Gearbox for mechanism or appendage

Good questions.

The backlash is somewhere around 1-3 degrees.

As for the efficiency, I really don’t have the right equipment to measure that right now. It is abs plastic and it’s not melting, if that’s saying anything.

Attached is a picture showing whats going on inside. The gearbox ended up being pretty simple, kind of ironic considering the path it took to get to this point.


gearbox inside.jpg

gearbox inside.jpg

Do you have any maximum torque figures?

How can we buy one?

2nd that.

Try Harbor Freight. Link to old thread. Other link.

This gearbox is known in the industrial robotics industry as an epicyclic gearbox and is very common in robots that require low backlash and high gear reduction.

A couple of business and technical points:

  1. There are MANY patents on this design. Several of them have expired but a few key ones still exist so do a patent search before selling this.

  2. This design looks to only have one stage going in a single direction. This will induce severe oscillations that cause torque ripple which are detrimental to high load arm applications. There are ways to solve this but it adds significant cost.

  3. In order to get acceptable efficiency the oscillating gear must have a very smooth surface finish. Grinding that surface is a must or I would not use it. Plastic gearing for this will severely limit the load carrying capability.

That’s all I can think of now.


Michael, I torqued the gearbox unpowered with some lever arms until it broke apart. I would estimate that it reached around 50 ft-lbs.

Adam, I built a website, I’m a mechanical engineer so no making fun of it looking like its from 2004, I need to go through the rules on mentioning trademarked names or just delete them, after I do that I’ll post the website address and they can be purchased through the Paypal payment system with a credit card.

Paul, I did some patent searching, I couldn’t find anything that the design violated, If you know something in particular that may be a problem I would welcome any information. I haven’t experienced oscillations besides having to counterweight the input cam/oscillator. Plastic gears do have their limits, obviously I wouldn’t recommend this for an industrial robot, but for FRC robot that needs to run for limited amount of time, I believe it can work.


Cool project! Very cool that you’ll be able to sell them.

It seems like whenever I talk about 3D printing at work (that’s all the time), someone always suggests that I use printed parts to make my own metal castings. It seems easiest to do something like lost-ABS investment casting, taking advantage of ABS’s low melting point.

Seems like this would be the perfect project to try this out. Where you have mating surfaces, you could add additional thickness and clean them up in a mill to get closer to the requisite surface finish.

These drives are very cool…I’ve been working to create a useful one for about three years.

I haven’t released a unit for FRC because of the immense challenges.

I won’t detail them here, but I will tell you that we have a “large” box of failed iterations. :eek:

I’ve attached a picture of a 400:1 unit with a 2.5" OD and an 8mm CIM input.

Here is a video…these are from May 2013.

1 Like

I’m in the sharing mood…table of previous design iterations. I’ve been working on these for a long time.



1 Like

Very cool. If these ever came on the market, I’d definitely be interested. Way more compact than a planetary.

I’ve been toying with the idea of a cycloidal gearbox for FRC manipulator applications for a while, but haven’t had the time or resources to really get very far. It’s exciting to see other people running with the idea, and that they have managed to get a lot further.

Sorry I haven’t responded this past week, I decided to write a provisional patent for the gearbox in case there was something worth protecting. It took a lot more work than I thought to get it all written out. Interesting learning experience, I would recommend students interested in design to write a mock or real patent. With a provisional being only $65 to submit, it’s really not very expensive.

I also got the website updated. You can find a little bit more info but the really important things aren’t nailed down yet, max torque, weight, etc. Check it out at

That is an awesome video ajlapp. Such a small form factor for a 400:1. Are you going to keep working on it?

Very cool stuff! I’ve been reading papers about these things and playing with designs for the past few days, I really want an excuse to build one.

Great looking gearbox. I’d love to hear more details on what kind of failures you ran into when going through different designs. I suspect that you need pretty tight tolerances on all of the parts so you land somewhere between “too tight to assemble” and “tore itself apart”.

For the ABS approach, I like the idea of cheap printable parts that can be remade when they wear out. We’ve seen the need for high reduction/low load gearboxes in the past where something like this would work. If we don’t need to replace it every few matches, then it might be worth it (entirely depends on what it’s being used for).


You want to learn about the kings of the industry for this technology? Go to this site:

They absolutely dominate the industrial robotics market. FANUC, ABB, and Motoman all use this gearbox for their main arm joints. I have first hand knowledge that there are at least three of these bad boys (RC-E or RV-C depending upon the need) on each general purpose FANUC robots.

They have cool videos and detailed cross sections of these products.

Why would they do this? They have pretty much all of the patents on this stuff. Many of them are expired, but some of them are still very active.

Anyway, this is the place to go to learn about the best use of Cycloidal gearboxes.


Hi All,

I did some measurements and testing on the gearbox, particularly on the 168:1.

Weight came to 1.1lb (0.5 kg)

Backlash was 1.1 degrees

Efficiency was a bit trickier. I have a supply that has 3 different power settings and a very course amperage meter. I put a volt meter across the leads when I ran with a bare motor and then retested with the gearbox (no load). I did this with 2 different motor / gearboxes and in both directions. The voltage across with the gearbox was around 9, 12, and 14 volts, so I think the efficiency was hurt a little bit by spinning faster than what would be seen in normal operation. Without putting a ton of data down, the average of the results for efficiency under no load came out to be 61%. All but 2 of the results were +/- %7 from this.

I would expect that the values for the 264 and 552:1 should be very similar to the above.

Next up, making an adapter plate for a torque wrench to test max torque.

If you could build in a FIRST standard bolt pattern, we could attach sprockets!

500:1 is probably a bit slow if the plan is to mate it to a CIM motor. At free speed, it would go down to 9 RPM or about 7 seconds per revolution. At peak power, it would be about 13 seconds per revolution. An arm sweeping 120 degrees would be between 2.5 and 4 seconds which would seem pretty slow.

The 125:1 & 250:1 arms have me pretty intrigued.

Are you measuring efficiency as “Actual speed / theoretical free speed”?

Richard Wallace and some others might be able to help you with a more accurate efficiency method that you could do.

A string and spool winch with a weight and stopwatch could probably be pretty effective way of measuring efficiency at several torque levels.

Isaac is right, a string, weight, spool, and stop watch are the best method for this kind of measurement when speeds are low; e.g., when it takes several seconds for the weight to drop.

On the original topic, I recommend this lugnut remover as cheap example of cycloidal gearing. It was easy to find at Harbor Freight years ago. Don’t know if you can find it there now. I posted some details and pictures a few years back, linked earlier in this thread.

Several of the four-arm-linkage lifts I saw this year only needed to span about 30 degrees of arc to pick up a game piece from the floor to quite a few feet off the floor, so 13 seconds per revolution would be a bit under 1 second. At these sorts of ratios, a CIM could compete for some jobs for which we would previously only have considered pneumatics.