top of page
  • Writer's pictureTretron

Kit robot part 3; shifting gears.

Updated: Apr 4, 2022

Well this project is due a long awaited update. Since the last update I have been struggling with designing the movement of the robot with the motors and bearings at hand and I need a rethink. I can't find the reductions I want and things have been ground to a hold so where does that leave me?

Well it leaves me with a robot with a finished end effector that I am not really happy with. Besides, I do not know what kind of weight budget I have to design the end effector with, so I want to try my hand on something new;

Using differential drives to power the Second and Third axis. (at the base, going up and down and rotating the arm around the base arm, the First axis is the base rotation) differential drives allow us to put the heavy engines on the base, where they have no influence in the weight. I am thinking to use differential drives further down the robot to allow the motors to be placed closer to the base, thus reducing the weight induced torque.

Further more, the engines themselves, I have done some digging; the current way I designed it is not compliant something that is a good feature to have when dealing with cobots (collaborative robots) furthermore with if we design on back drive capability we can program the robot to allow for operator movement but where it compensates its own weight. So lets get cracking; for back drive we need the following;

  • low gear reduction, between 1:5 and 1:20 (meaning high engine torque and holding power required)

  • gear reductions that are unidirectional (this excludes worm gears and strain wave gearing)

  • and we want it to be easily manufactured with cheap methods. Ideally 3D printing but i want to allow multi-manufacturing methods.

there are two designs that currently i know of that allow for high torque, efficiency and compactness; Planetary gear reduction and cycloidal gear reduction, so lets weight the pros and cons;

planetary gear reduction is often used and well understood within the hobby maker space. It can be made compact and light with only a few bearings. But I fear the wear and tear of this reduction. Is the strain of the tooth profile on the gears. For more accurate transference of the movement a finer tooth profile needs to be selected but this puts the whole strain induced by the robot on small tooth that are potentially 3D printed. Furthermore the fine tooth might be an issue with the technique, causing the printer to be put on a fine printing mode causing long manufacturing times for production. Laser-cutting is an option but that would limited the options of others to reproduce this work.

Cycliodal drive (by Petteri Aimonen, public domain)

The other option is Cyclical drives allow for an high reduction using an eccentric cam. Shown on the image on the left, the green input shaft forced and rotor against a mating surface. This can be rollers or half round shapes, what ever is desired by the designer (me in this case) the outer ring has one more 'tooth' then the inner one. The reduction is calculated as:

n(outer ring) - n(rotor)

So if you want to have a 10:1 reduction you need 10 'teeth' in the outer ring and 9 in the rotor. What is interesting in this design is that the pins/mating surfaces can be what ever we desire. We can use ball bearings for the design and allow the whole design to be 3D printed. Using ball bearings greatly induces the cost but reduces the wear. furthermore we can if we want make the rotor from any material and technique we want as that is the part that has the most strain. One side effect is that the eccentric motion induced vibration at high speeds. this might not be a problem for the movement of the axis but just in case I am going to compensate the weight of the rotor with a rotor 180 degrees out of phase.

but first; what is cycliodal?

Cycloid animation by Zorgit (CC BY-SA 3.0)

well a cycliod is a curve traced by a point by a circle that roles along a straight line, as shown on the right animation. This creates a shape that is ideal for a rolling motion of the rotor. However we do not want to use a pure cycloid. We want to have it offset by the rollers. Thus the shape actually used by the Cycliodal drive is an Epitrochoid. Gotta love those greek names.

Epitrochoid animation by Sam Derbyshire (CC BY-SA 3.0)

We trace a line along a circle with an off-set in the line. This shape might be familiar to those who ever heard of a Wankel engine, as it too uses an Epitrochoid shape for it's combustion chamber. Due to this a Wankel engine reach high rpm for prolonged periods of times but features an a-symmetric combustion chamber. (the combustion will only be introduced on one side which induced extreme vibrations if not properly balanced.

So lets decide our parameters for our drive. This will not be a solidworks tutorial how to build a drive but we will discussing the design parameters and what you want to think off about designing a drive and me showing off my design. Lets start off with a video from James Bruton who inspired the drive.

He iterates up on this design until he has a V3 but his designs, while opensource, are not that useful for me. I will be using some of this ideas and parts for my own drive but i iterate on it myself. First off, I am not a fan of the screwing in plastic. I prefer to have an drive that more maintainable, especially when my target for the end user to be students. Trough seeing how James uses his V3 design in robot dog V3, his design would be good for most project. I just prefer to have things build a bit more beefier (or over engineered depending on your point of view.)

So my design, it features the same motor and bearings but has some differences in size and mounting methods. I will upload a tutorial in designing a cycliodal drive in the coming week, I have also ordered the parts to make the drive.

So with all that, what are the design chances i made? Lets digitally gut the drive to find out.

from the side you see the following colours of parts;

Orange; output drive

Red; rotors

purple; input drive.

I made double threaded aluminum studs that I can easily make to lose weight, however a shouldered M4 bolt works also when you can source that part more easily. I use expansion (or thermoplastic) inserts to create a more secure connection. To increase the strength, where needed I brace the end of the inserts against plastic so it can't be easily pulled out. All fasteners are selected to be hex or torx drive as those fasteners are more easily manipulated. (Especially for those who suffer from RSI or carpal's tunnel)

Left is the exploded view of the part to show how many parts actually go into this whole design. And this marks an important moment in the design phase. If I complain in the future about the price of the robot, that price can be in part pinned down to this design and this design choice. The principle of this is explained in the MacLeamy curve

The MacLeamy curve

What this means if that during the design phase you have little actual cost and cost of design chances but can impact the most of the eventual cost. To relativist to this project, this designed drive and actuator will have the following cost:

  • € 209 ex VAT for the Odrive controller (controlls 2 actuators)

  • € 75 for the motor.

  • € 150 bearings and fasteners

  • € 20 euros for the encoder.

this comes out to around € 350 per actuator or a whopping € 2100 for all actuators if I use the same motors and transmission. Which I will not, but it is important to scale the cost at this moment. For now I will continue with the design, first off I will make a single drive, I made it modular so I can fit it where ever and make the second and third axis of the robot to test out. I will spread out the process in several weeks to spread the cost of designing this all.

The cads for this project can be found here:

21 views0 comments

Recent Posts

See All


bottom of page