The Rinnegan drive, aka HDP30: a strain-wave (harmonic) gearbox that bolts onto a standard NEMA 17 stepper, prints on a bog-standard bed slinger, and costs about $4 of hardware per unit. Half again the reduction of our HD20 in a thinner package, using the same screws, the same bearing and the same motor face.
Its standout spec is backlash: about 0.03° at the output (roughly 2 arcmin), measured on a load-cell rig at 5 RPM. That's close to what commercial strain-wave gearboxes quote, from printed PLA. See Test data.
| Reduction | 30:1 |
| Backlash | About 0.03° at the output (2 arcmin); no run of 12 slow runs showed more than about 0.1° |
| Tooth profile | Cycloidal, module 0.55, 60-tooth flexspline / 62-tooth circular spline |
| Working limit | Keep peaks under about 1.3 N·m. The flexspline wall starts cracking at 1.5-1.67 N·m |
| Measured torque | Held 1.49 N·m with no tooth skips. See Test data |
| Endurance | 5,744 flex cycles and 188 hard stops at 1.0-1.2 N·m with no damage |
| Efficiency | 47-60% static (stalled against a load cell), across two units |
| Torsional stiffness | 40 N·m/rad at 0.85 N·m, rising to 61 N·m/rad at 1.5 N·m |
| Motor | Any NEMA 17 with a 5 mm shaft (same interface plate as the HD20) |
| Lubrication | Super Lube synthetic PTFE grease |
| Printed parts | 11 pieces across 9 unique parts, including 3 shear pins, plus a printed alignment tool |
| Materials | PLA throughout, including the flexspline |
| Fasteners | 8 × M3×0.5 × 6 mm button head screws per drive, motor screws included |
| Hardware cost | ~$3.96 per actuator |
| Version | v2.0 (RevC, 2026-10-18) |
| License | CC BY-SA 4.0 |
⚠️ Keep peaks under about 1.3 N·m. In a 41-minute endurance run, one unit took 5,744 flex cycles and 188 hard stops at 1.0-1.2 N·m with no damage and no tooth skips. Then a single 1.67 N·m impact started a crack in the flexspline wall, and it broke twelve hits later. A second unit started cracking at about 1.5 N·m. Impacts count, so avoid slamming hard stops, and on a NEMA 17 like ours limit the motor current to about 0.6 A.
⚠️ Longer life testing is still to come. 5,744 cycles is about 40 minutes of running. If you run one for a long time, please open an issue and tell us how it held up. We'll credit you.
Three parts do the work. A wave generator, an ellipse riding on 11 loose 5 mm balls in a printed cage, pushes a flexible toothed ring (the flexspline) into an oval, so its teeth engage a rigid ring (the circular spline) at just two points. The circular spline has two more teeth than the flexspline, so one turn of the motor walks the output around by two teeth. With a 60-tooth flexspline, that tiny slip per revolution is the 30:1.
New to strain-wave gearing? We wrote a plain-language explainer: robrotics.web.app/learn.
v1.0 held 1.47 N·m and then started ratcheting: the flexspline teeth rode over the circular spline one tooth at a time, and every round of that left the drive weaker. RevC goes after that with two changes:
- The flexspline is now PLA instead of PETG. Same geometry, stiffer material. It also creeps about half as much under a steady load.
- The flexspline lines up better with the circular splines. A new printed alignment tool sets the wave generator at the right height on the motor shaft, so the flexspline teeth sit square in both circular splines. Both circular splines were also re-cut to take a third shear pin.
The result on the load cell: no tooth skips up to 1.49 N·m, about 40% less twist at the same torque, and the same efficiency. The tooth profile, the wave generator, the cage, the interface and both preloaders are unchanged from v1.0, and so is every piece of hardware.
We changed two things at once, so we can't fully split the credit. The way the stiffness rises with load points at the alignment, and the lower creep points at the PLA.
The HD20 got 20:1 out of a 40-tooth flexspline at module 0.8. The HDP30 keeps the same two-tooth difference but packs in 60 teeth at module 0.55, which buys half again the reduction at roughly the same pitch diameter, and lets the whole stack get thinner, hence pancake.
