Backlash is complex to measure. It has zero backlash with no load. With load, the springyness in the belts/3D printed parts give about 5mm @ 500m. This is why I like Capstan drives better. There is no springyness.
essentially they work with ropes and winds. check his channel! great stuff!👌 btw, a startup called haddington dynamics uses encoder on the output shaft and calculates force using the springyness of the belts. 🤯
@@YOUnoobGER ruclips.net/video/ENMZsPwCUcA/видео.html Technically not capstan drives as they rely on friction but it is the term that is being used in the robotics industry.
@@PaulGouldRobotics they also use 3d printed encoders and claim to utilize the sloppy slit quality to fingerprint each of the slits for absolute positioning. plus, then read each slit with an anlog sensor that allows them to read like a million (iirc) microsteps per slit. downside is that they need an 800$ fpga to do all those calculations in real time for their 5-dof robot... this info is from ~2018 when met them at a fair. not sure what changed since then, but you can find them on youtube :) Are you interested in "rotary" delta printers (as in delta robot, so without the linear rails and belts)? I was planning to build one with print speed as the main focus and found your channel when researching drives. Not sure if you saw my comment under your top 10 transmission video...?
Thanks. I get annoyed about the cables floating around, so doing cable management at the start helps. The cross section of the channels are trapezoidal.
Since you need more torque, you'll end up with pretty hot motors due to almost no air flowing alongside your motor. But still - like always - great! Thanks a lot for sharing. You're awesome!
The motor gets to 45-50degC. I usually limit the stepper to 15Watts but I wanted to test the strength of the timing belts. I thought about water cooling.
Clean and well designed as usual, I like the belt tension devices especially the sliding one. For positional feedback could you have a small gear on one arm and a rack meshing on the other , a magnet on the gear means you can use the encoder chip previously used. This has no load so there's no worries about torque slipping.Just a thought but I'm pretty sure you have a better idea. Thanks for sharing !
Cheers. I'm thinking about a donut shaped diametric magnet around the shaft of the stepper motor. Then 3 or 4 hall effect sensors around outside with a uC running some CORDIC (sine/cos to angle) software. This will give me absolute motor position. I could also use this for the joint position. I don't want to make the actuator wider.
Your designs are always so clean, really love that. This is just a proof of concept I know that but do you have already plans on how you could integrate a encoder? The only place that comes to mind is on the drive shaft for the second stage but that would still have the possibility of skipping. Even if that is very beefed up.
Thanks, I would like to add a a motor encoder and joint encoder with a custom controller. Or I could just skip that and use a brushless motor and the controller I have already got. The second stage has a really beefed up tensioning mechanism (go to 5:48). The bearings press the belt onto the pulley . It's not skipping any time soon. The first stage's GT2 still needs some work.
Your prints are so clean. Do you have an orientation that allows you to avoid supports yet achieve good inserts for nuts and bearings? Or other tricks?
Many other tricks. My printers are just Ender3s and are not tuned well. I have to file surfaces sometimes. I design with support in mind. Holes are drilled out and M3 nuts are slightly melted in place.
Cycloidal gearboxes are stronger and I haven't skipped a tooth yet. So for NEMA17 cycloidal are better because you can get a higher ratio in a smaller size. For a Brushless motor, I currently prefer timing belt and Capstan drives.
thats good. would be nice to make it more rigid. perhaps it was just the thing you had it mounted to seemed there was some wobble and backlash when moving fast?
Thanks a lot Paul for this amazing design, I will use model and modify it to make a great robotic arm, I was thinking to Design a great actuator but since i have my Exams etc. i didnt come to it, but your model sounds so promissing for my application :) keep the good work
Really nice design Paul. But i think it is not efficient for price-performance. Bearing and belts only cost $30 (in aliex) already i see there are planetary gearboxes for $25 with metal gear.
I agree, the bearing count/cost is too high. Belt cost is high too. I may try 3D printed bearings next. I am trying to achieve zero backlash which is not possible with planetary gearboxes.
Hey Paul, great design as always! What is the name of these type of bearings, i would like to make Scara robot joint and i just can't find these bearings
Backlash is complex to measure. It has zero backlash with no load. With load, the springyness in the belts/3D printed parts give about 5mm @ 500m. This is why I like Capstan drives better. There is no springyness.
Sorry to ask, but what are Capstan drives?
essentially they work with ropes and winds. check his channel! great stuff!👌
btw, a startup called haddington dynamics uses encoder on the output shaft and calculates force using the springyness of the belts. 🤯
@@YOUnoobGER ruclips.net/video/ENMZsPwCUcA/видео.html Technically not capstan drives as they rely on friction but it is the term that is being used in the robotics industry.
@@Mosenhosen I didn't know that did that. Cool
@@PaulGouldRobotics they also use 3d printed encoders and claim to utilize the sloppy slit quality to fingerprint each of the slits for absolute positioning. plus, then read each slit with an anlog sensor that allows them to read like a million (iirc) microsteps per slit. downside is that they need an 800$ fpga to do all those calculations in real time for their 5-dof robot... this info is from ~2018 when met them at a fair. not sure what changed since then, but you can find them on youtube :)
Are you interested in "rotary" delta printers (as in delta robot, so without the linear rails and belts)? I was planning to build one with print speed as the main focus and found your channel when researching drives. Not sure if you saw my comment under your top 10 transmission video...?
