I don't think the bushings are to blame for the noise. Bushings are virtually always quieter than bearings. Sounded to me like something was crunchy, like the bearings or bolts grating against each other, but hard to tell over video. Nylon bushes definitely shouldn't have made it louder. Also, I've been saying this for a while, but you could just print the outer gear instead of using bearings or bushings. There should only be rolling contact between two cycloidal gears, so if done properly the bearings and bushings would not rotate at all, thereby making them pointless. Would be much cheaper, lighter and easier to assemble if you just printed the outer cycloid. Gets rid of the bushings/bearings, their metal axles etc and it might even be stronger because all the stress isn't concentrated on the small pins and where they are embedded in the plastic. Also might have less lash because of having a full tooth shape to engage with.
Said the same in my comment. If you look at professional machines they use static metal parts with some lubrication, no bearings or even extra bushings. If anything some of the extra noise could be due to the bushings or bearings allowing for slipping and increased vibrations compared to if it was just a solid 3D print. Could also probably shrink the parts because he wouldn't need to make room for the bushings for external teeth. Shrink the rim around the casing to a thin wall just to keep dust out and move the fasteners to where the steel pins are now.
To me it sounds like the bushings are just clattering against the pla print. I would personally use gearbox lubricant, as it muffles sound a bit and lubricates.
That makes complete sense I hope he reads this because it will be a significant improvement over what he's currently doing which seems a bit redundant compared to your suggestion I never wouldve thought of it thanks for the comment
i designed a cycloidal drive and for the static part I used steel pins (actually from bearings, lol). and they are absolutely silent. the motor is louder than the gear.
Hey James. My experience from mechanical repairs tells me that gearbox would probably run 20x better if it was packed with industrial grease. That much mechanical force and movement going through a tiny layer of lubricant isn't the way to go (probably why it makes so much noise). I know it would be very messy, but I really think that if you could seal the gearbox and packed it with grease you could probably substitute all the bearings with nilon and it would work better then the bearing version. All the plastic on plastic contact would be going through a thick layer of grease increasing eficiency and massively reducing noise. I might be wrong but maybe it's worth a test.
It would certainly be quieter, but I doubt it would run very well. The cycloidal disks don't leave much space where the grease could move around, so you would get a lot of viscosity friction. No issue for low-speed high-torque operation, but not at all good for quick movements.
@@leftaroundabout If its sealed as suggested you don't need to use a high viscosity grease - even really thin compressor oil for instance could be used - don't even need to fill the whole thing either just fill it up enough that the gears will easily and rapidly carry the lubrication to the furthest points from the sump at the bottom. So the idea can be done - I would not personally suggest trying to go quite that thin - as a 3d print as a sealed chamber is the wrong approach, it is bound to loose its oil to easily - if you want to go very thin 3d print a former and frame and use fibreglass to create the sealed compartments - basically FDM 3d prints are generally rather porous so really light lubrication will be hard to use directly... The idea definitely has merit, and makes the design more practical - needing a small fortune of bearings when it really should not be required just adds costs. The only question is at what viscosity of lubricant does the porous nature of FDM prints mean your going to loose all the lubrication without some form of sealing the layers.. I think you will find that a thin grease will work fine, and not really need much in the way of redesign to keep in - but I've not tried FDM prints with every grade of grease/oil going so a good viscosity to get well lubricated with low enough resistance might not work as easily.
@@foldionepapyrus3441 3 layer 3D prints can easily be water tight so I don't think that will be a problem. He'd have to seal everything with gaskets tho. Tolerances won't be good enough for anything to slide without leaking oil
If James were to use industrial grease with the material he's printing with, capillary action would draw it into into the components and act as a solvent. He would in effect have created the robot equivalent of Rheumatoid arthritis. I don't think he should have to learn about material science to improve his design. It will work out and without a very messy solution.
I subscribed yesterday to help him reach 1M and out of curiosity! I watched him from a Mark Rober video and figured that his craft is interesting as well.
Also as the contact point (drive to bushing) changes radius (from motor center axis) as it turns, the bushing will change rotational speed slightly. The two drives may cause binding, and noise/wear.
You should definitely be able to replace the output shafts bearings with more nylon bushings. Also since all contact is supposed to be rolling contact by default, you could get away with 3D printing the external teeth instead of using nylon bushings at all. Looking at industrial designs they do just use bare metal with oil/grease and not bearings or anything in those areas. My gut says that those elements being able to rotate/slip especially freely causes extra noise and losses.
industrial designs can get away with that because of highly precise machining tolerances and very smooth/hard surfaces. I don't know if printed plastic would be able to hold up in that sort of situation.
@@turnerroll9431 Certainly won't last as long being plastic - but being plastic throw enough grease in there and just let it wear in as needed - as long as the tolerances in the design are not awful the 3d printer will make sufficiently accurate parts... Really should not need bearings for any of its teeth like bits, quite possibly not for anything internal at all - just the main shaft will definitely need bearing to deal with the robot dogs weight - it pretty much takes up all the nasty loads - so for all the rest of the internals the forces and friction inside the gears should be low enough to last fairly well as greased plastic... A nearly all plastic construction has a few advantages, pretty obvious ones really, and shouldn't really have much in the way of downsides for this use case. - It would be better to machine in a slippery plastic or metal the moving parts for better longevity and being able to keep the tolerance smaller and so cut down on slop in the movement a little more. - When the moving parts do start to wear out being a single integrated print means much more reprinting to do, any flaw in the 3d print could cause more rapid delamination around the layer lines. - run really hard on a hot day maybe you will get to the softening point of the plastic you used But none of these I would expect to be particularly bad, and to meet the goal of easy to replicate open source hardware its hard to argue with a well toleranced set of 3d prints for almost all the parts - less bearings, rods, bushings, screws to source if more of it can be printed on the relatively available, 'simple' and cheap FDM machines..
