Turning A Rotary Tool Into A CNC

Now that 3D printers are everywhere, electronics are cheap, and open source software is extremely capable, just about anyone can build a CNC machine. That’s exactly what [Nikodem] did by turning a Dremel tool into an extremely capable CNC machine that’s able to cut MDF and acrylic and can engrave aluminum.

The electronics for the build are just an Arduino Uno, a motor driver sheld running GRBL, a relay for the Dremel, a few motor drivers, and a big ‘ol 30 A power supply. The build uses NEMA 17 motors, two on the Y-axis and one each on the X and Z. The CNC has a fantastically strong frame despite the 3D printed parts. It is constructed out of aluminum extrusion, with the carriages riding on some nice straight rods.

As for how well this CNC machine works, it’s pretty good. With the Gcode to cut an 80mm diameter circle out of MDF, this machine managed to cut a circle that was 80.02 mm in diameter. That’s pretty good, and getting into the territory that the error is probably in the cheap set of calipers, not the finished part itself. It’s an awesome build, and [Nikodem] has everything documented in his four-part video series. You can check the end of that out below.

32 thoughts on “Turning A Rotary Tool Into A CNC

  1. That does look nicely done! If you could swap out the NEMA17s for some NEMA23s, and use a laminate trimmer instead of a Dremel, could then mill aluminum? I suppose by then the limiting factor would be the motor driver shield.

    This is seriously neat, though.

    1. I think bigger limiter here will be plastic construction than motor drivers. It also depends on milling bits used and how deep/fast you cut. If you had very small bits you could engrave some steel. I’m planning to make that mill myself someday exactly for some engraving.

  2. Nice project. Loved to see more about the construction and design considerations. How stable is it? Any backlash? What did it cost to make, all very interesting.

    I only have a small remark regarding the article:

    “the error is probably in the cheap set of calipers”
    ehmmm… this can go both ways, don’t assume that the error of 0.02 is a measurement error. For all you know it could in reality be 81.02. There is no way of knowing this exactly with just a single measurement and a single (non calibrated) tool.
    But seriously, what are we talking about, this kind of error is silly to even discuss for a project like this. In other words… it’s perfect. And if not then it is most likely caused by the wrong input of the tool diameter in the CAM stage.
    Anyway… measurement errors (especially at micrometer level) are in many cases caused by the operator and not the tool.

    1. This is absolutely true. The .020 “error” isnt actually error in the calipers. Its probably really that far off. Cheap calipers are actually fine. NOBODY knows how to measure with them, and pretty much nobody has a standard to test them with. Even if they did have a standard, they still probably wouldnt know how to use them properly on it. The calipers are rarely wrong, even the cheap ones. I have compared many cheap calipers against ultra high end Mitutoyo’s. In nearly every case the cheap ones were dead nuts, and in most cases they werent, was a simple gib adjustment.

      1. Here’s a teardown that Adafruit did 8 years ago on expensive Mitutoyo’s. https://www.flickr.com/photos/50241702@N04/ (scroll down to the end of the page for the 3 relevant pictures). Not really the most complicated technology out there. Been around since the latest 70s. Getting down both accuracy and precision into the thous with capacitive coupling construction isn’t difficult. The hardest part I’d imagine would be making the jigs to properly align the mag-strip parallel to the “feather-boarded” PCB (i.e. those stripes on the the moving jaw that form the other edge of capacitance). Even if it was off a few arcminutes, you’d still get the exact same repeatability as a Mitutoyo. That’s good enough for the home gamer (or even the job shopper, as long as you weren’t working against dimensions to send out a critical GD&T’d assembly). The most expensive part in labor would probably be getting a perfectly matched grind on both jaws made out of what I presume would be some sort of tool steel.

        When AvE tore the cheapo 8$ Harbor Freight calipers down, the construction IIRC was identical, other than the fact that the quiesent current draw was something absurd (no SRAM, so the micro could never go off,– the “off” button just turned off the display rofflez). You’d be going through coin cell batteries like a bastard but like you said, they’re pretty much dead nuts.

        Mitutoyo is still in the game for the same reason Fluke is– pros need two things – reliability (you don’t want to be questioning your readings) and safety (if you’re hung over as fuck, you don’t want to be constantly questioning whether your readings are right or that chinese meter decided to put jumpers across that 11A 1000v fast-blow fuse).

