I first heard about Project Escher a few months ago and found the idea really interesting: coordinating multiple print heads to introduce parallelism into additive manufacturing. Then, last week, there was a really interesting Q&A about the project in an Autodesk-internal newsletter. I’m reprinting it here, below, as I know it will be of interest to this blog’s readers.
When it came to looking for an image for this post, I decided to look for something by M. C. Escher that was in some way descriptive or representative. An obvious choice was his “Drawing Hands” lithograph, although it’s probably better suited to the RepRap project, in some ways. That said, I wouldn’t be surprised if the idea of “bots building bots” played some role in the choice of name for this fascinating project.
Here’s a video describing how Project Escher works:
Now for the Q&A. Many thanks to Kimberley and Cory for making this happen!
A Time Machine for 3D Printing: How Autodesk is making current 3D printing technologies faster
Kimberley Losey interviewing Cory Bloome.
Whether or not you have experience in 3D printing, you’ve probably heard that 3D printing is slow. With the recent announcement of Project Escher, Autodesk is giving 3D printer manufacturers the ability to create machines that can build bigger and faster without sacrificing precision.
Project Escher is a combination of control system, communication protocol and printer architecture. By using Escher technology and building machines that fit a loose set of architectural requirements, hardware manufacturers can build a fundamentally new class of faster printers.
Here’s a video to see it in action.
Below is an interview with Cory Bloome, Project Escher Hardware and Control Systems lead.
KL: Why is Project Escher important to Autodesk?
CB: Like many groups at Autodesk, we are working on ways to significantly advance additive manufacturing. The Escher system is a new way to control additive machines that have more than one print head. It’s a technology that, with some simple and some not-so-simple tricks, makes other existing technologies many times faster simply by adding more print heads to work on a given job.
We’re admittedly biased, but we think it’s an important addition to the Autodesk suite of products and it is a great example of deep integration of hardware and software making a better end product.
KL: Where did the idea for Project Escher come from?
CB: There were many people involved but the seed of the idea came from Greg Meess, Principal Engineer on our team. At the time our team was doing lots of brainstorming and exploring ways to improve FDM printing for industrial use. We were initially working on several technologies but have since narrowed to focus on this one.
When Greg proposed the concept we talked about it quite a bit. We saw the potential but needed to convince ourselves and others in the company that it worked. So we spent the next 2 1/2 weeks in a full court press, cranking the first prototype out- software, hardware, communication. We are fortunate to have a very diverse and well-suited group skills-wise, and that helped tremendously. Greg worked on algorithms. Matt Hovanec, an EE who codes, worked on communication and the motion system. Kenny Mejia our lead mechanical engineer, designed the unit, did the CAD, and built the first prototype machine. Everyone worked together to make a rough but functional unit in a really short period of time.
The 1.0 prototype printed parts cooperativelyusing two build heads that moved independently. That was enough to convinced us that the concept worked, and it’s been our team’s focus ever since.
KL: Some people question FDM for industrial use. Why have you chosen to build an industrial-grade control system for FDM?
CB: One of the cool things about the Escher control system is that the algorithms and control protocol are applicable to any deposition-based process, not just FDM. It could work for concrete printers, metal deposition, ceramics, bioprinters, just about anything where you’re depositing material. And a couple of days ago we had a very exciting conversation with Alex Oster, founder of Netfabb. Alex pointed out that some of the challenges faced by multi-laser DMLS (Direct Metal Laser Sintering) machines are strikingly similar, in terms of geometry, to multi-extruder deposition processes. We haven’t done a deep investigation yet but on the surface it looks like Escher control algorithms might work to speed up multi-laser DMLS machines.
What I’m getting at is that while we’re currently testing with FDM, the underlying algorithms and structure of the technology have wider application. FDM is a proving ground. That said, we’re huge fans of FDM. We think in some ways it still sucks, but each day to a lesser degree. In terms of cost, material selection, ease of implementation and adaptability to high-performance structures such as embedded continuous fibers, conductive paths etc., we think has enormous potential. There are several companies like Voxel8, Markforged, and Arevo Labs extending the limits of FDM technology, starting to tap into that potential.
KL: What are the biggest challenges in additive manufacturing and why do you think the Escher system addresses these challenges?
CB: There are many challenges in additive. We’ve chosen to focus on the trade-off between speed and precision, and the challenge of speed in general, in deposition-based processes. For deposition, detail is limited by the size of the hole that you’re squirting material out of. You can’t print a tiny feature on a part if your nozzle is the size of a fire hose. And, nozzles have speed limits. There is a ceiling to how fast you can accurately squirt material. The limits are different for different systems, and they’re a little fuzzy, but they’re always there.
The Escher system takes a really simple theoretical approach to these speed limits regardless of system details: add more independently moving nozzles. The devil, of course, is in the details. But software is good at details. We’re letting the software handle the movement, and we’ve created a guideline for printer manufacturers that teaches them to build hardware, in their own configurations, that plays nicely with the software.
KL: The system you’re building is something completely new. In the process of development if you could have called someone to ask a question, what question would you have asked?
CB: In addition to the algorithms themselves, there are two areas where we faced most of the challenges. The first area is that printing large parts with multiple heads is an entirely new process that has required many new approaches to printing itself. There aren’t any web forums or experts we can call on when we get stuck. We’ve had to come up with new processes to calibrate each of the print heads to each other, new ways to purge and start printing with nozzles that were waiting for other nozzles to finish, new methods to negate warp in large parts, and on and on. Sometimes we have had to create new processes or solutions where we expected smooth sailing. At times it’s frustrating. But it’s never boring!
The second group of challenges has been with the physical design of the test machines. One of our early learnings was that “nozzle density” as we call it, is king. Meaning that to maximize print speed, the many independently moving nozzles on a given machine must be able to get as close together as possible while printing. But to achieve this tight nozzle spacing we had to violate some well-established rules of thumb for good machine design. Those rules of thumb are there for a reason, so we’ve had to come up with some new mechanical tricks. It would have been nice if we could have had someone to call when we were breaking the rules in our machine design textbooks!
Recent Press on Project Escher: