

 

A 1/10 wave RMS OPD residual with a smooth mirror should be about 1/35 wave RMS and a Strehl in the realm of about 0.97. I think it's reasonable to call it quits after that, because no significant improvement will be noticed if we continue..

[BGRE]

How does Roddier stack up against interferometric results? Isn't it still subject to atmospheric variance, like any star test? How does one standardize that issue? Why is Roddier not more popular?

It's, a little less accurate than interferometry due to the implicit double integration required to go from the Laplacian to the wavefront.

In testing with a diffraction limited artificial star in a null test it compares well with a Bath at CoC with a numeric null.

The effect of seeing depends on the Fried parameter and the inner scale of turbulence. Low altitude turbulence can be helpful. 

[Gleb1964]

..
I am not sure how sensitive that method is, especially for slower mirrors, and for very fast mirrors it may not be sensitive enough. Can you tell if an F4 is a 100% parabola or 98%? I think star testing doesn't allow such precision. 
..

 

The sensitivity of star testing is not affected by the speed at which the wavefront converges (or how slow/fast is system).

 

When performing a star test with an eyepiece, the exit pupil produces an approximately flat wavefront with a small, controlled defocus (defocus = wavefront curvature).

 

The quantitative Roddier star test measures the curvature of the wavefront using a camera sensor. The sensor is defocused to capture star images on both sides of focus. Long integration times or image stacking help average out atmospheric fluctuations.

[MKV]

The sensitivity of star testing is not affected by the speed at which the wavefront converges (or how slow/fast is system).

 

When performing a star test with an eyepiece, the exit pupil produces an approximately flat wavefront with a small, controlled defocus (defocus = wavefront curvature).

 

The quantitative Roddier star test measures the curvature of the wavefront using a camera sensor. The sensor is defocused to capture star images on both sides of focus. Long integration times or image stacking help average out atmospheric fluctuations.

Thanks, Gleb. My question was in reference to Pinbout's description of how amateurs test mirrors using the "breakout" of the secondary obstruction shadow. 

[MKV]

In testing with a diffraction limited artificial star in a null test it compares well with a Bath at CoC with a numeric null.

What would be the advantage of testing with the Roddier test using an artificial star in a null configuration in the workshop? It requires more glass (eyepiece), and that means more potential equipment error contributions that need to be subtracted.. 

[BGRE]

The Roddier test uses a camera without a lens.

No eyepiece required.

As such its optically simpler than the star test when done by eye.

It works best near null.

Can be useful for testing surfaces such as an off axis part of a prolate spheroid without nulling optics.

A bath or Fizeau or Twyman-Green cannot do this.

A PDI with an external point source can.

There is also a geometric wavefront test that uses the same images processed differently that claims to measure the wavefront slope rather than its curvature.

 

A Roddier test can be used as a sanity check when using an interferometer to test a surface.

It often requires less equipment than an interferometer.

If all you have is a camera (and a point source) you can often measure wavefront errors using the Roddier test.

[MKV]

The Roddier test uses a camera without a lens.

No eyepiece required.

As such its optically simpler than the star test when done by eye.

It works best near null.

Can be useful for testing surfaces such as an off axis part of a prolate spheroid without nulling optics.

A bath or Fizeau or Twyman-Green cannot do this.

A PDI with an external point source can.

There is also a geometric wavefront test that uses the same images processed differently that claims to measure the wavefront slope rather than its curvature.

 

A Roddier test can be used as a sanity check when using an interferometer to test a surface.

It often requires less equipment than an interferometer.

If all you have is a camera (and a point source) you can often measure wavefront errors using the Roddier test.

Super! It's a shame not much is known here about it. I think it's time for a thread on the Roddier test. Do you know anyone who can and is willing? Are there any down sides to the Roddier method?

[BGRE]

The software uses a different enumeration/labelling for Zernike polynomials than the Wyant enumeration used by DFTFringe.

The (Noll) Zernike polynomials used are orthonormal, Wyant Zernikes are just orthogonal.

Conversion between Wyant Zernikes and the Noll Zernikes requires scaling that differs for each polynomial.

For some, the fact that the primary documentation is in French.

However, there is a good American English translation available.

It's less accurate than interferometry.

It has lower spatial resolution than interferometry. 

 

That said, if you are setup for a star test then you can do a Roddier test if you have a suitable camera.

The required camera resolution is relatively low so that a cheap camera can be used.

Pixel binning and ROI delineation can be used with higher resolution cameras to emulate such a camera.

 

Testing a small diameter spherical surface at its CoC using the Roddier test would be a simple way to become familiar with the test.

A fixed source together with a camera moved between intrafocal and extrafocal positions are required. 

 

A small refractor objective setup in autocollimation would also work as a testbed.

Once sufficient familiarity has been gained one can move on to testing with an artificial star at a suitable distance and to using real stars (likely requires tracking during exposure). The advantage of a testbed is that the results can be compared to those obtained with other methods such as interferometry etc.

 

The test optic need not be perfect especially if it can also be measured using a Bath or other interferometer.

