Cardboard CNC Machine Boxes Up Both A Tool And A Framework

Want to build up a desktop CNC machine without breaking your pocketbook? [James Coleman], [Nadya Peek], and [Ilan Moyer] of MIT Media Labs have cooked up a modular cardboard CNC that gives you the backbone from which you can design your own machine.

The CNC build comprises of design instructions for a single axis linear stage and single axis rotary stage with several ideas on how to combine multiple of these axes together to construct a particular machine. Whether your milling wood, laser-engraving your desk, or pipetting your bacteria samples, the designs [Dropbox] and physical components can be adopted for your end-application.

Perhaps the most interesting aspect of this project is that, at the high level, it is not just a cnc, but a framework known as Gestalt. This architecture enables users to develop their own machine configuration consisting of multiple software nodes linked together with high-level Python Code. Most of the high level computation is organized by a Python library that calls compiled C-code. This high-level framework processes instructions through the desired machine’s kinematics to output commands to the motor controllers. Finally, the top-level interface does away with the archaic GCode with two alternatives: a Python interface consisting of function calls to procedures and a remote interface to make procedure calls through http requests. While the downside of a motion control language is that commands have no standardization; they are, however, far more human-readable, a benefit that plays into the Gestalt Framework’s aim “to be accessible to individuals for personal use.”


In the paper [PDF], [Ilan] expresses the notion of a tool as an impedance-matching device, an instrument that extends the reach of our creativity to bend and morph a broader range of shapes into forms from our imagination. Where our hands fail in their imprecision and weakness, tools bridge this gap. Gestalt and the Cardboard CNC are first steps to creating a framework so that anyone can design and realize their own impedance-matching device, whether they’re weaving steel cables or carving wood.

The folks at MIT Media Labs a familiar heavy-hitters in this field of low-cost machinery, especially the kind that fit in a suitcase. We’re thrilled to see a build that reaches out directly to the community.

via []

33 thoughts on “Cardboard CNC Machine Boxes Up Both A Tool And A Framework

  1. This seems really, really cool (and something I would love to play with), the only part I haven’t quite figured out is how one goes about fabricating the Gestalt Nodes themselves? No, electronics are not my field of expertise at all, just curious as to how one could go about making them & actually trying this framework out (I can see a lot of potential uses for me in my everyday work with something like this).

  2. This is a very interesting project. All of the parts seem to be readily available / makeable except for the Gestalt nodes. I could not find a link to buy them or download the design files.

  3. Could be a cool project . But the grammatical errors, and confuse descriptions, don´t help its case, though. Also , the site could have avoided those video/animations of people plugging cables and been more descritive about the motors and axes. How do one fix the axes to the cardboard structure ? Maybe it is buried somewhere in the page, but I didn´t find it in a quick glance.

  4. Could be a cool project. But the grammatical errors and confusing descriptions don´t help its case. Also, the site could have avoided those videos/animations of people plugging cables and been more descriptive about the motors and axis. How does one fix the axis to the cardboard structure? Maybe it is buried somewhere in the page, but I didn´t find it in a quick glance.

    There, I fixed it for you.

    1. Thanks for the correction. The confuse x confusing I could argue about, but now that I read it again the use of “axes” instead of “axis” was really wrong …maybe if they decide to build an assassin cnc machine…

          1. “Axes” (ax-eez) is the plural of “axis”. As distinct from the big bladey things a very geeky James Bond villain might attach to his CNC machine.

  5. Am I the only person that doesn’t understand why they feel the need to change the language for the “controller”? What is wrong with G-code? It has been used for years. There is an actual Standard. Mills, lathes, 3d printers, for the most part*, all make use of G-code to program the moves. Will obfuscation help in the long run?

    *There are CNC machines that use conversational language(?). I believe many/most of them can still be programmed with G-code, if you are able to set the proper flag/configuration.

    1. The advantage of G-code, as I’m sure everyone knows, is the massive amount of existing software out there that supports it. If they want to create something better, fine, but for a project like this, keeping G-code would make it a thousand times more useful than getting rid.

      The “superior” language should be separate from the machine itself, it’s 2 different projects. Perhaps their new language could output G-code, if it’s possible (if G-code supports all the functions their new language uses), as a handy compatibility feature. That would be it’s proper place, as a new tool that fits in with the existing world around it.

  6. Interesting project. Certainly cool. But I see two main problems:

    1. cardboard is not stable. The huge spring used for their drawing pen speeks for the mechanical imprecision of the whole design. Why not just 3D print the mechanical parts? Same for shitty aluminium rods etc. I would only use their controller and rig the mechanical part myself.

    2. why replace G-Code? The huge success of 3D printers was only possible because of G-Code. Because there is a standard. So does MIT teach that it’s more important to be unique than to be useful but standardized?
    Every CAM program will be unable to produce toolpaths in their homebrew language. Why? G-Code is idiot simple!

    Also, I couldn’t really figure out how exactly their language was working… But anyway.

    1. From what I understood from the above links they seem to envision the use of cardboard as way of speeding up prototyping with the goal of replacing cardboard with more sturdy materials once the prototype is finished.

      As for replacing G-Code, that’s a mystery too me too.

      1. I assumed the idea is, with a proper end tool, that you’d use this to cut wood that would then be used to replace the cardboard parts. Once you’ve got your wooden version I suppose that could mill out metal for the same purpose. Same way 3D printers can print out parts for other 3D printers. This is a bootstrap.

    2. You need the huge spring because it is a pen, not an endmill. The paper isn’t exactly flat, etc. I’m currently building a large plotter and using fairly rigid t-slot stuff and found the pen needs pressure instead of accurate positioning to get nice pressure on the writing surface.

  7. This is utterly squared crap. Sure, milling wood, haha. Best way to scary off anyone interested in CNC.
    “impedance-matching device” wtf, did they ever heard about p.ex. mechanical oscillation feedback in the drive control loops? these are the true problems (among others also complicated), good luck with cardboard!
    Many have tried to improve G-Code in the past 40 years, nothing got through, continue to dream on…

    I understand people when they say “student can’t do real work, they’re just theoreticians”

  8. While I partially agree, Why re-invent G-Code? I also teach CNC (milling primarily) , and while In Theory, yes it’s standardized as a format for moves, practically EVERY machine implements or Doesn’t implement it’s own subset, and prefers it’s own non standard format. 3D printers have actually helped raise usage for machines that didn’t grandfather in older ways of handling more advanced functionality, but let’s be clear. G-Code isn’t Hard to learn –plenty of my students code their first parts by hand until more comfortable with CAD-to-CAM software– but it IS obtuse at times, and Isn’t a straight forward way to handle advances in kinematics and exotic hardware, especially for a new student building/tinkering with non-traditional computer controlled automation.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

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