IMPRESSIVE!!! since like a month ago there were like 15+ new videos on harmonic drives, i think you started a new movement! :P I recently bought an electronic screw driver from amazon and its been a blast! You should buy 1 too!!
hey dude, hope you're doing well! just came back to this to see whether you'd posted anything new. would be lovely to hear from you again. i'm very excited to see how this awesome project turns out!
Great stuff :-) I found your channel some years ago when looking into SCARA 3d printers, and thought it a shame that you weren't uploading regularly. Thrilled to see you have started uploading, thanks for keeping it up! The dynamic loads on the inner actuator of a SCARA are several times higher than on the outer one, so it doesn't make much sense to use the same unit for both as in your design. At least double them up on the inner pivot.
Although currently abandoned due to lack of motivation, the original idea was to have 2 proximal arms sandwiching the distal, so 2 drives for shoulder and 1 for elbow.
Reducing the reduction ratio and making the tooth shape more sine-wave like would both increase back drivability. Using a belt, tho, in theory makes it inherently harder to backdrive (but easier to skip) because the belt can deform between bearings so force applied at the outermost bearing may be lost to deformation before "reaching" the next. With a metal flex spline and/or a much tighter arrangment of bearings, force applied at the major axis, where teeth are fully engaged, should be much more distributed onto the rest of the gear teeth and strain wave gear bearings which gives better leverage to backdrive (in an industrial drive the flex spline often acts as the outer race of the bearing meaning the spacing comes down to the size of the ball bearings). To explain what I mean by leverage... The component of force acting on the strain wave gear, both magnitude and angle, is dependent on the tooth profile and tooth location relative to the major axis of the swg. At the point of full tooth engagement, the force is mostly compressive on the major axis of the swg and the horizontal force is minimal. As the force is distributed further from major axis, the horizontal force component applied to the major axis increases, but the tooth engagement decreases too so overall force is less. But its really the lesser engaged teeth that produce the backdriving force. So with a sparse interface (i.e. just a few bearings) between the flex spline and swg, the horizontal component just deforms the belt/flex spline rather than transfering it to the swg. Also metal harmonic drives would typically have grease/lube. Without lubrication, the backdriving forces will create a lot of friction between teeth and less of that force will be converted to a component acting on the flex spline/swg.
Solid iteration. Would like to see the torque tests re-done and see how much filling out the rest of the wave generator made. I worry about the strain on the output where it connects to z. I think making the top arm a triangle which would allow you to increase the distance between bearings, reducing flex in the printed parts. It could be lighter weight as well since, depending on the angle, the force on the hypotenuse is increasingly tension rather than flex. You could use relatively small width carbon tube. Removing the lower arm also means your min z height is lower and you can mount the hotend closer to the center of the wrist, reducing twists on high accel moves. A triangle also reduces any bouncing/oscillations on z moves. The main downside is that your overall height will always be a bit higher than max build height. But you can make the Z extremely sturdy with some thick alu extrusions and linear rails with 2 carriages (one for top bearing, one for bottom). Hope that makes sense. I can make a quick cad sketch if youre interested. Also while the placement of the stepper in the elbow is sleek. On the shoulder(?) joint it is adding to the rotational mass. Maybe minor given the radius, but mounting it stationary to the z gantry will allow you to use heavier, more powerful steppers if needed. Also, if youre set on using aluminum. A full square tube will definitely perform better than c channel. To save weight you could drill out pretty large holes in the top & bottom faces. You dont have to worry about twisting too much so traditional t or v slot extrusion is probably overkill on the arms.
Actually the twisting of the arms is very important when it comes to the vertical stability of the arm. If the two arms are perpendicular to eachother, a moment/torque can be easily induced in the distal arm which causes massive discrepancy in the z-height. This was the main issue with my previous design.
@@KoalDesigns Ah you are right. I was mainly thinking about reducing sag at full extension. I am really curious to see how much deformation there is in the 3d printed parts with such long arms and how the c channel holds up.
@@KoalDesigns Actually the twisting could also be mitigated by a triangulated top arm. But instead of a single "hypotenuse" tube, use 2 which meet at the upper z bearing and spread to 2 or 3 times the arm width at the elbow. 10-15mm square tube would probably be sufficient and 2 of them would probably still weigh less than the second alu arm. also, unless im thinking about it wrong, the bottom alu arm doesnt prevent these torsional forces very much unless you tie its z bearing into the shoulder drive. without, when the top arm twists, the bottom will twist AND rotate away. Im thinking as if it was all in petg or pla so in actuality, with high performance filaments, it could be negligible.
@@ComeToLyfe How about screwing some 1/16" aluminum plates to the sides of the proximal arm? Essentially making it into a large rectangular tube, which has much higher torsional stiffness than C channel.
@@dekutree64 If I'm understanding you correctly, you mean plates connecting the top and bottom arms? I think part of his thought process of having them separated is so they can be clamped really tight on the elbow joint. Having plates on the sides would prevent that/increase the stress on the printed parts where they join with the c-channel. Its also more weight.
swear you triggered a storm of new harmonic drive videos. have you seen James Burtons cycloidal drive version? I think he might be onto something. video: ruclips.net/video/KodT7lQyYkk/видео.html
Beautiful design! Can't wait to see it on the arm.
IMPRESSIVE!!! since like a month ago there were like 15+ new videos on harmonic drives, i think you started a new movement! :P I recently bought an electronic screw driver from amazon and its been a blast! You should buy 1 too!!
Amazing ! It is so relaxing to sit and design in cad, I need to get back to it!
This looks like a really good step forward. Please do share when you can test the full assembly.
