@@Skyentific i have mentioned you to him quite a few times. you are my two favourite channels. if he is too busy or successful to organise a collaboration with you, maybe you could do your own version of one of his projects, and treat it like a design challenge. then when both are done, you can make your robots compete. or, maybe you can design something useful to add to one of his robot platforms, which are all open source. i'm certain he could not say no to collaborating or at least featuring your work then. it would be very cool to see, you both make incredible things!
@@Skyentific Just buiid his really useful robot design with your creation of an 5 or 6 axis arm. And someone needs to tell James to rebuild the Base with a rectangular base and with Omni-Wheels.
This is an awesome design. Are you planning on making a larger one with your 9235 motor? Cycloidal gearboxes use a lot of bearings, I've been testing large 3D printed bearings with moderate success.
I tihink you've hit a homerun with this new design. The durability would probably improve with nylon printed gears. I'm interested to try this out with very hard resin like Siriya Tech Blue
When I wrote my bachelors thesis in mechatronics, I tried to develop a 3D-printable gearbox that; had high reduction, could handle high torque, was back-drivable and had high precision. To that end, I 'invented'(didn't know of any other at the time, though it turns out there were some similar ones) pretty much exactly this gearbox. In my attempts, I managed to make back-drivable gearboxes with up to 1:100 reduction that could handle a few Nm of torque. I reckon I could go way higher too, though I was very limited due to the pandemic. You mentioned that you wanted them to be back-drivable, I might have some tips if you're interested. It seems you have access to very high quality 3D-printing(at least better than what I did), so it would be really cool to see you try out some of my ideas/designs. Also, on a completely different note: I am currently taking a masters degree in cybernetics and robotics and would LOVE to see much more content on your control systems! Would be really interesting!
Wow you read my mind! I made pretty much the exact same thing a month or so ago. Although alot smaller and lower gear ratio! Makes it much more backdrivable. Good video! Nice work!
I would love to see a back-drivable version. I not sure but I think the most you can do to have a back-drivable gearbox, that is not too much static friction is less then 10 to1
Here you go ruclips.net/video/_VgXObcfXug/видео.html The carrier-driven variant has one less stage of tooth engagement, and thus higher efficiency (=more easily backdriveable). But it requires lots of bearings, and has higher planet RPM which may or may not be an issue depending on how fast you want to run it. Also you probably have to accept a little higher backlash to get backdriveability. Although there's one trick I read about in the comments on James Bruton's cycloid video that I really want to try, which is to use a dual-head printer to make gears where just the outer one or two walls are flex filament, and the rest is hard. The compliant surface allows tighter engagement without jamming, eliminating backlash. But it may still add enough friction to prevent backdriving. Would be great for robot arms though.
@@Nerdtronic By the look of it, high speed will probably heat up pretty quickly, but should be ok intermittently. Great for robot arms. If you want continuous high speed, print with a little more clearance.
yay, i was just hoping we would hear from you again soon! fantastic work, the cost considerations of this design are definitely a huge benefit but it also seems to work very well! of course it will wear eventually but that is the same for any 3d printed solution, and the beauty is that they can be replaced and replicated cheaply and easily when needed. and given the cost savings, one can spend a little more on a broad range of more expensive printable materials that i think can last a very decent time. great to see you! looking forward to the testing video! :-)
If printed from nylon, and given that such a diy robot arm transmission hardly makes the number of cycles as that of a city bus, I think wear is probably not a concern at all. Even PLA has been known to last for years in all kinds of gearing applications.
@@ericlotze7724 Metal casting is dimensionally challenging, and leaves a poor surface finish for gears. The slight give in plastic makes them work much more easily.
@@roryevans5032 Ah ok, well thanks for the insight, the metal idea is just an armchair one of mine. Would vibratory sanding help the finish or not even something like that?
@@ericlotze7724 to work properly metal gears need a smooth surface as they roll against one another. If the metal is rough they scrape and seize up, the only way to avoid that is make the tolerance lower, so there is more clearance, but that causes backlash. For the same reason you lose a lot of power to friction, and make lots of noise. Most gears are machined or ground, which gives a better surface quality than casting. Post processing is also difficult as you have to factor in the wear from the polishing step, so again to work you will generally have to lower tolerance and get backlash, noise, and low efficiency.
