hello there.. For the last 8 years i turn and measure optical mirrors in all shapes and sizes. The elevation pattern that you can see on your mirror profile is a sign of your tool not beeng precisely on center. This error looks either like an W , that would be your case, or and M. Some errors could also come from your diamond edge or your Axis but mostly i can see an error on your tool beeng of center and thats the error you can fix. Dependen on which side of your spindel your cutting tool stands and if the radius of your optic is concarv or convex you can deduce if your tool needs an correction in plus or minus. Cutting on the left of center in a concave radius getting a kind of a W shape i would say you need to correct your tool away from center to get a better shape. Correct your tool 10µ and cut another sphere and look what happens. when you measure many times with differend tool corrections you can find the sweetspot where there is no W or M error then you only see every other error :D The ring in the middle of your mirror was probably caused from some impuriti in your aluminium. Also you could get better results by taking the last two or three finisching cuts with the same depth and switching the side of your coolant so it blows away from center.
Kicking the dead constant sfm horse some more but you can keep the spindle speed constant while reducing the feedrate as you get to the center. In a big rigid setup, constant sfm doesn’t matter. On your smaller less rigid setup cutting forces may be coming in to play. Tool is deflected on the od because its spinning too fast and is unable to get a stable cut, digs in at the middle of the radius as it finds an optimal sfm to balance the forces, and finally deflects again at the center because it is feeding too fast for the spindle speed causing that tearing/rubbing finish. Another thing to check may be the tool height, make sure it is perfect. Take this with a huge grain of salt, I am only a .0001” machinist. This stuff is really cool, love the content.
How is the surface when making a plane mirror? I.e. with Z locked. Wondering if you can narrow the source of the defects again - I thought this clamping force test would have shown a larger impact. Awesome content, appareciate getting to see your great work.
Would have asked the same... In my eyes the results look like a pretty balanced toolpith with the exception of two areas where forces on the drive train add up and let go in a more spontaneous way. Does that make sense?😂
Outstanding work, as usual! I do R&D with an older Moore 450UPL at work. At one point I was investigating machining parameters (including constant surface speed [which is indeed over-rated for SPDT of aluminum], feed, tool geometry, alloys, etc). Decided to minimize variable by using 2" flats as my test standard, but found some residual power (PV error) much larger than expected. Thought it was vacuum pressure, but for these pucks that was an order of magnitude lower effect unless the mating surface wasn't lapped flat. However, one day I decided to chuck up a 6" 1/20 wave optical flat and ran a mahr electronic indicator across it (and clocked to rule out wedge) and found that there was waviness in the X-axis. During PM, Moore would adjust spindle axis perpendicularity with a test bar, but they'd only check one position along the axis so it went unnoticed by the production folks. So, if it's a geometry error, you should be able to find it without making a cut, unless it's dynamic. If that's the case and it passes the flat test, then find a high quality glass lens with a spherical curvature similar to yours to use as a "calibration artifact" and program a radial scan with an indicator and you'll get your answer. Measure it with your interferometer and make an error map of the artifact if you need to. That being said, if you can confirm with an artifact that it's not machine error, you still have many things to investigate. I've seen many befuddling errors due to material inhomogeneities, mounting interfaces, and residual stress from blank preparation. Heat treatment is important, as can be alloy selection. Obviously thermal can be a huge error source, and the longer the cut the more it can be an issue. There's also tool tip waviness - local irregularities in the tool's radius. Our supplier doesn't charge much more for a 1/10th wave tool tip waviness vs 1/4 wave and it helps on spherical parts when you're not using a B-axis to rotate the tool. Good luck!
Also, if it's a repeatable axis error you should be able to compensate your tool path (depending on your CAM package.) That worked well for the test flats I was talking about, but it was easier because it was a low-frequency error (i.e. power, not irregularity)
Just a comment on the Zernike settings. It's hard to see in the video but I think defocus is disabled. This automatically removes a large part of the difference in spherical shape. You might want to leave that checked for a better comparison.
I’m working through learning Zernike polynomials right now and that one had been confusing me. I feel like I’ve seen people keep it mostly unchecked so any literal defocus from the interferometer not being at the exact correct axial position is removed. How would one discriminate defocus due to alignment and defocus due to actual surface shape?
