Thanks for sharing!! This is great...crazy reducer!! I am currently looking into cyclodial reducers a lot, because of the compact size, and little backlash! Its cool to see how you are improving the design with each iteration!! :)
I get some pretty good gear meshing on my laser here in Auckland. I do it a couple of ways, stack 2 layers of 3mm abs together or you can flip thicker plastic and the kerf matches if the machine is set up well.
downgearing bldc motors is cutting edge for robotics field imo. however, rather than gears, I think pulley systems to downgear are the best route. quieter, can take up very low volume in total, can be made VERY small, can be VERY cheap to make. doing a lot of experimentation and design improvements on that area now. I recommend you look into this approach as an alternative.
This is impressive ! You do an incredible great job ! I'm looking for high gearbox ratio (200:1) for a star tracker project. I tried 3D printed planetary gear, but there were too much friction and my stepper motor (NEMA17) can't make it turn well. I found cycloidal gearbox to replace the previous gearbox, but my output is not linear and do back and forward. It seems your design with dual stage are really more stable. I will try to make a new version of my cycloidal gearbox with a dual stage.
Thanks. You can turn the input shaft quite easily at low torque. I've only made one prototype of this design. It needs a few improvements. I was trying to make it as small as possible. The plastic is quite springy at this small size. I suggest making it bigger for your star tracker, using bigger bearings. Good luck. Please show me your finished design
Paul Gould I printed my prototype of the cycloidal gearbox with two stages for a total reduction on 195:1. But after some tests, the output is not linear, there is a wavy movement that put the prototype totally unusable. So I will try with a 3D printed harmonic drive with a 50:1 ratio and a planetary gear 4:1 on second stage. The harmonic drive seems more linear than cycloid gearbox. I will try this new prototype and I comeback for the results 😉
I've had a few side projects, as I get distracted easily. I've made some servo quads to test the Inverse Kinematics and working on a cheaper/easier bldc controller (Arduino/ESP-32). The leg design mostly complete. Just have to print 3 more.
Hi Paul, thanks for putting this up with so much detail RE assembly and design. I have been waiting for this version to be posted. I am just wondering why did you get rid of the pins/linear bearings on the outside of the cycloid gear. These were present on your previous prototype- uploaded in April? Too expensive, heavy or some other reason?
I wanted to make an easy, compact design. It would need over 70 bearings and lots of bolts. This is just a prototype and would be much better if made out of nylon.
@@PaulGouldRobotics Yeah I thought that it had a lot of parts when I started shopping for them. I was thinking of using this design or similar scaled for an exoskeleton. I previously mucked around with trying to 3d print a harmonic drive, but plastic 3d printed gears aren't very practical. Rolling friction vs sliding etc. Thanks for the reply.
@@immersion-fx6650 immerse the 3d printed gear in grease and lube, lol. As long as the printed parts are not heated to the glass transform temp ( 60C for PLA), it wouldn't deform, for small and agile robot (low payload) it wouldn't be such a problem. The only problem in this technology, in my opinion, is dimension. FDM printers can't achieve good enough dimension accuracy for these type of part to work flawlessly. Resin printers can be the solution, but then the price of the 1 part is gonna be rocketed up
I think we would all like to see how it compares to previous gearbox joints you made. in terms of torque, speed, wear, stability, friction, accuracy, good points bad points.
This is the first prototype of this design. It has too many issues to make a fair comparison at this point. The simple cycloidal robot knee is the best for my robot quad, it is quickest to make, nicely backdrivable and medium torque. I haven't ran any of them too much to determine wear and stability. Accuracy will never that high with 3D printed parts.
@@PaulGouldRobotics Thanks for the video and the reply. I do think that your work can have future value in motorized joints in cost-effective robots but at some point the 3d-printed parts may need replacement by something stronger when it is ready to scale up.
@@richardbloemenkamp8532 3D printing is great for development and research of new ideas. Most of my design are not really for practical use, just for proof of concept.
