60 replies to this topic

[elliotwoods]

Dear All

 

I have learnt an enormous amount from the Amateur Telescope community. Seeing precision optics being made by hand in people's basements, and then used to look out into the cosmos is deeply inspiring.

From this inspiration, I am working on an artwork which incorporates 2 large handmade mirrors.

 

I must admit that the result will be an indoor art installation inspired by astronomy and Lorentz transformations, not an actual telescope.

I sincerely hope that it is not out of place to post my project here. 

Do let me know if this type of posting is not welcome.

 

Continuing...

 

I am 1 half of an artist collective based in South Korea called Kimchi and Chips.

I have some experience with optics (I studied Physics at university in Manchester) and with computer vision (some examples at my GitHub https://github.com/elliotwoods )

 

Here you can see one of our previous works Light Barrier which incorporates a 16m wide plenoptic mirror array.

 

Another example is our work Halo which incorporates an array of 99 heliostats:

 

If you are interested, you can see other works and videos at https://kimchiandchips.com/

 

 

 

For our next work, we intend to engineer 2 mirror flats, each 1.8m diameter.

If you think this is insane, then you'd be in good company.

 

Our P-V tolerance is generous. Up to 10 lambda is acceptable (although I would prefer to reach <1 lambda), so it might not be an absolutely insane endeavour. However, float glass is still not flat enough (Which has deviations as high as 10,000 lambda P-V across an 1800mm surface according to SCHOTT Flat Glass technical data).

 

My current plan is to build a fixed post turntable (Gordon Waite style), and use it to grind 3 surfaces:

A - Glass (1.8m)

B - Glass (1.8m)

C - Tool with ceramic tiles / pitch lap (sub-diameter, so possibly 1.0m)

 

Following the advice of Francis O'Reilly, I would consider to grind for each grit size:

 

B on top of A

A on top of B

C on top of B

B on top of C

A on top of C

C on top of A

[move onto next grit size]

 

However, it might not be practical to grind a larger surface over a smaller one on a fixed-post turntable. Furthermore, the concave/convex should 'come out in the wash' if I keep grinding the 3 surfaces against each other (negating the requirement to flip the pairs each time to keep the curvature neutral). In which case it might make more sense to do:

 

B on top of A

C on top of A

A on top of B

C on top of B

[repeat multiple times per grit size]

 

 

I understand that the process is quite unwieldy with such a large mirror, but we have the following to our advantage:

 

1. We have an engineering company we work with who can help build machines and mountings

2. We have use of a warehouse space with a gantry crane

3. Our tolerances are about 1 OOM weaker than normal parabolic mirrors (1/6 lambda) or real optical flats (1/20 lambda)

4. Since we don't have any curvature, some of the issues which arise when making a spheroid mirror are alleviated (e.g. accurately centering the glass on the turntable)

 

The mirrors will be displayed in a climate controlled environment, so large temperature changes are not so much an issue, so we are planning on using architectural float glass (not tempered). Also since it is a limited time exhibition, we currently plan to silver using the Angel Gilding technique.

 

I'm currently planning out how the procedure will go with the large mirrors, and will create a small version of the procedure which I will test on 6" (150mm) blanks which I have ordered from Firsthand discovery. i.e. I'll mock up a small fixed-post turntable and everything else that's required and try .

 

A few issues we are aware of and are working on:

1. The backing structure for the mirror (both for handling when grinding and mounting for final use)

2. Glass thickness (we're aiming for about 40mm at the moment. This is less than the 300mm for a 1/6 ratio, but in our FEM analysis this seems more than sufficient for the flatness specified above)

3. Lowering the glass on top of another piece of glass safely (i'm imagining the backing structure would have some 'release' threads which allow you to drop the final millimetres slowly and manually)

 

If this is something of interest to discuss here, then I will continue to post in this thread concerning design decisions, actual designs and production progress. We always try to open source our process as much as possible in case it is useful to others (e.g. the GitHub link above).

 

I would very much appreciate to hear your expertise, opinions and thoughts.

But again, is this project ok to discuss in this forum?

 

Thank you

Elliot

Edited by elliotwoods, 15 January 2021 - 05:41 AM.