Output torque tries to tumble the planets. This is probably a large source of friction. GearDownForWhat used three or more gear layers to balance out this force, but a planet carrier and bearings could do the same. larger pressure angle gears are more efficient under radial loads. (i.e. 35 or even 45 degree pressure angle) Finally, I expect that heat will be the weakest part of this gearbox so running it immersed in soapy water, or using higher temperature materials will help a lot. (heat is why Nylon, Delrin, and Torlon are the plastics of choice for gears)
I agree, the triple-layer structure of his strongest gearbox is probably quite important; he has some tests also showing this has a big impact on torque-efficiency. For pancake formfactor like this I suppose its not so important, but if scaled lengthwise, the torsional and flexural rigidity of the planets will also be crucial to torsion efficiency and backdrivability I think. GearDownForWhat pressed in steel rods; that solves the bending; but itd be better if they were epoxied in I think, so torsion can be transmissed efficiently along the length of the planet, rather than bothering your layer lines with it. Flex in torsion will cause stress to be concentrated on the teeth where the rings meet.
Thank you, this was really interesting stuff! I was expecting it to be very noisy but was fairly dampened compared to what I was imagining. Also from the video it sounds like the grease dampened some of the rattling noise a fair bit.
Nice As for the backdrivability, I've seen something about bilateral drives, which also use a compound planetary setup, but they are backdrivable, somehow
I made a 3d printed Cyclodial gearbox and am still preferring it over planetary gearboxes, the low backlash and high torque comes at the price of those bearings and also those bearings add weight...but I guess it is still the best option for a high torque, low backlash and backdrivable robotic application! But I am very tempted to give your design a try, nice and smooth and looks like there is very little backlash!! 😎
The 3D printer does not 3D print perfectly. There are always some micro imperfection. With my orientation all this micro imperfection distributed evenly. This makes everything rotate smoothly.
@@Skyentific Ah yes, clever. Makes sense since you have such low play in the gear mesh, to prevent binding. Did you try installing the planets in a random orientation before orient them this particular way, or was it just your intuition?
hello are the two sun gears rotate in the same direction and same speed? I am trying to understand for the 2nd planetary, it seems to have input speed from the carrier and output at the ring, and it looks like an overdrive?
This design is incredibly similar to a design patented in the US as a "gear bearing drive" (US8016893B2). Really the only difference is the use of helical gears as opposed to rings to hold everything together.
Yes. They are available on Patreon. Search for post which was published at the same day as video. There are two posts: one public, and second for Patrons with files (stl and Fusion 360)
Приятно видеть человека, занимающегося любимым делом) Вопрос автору - вы не хотели бы замутить общий проект с Сергеем Дорошем? Он занимается волновыми редукторами и актуаторами.
I have some questions: 1) Which of the three gearbox types (Harmonic, Cycloidal, and Planetary) are used in modern commercial robot arms? 2) Which of those three gearbox types has the lowest backlash? (low backlash being a requirement for precise and repeatable robot-arm end-effector positioning) Thank you.
1. Harmonic 2. Harmonic But: do you want to stay in the past with slow robots, or go into the future with fast robots? If the future is your choice, you should use planetary :)
Is it a good idea to use this actuator as the base of a robot arm? unfortunately we cannot pass the wires inside for better cable management. any ideas?
What software did you use to model the gear set, and did you do the maths manually, it did you use a plugin/script? I would like to design better gesr sets like this
Very good. Now make a smaller one, make it backdriveable, and use dirt cheap brushless motor and electronics to control it. I find that to be quite hard.
You can print small, you can change the muzzle on a printer and calibrate it to get good miniature resolution. You can strenghten the material with carbon fiber. Or level up and you even can print metal. But this is not the mass market anymore ;)
This is possible. I made this exact gearbox a month or so ago but much smaller. Wish I had shared it now, someones beat me too it :( But yes you can print a much smaller one!
Why did you remove the magnetic screw and put in a non magnetic one instead? I just missed a bit the explanation why you did certain things, but found in the comments also the reason for the special screw alignment
Magnetic screw would decrease the range of the magnetic field from the magnet for encoder. Most probably it would work with magnetic screw too. But with non-magnetic screw the magnetic field should be a little more intense.
Can you reverse the setup so that I could have instead of 1:28 I could have 28:1 and post where to obtain the 3D printed files for the completed setup...????
The only problem I foresee with design is that the encoder is at the motor not at the other end which means you lose the absolute nature of the measurement of the encoder. I suppose that can be resolved with limit switches, but means you cant simply turn it on and know where it is.
Superb package, loving the printable / afforable /accessible and compatct aspects of your design. I'm thinking back drivablilty is important and 28:1 is too high for that, is it possible to make a 10 or 15:1 ratio? Would a carrier for the plants help keep gears "aligned" and lower friction help back drivability? looking forward to some more Scientific tests of strength, backlash, durability etc. Cheers
Nice. Is it PLA ? How hot does the motor get ? Seems you can add air flow for cheap too. Because it's not backdriveable, you can use zero current for holding so the motor shouldn't get hot, and when you spin it, adding a small propeller to the motor should help cool it down.
I always watch your videos with great interest. Typically a brushless motor requires a controller with a reverse function. All my searches for such a controller are unsuccessful. If you can please link. Thank you in advance.
