Capstan Drive NEMA 17 Stepper Timing Belt

[@EnUsUserScreenname]

Not gonna lie, this is beyond clever and simply beautiful.

[@PaulGouldRobotics]

Prototype 3 is more aesthetic

[@moses5407]

​@@PaulGouldRobotics do you have a link to that video? I'm very interested in this capstan design

[@egeoeris]

My hat's off to the compactness of the system. Brilliant call to embed the stepper into the elbow.

[@thorgraum1462]

now, this is what i am looking for. looks like a reliable strong gear system with no backlash. beautiful!

[@dsg123456789]

I really love the designs that you make! Thank you for making all of the interesting documentation about building various gearboxes.

[@PaulGouldRobotics]

Thanks, There will be a few more video on Capstan in the next few weeks.

[@ErosNicolau]

Show me anything with real zero backlash and I will fall in love with it! Superb design! I believe Skyentific has also come up with an arm joint using wire, combining a capstan and a multi stage block and tackle system.

[@PaulGouldRobotics]

It's not perfect but it is my favorite gearbox right now but with the stepper replaced with a BLDC. Skyentific does some great work.

[@dekutree64]

Excellent work! Looks like a great candidate for robot arms and quadrupeds.

[@PaulGouldRobotics]

NEMA17/23 for robot arms, BLDC for quads

[@kurtnelle]

This is very elegant robotic drive solution. Nicely done.

[@gedr7664]

very very nice, thank you very much for your wonderful work! you are a master crafter of electromechanical design

[@PaulGouldRobotics]

Thanks

[@TaYpworm]

This is a neat mechanism. Thanks for sharing!

[@JonathanBeri]

Such an elegant and compact design!

[@PaulGouldRobotics]

Thanks

[@jeffschroeder4805]

The small capstan that broke wouldn't be too difficult to reproduce in metal (if you know someone with access to a lathe), different friction coefficient but MUCH stronger. Very nice design.

[@FilterYT]

This is nice Paul, thanks for sharing!

[@ethanroland6770]

Wow that's a really nice and practical solution

[@hysoon6167]

Man u are a genius

[@chuongnguyen4980]

This is very clever and compact design. Make it alsmot full rotation. Thanks you for sharing it. Best of Best 3D printed reduction

[@OMGWTFBBQSHEEP]

Such a clever design!

[@PaulGouldRobotics]

Thanks

[@minercraftal]

Simple and clever for sure! 😂 impressive idea!

[@besenyeim]

I love cable drives. That tensioning method needs some refinement. :) I wonder if belt can be substituted with cable too. Multiple turns need more cable, but for smaller forces thinner wire is usable.

[@PaulGouldRobotics]

I've changed the tensioning for both the timing belt and wire many times. It is getting better but far from ideal. I was thinking of two sets of capstans but it might be better to make one with a higher ratio.

[@arttupakarinen7642]

this is really good, but i need linear input/output with the same design, the possibility to relax. I think with a few adjustements i could get that out you your system. Thank you very much!

[@simonmerrett]

Great developments Paul. I can see plenty of mass optimisation opportunities with this - carbon fibre plates and tube, dyneema / kevlar cables etc. Wondering how to get another degree of reduction using additional capstan stages to allow lighter motors and remove dependence on timing belt.

[@PaulGouldRobotics]

I would like to try Dyneema. Yes, it is too heavy for what it is but I want to keep the parts 3D printed so anyone can make it. Two capstans in series is interesting, maybe one with a higher ratio is better. The brushless version is way more powerful and way lighter and have a lower ratio.

[@eelcohoogendoorn8044]

Not sure about the series capstan. I love capstans, but one of their unique features is a tradeoff between range of motion and compactness. The low-torque-fast-moving side of it isnt going to be particularly elegant I fear; I doubt it compares well with a continuous belt. Dyneema seems to be getting a lot of traction recently; ive seen a bunch of papers coming out proving some pretty impressive real world performance.

[@Sigmatechnica]

Another good one! :D

[@localhawk1]

Wow, great design in all aspects. I would love to see an aluminum version.

[@PaulGouldRobotics]

It shouldn't be too hard to make the capstan input shaft from Al with a Lathe. It's just thread cutting a different profile and stopping at the right location.

[@IainMace]

very nice design. interesting

[@PaulGouldRobotics]

Thanks

[@thereminoautomation4936]

Greath compliments!