[MitchAlsup]

Following the advice of Francis O'Reilly, I would consider to grind for each grit size:

 

B on top of A

A on top of B

C on top of B

B on top of C

A on top of C

C on top of A

[move onto next grit size]

Until the blanks are within grit-size of being flat you will want to stick with the largest grit size. This may take a dozen wets, each.

 

Once you have taken out the surface bumps and hills, then you might get away with the agressive grit size progression.

[coinboy1]

That is pretty cool! 10 waves is very generous. I think it should be doable. You probably need to make an accurate spherometer. If you get a dial indicator with enough resolution you should be able to see if it is concave/convex with a few waves of precision. The spherometer I made can measure to about 2 waves of resolution. Probably no need to test with other means if this for an art project. 

 

550nm wavelength of green light x 10 lambda = 5,500nm = .0055mm peak/valley easily measurable with a good spherometer. 

 

I don't think the 10,000 lambda figure is quite right. 

 

550nm wavelength of green light x 10,000 lambda = 5,500,000nm = 5.5mm peak/valley. That's nearly a 1/4" of deviation across a 70" surface. I know float glass is much flatter than this. Why don't you source some good ole' plate float glass? A lot of float glass table tops I have measured seems to be close to 10 waves of flatness when measured with my spherometer. 

 

On another note, I love the name "Kimchi and Chips". I am half Korean and just love Kimchi!!

Edited by coinboy1, 15 January 2021 - 03:54 PM.

[elliotwoods]

Until the blanks are within grit-size of being flat you will want to stick with the largest grit size. This may take a dozen wets, each.

 

Once you have taken out the surface bumps and hills, then you might get away with the agressive grit size progression.

Good call Mitch!

 

I'm curious if we are expecting P-V deviations at similar order of magnitude to the size of grit being applied.

e.g. if the first wet is with 60 grit Silicon Carbide (269 um), then are we expecting a P-V error across the surface to be similar to this value?

 

My instinct is:

 

* No for a parabola spheroid - It is not globally self-correcting, so the P-V error will scale linearly with the diameter of the mirror multiplied by the diameter of the grit (which can be considered the max gradient of 'noise' at surface)

* Yes for a flat - It is globally self-correcting, so the P-V error will be fixed across the entire surface, e.g. the P-V error will generally approach the RMS error (which again will be determined by the grit size)

 

These both presume sufficient grinding time has been performed.

My instincts might be wrong though.

Edited by elliotwoods, 16 January 2021 - 12:31 AM.

[mark cowan]

One doesn't grind a paraboloid usually, only a sphere (of which a flat is one example).  All of the spheres are self correcting as regarding the surface converging with grit sizes.

 

You'd do well to find some people who have handled that size of glass in the way you're contemplating, though.  Sag will be an issue with any conceivable glass you'd use.  The bottom piece will sag on its support, and the top piece will sag on the bottom piece.  That might be self cancelling.

[elliotwoods]

That is pretty cool! 10 waves is very generous. I think it should be doable. You probably need to make an accurate spherometer. If you get a dial indicator with enough resolution you should be able to see if it is concave/convex with a few waves of precision. The spherometer I made can measure to about 2 waves of resolution. Probably no need to test with other means if this for an art project. 

 

550nm wavelength of green light x 10 lambda = 5,500nm = .0055mm peak/valley easily measurable with a good spherometer. 

 

I don't think the 10,000 lambda figure is quite right. 

 

550nm wavelength of green light x 10,000 lambda = 5,500,000nm = 5.5mm peak/valley. That's nearly a 1/4" of deviation across a 70" surface. I know float glass is much flatter than this. Why don't you source some good ole' plate float glass? A lot of float glass table tops I have measured seems to be close to 10 waves of flatness when measured with my spherometer. 

 

On another note, I love the name "Kimchi and Chips". I am half Korean and just love Kimchi!!

Thank you @coinboy1! I appreciate the enthusiasm!