I'm not sure. There is no slop, just springy backlash. ABS plastic is naturally flexible. Is there a stiffer /better 3D filament?. A lower gear ratio would also help.
Nice! Too bad it's not as useful as some of your other designs. I'm looking into designing my own cycloidal gearbox, do you have any tips/things you found that could be useful?
I have not given up on this design yet. I think a lower reduction ratio and larger diameter will help. What are you making? Start with torque, work out the speed you need. This will help you determine gearbox ratio and the motor. Then choose outer bearings and every thing will work out from there. I suggest a bigger dia/ low KV motor and run it at low current. You can chat to me at hackaday or discord @gouldpa
@@PaulGouldRobotics I want to make a roughly human sized bipedal robot, and these reducers seem like a good low cost solution. I'm pretty sure a torque of ~200 NM would be needed for that. How do I contact you at discord?
Every cycloidal I've seen an tested has the same issue as the one seen in your current one: a wobbly / sinusoidal / non-linear output. Which, from what I managed to simulate so far, doesn't come from a geometric flaw in the design, but from a cyclic difference in teeth engagement pressure angle, resulting in a cyclic torque stiffness. This, to me, is much more disturbing / damaging than any eccentricity vibration, as it effectively impairs the system's angular precision under load. Do you have any ideas on how to mitigate THIS effect?
I don't fully understand what you are describing. I've only noticed once per revolution problems when something was made wrong. I have 6 of the "simple cycloidal knee" designs and only one of them wobbles. I will replace it at some point.
@@PaulGouldRobotics I've been noticing these wobbles ever since I've been experimenting with the hypocycloidal model. And I always wondered if this was a matter of a bad mathematical model or just a matter of weak construction. So now I finally gathered enough momentum to model a properly parametric model and to test its geometric properties. I have two conclusions: 1. the mathematical model is SPOT ON. The output results in a linear relation to the input. So I stand corrected there. But... 2. In terms of torsional rigidity, I found out that things are not as linear: whenever a pin gets to either a peak or a valley dead center, you have minimal torsional rigidity, which increases in a sine function while transiting towards the next peak/valley position, only to get back to a minimal value again. So for instance if you have a gearbox with 5 lobes and 6 pins, you have 10 such minimal torsional rigidity points and 10 maximal ones. This fades out to nothing with a construction that is both correct and well proportioned to handle the designed load, but rears its ugly face whenever the gearbox is designed a bit weaker than needed.
Hey, will you be using Gartt ML5010 motors for this project again? Since I am planning to build a stationary robot arm, I would like to rather use a power supply than lipo batteries! Could you give me a recommendation of which one would be good to use for this project? :)
The easiest solution would be to use a NEMA23 stepper for the lower joints and NEMA17-long for the upper joints. I would recommend this gearbox th-cam.com/video/pizl7i-uB68/w-d-xo.html. The triple dual stage is still in early development and has some work to complete.
I'll be making a bigger one of these. I would like to fix the current issues. I'm not sure what motor I'll be using, I don't really have a practical application for this project.
@@PaulGouldRobotics thats the one I already printed, still waiting for some parts to do the assembly...I also want to try brushless motors at some point, using the odrive...or hopefully your controller!!
Great content; but it would have been even better without the music. Being able to hear what happens during assembly and moving things around is a big part of groking what is going on for me.
Thanks. I've had this suggestion a few times before. My main focus is on the engineering and the videos are just a quick edit. I will try to have better production values.
It is an old(modified with covers abs) Geeetech iprusa clone. It is a very bad printer. I now have an ender 3 but it only prints pla. Bascally the printer doesn't matter. Just adjust tolerances in the drawings or printer settings.
Yes and no and it depends on your application. I see this being used as a hobby robot arm. Fast movements are not really required. This is a proof of concept and I think that a reduction ratio of 70:1 would be more useful. Typical RC/smart servos have ~300:1 gearboxes.