@@jamesbruton Bearings still shouldn't be needed for a lot of the areas you have been putting them. Cycloidal drives are supposed to be all rolling friction anyways, so adding bearings just adds unnecessary weight, cost, and size.
Fantastic iteration, loving the process of this era of OpenDog. Can't wait to see where your journey into 3D printed cycloidal drives takes you; this certainly isn't a drag and drop upgrade.
Some design upgrades to consider... - You can 3D-print your own roller bearings, and with a double taper (a taper on both sides) to boot, so that you can also combat axial loads. - A hybrid between a bearing and a bushing might be cool. - Tensegrity structures, compliant mechanisms and shapes of constant width could all potentially help somewhere in the design. - Herringbone gears. - Textured parts. - Hexagons are the bestagons! - In addition to O-Drives, also add some IMUs and other sensors. Some other tuning you can do; - Lubricate everything with silicone grease, teflon lubricant, graphite, or some combination thereof. - Make some tweaks to your 3D printer settings. (Teaching Tech, CNC Kitchen, Maker's Muse and Tom Stanton all have stuff that can help.) - Thread locker can help prevent bolts from coming loose, and also help distribute stress more evenly (bolts and screws are stress concentrators). - A very light adhesive between parts when you bolt them together can further even out stress.
Love your approach to R&D, James. Great video. The legs for OpenDog are looking better and better at each revision. Must admit I was sceptical about these motors but that’s with me not knowing anything about them. Very, very interesting. Thanks 🙏
Absolutely loving this work on cycloidal drives - the cleanness of your prototypes always blows me away. I can't imagine how much time and effort you must put in every video!
Would it make much of a difference to have two bushings stacked rather than just the one bushing supporting the two cogs? I feel like each cog would want to spin their bushing slightly different depending on the contact point.
I guess it would depend on the tolerances. Some diagrams show cycloidal drives to still be in contact with the post when at the peak of a cog tooth as well as the valley maintaining contact through the whole cycle I believe to reduce play in the system. I think that was the purpose of the plug-in that was mentioned to get the right tooth profile. You may be right though. I’m more of a hands on learner and haven’t played with one before, so it’s possible I’m wrong. It’s been known to happen.
8:35 i think the biggest issue here, in regards to the plastic bushings... has to be the fact that they're not tighter against the shaft, there's a lotta play that i see as you rotate the assembly.
A possible cause for your rattling issues is using threaded rod to support bearings. The play between the thread and the inner race will be significant. Maybe switch for ground shoulder bolts with the correct tolerance for a slight interference fit
Before resorting to the metal bearings, how about trying some silicone grease around the nylon bushings? It might help solve part of the noise issue, and might also help keep some of the costs down. Failing that, what about using split bushings in place of the full height ones currently in use? Well done so far. Looking forward to the next one. :)
This is something that’s always confused me. I build bikes and we use grease in bushing and oil in other areas. I don’t know why the usage is so sparse here.
@@tubularmonkeymaniac - If I had to guess, I'd suspect that he's trying to keep the mess to a minimum, since he's probably going to disassemble it several times for debugging purposes. But, that's only a guess. :)
I love watching your videos man, I mean I'm not sure I understand all of it but I like seeing all these motors and gearboxes being made and taken apart really helps me understand how the hell it all works. Particularly you video on the CVT ball drive as my tractor is driven by a hydraulic CVT and it was interesting to see another way to vario a drive, even our seed drill has some sort of vario drive which used sprag bearings and a walking motion. Anyway keep up the good work man!!!
Have you experimented with bushings instead of bearings for the larger input/output interfaces as well? Even with some custom machined brass bushings you would have significantly less weight and $$$
@@owensparks5013 By the time you source all those specific sized motors, nylon bushings, bearings, hardware, print all those parts...sourcing a machine shop to machine some brass wouldn't be all that different.
@@PowerScissor I disagree, a bill of materials with suppliers and partnumbers for off the shelf industry standard sized parts will always be easier to order than anything custom, and will have a shorter lead time. Chances are you'll get all the bearings from one supplier so some bearings and some custom bushings would be an extra supplier already.
The bushings should rotate since nylon on metal shaft gives less friction than nylon on plastic cyclodial disk. To fix this: 1. More clearance between bushings and outer shell so that both are not touching. 2. Cutting each bushing in two with a spacer between both so that both split parts can run in opposite direction.
Hi, i have a question: how does a sponsorship with jlcpcb works? Because I see them sponsoring basically every big or small diy channel on RUclips, but i have no idea on how to contact them and how the sponsorship works. Thanks
They have a "cooperations" link on their contact us page, I think that might be what you're looking for - support.jlcpcb.com/article/45-contact-jlcpcb It might not be, but I'm sure they'll point you to the right ppl. Good luck! :)
Are you sure its the bearings vs bushings? Seems more likely to be luck of the draw with the printing tolerances to me. Perhaps you have already done something like this, but Id print the discs with a succession of tiny offsets on the extrusion multiplier, to see if that matters.
Ah ok; yeah the sound is definitely less appealing; just not quite sure how the decreased rolling resistance of the bearing would be the essence of it from an acoustics pov. Unless there is play between the bushings and pins? Interestingly the V2 has this funny low frequency wobble going on indicating some tolerance issue that V3 seems much less affected by... Also the stacked cam axis seems like a tricky one from a tolerance pov...
@@jamesbruton I'm no expert but controlling machine tolerances will be controlling every variable during printing (room temperature till printing material batch) if possible consider visualsing the wear/friction ( IR camera/adding sooth and see where it sticks, maybe).