 

Spherometer : yes also need to think about testing / how to read the flatness. I've been considering:

* Spherometer

* Interferometer methods

* Direct computer vision methods (e.g. looking for subpixel deflections on reflections of known objects)

 

You're right, the figure of 10,000 lambda is a mistake. I was taking figures from here : https://www.schott.c...sh-18012017.pdf

 

 

Maximum deviation from flatness of 1 mm per 300 mm length measured
across the longest diagonal of the panel is permitted. The out-of-flatness
will be measured as the greatest right- angled distance between a fixed
vertical reference level and the panels.

Which i was scaling to 1800 / 300 = 9mm.

This obviously isn't the correct approach, and is contradicted in the next item which reads:

 

 

 

Maximum deviation from perpendicularity is 0.1°.
This corresponds to a deviation of 1.7 mm per meter.

These figures are certainly higher than 10 lambda (5um), although they may be worst cases.

 

It's good to know that your tests turned up < 10 lambda. I presume that is 10 lambda flatness across the baseline of your spherometer (e.g. 5um across 200mm)?

 

 

This leads onto 'what is the best way to measure flatness?'. 

We could imagine a a rough surface cross-section like so:

 

(the colours are just for labelling, not related to wavelength)

 

Width of section considered:

 

Red - 1000mm

Green - 500mm

Blue - 250mm

Purple - 125mm

Yellow - 72.5mm

 

P-V deviation within that section:

 

Red - 10 lambda

Green - 6 lambda

Blue - 4 lambda

Purple - 2 lambda

Yellow - 0.5 lambda

 

(these numbers are just made up)

 

i.e. the P-V deviation is a function of the size of region considered. If this were to be performed with a spherometer we would get the following results:

 

 

i.e. the spherometer measures the height deviation of the central point from the net gradient of the region.

 

I might be overcomplicating things here, but sometimes I have to get complicated first before I properly understand the simple things

 

 

I'm getting the impression that less than 1 lambda might be possible! Which is exciting ("Flatter than a wavelength of light")

Really we want it as flat at possible. Flatter than you can see with the naked eye.

 

And yes! Gotta love that kimchi !

Edited by elliotwoods, 16 January 2021 - 04:38 AM.

[elliotwoods]

One doesn't grind a paraboloid usually, only a sphere (of which a flat is one example).  All of the spheres are self correcting as regarding the surface converging with grit sizes.

 

You'd do well to find some people who have handled that size of glass in the way you're contemplating, though.  Sag will be an issue with any conceivable glass you'd use.  The bottom piece will sag on its support, and the top piece will sag on the bottom piece.  That might be self cancelling.

Sorry yes you're right. I meant sphere not paraboloid. Aiming for that infinite radius sphere!

 

We do have experience with handling large (>3m) pieces of glass at the studio for other projects, but I think you mean specifically in terms of mirror making here. We have reached out to professional mirror manufacturers, but haven't found a suitable partner so far.

 

I've been doing some FEM work regarding the sag (in Fusion 360). I'm investigating here the best structural approach for the glass and the mounting. I'll post more with images when I have clearer results.

[elliotwoods]

 All of the spheres are self correcting as regarding the surface converging with grit sizes.

Yes, thinking about this more, you're right. Thanks

[MKV]

My current plan is to build a fixed post turntable (Gordon Waite style), and use it to grind 3 surfaces:

A - Glass (1.8m)

B - Glass (1.8m)

C - Tool with ceramic tiles / pitch lap (sub-diameter, so possibly 1.0m)

Following the advice of Francis O'Reilly, I would consider to grind for each grit size:

B on top of A

A on top of B

C on top of B

B on top of C

A on top of C

C on top of A

[move onto next grit size]...

Hello,

 

I've made several sets of optical flats in the past 30 something years, and always using the ABC or 123 three-blank method, and your sequence is not how it's done. First, all three disks need to be the same size. You also need to remove any wedge in he disks. The backs of the flats should also  be flat or slightly concave. You may wish to prep these parameters first before grinding on the side to be used as a flat. 

 

The sequence is simple: A on B, Bon C, C on A. There's no need to do any additional flipping. The three disks will flatten simultaneously and will be quite flat (about 1/2 wave) after fine grinding by this method. You will have to make a separate polishing/figuring tools of the same or smaller size. The process is as old as ATM and is described in Ingall's Telescope Making Book One. There you will also find how the disks are tested against each other by contact interferences and how one goes about calculating their residual error in terms of fringes of power. It's basically solving for three unknowns simultaneously.