[@PaulGouldRobotics]

Thanks

[@christophermagorian9255]

Love the video! Got a link to the BOM and Step files?

[@MrZomhad]

Awesome design! Are you planning on sharing the files for this and your other capstans?

[@PaulGouldRobotics]

I will open source all of my designs. I generally don't put the first protypte up as it has so many error/problems. I can email a design if you want. Chat to me at hackaday, instagram or discord (gouldpa#9015)

[@RichardRamos]

Great Job

[@chrissmith82771]

I was wondering if using a tuning peg type tensioner system would work better it would let you get more even tension and make it easier to keep it.

[@PaulGouldRobotics]

Good idea. Will just need a good way to attach that to the plastic housing.

[@amoose136]

Saw a 2016 paper by Mazumdar et al. from Sandia national lab that talked about the benefits of using synthetic fibre cables over steel cables in situations where the winding ratio of curve radius to cord diameter is lower than 18:1. I’d want to try some dyneema or another UHMWPE cord on something like this.

[@PaulGouldRobotics]

I’m using 1mm 49 strand wire so it is pretty flexible and easy to work with. I will try dyneema in the future. I’m concerned that softer cable will not be round anymore and form a flat rectangleI. I’ll check out that paper. Thanks.

[@scienceraven1200]

very nice research

[@micultimy91]

just throw an AS5600 encoder breakout board on the back of the stepper motor, and BAAM! you got yourself a robot arm axis! Good luck!

[@bernhardtrzil6315]

What do you think about to scale this to a smaller NEMA Steppers? Would this possible?

[@PaulGouldRobotics]

Yes, it can scale down. If you had a round stepper motor then the output of the capstan could be the stepper housing.

[@swannschilling474]

Nice, makes me want to try it too... :)

[@PaulGouldRobotics]

It is easy to print but the wiring is a bit tricky.

[@arttupakarinen7642]

So to make that linear i just need to make 50-100mm tube with pulleys to output one cable

[@jpjay1584]

what size is the big bearing? 10mm x10mm? x OD and ID

[@ICGedye]

Hi. I don’t know if you find this of any use, but would it not be better to make the ‘grooves’ out of pre-wound and glued wire? Just thinking about to wear stress concentrations subjected to the plastic reel. Not knowing a lot about 3D printing, this may be a baseless idea. Lovely design you have.

[@PaulGouldRobotics]

This is a good idea. I'll see if it is possible in this design. Spreading the forces over a larger area is good when 3d printed plastic is used.

[@ErosNicolau]

Hmm, large bearings here too ;)

[@projected.5943]

Beauty😍😍😍

[@arnauaguilar8329]

Did this ever got open-sourced?

[@SidekickElements]

Great design, but i don't think the steel cable will withstand multiple cycles of being wrapped around that thin rod (speaking from experience). Dyneema cord should have low enough stretch and enough tensile strength (especially since you have 4 cables).

[@ifnotnowwhen6914]

Very similar to power window mechanisms on some cars. I'm curious about heat dissipation for the motor with it being enclosed with no air flow. Still an interesting idea/video.

[@PaulGouldRobotics]

I didn't know they were cable driven. I'll do some more research. The motor gets hot and it doesn't go well with PLA. It's only powered on for a few minutes. I'm thinking of adding a heatsink and/or cooling holes or a mini fan.

[@ifnotnowwhen6914]

@@PaulGouldRobotics For the record I wasn't trying to bash you. I'm pretty new to cnc & 3d printing. I noticed the nema17's on my cnc get seriously hot if I'm doing a long cut so I asked. And most ideas evolve before being finished. Thank you for sharing.

[@PaulGouldRobotics]

@@ifnotnowwhen6914 Your comments were positive. Heating is a problem that needs solving. I value the feedback from viewers, this is why I like open source designs. The brushless version of this is much better, cooler, stronger and faster.

[@h3Xh3Xh3X]

It looks like a cable tensioning part would help with assembly

[@PaulGouldRobotics]

Yes. Tensioning is the hardest part with these designs. I am still working on improvements.

[@LazarusAugment]

turn the outer housing into a pully housing and anchor point x at base of structure

[@jwstolk]

Would this have a lower friction that a 3D printed herringbone gear? It seems there would be plenty of area of a wide herringbone gear, and would be much simpler to assemble.

