Love your dedication to perfecting 3d-printed cycloidal gear design. Having affordable actuators will be so useful for making robotics accessible for the masses. Also that fist is just perfect! Keep up the awesome work.
i've been using dowel pins as roller pins for the ring gear and to make my own bearings. It works well if you want to have a really cheap cycloidal joint but you will have to struggle with tolerances.
My printers are pretty bad on the tolerances/roundness. I have found Needle and ball bearings drastically increase the efficiency of my 3D printed gearboxes.
Wow! Just awesome how much torque it reliably produces! Your efforts are great and that it's open source and modular is awesome! Cheers mate and enjoy the success!!
I've been gradually improving this over the past few months. I think a higher cycloidal ratio and lower timing belt ratio would be better. The tension on the timing belt is very high.
An epiphany I had recently regarding cycloidal design: With discs 180 degrees out of phase, the ring is distorted into an ellipse by the radial pressure. If you use 3 discs 120 degrees out of phase, or 4 discs 90 degrees out of phase, or even more discs equally spaced, the necessary ring stiffness goes down because the distance between pressure points is reduced. More vibration than the fully counterbalanced style you're using, but less than the common two disc style, and the ring is a large portion of the overall weight so this can save a lot. If you ever machine one from metal, another large weight reduction can be had by replacing those big output bearings with thin strips and circles cut from teflon sheet. 3D prints may not be precise enough to use this, but you could give it a try.
Thanks for your knowledge. I use the larger bearing to keep the ring from distorting. I've just some thinner wall bearings to lighten the joint. I have found that 3 discs at 120deg imposes a twist between the upper and lower housings. At some point, I will move beyond PLA 3D printed parts but I'm happy with the performance right now.
@@PaulGouldRobotics I see! Very clever using the bearings to prevent inward deformation of the ring, which is what actually pinches the discs. And that explains why twisting/asymmetrical deformation happens with equally spaced discs, since the ring itself can be relatively flexible then. I assume skipping of teeth is the limiting factor with this design, due to outward deformation of the ring? Another good place for plain bearings is the output rollers. I tried needle bearings first, then SF1 bushings pressed inside brass tube to cover the seam, and now use PTFE tube pressed inside brass or steel tube. High load capacity, small diameter, silent, cheap, and only slightly higher friction than needle bearings. Though even with a tubing cutter it is fairly time consuming cutting and deburring tubes to precise length. And not all PTFE tubes are concentric, so it can be trouble finding a good supplier. You could use this style for the ring pins too. I tried metal-on-metal steel tube sleeves over steel pins, but instead of the tube spinning over the pin, the pin itself spins along with the tube. So now I just use steel pins alone, which has the added advantage that the pins can be shorter length since they don't have to have their ends stick above/below the discs. Though in your case with the bearings functioning as ring stiffeners, having the pins supported by their ends right next to the bearings is an advantage.
Nice Design nicely built! Have you tried a design with a strain wave gearing that makes use of compliant plastic? I have seen some fairly small but effective designs and it would be interesting to see a comparison with a cycloidal drive of similar size and/or mass.
I have made a strain wave gearbox using TPEG. It was not that strong or efficient. Maybe a higher ratio would be better. Check out my Instagram page with the yellow and blue 3D printed parts.
Hi Paul, I found it interesting that you replicated an arm to show how this joint can lift a weight. It made me think how inefficient lifting weights is for stressing a real human joint, and how a mechanical joint would give various advantages. Question for you about possible practical implications. I've seen first hand some of the rehab practices of people recovering from limb surgery (specifically ACL knee rehab). A big thing during rehab is two fold. Gradually increasing range of motion in the limb, and also applying force (both flexion and extension) at different ranges/angles. It occurs to me that a robotic arm is probably very well suited to applying these controlled forces. Have you ever thought about building this joint into a type of "workout tool"?
