Why dont you have a mount on the bottom where the plug in is and put a case thing on the spinning bolts that goes around the entire thing and make that a joint.
scale that shit up into the Wrong Trousers is what I'm saying. use those fiberglass rods that they put in jumping stilts or something instead of that aluminum. hell just mechanize the jumping stilts so they fold up on the back of your calves. .... you know what? imma do that.
@@gokiburi-chan4255 ok, see you're now mixing the types of speech, yes the noun precision can refer to any level of precision, however when used as a noun modifier it means that the noun being modified is precise. So if I were to ask what the precision is then yeah it can be anything and still be considered precision, however if I describe something as being a precision thing then it is by definition a thing that is expected to have a high degree of precision. So by saying 'precision mechanical prints' it implies that the thing 'mechanical prints' is very precise. But even without knowing all the grammer rules and definitions (which I certainly do not, english is a mess) you can simply note that colloquially saying something is a 'precision instrument' generally implies a high degree of precision (not a variable degree of precision).
Love the fact you're exploring cycloidal drives, and I mentioned it in the previous video and I'll mention it again. The gear you have is not a proper cycloid, if it was the bearings around the perimiter would be unnessecary because there should be pure rolling contact between the cycloidal gear and a circular pattern of pins to begin with (i.e. those bearings shouldn't be turning at all). This would let you shrink the size of the thing down, save weight with all the bearings gone and make it cheaper. Second, you may or may not want to look into this, you can have the outer pins run free, if you fix/anchor the center pins so they don't spin. This might be handy since it moves the output to the outside perimeter. This means you can have the output supported between both ends, rather than cantilevered off the end of the gearbox which I think(?) you mentioned was a bit of an issues with the last design. Paul Gould here on youtube has some designs/videos with this configuration.
Seconded. The current design is using an "ordinary cycloid" which is easier to derive in CAD but lends itself to higher oscillating forces across the assembly and weaker cycloidal discs; the proper/alternative cycloid is called a "contracted cycloid" or "curtate cycloid" and it greatly improves the vibration and strength while allowing for a smaller assembly. Google "contracted cycloid tec-science" for a fantastic discussion and derivation of the improved drive geometry with nice animations. It's not something that can be easily geometrically derived in CAD but there are functions and generators out there for use in solidworks and fusion. I'd include links but youtube ate my last attempt to write this comment. :\
I mentioned this in the last video, but there's a variant where the outer housing rotates and the cycloidal disk(s) only wobble (but don't rotate). Since this is intended to be used as a shoulder or leg joint you would be able to print the armature and the outer housing as a single piece. This would seem to have some benefits in weight, overall size, and reduced side loading of the bearings in the whole leg. Also It looks like you have some space between your cycloidal gear and your bearing surfaces (both inner and outer), which is almost inevitable with 3d printed parts. In a perfect world the cycloid would touch every bearing at all times, and would not have any slip contact with the bearings (which then wouldn't turn). In the real world only a few points will contact at any one position, and you'll see some slip contact with the bearings (which is why they are turning slightly). If you continue using the bearings, consider a compliant membrane (rubber bands?) around the bearings. This will probably quiet the mechanism significantly as well as ensuring all the bearings share the forces, instead of one or two at a time.
I think the cycloidal disks themselves are only 18 degrees out of phase with each other (ie, the angle between two adjacent lobes). But the center cams they're mounted on are 180 degrees out of phase.
@@steveman1982 if something is more than 360 you can subtract x number of 360s and have the same offset so a phase offset of more than 359 doesn't make sense a 360 degree offset is a 0 degree offset when it comes to phases/relative position of peaks. I could be wrong but point me at something explaining what above 360 degree out of phase would mean
@@ShaunHusain That's the joke lmao. Outside of explaining phase (which is really -180 to +180°), having a value above 360° will tell you how many rotations PLUS the offset something has compared to a reference.
That looks very nice. You could concider brass inserts insted of roller bearings. Friction will be just bit bigger but you can significantly reduce diameter and mass of drive. (estim 50%). As well just as recomendation for knee joint to use smaller motor and drive as torqe will be much smaller on it. so again you can save some mass.
@@hzmeister9596 Not necesery. First of all sliding will be between Bolt and Brass insert. Plastic will be still working as it is now. But totall mas will dropp down significant.
@@mariuszhoscio870 I disagree. Bronze bushings are never used like that. There’s way too much friction and the point load of the smaller diameter would make it even worse.
have you thought about using a single cycloid with an off-center mass to counteract the vibrations, i have some ideas to make it pretty precise, also please look at the comment that suggests to use mathematically calculated cycloids that wouldn't slip on the exterior and probably could work witout the external rods with bearings or spacers, it would really save lots of weight and would be an interesting development, i think that mostly on the knee joint, witch will be a moving mass, the weight of the assembly w, ould be strictly related to the mobility of open-dog, i really believe in the project
This is interesting. A mathematically correct cicloid should not rub and would allow to decrease the outer weight of the drive, making it more inertial friendly
Some thoughts. Using dual cycloid offset by 180 degrees not only reduces vibration but also increases the strength of the gear. Strength is not always an issue, but given the high impulse load that using this to drive a leg joint makes me think it might be important, especially as it is being made out of PLA. And as for using off center mass to counteract vibration it seems to be pretty hard to achieve this with a good precision. Besides this has to work at varying RPM, not just at a specific speed, which makes it even harder to accomplish. But I'd be very happy if I'm proven wrong about that. Also where does the vibration really originate? It would seem to me that the high frequency vibrations is probably coming from the center cam as that rotates at the motor drive speed. The cycloid disk will oscillate at the same frequency, but rotates at one tenth the speed. It would be interesting to see a frequency spectrum breakdown. There should be spikes at engine RPM and at RPM/10. Question is how large these spikes are. Off center mass can be used to dampen the vibrations caused by the cam, but vibrations caused by the cycloid disk oscillating can't be countered as easily. Again, I would be happy to be proven wrong. To switch things around some, a single cycloid disk version should be more energy effective than the dual disk version. Less weight, smaller contact surfaces and less bearings do make a difference.
