Strain Wave and Hypocycloidal 3D Printed Gears Update

[@DocMicha]

Nice video. You know there is an online calculator for ratio and design of hypocycloidal gears? I do not want to make PR for the site, but it is a site with a lot of designs. Just google "hypocycloid gear calculator" and you find it at the first place. I managed to get the shapes into Fusion via excel and importing a csv Spline. It is a little bit tricky at first glance, but it works fine.

[@ericvauwee4923]

alternativly I can highly recommend to read this paper upon modelling hypocycloidal gears: www.researchgate.net/publication/303577116_Modelling_the_Meshing_of_Cycloidal_Gears I prefer this one over the otvinta one because it does not need any adaptation with offsets etc and the parametric functions are pretty easy. I use Excel to set the parameters and draw the profile to "play around" and import the r q and e values as parameters into my cad to constrcut the profile.

[@kenmagalnik6596]

I was making something very similar recently, and discovered that the bearing problem is caused by the groove with a half round cross section. This causes many portions of the ball to make contact, and each has a different linear speed due to different circumference. I changed the groove to a trapezoid cross sections, so that only the angled sides make contact with the balls, only in two points. The back edge has clearance, and does not make contact. There is a very drastic reduction in friction, the bearing spins freely while being tight enough to have no easily detectable slop. I haven't tried using a cage, but I suspect that it would work well.

[@scienteer3562]

Really impressive. The output speed of the hypercycloid seems much more consistent too.

[@simonmerrett]

Scienteer thanks, the trick will be keeping the output pins straight under load, especially with the higher ratio version. I suspect the flexibility of the belt could end up being an issue there too - may have to go with steel pins or printed pins.

[@i-make-robots]

Simon, your work is fantastic. Stay awesome!

[@simonmerrett]

Dan Royer thanks, it is may all turn out to be unnecessary when the right motors are made available cheaply but perhaps there's a period we'll be popping these into things, once they've reached a certain standard!

[@nikorPoulsen]

It's really cool to see all your experiments.

[@simonmerrett]

nikorPoulsen thanks, glad you enjoy it.

[@AppleCakes7]

Great stuff! I'm looking forward to future updates

[@simonmerrett]

R J that was quite a long period of development crammed into a 12 minute summary but I'll keep this going.

[@FilterYT]

Awesome work Simon, thanks for sharing!

[@FilterYT]

Well done Simon, thanks for sharing!

[@mikelaurie3850]

Do you have step or fusion 360 files for the hypocycloidial drive available

[@constructiverealities]

A thought regarding running the stepper with high reduction (and therefore high motor shaft speed)... The faster steppers move, the less effective torque they produce. So for a 60:1 the stepper really has to be zinging along, in a condition where it's most susceptible to stalling. I have a 40:1 globoid worm drive that I've been experimenting with and I've been disappointed in how easy it is to stall, despite the high mechanical reduction. my sense if that if this was a DC motor/servo, it would be *very* strong. Does this map to your experience? if the hypocycloidal is lower ratio, but producing high(er) effective torque... is this because of the style of motor? e.g. with a lower shaft speed it still produces sufficient torque but isn't on the verge of stalling all the time.

[@simonmerrett]

Constructive Realities I will pay attention to that as I go forward. As far as the versions I have already made, the grbl acceleration profile I'm using is pretty gradual over the course of a second or two, so I think that if the step rate / motor torque was an issue then it would overcome initially. Not sure if the motor stalls are instant or not (away from workshop for a couple of weeks). The other complication is that only the first hypocycloidal version and the elliptical strain wave version stalled the motor - the other versions failed due to the belt skipping. Good thinking though, thanks. For what it's worth on the motor type, I'm very interested in something like a gimbal bldc paired with something like an o-drive. The NEMA 17 is a good subject for development because it takes minimal investment for others to replicate. If you can do it with the NEMA 17 slowly, you will probably be able to do it faster with a bldc. And the power density of other motors will also help you reduce the mass of a robot arm and you have less of a torque requirement at each earlier joint.

