About the astigmatism, it's generally caused by stress in the material during the lathing proces. Either internal stress in the material or (more likely) due to the mounting method. The mounting method you use does not really eliminate the introduction of stress.
8:24 in the picture in the top right of the screen, you can even see the three blue marks caused by the three mounting "dogs" - if you imagine holding it using a centre draw bar to pull it against a ring, it will cause it to deflect into a concave surface, then when you machine it flat, then take it off, relieving the stress, it will deflect to a convex surface.
Don’t beat yourself up too much, this is one HELL of a project and you have made it very far! Now it’s just *troubleshooting hell* and maybe some modifications based off of that if need be, but you got this! Keep up the great work!
8 หลายเดือนก่อน +9
This is so exciting. Thanks for posting these updates!
Looking at main profile graph, my first guess it is some sort of undamped acustic vibration issue. No idea how to counter it, maybe use much more bigger mass for lathe and/or make special damper based on viscosous liquid with counterweight. Another enemy is temperature, maybe get some cheap but high res IR camera and ttermocouples al around lathe structure and track what happens during whole precess, maybe some unexpected local heat and bending. Anyway, very exciting to see this project, thanks for share.
A former friend had his SIP Jig borer set up in a separate basement room, with its own concrete base, completely isolated from the house, and a precision A/C system, with very high air flow in the room, in order to maintain an isothermal environment. Where there is a potential for step size errors in a system, one can add "noise" in the loop, to reduce "sticking" on sensor and actuator increments.
Yes my thoughts exactly, there is specific harmonics that are going to exist with the length of the tool, the vibration of the tool, and the feed speeds. So at least one of these are going to have to change. I also bet the distortions are also at very specific multiples of intervals.
Very interesting - thanks. Is the blank cut from plate? If so, the (symmetrical) astigmatism could be due to anisotropy wrt the drawing direction of the plate. I wonder if the rise in the surface inside ~R=13mm might be due to the blank distorting from your drawbar - actually machined flat, but sprang back when the drawbar was released? It doesn't explain anything to do with the feature inside R=5mm though. Can you tell anything about the X=0 position from the angle at the top of the curve in the centre? If X 0 was correct, then the two halves of the trace should be tangent in the centre. Looking forward to the next instalment. :)
I'm so glad I started following you back when you were experimenting with sugar rockets. I've learned so much about advanced metrology following the development of your ultra-precision machine tools. Great stuff, man.
I have an idea of how or what can be causing the blemishes. Wherever you are doing your maching (lathe setup location) could be getting vibrations from something as minute as a A/C unit on a house running could potentially cause blemishes on a micron level, so pretty much any other machines or appliances running or even talking to a friend in the vicinity of operation can result in acuastic resonance in the tooling or the oarts being machined. Thats why if you ever go to a professional manufacturer of these telescope mirrors they have clean rooms that are also acoustically cut off from the outside world that is if they are making mirrors in the same manner as diamond lathe, but there is also lapping of mirrors. But what im trying to get at is next time you machine a mirror turn off any other known electrical sources even the a/c on your house and see what the results are
You're using a drawbar, right? If so, did you measure the part constrained or unconstrained, and to the same torque? Also, if the tool was above center then the cutting forces likely increase the closer you get to the axis of rotation, which could contribute to errors.
What kind of environmental controls do you in place. For some projects Moore Tool baths the entire machine in the same oil used for machining. How often do you check the cutting tool geometry? Are you using a laser interferometer? Changes in humidity, barometric pressure and, of course, temperature can effect the parameters of the laser. Still, amazing dedication to the science, and practical use of, metro.ogy.
Great first result. As the tool moves toward the center of the piece, don't you want to maintain the linear speed of the cutting tool? ie: rotational speed increases as tool moves toward center.
For that mirror we discussed previously, a piece like you have at the start of this one is (scratches aside) more that adequate for my purpose. No rush, but it'd be useful to me and I'm happy to support your work some.
Sorry to add to your list of variables, but tool break in could also be a factor as well. Aluminum can also be very temperamental to turn because of the other embedded things in the material that can damage the tool edge or ruin the surface finish.
Do you have the rotation speed relative to the distance from the part center when cutting? Otherwise even if you have the perfect surface speed at one point at another it will be wrong.
How do you determine that the defects are caused by the spindle bearings rather than the servos that control the tool position? The error in the figure has built up over numerous machining passes, as though the errors in tool position have added coherently.
Spindle bearings are honestly the last suspect (cause PI is the best). However, you make a good point about the errors adding coherently. Spoiler alert, I had another go at it and saw a similar looking error profile in the new results. Trying to determine if this systematic error is on the metrology side or the machine side.
@@cylosgarage I was conflating this post with one of your earlier posts where you ascribed vibration problems of the spindle to airflow through the bearings. I assumed that post referred to a spindle of your own design. I should have asked if you were using the PI spindle here.
I think you really need to consider the overall stiffness of your flexture tool post / holder. The snap back could be the change in surface feet per minute as you are turning across the mirror face (outer towards inner) and causing the flexture to stress slightly. If the constant stiffness is not there throughout the whole cut, you're going to see turning rings as the monocrystalline diamond and flexture are fighting against each other.
I am not sure the surface will get better due to bearings at those scales. Metal doesn't cut the way you think. th-cam.com/video/aF7ltBT_atA/w-d-xo.html Taking a shallower depth of cut can actually give worse finish in some cases as there is not enough material being removed to produce a proper chip to evacuate. Instead it just mushes around. This might be what you are seing. The sudden change in cut at the inner area may be the stepover vs feed crossing some threshold causing it to suddently "bite" better. The speed of the main spindle would have to be controlled relative to the current cutting diameter to create a constant linear feedrate to eliminate such thresholds.
As soneone that writes software for devices my vote is that cause is rounding error in toolpath calculation. You are dealing with math where regular 80bit floating point isnt precise enough. Maybe inspiration for a video how whole proces goes from formula to toolpath generation to hardware used for driving and actuators. If you make second part you can see if pattern is repeatable or it is only random noise.
I have thought about this a little. I’m not sure how Linux cnc is doing the math in the Gcode behind the scenes, or how much it rounds. The benefit of having it in the Gcode is small file sizes. However, I have considered writing a matlab script to directly calculate the tool path to a specified precision and output a Gcode file of discrete points. The only downside with that method is your file size rapidly gets deep into the megabyte regime. Not sure the limit of what Linuxcnc can handle. I suppose another benefit of using the externally generated toolpath is the ability to create error maps and apply compensation to a given toolpath. hmmm...
@@cylosgarage I've done some pretty hairy adaptive paths and LinuxCNC doesn't care. I can't remember how much lookahead it has though, since a pause to refill the buffer would kill your process. All in all you're doing great work! I do worry about internal rounding errors in LinuxCNC proper though (or finding the right config entry)
Generally CSS is not used for diamond turning because changing the spindle speed would change the forces and hence deflections due to spindle imbalance as you move across the part. It’s a bit of a compromise in this regard.
About the astigmatism, it's generally caused by stress in the material during the lathing proces. Either internal stress in the material or (more likely) due to the mounting method. The mounting method you use does not really eliminate the introduction of stress.
Thanks for the tip!
8:24 in the picture in the top right of the screen, you can even see the three blue marks caused by the three mounting "dogs" - if you imagine holding it using a centre draw bar to pull it against a ring, it will cause it to deflect into a concave surface, then when you machine it flat, then take it off, relieving the stress, it will deflect to a convex surface.
Don’t beat yourself up too much, this is one HELL of a project and you have made it very far! Now it’s just *troubleshooting hell* and maybe some modifications based off of that if need be, but you got this!
Keep up the great work!
This is so exciting. Thanks for posting these updates!
Man that's really over my head but i really enjoy your enthusiasm. Thank you for sharing
How does the astigmatism line up with the bolt pattern and milled pockets?
Very impressive 1st result!!
Impressive results, keep at it!
Really looking forward to seeing how far you can improve this setup
Looking at main profile graph, my first guess it is some sort of undamped acustic vibration issue. No idea how to counter it, maybe use much more bigger mass for lathe and/or make special damper based on viscosous liquid with counterweight. Another enemy is temperature, maybe get some cheap but high res IR camera and ttermocouples al around lathe structure and track what happens during whole precess, maybe some unexpected local heat and bending. Anyway, very exciting to see this project, thanks for share.