The interface plate, the base preloader and every piece of hardware are the same as the HD20's: same 11 balls, same 30×42×7 bearing, same 8 screws, same 12 inserts. If you have already built an HD20, your leftovers cover this one.
The HD20 is still the stronger drive. It was tested to 3.3 N·m.
Not included in the BOM, bring your own:
- A NEMA 17 stepper with a 5 mm shaft. The interface plate is unchanged from the HD20, which was validated with an OMC StepperOnline 17HE12-1204S (42 × 42 × 30 mm, 26 N·cm, 1.2 A, 4-wire) under closed-loop FOC control.
- A soldering iron for the heat-set inserts, and hex keys.
Filament: PLA for everything, including the flexspline.
Building v1.0? Its flexspline is still PETG. The PLA advice only applies to the RevC geometry and alignment. v1.0's files and instructions are on the
v1.0tag.
Machine-readable source: bom.json. Unchanged from v1.0.
| # | Item | Qty | Pack | Pack cost | Per actuator |
|---|---|---|---|---|---|
| 1 | M3 heat-set inserts | 12 | 100 | $9.99 | $1.20 |
| 2 | M3×0.5 × 6 mm button head, stainless | 8 | 100 | $7.69 | $0.62 |
| 3 | 30 × 42 × 7 mm bearing | 1 | 10 | $16.39 | $1.64 |
| 4 | Set screw | 1 | 50 | $5.69 | $0.11 |
| 5 | 5 mm steel bearing balls | 11 | 200 | $7.20 | $0.40 |
| 6 | Super Lube synthetic grease | 1 | - | - | - |
| $46.96 buy-in | $3.96 each |
One tube of grease lasts many builds, so it isn't counted in the per-unit cost.
Parts are sold in packs, so the first actuator costs about $47 in hardware and every one after that costs about $4, and you'll have enough left over for eight more.
Affiliate disclosure: the purchase links in
bom.jsonand on our website are Amazon Associates links. If you buy through one, Robrotics earns a small commission at no extra cost to you. Every part listed is what we actually used; the links don't change the recommendation.
| Part | Qty | Material | File |
|---|---|---|---|
| Circular spline (base) | 1 | PLA | hdp30-circular-spline-base |
| Circular spline (output) | 1 | PLA | hdp30-circular-spline-output |
| Base preloader | 1 | PLA | hdp30-base-preloader |
| Output preloader | 1 | PLA | hdp30-output-preloader |
| Interface / motor mount | 1 | PLA | hdp30-interface |
| Wave generator | 1 | PLA | hdp30-wave-generator |
| Ball cage | 1 | PLA | hdp30-cage |
| Flexspline | 1 | PLA | hdp30-flexspline |
| Shear pin | 3 | PLA | hdp30-shear-pin |
| Alignment tool (assembly jig) | 1 | PLA | hdp30-alignment-tool |
All three shear pins are required, and they're the same part printed three times, so there's only one file. They carry shear load across the output joint directly, so the connection doesn't have to rely on friction from the preloaded screws to resist it. The screws clamp; the pins take the sideways load.
The HD20 was printed on a Bambu Lab P1S with Bambu PLA, and the HDP30 uses the same settings:
- Arachne variable-width wall generator
- One extra wall loop for strength
- Scarf seam turned on, so the seam blends in instead of leaving a ridge on the round surfaces
- Tree supports on the output circular spline and the interface. Everything else prints unsupported.
No custom temperatures, no special bed prep.
cad/step/ STEP, for slicing and for CAD work
cad/dxf/ The four tooth profiles, straight out of the harmonic maker
photos/ Build photos
manual/ LEGO-style build manual (PDF), rendered from the STEP files
bom.json Bill of materials (source of truth)
Grab the packaged bundles from the Releases page rather than cloning, because the STEP files are large. The v1.0 (RevB) bundles are still there too.
There are no pre-sliced files in the repo. Import the STEP into your slicer and use the print settings above, so the geometry you print is always the current one.
New in RevC: the alignment tool. Before you tighten the set screw on the wave generator, sit the alignment tool on top of the interface plate and slide the wave generator down the motor shaft onto it. That puts the wave generator at the right height, so the flexspline lines up with both circular splines. Tighten the set screw, then take the tool out.
