Dang, I was hoping for better backlash. That's some impressive torque though. And I didn't know you were using a printed belt. That means you can use different tooth size on each half to get proper meshing with both rings, improving efficiency for quadrupeds and other high power robots. Plus there may be a solution to the backlash problem: add some friction to the joint, after the gearbox. Vibration within the backlash range will cause ripples in the print, but if the joint has enough friction to hold its position until the actuator forces it in the opposite direction, then you can use the backlash compensation feature in Marlin and be all set.
This backlash test was after breaking it. When it wasn't broken, there was almost no backslash unless there was a large amount of force applied. I had an idea for pre-tensioning the drive by using a torsional spring.
Note: A fresh harmonic drive which hasn't been through the torque tests has almost no backlash. The backlash that is present could be solved with the use of a torsional spring.
I also thought that constraining the top of the bearing bolts on the wave generator would be worth a try. Also, can you use larger diameter bearings on the wave generator (in theory could go up to half belt inner diameter)? This would result, depending on the inner perimeter of the belt, in a more gradual meshing/unmeshing process and potentially increase the contact area, perhaps with some benefits for tooth wear/damage.
Hey just wanted to tell you that I didn't finish the video till the end but plastic is still flexible so clamping it like this can cause this problem specially with harmonic gears the clamping applies force from top and bottom and this can cause flex and more space in both left and right side, when the tow in bearings and belt reaches the sides the inner belt can't touch out one properly so it skips the tooth in these positions and causes this cracking sound. Just imagine squeezing a rubber ball with one hand
@11:04 might want to check that math unless you've invented antigravity lol Whats the advantage of using a custom printed belt as opposed to something commercially available? i.e. a gates belt flipped inside out. Seems like wider belts is a good easy change to increase the contact area. You could also consider adding the rest of the bearings on the wave generator or increasing the bearing size so that the belt pulls away at a shallower angle (or more smaller bearings). The belt no doubt can be flexed more/at sharper angles than a metal flex-spline of similar size so you can probably use a larger contact patch than possible with a metal spline. For keeping the weight low you could use a combination of bearings and PTFE "skis" which could much more precisely control the shape of the contact area and keep the belt from pulling away in the space between bearings. It could also be the bearings flexing away from the belt as the back-driving force causes the belt to ride up on the gears so increasing the length of the wafer screw or just tightening the tolerances on the captive nut so it's a press-fit may reduce the flex. Considering the depth of the strain wave gear I don't see a reason not to. Love the project. Hope to see it continue to improve!
I found that drive belts did exhibit belt skipping, probably due to their compressibility (rubber) and would expect nylon teeth to elastically deform much less under similar loads. So this seems like a better way than timing/drive belts for a high tooth-load application.
Compared to some other printed HD's I've seen, top-notch design! The fact that Nicholas Seward commented back at you, now that's impressive yet...
Dang, I was hoping for better backlash. That's some impressive torque though. And I didn't know you were using a printed belt. That means you can use different tooth size on each half to get proper meshing with both rings, improving efficiency for quadrupeds and other high power robots.
Plus there may be a solution to the backlash problem: add some friction to the joint, after the gearbox. Vibration within the backlash range will cause ripples in the print, but if the joint has enough friction to hold its position until the actuator forces it in the opposite direction, then you can use the backlash compensation feature in Marlin and be all set.
This backlash test was after breaking it. When it wasn't broken, there was almost no backslash unless there was a large amount of force applied. I had an idea for pre-tensioning the drive by using a torsional spring.
Note: A fresh harmonic drive which hasn't been through the torque tests has almost no backlash. The backlash that is present could be solved with the use of a torsional spring.
Really, really neat and compact design. Stunning
Good work, thorough testing. Shared to our Discord for all those into robots.
thanks for sharing this, wondering how you came up with the tooth profile? Any links or resources you can share?
So the blue parts were made on an FFF printer, what printer did you use? Also for the SLS part what if you tried high strength SLA parts?
Could you please also tell us the measurements in metric units?
Thanks a lot for your great video and greeting from germany. :D
American mind works in inch and yard.
Excellent! You might also add in a disk to register the tops of the wave generator bearings. I suspect at high torque they start deflecting.
I also thought that constraining the top of the bearing bolts on the wave generator would be worth a try. Also, can you use larger diameter bearings on the wave generator (in theory could go up to half belt inner diameter)? This would result, depending on the inner perimeter of the belt, in a more gradual meshing/unmeshing process and potentially increase the contact area, perhaps with some benefits for tooth wear/damage.
The ring bearing helps a lot, good design.
what are the number for the big belt
Hey just wanted to tell you that I didn't finish the video till the end but plastic is still flexible so clamping it like this can cause this problem specially with harmonic gears the clamping applies force from top and bottom and this can cause flex and more space in both left and right side, when the tow in bearings and belt reaches the sides the inner belt can't touch out one properly so it skips the tooth in these positions and causes this cracking sound. Just imagine squeezing a rubber ball with one hand
Also sorry for my poor english. And this might not be the problem but add it to your notes.
Check your camera angle.
What is material of flex gear .Or belt ?
SLS Nylon
Practice keeping things you show us in the frame.
Hanzhen harmonic drive gear , industrial robot arm gear reducer, over 30 years experience
You confused output and flexspline. The flexspline is, what you refer to as "the belt".
@11:04 might want to check that math unless you've invented antigravity lol
Whats the advantage of using a custom printed belt as opposed to something commercially available? i.e. a gates belt flipped inside out.
Seems like wider belts is a good easy change to increase the contact area. You could also consider adding the rest of the bearings on the wave generator or increasing the bearing size so that the belt pulls away at a shallower angle (or more smaller bearings). The belt no doubt can be flexed more/at sharper angles than a metal flex-spline of similar size so you can probably use a larger contact patch than possible with a metal spline. For keeping the weight low you could use a combination of bearings and PTFE "skis" which could much more precisely control the shape of the contact area and keep the belt from pulling away in the space between bearings. It could also be the bearings flexing away from the belt as the back-driving force causes the belt to ride up on the gears so increasing the length of the wafer screw or just tightening the tolerances on the captive nut so it's a press-fit may reduce the flex. Considering the depth of the strain wave gear I don't see a reason not to.
Love the project. Hope to see it continue to improve!
I found that drive belts did exhibit belt skipping, probably due to their compressibility (rubber) and would expect nylon teeth to elastically deform much less under similar loads. So this seems like a better way than timing/drive belts for a high tooth-load application.
@@simonmerrett interesting! Id seen your videos and a few others using belts which is why I asked.