[MKV]

Thanks, bruce! I think dedicated followers of the star testing method would benefit greatly from this. The transition doesn't seem to be too complicated. The software does all the number crunching and cool be able to make a switch very easily, ad the gain is obvious. They would be able to look at the actual wavefront and have quantitative data of it at the same without having to go through the trouble of learning more about and build  interferometers.

 

I found this 2018 article on this topic for those who have never heard or seen it.

 

https://www.cloudyni...4xxg/?p=8494358

 

PS It's the Mother of star tests. :o)

Edited by MKV, 14 April 2025 - 05:00 PM.

[Dale Eason]

One down side to using roddier to test your mirror is that the results are the combined effect of the diagonal and collimation.  A good thing when testing the telescope but perhaps not so good when trying to isolate the errors of the main mirror.

 

If you want to experiment remember that DFTFringe can create star tests that you may be able to feed into Roddier.

[BGRE]

If you have an autocollimation flat with a central hole no diagonal is required.

The errors of the diagonal can easily be measured and corrected for if required.

The artificial star distance can be reduced somewhat if the additional SA due to the finite object distance is corrected for,

Raytracing can be used to obtain the value of the corresponding numeric Null.

[Gleb1964]

Roddier software assumes a test wavelength, so it's advisable to use a green filter (or another filter) to isolate the wavelength range. Ideally, a monochrome camera with a linear response should be used. The defocus should be sufficiently large so that both defocused images are out of any ray caustic. The defocused star images on both sides of focus should be round and of equal size.

Defocused images can have about 20 to 30 waves of defocus. A series of images of different sizes can be taken on both sides of focus, and the resulting wavefront should converge.

[starspangled]

 Unless you've made lots of mirrors using the AC Foucault test and then star test them (forever, waiting for perfect seeing), how do you know when that faint shadow is insignificant?  Ceravolo said his ego forced him to figure (using an optical null) until he could see no defect.

 

Its a good question .The simple answer is that when you have a good quality null test , you do just that . Figure the surface until it cuts off cleanly and greys  evenly over . The mirror will be then beyond reproach under the sky . The smaller the error , usually the faster it is to remove . The first exercise you want to do if making a mirror is see what the foucault shadows for 1/4 wave wavefront look like either under or over . 

 

The knife edge shift to put the crest or trough at the 70% zone  is not hard to calculate . it is also 1 wave shift if you null the edge or the centre . Such an error is not hard to see in any test system in single or double pass .Any improvement on this shape is heading in the right direction . If you can use a star test in the system , with an eyepiece even better. You will see that small faint rings of high or low have no effect on the in focus star and barely perceptible in the fresnel rings . That is a kind of qualitative estimate on the point at which it is going to achieve little to continue .

 

You do not need textbook  perfect seeing to look at the distribution of light in the defocussed star.  Any significant errors you see in a null test will still be visible in a star test in less than perfect conditions.

 

So if you are an advanced amateur or vendor making optics , and you have a well characterized  system you do not have to test every mirror on the real sky in  perfect seeing . You already know how well it is going to perform . This does not help people who are setting up a null test as an exercise to test various 3rd part y mirrors they have around and they want to put numbers on the shadows they are seeing . 

Edited by starspangled, 14 April 2025 - 08:47 PM.

[MKV]

If you have an autocollimation flat with a central hole no diagonal is required.

Yes, of course, but then you're severely limited with the practical size of the mirror because of the required size of the flat. An interferometer doesn't need a nulled mirror, at least for most common ATM mirror sizes and focal ratios. Also, a relatively small (80-100 mm diameter Ross lens) can tackle mirrors in the 20-25 inch range and focal ratios of 3.5 or so. Such a lens won't be cheap but a 25 inch flat will cost a lot more -- and is  more difficult to make.

 

I can see the Roddier test being an advantage over the usual "Suiter" crowd's method, but not for amateurs who want their scopes to be big and fast. 

 

I still don't know what the sensitivity of the Roddier as well as star test method is. Can it distinguish be tween a fully corrected mirror (cc -1.0) and a a 98% parabola (cc -98)? In large mirrors that's a lot of waves of error.

[Dale Eason]

 

 

I still don't know what the sensitivity of the Roddier as well as star test method is. 

You should know the sensitivity of the star test method.  It is described in Suiter page 82.  1/20 wave for broad deformations on the wave front and 1/60 wave for sharp variations.   That for perfect conditions.  For usual conditions 1/10 wave on the wave front.

Edited by Dale Eason, 14 April 2025 - 11:23 PM.

[BGRE]

Yes, of course, but then you're severely limited with the practical size of the mirror because of the required size of the flat. An interferometer doesn't need a nulled mirror, at least for most common ATM mirror sizes and focal ratios. Also, a relatively small (80-100 mm diameter Ross lens) can tackle mirrors in the 20-25 inch range and focal ratios of 3.5 or so. Such a lens won't be cheap but a 25 inch flat will cost a lot more -- and is  more difficult to make.

 

I can see the Roddier test being an advantage over the usual "Suiter" crowd's method, but not for amateurs who want their scopes to be big and fast. 