In my experience (I'm in Sydney), rehab for ACL reconstruction is pretty much run by the physio profession. I saw you're a Perth Wildcats fan, you might be well placed to reach out to the medical team at the Wildcats (and/or an AFL team) and talk to them about it. Those two sporting codes would be bleeding edge in terms of ACL treatment. For an example of a brace that is pretty much already an exoskeleton, look at the appliances made by Bauerfeind, like the Sectec Genu. They are German engineered, very lightweight and strong, it would be a very good starting point to add controlled motion to.
I'm curious if you've tried split-ring planetary gearboxes or Wolfrom planetaries before (I believe they are similar)? They seem like an interesting alternative to cycloidal gearboxes.
Nice design. I have few questions I hope you can help me with. Why did you choose to go with belt system instead of making the cyclodial drive 2 stage with more reduction to acheive the same? Also there's a noticeable shake when the arm is being lifted. It's amazing as it is for a plastic gearbox to handle this weight but I just want to know this is a drawback in the cyclodial gearbox design or something else is contributing to this vibration. This will really help me in designing my gearbox so thank you very much in advance
I've tried a dual stage cycloidal and I couldn't solve the wobbling, see a previous video. Cycloidal work well because the forces are spread out over many teeth. The wobbling is caused by the center eccentric shaft bending. It is not the strongest part.
@PaulGouldRobotics thank you very much Paul. Exactly the answer I needed. Does that mean if it was made from metal parts the wobble issue could be solved without needing second stage? Still one part I hope you can answer, I'm not sure if your answer was referring to this but why did you favour the belt as a second stage instead of 2- stage cyclodial?
Just a though... have you tried a eccentrically cycloidal planetary gearbox? In theory you could make a planet gear with essentially "one tooth" As a boon the bearing could sit in the planet carrier so you could print the planet gear as a simple dumb plastic piece. This could drastically cut down on bearings.
Hi Paul, I would like to use this Elbow joint in a real prosthesis elbow design, but i need the size to be smaller. Large Dial(White) in your picture , in my case needs to be less than 30mm Dia meter, Motor should be smaller and For Arm portion should be smaller, can you please help? i need help with resizing, this will be a child prosthetic, so about 3/4/5kg/30cm is more than enough.. Please help.. thx in advance..
Changing to NEMA23 or a brushless motor should be easy. Changing to a higher ratio and larger diameter cycloidal gearbox would require a larger joint bearing and that gets heavy and $$$.
Love your dedication to perfecting 3d-printed cycloidal gear design. Having affordable actuators will be so useful for making robotics accessible for the masses. Also that fist is just perfect! Keep up the awesome work.
Thanks. I have a few more cycloidal designs to show.
Don't know how I missed this when you released it! Incredible engineering as always!
This was awesome, I loved the fist on the testing bar, lol. Great job on the cycloidal gearbox. :)
Super impressive - love the fist grip.
i've been using dowel pins as roller pins for the ring gear and to make my own bearings. It works well if you want to have a really cheap cycloidal joint but you will have to struggle with tolerances.
My printers are pretty bad on the tolerances/roundness. I have found Needle and ball bearings drastically increase the efficiency of my 3D printed gearboxes.
Paul, thanks for another great video. I love how your designs keep evolving and improving over time!
I have a few more cycloidal designs in the works.
Your channel is the best thing I found in TH-cam in the last few years
Awesome work Paul! You just keep refining and making more awesome and capable stuff… keep it up! (And props for open source :)
Open source is great.
Wow, that so smooth, lovely job Paul!
I'm very happy with it but It is not completely smooth. My 3D printers are a bit out of tolerance and are not exactly making square and round prints.
Your work is amazing! Keep doing this!
Thank you so much!
Clever and elegant, i love it!
Wow!
Just awesome how much torque it reliably produces! Your efforts are great and that it's open source and modular is awesome!
Cheers mate and enjoy the success!!
I've been gradually improving this over the past few months. I think a higher cycloidal ratio and lower timing belt ratio would be better. The tension on the timing belt is very high.