@@cylosgarage well, for example with a telescope mirror, the precise focal length is not that important, as long as the surface shape is correct for the optimal focal length. However, if you are interested in the actual deformation in the mirror shape due to mounting (like you are), you should include the effects of defocus in your evaluation. Because for example tightly mounting it in the centre could change focal length. If you ignore this effect, you are missing part of the actual surface deformation.
@@cylosgarage I concur with the great man - I hadn't appreciated that you were turning a parabola (I thought it was plane from the previous video). I think part of the explanation could be a small error in your X zero wrt the spindle axis. This generates a characteristic 'W' or 'M' pattern in the centre of the part wrt the best fit sphere, depending on whether the tool goes past centre, or stops short of the centre. This could explain everything within R=20mm.
Maybe it's just a hladini figure. The antinode at the resonance point.., a standing wave. The plate resonated at the time of processing. That is, to reduce the amplitude, it is necessary to reduce the speed. One of the theories.
Optics always seemed interesting to me but I feel like I'd be bumping into carefully aligned things all the time and living in an apartment chasing down vibrations from like upstairs neighbors stomping hard enough to shake my lamps almost 24/7
Maybe your air bearing guide suface has some dirt on it? You might just wanna try cutting on a different part of your x axis and see if your error moved with your tool or if its in your axis?
Your videos and progress are awesome as always! Do you see a chance that you could compensate for the error you measured in the next round of machining? ( I mean inentionally turning a wrong surfuce by the error you measured of the previous machining step.)
Neat. There is a lot of astigmatism but that is the least of your worries for now. It also looks like there is a concentric ripple pattern. If you wanted to compare two wave fronts to see how they are different you can use the DFTFringe diff (different) function to subtract one from the other.
Is the error you are seeing consistent over multiple cuts? If so, a way to get a better end product could be to "distort" your "inputshape" using your measured difference from the desired shape, i.e. cut deeper where your machine doesn't cut enough and cut less where the tool gets pulled into the surface too much.
Kind of new to your content but have you monitored the temperature of the cutter and item you are cutting? Even the slightest temperature changes especially the environment may effect the machine, cutter, and material at the precision to are trying to achieve. I'm actually impressed at what you have achieved so far. Your tooling geometry may matter and observing and measuring its temperature change if any during the cut could be of interest if you experiment with changing that variable. I had a thought on being able to measure any vibrations in the tooling itself during operation. Could the tooling be clamped with a say piezo material between the clamping mechanism and tooling and monitored during the cutting operation? Good luck.
Nice work!! The optical quality certainly is already quite good. While you were addressing the surface speed concerns, I was wondering if the tool angle to the workpiece might be a factor here. I have zero experience with diamond turning, but the same laws of physics still apply. As the tool traverses along the radius of the workpiece, the surface speed will differ at the cutting edge and if the tool is slightly out of alignment, the tool might cut more efficiently (i.e lower cutting forces) at certain speeds compared to others... Like I said, zero experience, just sharing my 5 cents worth...
diamond tools have a circular cutting edge (viewed from the top), and for a curved surface the contact point between workplace and tool keep changing. since it's nearly perpendicular to the work piece, the side rake and clearance angle don't really change unless there's an error in the shape of the tool (which does indeed happen.) you find differences in surface roughness as a function of cutting speed, but form errors aren't likely to occur since the cutting force is incredibly small
Averaging over 90 degree rotations is a bit questionable since your coupling has order-3 symmetry. When you average over 90 degrees (order-4), you might end up cancelling out deformations that are induced by the shape of your coupling. Although I doubt that such deformations are a major effect, I think some mild order-3 stuff was showing up in your wavefronts in the previous videos.
I should have been a little more specific there. I take 4 interferograms at 0,90,180, and 270 degrees, process them, derotate the last 3 interferograms -90,-180,-270 degrees and THEN average them. This is intended to average out any common mode error from the interferometer, and errors due to atmospheric turbulence. So basically I’m not averaging one at 0 and one at 90, I’m derotating the 90 back to 0 and then averaging.
Not an easy task. I guess you should monitor the behaviour of the moving axes. An autocollimator may help, not sure. I would try gluing the mirror to try reduce the clamping stresses.
Can it be the wiggle of the system that controls the diamond position (X and Y?)? I don't mean the stiffness of the construction, I mean the motors/controller/etc. Can the PID controller oscillate a bit around a new set point?
