1:6 reduction ratio www.thingivers... / gouldpa00 Inspired by Gustav Olsson - Inversed Timing Belt transmission • Inversed Timing Belt t... / @gustav.olsson
Thanks again Paul! I have built and tested this unit, which may not provide the torque necessary to do "root-joint" duty, but it is unbelievably smooth and quiet, making it PERFECT for silently panning a camera, so this is my chosen 1-DOF "wrist" joint for a robotic camera arm I'm building that is around 50% complete. In fact, the rest of my robotic camera arm uses your cycloidal joints (except, I regret, the center pivot (slew?), for which I chose a compound planetary - not back-drivable) , because I like the way you balanced the torque load among the cycloid disks and cylinder by splitting two typically thick disks, into four. Your cycloids are unbelievably precise, delivering maybe (on average) less than 0.25 degrees backlash (from my Ender 3 printers); bad precision for 3D printing or machine tooling maybe, but WAY better than ANYTHING I EVER expected. How did you do it? Anyway, except for printing some encoder and customized motor mounts (NEMA 17 steppers instead of BLDC) and adapters allowing me to connect YOUR cycloid joints to each other in a beautifully modular way, I will have a hard time taking any credit for the "design" of my arm, because its form is defined by your cycloidal joints. So, thank you Paul, for designing my robot arm for me. (So, where's my cycloidal pivot drive, hmmmm? JK...) Another very cool thing, because of your cycloidal joint designs, I also discovered a better way to use the ubiquitous AS5600 magnetic hall encoder, which is being universally mis-applied as a shaft encoder, (a job it CAN do, but only under certain restrictions), limiting the potential speed in their designs. YOUR joint designs revealed this fact, so hopefully before much longer, I'll be able to demonstrate this thing, but YOU get the credit for its major design.
Hi Mark, Sorry missed this comment. Can you please share a link to your work and chat with me on hackaday or Instagram. I like sharing these designs and seeing people making improvements. Tell me about how to use the AS5600 better.
@@PaulGouldRobotics I will get the links to you, but I didn't even have to make any "improvements"; your designs worked amazingly well "right out of the box". Now, I have to wait until I can get the remaining boards for the controller (at a reasonable price). "Better use" of the AS5600 means people need to stop trying to use it as a shaft encoder, using it to sense joint position instead. It isn't fast enough to reliably detect high RPM's, and nobody needs the insane resolution this method provides anyway (on a plastic joint), but it does have more than enough resolution to precisely measure joint angle, and since it is an absolute encoder, once set, no "homing" routines are ever necessary. I'm putting "See Paul Gould's Channel" stickers on this thing before I record it....
If the roller (seen @ 1:27 in the nine o'clock position) were narrower to fit inside the flange of the toothed-pulley, it would serve as a 'pinch-roller' and prevent slippage. The roller could be mounted on a spring, and a sort of mechanical over-load threshold would be realized. Or mount it solid! Belt teeth have a surprisingly high 'tear-out' resistance, but they will hop over pulley teeth given 0.5 of a chance. With a pinch-roller, the chance of belt-hopping is reduced.
This is a wonderful design! It also looks like, in rough theory, a continuous belt could be used. That doesn't serve much purpose for a robot arm, of course, but it might be interesting for some other applications for sure :> It's always a nice creative stimulus to see what can be done with just a few smartly placed pulleys, in any case.
This design went quite smoothly. I just added the third idle pulley so that the belt would go around the centre pulley 180deg. I'm yet to work out a good tensioning mechanism. I just had to make sure that the 45deg twists in the pulley were centred around the cross-sectional centre of the belts.
What about a 4th pulley on a swivel that would hit between the two close pulleys with the straight belt section. Could make it be adjusted with a screw and could keep the integrity impact of the overall design to a minimum. At least from what I'm thinking
Elegant! The only question mark i see is the bearing running on the stepper case which is neither a precision surface nor guaranteed coaxial with the shaft, but regardless, it's elegant!
There the "Tails" are secured try printing a part with groves for the belt teeth to fit in. That should be better than just the pressure/friction of a flat surface. Nice project!
Awesome design! I was thinking that a plate over the shaft and pulleys would allow the pulley axles to be supported on both ends. This should reduce deflection and possibly reduce the slippage under load.. Another (impractical) idea is to put teeth around the joint and use a closed loop belt to reduce the length of unconstrained belt and hopefully reduce the stretching, but God help you trying to install and tension that😋
These are all great suggestions. I have a closed loop belt coming ~400mm. I have designed an additional top plate. I just need to be able to tension the third pulley, somehow.
@@PaulGouldRobotics Oh, cool! I can't wait to see if the top plate helps👍 Would an eccentric shaft work for tensioning? You would need a bigger bore for the pulley.. not sure how feasible it is, given the space available. The low-tech solution could be to print a few pulleys in increasing size and swap them out till you find the right one (not so convenient 😋).. Anyways, just spit-balling ideas here.. I hope it helps😉 I'm really impressed with this design, totally out of the box! Keep up the great work!
Paul, if you were to pick one of your many beautiful actuators/gearboxes, what method of reduction would you pick? I ask because I'm still playing with my robot arm actuator design and I'm still tending towards a compound planetary gearbox with brackdrivability. I like the cycloid drives too especially the few you've shown, but they seem to be finicky. I like harmonic, but that still seems no good with 3d printing. I like your timing belt in this video, but those seem to have their own issues. What would you pick?