The design looks way better now :) With the “actual“ cycloidal shape you increase the number of contact points, thus it more difficult to get the fit of the discs right. Despite what’s written here in different comments, there is not just rolling contact on the cycloidal discs and the pins! So rollers definitely help here! (IMPORTANT) The generated noise is mainly driven by you disc design and the load distribution that comes with it (load distribution; tolerances..). Try to play around with your printed discs tolerances. A slight preload can help a lot, and yes this reduces the efficiency on the friction side a bit, the load transmission could be improved though.. ans add a bit more grease as well - choose a proper one though. For position control of those joins keep in mind that the damping is quite high - that’s good. On the other hand though, there is quite a big hysteresis compared to your former belt drives. In this application it could be fine though, since it doesn’t have to be too precise
That's what I thought - when the 'peak' of the disc comes past the pins while the gear is engaged with the other side it's definitely not rolling. I had to tolerance the disc down to get it to work at all. I just bought 400 bearings for about $70 from China though, so that'll be the way forward.
@@jamesbruton Exactly, it‘s just pure rotation in one point at a time. Since this kinda transmission is designed to have multiple contact points to distribute the load, all the other points are not just rolling - as you said. Yeah, the disc shape and tolerance is quite demanding. To get it working with multiple contact points it might need a lot of testing and matching of the parts - 3D printing has its limits here. The most critical part is the friction: By reducing the size of the disc to get it working without load (not multiple contact points), the amount of contact pressure in the one point under load increases and reduces the capability of the transmission. Your bearings would definitely help here, but it‘s quite some weight and parts. I have also thought about multiple ways of designing and building a ‘simple‘ transmission thats still back drivable. My best approach yet is a two stage belt transmission where the motor sits inside the large output pulley and two different pulley sizes to keep it as small as possible. I achieved a 28:1 transmission ratio and it‘s still backdrivable…(and yes it can handle high torques: tested with 30 Nm and max speed 720 deg/s (at lower torque..)) You can see it in my 6 axis robot vid in case you’re interested. Such a design might be suitable for your openDog
I was honestly surprised he didn't use graphite. I could very well be wrong, but it seems to me that it would do a better job than his runny silicon oil here?
Very cool James, I'm surprised the cycloidal profile didn't help matters too much. Would it be worth replacing the bushings with bearings (without reprinting anything else) to see if the bushings really are the root cause of the noise/friction?
Why would you worry about being able to back drive the motor assembly, when designing the legs themselves to accept shock and give a natural bounce there may allow for stronger motors, more lift capacity, and better shock absorption through the leg structure?
Fascinating project, love to see it develop. In the interest of Open Source, will you also publish the F3D/F3Z files? STP files aren't source, I'm afraid.
Agreed. I was surprised when I went to check out the files and they were all STP files. It's kinda like providing precompiled exes and claiming it's Open Source.
In the interest of open source, perhaps he should be using FreeCAD. Where's the logic in locking one's work in a format that can only be opened with subscription-based software?
If you could simulate backdrivability the harmonic drive would allow much longer battery life since it doesn't have to hold its own weight when standing still.
Interesting fact about speed reducers (ie: planetary, cycloidal, strain wave, ect) if you rotate the outer housing the output will be a differential of the outer housing rotating speed reguardless of the direction of the input rotation. We use very similar speed reducers where i work to achieve differential speeds in various rotating equipment.
I really like the idea of those in between spacers, my two disk are having a problem because the twisting forces will get them into contact. This is of course is causing friction and heat, those washers could be a very good solution to my problem!!
As I understand the workings of a cycloidal (adjusted epicycloidal) drive the contact between the lobes of the 'cycloidal' gear roll on the housing 'pins' and so there is no need at all for bearings or bushings! There should be no relative motion between correctly sized and spaced outer pins; you can print them in the housing (if the surface is smooth enough - which it should be in PLA at appropriate printing temperatures)
How mutch is the clearing between the red part below the locknut and the bearing at 5:16? There should be at least some also overtensioning the locknut might affect the system. Good work btw! I'm loving this stuff
As some other people have commented I think you should print the outer gear to make the drive cheaper and nicer sounding and makes the bearings and bushings redundant
Thinking about this; you could probably use wormgears with actual springs to allow the wormgear to move on a shaft, or the gear driven by the wormgear could be mounted in some kind of springy fasion for the "springyness" you want.
Look for nylon bearings. I have found them in both ball bearing and needle bearing designs. They ought to be quieter and smoother than the nylon spacers. Also, give that test leg a good work out. It may smooth out with use and it's rougher now because of the ridges from 3d printing.
Have you thought about using machined delrin for the cycloidal discs (Not laser cut, due to elephants foot)? Should be very accurate and delrin has low sliding friction.
Do you think nylon would work for a low friction surface? I think laser cutting may work for this depending on disc thickness and focal point. The kerf does have an angle to it but, I've seen worse in 3d prints having elephants foot.
@@4theloot638 Nylon is great, but I wonder if it might be too compliant. Obviously depends on the nylon, but delrin sheets are very rigid in the plane of the sheet. Yeah, laser cutting delrin can be fine if you dial it in - I've just had issues with it really liking to slump down behind the path of the laser
Maybe a bit more grease around the bushings to reduce residual vibration? Also, thin metal spacers between the drive's gearing will probably increase durability since plastic isn't rubbing on plastic anymore.
Very cool! Love seeing your progress on this project. If you're ever looking for a new project, I was thinking a machine that makes your bed could be a cool challenge, because it would require the manipulation of fabric. Just an idea! I always love your content, and I look forward to seeing your next video
Jesus Christ this guy is brilliant. Whenever I think of starting my own tinkering RUclips channel I compare myself to this guy and realise I'm an idiot....