 

Good luck, Elliot!.

 

Mladen

 

[heartofglass]

Something is wrong. I suspect some miscommunication or misinterpretation. Perhaps someone looked at the allowable thickness variation plus/minus for the thickest glass and just gave you that number as a flatness spec. I would suggest first looking at the mounting of a piece of the size you are investigating on a horizontal plane.

https://www.davidlew...ronto.com/plop/

And tilted up to vertical.

https://www.cruxis.c...ecalculator.htm

Obtaining such a float glass blank, then measuring its flatness with a spherometer. You can also look up University of Arizona methods to measure a large flat with commercially available instruments. Yellowhair has some other papers measuring large flats as well.

http://www.loft.opti...lat_mirrors.pdf

But start with a spherometer.

https://www.youtube....h?v=htBx5SRFeMI

I believe you may get a big surprise at how flat the glass actually is and just how much the mounting and support make a difference.

 

Jack

 

Edit. The University of Arizona paper has "loft" in its address. That is the large optical flat testing program. They have a lot of good information.

Edited by heartofglass, 16 January 2021 - 03:54 AM.

[elliotwoods]

Something is wrong. I suspect some miscommunication or misinterpretation. Perhaps someone looked at the allowable thickness variation plus/minus for the thickest glass and just gave you that number as a flatness spec. I would suggest first looking at the mounting of a piece of the size you are investigating on a horizontal plane.

https://www.davidlew...ronto.com/plop/

And tilted up to vertical.

https://www.cruxis.c...ecalculator.htm

Obtaining such a float glass blank, then measuring its flatness with a spherometer. You can also look up University of Arizona methods to measure a large flat with commercially available instruments. Yellowhair has some other papers measuring large flats as well.

http://www.loft.opti...lat_mirrors.pdf

But start with a spherometer.

https://www.youtube....h?v=htBx5SRFeMI

I believe you may get a big surprise at how flat the glass actually is and just how much the mounting and support make a difference.

 

Jack

 

Edit. The University of Arizona paper has "loft" in its address. That is the large optical flat testing program. They have a lot of good information.

 

Thanks Jack

 

This paper is fantastic. And looking up the University of Arizona group brings up many leads. I will try to reach out to them once I feel prepared enough to ask the right questions.

 

This also brings back the idea of creating a fully automated system with a gantry robot that iterates between measuring the flatness of the mirror, and then fine grinding or polishing as required. 

 

Concerning the Spherometer approach. This tool can only measure relative sphericity(?) of 2 surfaces, and therefore requires a reference surface. coinboy1 above mentions the tool accuracy being ~2 lambda, but is a surface plate accurate to that level when calibrating? It seems that common surface plates are accurate to 3um (~5 lambda). This doesn't stop me seeing variances in sphericity between different parts of the glass, but is there a method for knowing the absolute flatness? I could compare A and B (presuming they are well ground vs each other).

 

 

 

Concerning the mirror cell (aka mirror mount), thank you for posting those links. I feel that I often see screenshots of the results from plop, but it's good to know the actual software behind them. 

 

Some great materials here to look into. Thank you

[elliotwoods]

Hello,

 

I've made several sets of optical flats in the past 30 something years, and always using the ABC or 123 three-blank method, and your sequence is not how it's done. First, all three disks need to be the same size. You also need to remove any wedge in he disks. The backs of the flats should also  be flat or slightly concave. You may wish to prep these parameters first before grinding on the side to be used as a flat. 

 

The sequence is simple: A on B, Bon C, C on A. There's no need to do any additional flipping. The three disks will flatten simultaneously and will be quite flat (about 1/2 wave) after fine grinding by this method. You will have to make a separate polishing/figuring tools of the same or smaller size. The process is as old as ATM and is described in Ingall's Telescope Making Book One. There you will also find how the disks are tested against each other by contact interferences and how one goes about calculating their residual error in terms of fringes of power. It's basically solving for three unknowns simultaneously.