For a robot arm I like the 32.5:1 timing belt and dual capstan with internal NEMA17. For a quad I like the 10:1 timing belt and dual capstan with internal ML5010 brushless. Both are cheap, easy to print/low tolerances, zero backlash, high efficiency / backdriveability. It is also light. Tensioning is a bit of a pain but all timing belt designs are. It also looks good (I think) and is different.
I watched Gustav's video as well. very neat idea. I though maybe switching to 5mm pitch belt would help with the skiping and loads. the idlers needs to be gear and not smooth. teeth warp on the smooth face. you can use an eccentric nut on the middle puller for belt tension. larger size pullers are needed for skip reduction or a bearing pressing on the belt outside the pulleys to stop skipping.
@5:44 you can see the slack on your belt. you need to have stonger housing material to keep the tension on the belt. you will probably break the reversing pulleys out of their location if you put too much tension on.
I have some T5 belt, so I may try that. The idlers that have the belt's side tooth inwards have gears. The others do not. Yes, the middle pulley will be good for tensioning. There is not much space. I may have to make the overall diameter larger.
I always love your videos, simple designs for high torque density joints. Why don't you add pulley teeth to the outside of the black 3D printed "Rotor" of the assembly? It would be able to provide a holding force for the belt. Could be difficult for modeling, but it could be well worth it in the end.
I thought about that and it would make for an exact reduction ratio. It is currently just under 6:1. With teeth there I would have come up with a tension mechanism in the centre section. I have other projects with GT2 timing teeth with a parametric model that work well.
@@PaulGouldRobotics Could instead insert the ends of the belt into the body itself, the belt would have to make a 180 degree turn around a rather small radius, but it would be trivial to increase the thickness of the plastic to make it viable. It would mean the black plastic rotor would no longer be perfectly circular, but that doesn't entirely matter since it can't do a 180 degree rotation. Could be promising!
What about using a dual shaft motor and installing the timing belt+pulley mechanism at the back of the motor too? That way it would be double the contact area and would probably decrease the chances of belt slippage.
I have a dual shaft motor. I may try it, thanks. I will try to improve the tensioning first. I think I'm getting close to the max torque of the stepper motor.
that white one would work 4x better of the belt was tight around the rollers and was under tension for better control and drive at 4 points instead of 1 point. The belt isn't doing anything but laying against the inside gear. put a smaller belt on it. it's kind of under compression to stay in place the way it is
@@PaulGouldRobotics I think it would make a great knee joint for an exto-skeleton. If some guy by the name of Miles Dyson starts asking you for advice, whatever you do, don't help him!
@@PaulGouldRobotics haha too funny, I literally just found your channel. I an been thinking about making a exoskeleton for my dog. She is starting to lose the use of her back legs (degenerative myelopathy). I am not sure how I would get it all to work, but went RU-vid showed me this video, I recognized that your design is awesome.
@@PaulGouldRobotics was thinking of something like a pull chain for a lamp. Guess that wouldn't work either though. Great design and I love the work and can't wait to see more.
Could it be beneficial to turn the belt inside out, by doing so you can lose one of the three idlers at the center and it should still do the exact same.
That was the first prototype. The belt had a very small contact angle with the pulley. Now it is almost 180degs. Please send a drawing if I’m not understanding your comment correctly. All improvements are welcome.
@@PaulGouldRobotics I almost want to apologize for my comment, I thought I had an easy improvement. But after a bit of trial and error all I am left with is an enormous amount of respect for the original design. Very impressive!
Jak na moj gust, to mozna to samo zrobic zwylka przekladnia z silnikiem krokowym. Nie rozumiem poziomu niepotrzebnego skomplikowania takiej przekladni.
I have a few different cable designs which are currently my favorite reduction mechanism. I can not work out how to make this one with cables and maintain a constant reeling and unreeling difference length.
@@PaulGouldRobotics Ah, looking forward to seeing those! I've also found this: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-utDagouxM5U.html which can be quite interesting for some robots, as the motor doesn't have to be at the position of the joint
@@simonedallai48 Set the output to 0degrees and record the multiturn encoder value, move the output to 90degrees and record the value. Do maths to get the scale factor. You will still need to reference every power up. I would suggest just moving the encoder to the output.
I like using the absolute magnetic encoders because you know the exact joint position on power up and you don't need to reference every time. MA730/MA732 are also very cheap.
Nice method! That’s a robot arm now😂 Have some idea, the method of the pulley, you can try to add this design on the big black ring, to make a much stronger output! Thus it reduced the speed, don’t know what’s your goal😂
@@PaulGouldRobotics one puller will make the ratio doubled 😂, but the problem I know about the stepper motor is when it spins fast, it’s output decreases... well failed to find somewhere send pictures, I was thinking about make a drawing of the belt line, of how it can make that...
The pulleys are bolted to printed plastic parts and loaded in single shear ... you shouldn't be surprised the belt skips at such low loads. Make a brace for the pulleys, get proper Gates belt and pulleys, not the crappy china stuff, and the difference will be like night and day. Still, a clever design, thx for sharing.
This is a first prototype and I'm using the viewers to give improvements. I will add dual sided pulley support in the next one. I didn't know about Gates belt, nice.