AGS Door Ease Lube Stick is a very effective lubricant for plastic-on-plastic parts. Just apply like a crayon onto the 3d printed parts and exercise the joint. In a few motions it is super slick without the liquid residue. I simply don't use oils or greases any more on 3d printed parts.
A backdrivable £50 20Nm 200g 200W actuator would change the world of dynamic jumping robots. I'm not convinced it would be possible with anything less than 50:1 reduction. This means you 'only' need 10A at 20V and a 200kv motor delivering 0.4N at 5000rpm. The "Open Dynamic Robot Initiative" (website on github) is also quite impressive. They save weight by integrating the reduction into the leg instead of a separate actuator. 12 x 1100g is quite a weight budget before you even start on the body.
You might have seen this before but the VESCs are able to connect to hall effect sensors for positional accuracy if your motor has them. It solves (or reduces) lots of the issues with stall torque and can safely dump a lot more current in at low speeds
Very nice! That looks much smoother and more durable than the last one. I still think that if you made the outside of the disks slightly concave and the outside of the bearings slightly convex that you would get a smoother transfer of power. A) It would make the disks self-centering; reducing wobble and vibration. B) It would reduce side-load on your bearings; increasing longevity and reducing friction. C) It would spread load over a larger surface area; increasing longevity of the disks. Also, maybe a grease channel in the center of the disks to hold a very small amount of lubricant?
Hey James! Loved the video. Just a thought for the final build, you should look into precision shoulder bolts... They have a precisely ground diameter and a threaded section on the end. Should work really well for a bearing fit without much redesign. They're not too expensive either. Cheers!
An simple trick to increase layer line strength is to just print at an angle. Even a slight angle will help. With some of these parts that are held on the inside of a bearing, you could put a flat spot on the OD of the part and print it on its side as to not use any support material. Another easy trick is with counter sink screws. If you use a hand drill to spin the screw fast in the hole before assembly, you can friction form the counter sink to the shape of the screw. This not only more evenly distributes the force of the screw head to the plastic but also bonds the layers in the countersink together better. Also, this helps when you don't know the dimensions of the screw head. (some) Wood screws have a compound angle on the head and instead of trying to model it perfectly, get it close and melt it with friction the rest of the way.
Interesting application. You could add bearing tracks to each cycloidal disk that would allow for adding individual ball bearings (also nylon or ceramic?) and would also serve to space the discs apart; thus reducing contact area and friction.
If you put lips on the cycloidal gears, kind of like the flanges on the side of a timing pulley, then they would self center on the 11 little bearings. This would help the 2 gears not rub on each other. alternatively maybe use bearings with snap-rings built in to hold them centered?
PLA is probably the best choice for gears. Because of its hardness and also because its very smooth. ABS is maybe more heat resistent but because it is more flexible and not as smooth (still very smooth), also would generate more heat.
If you were to make the assembley water tight, you could fill it with oil to lubricate everything nicely, and ensure slippage between the cycloidal discs.
I'm really impressed with what you have done here. I have done a modest dive into the cycloidal drive system. I have come to wonder if it might be better, for the drive/reduction to be in the pulley. Reasoning that even industrial robots use the belt drives to creatively position the motors. As you mentioned, if tension-ed correctly they have incredible holding power and minimal backlash. This also provide modularity to the system; a motor, pulley or belt can be changed without removing and disassembling an entire sub-assembly. Therefore; I postulate that building the cycloidal reduction into the pulley and retaining the belts is going to be your best approach.
You can attach three hall effect sensors to the VESC, and just glue them near the motors magnets. Impressive design! I think not having bearings would be a mistake. Rolling friction is in a whole different ballpark than sliding.
The 5 holes in the cycloidal disk could be lined with high tolerance metal rings instead of bare plastic. This could massively reduce wear and tear and also allow you to more tightly control the tolerances to reduce backlash. Just a thought! Looooving this project!
This is some pretty good engineering. I wish my body was still able to handle doing the things my mind can come up with. It is nice to watch videos of like minded individuals like yourself! Thanks for the wonderful video!
Well done so far. The next step is to test till destruction, and identify the weak links, Im guessing the central cam shaft pieces. For these you could make a mold from the 3d printed parts in silicone, then cast two part polyurethane copies of the original ( pick a material in the 65-75 D shore hardness range) hexcel modipur. In fact, if you need to make several of these, you could just make a mold set for all of the parts. The good thing about this is that you can re-inforce the polyurethane with chopped strand fiber glass, or carbon fibre / mineral filler. This tech is a perfect fit for the type of prototyping you do
You can replace the bearings with 3D printed roller design bearings, they work perfectly and you can replace the big bearings easily they are perfect for this size. Of course you still have the cost of the small ones but as other people pointed out a different design of the cycloidal gear would eliminated those as well. So you could theoretically reduce the cost by a lot, and the 3d printed ones are just a simple swap so testing them wouldn't require so much effort. (also i think 3d printed bearings can be made lighter than what you have with some expirements and the proper material,infil,etc.) Now for an extra thought you can 3d print very small bearings with another design which uses airsoft bbs and a housing, but they tend to fall apart at high rpms due to friction, in contrast with the roller design which uses friction to break it in and make it smooth. it would be intresting though to replace all the bearings with the 2 designs above though just to produce a very cheap gearbox and test it's limits.(the airsoft design would be OK for the bearings the surround the main gear as they don't spin at high rpms.
This is starting to look super promising! Another alternative to the bearings might be bronze bushings. Could probably be fashioned out of some bronze pipe or whatever that's available at a local hardware store for plumbing, though their inner diameter might be too large. Nylon would certainly be better for weight reduction, but I'm sceptical about their endurance (based only on gut feeling, to be fair). With either you could try making the cycloidal disks either thicker or thinner. I wonder how it'd hold up with less or more surface area, especially once you work on opendog v3 where you might want to have the knee joint be a thinner design than the hip joint or something like that.
