An open hardware project has documented the design of a 3D-printed cycloidal gearbox alongside the Python script used to generate its central geometry. The repository presents three iterations, including a hand-cranked test model and a later version designed to operate with a NEMA 17 motor.
Cycloidal gearboxes convert input rotational speed into increased torque. In this project, the first prototype was hand cranked and used a 1:9 reduction ratio. Its purpose was to test whether the author’s Python-based cycloidal profile generator produced a viable shape rather than to serve as a finished motorized assembly.
The second design aimed to fit within the footprint of a NEMA 17 while preserving the same 1:9 ratio. That compact version did not work. The repository attributes the failure to the tight tolerances required by a small cycloidal drive and the limited precision available from the chosen 3D-printing process. The result is included as part of the engineering record rather than omitted from the project history.
A third version increased the footprint, allowing more generous tolerances. The author describes it as the first functioning version driven by a NEMA 17. Its CAD files are organized under a dedicated Version 3 directory, making the design available for inspection or further modification.
The generator is based on equations described in a SolidWorks article about building cycloidal drives. In the repository’s formulation, the rotor’s reduction ratio is one to the number of pins minus one, and the rotor turns in the direction opposite the input shaft. Publishing the script with the mechanical files makes the profile parametric: dimensions can be generated from inputs rather than being available only as a fixed mesh or drawing.
The project also identifies possible substitutions for builders who want to reduce friction or increase rigidity. Housing pins may be replaced with MR128 bearings, which the author says can improve efficiency. The output pins can instead use M2 screws with metal coverings, intended to strengthen the mechanism and improve its torque handling and efficiency. The documented print setup uses PLA on a Bambu Lab P1S, with support PLA specified at interfering support surfaces.
The repository does not provide independent performance measurements, a rated maximum torque or long-duration reliability results. Its strongest contribution is therefore the transparent iteration trail: a validation model, a miniaturization attempt that exposed manufacturing limits, and a larger working design. By pairing those outcomes with CAD assets and the generating equations, it offers other builders a reproducible starting point while clearly recording the unsuccessful step that informed the final geometry.