Nice work, have you considered 3d printed sintered gears ? (You only need to print the teeth in metal, like a hollow shell, and then fill with epoxy for structure.
if gear box not back drive , the efficiency will be reduce . this mean your reduction will be reduce alot. i hope you can compare efficiency normal and shift teeth gear box . thank you so much
Something tells me your finger would get mangled if you accidentally put it in there 😂 I assume for actual work it's gonna get covered anyway, but for testing purposes, maybe you cut cut out a sheet of acrylic so it would be safe to handle while still see-through?
So, what you have there is what is commonly referred to as a "Wolfrom" (not WolfrAm) planetary gearbox. While they are attractive for their compactness, they are also notoriously inefficient and have fooled many, many, many robot designers. Wolfrom's look like they would be the perfect robot gearbox, but they are not, due to the phenomenon of "recirculating power". It's a subtle, but often fatal, flaw. The recirculation basically knocks the efficiency down to 50%, maybe 60% on a well-made metal gearbox. A plastic, or worse, printed, Wolfrom will likely be horrifically bad - likely below 50% and maybe 30-40%. The fact that you can't backdrive it is the first clue that something is wrong. A two-stage, 30:1 gearbox SHOULD be backdriveable unless it's a wormgear or similar drive with loads of sliding friction or otherwise poor efficiency. If I had a Swiss Franc for every Wolfrom gearbox made by an optimistic robot engineer, I could buy a Harmonic Drive. The insidious thing is that Wolfroms tend to spin well when NOT under load, but once you load them down, the efficiency drops through the floor. The planets also have a fierce twisting moment on them which tends to make them even less efficient, or worse, just bind up solid. Anyway, all that to say, please run some LOADED tests and see what kinds of efficiency you are seeing before investing in a full arm design. I predict that you will be disappointed. The only Wolfrom drive that I know of that isn't horrible is a traction-planetary version (Archimedes Drive) made by a Dutch company called IMSystems. They work, and they work well, but only because of the very high rolling efficiency of their traction elements - even higher than involute gears - that makes the math work out. Even so, I think Archimedes drives are still only the 80% efficient range - good, but not great. An equivalent (sequential) 2-stage planetary will have efficiency greater than 90% and maybe >95% for a really high quality gearbox. The poor backdriveability that you are seeing also means that the drive will have poor "transparency" so doing sophisticated torque-based control is a non-starter. If you don't have a mobile (battery powered) or force-controlled application, and you don't care about having to oversize motors and drives, then yes, they can work.
Thanks for the thoughtful critique; I was thinking of making something similar but this is food for thought, for sure. GearDownForWhat was also reporting rather low torque efficiencies in his test of his printed gearboxes. Still I just recently saw a video of a backdrivable 100:1 along these lines. The fundamental compromise with this design is that you will have fast moving contacts under high load at the ring-planet contact; which is something many other transmission like a stacked low ratio gearbox avoids. But as long as the losses there are optimized sufficiently well, it seems to me there are legitimate applications for this. High duty cycle applications are not ideal; but unless your robot arm is actually employed on an assembly line, its duty cycle can be very very low. To open up robotics to applications other than assembly lines, things like cost can be much more the deciding factor than even double digits of percentages of efficiency.
@@eelcohoogendoorn8044 Yes, that's exactly the issue. This style of drive is notorious for having "output loads at input speeds" - which is the worst possible combination. This leads to the low forward and backward efficiencies and poor transparency. I don't follow your duty cycle comment - efficiency is efficiency regardless of how much you use it. Low duty cycle will let you avoid having heat buildup that might melt your printed gearing (a secondary failure mode), but it won't change anything related to the primary "failure" of terrible efficiency. However, as I mentioned in my first post, if you don't care about transparency (i.e. you don't aspire to teleoperation or force control uses) or power consumption or having to have larger motors and drives (and batteries if you are mobile), then Wolfrom drives might still be useful to you. They DO work, they just don't work WELL.
@@bschena My point is, priorities are different depending on the application. A robot that is working 24/7 will most likely have a higher added value and thus higher budget than an actuator that needs to perform perhaps only a few seconds per day. While the Archimedes Drive is pretty cool, it sets you back 2.5k, for instance. If the choice was between this printed actuator or a sub-100$ with the efficiency of the archimedes drive, itd be an easy choice, but we dont face any such choice sadly. Check out this video: ruclips.net/video/vQqyGsqPwb0/видео.html you can find his papers too with some googling. Its just a more well optimized version of the same concept discussed here. Im guesstimating with some good design optimizations you could get a printed affordable high ratio version with 60% torque efficiency; and there are plenty of imaginable applications for that, even if kuka will scoff at it. GearDownForWhat also did a video on a printed backdrivable 20:1, btw.