I love clicking on vids with less than 1k views I have no clue whats going on but I get to experience more I have no clue what hes talking about Im stupid but it looks cool heres a sub Ill try and learn when I come down
Good method. Also suggest @ROBRENZ for more high precision machining, and @HuygensOptics for the optics. They explain a lot. And the comments. These kind of channels, you mentioned, are watched by enthusiast (some professionals). They can add greatly to a topic. Over the years, I learned as much from the comments as from the videos. (Sometimes just by going down a rabbit hole led by a single keyword.)
Is there any chance that some of the distortion is thermal? Some sort of latent stress in the base material that leads it to distort slightly with however many degrees the room temperature might vary.
@@AlJay0032 500 rpm. Centrifugal forces are indeed a factor in diamond turning. This part is pretty small so I’ve assumed it to be negligible. But who knows, might be worth the analysis
it's a good thought, but the cutting forces are extremely low and his setup is very compact so it's not likely to explain this problem. Stiffness issues may show up in micro-roughness, but not overall form accuracy. Diamond turning doesn't work like regular machining - all macroscopic intuition goes out the window! And just to second Mister_G's comment, below - zero rake works best for diamond turning aluminum.
Moore Nanotech's and Precitech Nanoform's don't have flexture tool posts like this ultra-precision home built lathe does. It's like comparing apples to oranges with the stiffness order of magnitudes achievable in commercial SPDT lathes.
hello there.. For the last 8 years i turn and measure optical mirrors in all shapes and sizes. The elevation pattern that you can see on your mirror profile is a sign of your tool not beeng precisely on center. This error looks either like an W , that would be your case, or and M. Some errors could also come from your diamond edge or your Axis but mostly i can see an error on your tool beeng of center and thats the error you can fix. Dependen on which side of your spindel your cutting tool stands and if the radius of your optic is concarv or convex you can deduce if your tool needs an correction in plus or minus. Cutting on the left of center in a concave radius getting a kind of a W shape i would say you need to correct your tool away from center to get a better shape. Correct your tool 10µ and cut another sphere and look what happens. when you measure many times with differend tool corrections you can find the sweetspot where there is no W or M error then you only see every other error :D The ring in the middle of your mirror was probably caused from some impuriti in your aluminium. Also you could get better results by taking the last two or three finisching cuts with the same depth and switching the side of your coolant so it blows away from center.
Nothing I do is measured in nanometers but I am profoundly fascinated by such work and the fact that it is possible.
Kicking the dead constant sfm horse some more but you can keep the spindle speed constant while reducing the feedrate as you get to the center.
In a big rigid setup, constant sfm doesn’t matter.
On your smaller less rigid setup cutting forces may be coming in to play.
Tool is deflected on the od because its spinning too fast and is unable to get a stable cut, digs in at the middle of the radius as it finds an optimal sfm to balance the forces, and finally deflects again at the center because it is feeding too fast for the spindle speed causing that tearing/rubbing finish.
Another thing to check may be the tool height, make sure it is perfect.
Take this with a huge grain of salt, I am only a .0001” machinist.
This stuff is really cool, love the content.
How is the surface when making a plane mirror? I.e. with Z locked. Wondering if you can narrow the source of the defects again - I thought this clamping force test would have shown a larger impact.
Awesome content, appareciate getting to see your great work.
Would have asked the same... In my eyes the results look like a pretty balanced toolpith with the exception of two areas where forces on the drive train add up and let go in a more spontaneous way. Does that make sense?😂
Outstanding work, as usual!
I do R&D with an older Moore 450UPL at work. At one point I was investigating machining parameters (including constant surface speed [which is indeed over-rated for SPDT of aluminum], feed, tool geometry, alloys, etc). Decided to minimize variable by using 2" flats as my test standard, but found some residual power (PV error) much larger than expected. Thought it was vacuum pressure, but for these pucks that was an order of magnitude lower effect unless the mating surface wasn't lapped flat. However, one day I decided to chuck up a 6" 1/20 wave optical flat and ran a mahr electronic indicator across it (and clocked to rule out wedge) and found that there was waviness in the X-axis. During PM, Moore would adjust spindle axis perpendicularity with a test bar, but they'd only check one position along the axis so it went unnoticed by the production folks.
