Summary

Description

Split Ring gearboxes are known for being able to get very large gear
reduction with very few stages. The details for the design captured
here were mostly provided by David Hartkop. See:
https://github.com/IdeaPropulsionSystems/CoolEpicyclicGearing
https://www.youtube.com/watch?v=-VtbSvVxaFA
I really liked this design because it kept the planet teeth
aligned with one another making them easier to 3D print as well
as allowing for an output sun gear to help prevent the planet
gears from twisting under load. While I can get very high
reduction very easily with this design, I tended to need much
less reduction in practice but instead used it as the final output
stage for various arms which need to support more torque. The
secret to that was to print them big and use lots of planets.
Maintaining the input/output modulus greater than 1 is likely to be
helpful as well.

I wanted to build this gearbox for a robotics application but I
needed bearings to support the loads involved. I discovered this
excellent design for 3D printed bearings by Positive Attitude
that I thought would be perfect for this application. See:
https://www.youtube.com/watch?v=NIIVZcgJWPs&t=46s
By merging these bearings with David's original gearbox, I came
up with the following design. I ended up changing the output
stage to have the mounting holes outside the ring teeth. This
works better for carrying high loads and mounting bearings but
it does reduce the input/output ring diameters which I'm not
sure was the best trade-off.

Since I was using this for a robotics application, I wanted to
include in the design and absolute encoder on the final output.
I originally expected to use a 10-turn pot but I never quite
liked how it fit in the design. Instead, I decided to go with
a rotary encoder module (CJMCU-103, available on Amazon/AliExpress
for ~ $1-2 per unit). This module only encodes 1-turn so I needed
to come up with a high reduction ratio gearbox for the encoder.
The gearbox I came up with can support an output sensor ring
with up to 270-teeth. It's a traditional multi-stage and has
some backlash that I expect is difficult to remove. The gears
are really small and I even 3D printed the "shaft nubs".
Seems to work OK but I'm not sure about durability. If I run
into durability issues, I'll replace the printed "shaft nubs"
with holes and put solid 2-3mm metal shafts/wires in instead.

I expected that many applications would want to have 2 identical
gearboxes driving some arm on either side of the arm but driven
by the same shaft/motor. This requires that the gearboxes
be assembled identically. This can be challenging so I added
some index marks to aid in this effort but I'll admit it is
likely a bit challenging to assemble them identically.

Configurablity and Printing

The OpenSCAD file provided supports the configurator and
there are messages that show the computed values and asserts
to check that the settings you picked work. Most of these came
from David's spreadsheet (thanks for that) but some I added.

The Assembly view shows all the parts together as they would
be in the final Assembly but I did cut out a section view so
you can see inside a bit and added the ability to Explode the
assembly to see the guts even better. To print your design, you
would need to dump the STLs of all the individual parts. I
usually dumped the STL file from OpenSCAD from the exploded
diagram (without the CrossSection :) ). Then I just split the
parts across a number of build plates in my slicer. I liked the
look of printing the stationary parts (Base and OuterRace) in
grey and the moving parts (all the gears, cages, and OutputRing)
in white.

Slicer Configurations

I printed mine in PETG with a 0.4mm nozzle diameter and 0.2mm layer
height. The torque of the gearbox is carried by the planet gears
and the 2 ring gears all near the outer diameter of the gearbox.
To make these strong, I printed them with lots of wall thicknesses
(5-10 which might be overkill). I used 15%, Gyroid infill for what
remained but I'll admit there wasn't much infill needed with all
those wall layers.

