A 3D-printable 20:1 strain wave (harmonic) gearbox that bolts onto a standard NEMA 17 stepper. Prints on a normal printer, assembles with about $4 of hardware, and has been tested to 3.3 N·m.
📖 Full build guide: robrotics.web.app/actuators/hd17
Specs
| Reduction | 20:1 |
| Tooth profile | Cycloidal, module 0.8 |
| Measured torque | 2.1–2.6 N·m @ 0.7 A · 3.0–3.3 N·m @ 1.4 A (peak 3.5 N·m) |
| Efficiency | 48–58% |
| Backlash | measurement in progress |
| Motor | Any NEMA 17 with a 5 mm shaft (tested: OMC 17HE12-1204S, 26 N·cm, 1.2 A) |
| Printed parts | 9, including the shear pin |
| Hardware cost |
Which download do I want?
| Bundle | Use it for |
|---|---|
hd17-v1.0-step.zip |
Start here. Import into your slicer to print, or into CAD to remix. |
hd17-v1.0-mjf-nylon-set.zip |
Ordering the nylon flexspline from a service bureau. |
hd17-v1.0-everything.zip |
The lot, including photos and the test report. |
Note: the 3MF bundle has been withdrawn. Those exports no longer matched the
current geometry, so they were removed rather than left to mislead builders.
Slice from the STEP files, using the print settings in the
README.
Prefer to change the design itself? The parametric model is public on Onshape — copy it to your workspace and go.
Read this before you build
- Print the flexspline in PETG, not PLA. It flexes on every revolution; PLA has almost no fatigue life in that duty and will crack.
- The MJF nylon option is a matched set of three — flexspline, cage and wave generator. Their geometry differs from the printed versions. Mixing them will not fit correctly.
- Torque drops ~10% after running in. Expected. Plan around the run-in figure.
- The 1.4 A numbers are above the test motor's 1.2 A rating — treat them as a short-burst ceiling, not a continuous rating.
- Unit-to-unit variation is real — two units of the same design differed by up to 18%.
Full list in Known behaviour.
Test data
The torque figures come from load-cell runs, not motor-spec arithmetic: seven configurations, 1,500–3,000 samples each, 100 mm lever arm, with a no-gearbox baseline for the efficiency maths. The full report is in docs/torque-test-report.html and readable online.
Testing is still being revised, so expect these numbers to move.
Known gaps in this release
Being upfront: an assembly video and written steps are still to come, backlash is unmeasured, and a few fastener sizes aren't documented yet (noted inline in bom.json). Issues welcome — tell us your version, filament and printer.
Licensed CC BY-SA 4.0 — build it, change it, sell it. Just credit Robrotics and share derivatives alike. No warranty; printed parts fail, so don't put this anywhere a failure could hurt someone.