pic: HGX-102, 36:1 harmonic VP stage



This is HGX-102, a harmonic drive prototype I’ve been working on for a few weeks. I printed a new gear with slightly different tolerances every few days until I got the fit I wanted. Printing was done with an Ultimaker 3 Extended; I would not recommend printing this on any lower-end printers.
The hub with the bearings is a machined part I got from Plethora Machining. It’s possible to 3D print the hub, but the bearings and screws need to be bought.

The final gear ratio is 36:1, or (#of teeth in ring gear)/2; this is true for most (or all, I believe) harmonic/strain wave drive systems.
I was hoping to get good anti-backdrive results with this, preferably better than the cycloidal stage, but ultimately even the 3D printed version is very easy to backdrive. On one hand, that means it could have good efficiency, but on the other hand, it also means that the extra manufacturing costs and still-worse-than-planetary efficiency will be a hard tradeoff to make for just a bit more gear ratio.

I think I’ll lay harmonic and cycloidal drives to rest for now. They’ve had a good run, but the only place where I would seriously want to use them is a larger version on an arm or something. The quest for a non-backdriveable non-ratchet VP stage continues.

Link to photo album and a short backdrive test

Link to CAD. Look in the HGX-102 folder. The Zip folder is a PackAndGo and has the printable parts and a STEP file as well of the whole assembly.

1 Like

That’s really cool. I appreciate you putting in the effort to try to make a harmonic VP stage work. Aside from the non-backdrivability, the harmonic gear system should have really low backlash. How was the backlash on your prototype?

Any CAD files I could look at? This is really cool.

How is the max torque output? We know the 10:1 VersaPlanetary has a problem with shearing gears if used at the end of the geartrain. Would you say the 36:1 harmonic is better, about the same, or worse in that regard?

Pardon my ignorance on this subject, but is the effective ratio of any harmonic stage utilizing existing VP ring gears fixed, or is there some flexibility to go to a higher or lower reduction?

I one thing you also have to consider is that a reduction like this could substantially reduce the size and weight of a VP assembly. This could potentially allow certain 2 stage reductions to be consolidated into 1 stage, and 3-stage reductions down to 2 stages. Backdrive prevention would be a nice bonus, but, at least from my perspective, not the main reason to use something like this.

I’ll second this statement. Real cool use of a VP stage! CAD files would be really handy especially how and what the flex spline (ring gear) is made.

Yes, as noted, for harmonic gearboxes, the ratio is usually half the number of teeth in the ring gear:1. You could get less reduction by having a smaller inner gear, but why? Also, this would induce even more wobble as this gear’s center of gravity moves away from the center of rotation.

Also, 3:1 and 10:1 are practical limits for a planetary gearbox with at least three planet gears, driven on the sun gear with fixed ring gear output on the planetary carrier. If you gear down much farter, the planet gears grind on each other; if you go up much farther, the planets have too few teeth to be driven properly.

The theoretical limits for very small teeth are 14.9:1 and 2:1; see STEM Quiz #1 for more details.

Thank you, it was a good experience. The backlash was indeed super low. I would estimate that it was no more than 1* or less, and a steel final version would have even better capabilities. Ultimately, it’s lower bound by the tolerances on the Vex ring gears.

I’ll have CAD on my GrabCAD in a few minutes, Pack&Go and STEP. Be aware that there are two strain wave gears in the assembly; one that has been deformed into an ellipse shape, and one that is circular. If anybody seeks to print one for themselves, use the “HGX-102 flexspline 2” part in the “toPrint” configuration. You may need to tweak the teeth dimensions slightly.
EDIT: The HGX-102 flexspine 2 toPrint configuration is broken. I put the .stl file for printing in the packAndGo zip folder. Currently, it uses an offset of 0.003" on the gear teeth to make it larger, and a 0.008" offset on the output spline to make it larger as well. The input spline stl should be workable and has similar edits made, but is not broken.

This definitely withstood at least 10ft-lbs just with plastic. If it was steel, it would easily beat the 10:1 in that regard.

There is a way to get a lower reduction I believe.
If you look at the “wave generator” right now (the hub with bearings on it) you’ll notice how it forms an ellipse. If that were switch out for a triangle-like shape (I don’t know the actual term) then your ratio would be (#ofTeethInRingGear)/3. Similarly, a square could decrease it to /4. That means 36:1, 24:1, 18:1, etc. are all possible I think, although I’m not 100% certain. I only say this because I’ve seen triangular versions, but I don’t know if that’s really decreasing the ratio.
I don’t believe it’s possible to get 72:1 without making it unbalanced, or at all.
EDIT: GeeTwo’s idea would work as well.

It’s certainly true that you can get higher reductions in a smaller package, but this will probably have a high efficiency drop. I measured the free current on a BAG motor to be about 5 amps with this on, compared to <2 amps with just motor + 10:1. I don’t know what loaded testing would reveal, but I doubt this would be more than 70% efficient in the best case.
A 10:1 at 90% efficiency on an input of 1ft-lb will create 9ft-lb on the output.
A 9:1 and a 4:1 together at 90% efficiency each would create 29ft-lb on the output.
A 36:1 at 70% with an input of 1ft-lb would create 25ft-lb on the output. That’s actually a pretty fair tradeoff (although VP stages are probably better than 90% efficient), but the costs to manufacture the strain wave gear and hub are considerably higher than the costs for a regular VP carrier plate and gears. 8 bearings, a 70t gear that’s been shelled out with the spline pattern, and the carrier plate with tapped holes won’t be cheap; I estimate it would be about $25 for a gear stage on the market.

Thank you! Link to CAD and a small photo album will be in the main post in just a minute.

Speak of the devil. I don’t know if it is a coincidence or I overheard someone mention this topic about a year ago but I have been looking into and tinkering with this exact concept for a cycloidal drive. This one is a 12:1 reduction arm joint that I have been working on since we got back from St. Louis champs. One of those ideas I had kicking around in my head and felt the need to get it into CAD and 3D print as a proof of concept. It works well enough in PLA but requires a fair bit more development before I would ever consider using it on a competition robot.

I’d like to at least finish the rest of this initial concept before releasing any CAD but if you’ve got any questions just shoot me a PM.

Very interesting! That gives me some ideas for how to incorporate this into an arm. My other iterations on that concept leave some things unsolved.

Still salty you built it before we did ;). The 3d printed flex spline was a good idea. Really happy you got to use Plethora for such a cool part. Plethora is an awesome company and awesome sponsor of us. They truly take an interest in the stuff we do. Highly recommend them for anyone who needs a part milled.

Hey, I gave you like 2 months to build it! :stuck_out_tongue: If you are still up for making the steel version that would be neaterino.
Plethora is actually pretty cheap for milled prototypes too. Even if I was a company I would look at them for prototyping needs.

I really like this idea! I might have to take a crack at this kind of design myself. Among other things, I really like that this design would not require a large amount of custom parts.