How NOT to make a CNC Mill (part 1)


Computer Numeric Controlled Mills (CNC for short) are great tools to have around.  Wouldn’t it be great if we could make one from “scrap” parts?!

This series was originally intended to be a tutorial on how to make a CNC mill “on the cheap”.  This build was pieced together mainly from different bits and pieces gathered from some old projects, including an original controller PC that was over ten years old.  The total mechanical parts cost was $60-$90 to start (not including electronics or the re-purposed parts).  It’s been an on-off project for quite a while, many things have been learned – but, currently, it just doesn’t work as required.  That’s what these articles are about – detailing mistakes made so others can learn from them!


The Beginning


CNC mills have been popping up all over the place the last several years.  One that was especially of interest was a Mantis 9 Mill, created by David Carr over at the MIT Media Lab (definitely check out the media lab – they provide a constant stream of really cool projects).  The reason for looking in detail at the Mantis 9 was because it was inexpensive and shouldn’t require any serious machining to create (more on this later).

Mantis 9 Gotcha’s

First, it should be noted that this is in no way griping about the Manits 9, but merely observations – many of which are specific to this implementation.  The mantis 9 project is all about a cheap CNC Mill.  The project page says this was accomplished in around $90 per mill for the mechanical portion.  While this was true for raw mechanical BOM, when amortized over several mills (and not including shipping charges), you’d be quite hard-pressed to actually buy all of the materials needed for under $100 after shipping charges.  There were also some details that were left out – like the ACME threaded rods that require a fairly precise bore through the end (which should be done by a machinist), the delrin lead screw nuts that were tapped with an ACME tap (definitely not included in the $100).

What’s On Hand?

The main reason for getting into this project to begin with was a few precision rods – a dusty bag of precision rods (5/8″ 24″ long) – enough to make an entire mill!  Looking around a bit more uncovered an assortment of discarded NEMA 17 and 23 stepper motors – sounds like the making of a mill!  So – that’s how/why this entire project got started – because some old “junk” turned up in a bin and sparked a memory of “that cheap Mill MIT put together”.

First, the mill was modeled in Google Sketchup.  Since the linear rods were significantly longer, an entirely new model needed to be created.  This mill was going to be fairly large (well, large for a “small” mill ;-)), having an original working envelop of around 20″ x 10″ x 3″.

pcbMill model 1

The material chosen was 1/2″ MDF because it should eliminate concerns with warped plywood.  You’ll see this used quite a bit for things that need to be cut cleanly, it has a uniform consistency that doesn’t contain any voids like plywood and doesn’t splinter.  MDF is quite strong along it’s length, but is surprisingly worse than plywood for holding on its “end-grain”.  This created quite a few issues during a few stages.

What’s Still Needed?

So, we’ve got the precision rods for linear motion, some motors, and some MDF.  We still need some bushings (to ride along the precision rod), threaded ACME rod, and lead-screw nuts – McMaster Carr, here we come! Here’s what was ordered:

2868T104 12 Each Sae 863 Bronze Sleeve Bearing, For 1/2″ Shaft Diameter, 5/8″ Od, 5/8″ Length
98935A817 2 Each Plain Steel General Purpose Acme Threaded Rod, Right Hand, 3/8″-12 Acme Size, 3′ Length
95155A111 3 Each Brass General Purpose Acme Hex Nut, Right-hand, 3/8″-12 Acme Size
3201T61 1 Each Black-oxide Steel U-bolt, W/plate,5/16″-18 X 1″l Thrd,for 2-1/2″od,610# Wll
92105A660 1 Pack Zinc-alloy Knife-thread Insert For Wood, Hex Drive, W/out Flange, 8-32 Int Thrd, 25/64″ L
91770A197 1 Pack 18-8 Ss Truss Head Phillips Machine Screw, 8-32 Thread, 3/4″ Length
6061K23 2 Each Hardened Precision Steel Shaft, 1/2″ Diameter, 9″ Length

Now all that is needed are the control electronics.

In How Not to make a CNC Mill (part 2), we’ll take a look at what went wrong with the first build.