 

Good luck, Elliot!.

 

Mladen

 

three flats.jpg

Thank you Mladen!

 

Incredible to get some input from somebody with decades of experience making optical flats.

Great to know that you can achieve 1/2 lambda before figuring.

 

Can I ask a couple questions about what you mentioned:

 

1. Can you explain more about why the backs of the glass need to be flat or slightly concave? (For structural reasons, I'm actually planning on the backs of the glass being a bit complicated). Is this because you need the rear to be flat when you're performing flatness tests on the front side later? Or to do with structural consistency / weight whilst grinding being the same?

2. Is there a specific reason why the third surface cannot be a ceramic tile tool / a tool of a different size? I presume it's to do with having equivalent rates of material removal per surface?

3. Have you used a motorised turntable whilst creating optical flats?

 

Thank you

Elliot

[coinboy1]

Hello Elliot,

 

I "zeroed" my spherometer on a 1/20 wave and 1/4 wave optical flat. You have to have a good calibration flat to "zero" your spherometer. 

 

There certainly is a "two-disc" method of grinding optical flats. I have done it before, as mentioned in Tips for Making Optical Flats  p.69 by Robert Piekiel. 

 

This will save a huge amount of money by not having the 3rd blank but you will need to grind four surfaces against each other. PM me if you need the grinding sequence for the "two-disc" method.

 

Mladen and others knows a great deal about optics, I am still learning but just throwing out what I know. Good luck!

[coinboy1]

Also it probably is possible to just buy one blank, and have 2 other full size tile tools for your required 10 wave flatness for the "three-disc" method. That really would save the most amount of money. 

 

If this is really for an art project, I don't see why you would need to test with interferometry or other methods. I highly doubt the eye could see any deviations down to 10 waves of flatness.

[coinboy1]

Just read your initial post again, you state you want to engineer two 1.8m flats. I would probably say the "two-disc" method is certainly your best best and most cost effective. All the surfaces would be ground against each other and the benefit would be all four surfaces would have nearly the same surface accuracy. Perhaps mount these flats on a spindle to showcase both sides and that would be pretty impressive. I could see some artistic value in this...

[elliotwoods]

Also it probably is possible to just buy one blank, and have 2 other full size tile tools for your required 10 wave flatness for the "three-disc" method. That really would save the most amount of money. 

 

If this is really for an art project, I don't see why you would need to test with interferometry or other methods. I highly doubt the eye could see any deviations down to 10 waves of flatness.

It might be worth to note that the 2 mirrors will be facing each other, and therefore any imperfections will be multiplied 'downstream'. We're aiming for an incredibly clean (and hopefully bright*) infinite reflection. This means that distortions that might be imperceptible on a single reflection would be visible in the multiplied reflection.

 

* with the silver coating, front silvering, and high polish

[chipe450]

A methode use in mechanical inspection to measure a flat surface without the need to calibrate the tooling is to use a high gage mounted with high resolution dial indicator or electronic dial indicator.
The idea is to scan the surface with the indicator. Since there is an offset of about 250 mm between the high gage and the indicator, if there is a bump in your surface, it will show. So you can map the form within the accuracy of the indicator

Andre

[davidc135]

Could you give a rough sketch of the layout and observer view? I'm getting the impression that the mirrors are nearly parallel to each other and that the audience walks between them?

 

Would it be worthwhile doing a practice, small scale version, say 30cms using the same material, silvered as is, before any working? 

 

The beams of light from each point of an object would be the width of the eye pupil reflecting to and fro between each mirror and with each reflection hitting the surface at a slightly different angle according to surface errors and the image thus being distorted or displaced slightly. Would the larger mirrors behave differently to the smaller other than giving a broader and more impressive effect? But the same distortion?

 

But I'm making assumptions about your design.

 

If you do find that the discs need grinding and polishing it may turn out to be enough that only one surface of each disc could be ground against the other to end up with both being sufficiently flat, given their method of production. But it would be good to be able to measure.

 

It's likely that the mirrors will both have astigmatic flexure when placed on their sides plus any overall curvature. Maybe these matter less than local irregularities.