Assembly

To assemble the default configuration, you will need all the printed
the parts plus:

• 4, #6 flat-head screws, nuts, washers, lock-washers at least 3/4" long
• 8, #10 screws, nuts, washers, lock-washers at least 1.5" long
• 6, 1/4x20 hex-head bolts, nuts, washers, lock-washers at least 1.0" long
• 57, 4mm ball bearings
• 1 CJMCU-103 variable resistor encoder module

Assemble the sun gear bearings by placing the appropriate number
of ball bearings in the outer race (all on 1 side) and then inserting
the sun bearing in the appropriate orientation. Hold the sun gear
against the ball bearings as you begin distributing the bearings
around the race. Once you get the bearings roughly distributed,
the inner and outer races will hold the bearing together as you
work to evenly distribute the bearings. Once the bearings are
close to evenly distributed, use the associated cage under the
bearing as a guide to get the balls distributed evenly. Once
every cage position has a ball ready to go in, press down on
the bearing to snap the cage into place. Follow the same
procedure with the output bearing.

After all the bearings are assembled, insert the 6 1/4-20 hex head
bolts through the cavities in the output ring. You may need to put
the nuts on loosely to keep them from falling out as you complete
the assembly.

Line up the index mark on the input sun gear with the index mark
on the base. Then insert the planet gears using the alignment
marks on both the input sun gear and the base. Once all the
planets are inserted and aligned so that they are evenly spaced,
put the output bearing assembly on top of it ... spinning the output
sun gear and output ring gear as needed to engage with the planet
gears. Secure the output assembly to the base using the #10 bolts.

Now that the gearbox is assembled, you can assemble the
encoder gearbox. Prior to assembling the encoder gearbox, you will
want to solder either wires or the provided jumper pins onto the
CJMCU-103 board with the wires or longer side of the jumper pins
coming out the bottom of the board. Slide the CJMCU-103 into the slot
with the pins/wires able to come out the access hole on the underside
of the mount. Next, insert the encoder gear with the longer shaft/key
so that it engages with the variable resister module on the CJMCU-103.
Next, place the 10:20 tooth gear so that the 10-tooth side engages
with the encoder gear and axel nub is in the hole on the mount.
Next, place the 20:30 tooth gear so that the 20-tooth side engages
with the 20-tooth side from the previous step and the axel nub is in
the hole on the encoder gear. Next, place the 10:30 tooth gear so
that the 10-tooth side engages with the 30-tooth side from the
previous step and the axel nub is in the hole on the 20:30 gear.
Last, place the encoder cover over the assembly making sure
the axel nubs on the cover align with the holes on the top two
gears and assemble with #6 screws/nuts. You may have to
take the cover off and align the encoder so that the useful
range is aligned with the useful range for your gearbox.

Load Testing

I built the 33:1 configuration and was able to load it with 84ft-lbs (114Nm)
before the hex spline of the input sun gear stripped out. With
a different power delivery or a higher gear ratio, I expect the
gearbox itself could carry more torque. The gearbox did back-drive
with around 14ft-lbs (19Nm) of torque applied. The bearings did a
great job at handling the radial loads and even a reasonable job
handling the axial loads but I was able to pull the output ring out
of its bearing with a caparatively smaller moment load (though the
planet gears were not in place when I was testing the bearings).

I built the 56:1 configuration as well and tested it to failure.
This configuration did not seem to back-drive. It failed at 70ft-lbs (94.9Nm)
with a couple of the teeth in the output ring failing (and a couple
of the planet gears were slightly deformed as well). With 6 planet
gears, each tooth was able to carry 94.9/6=15.82Nm. The radius to
the tooth pitch was roughly 48.7mm, which means that an 8mm tall tooth,
printed in PETG can hold roughly 320N before it will fail. I'll
add an output message showing the rough torque estimate for a
given configuration using this data.

I also played around with the configurator and uploaded a couple
different configurations of the gearbox to show the range of options:

• Default: 33:1
• 91:1
• 165:1
• 246:1

History

1.0 Initial Revision
1.1 Fixed partial encoder hole
1.2 Changed sun gear bearings to allow for them to be disassembled
1.3 Added counter-sunk holes for encoder cover
1.4 Added assembly directions
1.5 Took better control of $fn to improve quality and reduce render time
1.6 Added markings for encoder and results from testing
1.7 Enlarged EncoderGear axel to be stronger/print better
1.8 Added torque estimate as output