А что насчёт температуры двигателя в разных режимах? Как известно PLA уже при 60 градусах начинает плыть. Или этот актуатор не предназначен для работы в режиме удержания положения?
Does the output sun gear contribute anything other than mechanical stability? I noticed it's teeth count is not included in the ratio calculation. would it work if it were just omitted?
great video! I was wondering if you could shed some light on how you calculated the teeth required? Let me clarify: at the moment I made an excel spreadsheet where I provide a sun gear with a set amount of teeth and diameter. Based on this I calculate the module by dividing the diameter by the amount of teeth. Now I can calculate the planet gear amount of teeth and the ring gear amount of teeth. Planet gear teeth = module/planet gear diameter ; ring gear teeth = module / ring gear diameter. Then I deduct the diameter of both the ring and planet gear by using the formula 1+rR/rS where rR = radius of ring gear and rS= radius sun gear. I have a feeling that this is where I make a mistake but this is also where I'm asking for your help please. If I do it my way I end up with numbers that end up in dividing by zero if I use your formula for calculating the ratio.
I'm brand new to robotics. I'm enjoying your videos. I'm hoping to build a small table top 6 axis arm with at least a 2 kg payload for deburring parts that come off my CNC mill. I really like this mechanism. What would it need to be back-drivable?
If you are just controlling position then backdriveability doesn't matter. However, if you want to do high quality, open-loop force control (by say controlling motor current) then you will be in trouble. Backdriveability, often referred to as "drive transparency" by roboticist is the Holy Gear for a well controlled, force feedback robot arm. Say, for example, you need to control contact force for, say, a deburring operation, then transparency/backdriveability will likely be required - unless you have a lot of compliance (springiness) between the robot arm and deburring device.
Awesome video! I'm planning on making a 6-axis robot arm and I'd like it to be able to 3d print. Are 0.25° backlash reductors from stepperonline precise enough? Or I'll have to use belts to reduce backlash?
Why do you give like to the video if you don't watch it still?
Me: Because I know that the video it's incredible. Regards Skyentific
Videos like this, are why I'm a Patron.
Thank you for been long time Patron!
Nice work man! I like how you did the planets. That removed the need to split the rings for assembly.
Thank you for watching my channel and for the inspiration!
This gearbox even can be the part of arm, that's amazing!
very good! i wish to see collaboration with you and james bruton!
I wish too! :)
@@Skyentific i have mentioned you to him quite a few times. you are my two favourite channels. if he is too busy or successful to organise a collaboration with you, maybe you could do your own version of one of his projects, and treat it like a design challenge. then when both are done, you can make your robots compete.
or, maybe you can design something useful to add to one of his robot platforms, which are all open source. i'm certain he could not say no to collaborating or at least featuring your work then. it would be very cool to see, you both make incredible things!
@@Skyentific Just buiid his really useful robot design with your creation of an 5 or 6 axis arm. And someone needs to tell James to rebuild the Base with a rectangular base and with Omni-Wheels.
He’s just developed a back-drivable cycloidal gearbox.
beautifully engineered, I love how quiet it is when it is running too.
Write comments to push video to recommendations section!
Thank you!
This is an awesome design. Are you planning on making a larger one with your 9235 motor?
Cycloidal gearboxes use a lot of bearings, I've been testing large 3D printed bearings with moderate success.
I tihink you've hit a homerun with this new design.
The durability would probably improve with nylon printed gears.
I'm interested to try this out with very hard resin like Siriya Tech Blue
When I wrote my bachelors thesis in mechatronics, I tried to develop a 3D-printable gearbox that; had high reduction, could handle high torque, was back-drivable and had high precision. To that end, I 'invented'(didn't know of any other at the time, though it turns out there were some similar ones) pretty much exactly this gearbox. In my attempts, I managed to make back-drivable gearboxes with up to 1:100 reduction that could handle a few Nm of torque. I reckon I could go way higher too, though I was very limited due to the pandemic. You mentioned that you wanted them to be back-drivable, I might have some tips if you're interested. It seems you have access to very high quality 3D-printing(at least better than what I did), so it would be really cool to see you try out some of my ideas/designs.
Also, on a completely different note: I am currently taking a masters degree in cybernetics and robotics and would LOVE to see much more content on your control systems! Would be really interesting!
Wow you read my mind!
I made pretty much the exact same thing a month or so ago. Although alot smaller and lower gear ratio! Makes it much more backdrivable.
Good video! Nice work!
I would love to see a back-drivable version. I not sure but I think the most you can do to have a back-drivable gearbox, that is not too much static friction is less then 10 to1
Completely agree!
Here you go ruclips.net/video/_VgXObcfXug/видео.html
The carrier-driven variant has one less stage of tooth engagement, and thus higher efficiency (=more easily backdriveable). But it requires lots of bearings, and has higher planet RPM which may or may not be an issue depending on how fast you want to run it.
