Clank: Fabricatable, Modular CNC
Clank is a highly fabricatable, modular, tool-changing CNC machine from the MIT CBA MTM Project.
Clank is an open, highly fabricatable1 CNC machine that implements an automated toolchanger2, so that we can use a common motion platform to accomplish a variety of CNC tasks. This design serves the “mid-macro” length scale, fabricating almost anything with characteristic lengths of ~ 10 - 1000mm, with feature sizes down to 0.25mm at a resolution near 0.05mm3.
Clank refers to a machine platform consisting of three variants; the -corexy (which is the most up-to-date), the -stretch and the flying-xy, or -fxy, which serve “wide-and-flat” and “tall-and-skinny” tasks respectively. All variants implement similar design languages, very similar BOMs, and use the same toolchanger and tools.
The machine platform is entirely 3D Printed and a complete Hardware BOM (not including controllers) costs about $400 USD. The tools required to build the machine total an additional $500, including the 3D Printer used to fabricate parts.
The Tool Changer
The heart of this project is really the tool changer: it is also completely 3D Printed, and differs from other designs in its high-stiffness clamping mechanism. The design qualifies as a kinematic mount but uses a non-standard layout in order to automatically clamp and release tools.
I’ve tested the toolchanger for repeatability, it’s better than I had hoped at ~ 5 microns in any axis. Errors are well centered as well.
|axis||standard deviation after pickup, mm (inch)||average from center, mm (inch)|
|x||0.0047 (0.0002)||5.921e-16 (2.331e-17)|
|y||0.0033 (0.0001)||-2.220e-16 (-8.742e-18)|
|z||0.0047 (0.0002)||-1.480e-16 (-5.828e-18)|
Hysteresis on the motion system is not as killer, but also not egregious, ~ 50 microns-ish, worse on the Z axis, which I would attribute to the printed pulley having some poorly manufactured geometry.
|axis||hysteresis mm (inch)|
Qualitatively, Clank can 3D Print, mill PCBs, and mill plywood. It can also heft around ~ 3kg of clay. The platform stiffness remains to be tested, and will likely vary greatly across instance sizes and setups.
The global movement of fab labs and makerspaces has the potential to scale exponentially if the goals of fabricatability1 are achieved for a wide variety of CNC machines. Namely, machines should be cheap, easy to build using a small set of tools, and easily reproduced globally.
In particular, FDM4 3D Printers have become globally ubiquitous, with “good enough” models approaching the $100 USD cost mark. This is no small part thanks to the open source hardware movement for (1) developing designs and (2) creating market demand. Or it’s the other way around, who knows.
However, fabrication is fundamentally multi-process, and other tools like Laser Cutters, CNC Mills, Pick-and-Place machines, PCB Mills, and Vinyl Cutters are less common in the open source, and many designs are not highly fabricatable.
Clank aims to bridge this gap, providing a boot-strappable motion platform that is modular by default, implementing a tool-changer that allows almost any process-specific tool to be mounted. Using Clank as a common platform, we can quickly develop new end-effectors for the “long tail” of specialized CNC processes, and - to boot - we could develop the “fab-lab in a box” wherein one machine, with one small BOM, can provide all of the functionality usually acquired with an ~ $50k USD investment across five or more machines.
There’s one more thing: because our toolchanger is automated, it becomes possible to develop multi-process operations, i.e. parts can be FDM printed and then CNC milled where tolerances require it as in E3D’s ASMBL system, or garment net-shapes can be marked up with a pen plotter before being knife cut, or PCBs could be milled, and then components placed and soldered all in one ‘job’ on the same machine.
The open source 3D Printing community is alive with modular tool-changing designs. Most are focused around FDM-sized tools, whereas Clank focuses just one step up in the size / stiffness spec. Other tool changers also require some machining or otherwise non-FDM fabrication steps, making Clank’s design the easiest (and cheapest) to implement.
There are also handfuls of other “mostly-printed” CNC machine designs, most notably the mpcnc. I’ve got to shout out the voron project which is certainly not cheap-and-cheerful, but is a gorgeous design nonetheless. The OSHW communities / internet machine nerds are hard at work!
Mechanical engineers will normally optimize designs to hit some well defined performance specification; in the case of CNC machines these specifications would be process specific, i.e. a CNC mill should be highly rigid and can trade weight for stiffness without much repercussion. A 3D Printer should be extremely fast (light) but doesn’t need to be terribly stiff.
Clank is in an odd spot, because it wants to be “pretty good” at a wide variety of tasks. It’s also been designed under the heavy constraint that every custom part should be FDM printed, meaning thermoplastics - where the high-end flexural modulus is about 3 GPa5, whereas i.e. aluminum’s modulus is around 60 GPA, for the same weight6.
As such, the design uses big, blocky pieces of plastic to eek out whatever stiffness we can. It also focuses on maintaining small load paths through plastic bits, quickly routing loads from extrusion-to-extrusion.
The designs are parametric, meaning that you can build a Clank of whatever size you’d like. This desirement, along with the minimal-BOM desire, precludes the use of COTS7 linear rails. As a result, you’ll see a repeating pattern of roller-bearings on extrusion in the designs.
There’s a few little nuggets of design wisdom in the mechanical details section of this website, check them out!
Project Status and Roadmap
Clank is mostly a research mule for myself, but I’m keeping designs and docs in the open as best I can. At the moment, the “corexy” variant is fairly stable, quite cheap and as simple as I can manage.
Controllers for the machine are an ongoing research project: we are trying to figure out out how to build machine systems in a principled, layered manner. Our best attempt with that is the modular-things project. Soon enough, clank-trollers will be modular-things examples.
For the time being, though, you can find success with off-the-shelf controllers, i.e. TinyG, Duet, Smoothie, Klipper, or whatever control solution you see fit to implement.
Besides stable and available controllers, the development roadmap looks towards expanding the number of tools available for use with Clank. In that vein, collaborators who would like to develop end-effectors (mechanical wizards) or who would like to collaborate on modular controllers (embedded development / networking / controls savants) should get in touch with me directly.
“Fabricatability” is a fun term popularized by our friend Jens Dyvik and formalized in this paper …. I figure the word mostly speaks for itself. ↩ ↩2
Many CNC machines implement only one process, i.e. a 3D Printer can 3D Print, a Laser Cutter can Laser-Cut, etc - somewhat unsurprisingly. However, we have for a while been collectively dreaming of machines that separate the generalizeable aspect of CNC machining i.e. computer-numeric motion control from the specific aspect, i.e. process-specialized tools, or end-effectors. ↩
That’s about 1/2” -> 40” - and this statement should be qualified with a few caveats: (1) at the time of writing (8th of Jan 2022) I haven’t rigorously tested the feature size & resolution claim. (2) the machine has configurable sizes, and the “default” spans only a 600mm build area, however, I’m pretty confident the thing can be scaled out to a 4x8’ bed. Tests are on the roadmap. ↩
Fused Deposition Modelling, or as I like to call it, the “hot glue gun” style of 3D Printer, where thermoplastics are squished out of a hot nozzle to build things layer-by layer. ↩
HDPE is ~ 1 GPA, Nylons are ~ 3, PLA is somewhere in the 1.6-2.5 range, etc. High performance plastics, like those used in an off-the-shelf cordless drill, are normally glass-fiber reinforced up to ~ 60% or so - those parts can reach towards 6 GPA, but we can’t easily FDM print with them. ↩
There’s a little table of short-hand material specs here on the how-to-build-machines MIT class page. ↩
Commercial Off-The Shelf: things we can buy, rather than build. ↩