Temp controlled environment and much bigger lathe 100%
A former friend had his SIP Jig borer set up in a separate basement room, with its own concrete base, completely isolated from the house, and a precision A/C system, with very high air flow in the room, in order to maintain an isothermal environment.
Where there is a potential for step size errors in a system, one can add "noise" in the loop, to reduce "sticking" on sensor and actuator increments.
Yes my thoughts exactly, there is specific harmonics that are going to exist with the length of the tool, the vibration of the tool, and the feed speeds. So at least one of these are going to have to change.
I also bet the distortions are also at very specific multiples of intervals.
Wow! I’ve got a new favorite channel! This is some amazing work. I’ll be following along to see how everything goes
Very interesting - thanks. Is the blank cut from plate? If so, the (symmetrical) astigmatism could be due to anisotropy wrt the drawing direction of the plate.
I wonder if the rise in the surface inside ~R=13mm might be due to the blank distorting from your drawbar - actually machined flat, but sprang back when the drawbar was released? It doesn't explain anything to do with the feature inside R=5mm though.
Can you tell anything about the X=0 position from the angle at the top of the curve in the centre? If X 0 was correct, then the two halves of the trace should be tangent in the centre.
Looking forward to the next instalment. :)
I’m new to this channel and GOD DAMN! This is insane level of accuracy ! Awesome!
I'm so glad I started following you back when you were experimenting with sugar rockets. I've learned so much about advanced metrology following the development of your ultra-precision machine tools. Great stuff, man.
Ah, the good old days. Thanks for sticking around!
For the build up of error and sudden snap back... I'm just guessing, but could that be indicative of a quantization step in the servo loop?
I guess to check for that, he could just cut a flat mirror. With no z feed comes no z quantization...
very nicely done.
7:30 What software is that?
DFTFringe, absolutely wonderful piece of open source interferometry software
I have an idea of how or what can be causing the blemishes. Wherever you are doing your maching (lathe setup location) could be getting vibrations from something as minute as a A/C unit on a house running could potentially cause blemishes on a micron level, so pretty much any other machines or appliances running or even talking to a friend in the vicinity of operation can result in acuastic resonance in the tooling or the oarts being machined. Thats why if you ever go to a professional manufacturer of these telescope mirrors they have clean rooms that are also acoustically cut off from the outside world that is if they are making mirrors in the same manner as diamond lathe, but there is also lapping of mirrors. But what im trying to get at is next time you machine a mirror turn off any other known electrical sources even the a/c on your house and see what the results are
That's great advice
I like the suggestion of mass to reduce vibration although I have no idea if it's practical for you design.
You're using a drawbar, right? If so, did you measure the part constrained or unconstrained, and to the same torque?
Also, if the tool was above center then the cutting forces likely increase the closer you get to the axis of rotation, which could contribute to errors.
I did this test today, it has some effect but unfortunately not really close to being the dominant error source
What kind of environmental controls do you in place. For some projects Moore Tool baths the entire machine in the same oil used for machining. How often do you check the cutting tool geometry? Are you using a laser interferometer? Changes in humidity, barometric pressure and, of course, temperature can effect the parameters of the laser. Still, amazing dedication to the science, and practical use of, metro.ogy.
Great first result. As the tool moves toward the center of the piece, don't you want to maintain the linear speed of the cutting tool? ie: rotational speed increases as tool moves toward center.
For that mirror we discussed previously, a piece like you have at the start of this one is (scratches aside) more that adequate for my purpose. No rush, but it'd be useful to me and I'm happy to support your work some.
Shoot me an email at cyruslloyd4@gmail.com and let’s talk more
Sorry to add to your list of variables, but tool break in could also be a factor as well. Aluminum can also be very temperamental to turn because of the other embedded things in the material that can damage the tool edge or ruin the surface finish.
Ah, the joys of precision engineering!
if a butterfly would land on the moving axis while the lathe is turning something, would it leave a mark on the surface?