@@PaulGouldRoboticsyes it formats is solidworks and if you share with us I am very happy for this format is not important which if you want only its draw format its enough
@@yusuftasdemir6722I can create a step file for you. I am trying to find an equation for a dual stage cycloidal gearbox but I can not find one. If N1-N2 =1 then the gear ratio is 1: (N1xN2). But is N1=N2 then there is not reduction ratio. I think that the ratio is 1 : ((N1*N2)/(N1-N2)). Please check this. If true then N1=21 and N2=25 gives exactly 1:100
Quick answer is no. The min reduction ratio is about 80:1. It is not designed for continuous operation. The design is a prototype and more for robotics.
@@ramadavince6611 I use the Gartt ML5010 (850W) from gogorc.com (cheapest I can find them). I haven't seen this type of motor used on scooters/skateboards. They are usually 5050 in size at 2-3kWatts.
@@PaulGouldRobotics 50** I know. But my problem is I live in one of India. The only thing I can get is this 5010 motor in national online shopping 😟. And small once like 2210 etc . Can get like 6384 ect from AliExpress,bangood... But they are very risky on shipping in with this Corona thing. My guess is if a 250 watts motor like 'my 1016 motor' can run ebike and scooter. A 800 watts motor may run may get the job done 😁? So.. I'm just asking as you have more experience with the motor😃
This one has a too higher ratio for that. It could be redesigned to be less than 70:1 and I think that it would be good. Also look at the simple stage th-cam.com/video/XWp5dkeHgTY/w-d-xo.html
Thanks for sharing!! This is great...crazy reducer!! I am currently looking into cyclodial reducers a lot, because of the compact size, and little backlash! Its cool to see how you are improving the design with each iteration!! :)
Thanks
I think this can also be done a hobby CNC mill. POM is often used for gears. Great design
I sure it could be made on a hobby cnc mill.
Very inspiring work. Have you considered laser cutting some parts for faster fab and better tolerances? I am curious as to where are you located
Thanks. I'm not sure how to deal with the 4 degree laser kerf on standard laser cutters. I'm located in Darch, Perth, Western Australia.
I get some pretty good gear meshing on my laser here in Auckland. I do it a couple of ways, stack 2 layers of 3mm abs together or you can flip thicker plastic and the kerf matches if the machine is set up well.
That's cool what sort of tolerances can a well set up cuter achieve?
around 50 microns but the smallest inside radius is around 200 microns
mindblowing. seriously.
Thanks. It is my favorite design yet. If only I had a practical use for it.
that is such a cool design!
downgearing bldc motors is cutting edge for robotics field imo. however, rather than gears, I think pulley systems to downgear are the best route. quieter, can take up very low volume in total, can be made VERY small, can be VERY cheap to make. doing a lot of experimentation and design improvements on that area now. I recommend you look into this approach as an alternative.
Check out my other videos and on Instagram. Many other gearbox designs there. Each has their place. I do this mainly to learn, no real end game.
Great design , i have printed the parts waiting for the bearings , please provide list of hardware . Thank you
The Thingiverse page should have everything you need.
Thank you,i will update you
This is impressive ! You do an incredible great job !
I'm looking for high gearbox ratio (200:1) for a star tracker project. I tried 3D printed planetary gear, but there were too much friction and my stepper motor (NEMA17) can't make it turn well. I found cycloidal gearbox to replace the previous gearbox, but my output is not linear and do back and forward.
It seems your design with dual stage are really more stable. I will try to make a new version of my cycloidal gearbox with a dual stage.
Thanks. You can turn the input shaft quite easily at low torque. I've only made one prototype of this design. It needs a few improvements. I was trying to make it as small as possible. The plastic is quite springy at this small size. I suggest making it bigger for your star tracker, using bigger bearings. Good luck. Please show me your finished design
Paul Gould I printed my prototype of the cycloidal gearbox with two stages for a total reduction on 195:1. But after some tests, the output is not linear, there is a wavy movement that put the prototype totally unusable. So I will try with a 3D printed harmonic drive with a 50:1 ratio and a planetary gear 4:1 on second stage. The harmonic drive seems more linear than cycloid gearbox. I will try this new prototype and I comeback for the results 😉
@@edwinripaud7724 I had this same problem. The maths is looks correct, so I can't work out where the non-linearity is coming from.
