3D Printed Harmonic Drive / Strain Wave Gear Using Timing Belt

After being inspired by a recent Hackaday article on hypercycloidal gears I looked into harmonic drives / strain wave gears. Other people have made their own and even 3D printed their parts. However, the flexible splined "cup" looked like the hardest part to make so I wondered if it could be made by constraining a closed timing belt against a flat disk. Seems like it works in principle!
I know there are limitations as to what desktop FDM printing can do but given the proliferation of 3D printable small robot arms, with the right bearings this might be a nice little mechanism to open up the range of motors that can power these hobby robot arms. There may be other uses such as a high resolution panning mechanism (photography, 3D scanning...).
If there's any interest I'll put the files on Thingiverse but I'm sure others would do a better job of turning this into a load-bearing design. An openscad or other parametric part generator would be good if it got to the point of being usable design.

[@PaulJurczak]

Very clever use of timing belt!

[@simonmerrett]

Paul Jurczak thank you. I've just made a gear without having to constrain the belt so there's no distortion along the teeth. I do this by simply having one outer ring with n teeth and the other with n+2. Will try and make a short video to show at some point. The key to making this actually useful is being able to constrain the assembly so it can transfer loads while dealing with all the sideways moments etc. Anyone know any good larger diameter bearings for this? I've got a lazy Susan bearing but it requires a clamping force to hold the faces together and that makes the mechanism harder to design.

[@ompatel5226]

Amazing u have created a simplest and best design on whole KZbin

[@MCRuCr]

It depends on the requirements of you solution

[@ahobimo732]

I was just watching videos on strain wave gears, and I had the same thought (to use a timing belt as the deformable element). But I could never have put the idea into practice as elegantly as you did here. This is fantastic work!

[@simonmerrett]

Mark Osmond thank you for your kind words. I have improved the design from a mechanical perspective in my more recent video on the strain wave gear.

[@das250250]

That is fantastic , thanks for sharing ..another practical 3d print exhibition. it also could be used to convert a microstepping stepper to no micro steps .

[@lucianolizana446]

This is really cool !

[@joshmnky]

It's amazing this isn't more popular. It seems like using a manufactured belt in lieu of a 3D printed flexspline could greatly increase durability. Friction isn't as much of an issue here as in 3D printed gears as well. I really think the potential cost savings and design flexibility could make this revolutionary for diy robotics.

[@simonmerrett]

joshmnky thank you for your encouragement. There's a lot of development needed to get this as a functional gear. One idea that excites me is that if the mechanics can be sorted out, the wave generator arm might be able to be driven by a metal gear servo. If you modify it for continuous rotation and put the feedback potentiometer on the outer slow moving gears you could have a cheap closed loop gear.

[@joshmnky]

+Simon Merrett I've been thinking about this, and it seems like I might be able to make a version with just my dremel drill press (don't have a 3D printer) and dremel circle cutter attachment. Some belt profiles have essentially round teeth that would fit perfectly into a drilled hole. If the gears were aluminum, do you think the belt would last under a great deal of torque? I want giant servo! haha

[@simonmerrett]

I'm definitely no expert on the belts. Bear in mind they are flexed in the opposite direction they were designed for - I don't know if that's significant. If you can drill, cut and finish a tooth profile which prevents slip and isn't sharp (so it doesn't cut the belt), I can't see a reason this wouldn't work for a decent amount of time. Consider that I haven't tried spinning the wave generator quickly yet and others who've 3D printed their strain wave gears have seen the mechanism heat up in normal use. The profile of the wave generator arm that is in contact with the inside surface of the belt (the bearing edges) is also probably a factor in the torque you can transmit with this design because it determines how many effective belt/gear teeth are in contact with each other. I think Barton Dring used an array of bearings to create an oval, rather than the two I have at each end of the wave generator arm and his way may allow better contact area. Please post a link to your servo when you've built it!

[@CatNolara]

I don't know how good this would hold up with high torque or with higher rpms. Some ideas to the design: -use bigger bearings to push against the belt so you get more teeth into contact -use a smaller belt, this way the teeth will only rub against eachother where the bearings aren't pressing on the belt -try to find a belt as wide as possible for maximum contact -preload the bearings with some kind of spring mechanism Anyway, in normal belt applications you have really good contact conditions because it literally wraps around the pulley. In this application though it has only proper contact to a couple of teeth and some just on the corners.

