Making a 3D Printed Harmonic Drive Using a Timing Belt

A harmonic drive reduces the RPMs of a motor output shaft and increases the torque proportionally. I created a proof-of-concept harmonic drive using 3D printed parts and a small, closed-loop timing belt.
The motor I used - amzn.to/2LeZzQi
Disclaimer: I may earn an affiliate commission when items are bought through these links.
Support my projects - / levijanssen
Get more frequent project updates - / leviajanssen

[@BenryanALS]

The original 1955 design is genius but I have to say this is equally impressive. You need to pay closer attention to the elliptical geometry so that 30% of the teeth are always engaged. As has been suggested a larger set of bearings in the wave generator will improve things.

[@unionse7en]

Non back driveable is a drawback for arms that you would like to teach via manual positioning. Life is full of compromises . I like the belt as spline idea!

[@gordonjones1516]

From what I've seen and understand, the flex spline (the timing belt here) is actually physically attached to the output housing in an actual harmonic drive. It's like a bell shape with teeth on the outer rim, with the entire bell flexing. In your design they are separate. Essentially you've created an interesting variation of a harmonic drive with a floating internal flexure. Try using PET to print the output housing as a single unit. This should give you the rigidity to achieve the calculated output torque. I'm gonna try it myself, this is a nifty print.

[@LeviJanssen]

You are correct. I didn't opt to print the flex spline because I didn't think my printer could handle flexible filament.

[@BenryanALS]

@@LeviJanssen I've been playing with colorfabb vario shore. Its so clean to print with. I make prosthetics with it

[@thomasmills6789]

Beautifully simple design, I love the belt idea.

[@gsegallis]

Cool project, but a couple of things... Your gear ratio is actually 25:1, not 26:1. (it's (50-52)/50 wrt the motor rotation) In a typical harmonic drive, the flex gear would be the output stage. You actually have a split-ring gear harmonic drive which is a little different. In the case of a split ring gear, the tooth count of the flex gear is actually irrelevant - it is the ratio of the output ring gear to the stator ring gear tooth count that determines the gear ratio. You didn't state it, but your output ring gear appears to have 50 teeth, which is why you get a 25:1 reduciton. If you kept the 52 and 50 tooth ring gears but changed the flex belt to, say, 48 teeth, you would still observe a 25:1 reduction. And if your output ring gear had 52 teeth, it would not rotate at all (regardless of flex belt tooth count, though the flex belt will still rotate).

[@mprotec1]

just use a bigger diameter bearing...to have better gear torsion! this will improve your torque!

[@radarmusen]

This is brilliant. I guess it can end to be pretty silent with a rubber belt.

[@1-t254]

I have wondered on these why the flex spline couldn't be replaced by 2 smaller cogs. I really need to get into 3D printing.

[@londonnight937]

I don't understand why the output ring gear doesn't allow the belt's teeth to "walk" past its teeth, like it happens with the stationary ring gear.

[@ThijnKolk]

I watched a video from Jetpack Academy called Harmonic Drives explained, he explains very well how this system works

[@paulkahler6373]

The upper ring probably has a different number of teeth compared to the bottom ring. It must right?

[@kaden56]

PLEASE keep this project up! I will be entering the harmonic drive realm shortly and I would love a current community online to collaborate with or at least study what their doing. Great work and great video!

[@LeviJanssen]

I certainly will! I’m looking into 3D printing a flex spline for future iterations, and I’d love to see your future work and ideas.

[@jimmymark8232]

Wow, awesome work Levi..thanks for sharing. I have been looking at planetary gears which are expensive. This is a very good alternative. Thanks bro👍👍👍

[@ekaggrat]

does the top lid have the same number of teeth as the bottom or 2 less like the belt?

[@paulkahler6373]

I came to ask the same question. They must have different number of teeth, or the belt teeth would slot into the upper ring teeth the same as the lower and the top piece wouldn't move. So the belt won't fit both sets of teeth the same. Not sure if that's a problem. It would be nice to see the poster confirm differing tooth counts.

[@THRILL_MANJUM]

You're a genius. I'm gonna make a harmonic decelerator.

[@willschmit436]

Consider this: buy two belts (one with 2 more teeth), turn it inside out, and crazy glue it to a 3d printed cylinder...

[@kyleeames8229]

It looks like it could be a fully functional design with only a couple of simple modifications that you already intend to implement. It’s a very elegant feat of mechanical engineering. This is just a small semantic nitpick, but can we call these devices strain wave gears? It’s much more descriptive and It’s already a widely used term.

[@hipphipphurra77]

The belt is less of a good idea because it deforms in a non elastic way consuming energy and reducing the efficiency. Another issue with the belt is that it introduces backlash.This part should be made of a rigid elastic material. Just print it and it will be much better.

[@j.o.5957]

These are great, didn't know they existed, thank you

[@AdrianStaicu82]

How many teeth does the top part have? From my understanding of harmonic drives bottom part should have 52 teeth. Belt and top part should be one unit, so I don't see the point of having bearing to push against that.