[@PaulGouldRobotics]

3D printed gears are not that efficient, can not handle high torque and have backlash which increases over time. Timing belt and capstan drives are highly efficient even 3D printed ones. They can be tensioned for zero backlash and it can be tighten over time. The assembly of capstans is not easy and I don't know how to make it better right now. Single capstans will be easier that dual capstans.

[@LanceThumping]

Could you replace the 3 separate wires with one wire since they loop back on each other anyways at the anchor point?

[@PaulGouldRobotics]

It could be possible but it might make assembly difficult.

[@alanearl9354]

Hi Paul, awesome design, well done!. Do you have the mechanical parts in STEP form as I use F360?

[@PaulGouldRobotics]

Message me through hackaday or Instagram and I can send it to you.

[@chuongnguyen4980]

Can you please share the cad file? thank you.

[@haenselundgretel654]

This is absolutely awesome. What do you need the Backdrive-ability for? I'd consider it mostly a negative side effect. Cheers mate! You're absolutely great!

[@PaulGouldRobotics]

High Efficiency/low friction is needed for Backdrive-ability. I actually need to add a brake to this design in the future. Robot arms don't need to backdrive but quadrupeds do.

[@prlancast]

@@PaulGouldRobotics Some robot arms can be pushed around by hand to learn tasks. Only when they can be back driven though. I was wondering about cooling of the motors. Do they get hot? I guess if they were made of metal they would act as a heat sync.

[@PaulGouldRobotics]

@@prlancast The motors are warm to tough at 12W. They are running open loop. I will add an encoder and run closed loop so hopefully this drops the average wattage. It would be possible to add heat dissipating pads and/or water cooling.

[@prlancast]

@@PaulGouldRobotics Great work btw there are not many actuators that are so small containing the motor in the centre!

[@rok1475]

I suggest reworking the design to incorporate crimp on threaded cable plugs to tension the cables. Pulling the cables by hand and securing them with screws to plastic will not get you consistent results and will not scale up.

[@DerSolinski]

Noice. That belt skipping should be easily fixable with a wider belt, higher tension or bigger drive pully. Did you do a reliability test yet? Since I worry that it will run in and deform very quickly.

[@PaulGouldRobotics]

I hate timing belts in general. I fixed the belt problem in prototype 2. I have redesigned the whole thing and have just finished making prototype 3. So I haven’t done much testing. All contacts are rolling so it should wear to much. Point forces on plastics maybe a problem.

[@besenyeim]

@@PaulGouldRobotics similarly, I hate stepper motors in general. But I have to admit, they have their place, and not completely useless. Timing belts are good too, but not universal.

[@williammische2409]

Well done! Do you put some kind of coating on the tips of the wires?

[@PaulGouldRobotics]

Yes, my daughter’s mail polish. It stops the wires from fraying.

[@deplorablesecuritydevices]

Thanks very much!

[@SSSuzd]

трос из-за малого радиуса изгиба очень быстро выйдет из строя. Наверно лучше шестерни использовать. Или, как вариант, использовать еще тонкие тросы, но увеличить их количество. Возможно, тогда можно вообще использовать другой материал в место стальных тросов.

[@SSSuzd]

а, так вы уже делали так...😅Нашел видео 👍

[@oracid]

Thanks for your videos. Are you sure this is a capstan? Maybe a block and tackle. en.wikipedia.org/wiki/Block_and_tackle Or simply a personal and original achievement.

[@PaulGouldRobotics]

Google search for "Capstan robot". The term is not new (and possibly not correct) but it being used to describe this mechanism. There has been some discussion on discord about this terminology. Also check out my friends hackaday.io/project/176726-stanley-the-capstan-based-quadruped

[@phenioxrj]

@@PaulGouldRobotics Which Discord do you hang out on?

[@PaulGouldRobotics]

@@phenioxrj Odrive, mjbots, quadrupedalism, tinymovr

[@TheRainHarvester]

5:00 did you start they middle cable by going thru the axle? Or is that going under the axle? Great mechanism!

[@TheRainHarvester]

8:48 what is the clicking, and how did you fix it?

[@PaulGouldRobotics]

It goes though the axle.

[@PaulGouldRobotics]

@@TheRainHarvester It is the timing belt rubbing on the pulley because the tensioning mechanism is not aligned perfectly.

[@PaulGouldRobotics]

Check the cable layout at 2:00.

[@ohbaewan3147]

Hi Paul, You often say this is your favorite drive technique, any chance we will see a release of the models on Thingiverse? Love your work!