Also you probably have to accept a little higher backlash to get backdriveability. Although there's one trick I read about in the comments on James Bruton's cycloid video that I really want to try, which is to use a dual-head printer to make gears where just the outer one or two walls are flex filament, and the rest is hard. The compliant surface allows tighter engagement without jamming, eliminating backlash. But it may still add enough friction to prevent backdriving. Would be great for robot arms though.
Wonderful work. I watched with pleasure.
Thank you!
Would like to have seen a faster input speed to see how quick you could get the output shaft to move. Great video as always!
I second that. I'm always interested in how fast the actuator can move from one position to another
@@Nerdtronic By the look of it, high speed will probably heat up pretty quickly, but should be ok intermittently. Great for robot arms. If you want continuous high speed, print with a little more clearance.
controller board has really made the difference in terms of cmpactness
yay, i was just hoping we would hear from you again soon! fantastic work, the cost considerations of this design are definitely a huge benefit but it also seems to work very well! of course it will wear eventually but that is the same for any 3d printed solution, and the beauty is that they can be replaced and replicated cheaply and easily when needed. and given the cost savings, one can spend a little more on a broad range of more expensive printable materials that i think can last a very decent time.
great to see you! looking forward to the testing video! :-)
If printed from nylon, and given that such a diy robot arm transmission hardly makes the number of cycles as that of a city bus, I think wear is probably not a concern at all. Even PLA has been known to last for years in all kinds of gearing applications.
this looks fantastic! good job on the design! very well thought out mechanism, playing to the strengths of 3dp and leaning away from it's weaknesses.
Great to see a video from you again! Interessting as always.
Impressive I have seen one of the other RUclips who made a way to test the torque on the gear, yet it is a fantastic design 👍
I cannot wait to see a durability test of these plastic gears in action. As well as a max torque and destructive testing of the gearbox as well.
A "FDM 3D Printed Part -> Metal Cast (optional -> heat treatment) -> Filing for dimensional accuracy" workflow would make it even more durable!
Also some crazy filaments like Nylon, PC, (And composite variants), etc
@@ericlotze7724 Metal casting is dimensionally challenging, and leaves a poor surface finish for gears. The slight give in plastic makes them work much more easily.
@@roryevans5032 Ah ok, well thanks for the insight, the metal idea is just an armchair one of mine.
Would vibratory sanding help the finish or not even something like that?
@@ericlotze7724 to work properly metal gears need a smooth surface as they roll against one another. If the metal is rough they scrape and seize up, the only way to avoid that is make the tolerance lower, so there is more clearance, but that causes backlash. For the same reason you lose a lot of power to friction, and make lots of noise. Most gears are machined or ground, which gives a better surface quality than casting. Post processing is also difficult as you have to factor in the wear from the polishing step, so again to work you will generally have to lower tolerance and get backlash, noise, and low efficiency.
Спасибо, что по русски тоже понятно! Когда-нибудь соберу 3d принтер и разберусь с CAN шиной и сделаю робот арм для работы на конвейре.
Awesome video and great ingenuity !
This is great! Finally might have somewhat cheap 3 axis robo arms!
Output torque tries to tumble the planets. This is probably a large source of friction. GearDownForWhat used three or more gear layers to balance out this force, but a planet carrier and bearings could do the same. larger pressure angle gears are more efficient under radial loads. (i.e. 35 or even 45 degree pressure angle) Finally, I expect that heat will be the weakest part of this gearbox so running it immersed in soapy water, or using higher temperature materials will help a lot. (heat is why Nylon, Delrin, and Torlon are the plastics of choice for gears)
I agree, the triple-layer structure of his strongest gearbox is probably quite important; he has some tests also showing this has a big impact on torque-efficiency. For pancake formfactor like this I suppose its not so important, but if scaled lengthwise, the torsional and flexural rigidity of the planets will also be crucial to torsion efficiency and backdrivability I think. GearDownForWhat pressed in steel rods; that solves the bending; but itd be better if they were epoxied in I think, so torsion can be transmissed efficiently along the length of the planet, rather than bothering your layer lines with it. Flex in torsion will cause stress to be concentrated on the teeth where the rings meet.
Thank you very much
Thank you, this was really interesting stuff!
I was expecting it to be very noisy but was fairly dampened compared to what I was imagining.
Also from the video it sounds like the grease dampened some of the rattling noise a fair bit.
Yes please show us more tests on 3d printed actuators and show us which printer I should buy jiji. Everything change if I can print it.
I use prusa printers.
Great design looking forward to see the torque test
Genius! No other words.