Awesome how could I have possibly missed this!! Hoping to see more of it soon!!
Great work, nice video. Thanks for sharing
hey dude, hope you're doing well! just came back to this to see whether you'd posted anything new. would be lovely to hear from you again. i'm very excited to see how this awesome project turns out!
Great stuff :-) I found your channel some years ago when looking into SCARA 3d printers, and thought it a shame that you weren't uploading regularly. Thrilled to see you have started uploading, thanks for keeping it up!
The dynamic loads on the inner actuator of a SCARA are several times higher than on the outer one, so it doesn't make much sense to use the same unit for both as in your design. At least double them up on the inner pivot.
Although currently abandoned due to lack of motivation, the original idea was to have 2 proximal arms sandwiching the distal, so 2 drives for shoulder and 1 for elbow.
I admire your work
I heard that original Harmonic Drive™ reductors are actually backdriveable.
Any way to achieve this with diy design?
Reducing the reduction ratio and making the tooth shape more sine-wave like would both increase back drivability. Using a belt, tho, in theory makes it inherently harder to backdrive (but easier to skip) because the belt can deform between bearings so force applied at the outermost bearing may be lost to deformation before "reaching" the next. With a metal flex spline and/or a much tighter arrangment of bearings, force applied at the major axis, where teeth are fully engaged, should be much more distributed onto the rest of the gear teeth and strain wave gear bearings which gives better leverage to backdrive (in an industrial drive the flex spline often acts as the outer race of the bearing meaning the spacing comes down to the size of the ball bearings). To explain what I mean by leverage... The component of force acting on the strain wave gear, both magnitude and angle, is dependent on the tooth profile and tooth location relative to the major axis of the swg. At the point of full tooth engagement, the force is mostly compressive on the major axis of the swg and the horizontal force is minimal. As the force is distributed further from major axis, the horizontal force component applied to the major axis increases, but the tooth engagement decreases too so overall force is less. But its really the lesser engaged teeth that produce the backdriving force. So with a sparse interface (i.e. just a few bearings) between the flex spline and swg, the horizontal component just deforms the belt/flex spline rather than transfering it to the swg.
Also metal harmonic drives would typically have grease/lube. Without lubrication, the backdriving forces will create a lot of friction between teeth and less of that force will be converted to a component acting on the flex spline/swg.
I don't understand how output ring is attached to the flexi spline.
Any progress I’d love to see more!
Solid iteration. Would like to see the torque tests re-done and see how much filling out the rest of the wave generator made.
I worry about the strain on the output where it connects to z. I think making the top arm a triangle which would allow you to increase the distance between bearings, reducing flex in the printed parts. It could be lighter weight as well since, depending on the angle, the force on the hypotenuse is increasingly tension rather than flex. You could use relatively small width carbon tube. Removing the lower arm also means your min z height is lower and you can mount the hotend closer to the center of the wrist, reducing twists on high accel moves. A triangle also reduces any bouncing/oscillations on z moves. The main downside is that your overall height will always be a bit higher than max build height. But you can make the Z extremely sturdy with some thick alu extrusions and linear rails with 2 carriages (one for top bearing, one for bottom). Hope that makes sense. I can make a quick cad sketch if youre interested.
Also while the placement of the stepper in the elbow is sleek. On the shoulder(?) joint it is adding to the rotational mass. Maybe minor given the radius, but mounting it stationary to the z gantry will allow you to use heavier, more powerful steppers if needed.
Also, if youre set on using aluminum. A full square tube will definitely perform better than c channel. To save weight you could drill out pretty large holes in the top & bottom faces. You dont have to worry about twisting too much so traditional t or v slot extrusion is probably overkill on the arms.
Actually the twisting of the arms is very important when it comes to the vertical stability of the arm. If the two arms are perpendicular to eachother, a moment/torque can be easily induced in the distal arm which causes massive discrepancy in the z-height. This was the main issue with my previous design.
@@KoalDesigns Ah you are right. I was mainly thinking about reducing sag at full extension. I am really curious to see how much deformation there is in the 3d printed parts with such long arms and how the c channel holds up.
@@KoalDesigns Actually the twisting could also be mitigated by a triangulated top arm. But instead of a single "hypotenuse" tube, use 2 which meet at the upper z bearing and spread to 2 or 3 times the arm width at the elbow. 10-15mm square tube would probably be sufficient and 2 of them would probably still weigh less than the second alu arm.
also, unless im thinking about it wrong, the bottom alu arm doesnt prevent these torsional forces very much unless you tie its z bearing into the shoulder drive. without, when the top arm twists, the bottom will twist AND rotate away. Im thinking as if it was all in petg or pla so in actuality, with high performance filaments, it could be negligible.
@@ComeToLyfe How about screwing some 1/16" aluminum plates to the sides of the proximal arm? Essentially making it into a large rectangular tube, which has much higher torsional stiffness than C channel.
@@dekutree64 If I'm understanding you correctly, you mean plates connecting the top and bottom arms? I think part of his thought process of having them separated is so they can be clamped really tight on the elbow joint. Having plates on the sides would prevent that/increase the stress on the printed parts where they join with the c-channel. Its also more weight.
Did you ever do a torque test?
Hanzhen harmonic drive gear , robot joint,
strain wave reducer, over 30 years experience
Wow! Do you have plans to share your great designs?
swear you triggered a storm of new harmonic drive videos. have you seen James Burtons cycloidal drive version? I think he might be onto something. video: ruclips.net/video/KodT7lQyYkk/видео.html
You're saying "aah, ugh, uhm"s 99 times during this video, that's one every 4.7 seconds. Work on it, please.