Precision 3d Printer Project

[CM800]

Hi All,

I'm currently working on building a precision 3d printer / laser cutter unit.

Just tested one of the Parker MX80L Linear stages I brought off ebay, Wanted to share the sucess I'm getting:




Not bad for 3 minutes of PID tuning work (+ 5 minutes of configuring)

Cost £1k for 2 stages second hand, pretty damn good price for what I'm getting.

[jeremy]

Is there a subforum for distance-nuts yet?  Can't let the time-nuts and volt-nuts have all the fun.

As someone who was eyeing off those stages too, the problem with "chasing microns" in these encoders is that rarely is the encoder actually measuring what you want. I don't want to rain on your parade, but perhaps you might consider thinking further about your sources of error.

In addition, I'd be extremely surprised if that encoder actually had absolute accuracy anywhere near 0.2um. You'd need to calibrate it with a stabilised interferometer to get close to that.

For example, the coefficient of thermal expansion of 6061 aluminium is about 25um/C. I have no idea what material the stage is made of, but if it was 6061, you can expect an expansion along the long dimension of 3um for a 150mm stage for each degree C. Where I live, that would give uncertainty of something like 30-45um over the day. To beat this problem, the usual approach is either very good climate control, or make everything out of invar. Often in really serious stuff like interferometer design, you have to do both. Temperature is such a big problem that the mitutoyo Kiyohara factory where they make fancy scales is actually underground to avoid temperature cycling.

I'm fairly sure that FDM 3D printing is 90% tuning temperature and material parameters, and 10% accurate motion (I think anything beyond steppers is overkill). Worst part is that FDM has heaters and large temperature gradients all over the place! SLS needs the "bed" to be still during XY motion, because it's full of fluid and that will slosh around. Laser marking could work, but you'd need a very good fibre laser (maybe a $3-5k 2nd hand) to get the spot size anywhere near a few micron. A low cost violet laser from the usual suspects would likely have a poor beam shape which is not possible to focus tightly, and CO2 lasers have a wavelength of 10um, so obviously you are going to be diffraction limited before you get anywhere near the resolution of the encoder.

With all this in mind, I wish you the best of luck! I'll leave you with something a colleague of mine once said: "Building optical systems is like playing with adult lego. Really, really, really expensive lego".

[SiliconWizard]

Well, accuracy of positioning is important, but the overall precision you'll get with deposition through an extruder will still severly limit you, so I'm not sure a µm-precision would make sense. Likewise with a laser, you'd have to use a very expensive laser unit to be able to leverage this kind of positioning accuracy.

[CM800]

Is there a subforum for distance-nuts yet?  Can't let the time-nuts and volt-nuts have all the fun.

As someone who was eyeing off those stages too, the problem with "chasing microns" in these encoders is that rarely is the encoder actually measuring what you want. I don't want to rain on your parade, but perhaps you might consider thinking further about your sources of error.

In addition, I'd be extremely surprised if that encoder actually had absolute accuracy anywhere near 0.2um. You'd need to calibrate it with a stabilised interferometer to get close to that.

For example, the coefficient of thermal expansion of 6061 aluminium is about 25um/C. I have no idea what material the stage is made of, but if it was 6061, you can expect an expansion along the long dimension of 3um for a 150mm stage for each degree C. Where I live, that would give uncertainty of something like 30-45um over the day. To beat this problem, the usual approach is either very good climate control, or make everything out of invar. Often in really serious stuff like interferometer design, you have to do both. Temperature is such a big problem that the mitutoyo Kiyohara factory where they make fancy scales is actually underground to avoid temperature cycling.

I'm fairly sure that FDM 3D printing is 90% tuning temperature and material parameters, and 10% accurate motion (I think anything beyond steppers is overkill). Worst part is that FDM has heaters and large temperature gradients all over the place! SLS needs the "bed" to be still during XY motion, because it's full of fluid and that will slosh around. Laser marking could work, but you'd need a very good fibre laser (maybe a $3-5k 2nd hand) to get the spot size anywhere near a few micron. A low cost violet laser from the usual suspects would likely have a poor beam shape which is not possible to focus tightly, and CO2 lasers have a wavelength of 10um, so obviously you are going to be diffraction limited before you get anywhere near the resolution of the encoder.

With all this in mind, I wish you the best of luck! I'll leave you with something a colleague of mine once said: "Building optical systems is like playing with adult lego. Really, really, really expensive lego".

Oh for certain, I'm well aware of those numbers, I know that this is mostly on the resolution side of things.

Still, it's fun putting the finest dial indicator you can find, and be able to control the position on finer degrees of accuracy you can see on it.

I didn't know that Mitutoyo's facility was underground... that's pretty damn impressive.

I'd love to get hold of an interferometer and give it a test.

There's small lies, big lies, those that are on your oscilliscope... and those that come from an interpolated encoder. 

Considering how small the movements are, it's impressive the drive is able to keep it within a good few counts even with me wiggling my hand over the unit.

It really is mostly 'big boys toys' expecially considering I'm depositing molten plastic ontop of the plate. 

[metrologist]

Eight millionths of an inch or two ten thousandths of a mm is a very small increment, laboratory grade rather than production.

I recall trying to use a DIY laser interferometer to measure distance in CNC machining - variations due to temperature never came up, but vibrations were a huge problem. The entire machine had to be built with large granite surface plates and you still had to tip-toe around. It was not workable.

Do you know how these encoders work?

[jeremy]

Eight millionths of an inch or two ten thousandths of a mm is a very small increment, laboratory grade rather than production.

I recall trying to use a DIY laser interferometer to measure distance in CNC machining - variations due to temperature never came up, but vibrations were a huge problem. The entire machine had to be built with large granite surface plates and you still had to tip-toe around. It was not workable.

Do you know how these encoders work?

Unless you were using a temperature stabilised gas laser, for sure the laser would drift in wavelength more than the material would move unless you gave it a nice long warmup time. Usual practice today is to use relatively light tables (still >100kg though) and active damping. Interferometers are a bit painful like that... if you want to measure nm displacement, you also have to worry about nm vibrations. We banned a guy from the lab because he liked a good joke but his laughter was too deep; you could pick it up in all of the optical setups.

Pretty sure it’s a grating style encoder: http://www.renishaw.com/en/rgh22-incremental-encoder-system-with-rgs-linear-scale--6443

[CopperCone]

Oooonoooo it godzillaaaa

[CM800]

Dang, must have failed to send my reply when I left work...


The axis itself is this:

www.parkermotion.com/manuals/mx80/MX80LManualRev6.pdf

and the drive being used:

http://www.elmomc.com/products/DC-whistle-servo-drive-gold.htm

The encoder is indeed a renishaw diffraction grating based encoder.

20um sincos interpolated to 10nm

The good thing with an ultra-high resolution, is that the servo drive has more data to work with, allowing much better control.

If I was to have 1um resolution, I would be unable to make much use of a 10um accurate encoder without painfully extensive tuning.

[CopperCone]

I think you need air bearings to make use of that stuff in the very least.

[SiliconWizard]

I think you need air bearings to make use of that stuff in the very least.

I guess so.   

But I must admit it's a great toy for grown-ups!

[CopperCone]

I heard semiconductor industry air bearings are not too expensive used