[@TheRainHarvester]

I counted 52 teeth on the part attached to motor. 1:26 The top was harder to count but I counted 50 teeth. The belt had 50 teeth.

[@AdrianStaicu82]

@@TheRainHarvester That's impressive that you actually counted. If you look at 5:56 you will see that there is a bearing on the top part of the belt. In my opinion top cover and belt should match exactly, in size as well, with no actual room for movement and could be glued. At least all the diagrams I saw were one solid piece of metal. In this design the top cover might have the same radius as the bottom part, but have it's teeth spread apart more not matching with the belt unless it's deformed by the top bearings and has two points of contact. Some friction may occur.

[@macksoneh]

Can you please post a link for the small timing belt you used for this demo?

[@ray-charc3131]

Timing belt! You need to know why there is little backlash and long usage life time of it when it is under normal use. The tension force acts on it is along the length of the belt and it is counter-acted or withstood by the fibre strands inside. In yours application, it didn't use any of this benefits, instead it just uses a couples of rubber tooth to hold the two internal ring gears together. Flexible rubber creates backlash when they are under the torsional force, and belt will break apart after a long time high loading usage.

[@ninehundreddollarluxuryyac5958]

Belt is better suited to the flexing a flexible splined cup sees. Is this a two stage gearbox? It looks like the belt should add a second reduction when coupling to the output. What is the reason you didn't continue with this instead of going on to cycloidal gears?

[@listerdave1240]

I think it might help if you used larger diameter bearings to press against the belt, so as to get more teeth to engage. That should reduce the chance of slippage and also make it run smoother.

[@JohannesLauesen]

This is really a good design, Impressive.

[@AliIsmaeltyphoon]

Wishing you good luck for your future projects. Excellent explanation.

[@COSMICAMISSION]

great project, thanks for sharing!

[@p4wellis]

why does the top rotate, isn't it just a (more or less) mirror of the stationary piece attached to the motor?

[@joshuahuang07]

the top has 2 less teeth which is why it rotates.

[@EmmanuelRobert1]

Pb with the nema is that it has no torque at high speed. So definitely either you should consider another motor. Nevertheless great job with this harmonic drive : very inspiring

[@miron__]

That's the point of the reduction tho

[@franciscoanconia2334]

That's an excellent idea!.

[@DuncanAitken]

Ingenious!

[@bogdanluncasu5772]

Good video, thanks for posting it. Although it’s definitely simpler to make this drive with a rubber belt, ideally the three components would be made from the same material to reduce wear, ideally something with low friction coefficient. The shape of the teeth should also be revolute, that will increase the force of the alignment. I will cut one out of HDPE on my CNC, the design is definitely CNC-able and I find that HDPE or nylon could be the ideal materials for a cost-effective build.

[@benjaminmuller9348]

What are revolute teeth? Wasn't able to find anything on the internet

[@MJ-iy4fb]

Nice. Thanks for posting the video.

[@Robotinnovations]

Really awesome, have you made any new versions of this, that can be used for robot arm

[@mahdijavanbakht642]

Haven't you change the bearings yet? Did larger bearings solve the vibration and torque problem? I want to use it as a functional part in my telescope mount. Can you make a video of how to design it in Fusion 360, like the one you did for cycloidal gearbox? I don't know how to design an internal gear to pairing the HTD belt.

[@scp-yearsago]

I am trying to build an equatorial mount. Did you find what you were looking for?

[@mahdijavanbakht642]

@@scp-yearsago No, my project was abandoned. But as soon as I buy a 3D printer, I will try to print a harmonic drive.

[@diego.almeida]

You said it is easy to make, cheap and doesn't require tight tolerances... well industrial grade harmonic drive is just the opposite of it, since it uses elliptical bearing which is very hard to make one with decent quality, good harmonic drives have price tag of 700 USD and up.Nice video though. Just curious, why did you abandoned the harmonic drive project and focused on the cycloidal drives? cheers

[@BenryanALS]

I am struggling to get my head around the elliptical bearing too.

[@jimmymark8232]

Is it possible to use triangular centre hub with 6 total bearings? 🤔

[@LeviJanssen]

Yes, it’s fairly common.

[@AbdolazimHasseli]

Great! What is the number of teeth on the upper part?

[@772777777777777]

Harmonic drive vs cycloidal drive which better for robotic arm

[@asgharrezaei0138]

Good job , you are doing fine Good luck

[@bricksanddirt1917]

Wait... Was this only a test?

[@FPChris]

Would TPU work to print the belt? I bet it would

[@dieselphiend]

I had previously thought planetary gears were king. I now know better.

[@sermadreda399]

Great vedio ,thank you for sharing

[@yiptastic12]

gear ratio calculation: flex spline teeth = 50; circular ring teeth = 52. Ratio = (Flex spline teeth) / [(circular ring teeth) - (flex spline teeth)] Ratio = (50) / (52 - 50) = 25 how did you get 26?