Do you have the rotation speed relative to the distance from the part center when cutting? Otherwise even if you have the perfect surface speed at one point at another it will be wrong.
10:00 you've also got a fairly large error related to drawbolt clamping pressure
Yea, looks like i should have made my groove even deeper
@@cylosgarage though that would reduce the stiffness during the turning operation even further...
The upper right picture at the nine minute mark looks like the effects of distortion caused by your fixturing method. Could that be the case?
How do you determine that the defects are caused by the spindle bearings rather than the servos that control the tool position? The error in the figure has built up over numerous machining passes, as though the errors in tool position have added coherently.
Spindle bearings are honestly the last suspect (cause PI is the best). However, you make a good point about the errors adding coherently. Spoiler alert, I had another go at it and saw a similar looking error profile in the new results. Trying to determine if this systematic error is on the metrology side or the machine side.
@@cylosgarage I was conflating this post with one of your earlier posts where you ascribed vibration problems of the spindle to airflow through the bearings. I assumed that post referred to a spindle of your own design. I should have asked if you were using the PI spindle here.
I think you really need to consider the overall stiffness of your flexture tool post / holder. The snap back could be the change in surface feet per minute as you are turning across the mirror face (outer towards inner) and causing the flexture to stress slightly.
If the constant stiffness is not there throughout the whole cut, you're going to see turning rings as the monocrystalline diamond and flexture are fighting against each other.
make a catadioptric and don't worry about the center 😅
Pretty cool, but what are you going to do with it?
A telescope is the goal, but this project has been more about the journey than the destination
Whittaker rings, maybe? Would only be a possibility if you machined at constant speed/feed.
The center gives you so little benefit, I'm curious why you didn't stop short and put a black dot in the middle
That’s what I’ve done up until now, for this part I wanted to see how good I could I could set the tool height just out of curiosity.
My guess is its probably the aluminum sub plate meeting the steel other half. Grind the matting surfaces
At this level of precision the surfaces should be lapped.
I am not sure the surface will get better due to bearings at those scales. Metal doesn't cut the way you think.
th-cam.com/video/aF7ltBT_atA/w-d-xo.html
Taking a shallower depth of cut can actually give worse finish in some cases as there is not enough material being removed to produce a proper chip to evacuate. Instead it just mushes around. This might be what you are seing. The sudden change in cut at the inner area may be the stepover vs feed crossing some threshold causing it to suddently "bite" better. The speed of the main spindle would have to be controlled relative to the current cutting diameter to create a constant linear feedrate to eliminate such thresholds.
You've made a CD.
Cool!
Thanks Dave 🙂
As soneone that writes software for devices my vote is that cause is rounding error in toolpath calculation. You are dealing with math where regular 80bit floating point isnt precise enough. Maybe inspiration for a video how whole proces goes from formula to toolpath generation to hardware used for driving and actuators. If you make second part you can see if pattern is repeatable or it is only random noise.
I have thought about this a little. I’m not sure how Linux cnc is doing the math in the Gcode behind the scenes, or how much it rounds. The benefit of having it in the Gcode is small file sizes. However, I have considered writing a matlab script to directly calculate the tool path to a specified precision and output a Gcode file of discrete points. The only downside with that method is your file size rapidly gets deep into the megabyte regime. Not sure the limit of what Linuxcnc can handle. I suppose another benefit of using the externally generated toolpath is the ability to create error maps and apply compensation to a given toolpath. hmmm...
@@cylosgarage I've done some pretty hairy adaptive paths and LinuxCNC doesn't care. I can't remember how much lookahead it has though, since a pause to refill the buffer would kill your process. All in all you're doing great work! I do worry about internal rounding errors in LinuxCNC proper though (or finding the right config entry)
way cool
Does diamond turning typically try to use constant surface speed? Would that avoid some sources of tool pressure variance across the optic?
Generally CSS is not used for diamond turning because changing the spindle speed would change the forces and hence deflections due to spindle imbalance as you move across the part. It’s a bit of a compromise in this regard.
GREAT video, now drop the imperial crap, the rest of the world moved on about a century so ago 😂
🖐️😛🤚
@@cylosgarage okay, okay, it’s still a great video. Well done and keep going 🫅