Shouldnt need any bearings are bushings on the outside. I have designed and printed cyclodial drives with 0 moving parts on the "outer" ring and they work great.
thats cool thanks for sharing just wondered if maybe put small holes aside each bushing for easy maintenance I know this grease has a short shelf life even if you use EP grease, if one doesn't mind taking it apart once a year I guess it wouldnt matter
It really comes down to how accurate your 3d printer is. I use 0.1mm offset as a default value for a press fit. So, for a 2mm hole, it's modeled as 2.2mm
I noticed you used one bushing for each shaft that goes across both of the gear things in the cycloidal drives. Have you considered using two per shaft like cutting them in half? It wouldn't fix the issue and it would certainly still be worse than bearings but I think it could help. Another thing that could help is using smooth rods for those shafts instead of threaded but I'm not sure how you could do that and it probably wouldn't be worth the decreased strength and increased complexity to add more other screws or threaded rods.
I'm watching all the cycloidal drive serie and I'm asking myself why did you use grid infill for the parts? There are better options for maximizing the strength in all directions. Amazing work btw
B&B manufacturing US have a online belt calculator that is pretty accurate. Used it to design and didn't need tension pulleys. Long belts/ high load may be a different story.
Have you thought about looking into metal filament? It is a little more expensive, but it's cost includes sintering tickets, to send your parts out to be sintered at to additional cost. That would allow you to make cheap metal gears
Not an engineer but I think you can get a self lubricating nylon and then they shouldn't need to rotate. Make them a press fit on those pins so there is no movement for noise to occur. My guess is the tiny gap between the ID of the bushes and the pins is getting slapped each time the cycloid comes around.
Hi, I've been watching your videos for a couple years and I love it. I'm considering using this cycloidal gearbox on my sumo 3kg robot, but I'm not sure if it can handle the torque and speed of the motor. We'd probably use metal parts instead of PLA, and the KV of the motor would be around 400KV with 4s or 6s. Do you think the gearbox could handle it?? Is it worth trying? Cause the reduction is very high for the size of it.
Have you tried using 2 half height bushings instead of full height ones? if the cycloidal discs are moving in opposite directions, the discs may be trying to move the top and bottoms of the bushings in opposite directions as well.
James, myself and a good friend of mine have just graduated with mechanical engineering degrees. In our free time, we have undertaken the task of trying to build human exoskeletons. I want to use cyclical drives for joints that have one degree of motion. I want to know if you think the drives you've made would be applicable. Thank you for the videos. I've been following you since your hulk buster build. Keep up the amazing work.
idk if you did already but you need to use evolutionary algorithms on the large functional pieces and some of the gears too to reduce weight and increase strength instead of geometrical infill which is wasting weight and not adding much benifits compared to a solid or hollow part also, have you considered using coiled nylon fishing wire and a healing element for artificial muscles instead of motors?
Hey James, I'm really struggling to actually discover a source of brushless motors like what you've used here. A lot of is due to not knowing the exact size / kv that are available making it hard to search for what I need. Could you give me some pointers?
Im wondering myself! If im not wrong (and im totally New to this topic...), you need a motor able to handle enough V, so you can lower A and heat for the same Watts. Kv should be as low as possible i think. Since i dont know where to Look, i just scroll all shops 😆 Correct me if im wrong pls
@@cycust "Kv should be as low as possible i think" makes sense. I came here wondering why a multi-rotor motor is used since it is designed to give high speed but low torque. A dog trying to walk needs the exact opposite. Seems its compact-ness is the reason.
I love the: "just print a few extra parts" outlook... 27 parts later
I wish I had that kind of capability. I can't get one print that size to work well.
Not to mention all the prototypes and failed prints left off camera!
Could you make an entire robot dog with less parts?
And how many printing time 😅
I don't think the bushings are to blame for the noise. Bushings are virtually always quieter than bearings. Sounded to me like something was crunchy, like the bearings or bolts grating against each other, but hard to tell over video. Nylon bushes definitely shouldn't have made it louder.
Also, I've been saying this for a while, but you could just print the outer gear instead of using bearings or bushings. There should only be rolling contact between two cycloidal gears, so if done properly the bearings and bushings would not rotate at all, thereby making them pointless. Would be much cheaper, lighter and easier to assemble if you just printed the outer cycloid. Gets rid of the bushings/bearings, their metal axles etc and it might even be stronger because all the stress isn't concentrated on the small pins and where they are embedded in the plastic. Also might have less lash because of having a full tooth shape to engage with.
Said the same in my comment. If you look at professional machines they use static metal parts with some lubrication, no bearings or even extra bushings.
If anything some of the extra noise could be due to the bushings or bearings allowing for slipping and increased vibrations compared to if it was just a solid 3D print.
Could also probably shrink the parts because he wouldn't need to make room for the bushings for external teeth. Shrink the rim around the casing to a thin wall just to keep dust out and move the fasteners to where the steel pins are now.
To me it sounds like the bushings are just clattering against the pla print. I would personally use gearbox lubricant, as it muffles sound a bit and lubricates.
That makes complete sense I hope he reads this because it will be a significant improvement over what he's currently doing which seems a bit redundant compared to your suggestion I never wouldve thought of it thanks for the comment
i designed a cycloidal drive and for the static part I used steel pins (actually from bearings, lol). and they are absolutely silent. the motor is louder than the gear.
There is not just rolling contact same as for a conventional gear - rollers are a common way :)
Hey James. My experience from mechanical repairs tells me that gearbox would probably run 20x better if it was packed with industrial grease. That much mechanical force and movement going through a tiny layer of lubricant isn't the way to go (probably why it makes so much noise). I know it would be very messy, but I really think that if you could seal the gearbox and packed it with grease you could probably substitute all the bearings with nilon and it would work better then the bearing version. All the plastic on plastic contact would be going through a thick layer of grease increasing eficiency and massively reducing noise. I might be wrong but maybe it's worth a test.