I was thinking just make the entire thing smaller. It looks really solid already. If there is a problem with structural integrity for a smaller version, just try a stronger material.
I Seriously think this is one of the most underrated science-mechanics-robotic-tech stuff in youtube. Been following since you were 23k subs James and you were doing the First Iron man stuff. Such a great channel. Thanks for existing and creating all of these videos.
VESC can use Sensors and the newer firmware version also have HFI that allowes you to start pretty good even without sensors. Its still probably not the best controller for your application but if you configure everything properly it will be a lot better than in your current test.
to space the two drives apart from one another and casings you may want to look into replacing the bearings on the center shaft with ones with a cir clip grove in the outer race or a shouldered bearing.
I am a homeschooler, but I have a coop on Tuesdays that starts early. Every Tuesday morning, I make it a priority to watch your new videos before I leave for school! Thanks!
You could reduce friction on the disc's by making some small dimples or grooves in the surface (reduced surface contact). Would also serve as a reservoir for extra lube.
Friction has almost nothing to do with contact area. Just force and friction coefficient. A reduced contact area will most likely increase wear and a thin grease film is sufficient.
Holy shit. That is some really impressive engineering for an "amateur" (and that is a complete understatement when describing your eclectic skillset) fabricator working out of your spare bedroom. Well done.
Hey if you are still using your cnc machine I would recommend that you submerge the cutting bed fully in water to help disperse the heat and to help with the lubrication and the particulates.
I think there are still enthusiasts in mechanics. In fact, cycloidal gearboxes are quite capricious to deformation, backlash and centering. In large deformations they are prone to jamming, and in large efforts they tend to drop their efficiency by up to 15%, although they have the highest efficiency in the gearboxes. Personally, I deal with the development of such industries and my advice is to eliminate the deformations in the mechanism to a minimum, as well as to strengthen the bearing of the shafts. In the laboratory we managed to achieve a 540 gram all-steel reducer with an efficiency of 97% and a hardening of the steel of 60 Hz and an output dynamics of 430 Nm. To find things, look at the places where you should not minimize the deformations and centering of the whole system. I wish you good luck!
James: I HIGHLY reccomend Polymaker materials. Their polycarbonate prints amazingly and is very stiff tough, and the natural coloured nylon (not tried the black) is incredibly tough. I print them with an insulated enclosure; but no active heating, with no trouble whatsoever.
This new motor design is fantastic. Doing two internal gears to reduce vibration is genius. I am excited to see what version 3 looks like, i predict getting rid of all the crazy drives is going to allow you to make something super sleek!
If you use HFI sensor mode (High Frequency Injection) on the ESC (Technically it's not a VESC if it's not manufactured by trampa), it will actually outperform hall sensors, and have great holding torque at the expense of noise. However HFI is picky about what motors it will run on, it's pretty good on skateboard motors, but useless on hub motors, and yours seems somewhere in between. Oh, and HFI is only available for FOC mode, which supports encoders as well.
It's so cool that you're making a cycloidal drive at the same time Levi Janssen is making a harmonic drive. They are very different approaches to the same problem, compact gear reduction, and I'm interested in seeing how they compare. It seems like the harmonic drive has a larger reduction and might not be as easy to backdrive, but it looks way more compact and is simpler to make.
Printing the whole thing in Nylon or some other hard wearing filament (igus filament, Ninjatek Armadillo), and packing it with some heavier grease will do a good job in reducing the parts count. Potentially using some liners and bushings instead of the large section bearings well might cut down on the cost without compromising the accuracy and strength. Bearings are great for higher speed, but aren't as necessary for low speed applications like a robot joint.
James, i just wanted to thank you for your work! I am currently not in the financial situation to support you, but when i am, i want to do it! Because its fantastic that you publish a bunch of your incredible work open source to work with. Thank you so much!
Armchair engineer here (in other words, zero knowledge). With regards the discs rubbing together, could you add some recessed ball bearings between them? I guess though it could cause wear and tear as well a little flex.
@@jamesbruton What are your thoughts on having three disks? Is your design rigid enough to have the output on one side as opposed to both sides of the gearbox?
cycloidal drives offer more reduction in a small package. he's using a 5:1 reduction and could probably get away with a belt which would be FAR simpler with a belt, and zero backlash too. I made a 30:1 drive that's smaller than what he made (using two for a satellite dish rotator)
Molybdenum grease is incredible, double the lobes, no rotary bearings just double metallic sheet between could have an optical gradient for position, halves the size using a stronger substrate.
It would be pretty easy to 3D print bearing "replacements" for the outer ring of bearings, just to test it. Since they are turning very slowly I would thing lubricated PLA wold be fine for the job, as there is a fairly large contact area between the cycloid gears and the rollers.
Is there a way to add something like a friction plate based clutch to a gearbox like this? To introduce a failure point instead of breaking your hark work? Haha I've been wanting to make a cool clunky automatic door opener, but I don't want someone just throwing the door open and shattering my work... I figured maybe if it gets overloaded the friction plates slip instead of breaking the gears. It would only need to be adjustable and not actuated, as it only acts as an emergency safety device. I'd like to see someone build a fail safe or something like that.
Nice work! Maybe you could have a look at strain wave gear / harmonic drive reduction? Not sure how feasible it would be to make but it might reduce the size and weight of the gearbox
Could you improve on the single cycloidal design by adding a counterweight instead of a whole new cycloidal disc? and also improving on the strength. This would keep the weight down by a huge amount - half the bearings, but still the same weight from the weights being equal to that of a second cycloidal disc
no need for odrive, vesc can handle foc bldc aswell. vedder even works on a way to measure the motor imperfections with current pulses into the coils of the motor, so that afterwards he can use the floating coil as kind of a hall sensor itself. this guy is a genius! also nice video btw, really enjoyed.