📘 LEGO-style build manual (PDF): 10 illustrated steps, every part rendered straight from the RevC CAD, with a parts list and a callout for what goes in at each step.
📹 Quick assembly run-through (v1.0): instagram.com/reel/DclcvVig55a
That one is fast, and it predates the alignment tool. A full, properly paced build video is in the works and will go up on youtube.com/@robrotics and be linked here.
Two ways in, depending on how deep you want to go.
Change the tooth profile. Every profile in this drive came out of our own free generator, the harmonic maker. This link opens it with the exact settings used for the HDP30, so you can nudge one number and export a new DXF:
Open the HDP30 profile in the harmonic maker →
The four DXFs it produces are checked into cad/dxf/ if you'd
rather start from ours. They didn't change in RevC.
Change the mechanics. The full parametric model is public on Onshape:
Copy it to your own workspace and change what you like: a different motor face, a different output interface, a different ratio. If you build a variant, we'd genuinely like to see it: open an issue or tag @robrotics.
Measured on this drive:
- Cracking starts at 1.5-1.67 N·m. Past that, the PLA flexspline wall cracks. Once a crack starts, the drive softens and gives way a little more on every hit until the wall breaks. We never saw RevC ratchet, so don't count on v1.0's buzzing as a warning. Limit the motor current.
- More current stops buying torque past about 0.8 A. Held torque levels off near 1.35 N·m, but impacts keep getting harder. About 0.6 A is the sweet spot on a NEMA 17 like ours.
- It holds its position unpowered. With the motor off, it kept about 1 N·m on the load cell. It doesn't backdrive at that load.
- It softens a little when warm. After 20 minutes of running it was warm to the touch and about 12% softer.
- It twists, but it doesn't rattle. Expect about 1° of elastic twist at the output at 0.85 N·m, and about 1.6° at 1.5 N·m. That's give, not slack: the actual backlash is only about 0.03° (2 arcmin).
- Torque relaxes a little under a steady hold. With the motor held still, output torque dropped about 5% over two seconds and then levelled off.
Carried over from the HD20:
- PLA creeps under sustained load. Holding a heavy static load for hours will slowly deform the splines, and that now includes the flexspline.
- Unit-to-unit variation is real. Two HD20s of the same design differed by up to 18% at the same current. Printed gearboxes are not precision parts.
- Output spline preload is the biggest tuning knob. On the HD20, tightening it further took one unit from 2.37 to 2.55 N·m. Expect it to matter here too.
Not yet known for this drive: life beyond a few thousand cycles, running efficiency under load, and long-term wear below the cracking limit.
Found something we haven't listed? Please open an issue and include your version, filament, and printer.
Tested on 18 October 2026 with a 100 mm arm between two load cells, against a no-gearbox baseline of the same motor. A second unit then ran 41 minutes of endurance testing at rising current until the flexspline cracked. The full report, with plots, the method, the caveats and the raw CSVs, is at robrotics.web.app/actuators/hdp30-revc-torque-test.
| v2.0 (RevC) | v1.0 (RevB) | |
|---|---|---|
| Backlash | about 0.03° (2 arcmin) | under 0.3° (not resolved) |
| Torque | Held 1.49 N·m, no skips; cracks from 1.5-1.67 | Held 1.47 N·m, then ratcheted |
| Endurance | 5,744 flex cycles at 1.0-1.2 N·m, no damage | Not tested |
| Efficiency (static) | 47-60% | 43-55% |
| Stiffness | 40-61 N·m/rad | 31-44 N·m/rad |
| Creep in a 2 s hold | about 5% | about 10% |
The v1.0 report is archived at robrotics.web.app/archive/hdp30-revb-torque-test.
We're working on a cup-style harmonic drive, the classic layout where the flexspline is a thin-walled cup instead of a ring. Follow @robrotics or watch this org on GitHub to hear when it lands.
CC BY-SA 4.0. Use it, change it, sell it. Just credit Robrotics and share your derivatives under the same license.
No warranty. Printed parts fail. Don't put this anywhere a failure could hurt someone.