An epiphany I had recently regarding cycloidal design: With discs 180 degrees out of phase, the ring is distorted into an ellipse by the radial pressure. If you use 3 discs 120 degrees out of phase, or 4 discs 90 degrees out of phase, or even more discs equally spaced, the necessary ring stiffness goes down because the distance between pressure points is reduced. More vibration than the fully counterbalanced style you're using, but less than the common two disc style, and the ring is a large portion of the overall weight so this can save a lot.
If you ever machine one from metal, another large weight reduction can be had by replacing those big output bearings with thin strips and circles cut from teflon sheet. 3D prints may not be precise enough to use this, but you could give it a try.
Thanks for your knowledge. I use the larger bearing to keep the ring from distorting. I've just some thinner wall bearings to lighten the joint. I have found that 3 discs at 120deg imposes a twist between the upper and lower housings. At some point, I will move beyond PLA 3D printed parts but I'm happy with the performance right now.
@@PaulGouldRobotics I see! Very clever using the bearings to prevent inward deformation of the ring, which is what actually pinches the discs. And that explains why twisting/asymmetrical deformation happens with equally spaced discs, since the ring itself can be relatively flexible then. I assume skipping of teeth is the limiting factor with this design, due to outward deformation of the ring?
Another good place for plain bearings is the output rollers. I tried needle bearings first, then SF1 bushings pressed inside brass tube to cover the seam, and now use PTFE tube pressed inside brass or steel tube. High load capacity, small diameter, silent, cheap, and only slightly higher friction than needle bearings. Though even with a tubing cutter it is fairly time consuming cutting and deburring tubes to precise length. And not all PTFE tubes are concentric, so it can be trouble finding a good supplier.
You could use this style for the ring pins too. I tried metal-on-metal steel tube sleeves over steel pins, but instead of the tube spinning over the pin, the pin itself spins along with the tube. So now I just use steel pins alone, which has the added advantage that the pins can be shorter length since they don't have to have their ends stick above/below the discs. Though in your case with the bearings functioning as ring stiffeners, having the pins supported by their ends right next to the bearings is an advantage.
fantastic, as usual!
Thanks
WONDERFUL!
That hand grip is hysterical. Love it. How's the backlash with this joint? Also is it backdrivable?
There is currently about 1 degree of backlash but it is increasing the more I run it. It is just back drivable with 2kg of weight.
Waw, very impressive ! 10 Kg is massive !
The limitation is the timing belt. I'm thinking of a dual timing belt reduction with GT2 then HTD5
hi is this considered compound gearing? one from Pulley and another from cycloidal ? without the pulley will it able to produce the same torque?
Your super fan
Nice Design nicely built!
Have you tried a design with a strain wave gearing that makes use of compliant plastic? I have seen some fairly small but effective designs and it would be interesting to see a comparison with a cycloidal drive of similar size and/or mass.
I have made a strain wave gearbox using TPEG. It was not that strong or efficient. Maybe a higher ratio would be better. Check out my Instagram page with the yellow and blue 3D printed parts.
Can you please link to your work?
Hi Paul,
I found it interesting that you replicated an arm to show how this joint can lift a weight. It made me think how inefficient lifting weights is for stressing a real human joint, and how a mechanical joint would give various advantages.
Question for you about possible practical implications. I've seen first hand some of the rehab practices of people recovering from limb surgery (specifically ACL knee rehab). A big thing during rehab is two fold. Gradually increasing range of motion in the limb, and also applying force (both flexion and extension) at different ranges/angles. It occurs to me that a robotic arm is probably very well suited to applying these controlled forces.
Have you ever thought about building this joint into a type of "workout tool"?
I have started a thin joint for an exoskeleton (See my Instagram). I need to contact a biomedical company to learn more about what is required.
In my experience (I'm in Sydney), rehab for ACL reconstruction is pretty much run by the physio profession. I saw you're a Perth Wildcats fan, you might be well placed to reach out to the medical team at the Wildcats (and/or an AFL team) and talk to them about it. Those two sporting codes would be bleeding edge in terms of ACL treatment.