Posted in CNC Tagged with: , ,
  • Jeremy

    I see you’ve got your Dremel router in what looks like a sound enclosure?  I used a Dremel with mine, a Zen Toolworks 7×12 –  initially, but it was super-loud!  The Zen spindle isn’t cheap, but it’s much quieter.

    I keep thinking about building a bigger router, but mine does the job so far!

    • bamos

      The dremel is definitely noisy, which was the original reason for the sound enclosure.  The enclosure also doubles as a way to keep all of the dust contained (this sits in my office).  The Zen CNC’s looked decent – how do you like yours?

      • Jeremy

        For the price, I really like mine.  I really like the new Zen spindle after trying it out (after using the Dremel 200).

        Assembly isn’t too hard, but the real learning curve is setting up the electronics and learning to use the software.  I haven’t tried it, but the Shapeoko looks really neat too, especially since it can supposedly be expanded.The support has been fantastic.  It’s a pretty small company, so I could email the owner (Xin) and get a response.  Their forum can be pretty helpful too, and people seem more than willing to help with whatever you’re working on ZTW or not – if you want to check it out.

  • Larry

    Funny that you talk about the Mantis using parts not easily available or cheap and then you proceed to order parts from McMaster Carr, who doesn’t ship outside of the USA.

    • bamos

      I didn’t even realize McMaster only sold within USA – probably makes sense due to shipping costs of the parts though. . .

      The parts themselves (for me) were easily available, but required modification that was done on equipment I didn’t have. My main point was that the $100 BOM cost for mechanical components couldn’t be realized if you were only building one mill – there were many hidden costs.

      I do agree that easy availability is very relative – heck, I’ve needed to jump through hoops to get a silly M2.5 screw (which I imagine is available at local hardware stores anywhere other that the USA. Even McMaster’s selection of simple metric screws is pretty pitiful (and quite expensive). It shouldn’t be too hard to cross source the components based on their descriptions alone to something that’s commonly available in your area. If you’d like to talk about anything in depth, feel free to start a thread in the forum or just reply here. I’d be happy to help in any way I can!

  • SCSInet

    I enjoyed this writeup… not often do people document projects that went wrong… good to see someone sharing the learning that they got along the way rather than just sharing the successes.
    Two things. First, your choice of a dremel might also cause you some problems. When using engraving bits, runout is a huge concern, you usually want it very low… a few mils at the very most. Dremels just aren’t built for that. I had tremendous problems with my PCB routing when using engraving bits, I was chewing through bits like crazy, and after talking with a couple of engraving bit vendors, I checked the runout of my spindle (A Bosch router) and found my runout to be almost 1/100th of an inch. I switched to a “real” CNC spindle and now I’m not wasting tooling anymore. Compared to the cost of trashed boards, wasted time, hair pulling, and replacing tools, it becomes a wise investment. I am using a water-cooled 3hp spindle (I use my mill for other stuff), but for a CNC mill, the Wolfgang spindles or similar types might be a thing to look at.

    Also, your comments about the presser foot…. I’m not using one, I use thicker PCB substrate and my boards do not need very fine pitch, so I can cut a little deeper to make up for slight variances and flex (also with a good spindle, you can use 45 or 30 degree tools and not trash them to mitigate the isolation width). However, with the presser foot, you could secure your board with double sided tape, and keep some extra stock laying around to set up around the sides of your workpiece to offer a wider margin around it to allow your presser foot to work without constantly having to use a much larger than necessary workpiece.

    Just some food for thought.

    • bamos

      Thanks for the comment – and advice!

      ==>” (A Bosch router) and found my runout to be almost 1/100th of an inch.”
      Wow, I see those Bosch routers used everywhere on CNC builds – I assumed they’d be much better than that. . .I’ve also seen the Wolfgang spindles before – they look great, although the spindle alone is somewhere around 2-3x the current cost of materials :(

      I went with 60 degree bits since they seemed to be the most durable, good to hear you’ve had decent results with the steeper angles (although with a much better spindle). What isolation have you done with your setup?