Tips for modeling worm gears that I wish I'd known sooner

Oct 10, 2026 Last reply: 1 day ago 4 Replies
Jump to replies (4)

Been building worm drives in CAD for a few years now, mostly for small robotics projects, and there are a couple of things that tripped me up early on that might save someone else the headache. First, lock down your center distance before you draw anything. I used to model the worm and wheel separately and then try to nudge them into mesh, which never quite worked. Pick your center distance first, then build both parts around it. And leave a bit of backlash in there, more than feels right. Printed gears always come out tighter than you expect and a mesh that looks perfect on screen will bind on the bench.



Second thing is testing the mesh before you commit to a print. Rotate the worm in the assembly thru a full turn and watch how the wheel teeth track. If anything catches or the contact looks lopsided, fix it in CAD, not with a file later. Also worth checking the wheel tooth profile against the worm at a few points, not just one cross section.



If you're using a script or custom feature to generate the geometry, resist the urge to accept the defaults blindly. Change one parameter at a time and remesh each time so you know exactly what broke if it stops fitting.


I went through the same pain on a chicken coop door opener I built last winter. My first wheel bound up so badly the little motor stalled, and it took me a while to realize the teeth were actually fine, the housing had crept in because I printed the mount in PLA and it warped a touch. Reprinted the carrier in PETG and the same gears ran smooth. So these days I check the mount almost as much as the gears themselves. One thing that helped me was printing a throwaway pair in PLA before committing to the final material. It costs me maybe two hours of printer time and it catches most fit problems. I also bump my tooth clearance way up compared to what the generator suggests, something like double the default, because my Ender always squidges holes slightly small.

On the testing side, I do what you describe but I turn the worm by hand with the wheel held loosely in a jig, no motor attached. If there is any spot that grabs, I can feel it in my fingers before I ever see it on screen. The middle of rotation is where mine always went wrong, not the ends.

One thing that bit me was greasing the mesh before the hand test. Dry PLA on PLA feels smooth enough untill it warms up, then it gets gummy and you get a false bind that isn't really a geometry problem. A dab of silicone grease on the teeth tells you pretty quick if a catch is real or just friction. Also, does anyone else's worm wheel come out slightly conical? Mine always has a hair of taper from the bed not being level, wich throws the contact pattern off center. Leveling the bed better than I thougt I needed fixed half my mesh issues, honestly.

The custom feature route saved me on a feed mechanism for a camera slider I was messing with. I kept rebuilding the wheel by hand every time I changed the tooth count, which was dumb, so I finally made a spreadsheet that spits out the sketch dimensions and I paste them straight into the profile. Now when somthing doesn't mesh I can regenerate both parts in a minute and try again the same evening instead of putting the project down for a week. My advice on top of yours is to log every parameter change somewhere, even just a note on your phone. When I got a good mesh I couldn't remember what clearance value I'd used, had to reverse engineer my own part from a caliper measurement. Wrote it down after that.

Not sold on just doubling the clearance numbers across the board. On a small module worm wheel the teeth are thin enough allready, and if you open them up that much you end up with a sloppy mesh that backlash-hops under load, which is its own headache. My wheel skipped a tooth doing exactly that on a curtain motor once. Better plan is to print one test wheel at stock clearance, measure where it actually binds, and only add clearance where the plastic squished, not everywhere. Also, on the hand test, try pausing at each tooth instead of turning steadily, slow spots hide if you keep moving.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required