        If I’m a linesman ohming out a step-down 14kv transformer or whatever, I sure as hell am going to pay an extra $400 for those big ol 80k40 HV Fluke attenuating probes and an extra couple hundo for that red-cased 28-ii instead of the yellow, if only to give me the warm fuzzies. Could I make a safe, reliable probe that stays linear up to a few kHz myself? Yeah sure, in fact, I’d probably be able to do it for 1/10th as much in parts from digikey and stay 3dB up to a meg AC. But, that’d just be for funsies. if I were a field tech, no question I’d pay money which has full tracability down to the serial # of each unit and will issue a recall the second a minor safety issue pops up. Everyone is like ‘oh man another Fluke recall?’ as though the quality of their units were shit. I take it as a testiment to their continued diligence and commitment to safety. (It seems to me as though) they’ll issue a recall even if it’ll cost them an arm and a leg for that specific set of serials. They’ll make up for it in sales in the long run, for the same reason I get comfort from a handwritten cert when I open a new Mitutoyo tool (rather than a photocopied chinesium tolerance sheet lawlz).

        Times change, it’s good to have options depending on what you need.

  3. I turned a faux Dremel into CNC router by making a custom 3D printed mounting bracket for cheap 3018 CNC. Actually I had to make two of them, because first one was printed badly, every hole was 0.5mm too small. New part made by different company worked great. Original router was made from DC motor with crappy shaft adapter – holding screws got loose constantly due to vibration…

  4. Just curious. In every build using a Dremel like tool the tool itself is put into the mounting bracket. I have a Dremel with a whip ( the extension that allows the tool to hang from a stand while you hold the actual bit in a smaller holder at the end of the whip – I’m trying to be as descriptive as possible here) . My intention with this was to replace the weaker Dremel tool with a Mixmaster ( handheld mixer you would use for bleeding soup, drinks with ice, etc) and have the more powerful motor ( more tourque) so I could be a little more ” heavy handed ” in my carving. Would such a tool not be easier to manipulate than trying to move the entire Dremel, motor and all along three axis in a cnc type setup? Would it not make it easier to add a so called fourth dimension as well allowing for a tilt (yaw) to the tool to cut a more rounded top surface? Am I the only one to consider that the cutting tool itself doesn’t have to actually point down?

    1. That “whip” won’t last too long in such a setup. Especially if you want to get much more torque. This is the reason why such linkages are not used in industry. And don’t forget the losses caused by the core scraping on the outer shell when it’s bent.

      I’d suggest grabbing a CNC router motor with ER11 collar. If I had $100-200, I’d get that for my CNC instead of cheap faux Dremel…

      1. Yes, I agree. Your correct about the losses in power due to the core and the outer shell. there was a fair bit of heat generated as well when i first tried it. I packed mine (and i mean packed) with a high temp. bearing grease. It remains to be seen if it will leak out all over the place when used for long periods of time but for now it has become very usable as long as the whip itself is kept reasonably straight. My thoughts were on the benefits of a larger motor being available… I suppose for a hobbyist it might be ok, but i definitely see your point about such a thing not being present in an industrial installation. The closest there i think might be a flexible spring wound shaft that is capable of taking a bit of play in the linkage as opposed to something that can carry rotation around a bend… My final thought however has to do with industry… Just as you can buy an industrial sewing machine that for a regular consumer would last longer than forever ( my mom has one – a Bernina, and the electronic clutch alone on that thing is worth the price of the machine) I find myself wondering if such a “whip” is used anywhere in a more industrial environment. I wonder if someone somewhere has adapted one to be used in a more 24 hour a day situation. Most likely by a worker in a plant somewhere grinding a part or casting on a production line?

        1. One could use two universal joints and long sleeve coupling between them, the motor would have to be mounted above the CNC machine at the height that lets th sleeve coupling to extend and retract at any given position of the cutting bit. With good bearings and grease it should be quite robust, but the machine would be rather tall. For this to work the motor would have to be mounted at the height that is at least 3 times the Z axis length. Possibly more depending on the XY dimensions…

          1. True, and there is also a limit on RPMs they can spin. So one would hae to add sets of planetary gears, one to reduce the speed, other to increase it back. Not a very good idea.

            What about pneumatic or hydraulic drive? The kind used by dentists but scaled up?

          2. As someone mentioned U-joints can vary in speed according to where they are in their rotation and their angle (It is a very small difference, but it is there). CV joints are a thing because of this very issue, depending on the size machine you were talking about these can probably be sourced easily from car parts etc.

    2. It’s been done before. Usually that’s how 3d-printer “Dremel” attachments work. There are examples in operation on Youtube videos. I don’t know how long they tend to hold up for though.

  5. The whole “Mixmaster ” motor thing is actually irrelavent. My point t being about the”whip” as opposed to the motordrive itself. A standard Dremel with a whip would be suitable for light work. However I’m willing to bet that aluminum would be easier to work with with a stronger motor.with that in mind could you not drive the whip with a 1/4 hp motor if you didn’t have to wave it in the air as you cut? Just saying… I filled that whip with axlegrease so as to reduce the friction inside. Aw hell, it’s designed to spin at 15000 rpm max anyway …

    1. I’m thinking the “whip” might be used for carving foam and maybe PCBs. If you want to carve blocks of metal, even aluminum then yah, I would certainly go with something a bit more professional!