It looks great, though there will still be a tiny amount of imbalance off-axis, but it's not worth the extra expense to make it perfectly balanced. It may be worth the effort to get 4mm steel shafts, as they are easy to get on Amazon (uxcell is the first seller that comes up). Just be forewarned that they are hardened steel and thus do require tools to cut, and it might help to have some of the basic tools for mounting bearings because it'll be a close fit.
You may want to consider ground dowl pins and bushes instead of bolts and bearings, it may save you weight and be more compact. Added bonus Extractable dowel pins DIN7979C have tapped holes in the end that could help hold everything together
Just print on glass and under cura enable ironing for perfect smooth as glass parts.. His printer is under excruding those parts are very stringy.. 100% infill on some key parts and petg would make it insane. I want 4 so I can make a 4wd MX5 groceries getter picker upper dinner supper~~~
hey james, just want to say your videos and projects are super inspiring and waiting for your upload is one of my weekly highlights, thank you for all the amazing work you do!
Nylon rod on a lathe to machine down to precise diameter and replace all of the small bearings with cuts of it, using lubricant of course. You could embed bushings in them to sleeve the bolts to further reduce friction and wear.
That cycloidal drive looks even faster then the last one and I'm really sure that would produce a lot of thrust if someone attached a propeller of the right size to it and it might even have the potential to produce enough thrust to propel a paraglider.
I don’t think so; rctestflight needed to lubreacte his a lot because he had plastic on plastic for the gears, but this one only has roller bearings on moving surfaces, so it doesn’t need as much or any, for that matter.
Great design. Just a thought: If you're just testing the gearbox, why not just take the motor out and lock the motor input. Then you can put torque on the output until failure.
I once had a cyclo drive where the profile of each of the two cyclo wheels was the xor of two cyclo wheels where one was turned by half a "tooth". If I remember correctly, this doubled the contact points
Seeing James Bruton grab the bar rotating bar makes me think hes lucky to still have all his fingers parts. If he had kept holding onto it for too long the bar would have slammed his hand against the 2x4 and he could have seen how much damage the shearing action would do against his hand. If your doing stupid crap, plan it so that when eventually you screw up you don't pay too steep a price.
Grab Atlas VPN for $1.39/mo before the deal expires: atlasv.pn/JamesBruton 💥
@@occhieanimation2972 Plastics are hard to predict and tend to preform poorly under constant deformation.
@James Bruton The use of 3d Printed cycloidal gears is the focus of my undergrad research. I'd be interested in sharing my research!
Why dont you have a mount on the bottom where the plug in is and put a case thing on the spinning bolts that goes around the entire thing and make that a joint.
get an SLS .. that nozzled crap is really only for 3d illustration of prototypes
JAMES : "The cap fits perfectly " while he is just continuously hitting it with a screwdriver.
James: "I made a leg that hops to test shock loading."
Meanwhile the neighbors: 😳
A robot that gains you some street cred... nice.
scale that shit up into the Wrong Trousers is what I'm saying.
use those fiberglass rods that they put in jumping stilts or something instead of that aluminum. hell just mechanize the jumping stilts so they fold up on the back of your calves.
.... you know what? imma do that.
@@kingmasterlord DUDE POST THE LINK HERE PLS!
This is rev1
Those spacers need to be printed in red. Ask Ivan Miranda. He knows.
SPACEEEEEEEERS
Ivan would also approve of using a screwdriver as a hammer!
Why?
That's amazing. There is something really statisfying about seeing precision mechanical prints that work.
'Precision' isn't very applicable here 😅
@@xxportalxx. would you care to explain? :3
@@gokiburi-chan4255 it's made from 3dp fdm parts using bolts as axles... do I really need to explain the lack of precision here?
@@xxportalxx. but isnnt lack of precision in this case still a level of precision ?
i'm really asking honestly, forgive me for being an idiot.
@@gokiburi-chan4255 ok, see you're now mixing the types of speech, yes the noun precision can refer to any level of precision, however when used as a noun modifier it means that the noun being modified is precise. So if I were to ask what the precision is then yeah it can be anything and still be considered precision, however if I describe something as being a precision thing then it is by definition a thing that is expected to have a high degree of precision.
So by saying 'precision mechanical prints' it implies that the thing 'mechanical prints' is very precise.
But even without knowing all the grammer rules and definitions (which I certainly do not, english is a mess) you can simply note that colloquially saying something is a 'precision instrument' generally implies a high degree of precision (not a variable degree of precision).
Love the fact you're exploring cycloidal drives, and I mentioned it in the previous video and I'll mention it again. The gear you have is not a proper cycloid, if it was the bearings around the perimiter would be unnessecary because there should be pure rolling contact between the cycloidal gear and a circular pattern of pins to begin with (i.e. those bearings shouldn't be turning at all). This would let you shrink the size of the thing down, save weight with all the bearings gone and make it cheaper.
Second, you may or may not want to look into this, you can have the outer pins run free, if you fix/anchor the center pins so they don't spin. This might be handy since it moves the output to the outside perimeter. This means you can have the output supported between both ends, rather than cantilevered off the end of the gearbox which I think(?) you mentioned was a bit of an issues with the last design. Paul Gould here on youtube has some designs/videos with this configuration.
hope it does this, opendog v3 wuold be wasted with 15kgs of motors and gearboxes
I would definitely check out Paul Gould, James.