 

David

Edited by davidc135, 16 January 2021 - 12:28 PM.

[MKV]

There certainly is a "two-disc" method of grinding optical flats. I have done it before, as mentioned in Tips for Making Optical Flats  p.69 by Robert Piekiel. This will save a huge amount of money by not having the 3rd blank but you will need to grind four surfaces against each other. 

Tony, unless you have a reference/master flat, you can't certify only two disks with unknown surface quality because the contact interference test (Newton's rings) shows the profile of both flats combined. If one is a reference flat then we know the profile we see is that of the test flat, but if both are unknown we can't know which or how much of each is what. That's why you need the third unknown of the same size.

 

For absoluter certification (as has been used by the US Bureau of Standards for many, many decades) the three disks need to be the same size and same substrate; if any of them is a ceramic tool, it can't be used to test against others. Besides, the wear and tear of the ceramic vs glass won't be the same. 

 

Mladen

[davidc135]

If it turns out that the mirror surfaces need to be worked how feasible is it to fine grind two 1.8m surfaces against each other? Would there be too much friction? An alternative would be to use a smaller tool to work each surface along with another way to measure flatness.

 

Considering the cost of a third full size glass disc could a 30cm flat, say of 1/10 wave surface accuracy, be hired for not too much? Used to calibrate a spherometer the whole surface could be measured in steps- but that doesn't sound very accurate- maybe OK for 10 lambda. Doing the same over both mating surfaces would help reduce measurement error.

 

As a matter of interest how would you measure a smooth curvature over 1.8m to an accuracy of a few waves? If fine ground? flash polished?

 

David

Edited by davidc135, 16 January 2021 - 01:54 PM.

[mark cowan]

A rough (500 grit SiC) grind to randomize the surface irregularities followed by fast polishing (felt or pads, avoiding onion peel surfaces at least) might be more than sufficient, if the goal is to not be able to see reflection errors with the eye.  Consider obtaining building size window stock and silver the front of these for a test setup though.

[Pinbout]

 

However, float glass is still not flat enough

Why?

 

I'm an artist that makes optics and designs architectural exhibits.

 

your pieces above are awesome, saw them before.

 

but I would like to hear why its necessary to lower the flatness to below wavelengths of light?

 

what is your expectations?

 

I'm using float glass to test another flat to see if its an optical flat...

the float glass is good enough to perform this test.

 

https://www.youtube....45PT43k&index=6

Edited by Pinbout, 16 January 2021 - 09:01 PM.

[MKV]

As a matter of interest how would you measure a smooth curvature over 1.8m to an accuracy of a few waves? If fine ground? flash polished?

The OP says he has access to engineers who can make equipment needed. I suggest using this method

 

https://spie.org/new...t-mirrors?SSO=1

 

Another method could be water immersion test -- a lot cheaper but never tried on very large flats.

 

Mladen

[heartofglass]

Elliotwoods,

  To answer your question, yes you will need a reference to use a spherometer. I guess mentioning a spherometer will get you acquainted with a fundamental tool for measuring flatness or curvature of optical surfaces. But I was also somewhat subversively trying to lead you back to questioning your assumptions about the flatness of float glass. You see, a standard practical flat reference for a spherometer is a thick piece of float glass as Pinbout mentions above. Considering you may have these mirrors in some kind of environment where temperature can change and vary and where you may have an air gap between these mirrors where currents of air with differing temperatures can refract light, I strongly suspect that further inquiry (perhaps with the architectural glass supplier) will make it apparent that simply silvering the existing glass surface as it comes from the supplier will suffice. But temperature control is another factor to complicate things especially if strong lighting is used as many lights generate significant heat. What I'm getting at is that environmental factors will possibly overwhelm any influence from imperfect surfaces of the float glass surfaces. If people are going to walk in between the mirrors then you may create an interesting Schlieren chamber given the creation of a very long light path that may mimic collimated light. At any rate, it may save you great amounts of trouble and expense to inquire about the flatness of float glass a bit more.

[coinboy1]

I agree with above post. I bet a lot of float glass will be within about 10 waves of flatness. Probably would work fine for this project and a lot less effort.