Nice
As for the backdrivability, I've seen something about bilateral drives, which also use a compound planetary setup, but they are backdrivable, somehow
Great Design !!! 👍😃
Cannot wait to see your praktibilety testing and using it in one of my own projects. 🤩
Thank you a lot. 💐
That was very good to me I leaned something new
I made a 3d printed Cyclodial gearbox and am still preferring it over planetary gearboxes, the low backlash and high torque comes at the price of those bearings and also those bearings add weight...but I guess it is still the best option for a high torque, low backlash and backdrivable robotic application! But I am very tempted to give your design a try, nice and smooth and looks like there is very little backlash!! 😎
What was the reason for orienting the planet bolts as you did?
The 3D printer does not 3D print perfectly. There are always some micro imperfection. With my orientation all this micro imperfection distributed evenly. This makes everything rotate smoothly.
@@Skyentific Ah yes, clever. Makes sense since you have such low play in the gear mesh, to prevent binding. Did you try installing the planets in a random orientation before orient them this particular way, or was it just your intuition?
hello are the two sun gears rotate in the same direction and same speed? I am trying to understand for the 2nd planetary, it seems to have input speed from the carrier and output at the ring, and it looks like an overdrive?
1. What motor is this?
2. Where do I get the electronics, how much is the cost?
3. What is the magnet for, and how does it work?
Really nice design. ;)
This design is incredibly similar to a design patented in the US as a "gear bearing drive" (US8016893B2). Really the only difference is the use of helical gears as opposed to rings to hold everything together.
You are a master sir. Just inspired me to build more stuff….
Controllers are cool. Actuators are cool but logic and sensing is even cooler... for example following robots
This is SO cool!! Keep up the good work.
Awesome video as always! I would suggest to print using Nylon instead of PLA. Much more durable, strong and resistant to abrasion. Thank you!
The whole point of the design is to be printable by Joe Lambda. Nylon is not trivial to 3D print.
@@ArnaudMEURET Nylon tends to warp and require higher temperatures, but isn't that hard, specially for round parts. But I got your point.
You should do a collab with James Bruton to make a gearbox for his robot dogs
Time for robots that can sense us and collaborate with us
So nice! Very inspiring to see!! Thank you, sir!
Great work, Thanks for posting you are a genius
Very nice video .. please keep up the great stuff your are doing
Do a video on several methods for robots to sense. Led, radar, load sensor, back emf ect
Very cool!
Amazing ! Can you tell me which CAN to USB your are using in this video ?
what about backlash ?
Nice gearbox, what is theoretical maximum rotation speed and torque?
First lol
I'm working currently on a tiny cycloidal drive for gimbal brushless motors, any tips?
I joined the $10 Patreon so I could get the files for this but they are not available. Are the files for this actuator available anywhere?
Yes. They are available on Patreon. Search for post which was published at the same day as video. There are two posts: one public, and second for Patrons with files (stl and Fusion 360)
Приятно видеть человека, занимающегося любимым делом)
Вопрос автору - вы не хотели бы замутить общий проект с Сергеем Дорошем? Он занимается волновыми редукторами и актуаторами.
Could eliminate the screws using herringbone gearing and print the whole lot as a single print!
Yes, but in this case it is difficult (almost impossible) to have very tight tolerances.
I have some questions:
1) Which of the three gearbox types (Harmonic, Cycloidal, and Planetary) are used in modern commercial robot arms?
2) Which of those three gearbox types has the lowest backlash? (low backlash being a requirement for precise and repeatable robot-arm end-effector positioning)
Thank you.
1. Harmonic
2. Harmonic
But: do you want to stay in the past with slow robots, or go into the future with fast robots? If the future is your choice, you should use planetary :)
Amazing!
I‘d say, make a bigger one... 😆👍
And I want to say make a smaller one. :D How small can he go... Could he make one that's just the diameter of the motor?
Is it a good idea to use this actuator as the base of a robot arm? unfortunately we cannot pass the wires inside for better cable management. any ideas?
What software did you use to model the gear set, and did you do the maths manually, it did you use a plugin/script? I would like to design better gesr sets like this
I used plugin in Fusion360. The name of plugin is simply: “helical gear”
Ваш английский стал гораздо лучше :)
Спасибо. А то все только ругают мой английский :)))
English is perfectly understandable and quite good, no problems at all.
@@toomaspruuden3940 thank you! At least some people tolerate it :)))
Very nice. Very very nice.
Have you considered winding your own motors yet?
Love this channel
Very good. Now make a smaller one, make it backdriveable, and use dirt cheap brushless motor and electronics to control it. I find that to be quite hard.
very interesting.
i wonder if a very small actuator one can be 3d printed . Nice work btw :)
You can print small, you can change the muzzle on a printer and calibrate it to get good miniature resolution. You can strenghten the material with carbon fiber.