Nice design, what material should these be printed out of?
This one is made of ABS. Nylon would be the best.
Where the heck did you hide??? A mazing!
Very cool. Hows your dog project coming? Have you considered a resin printer for higher tolerances parts?
I've had a few side projects, as I get distracted easily. I've made some servo quads to test the Inverse Kinematics and working on a cheaper/easier bldc controller (Arduino/ESP-32). The leg design mostly complete. Just have to print 3 more.
Hi Paul, thanks for putting this up with so much detail RE assembly and design. I have been waiting for this version to be posted. I am just wondering why did you get rid of the pins/linear bearings on the outside of the cycloid gear. These were present on your previous prototype- uploaded in April? Too expensive, heavy or some other reason?
I wanted to make an easy, compact design. It would need over 70 bearings and lots of bolts. This is just a prototype and would be much better if made out of nylon.
@@PaulGouldRobotics Yeah I thought that it had a lot of parts when I started shopping for them. I was thinking of using this design or similar scaled for an exoskeleton. I previously mucked around with trying to 3d print a harmonic drive, but plastic 3d printed gears aren't very practical. Rolling friction vs sliding etc. Thanks for the reply.
@@immersion-fx6650 immerse the 3d printed gear in grease and lube, lol. As long as the printed parts are not heated to the glass transform temp ( 60C for PLA), it wouldn't deform, for small and agile robot (low payload) it wouldn't be such a problem. The only problem in this technology, in my opinion, is dimension. FDM printers can't achieve good enough dimension accuracy for these type of part to work flawlessly. Resin printers can be the solution, but then the price of the 1 part is gonna be rocketed up
I think we would all like to see how it compares to previous gearbox joints you made. in terms of torque, speed, wear, stability, friction, accuracy, good points bad points.
This is the first prototype of this design. It has too many issues to make a fair comparison at this point. The simple cycloidal robot knee is the best for my robot quad, it is quickest to make, nicely backdrivable and medium torque. I haven't ran any of them too much to determine wear and stability. Accuracy will never that high with 3D printed parts.
@@PaulGouldRobotics Thanks for the video and the reply. I do think that your work can have future value in motorized joints in cost-effective robots but at some point the 3d-printed parts may need replacement by something stronger when it is ready to scale up.
@@richardbloemenkamp8532 3D printing is great for development and research of new ideas. Most of my design are not really for practical use, just for proof of concept.
Great work, if this was made with better tolerances would the backlash be zero, or would we have to switch materials?
I'm not sure. There is no slop, just springy backlash. ABS plastic is naturally flexible. Is there a stiffer /better 3D filament?. A lower gear ratio would also help.
Thank you for sharing! I might use it for my project.
This is the first prototype of a proof of concept design. There are still some improvements needed.
@@PaulGouldRobotics I have started my own design for my needs in terms of reduction ratio. Can you address the areas that needs improvement?
Sure, always happy to help. You can chat to me at hackaday.io or discord
If you need any bearings showing in the video, please visit our web: www.bobibearing.com
Nice!
Too bad it's not as useful as some of your other designs.
I'm looking into designing my own cycloidal gearbox, do you have any tips/things you found that could be useful?
I have not given up on this design yet. I think a lower reduction ratio and larger diameter will help.
What are you making? Start with torque, work out the speed you need. This will help you determine gearbox ratio and the motor. Then choose outer bearings and every thing will work out from there. I suggest a bigger dia/ low KV motor and run it at low current. You can chat to me at hackaday or discord @gouldpa
@@PaulGouldRobotics I want to make a roughly human sized bipedal robot, and these reducers seem like a good low cost solution. I'm pretty sure a torque of ~200 NM would be needed for that.