So, if it's a geometry error, you should be able to find it without making a cut, unless it's dynamic. If that's the case and it passes the flat test, then find a high quality glass lens with a spherical curvature similar to yours to use as a "calibration artifact" and program a radial scan with an indicator and you'll get your answer. Measure it with your interferometer and make an error map of the artifact if you need to.
That being said, if you can confirm with an artifact that it's not machine error, you still have many things to investigate. I've seen many befuddling errors due to material inhomogeneities, mounting interfaces, and residual stress from blank preparation. Heat treatment is important, as can be alloy selection. Obviously thermal can be a huge error source, and the longer the cut the more it can be an issue.
There's also tool tip waviness - local irregularities in the tool's radius. Our supplier doesn't charge much more for a 1/10th wave tool tip waviness vs 1/4 wave and it helps on spherical parts when you're not using a B-axis to rotate the tool.
Good luck!
Also, if it's a repeatable axis error you should be able to compensate your tool path (depending on your CAM package.) That worked well for the test flats I was talking about, but it was easier because it was a low-frequency error (i.e. power, not irregularity)
Just a comment on the Zernike settings. It's hard to see in the video but I think defocus is disabled. This automatically removes a large part of the difference in spherical shape. You might want to leave that checked for a better comparison.
I’m working through learning Zernike polynomials right now and that one had been confusing me. I feel like I’ve seen people keep it mostly unchecked so any literal defocus from the interferometer not being at the exact correct axial position is removed. How would one discriminate defocus due to alignment and defocus due to actual surface shape?
@@cylosgarage well, for example with a telescope mirror, the precise focal length is not that important, as long as the surface shape is correct for the optimal focal length. However, if you are interested in the actual deformation in the mirror shape due to mounting (like you are), you should include the effects of defocus in your evaluation. Because for example tightly mounting it in the centre could change focal length. If you ignore this effect, you are missing part of the actual surface deformation.
@@cylosgarage I concur with the great man - I hadn't appreciated that you were turning a parabola (I thought it was plane from the previous video). I think part of the explanation could be a small error in your X zero wrt the spindle axis. This generates a characteristic 'W' or 'M' pattern in the centre of the part wrt the best fit sphere, depending on whether the tool goes past centre, or stops short of the centre. This could explain everything within R=20mm.
@@cylosgarage Worth looking at the effect that an error in tool radius generates too.
Maybe it's just a hladini figure. The antinode at the resonance point.., a standing wave. The plate resonated at the time of processing. That is, to reduce the amplitude, it is necessary to reduce the speed. One of the theories.
"back to work"... reminds me of another engineer, Mike Patey. Keep up the great work - very interesting.
Optics always seemed interesting to me but I feel like I'd be bumping into carefully aligned things all the time and living in an apartment chasing down vibrations from like upstairs neighbors stomping hard enough to shake my lamps almost 24/7
Maybe your air bearing guide suface has some dirt on it? You might just wanna try cutting on a different part of your x axis and see if your error moved with your tool or if its in your axis?
Very impressive. Thank you for responding to my question about linear cutting speed as the tool approaches the center. Now, back to work!
Your videos and progress are awesome as always! Do you see a chance that you could compensate for the error you measured in the next round of machining? ( I mean inentionally turning a wrong surfuce by the error you measured of the previous machining step.)
Can you feed the inverse of the surface error into the Z axis setpoint?
Neat. There is a lot of astigmatism but that is the least of your worries for now. It also looks like there is a concentric ripple pattern. If you wanted to compare two wave fronts to see how they are different you can use the DFTFringe diff (different) function to subtract one from the other.
Is the error you are seeing consistent over multiple cuts? If so, a way to get a better end product could be to "distort" your "inputshape" using your measured difference from the desired shape, i.e. cut deeper where your machine doesn't cut enough and cut less where the tool gets pulled into the surface too much.
Kind of new to your content but have you monitored the temperature of the cutter and item you are cutting? Even the slightest temperature changes especially the environment may effect the machine, cutter, and material at the precision to are trying to achieve. I'm actually impressed at what you have achieved so far. Your tooling geometry may matter and observing and measuring its temperature change if any during the cut could be of interest if you experiment with changing that variable. I had a thought on being able to measure any vibrations in the tooling itself during operation. Could the tooling be clamped with a say piezo material between the clamping mechanism and tooling and monitored during the cutting operation? Good luck.
