Autodesk Open Sources Ember 3D Printer

If you’ve ever been interested in what goes on inside a (roughly) $6000 DLP stereolithography printer, you might want to check out the recent announcement from Autodesk that open sources their electronics and firmware for their Ember 3D printer. The package includes the design files and code for their controller (which is more or less a BeagleBone black with a USB hub, and more memory. It also has two AVR controllers for motor and light control.

The mechanical design has been open since May. Autodesk point’s out that many of the parts are injection molded and might be hard to duplicate by other means (although apparently their prototype machines used 3D printed parts). The DLP projector, too, is something you’d probably source and not duplicate from these files. Autodesk says, “Our thought is not that you would duplicate Ember, but extend it.”

That may be true, but our guess is someone is going to build one, perhaps making the design cheaper and easier to build in the process. Autodesk seems to be very committed to openness given that they also make their resin formulation open source, as well.

Not that there aren’t other stereolithography designs out there. The Peachy, for example, is famous for using a water drip to control layer height. If you are an accomplished scrounger, you might even be able to build one for under $100. However, Ember is a professional product and having its internals exposed makes it very educational, even if you won’t build an exact duplicate of it. Whatever you do end up building from this info, we’d like to hear about it!

20 thoughts on “Autodesk Open Sources Ember 3D Printer

  1. Yes, the one big ‘drawback’ to the present design is the optics system (at least the last time I checked the Ember’s source out) is basically a TI DLP EVM. While this does away with a great deal of upfront ‘complexity’ and despite the massively decreased cost of the ‘so-called’ hobbyist boards (i.e. BBB), ‘professional’ level EVMs still sell for a pretty penny premium– and further, for a use case such as this, offer much more functionality than is actually needed.

    It is great that they are open-sourcing everything like this, but the first task would to be to create a custom DLP board/setup for this application (Smirk. Using existing DLP/DMD chips of course… that part still not quite DIY). It would make a big difference in cost overall.

  2. Is there something in that machine design that the DIY community wants? If it takes $3000 (just spit-balling) to build one, and has a tiny build volume for machine volume, then why would we want it?

    1. On the one hand if there weren’t some appeal about the design I’d think Autodesk wouldn’t have pursued it this far.

      On the other I wonder if maybe there’s a reason this is getting open-sourced: they don’t see a way forward and are hoping to crowd-source the turd-polishing stage of R&D.

    2. Compared to a laser SLA, a DLP projector setup is MUCH easier to program for and work with. I’m not sure if the $3k mark is quite right – you could pick up a new projector in the $400 range easy and then the hardware and motor will cost you around $100 new. You could do this for $500 if you wanted to no problem.

    3. The entire community wants the software that turns 3d objects into beams of light on the bottom of the resin tank. When Autodesk announced they were open sourcing *everything*, they didn’t bother to include the control software.

      This is the same case with the Form1; the control software isn’t open source. Any other resin printer…. the control software isn’t open source. Compare this to filament printers where you have Cura, repetier, skeinforge, and a dozen other hosts and slicers that are open source, and you start to realize why resin printing hasn’t taken off. Autodesk refuses to open source the one thing the technology needs for wider adoption.

      Yes, there’s Creation Workshop, but there were grumblings from the dev about using it for commercial purposes. That’s relatively resolved now, but not quite. Yes, there are people working on projector-based resin printer hosts, but they’re very rough right now. Autodesk has a complete solution that would get this technology off the ground, and they will open source everything except what would be most useful to the community.

      I heard the Ember was ‘open source’ two weeks ago at OHS. The host software was not open sourced. There’s a reason I didn’t write about the Ember then.

        1. They have open sourced the software in the printer. [Brain’s] point is they have not open sourced the “slicer” software that runs on the host. Granted, that would be nice, just as if someone open sourced a Postscript laser printer it would be nice if they also gave me software to create PostScript. But, to me, it doesn’t devalue that they are opening what they have up for scrutiny. Sounds like [Brian] probably knows more about it, but I don’t know if they have said they will not open source the rest or if they can’t or if it just hasn’t happened yet. After all, some months ago it was just the resin, then the resin and the hardware, and now the resin, the hardware, and the firmware. As I mentioned (as have others) you’d still need to handle the light source, etc. so it isn’t clear how reproducible this will be in practice for several reasons.