He was featured in part 1
Seconded. The current design is using an "ordinary cycloid" which is easier to derive in CAD but lends itself to higher oscillating forces across the assembly and weaker cycloidal discs; the proper/alternative cycloid is called a "contracted cycloid" or "curtate cycloid" and it greatly improves the vibration and strength while allowing for a smaller assembly. Google "contracted cycloid tec-science" for a fantastic discussion and derivation of the improved drive geometry with nice animations. It's not something that can be easily geometrically derived in CAD but there are functions and generators out there for use in solidworks and fusion. I'd include links but youtube ate my last attempt to write this comment. :\
@@kimtae858 This is how I designed my cycloidal gearboxes and it is actually quite easy. th-cam.com/video/Nk3aaVcvbpA/w-d-xo.html
I mentioned this in the last video, but there's a variant where the outer housing rotates and the cycloidal disk(s) only wobble (but don't rotate). Since this is intended to be used as a shoulder or leg joint you would be able to print the armature and the outer housing as a single piece. This would seem to have some benefits in weight, overall size, and reduced side loading of the bearings in the whole leg.
Also It looks like you have some space between your cycloidal gear and your bearing surfaces (both inner and outer), which is almost inevitable with 3d printed parts. In a perfect world the cycloid would touch every bearing at all times, and would not have any slip contact with the bearings (which then wouldn't turn). In the real world only a few points will contact at any one position, and you'll see some slip contact with the bearings (which is why they are turning slightly). If you continue using the bearings, consider a compliant membrane (rubber bands?) around the bearings. This will probably quiet the mechanism significantly as well as ensuring all the bearings share the forces, instead of one or two at a time.
this
lol do you mean harmonic drives?
@@qwertzuiopqwertzuiop2107 harmonic drive would be a better idea
@@qwertzuiopqwertzuiop2107 not at all. I dont advocate harmonic drives for this application because they are rarely back-driveable
“They’re actually slightly out of phase with each other, by 180°” 😂
I think the cycloidal disks themselves are only 18 degrees out of phase with each other (ie, the angle between two adjacent lobes). But the center cams they're mounted on are 180 degrees out of phase.
That's less than 540
@@steveman1982 if something is more than 360 you can subtract x number of 360s and have the same offset so a phase offset of more than 359 doesn't make sense a 360 degree offset is a 0 degree offset when it comes to phases/relative position of peaks. I could be wrong but point me at something explaining what above 360 degree out of phase would mean
yeah, each nub is only 18 degrees out of phase
@@ShaunHusain That's the joke lmao. Outside of explaining phase (which is really -180 to +180°), having a value above 360° will tell you how many rotations PLUS the offset something has compared to a reference.
@ 6:04 - Every tool is a hammer ~Adam Savage
I've bapped some stuff in place with power tools definitely not designed for it.
Ah yes, my life savings worth of bearings
Jesus dude, calm down the flexing. I only have so many girls you can steal.
Just buy em bulk my man. They are cheap as shit in bulk like 10 cents per
@@rrtsduf do you have any good sources you use for bulk bearings?
AliExpress is great for that kind of thing
A cheap source of bearings is fidget spinners
That looks very nice. You could concider brass inserts insted of roller bearings. Friction will be just bit bigger but you can significantly reduce diameter and mass of drive. (estim 50%). As well just as recomendation for knee joint to use smaller motor and drive as torqe will be much smaller on it. so again you can save some mass.
Brass inserts are fine between two harder surface. they would cut into the soft plastic too easily.
@@hzmeister9596 Not necesery. First of all sliding will be between Bolt and Brass insert. Plastic will be still working as it is now. But totall mas will dropp down significant.
@@mariuszhoscio870 I disagree. Bronze bushings are never used like that. There’s way too much friction and the point load of the smaller diameter would make it even worse.
have you thought about using a single cycloid with an off-center mass to counteract the vibrations, i have some ideas to make it pretty precise, also please look at the comment that suggests to use mathematically calculated cycloids that wouldn't slip on the exterior and probably could work witout the external rods with bearings or spacers, it would really save lots of weight and would be an interesting development, i think that mostly on the knee joint, witch will be a moving mass, the weight of the assembly w, ould be strictly related to the mobility of open-dog, i really believe in the project
This is interesting. A mathematically correct cicloid should not rub and would allow to decrease the outer weight of the drive, making it more inertial friendly
@@Otakutaru great point
This is a common solution. This is the design I'm currently testing in my research.
I think doing a non-symmetrical infill could offset the balance of gear
Some thoughts. Using dual cycloid offset by 180 degrees not only reduces vibration but also increases the strength of the gear. Strength is not always an issue, but given the high impulse load that using this to drive a leg joint makes me think it might be important, especially as it is being made out of PLA.
And as for using off center mass to counteract vibration it seems to be pretty hard to achieve this with a good precision. Besides this has to work at varying RPM, not just at a specific speed, which makes it even harder to accomplish. But I'd be very happy if I'm proven wrong about that. Also where does the vibration really originate? It would seem to me that the high frequency vibrations is probably coming from the center cam as that rotates at the motor drive speed. The cycloid disk will oscillate at the same frequency, but rotates at one tenth the speed. It would be interesting to see a frequency spectrum breakdown. There should be spikes at engine RPM and at RPM/10. Question is how large these spikes are.
Off center mass can be used to dampen the vibrations caused by the cam, but vibrations caused by the cycloid disk oscillating can't be countered as easily. Again, I would be happy to be proven wrong.
To switch things around some, a single cycloid disk version should be more energy effective than the dual disk version. Less weight, smaller contact surfaces and less bearings do make a difference.
I just found your channel yesterday from the Cycloidal Drive v1 video and subscribed. Needless to say, I am thrilled with your timing.
AGS Door Ease Lube Stick is a very effective lubricant for plastic-on-plastic parts. Just apply like a crayon onto the 3d printed parts and exercise the joint. In a few motions it is super slick without the liquid residue. I simply don't use oils or greases any more on 3d printed parts.
woulnt graphite just work as well?
Yes, but very messy. The Door Ease is clear and odorless.
A backdrivable £50 20Nm 200g 200W actuator would change the world of dynamic jumping robots. I'm not convinced it would be possible with anything less than 50:1 reduction. This means you 'only' need 10A at 20V and a 200kv motor delivering 0.4N at 5000rpm.