[@constructiverealities]

Simon Merrett i think in terms of apple to apple toque comparisons for the units you have, testing with the input shaft speed consisent will probably yield different results. My hypothesis is that high speeds with the steppers results in poor torque measurement, due to the low load requiired to stall fast moving stepper. Im also interested in bldc for the reasons listed.

[@roshanbirkodi]

Great work Thanks for sharing

[@scalemog2051]

Great work. What I didn't catch from your speech, what was the reduction ratio of the final hypocycloidal gear working with the belt? And the numbers you showed for the hypocycloidal in the spreadsheet, they were taken with the first version with the roller bearings on the outer circle, right?

[@simonmerrett]

Scale Mog 69:1 (I think but rusty on the calcs at the moment - 69 teeth on the cycloidal discs and 70 teeth on the belt). The numbers in the spreadsheet cover a range of my designs but the one with the outer roller bearings had a 12 teeth ring and 11 teeth cycloidal discs, with a torque multiplier of 13.6 before the NEMA 17 stalled (likely due to the relatively low reduction ratio).

[@Graciejojo1]

Can you make a Video about The rigidity of The gear, maybe with a lever arm? Please

[@simonmerrett]

Jojo1 my video setup won't really accommodate filming the gears with the beam I use for testing. The motor is just clamped to the desk. That's why I decided to calculate the torque and compare different designs that way. Perhaps one day but I'd like to get the gear performance better first. hackaday.io/project/19405-strain-wave-gear-with-timing-belts/log/88278-teardown-rough-first-prototype-htd-3m-hypocycloidal gives you an idea of what my test looks like.

[@Graciejojo1]

oke. thank's for feedback!

[@claudekim7876]

whatever happend to the morexy?

[@simonmerrett]

It's sitting here in my office, waiting for a move to a yet-to-be-made enclosure. I just discovered repetier server so will be using that to send prints in future.

[@createthiscom]

This is awesome. Has this project stalled? I haven't seen an update in about a year.

[@simonmerrett]

Yes, stalled in progress but not in spirit. There are several projects that seem to be doing very well at this kind of gear, so without a strong personal use case to drive the development, I'm letting others progress the field. E.g. Open HD (harmonic drive) and Paul Gould's work. Both on hackaday.io.

[@createthiscom]

Simon Merrett I’ve been looking at Open Torque. Something like that seems to be the sweet spot as it would allow for natural feedback at the expense of some backlash maybe.

[@simonmerrett]

Ooh, I hadn't seen that before. Thanks. Seems similar mechanically to this recent endeavour hackaday.io/project/164175-custom-motor-for-robotics

[@davidprock904]

I have an idea without the flexing needed.

[@simonmerrett]

Will you share it with everyone please? Or show us the results?

[@davidprock904]

@@simonmerrett let me get ok enough with FreeCAD and ill put it out there.

[@T70781]

Great content. Thanks for sharing.

[@websonic1000]

You have way to many bearing on roller thats what is killing your wave. Use no more than one on each side.

[@JakeRobertRead]

heyo, if you're interested in making a robot arm w/ this project, you might find that 12:1 is plenty of reduction if you can use a BLDC motor on the input side. steppers are pretty low torque density! jakeread.pages.cba.mit.edu/actuators/

[@simonmerrett]

Jake Read nice tool you have there! I agree the BLDCs are a good way to go eventually but I like Dan Royer's approach to making this as accessible as possible. So far I see drivers like o-drive which could do this but aren't yet as common and cheap as a MEGA2560 and Ramps with stepper drivers. Specifically because I'm looking at the highest torque joint for these designs/tests, it's right above the base in many arm configurations, so the torque density wouldn't have much impact on the spec for other motors in the system. Of course, once you start working your way out to the effector, people should go and check out your website (and Dan Royer's github.com/MarginallyClever/armTorqueCalculator) because it gets increasingly worth their while to invest in torque density and you've made it really easy to see how to carve the budget up. Thanks!

[@Build_the_Future]

I may try something like that in my robot. I'm building a 2 legged full size walking robot on my channel check it out and let me know what you think thanks.

[@vitezhrabri4054]

999 sub dam so irritating