It would certainly be quieter, but I doubt it would run very well. The cycloidal disks don't leave much space where the grease could move around, so you would get a lot of viscosity friction. No issue for low-speed high-torque operation, but not at all good for quick movements.
@@leftaroundabout If its sealed as suggested you don't need to use a high viscosity grease - even really thin compressor oil for instance could be used - don't even need to fill the whole thing either just fill it up enough that the gears will easily and rapidly carry the lubrication to the furthest points from the sump at the bottom. So the idea can be done - I would not personally suggest trying to go quite that thin - as a 3d print as a sealed chamber is the wrong approach, it is bound to loose its oil to easily - if you want to go very thin 3d print a former and frame and use fibreglass to create the sealed compartments - basically FDM 3d prints are generally rather porous so really light lubrication will be hard to use directly...
The idea definitely has merit, and makes the design more practical - needing a small fortune of bearings when it really should not be required just adds costs. The only question is at what viscosity of lubricant does the porous nature of FDM prints mean your going to loose all the lubrication without some form of sealing the layers.. I think you will find that a thin grease will work fine, and not really need much in the way of redesign to keep in - but I've not tried FDM prints with every grade of grease/oil going so a good viscosity to get well lubricated with low enough resistance might not work as easily.
@@foldionepapyrus3441 3 layer 3D prints can easily be water tight so I don't think that will be a problem. He'd have to seal everything with gaskets tho. Tolerances won't be good enough for anything to slide without leaking oil
If James were to use industrial grease with the material he's printing with, capillary action would draw it into into the components and act as a solvent. He would in effect have created the robot equivalent of Rheumatoid arthritis. I don't think he should have to learn about material science to improve his design. It will work out and without a very messy solution.
Que
Man can we get this man to 1 Mil he is so close!
Yup! We will asap
I subscribed yesterday to help him reach 1M and out of curiosity! I watched him from a Mark Rober video and figured that his craft is interesting as well.
Ok im In, lets go 1.000.000
Here's a +1 : )
@@rocks813 same
I recommend cutting the nylon bushings in half, so that each disc can spin the bushing around the shaft in the direction they spin.
Also as the contact point (drive to bushing) changes radius (from motor center axis) as it turns, the bushing will change rotational speed slightly. The two drives may cause binding, and noise/wear.
AIUI the bearings/bushings aren't supposed to spin at all.
@@cheaterman49 key term: supposed. Observations of V2 show that they do rotate.
Very nice progress, love the potential of these kind of actuators.
You're an absolute beast. I dont understand how you do all of this so quickly. Keep it up.
You should definitely be able to replace the output shafts bearings with more nylon bushings.
Also since all contact is supposed to be rolling contact by default, you could get away with 3D printing the external teeth instead of using nylon bushings at all.
Looking at industrial designs they do just use bare metal with oil/grease and not bearings or anything in those areas.
My gut says that those elements being able to rotate/slip especially freely causes extra noise and losses.
industrial designs can get away with that because of highly precise machining tolerances and very smooth/hard surfaces. I don't know if printed plastic would be able to hold up in that sort of situation.
The industry solves that with tempered hand polished surfaces. The bushings are mainly so the discs don't wear too fast.
@@turnerroll9431 Certainly won't last as long being plastic - but being plastic throw enough grease in there and just let it wear in as needed - as long as the tolerances in the design are not awful the 3d printer will make sufficiently accurate parts... Really should not need bearings for any of its teeth like bits, quite possibly not for anything internal at all - just the main shaft will definitely need bearing to deal with the robot dogs weight - it pretty much takes up all the nasty loads - so for all the rest of the internals the forces and friction inside the gears should be low enough to last fairly well as greased plastic...
A nearly all plastic construction has a few advantages, pretty obvious ones really, and shouldn't really have much in the way of downsides for this use case.
- It would be better to machine in a slippery plastic or metal the moving parts for better longevity and being able to keep the tolerance smaller and so cut down on slop in the movement a little more.
- When the moving parts do start to wear out being a single integrated print means much more reprinting to do, any flaw in the 3d print could cause more rapid delamination around the layer lines.
- run really hard on a hot day maybe you will get to the softening point of the plastic you used
But none of these I would expect to be particularly bad, and to meet the goal of easy to replicate open source hardware its hard to argue with a well toleranced set of 3d prints for almost all the parts - less bearings, rods, bushings, screws to source if more of it can be printed on the relatively available, 'simple' and cheap FDM machines..
Happy to see you come back to build another successor to open dog after testing a few concepts in standalone projects.
Me and a friend are going to build the v3 can not wait for more episodes!!
‘I didn’t want to have to buy metal gears so instead I’m using 64 bearings per leg’
I just bought 400 bearings for about $60 though, so it's still a lot cheaper.
64 bearings = a weapon to surpass Metal Gear
Exactly my thoughts. At certain point, the 3D printing just cloaks your judgement and you want to print everything.
@@Jakub1989YTb that's open source, he just want us to be able to do it
@@jamesbruton Bearings still shouldn't be needed for a lot of the areas you have been putting them. Cycloidal drives are supposed to be all rolling friction anyways, so adding bearings just adds unnecessary weight, cost, and size.
Fantastic iteration, loving the process of this era of OpenDog. Can't wait to see where your journey into 3D printed cycloidal drives takes you; this certainly isn't a drag and drop upgrade.
Really loving each of these designs! Looking forward to seeing where this goes!
Some design upgrades to consider...
- You can 3D-print your own roller bearings, and with a double taper (a taper on both sides) to boot, so that you can also combat axial loads.
- A hybrid between a bearing and a bushing might be cool.
- Tensegrity structures, compliant mechanisms and shapes of constant width could all potentially help somewhere in the design.