 

I still don't know what the sensitivity of the Roddier as well as star test method is. Can it distinguish be tween a fully corrected mirror (cc -1.0) and a a 98% parabola (cc -98)? In large mirrors that's a lot of waves of error.

It's good enough that the Rubin observatory telescope formerly known as the LSST (8.4m F/1.18 primary) uses a number of them for wavefront sensing at multiple field points.

There's no inherent issue in detecting aberrations with a few waves of amplitude. As in all wavefront sensors only the wavefront error is measured. The root cause of the wavefront error has little effect on detection, 

[MeridianStarGazer]

How certain are you that the dpac flat is perfectly spherical and smooth? Could it make the zambuto mirror appear worse than it is?

[BGRE]

One of the few affordable tests for measuring the flat in situ uses a Ritchey-Common setup with preferably a high resolution on axis spherical wave interferometer so that the residual errors in the spherical mirror and interferometer can be measured and subtracted from the result.

[MKV]

One of the few affordable tests for measuring the flat in situ uses a Ritchey-Common setup with preferably a high resolution on axis spherical wave interferometer so that the residual errors in the spherical mirror and interferometer can be measured and subtracted from the result.

The Ritchey-Common test measures only flatness, which is even not critical in autocollimation flats. It says nothing about the optical figure (profile). AC flats can have quite a bit of power (convexity or concavity) even when used for null testing of very fast optics.

 

Whether the figure is a sphere or not is theoretical (it should be spherical), because the radius of curvature is miles long, even tens of miles or more, but it has to be a single conic figure of revolution, and not a mixture of more than one (as is the case of the Schmidt corrector). Flats mustn't be zoney.

 

There is really very little difference between a spherical surface and an aspheric at extremely long radii of curvature compared to the aperture. The optical power is the invers of the focal length, so optical flats essentially have zero power for all practical purposes. One can think of them as reflective optical windows. 

[BGRE]

Nonsense, when a spherical wave interferometer is used to test a flat using a Ritchey-Common setup measurement of the actual figure is inherent using appropriate software.  

 

https://www.research..._interferometry

[hamishbarker]

You should know the sensitivity of the star test method.  It is described in Suiter page 82.  1/20 wave for broad deformations on the wave front and 1/60 wave for sharp variations.   That for perfect conditions.  For usual conditions 1/10 wave on the wave front.

I would say it's very sensitive for astigmatism. Whenever I set up my big dob, first thing I do is a quick star test on the local radio tower anticollision light. semi monochromatic, so eyepiece/eye chromatic error isn't an issue. if there is any astigmatism, I go back to the tailgate and check that all the wiffle tree triangles are properly seated - and every time this has been the cause.

 

When I have been figuring various spheres and long-focus mirrors and test plates for Stevick Paul telescopes, I use the star test at the centre of curvature for very sensitive detection of astigmatism. Texerau wrote "if astigmatism of even 1/10 wave is present, the image appears distinctly elliptical."

Edited by hamishbarker, 15 April 2025 - 09:52 PM.

[MKV]

You should know the sensitivity of the star test method.  It is described in Suiter page 82.  1/20 wave for broad deformations on the wave front and 1/60 wave for sharp variations.   That for perfect conditions.  For usual conditions 1/10 wave on the wave front.

I used DFTFRinge for a D500 F3 mirror and simulated a perfect 100% paraboloid, one which was only 99% corrected, and one 98%.

 

These are the DFTFRinge results:

 

Conic  Strehl   RMS    PV

 

-1.000  0.999  0.004  λ/71

-0.990  0.930  0.043  λ/7

-0.980  0.752  0.085  λ/3

 

 

Thus, the total change in the conic of 0.02 (!) resulted in a change of residual PV error from 0 (perfect) to 1/3 wave (unacceptable).  Yet, the star test simulation shows barely perceptible differences, the best being the focused image (in the eyepiece). And don;t forget these are stationary, rather than dancing all over.

 

 

This tells me that an observer could have difficulty distinguishing between a mirror that is perfect and one that has a conic of -0.98, except maybe in extraordinarily perfect observing conditions. That's not something I would use to assess a mirror, at least not these big and fast mirrors everyone is in love with.

 

However, compared to these results, interferometric results show most unambiguous difference for each value of the conic, many times smaller than 0.02. No guesswork needed.

[MKV]

Nonsense, when a spherical wave interferometer is used to test a flat using a Ritchey-Common setup measurement of the actual figure is inherent using appropriate software.  

https://www.research..._interferometry

Hardly. The original Ritchie-Common test described in ATM Book I was solely used to measure the sag of the surface. If any amateurs today actually use that test for their flats, they are surely using it exactly as it was described more than 80 years ago. No software, no spherical wave interferometers. Nonsense is telling someone to use something they don't have. They want to know what they can do in their workshops. 

[BGRE]

If you read my post (#68) I specified using a spherical wave interferometer in a Ritchey-Common setup not the original Ritchey-Common test!

[MKV]

If you read my post (#68) I specified using a spherical wave interferometer in a Ritchey-Common setup not the original Ritchey-Common test!

Yes, I know. That test is so different from the original test that it should have a different name, imo.