I love the details which are so transferable to other designs - such as printing in interference fit cable management channels. Brilliant.
Thanks. I get annoyed about the cables floating around, so doing cable management at the start helps. The cross section of the channels are trapezoidal.
Paul, your consistently impressive output is always inspiring. A beautiful actuator as always!
Yours as well, mate!
I love how compact and thought through this design is. Great job!
I can't see how this could be made any smaller with off the shelf parts. Superb!
I would like a higher ratio. Maybe another reduction stage.
Since you need more torque, you'll end up with pretty hot motors due to almost no air flowing alongside your motor. But still - like always - great! Thanks a lot for sharing. You're awesome!
The motor gets to 45-50degC. I usually limit the stepper to 15Watts but I wanted to test the strength of the timing belts. I thought about water cooling.
Can't wait to see the brushless version ! Great video ! I love your work !
Thanks.
Clean and well designed as usual, I like the belt tension devices especially the sliding one.
For positional feedback could you have a small gear on one arm and a rack meshing on the other , a magnet on the gear means you can use the encoder chip previously used. This has no load so there's no worries about torque slipping.Just a thought but I'm pretty sure you have a better idea. Thanks for sharing !
Cheers. I'm thinking about a donut shaped diametric magnet around the shaft of the stepper motor. Then 3 or 4 hall effect sensors around outside with a uC running some CORDIC (sine/cos to angle) software. This will give me absolute motor position. I could also use this for the joint position. I don't want to make the actuator wider.
Amazing video! Thank you for sharing this
i love the part when you thought about the wires =)
Your designs are always so clean, really love that.
This is just a proof of concept I know that but do you have already plans on how you could integrate a encoder?
The only place that comes to mind is on the drive shaft for the second stage but that would still have the possibility of skipping.
Even if that is very beefed up.
Thanks, I would like to add a a motor encoder and joint encoder with a custom controller. Or I could just skip that and use a brushless motor and the controller I have already got.
The second stage has a really beefed up tensioning mechanism (go to 5:48). The bearings press the belt onto the pulley . It's not skipping any time soon. The first stage's GT2 still needs some work.
Your prints are so clean. Do you have an orientation that allows you to avoid supports yet achieve good inserts for nuts and bearings? Or other tricks?
Many other tricks. My printers are just Ender3s and are not tuned well. I have to file surfaces sometimes. I design with support in mind. Holes are drilled out and M3 nuts are slightly melted in place.
I like your compact design.👍
Thanks for sharing Paul :) Sorry to ask: why different belts in different stages? Best Regards,
Nice! When the belts are tight what is the angle of backlash?
I'll do some testing and put up the result.
What about using S42B closed loop stepper drivers (from AliExpress) you can easily put in place instead of using normal stepper drivers?
I might buy one of them for testing or get one and rewrite the software.
Makes me want to try timing belts now!! 😊
Compared to cycloidal gearbox, which one is better?
Cycloidal gearboxes are stronger and I haven't skipped a tooth yet. So for NEMA17 cycloidal are better because you can get a higher ratio in a smaller size. For a Brushless motor, I currently prefer timing belt and Capstan drives.
@@PaulGouldRobotics Thank you for your kind reply. Can I ask you one more question? Why don't you prefer cycloidal gearbox on the BLDC motor?
@@handonghee1 Capstan drives are zero backlash, more efficient, cheaper (less bearings) and smaller. Ratios aren't as high.
@@PaulGouldRobotics Thanks for the easy and accurate explanation!
@@handonghee1 Capstans have a higher stiffness, less springiness.
thats good. would be nice to make it more rigid. perhaps it was just the thing you had it mounted to seemed there was some wobble and backlash when moving fast?
The mounting was not good. Even thought the belts are very tight they are still a bit springy.
awesome
Thanks a lot Paul for this amazing design, I will use model and modify it to make a great robotic arm, I was thinking to Design a great actuator but since i have my Exams etc. i didnt come to it, but your model sounds so promissing for my application :) keep the good work
Thankyou
awesome! pd: which is the size / code of those bearings?
All files and details are on the thingiverse page.
Really nice design Paul. But i think it is not efficient for price-performance. Bearing and belts only cost $30 (in aliex) already i see there are planetary gearboxes for $25 with metal gear.
I agree, the bearing count/cost is too high. Belt cost is high too. I may try 3D printed bearings next.
I am trying to achieve zero backlash which is not possible with planetary gearboxes.
Hey Paul, you know this is world-class, don't you?
You are so kind to say that. There are so many improvements I’d like to do.
Finally inner peace 😌 no nagging sound. What about bldc one !!
I'm not good at making the videos. The BLDC video is a few weeks away.
@@PaulGouldRobotics well you are good at making awesome gear boxes ! 😀
@@shashankshekhar6549 Many thanks.
Nice fitness for everything!😂
Hey Paul, great design as always! What is the name of these type of bearings, i would like to make Scara robot joint and i just can't find these bearings
Aliexpress
6811 Bearing
6809 Bearing
@@PaulGouldRobotics Thank you friend!!
Superb! 👍
Great job!
Where do you buy the parts like gears, box and axels?
These came mostly from and old ender 3. Ali express is where I shop. Just have to buy in advance.
God jop!
I need help me for about siz belt timing
what's that device on the right at 0:11
It is slider pot connected to an Arduino for the speed control of a stepper motor driver.