- Herringbone gears.
- Textured parts.
- Hexagons are the bestagons!
- In addition to O-Drives, also add some IMUs and other sensors.
Some other tuning you can do;
- Lubricate everything with silicone grease, teflon lubricant, graphite, or some combination thereof.
- Make some tweaks to your 3D printer settings. (Teaching Tech, CNC Kitchen, Maker's Muse and Tom Stanton all have stuff that can help.)
- Thread locker can help prevent bolts from coming loose, and also help distribute stress more evenly (bolts and screws are stress concentrators).
- A very light adhesive between parts when you bolt them together can further even out stress.
Love your approach to R&D, James.
Great video. The legs for OpenDog are looking better and better at each revision.
Must admit I was sceptical about these motors but that’s with me not knowing anything about them. Very, very interesting.
Thanks 🙏
Absolutely loving this work on cycloidal drives - the cleanness of your prototypes always blows me away. I can't imagine how much time and effort you must put in every video!
The rattling of the bushings could maybe be reduced by putting grease on the stainless steel rods.
Would it make much of a difference to have two bushings stacked rather than just the one bushing supporting the two cogs? I feel like each cog would want to spin their bushing slightly different depending on the contact point.
Unless I'm mistaken, any given bushing is in contact with at most one of the cogs at any given time, so I don't think it's a concern.
I guess it would depend on the tolerances. Some diagrams show cycloidal drives to still be in contact with the post when at the peak of a cog tooth as well as the valley maintaining contact through the whole cycle I believe to reduce play in the system. I think that was the purpose of the plug-in that was mentioned to get the right tooth profile.
You may be right though. I’m more of a hands on learner and haven’t played with one before, so it’s possible I’m wrong. It’s been known to happen.
8:35 i think the biggest issue here, in regards to the plastic bushings... has to be the fact that they're not tighter against the shaft, there's a lotta play that i see as you rotate the assembly.
Congrats on being so close to 1M! You deserve all of the attention and praise for these projects that are so insanely complicated
soon!
You're a extremely clever dude, I don't understand much of what you say but I'm amazed by what you can do
So happy for you to be hitting 1,000,000!! Congratulations!
divide the bushings into 2 equal pieces!
I really admire the fact that u do all of this for open source :) thank you very much
A possible cause for your rattling issues is using threaded rod to support bearings. The play between the thread and the inner race will be significant. Maybe switch for ground shoulder bolts with the correct tolerance for a slight interference fit
Before resorting to the metal bearings, how about trying some silicone grease around the nylon bushings? It might help solve part of the noise issue, and might also help keep some of the costs down. Failing that, what about using split bushings in place of the full height ones currently in use? Well done so far. Looking forward to the next one. :)
This is something that’s always confused me. I build bikes and we use grease in bushing and oil in other areas. I don’t know why the usage is so sparse here.
@@tubularmonkeymaniac - If I had to guess, I'd suspect that he's trying to keep the mess to a minimum, since he's probably going to disassemble it several times for debugging purposes. But, that's only a guess. :)
Everytime I see one of your videos, it happens to be working on a leg. Amazing work on the open dog
I love watching your videos man, I mean I'm not sure I understand all of it but I like seeing all these motors and gearboxes being made and taken apart really helps me understand how the hell it all works. Particularly you video on the CVT ball drive as my tractor is driven by a hydraulic CVT and it was interesting to see another way to vario a drive, even our seed drill has some sort of vario drive which used sprag bearings and a walking motion. Anyway keep up the good work man!!!
I feel like that "leg" design is very flexible as a platform for its applicable utility.
Have you experimented with bushings instead of bearings for the larger input/output interfaces as well? Even with some custom machined brass bushings you would have significantly less weight and $$$
Custom machined brass rather defeats the open source concept.
@@owensparks5013 By the time you source all those specific sized motors, nylon bushings, bearings, hardware, print all those parts...sourcing a machine shop to machine some brass wouldn't be all that different.
@@PowerScissor
I disagree, a bill of materials with suppliers and partnumbers for off the shelf industry standard sized parts will always be easier to order than anything custom, and will have a shorter lead time. Chances are you'll get all the bearings from one supplier so some bearings and some custom bushings would be an extra supplier already.
The bushings should rotate since nylon on metal shaft gives less friction than nylon on plastic cyclodial disk. To fix this:
1. More clearance between bushings and outer shell so that both are not touching.
2. Cutting each bushing in two with a spacer between both so that both split parts can run in opposite direction.
I love those robotic projects. Unbelievable what is possible nowadays.
Hi, i have a question: how does a sponsorship with jlcpcb works? Because I see them sponsoring basically every big or small diy channel on RUclips, but i have no idea on how to contact them and how the sponsorship works. Thanks
They'll usually contact you if they're interested
They have a "cooperations" link on their contact us page, I think that might be what you're looking for - support.jlcpcb.com/article/45-contact-jlcpcb
It might not be, but I'm sure they'll point you to the right ppl. Good luck! :)
Are you sure its the bearings vs bushings? Seems more likely to be luck of the draw with the printing tolerances to me. Perhaps you have already done something like this, but Id print the discs with a succession of tiny offsets on the extrusion multiplier, to see if that matters.
This! these are machine tolerances your after, additive manufacturing will not get get to the thousandths consistently.
Yes I already had to tolerance the discs down to make it run at all. The sound it makes though...
Ah ok; yeah the sound is definitely less appealing; just not quite sure how the decreased rolling resistance of the bearing would be the essence of it from an acoustics pov. Unless there is play between the bushings and pins? Interestingly the V2 has this funny low frequency wobble going on indicating some tolerance issue that V3 seems much less affected by... Also the stacked cam axis seems like a tricky one from a tolerance pov...