Or level up and you even can print metal. But this is not the mass market anymore ;)
This is possible. I made this exact gearbox a month or so ago but much smaller. Wish I had shared it now, someones beat me too it :(
But yes you can print a much smaller one!
Интересно. хорошая реализация.
Now I'm very wary of expensive gearboxes.
Why did you remove the magnetic screw and put in a non magnetic one instead? I just missed a bit the explanation why you did certain things, but found in the comments also the reason for the special screw alignment
Magnetic screw would decrease the range of the magnetic field from the magnet for encoder. Most probably it would work with magnetic screw too. But with non-magnetic screw the magnetic field should be a little more intense.
Can you reverse the setup so that I could have instead of 1:28 I could have 28:1 and post where to obtain the 3D printed files for the completed setup...????
What is the efficiency of the gearbox?
That is, indeed, the key question. See my comment above about circulating power.
Leaving a cookie for the algorythm :)
What about torquimeter? Let's add it. And an acceletometer also. Like HEBI actuators yes! Thank for your videos.
How's the backlash?
Brilliant! Herringbone gears are sexy! Subscribed!
The only problem I foresee with design is that the encoder is at the motor not at the other end which means you lose the absolute nature of the measurement of the encoder. I suppose that can be resolved with limit switches, but means you cant simply turn it on and know where it is.
What were these Igus actuators you were referring to? (I wanna see if I can CNC them)
They sell worm gear actuators. Also they have plastic harmonic drive, but it is not yet available.
Superb package, loving the printable / afforable /accessible and compatct aspects of your design. I'm thinking back drivablilty is important and 28:1 is too high for that, is it possible to make a 10 or 15:1 ratio? Would a carrier for the plants help keep gears "aligned" and lower friction help back drivability? looking forward to some more Scientific tests of strength, backlash, durability etc. Cheers
Nice. Is it PLA ? How hot does the motor get ? Seems you can add air flow for cheap too. Because it's not backdriveable, you can use zero current for holding so the motor shouldn't get hot, and when you spin it, adding a small propeller to the motor should help cool it down.
Single pay by PayPal allows me to get this 3d print files ?
I always watch your videos with great interest. Typically a brushless motor requires a controller with a reverse function. All my searches for such a controller are unsuccessful. If you can please link. Thank you in advance.
Whooot! good project!
Nice work, have you considered 3d printed sintered gears ? (You only need to print the teeth in metal, like a hollow shell, and then fill with epoxy for structure.
Also you can print stators and wind it yourself then insert magnets in the ring gear. Only motor's bearing cannot be printed
I need 3d printing design can you give me this file and what is the cost
You need lower gear ratios for backdrivability, maybe 1:8 or 1:6 :P
great video !
if gear box not back drive , the efficiency will be reduce . this mean your reduction will be reduce alot. i hope you can compare efficiency normal and shift teeth gear box . thank you so much
Здорово получилось. Интересно какой момент выдержит. Удачи
Hi , great work .Are the cad models available?
Something tells me your finger would get mangled if you accidentally put it in there 😂 I assume for actual work it's gonna get covered anyway, but for testing purposes, maybe you cut cut out a sheet of acrylic so it would be safe to handle while still see-through?
Love your channel, learn so much from the videos. I am trying to build a robot arm similar to the one from innfos. Any advice on what to keep in mind?
So, what you have there is what is commonly referred to as a "Wolfrom" (not WolfrAm) planetary gearbox. While they are attractive for their compactness, they are also notoriously inefficient and have fooled many, many, many robot designers. Wolfrom's look like they would be the perfect robot gearbox, but they are not, due to the phenomenon of "recirculating power". It's a subtle, but often fatal, flaw. The recirculation basically knocks the efficiency down to 50%, maybe 60% on a well-made metal gearbox. A plastic, or worse, printed, Wolfrom will likely be horrifically bad - likely below 50% and maybe 30-40%.
The fact that you can't backdrive it is the first clue that something is wrong. A two-stage, 30:1 gearbox SHOULD be backdriveable unless it's a wormgear or similar drive with loads of sliding friction or otherwise poor efficiency. If I had a Swiss Franc for every Wolfrom gearbox made by an optimistic robot engineer, I could buy a Harmonic Drive. The insidious thing is that Wolfroms tend to spin well when NOT under load, but once you load them down, the efficiency drops through the floor. The planets also have a fierce twisting moment on them which tends to make them even less efficient, or worse, just bind up solid. Anyway, all that to say, please run some LOADED tests and see what kinds of efficiency you are seeing before investing in a full arm design. I predict that you will be disappointed.
The only Wolfrom drive that I know of that isn't horrible is a traction-planetary version (Archimedes Drive) made by a Dutch company called IMSystems. They work, and they work well, but only because of the very high rolling efficiency of their traction elements - even higher than involute gears - that makes the math work out. Even so, I think Archimedes drives are still only the 80% efficient range - good, but not great. An equivalent (sequential) 2-stage planetary will have efficiency greater than 90% and maybe >95% for a really high quality gearbox.