How do I contact you at discord?
@@Skythedragon discord.gg/k3ZZ3mS
@@PaulGouldRobotics Thanks
Super Cool Paul!
Can you add the link to where you source your 6mm shafts to the thingiverse BOM?
mcjing.com.au/silver-steel-rod.html
@@PaulGouldRobotics Thanks a ton!
have you tried printing your parts on an SLA printer for much greater precision? would suggest using Elgoo Mars or Saturn
I have found SLA/resin parts very brittle and not very strong. I could do with the precision though.
Every cycloidal I've seen an tested has the same issue as the one seen in your current one: a wobbly / sinusoidal / non-linear output. Which, from what I managed to simulate so far, doesn't come from a geometric flaw in the design, but from a cyclic difference in teeth engagement pressure angle, resulting in a cyclic torque stiffness. This, to me, is much more disturbing / damaging than any eccentricity vibration, as it effectively impairs the system's angular precision under load. Do you have any ideas on how to mitigate THIS effect?
I don't fully understand what you are describing. I've only noticed once per revolution problems when something was made wrong. I have 6 of the "simple cycloidal knee" designs and only one of them wobbles. I will replace it at some point.
@@PaulGouldRobotics I've been noticing these wobbles ever since I've been experimenting with the hypocycloidal model. And I always wondered if this was a matter of a bad mathematical model or just a matter of weak construction. So now I finally gathered enough momentum to model a properly parametric model and to test its geometric properties.
I have two conclusions:
1. the mathematical model is SPOT ON. The output results in a linear relation to the input. So I stand corrected there. But...
2. In terms of torsional rigidity, I found out that things are not as linear: whenever a pin gets to either a peak or a valley dead center, you have minimal torsional rigidity, which increases in a sine function while transiting towards the next peak/valley position, only to get back to a minimal value again.
So for instance if you have a gearbox with 5 lobes and 6 pins, you have 10 such minimal torsional rigidity points and 10 maximal ones. This fades out to nothing with a construction that is both correct and well proportioned to handle the designed load, but rears its ugly face whenever the gearbox is designed a bit weaker than needed.
Are you taking about a per tooth wobble or a per rev wobble?
@@PaulGouldRobotics per tooth
Wait... did you just offset the I/O from the gearbox so it is O O and not I -> O ?
That is clever mate, awesome design.
It is just mirrored.
It is a combination of the triple cycloidal and double dual stage. It is an iterative design process.
This is amazing! Now I feel a bit annoyed I didn't find this sooner, I already used a (really crappy) harmonic drive for my scara robot...
Neat design! Will you share the files somewhere?
You were too quick. www.thingiverse.com/thing:4412347 is up now.
@@PaulGouldRobotics awesome! Thanks.
@@uint16_t What are you planning on making with this?
@@PaulGouldRobotics , Hi it seems that the smaller rings are missing. And what are the dimensions of the screws ? Thank you
@@bravojul7351 Thanks for the feedback. I just updated the Thingiverse page with the little gear and the bolt info.
How do you generate your cycloidal discs? is it parametric or do you do something else
hackaday.io/project/157812/log/145872-mechanical-design-cycloidal-transmission
others have written some for fusion360
What kind of bearing is this sir?
Please see the Thingiverse page, it has additional information.
Hey, will you be using Gartt ML5010 motors for this project again? Since I am planning to build a stationary robot arm, I would like to rather use a power supply than lipo batteries! Could you give me a recommendation of which one would be good to use for this project? :)
The easiest solution would be to use a NEMA23 stepper for the lower joints and NEMA17-long for the upper joints. I would recommend this gearbox th-cam.com/video/pizl7i-uB68/w-d-xo.html. The triple dual stage is still in early development and has some work to complete.