@@jamesbruton I'm no expert but controlling machine tolerances will be controlling every variable during printing (room temperature till printing material batch) if possible consider visualsing the wear/friction ( IR camera/adding sooth and see where it sticks, maybe).
Congratulations on hitting 1M Subs. Really enjoy watching your videos
Thanks!
The design looks way better now :)
With the “actual“ cycloidal shape you increase the number of contact points, thus it more difficult to get the fit of the discs right. Despite what’s written here in different comments, there is not just rolling contact on the cycloidal discs and the pins! So rollers definitely help here! (IMPORTANT)
The generated noise is mainly driven by you disc design and the load distribution that comes with it (load distribution; tolerances..). Try to play around with your printed discs tolerances. A slight preload can help a lot, and yes this reduces the efficiency on the friction side a bit, the load transmission could be improved though.. ans add a bit more grease as well - choose a proper one though.
For position control of those joins keep in mind that the damping is quite high - that’s good. On the other hand though, there is quite a big hysteresis compared to your former belt drives. In this application it could be fine though, since it doesn’t have to be too precise
That's what I thought - when the 'peak' of the disc comes past the pins while the gear is engaged with the other side it's definitely not rolling. I had to tolerance the disc down to get it to work at all. I just bought 400 bearings for about $70 from China though, so that'll be the way forward.
@@jamesbruton Exactly, it‘s just pure rotation in one point at a time. Since this kinda transmission is designed to have multiple contact points to distribute the load, all the other points are not just rolling - as you said. Yeah, the disc shape and tolerance is quite demanding. To get it working with multiple contact points it might need a lot of testing and matching of the parts - 3D printing has its limits here. The most critical part is the friction: By reducing the size of the disc to get it working without load (not multiple contact points), the amount of contact pressure in the one point under load increases and reduces the capability of the transmission. Your bearings would definitely help here, but it‘s quite some weight and parts.
I have also thought about multiple ways of designing and building a ‘simple‘ transmission thats still back drivable. My best approach yet is a two stage belt transmission where the motor sits inside the large output pulley and two different pulley sizes to keep it as small as possible. I achieved a 28:1 transmission ratio and it‘s still backdrivable…(and yes it can handle high torques: tested with 30 Nm and max speed 720 deg/s (at lower torque..)) You can see it in my 6 axis robot vid in case you’re interested. Such a design might be suitable for your openDog
Congrats on reaching 1m subscribers, James. It's very much deserved.
Thanks!
Lube everything with graphite and it will tighten all the gaps and reduce friction, which will in turn reduce the noise a ton.
I was honestly surprised he didn't use graphite. I could very well be wrong, but it seems to me that it would do a better job than his runny silicon oil here?
@@cheaterman49 especially with such an open system the graphite would do way better. The silicone lube will do nothing but collect dust.
Can't wait till this becomes one of those monthly subscription packs :)
I miss the old days of magazines with build parts.
One step closer to printing my wife in PLA.
Very cool James, I'm surprised the cycloidal profile didn't help matters too much. Would it be worth replacing the bushings with bearings (without reprinting anything else) to see if the bushings really are the root cause of the noise/friction?
Why would you worry about being able to back drive the motor assembly, when designing the legs themselves to accept shock and give a natural bounce there may allow for stronger motors, more lift capacity, and better shock absorption through the leg structure?
Because I can change the compliance in real time through software and do force control.
I really appreciate work work man 💪💪💪 that's why I have given u 10-15 subs(although it's nothing in front of ur great work)...Love u James 😇
Fascinating project, love to see it develop. In the interest of Open Source, will you also publish the F3D/F3Z files? STP files aren't source, I'm afraid.
Agreed. I was surprised when I went to check out the files and they were all STP files.
It's kinda like providing precompiled exes and claiming it's Open Source.
@@LanceThumping Dunno about fusion, but Solidworks lets you load STP (not STL) and have it work very well...
In the interest of open source, perhaps he should be using FreeCAD. Where's the logic in locking one's work in a format that can only be opened with subscription-based software?
@@thorjohnson5237 STEP files do work well for what they are. But you can't usually change dimensions or features in them very easily.
@@AaronGrogan Yes, that would be the next step ;-)
boston dynamics has left the group!
I remember you having a CNC. Have you considered CNCing the discs to make them more accurate. Im sure that would improve the drive alot.
If you could simulate backdrivability the harmonic drive would allow much longer battery life since it doesn't have to hold its own weight when standing still.
Interesting fact about speed reducers (ie: planetary, cycloidal, strain wave, ect) if you rotate the outer housing the output will be a differential of the outer housing rotating speed reguardless of the direction of the input rotation. We use very similar speed reducers where i work to achieve differential speeds in various rotating equipment.
Such a great video! I'm very interested in seing the future videos for OpenDogV3
Such a nice project! Thanks for sharing it with us
Your engineering is something special. Keep up the great work-
Congratulations on almost 1 million subscribers! Loved the video super cool!
I really like the idea of those in between spacers, my two disk are having a problem because the twisting forces will get them into contact. This is of course is causing friction and heat, those washers could be a very good solution to my problem!!
look forward to seeing the full robot. Congratulations on million subscribers (in next few hours)
As I understand the workings of a cycloidal (adjusted epicycloidal) drive the contact between the lobes of the 'cycloidal' gear roll on the housing 'pins' and so there is no need at all for bearings or bushings! There should be no relative motion between correctly sized and spaced outer pins; you can print them in the housing (if the surface is smooth enough - which it should be in PLA at appropriate printing temperatures)
How mutch is the clearing between the red part below the locknut and the bearing at 5:16? There should be at least some also overtensioning the locknut might affect the system. Good work btw! I'm loving this stuff
Let's get him to 1 mill
As some other people have commented I think you should print the outer gear to make the drive cheaper and nicer sounding and makes the bearings and bushings redundant
HAPPY 1 MILLION SUBS JAMES!!!!!
wait i just saw yur subs congrats on a million subs it is actually a big thing keep it up
plz make a separate vid for the unboxing
Best intro/outro music EVER !!!