The "Open Dynamic Robot Initiative" (website on github) is also quite impressive. They save weight by integrating the reduction into the leg instead of a separate actuator. 12 x 1100g is quite a weight budget before you even start on the body.
You might have seen this before but the VESCs are able to connect to hall effect sensors for positional accuracy if your motor has them. It solves (or reduces) lots of the issues with stall torque and can safely dump a lot more current in at low speeds
Very nice! That looks much smoother and more durable than the last one. I still think that if you made the outside of the disks slightly concave and the outside of the bearings slightly convex that you would get a smoother transfer of power. A) It would make the disks self-centering; reducing wobble and vibration. B) It would reduce side-load on your bearings; increasing longevity and reducing friction. C) It would spread load over a larger surface area; increasing longevity of the disks. Also, maybe a grease channel in the center of the disks to hold a very small amount of lubricant?
Hey James! Loved the video. Just a thought for the final build, you should look into precision shoulder bolts... They have a precisely ground diameter and a threaded section on the end. Should work really well for a bearing fit without much redesign. They're not too expensive either. Cheers!
An simple trick to increase layer line strength is to just print at an angle. Even a slight angle will help. With some of these parts that are held on the inside of a bearing, you could put a flat spot on the OD of the part and print it on its side as to not use any support material.
Another easy trick is with counter sink screws. If you use a hand drill to spin the screw fast in the hole before assembly, you can friction form the counter sink to the shape of the screw. This not only more evenly distributes the force of the screw head to the plastic but also bonds the layers in the countersink together better. Also, this helps when you don't know the dimensions of the screw head. (some) Wood screws have a compound angle on the head and instead of trying to model it perfectly, get it close and melt it with friction the rest of the way.
Interesting application. You could add bearing tracks to each cycloidal disk that would allow for adding individual ball bearings (also nylon or ceramic?) and would also serve to space the discs apart; thus reducing contact area and friction.
Would be cool to see a v3 with nylon bushings instead of bearings, but they would probably be better of turning on a smooth rod than a threaded rod
If you put lips on the cycloidal gears, kind of like the flanges on the side of a timing pulley, then they would self center on the 11 little bearings. This would help the 2 gears not rub on each other.
alternatively maybe use bearings with snap-rings built in to hold them centered?
PLA is probably the best choice for gears. Because of its hardness and also because its very smooth. ABS is maybe more heat resistent but because it is more flexible and not as smooth (still very smooth), also would generate more heat.
You are absolutely a gem for releasing these files as open source. Do you have a bill of materials for the bearings?
When he was trying to stop the arm with his hands I was terrified 😳
"It would probably take my fingers off it I wasn't careful..." Proceeds to stick fingers near pinch point.
Super cool! Can't wait to see version 3 with nylon rings instead of bearings.
If you were to make the assembley water tight, you could fill it with oil to lubricate everything nicely, and ensure slippage between the cycloidal discs.
The VESC does have hall sensor ports and can handle encoders like the AS5047 - not sure why the motor driver was blamed so much.
I use VESC with hall sensors or a AS5047D encoders all the time.
With an AS5047D the VESC supports position control, too.
All this professional production and the man is using a screwdriver as a hammer
I'm really impressed with what you have done here. I have done a modest dive into the cycloidal drive system. I have come to wonder if it might be better, for the drive/reduction to be in the pulley. Reasoning that even industrial robots use the belt drives to creatively position the motors. As you mentioned, if tension-ed correctly they have incredible holding power and minimal backlash. This also provide modularity to the system; a motor, pulley or belt can be changed without removing and disassembling an entire sub-assembly. Therefore; I postulate that building the cycloidal reduction into the pulley and retaining the belts is going to be your best approach.
You can attach three hall effect sensors to the VESC, and just glue them near the motors magnets. Impressive design! I think not having bearings would be a mistake. Rolling friction is in a whole different ballpark than sliding.
The 5 holes in the cycloidal disk could be lined with high tolerance metal rings instead of bare plastic. This could massively reduce wear and tear and also allow you to more tightly control the tolerances to reduce backlash. Just a thought! Looooving this project!
This is some pretty good engineering. I wish my body was still able to handle doing the things my mind can come up with. It is nice to watch videos of like minded individuals like yourself! Thanks for the wonderful video!
This has been very informative. Thanks James. I'm working on my own robot, and this tutorial has given me some new ideas.
Well done so far. The next step is to test till destruction, and identify the weak links, Im guessing the central cam shaft pieces. For these you could make a mold from the 3d printed parts in silicone, then cast two part polyurethane copies of the original ( pick a material in the 65-75 D shore hardness range) hexcel modipur. In fact, if you need to make several of these, you could just make a mold set for all of the parts. The good thing about this is that you can re-inforce the polyurethane with chopped strand fiber glass, or carbon fibre / mineral filler. This tech is a perfect fit for the type of prototyping you do
You can replace the bearings with 3D printed roller design bearings, they work perfectly and you can replace the big bearings easily they are perfect for this size. Of course you still have the cost of the small ones but as other people pointed out a different design of the cycloidal gear would eliminated those as well. So you could theoretically reduce the cost by a lot, and the 3d printed ones are just a simple swap so testing them wouldn't require so much effort. (also i think 3d printed bearings can be made lighter than what you have with some expirements and the proper material,infil,etc.)
Now for an extra thought you can 3d print very small bearings with another design which uses airsoft bbs and a housing, but they tend to fall apart at high rpms due to friction, in contrast with the roller design which uses friction to break it in and make it smooth. it would be intresting though to replace all the bearings with the 2 designs above though just to produce a very cheap gearbox and test it's limits.(the airsoft design would be OK for the bearings the surround the main gear as they don't spin at high rpms.