For an example of a brace that is pretty much already an exoskeleton, look at the appliances made by Bauerfeind, like the Sectec Genu. They are German engineered, very lightweight and strong, it would be a very good starting point to add controlled motion to.
I see you are a building a robot to workout for you.
Smart!
It can do way more reps than me.
maestro
I'm curious if you've tried split-ring planetary gearboxes or Wolfrom planetaries before (I believe they are similar)?
They seem like an interesting alternative to cycloidal gearboxes.
I've tried many different gearbox types (Magnetic, capstan, harmonic, cable, twist) but not planetary. Seams like other makers have this covered.
I really like the belt before the cyclodial drive!!! What do you think the efficiency might be?
I'm guessing about 70%. It is all rolling contacts.
What specific motor are you using?
Amazing !!!
Thanks
Nice design. I have few questions I hope you can help me with. Why did you choose to go with belt system instead of making the cyclodial drive 2 stage with more reduction to acheive the same? Also there's a noticeable shake when the arm is being lifted. It's amazing as it is for a plastic gearbox to handle this weight but I just want to know this is a drawback in the cyclodial gearbox design or something else is contributing to this vibration. This will really help me in designing my gearbox so thank you very much in advance
I've tried a dual stage cycloidal and I couldn't solve the wobbling, see a previous video. Cycloidal work well because the forces are spread out over many teeth. The wobbling is caused by the center eccentric shaft bending. It is not the strongest part.
@PaulGouldRobotics thank you very much Paul. Exactly the answer I needed. Does that mean if it was made from metal parts the wobble issue could be solved without needing second stage? Still one part I hope you can answer, I'm not sure if your answer was referring to this but why did you favour the belt as a second stage instead of 2- stage cyclodial?
Just a though... have you tried a eccentrically cycloidal planetary gearbox?
In theory you could make a planet gear with essentially "one tooth"
As a boon the bearing could sit in the planet carrier so you could print the planet gear as a simple dumb plastic piece.
This could drastically cut down on bearings.
Question: would it be possible to replace the 3 discs with 1, and print some counterweights to avoid wobble instead?
Three keeps forces on the eccentric shaft balanced.
I like it! What's about backlash?
It is about 1deg and gradually getting more as I run it for longer.
hey Paul,
what size bushings are you using ??
Check the Thinkiverse link.
When is the next video
Hi Paul, I would like to use this Elbow joint in a real prosthesis elbow design, but i need the size to be smaller. Large Dial(White) in your picture , in my case needs to be less than 30mm Dia meter, Motor should be smaller and For Arm portion should be smaller, can you please help? i need help with resizing, this will be a child prosthetic, so about 3/4/5kg/30cm is more than enough.. Please help.. thx in advance..
90 degree slip discs, simple gear ratio change by linear radius position move, axial transmission gear box axle drive, for bikes and bicycles
talk to the hand, because face is not listening :)
flexing, are you :)
What are "90 degree slip discs"? Google search didn't help.
@@PaulGouldRobotics two discs 90 degrees angle to each other, one running on other, changing the radius it runs on, making cvt
@@PaulGouldRobotics or just 90 degree joint high gear ratio tooth, but no cvt in that
How well does your design scale up?
Changing to NEMA23 or a brushless motor should be easy. Changing to a higher ratio and larger diameter cycloidal gearbox would require a larger joint bearing and that gets heavy and $$$.
Nice. Ill need your files! XD
How much did it cost to make this amazing technology? Can you sell it to me, I'm based in Rsa
It is open source, you can make it yourself. Cost about US$50. www.thingiverse.com/thing:5595805
Is it still backdrivable?
Yes. It can hold about 2kg unpowered, any more than that and it falls down.
How much weight is that?
The joint weighs 1.3kg and it is lifting 10Kg @ 300mm.