Nice work!! The optical quality certainly is already quite good.
While you were addressing the surface speed concerns, I was wondering if the tool angle to the workpiece might be a factor here. I have zero experience with diamond turning, but the same laws of physics still apply. As the tool traverses along the radius of the workpiece, the surface speed will differ at the cutting edge and if the tool is slightly out of alignment, the tool might cut more efficiently (i.e lower cutting forces) at certain speeds compared to others...
Like I said, zero experience, just sharing my 5 cents worth...
diamond tools have a circular cutting edge (viewed from the top), and for a curved surface the contact point between workplace and tool keep changing. since it's nearly perpendicular to the work piece, the side rake and clearance angle don't really change unless there's an error in the shape of the tool (which does indeed happen.)
you find differences in surface roughness as a function of cutting speed, but form errors aren't likely to occur since the cutting force is incredibly small
Any chance of keeping the on-machine interferometry setup in-place and using adaptive control of the cutting tool to minimise the errors?
how much vibration are you getting in the tool, getting toward the center would i think increase the amplitude?
That building is no comet, but I suppose it shares a similar name to one. Did a double take! :)
Haha, nice 😄
Averaging over 90 degree rotations is a bit questionable since your coupling has order-3 symmetry. When you average over 90 degrees (order-4), you might end up cancelling out deformations that are induced by the shape of your coupling. Although I doubt that such deformations are a major effect, I think some mild order-3 stuff was showing up in your wavefronts in the previous videos.
I should have been a little more specific there. I take 4 interferograms at 0,90,180, and 270 degrees, process them, derotate the last 3 interferograms -90,-180,-270 degrees and THEN average them. This is intended to average out any common mode error from the interferometer, and errors due to atmospheric turbulence. So basically I’m not averaging one at 0 and one at 90, I’m derotating the 90 back to 0 and then averaging.
Not an easy task. I guess you should monitor the behaviour of the moving axes. An autocollimator may help, not sure. I would try gluing the mirror to try reduce the clamping stresses.
Can it be the wiggle of the system that controls the diamond position (X and Y?)? I don't mean the stiffness of the construction, I mean the motors/controller/etc. Can the PID controller oscillate a bit around a new set point?
I want to see you sharpen a knife with this kind of precision. I think you could make the sharpest knife on youtube. with the ultimate polished edge.
I love clicking on vids with less than 1k views I have no clue whats going on but I get to experience more I have no clue what hes talking about Im stupid but it looks cool heres a sub Ill try and learn when I come down
Good method.
Also suggest @ROBRENZ for more high precision machining, and @HuygensOptics for the optics. They explain a lot.
And the comments. These kind of channels, you mentioned, are watched by enthusiast (some professionals). They can add greatly to a topic. Over the years, I learned as much from the comments as from the videos. (Sometimes just by going down a rabbit hole led by a single keyword.)
Is there any chance that some of the distortion is thermal? Some sort of latent stress in the base material that leads it to distort slightly with however many degrees the room temperature might vary.
How fast is the disk spinning? Could vibrations or standing waves or stretching by centrifugal forces explain any of this?
@@AlJay0032 500 rpm. Centrifugal forces are indeed a factor in diamond turning. This part is pretty small so I’ve assumed it to be negligible. But who knows, might be worth the analysis
@@cylosgarage Your work is really amazing.
William "The Goodest Boy" Hobbs? 😂😂😂😂
Subtract your measurement file from your nominal tool path to compensate.
If it is a repeatable machine geometry, then reverse it in your program...
insane! keep it up!
Awsome video!
I'm telling you, it's not a stiff enough tool post flexure and the monocrystalline diamond needs a steeper negative rake to help it cut smoother.
Zero rake for aluminium with diamond tools. Negative rake for other materials like germanium, zinc sulphide, etc.
it's a good thought, but the cutting forces are extremely low and his setup is very compact so it's not likely to explain this problem. Stiffness issues may show up in micro-roughness, but not overall form accuracy. Diamond turning doesn't work like regular machining - all macroscopic intuition goes out the window!
And just to second Mister_G's comment, below - zero rake works best for diamond turning aluminum.
Moore Nanotech's and Precitech Nanoform's don't have flexture tool posts like this ultra-precision home built lathe does. It's like comparing apples to oranges with the stiffness order of magnitudes achievable in commercial SPDT lathes.