This is starting to look super promising! Another alternative to the bearings might be bronze bushings. Could probably be fashioned out of some bronze pipe or whatever that's available at a local hardware store for plumbing, though their inner diameter might be too large. Nylon would certainly be better for weight reduction, but I'm sceptical about their endurance (based only on gut feeling, to be fair). With either you could try making the cycloidal disks either thicker or thinner. I wonder how it'd hold up with less or more surface area, especially once you work on opendog v3 where you might want to have the knee joint be a thinner design than the hip joint or something like that.
I was thinking just make the entire thing smaller. It looks really solid already. If there is a problem with structural integrity for a smaller version, just try a stronger material.
I Seriously think this is one of the most underrated science-mechanics-robotic-tech stuff in youtube.
Been following since you were 23k subs James and you were doing the First Iron man stuff. Such a great channel. Thanks for existing and creating all of these videos.
VESC can use Sensors and the newer firmware version also have HFI that allowes you to start pretty good even without sensors. Its still probably not the best controller for your application but if you configure everything properly it will be a lot better than in your current test.
Yep not sure if HFI would would work on a motor like that but is worth a test! It can use many encoders as well.
to space the two drives apart from one another and casings you may want to look into replacing the bearings on the center shaft with ones with a cir clip grove in the outer race or a shouldered bearing.
I just kept saying "that's fucking awesome" over and over during this video.. amazing work.
oilite bronze bushings would also be good to replace the small bearings as they are ideal for low rpm high load applications and are self lubricating
I am a homeschooler, but I have a coop on Tuesdays that starts early. Every Tuesday morning, I make it a priority to watch your new videos before I leave for school! Thanks!
This is your next generation of actuator!
This might be the coolest concept, execution, and overall video I have seen in a long time. Captivating.
You could reduce friction on the disc's by making some small dimples or grooves in the surface (reduced surface contact). Would also serve as a reservoir for extra lube.
Friction has almost nothing to do with contact area. Just force and friction coefficient. A reduced contact area will most likely increase wear and a thin grease film is sufficient.
@@AlmightyMister look up machine scraping vs surface grinding and tell me that's still true.
I cannot overstate how impressed I am with this - it brings me great joy to watch this
Holy shit. That is some really impressive engineering for an "amateur" (and that is a complete understatement when describing your eclectic skillset) fabricator working out of your spare bedroom. Well done.
Gosh James.. You are so brilliant my man. I'm going to join your patreon. Enough following you behind the scenes.
Hey if you are still using your cnc machine I would recommend that you submerge the cutting bed fully in water to help disperse the heat and to help with the lubrication and the particulates.
I think there are still enthusiasts in mechanics.
In fact, cycloidal gearboxes are quite capricious to deformation, backlash and centering. In large deformations they are prone to jamming, and in large efforts they tend to drop their efficiency by up to 15%, although they have the highest efficiency in the gearboxes. Personally, I deal with the development of such industries and my advice is to eliminate the deformations in the mechanism to a minimum, as well as to strengthen the bearing of the shafts. In the laboratory we managed to achieve a 540 gram all-steel reducer with an efficiency of 97% and a hardening of the steel of 60 Hz and an output dynamics of 430 Nm. To find things, look at the places where you should not minimize the deformations and centering of the whole system.
I wish you good luck!
James: I HIGHLY reccomend Polymaker materials. Their polycarbonate prints amazingly and is very stiff tough, and the natural coloured nylon (not tried the black) is incredibly tough. I print them with an insulated enclosure; but no active heating, with no trouble whatsoever.
Try enabling HFI on the vesc. It will track the motor perfectly, although it is a little noisy.
The vesc can also use an encoder :)
This new motor design is fantastic. Doing two internal gears to reduce vibration is genius. I am excited to see what version 3 looks like, i predict getting rid of all the crazy drives is going to allow you to make something super sleek!
I was expecting the table legs to break while testing the torque
Those finger torque tests always scare the shit out of me, but maybe im just overestimating the power of those motors.
especially here where it will essentially guillotine between the extrusion and 4x2, there's a bit of clearance there but not enough for his forearm
The torque isn't that great vs an human arm!
That's some serious designing
If you use HFI sensor mode (High Frequency Injection) on the ESC (Technically it's not a VESC if it's not manufactured by trampa), it will actually outperform hall sensors, and have great holding torque at the expense of noise. However HFI is picky about what motors it will run on, it's pretty good on skateboard motors, but useless on hub motors, and yours seems somewhere in between. Oh, and HFI is only available for FOC mode, which supports encoders as well.
hehehe I see you
and also, HFI can be tunned, but that is a hella ride :D and when it does not like something, the noises it makes...
It's so cool that you're making a cycloidal drive at the same time Levi Janssen is making a harmonic drive. They are very different approaches to the same problem, compact gear reduction, and I'm interested in seeing how they compare. It seems like the harmonic drive has a larger reduction and might not be as easy to backdrive, but it looks way more compact and is simpler to make.
Hey James, I'm gonna put this on my large format 3d printer. Love what you've done here.
hmmm, i might replicate this design for a compact hand crank reduction gear box for a cardboard, paper, plastic, ect shredder. would certainly be cool
Printing the whole thing in Nylon or some other hard wearing filament (igus filament, Ninjatek Armadillo), and packing it with some heavier grease will do a good job in reducing the parts count. Potentially using some liners and bushings instead of the large section bearings well might cut down on the cost without compromising the accuracy and strength. Bearings are great for higher speed, but aren't as necessary for low speed applications like a robot joint.
James, i just wanted to thank you for your work! I am currently not in the financial situation to support you, but when i am, i want to do it! Because its fantastic that you publish a bunch of your incredible work open source to work with. Thank you so much!
Excellent improvements! The 180 degree out of phase worked great! Looking forward to a complete legs setup and test!
Even if I don't know about things in this field its still really interesting to see. That's a cool gear box.
Armchair engineer here (in other words, zero knowledge). With regards the discs rubbing together, could you add some recessed
ball bearings between them? I guess though it could cause wear and tear as well a little flex.