The poor backdriveability that you are seeing also means that the drive will have poor "transparency" so doing sophisticated torque-based control is a non-starter. If you don't have a mobile (battery powered) or force-controlled application, and you don't care about having to oversize motors and drives, then yes, they can work.
Thanks for the thoughtful critique; I was thinking of making something similar but this is food for thought, for sure. GearDownForWhat was also reporting rather low torque efficiencies in his test of his printed gearboxes. Still I just recently saw a video of a backdrivable 100:1 along these lines. The fundamental compromise with this design is that you will have fast moving contacts under high load at the ring-planet contact; which is something many other transmission like a stacked low ratio gearbox avoids. But as long as the losses there are optimized sufficiently well, it seems to me there are legitimate applications for this. High duty cycle applications are not ideal; but unless your robot arm is actually employed on an assembly line, its duty cycle can be very very low. To open up robotics to applications other than assembly lines, things like cost can be much more the deciding factor than even double digits of percentages of efficiency.
@@eelcohoogendoorn8044 Yes, that's exactly the issue. This style of drive is notorious for having "output loads at input speeds" - which is the worst possible combination. This leads to the low forward and backward efficiencies and poor transparency.
I don't follow your duty cycle comment - efficiency is efficiency regardless of how much you use it. Low duty cycle will let you avoid having heat buildup that might melt your printed gearing (a secondary failure mode), but it won't change anything related to the primary "failure" of terrible efficiency.
However, as I mentioned in my first post, if you don't care about transparency (i.e. you don't aspire to teleoperation or force control uses) or power consumption or having to have larger motors and drives (and batteries if you are mobile), then Wolfrom drives might still be useful to you. They DO work, they just don't work WELL.
@@bschena My point is, priorities are different depending on the application. A robot that is working 24/7 will most likely have a higher added value and thus higher budget than an actuator that needs to perform perhaps only a few seconds per day. While the Archimedes Drive is pretty cool, it sets you back 2.5k, for instance. If the choice was between this printed actuator or a sub-100$ with the efficiency of the archimedes drive, itd be an easy choice, but we dont face any such choice sadly. Check out this video: ruclips.net/video/vQqyGsqPwb0/видео.html you can find his papers too with some googling. Its just a more well optimized version of the same concept discussed here. Im guesstimating with some good design optimizations you could get a printed affordable high ratio version with 60% torque efficiency; and there are plenty of imaginable applications for that, even if kuka will scoff at it. GearDownForWhat also did a video on a printed backdrivable 20:1, btw.
А что насчёт температуры двигателя в разных режимах? Как известно PLA уже при 60 градусах начинает плыть. Или этот актуатор не предназначен для работы в режиме удержания положения?
Love it. Well done! 🤭
Does the output sun gear contribute anything other than mechanical stability? I noticed it's teeth count is not included in the ratio calculation. would it work if it were just omitted?
great video! I was wondering if you could shed some light on how you calculated the teeth required? Let me clarify: at the moment I made an excel spreadsheet where I provide a sun gear with a set amount of teeth and diameter. Based on this I calculate the module by dividing the diameter by the amount of teeth. Now I can calculate the planet gear amount of teeth and the ring gear amount of teeth. Planet gear teeth = module/planet gear diameter ; ring gear teeth = module / ring gear diameter. Then I deduct the diameter of both the ring and planet gear by using the formula 1+rR/rS where rR = radius of ring gear and rS= radius sun gear. I have a feeling that this is where I make a mistake but this is also where I'm asking for your help please. If I do it my way I end up with numbers that end up in dividing by zero if I use your formula for calculating the ratio.
I'm brand new to robotics. I'm enjoying your videos. I'm hoping to build a small table top 6 axis arm with at least a 2 kg payload for deburring parts that come off my CNC mill. I really like this mechanism. What would it need to be back-drivable?
If you are just controlling position then backdriveability doesn't matter. However, if you want to do high quality, open-loop force control (by say controlling motor current) then you will be in trouble. Backdriveability, often referred to as "drive transparency" by roboticist is the Holy Gear for a well controlled, force feedback robot arm. Say, for example, you need to control contact force for, say, a deburring operation, then transparency/backdriveability will likely be required - unless you have a lot of compliance (springiness) between the robot arm and deburring device.
Awesome video! I'm planning on making a 6-axis robot arm and I'd like it to be able to 3d print. Are 0.25° backlash reductors from stepperonline precise enough? Or I'll have to use belts to reduce backlash?
need to take a weight test..
backlash test backlash test :D how many kilograms per meter can he lift?