  6. Last point. In my design for my CNC machine ( yet to be built) I used the inside pieces of a PC that we’re designed to hold the drives as my side pieces and the metal rods/ gears from a few inkjet printers as the rods to go along the x and y axis.all readily available and cheap. At the dump. There are so many desktop PC cases out there ( at the dump) that you can pick and choose the brackets to suit your design. Just drill out the rivets and your done. Thesheet metal sides of the cases are also handy. Where else can you get sheet metal pieces that cheap? (Free).

  7. The error could be due to the play in the end of the Dremel between the tool and the end of the bit. Too much forceon the tool and there is a variation. They we’re never designed for that level of precision. However a little lighter on the tool and I’ll bet that tiny amount would disappear. What happens when another exact circle is cut? what would the results be with ten identical cuts? To assume the cause of the error is a waste of time. A precise machine can be used to make a slightly more precise machine with the intervention of human hands and eyes. Then another more precise machine can be created from that one…

  8. Milling an 80 mm circle with such an accuracy with an machine of this class is indeed just perferct.
    But I do find the speculation about the sources of the 20um error quite silly.
    Just throwing a possible cause into the air without any possible way to back it up is less than useless.

    Milling a circle with a CNC machine is a usefull first test in itself though.
    You can use it to check accuracy of (discrepancies between) the X and Y axis by measuring along those axis and comparing them.
    But when you also measure along the diagonals you can check for squareness of the axes of the CNC machine.
    If it’s out of square, one diagonal will get bigger, while the other gets smaller.

    If it measures as a perfect circle then causes for errors must be sought elsewhere.
    But for MDF it could just be a change in the relative humidity in the air or surface roughness from loose fibers sticking out.
    With MDF anything within a tenth (of a mm) is just perfect, So it’s 5x better than perfect :)

  9. I have been using a Chinese ripoff of the foredom flex shaft tool for two years now. Modifications to the hand piece include adding a brass sleeve to the space on the inside (the mass keeps the tool from grabbing and stalling) all axis motor shafts have support bearings. O and don’t stress about work holding use a heated bed and a $1.00 hot glue gun heat the bed @ 80C place some hot glue on the part place it where you want and let it cool. Machine on it for several days and when you want it off heat the bed and pull your part off . My Anet A8 is now a very rigid 3D printer and there’s not a single hole in the print bed. O yea the flex shaft tool has a small drill chuck 3/16” to zero and a1/4HP motor rpm around 20,000 just take small cuts use spring passes if a surface is critical! HAPPY DESKTOP MILLING!

  10. Re: industrial use of flex shafts
    To skive out cuts and damage in the tread or sidewalls of truck tires we use about an eight foot armored flex shaft driven by a 1/2 hp motor. Carbide grit grinding wheels from 3-6″ in diameter are the main tools. The interior flex is around 5/8″ diameter and the exterior is about an inch and a half. As long as we keep it well greased (as Jason Marsh says above) it has lasted for years. It is heavy and does take a man with (or willing to last until he gets) significant wrist and arm strength to operate it all day.
    I’ll check when I go back to work, but I’m assuming it runs at 1725 rpm.

  11. A pneumatic tool sounds interesting. The connecting source/air supply tube would be light. Torque would be interesting though. It would increase as required when there was a speed reduction due to the load on the cutting tool but I wonder if that drop in speed would effect the work being done. There would there have to be some kind of communication between the tool and the controller or would a higher air pressure overall do the trick? It would also potentially be noisy unless a decent muffler was used. … Is there some kind of commercially available drive that could be modified to try this out?

  12. Re: industrial use of flex shafts
    That sounds like one enourmous tool to move around. I used to do industrial vaccuming. We would sometimes have a 4″ or 6″ rubber line to manipulate around by hand. If you taped a broomstick sized handle to the end and again one foot up it would allow the line to be manipulated easier. You would still leave with mighty sore arms at the end of the day but it would lighten the work for the hands…

  13. I did this with a Dremel (395?) some 14+ years ago, slapped some drawer slides to some plywood sheet, some salvaged steppers with 1/4″ brass threaded rod (tweaked backlash by angling the nut as press-fit in a manually milled block – it was never intended to last), built my own stepper driver board controlled via LPT (step and direction, PIC16F something if I recall, FETS may have been IRL530s), and ran G-Code via TurboCNC under DOS. For about $200 total, I built my first CNC machine before converting my mini mill to CNC (I bought a custom kit for the hardware for that conversion, was pricier, but still working great! Ballscrews and some honking NEMA34 steppers on Geckodrives). I used two steppers on Z so as not to rack. I have a pic of that setup somewhere. That said, I don’t recall checking how accurate it was, but I did some engraving and routing as proof of concept, was impressed with the results for such a cheap build. I wanted to get a feel for CNC before converting my mill. Oh – and I used a Dremel router base and some large-ish L brackets to mount the tool.

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.