[@FlexibleBladeTurbine]

It obviously depends on the side the stepper is mounted. So: Ratio = (circular ring teeth) / [(circular ring teeth) - (flex spline teeth)] = 26

[@Sega-and-Mao]

Nice job!!! Any link to gcode for the parts printed or any model to share?

[@semphony100]

did you make the robotic arm with it

[@davidbroadway3012]

Use a cam nut (eccentric bearing) as one of the rotor bearings.

[@TheRainHarvester]

Can you explain what it would do? (Trying to understand)

[@rogerlarge484]

@@TheRainHarvester It would allow easy adjustment of the tension inside. Make it tighter without reprinting everything.

[@n5sdm]

Use a power transmission belt. See if that makes a difference. Timing belts slip too easy

[@facudejota]

Greatt!!!! Want to see more

[@artantme]

Cool man! Have you made any further progress? I'm thinking of making some, and i really like that you used standard timing belt.

[@LeviJanssen]

No, I haven’t. I much prefer cycloidal drives, so I’ve been spending time working on those.

[@artantme]

@@LeviJanssen any particular reasons?

[@LeviJanssen]

I don’t like the idea of using soft bodies in a high torque system. Doesn’t seem reliable. And I just think cycloidal drives are beautiful.

[@artantme]

Levi Janssen That’s an inspiring answer :) I guess i’ll try both, but my internal feeling tells me that it is actually not that bad to have something rubbery inside a 3d printed mechanism, so it absorbs some shock instead of breaking a piece.

[@BenryanALS]

@@LeviJanssen Im looking at making silicone moulds for PU parts. They should be plenty strong enough and allow you to replace the timing belt with a cup type part which will eliminate the issue of belt travel

[@ameliabuns4058]

clever...... how do people have such awesome smart ideas ?

[@crazyhank99]

Awesome! Wish I could give 10 likes

[@stanley6602]

Hi, Can you share us where you purchase that timing bell?

[@cc950]

NO

[@fdavpach]

Look on Aliexpress for an HTD 3M 150, it's a 50 teeth, 150mm tall with a 3mm pitch, you just have to turn it inside out.

[@mahdijavanbakht642]

Hi Levi I really need your helps about this design. Has its torque improved after the bearings have been replaced? How much is its practical reduction ratio? Does the friction between the belt and teeth cause the belt to wear out?

[@LeviJanssen]

I never replaced the bearings. I didn’t need to. The reduction in a conventional harmonic drive is half the teeth of the flex spline (25:1 in my case). I’d imagine the belt would wear some. Mine really wasn’t used for anything, as it was just a test, so I wouldn’t know exactly.

[@mahdijavanbakht642]

@@LeviJanssen Sorry, You said in the video you want to use a longer internal bar. Did you test it? How can I design the internal gear tooth profile?

[@LeviJanssen]

I didn’t test it. To design the tooth profile I just took some measurements and approximated. Nothing too scientific. It took a couple tries but it worked.

[@mahdijavanbakht642]

@@LeviJanssen I have to make myself one and see how it works. Thanks

[@mahdijavanbakht642]

I came back with few questions. 1-Is the top piece has same teeth number as the belt? 2-Should the lower piece teeth have same profile as belt teeth or there should be much more clearance? If so, how much clearance is enough?

[@flassmaximusgaming9950]

how many teeth in the top gear there mate?

[@ilgazaskin]

Nice work! Is it possible to share the CAD files?

[@LeviJanssen]

It’s certainly possible, but I’ve never done that before, and don’t know how exactly I’d do it.

[@BenryanALS]

@@LeviJanssen export as .f3d under save menu

[@davidjohnston4240]

@@LeviJanssen Thingiverse is a good option. I put mine stuff there.

[@RobertSzasz]

Wait... It's 50:52:50? Why would you get any output at all?

[@RobertSzasz]

I guess you could mount the output eccentric and make it a 51 or 53 tooth?

[@hansgunther6522]

I think it is 52:50:50 (Input:Belt:Output). So the belt moves 2 teeth per cycle of the motor which results in a 1:25 gear. (2/teeth of belt). The output moves with the same speed of the belt as they have an equal numer of teeth.

[@e.k.anathorist6910]

Main usage of a harmonic drive is not gear reduction, it is to reduce back slash

[@sergusy7005]

wow. 0.9 degree motor...

[@jmile69]

You're better off if you can make an elliptical greased mechanism rather than the bar you have in there.

[@TheRainHarvester]

Why is that better? Just curious

[@dr.rajasaurusandunclebonec6526]

who else was thinking about the girl from Harry Potter!

[@huaahhggg9507]

Why deteriorate your channel with a half assed project?

[@hikolanikola8775]

this is wrong.... you are not putting the correct number of teeth