You can also use sensors on the vesc
Great design and nice improvements over V1.
thanks, more testing coming up over the next few weeks.
@@jamesbruton What are your thoughts on having three disks? Is your design rigid enough to have the output on one side as opposed to both sides of the gearbox?
I read that having only 2 discs is not enough to cancel all vibrations. But adding more out of phase discs adds friction. So only good at low speeds.
cycloidal drives offer more reduction in a small package. he's using a 5:1 reduction and could probably get away with a belt which would be FAR simpler with a belt, and zero backlash too. I made a 30:1 drive that's smaller than what he made (using two for a satellite dish rotator)
Molybdenum grease is incredible, double the lobes, no rotary bearings just double metallic sheet between could have an optical gradient for position, halves the size using a stronger substrate.
It would be pretty easy to 3D print bearing "replacements" for the outer ring of bearings, just to test it. Since they are turning very slowly I would thing lubricated PLA wold be fine for the job, as there is a fairly large contact area between the cycloid gears and the rollers.
Probably, I'd quite like to find something off the shelf though.
Is there a way to add something like a friction plate based clutch to a gearbox like this? To introduce a failure point instead of breaking your hark work? Haha
I've been wanting to make a cool clunky automatic door opener, but I don't want someone just throwing the door open and shattering my work... I figured maybe if it gets overloaded the friction plates slip instead of breaking the gears. It would only need to be adjustable and not actuated, as it only acts as an emergency safety device.
I'd like to see someone build a fail safe or something like that.
just writing to let the YT algo know how awesome is ur Channel 🔥🔥🔥
That's well engineered! This video (especially after having seen version 1) is "very satisfying." lol
Yet another project that makes me want to get a 3d project. Hahaha, awesome!!!
10:00 It's perfect. It whimpers like a dog.
Almost 1m subs. Its been a crazy durney watching your progress over the years
5:04 -- How my brain feels when trying to follow every detail of these builds.
Nice work! Maybe you could have a look at strain wave gear / harmonic drive reduction? Not sure how feasible it would be to make but it might reduce the size and weight of the gearbox
Yep next week!
@@jamesbruton Ah excellent!
@@jamesbruton Have you seen the Abacus drive from SRI?
James Bruton, You are a real life Stu Pickles. Really loving your channel!!
Have you considered replacing the smaller bearings with 3D printed rollers on metal pins? Might help reduce costs if it works.
Could you improve on the single cycloidal design by adding a counterweight instead of a whole new cycloidal disc? and also improving on the strength. This would keep the weight down by a huge amount - half the bearings, but still the same weight from the weights being equal to that of a second cycloidal disc
Yes, this is a common solution. This is the design I'm currently testing in my research.
no need for odrive, vesc can handle foc bldc aswell. vedder even works on a way to measure the motor imperfections with current pulses into the coils of the motor, so that afterwards he can use the floating coil as kind of a hall sensor itself. this guy is a genius! also nice video btw, really enjoyed.
It looks great, though there will still be a tiny amount of imbalance off-axis, but it's not worth the extra expense to make it perfectly balanced. It may be worth the effort to get 4mm steel shafts, as they are easy to get on Amazon (uxcell is the first seller that comes up). Just be forewarned that they are hardened steel and thus do require tools to cut, and it might help to have some of the basic tools for mounting bearings because it'll be a close fit.
You may want to consider ground dowl pins and bushes instead of bolts and bearings, it may save you weight and be more compact. Added bonus Extractable dowel pins DIN7979C have tapped holes in the end that could help hold everything together
Thank you! This is very good starting point for further improvements!
fantastic work! I did not think the offset would be enough but it seemed to work!
great work! Maybe a bit of sanding with some rough then fine paper on the discs might help with longevity.
Just print on glass and under cura enable ironing for perfect smooth as glass parts.. His printer is under excruding those parts are very stringy..
100% infill on some key parts and petg would make it insane.
I want 4 so I can make a 4wd MX5 groceries getter picker upper dinner supper~~~
Bushings would be a good replacement for the bearings.
Yep, probably Nylon ones like I mentioned
@@jamesbruton hey now. Are you familiar with the work of Buckminster Fuller? He really did do a lot of thinking ahead of time...
hey james, just want to say your videos and projects are super inspiring and waiting for your upload is one of my weekly highlights, thank you for all the amazing work you do!
thanks!
Nylon rod on a lathe to machine down to precise diameter and replace all of the small bearings with cuts of it, using lubricant of course. You could embed bushings in them to sleeve the bolts to further reduce friction and wear.
That asynchronous and continues motion of gears is troubling my brain, but it looks mesmerising.
That cycloidal drive looks even faster then the last one and I'm really sure that would produce a lot of thrust if someone attached a propeller of the right size to it and it might even have the potential to produce enough thrust to propel a paraglider.
this is the best video yet that ive seen on cycloidal drives.
I love it, glad u got part 2 so quickly after the first one. Also i believe you should lubricate that gearbox A LOT more. A LOT A LOT more
I don’t think so; rctestflight needed to lubreacte his a lot because he had plastic on plastic for the gears, but this one only has roller bearings on moving surfaces, so it doesn’t need as much or any, for that matter.
Please remember to release the github file to the public
it is already done
Great design. Just a thought: If you're just testing the gearbox, why not just take the motor out and lock the motor input. Then you can put torque on the output until failure.
I once had a cyclo drive where the profile of each of the two cyclo wheels was the xor of two cyclo wheels where one was turned by half a "tooth". If I remember correctly, this doubled the contact points
Killin it lately, James! Love this stuff.
Seeing James Bruton grab the bar rotating bar makes me think hes lucky to still have all his fingers parts.
If he had kept holding onto it for too long the bar would have slammed his hand against the 2x4 and he could have seen how much damage the shearing action would do against his hand.
If your doing stupid crap, plan it so that when eventually you screw up you don't pay too steep a price.