[@mgiaros]

If this design could handle 500 w, 3000 rpm rc outrunner motor and a gear ratio of 100:1 it would be a revolution in mid drive e-bikes.

[@janickjorgensen2964]

As of now, they are very expensive to make

[@adisharr]

Very clever using a timing belt, this looks like one of the better designs. I would love to know how ling the belt would last, they;re usually pretty durable.

[@BenryanALS]

Sleek engineering man!

[@JaredAllaway]

very cool!

[@Pompiduskus]

This is awesome !

[@simonmerrett]

Pompiduskus thank you. I realise I still owe people the files on thingiverse. Hopefully we can get a stiffer version, able to be used in robot arms. Barton Dring has a project on the go which looks promising. Check out his video of the prototype.

[@Pompiduskus]

I saw same same principle but with 3 or 4 arms, and i like it to, because you have fore time the surface grip. If you could share the model files, i would like to help develop this model. In last couple days i saw something similar, but ridget body.

[@dorianmccarthy7602]

In the cases where you are referring to either side of the arm, you're referring to each side of the arm. Either is one or the other and each, refers to both.

[@simonmerrett]

270mm circumference, not diameter.

[@3D_Printing]

"Speed Rings" are like washes but do not bind against the skate bearing

[@simonmerrett]

The standard M8 washers seemed to do fine without binding but thanks for the tip

[@FPChris]

I’ve been trying to print the belt with TPU

[@simonmerrett]

Here are the files for this version: www.thingiverse.com/thing:2013137

[@SebastianOsterfeld]

Very nice! Not having to constrain the belt makes this gear more simple and intuitive. Is it possible to buy these parts somewhere as a kit? I would like to buy one, if it's not too expensive.

[@simonmerrett]

Sebastian Osterfeld I'd wait until it gets developed into a more rigid design as it's quite "floppy". If you're happy with that and can't wait for something stronger, perhaps get a local 3D printing service like 3D hubs to make one for you. Use the thingiverse link in another comment on this video for the files.

[@mirceaar]

Fantastic job. How did you get the HTD 3M profile into the design? Using some OpenScad pulley design and revesing/subtracting it from a large block won't do the trick as it's not a reversible/symmetrical belt design.

[@simonmerrett]

It's been so long I can't remember, sorry. But I do remember using the dimensions of the belt profile taken from diagrams. That said, my printer wasn't that accurate, so it may have been quite a rough estimated profile and some trial and error with print tolerances. The onshape files are probably still lurking somewhere publicly, although I don't touch onshape anymore.

[@mirceaar]

@@simonmerrett Thank you for the answer, I sketched the reverse tooth profile in F360 and managed to create a toothed crown. Best of luck.

[@MatchaMakesThings]

Did anyone ever find the stl/cad files for this?

[@simonmerrett]

This is now a project on Hackaday.io for anyone who wants to join in: hackaday.io/project/19405-strain-wave-gear-with-timing-belts

[@amantin]

Questions : 1)Could you add add a third circle so that you have a compound ratio? 2) do you know any bicycle that works with this at its gears? Thank you!!

[@simonmerrett]

Yes. I did consider doing this but got distracted away after making an unsuccessful initial attempt. I don't know any bikes which use this but perhaps their priority is being able to change gears and I don't know how you would change the ratio on this.

[@cyberspider78910]

Can you put files on thingverse ?

[@NSAwatchesME]

nice

[@simonmerrett]

thanks!

[@budasfeet]

Awesome! Looks great! 👍 Where can we find it on thingverse?

[@simonmerrett]

budasfeet I'll pop it on soon!

[@SebastianOsterfeld]

One issue is that the red and black gears have slightly different diameters, so the timing belt will be twisted. Is there a way to improve that?

[@simonmerrett]

Sebastian Osterfeld yes there is! I've tried to explain in the project log of my hackaday.io pages. The link is in a different comment on this video.

[@simonmerrett]

Here's the link to a video explaining: kzbin.info/www/bejne/nprUZoR6mr2qY5I

[@jamesmatheson5813]

I need a harmonic drive I can backdrive if you can recommend one Thanks for the video

[@353click]

You can't back drive harmonic drives.

[@3D_Printing]

Looks like high Torque :can't work it out but 90 to 92 (45 to 46)

[@simonmerrett]

Not sure about torque transmission yet but gear ratio is 46:1 (I think).