I'll be making a bigger one of these. I would like to fix the current issues. I'm not sure what motor I'll be using, I don't really have a practical application for this project.
@@PaulGouldRobotics thats the one I already printed, still waiting for some parts to do the assembly...I also want to try brushless motors at some point, using the odrive...or hopefully your controller!!
Great content; but it would have been even better without the music. Being able to hear what happens during assembly and moving things around is a big part of groking what is going on for me.
Thanks. I've had this suggestion a few times before. My main focus is on the engineering and the videos are just a quick edit. I will try to have better production values.
Are you printing with abs?
Yes, just cheap abs.
would t be possible to make a multi stage gear reduction by increasing the diameter (instead of making it longer, make it fatter)
I have often though about this and tried drawings something up in CAD but nothing has worked yet. If you have any designs, I'd love to see them.
What kind of printer do you use?
It is an old(modified with covers abs) Geeetech iprusa clone. It is a very bad printer. I now have an ender 3 but it only prints pla. Bascally the printer doesn't matter. Just adjust tolerances in the drawings or printer settings.
isn't 240:1 a lot of reduction for fast movements?
Yes and no and it depends on your application. I see this being used as a hobby robot arm. Fast movements are not really required.
This is a proof of concept and I think that a reduction ratio of 70:1 would be more useful.
Typical RC/smart servos have ~300:1 gearboxes.
Is there any backlash in this design?
It is about 1 degree. But most of that is springyness in the 3d printed parts. Better tolerances will reduce that.
Hi do you have sldprt doc?
What is that format? I think that it is only for Solidworks, which I don't have. I can create some other formats if you like.
@@PaulGouldRoboticsyes it formats is solidworks and if you share with us I am very happy for this format is not important which if you want only its draw format its enough
I want to ask a question please can you tell me how can we setting gear rate for example you did 240/1 but O need 100/1 how can I set it
@@yusuftasdemir6722I can create a step file for you. I am trying to find an equation for a dual stage cycloidal gearbox but I can not find one. If N1-N2 =1 then the gear ratio is 1: (N1xN2). But is N1=N2 then there is not reduction ratio. I think that the ratio is
1 : ((N1*N2)/(N1-N2)). Please check this. If true then N1=21 and N2=25 gives exactly 1:100
@@PaulGouldRobotics thanx but I didnt understand your calculate
I found that it was not 240, it may be should be 15×17=255 : )
I have never seen an equation for this type of gearbox. Please link if you can find one.
Dude can I used this motor on my diy scooter 10" wheel at 20:1 or 12:1 gear ratio my weight is around kg50
Quick answer is no. The min reduction ratio is about 80:1. It is not designed for continuous operation. The design is a prototype and more for robotics.
@@PaulGouldRobotics ok. Thanks for saving my time and money 😊
@@PaulGouldRobotics o. Dude I mean the "motor you used in another video 5010😀". That motor.
Not this gear
@@ramadavince6611 I use the Gartt ML5010 (850W) from gogorc.com (cheapest I can find them). I haven't seen this type of motor used on scooters/skateboards. They are usually 5050 in size at 2-3kWatts.
@@PaulGouldRobotics 50** I know. But my problem is I live in one of India. The only thing I can get is this 5010 motor in national online shopping 😟. And small once like 2210 etc
. Can get like 6384 ect from AliExpress,bangood... But they are very risky on shipping in with this Corona thing.
My guess is if a 250 watts motor like 'my 1016 motor' can run ebike and scooter. A 800 watts motor may run may get the job done 😁? So.. I'm just asking as you have more experience with the motor😃
Great job. It's possible to build a six-leg robot with this actuator? (a machine like a ant, for example?)
This one has a too higher ratio for that. It could be redesigned to be less than 70:1 and I think that it would be good. Also look at the simple stage th-cam.com/video/XWp5dkeHgTY/w-d-xo.html
Thin section sealed ball bearings, visit link: www.bobibearing.com , we would like to give any help.
Cool