      1. Could be me but the only thing you need to do is show an image on the DLP projector for a certain amount of time? Cura engine has debug features that output all the slices as images, which is probably all you need. Granted, it won’t have support, but slicing those images is not that hard.

        1. Slicing is a ‘solved’ problem. That’s not to say there’s still not work to be done, but slicing isn’t the problem.

          Yes, resin host software is ‘just’ showing an image on a DLP, but you need to do that in conjunction with moving steppers and servos. There’s also the issue of calibrating the size/shape of the projector, and maybe even using galvos and a laser instead of a projector.

          It’s not a *hard* problem – Formlabs did it, autodesk did it, a *lot* of people have done it. Not many people are doing it open source, and when you look at the likes of Cura and repetier, you start to see this is why we don’t have resin printers yet.

  3. Oligomer: Allnex Ebecryl 8210 39.776%, Sartomer SR 494 39.776%

    Photoinitiator: Esstech TPO+ (2,4,6-Trimethylbenzoyl-diphenyl-phosphineoxide) 0.400%

    Reactive diluent: Rahn Genomer 1122 19.888%

    UV blocker: Mayzo OB+ (2,2’-(2,5-thiophenediyl)bis(5-tert-butylbenzoxazole)) 0.160%

    Great, now these above companies are going to have to send samples out to everybody. Maybe they will sell it in smaller than drum quantities for most of the raw ingredients (photoinitiator and UV blocker excluded)? Plus, this is hardly an optimized resin but at least it’s a step in the right direction. Stratasys / 3D Systems are in for a bad time if they keep up business as usual in light of this newer, open competition.

    1. Also:

      Esstech: Exothane 10 24.9 % wt./wt.

      Rahn: Tripropylene glycol di-acrylate 74.8 % wt./wt.

      Various suppliers: TPO (Esstech,Diphenyl(2,4,6-trimethylbenzoyl)phosphine oxide) 0.20% wt./wt.

      Mayzo: Benetex OB+ (2,2’- (2,5-Thiophenediyl) bis [5-tert-butylbenzoxazole]) 0.016 wt./wt.

      Making this based on tripropylene glycol di-acrylate makes it a pretty basic resin! Not bad and fairly cheap but pretty run of the mill performance wise.

      1. For the TPO, the above should read this instead. Esstech is one but there are many different suppliers.

        Various suppliers: TPO (Diphenyl(2,4,6-trimethylbenzoyl)phosphine oxide) 0.20% wt./wt.

  4. A few years back, I had started to create a web-controlled DLP print controller. I found FreeSteel slicer to be very good :http://www.freesteel.co.uk. Looks like their IP was purchased by Autodesk, though…not sure if their slicer is still available.

    The idea was to upload images to a web service. The web service would slice the model into its constituent layers as a single png per slice. Javascript would consume a JSON file, and display each slice for a predetermined amount of time. The web page would then play audio files to send command data to the controller, with clock on one audo channel and data on the other. The beauty of this setup is that it’s completely cross-platform and could turn even a tablet into a print controller (if it supported HID over USB) when connected to a DLP. Biggest challenge would be calibration…but not insurmountable.

    I spun a PCB with an ATMega88 processor and H-Bridge driver. The processor would interpret the audio signal as a control signal, moving the print platform via stepper motor with each control message. In response, the hardware could then output a keystroke via V-USB HID interface. The web browser would capture the emulated keystroke, allowing bi-directional communications to happen between web page and controller. The command / response nature of these transactions is very low bandwidth, and this proved to be a reasonable mechanism.

    I was working with someone else responsible for the mechanicals, but he stopped short of fully developing the mechanical bits of the project…so it just sort of died, unfortunately.

    Demo video of the hardware at https://www.youtube.com/watch?v=s5qACvU-C28

    Software proof-of-concept online demo at : http://nhsm.dizot.net – select a model and hit the “Z” key twice on the keyboard to get the slices moving when presented with a black screen (simulates an ACK of the HOME and START commands sent by the browser).

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.