[@PaulGouldRobotics]

I haven't got a chance to put in on Thingiverse yet. Here are the raw files. github.com/gouldpa/Odd-Mech-Assemblies/blob/main/stepper_timer_capstan_REV3.zip

[@ohbaewan3147]

@@PaulGouldRobotics Thank you! Your work is inspiring!

[@buder5116]

i wonder what would happen with a coolstep & stall guard activated LUL + the anti loose steep thing i found (S42B closed loop stepper motors) i am intrigued you must try it ( the S42B closed loop stepper motors) while its overloaded with weight !!

[@PaulGouldRobotics]

That is exactly what I’m working on now plus an extra absolute encoder for joint angle. I’ve already made a brushless version.

[@buder5116]

@@PaulGouldRobotics nice :D

[@tuliopireso]

My friend...Congrats! What plastic are you using, PLA , ABS or PETG?

[@PaulGouldRobotics]

The cheapest PLA Plus I can find.

[@ray-charc3131]

Are you sure they (the ropes and the spinning rod) won't slip over time? Anyway, it is really great!

[@PaulGouldRobotics]

The wire is attached to the spinning rod. It is technically not a capstan (friction drive), more like a winch. Capstan is the term being used in the hobby robots area.

[@ray-charc3131]

Understand, thanks.

[@ray-charc3131]

By the way, will the power requiref to rolling up the thick wire ropes on the spinning rod is large and the angle travel is not so precised are the issues?

[@PaulGouldRobotics]

@@ray-charc3131 The wire is 1mm dia with 49 strands. It is extremely flexible. I'm not sure about angle travel but the joint angle is measured by an magnetic absolute encoder.

[@Праведныймиротворец]

Hey, how do you prevent the steel wire strands become loose?

[@PaulGouldRobotics]

I use nail polish to keep the stands together at the ends. It works ok. This design doesn't have a good tensioning mechanism. The next prototypes are much better. New video soon. Instagram has more details.

[@Праведныймиротворец]

@@PaulGouldRobotics ok thank you so much 👍

[@ДмитрийКоростелев-х5н]

Лайк из России. 👍

[@DIYMaker76]

Hi guy, what's name for bearing or dimension? I need bearing for my projects. Pls and thank you! 😁

[@PaulGouldRobotics]

Bearings are 6810RS and 6811RS

[@DIYMaker76]

@@PaulGouldRobotics thanks 😍

[@docstrench948]

Hola, tu invento está genial pero hasta que no podáis eliminar el golpe de arranque y el de parada no tenéis futuro en la robótica

[@tadashi_hamada]

Hello Paul, i was unable to find any theory or research behind capstan drives. Which sources is your design based on?

[@PaulGouldRobotics]

kzbin.info/door/oKZJmC6Bzh7UDMV29pZTLQ hackaday.io/project/176726-stanley-the-capstan-based-quadruped google search "capstan robot"

[@dawmro]

Give me a place on which to stand, and I will move the earth.

[@BHARGAV_GAJJAR]

Cool maybe you can build one of my designs

[@PaulGouldRobotics]

Please send a link or chat at Instagram/Hackaday.

[@lukasskymuh5910]

What is the reason for 4 wires? Would it work with 2?

[@PaulGouldRobotics]

It would work with 2 wires. I like 4 wires because it balances the load left and right, it spreads the load over a larger area and it is different. It does make assembly better is some places and harder in others.

[@agzt29]

Hi, I like your designs, thanks for doing this. Just a question, what type of bearings are them? where do you buy them? I want to do something like this, check my channel if you can, I am just starting. Thanks again

[@PaulGouldRobotics]

Search "6809 bearing" on aliexpress or google. I use 6808, 6809,6810 for the larger bearings. And 6704-2RS ball bearing 27x20x4 or 6701 for the medium size ones. I checked out your channel last week. You need bearings :-) and more rigid joints. Also check out the smart servos.

[@agzt29]

@@PaulGouldRobotics Thanks a lot!!! yeah I need bearings XD my joints sucks haha I just checked the smart servos, looks like they are perfect for the SCARAs Thanks a lot again!

[@владвладов-ы1ь]

Хрень так как двигатель постоянно под нагрузкой, червячный привод нужен

[@PaulGouldRobotics]

Worm drives are very in-efficient, have huge backlash and are generally not backdriveable.