Almost a million subs! Congrats. Nice work. Love you videos!
Thinking about this; you could probably use wormgears with actual springs to allow the wormgear to move on a shaft, or the gear driven by the wormgear could be mounted in some kind of springy fasion for the "springyness" you want.
Tuning it would be hard, and so would changing compliance in real time for force control.
Look for nylon bearings. I have found them in both ball bearing and needle bearing designs. They ought to be quieter and smoother than the nylon spacers. Also, give that test leg a good work out. It may smooth out with use and it's rougher now because of the ridges from 3d printing.
Have you thought about using machined delrin for the cycloidal discs (Not laser cut, due to elephants foot)? Should be very accurate and delrin has low sliding friction.
Do you think nylon would work for a low friction surface? I think laser cutting may work for this depending on disc thickness and focal point. The kerf does have an angle to it but, I've seen worse in 3d prints having elephants foot.
@@4theloot638 Nylon is great, but I wonder if it might be too compliant. Obviously depends on the nylon, but delrin sheets are very rigid in the plane of the sheet. Yeah, laser cutting delrin can be fine if you dial it in - I've just had issues with it really liking to slump down behind the path of the laser
Maybe a bit more grease around the bushings to reduce residual vibration? Also, thin metal spacers between the drive's gearing will probably increase durability since plastic isn't rubbing on plastic anymore.
Hurry make 3 more give them to dog this is awesome love this
Very cool! Love seeing your progress on this project. If you're ever looking for a new project, I was thinking a machine that makes your bed could be a cool challenge, because it would require the manipulation of fabric. Just an idea! I always love your content, and I look forward to seeing your next video
Jesus Christ this guy is brilliant. Whenever I think of starting my own tinkering RUclips channel I compare myself to this guy and realise I'm an idiot....
it comes with practice, you should be inspired, not deterred!
I agree you should not be dissuaded, be sure you’re not an idiot and that you’d likely find an audience that loves you just like you love James.
You always inspire me with these videos , someday i would like to be able to do stuff like this !! Great video man!
Shouldnt need any bearings are bushings on the outside. I have designed and printed cyclodial drives with 0 moving parts on the "outer" ring and they work great.
thats cool thanks for sharing just wondered if maybe put small holes aside each bushing for easy maintenance I know this grease has a short shelf life even if you use EP grease, if one doesn't mind taking it apart once a year I guess it wouldnt matter
How did you get the encoder output from the motor shaft to match that of the actual joint position since there is a gear reduction?
When making parts that have to press into each other, such as bushings, what sort of tolerances do you usually have in the model to get the right fit?
It really comes down to how accurate your 3d printer is. I use 0.1mm offset as a default value for a press fit. So, for a 2mm hole, it's modeled as 2.2mm
@@MrVolt you just have to calibrate your printers step/mm and horizontal expansion in the slicer
About that much
Amazing project 👌
Hey almost 1mil 😀 congratulations.
I noticed you used one bushing for each shaft that goes across both of the gear things in the cycloidal drives. Have you considered using two per shaft like cutting them in half? It wouldn't fix the issue and it would certainly still be worse than bearings but I think it could help. Another thing that could help is using smooth rods for those shafts instead of threaded but I'm not sure how you could do that and it probably wouldn't be worth the decreased strength and increased complexity to add more other screws or threaded rods.
You need those small bearings that we used to find in the V-YoYos!!!!
I'm watching all the cycloidal drive serie and I'm asking myself why did you use grid infill for the parts? There are better options for maximizing the strength in all directions. Amazing work btw
B&B manufacturing US have a online belt calculator that is pretty accurate. Used it to design and didn't need tension pulleys. Long belts/ high load may be a different story.
Have you thought about looking into metal filament? It is a little more expensive, but it's cost includes sintering tickets, to send your parts out to be sintered at to additional cost. That would allow you to make cheap metal gears
It shrinks though, I have a CNC machine, but I'd really like to keep everything accessible
Cycloidal gearboxes, great thing for backlash less drives on Japanese cnc machinery. Expensive as hell, but work great.
Love your vids and great progress! Soon 1million subs!
This dude must go through so much PLA
I was thinking about a poor man's pvc robot dog design using 90° bend and straight pvc pipe
James Bruton : the one man boston dynamics...
Very cool sir I do approve
Very close to 1M
at least you have now bushings for the rest of your life :D
maybe give pla+ a try. still very accessible and has better resistance over normal pla
Not an engineer but I think you can get a self lubricating nylon and then they shouldn't need to rotate. Make them a press fit on those pins so there is no movement for noise to occur. My guess is the tiny gap between the ID of the bushes and the pins is getting slapped each time the cycloid comes around.
Hi, I've been watching your videos for a couple years and I love it.
I'm considering using this cycloidal gearbox on my sumo 3kg robot, but I'm not sure if it can handle the torque and speed of the motor. We'd probably use metal parts instead of PLA, and the KV of the motor would be around 400KV with 4s or 6s. Do you think the gearbox could handle it?? Is it worth trying? Cause the reduction is very high for the size of it.
Have you tried using 2 half height bushings instead of full height ones? if the cycloidal discs are moving in opposite directions, the discs may be trying to move the top and bottoms of the bushings in opposite directions as well.
yooo looks like we are gonna hit 1 million subs soon
what if you just used a chain, like in a bicycle.
It wouldn't slip that easy and maybe saved some space and weight.
Most of the time I don't understand what's going on, just here for the 3d printing implementations
James, myself and a good friend of mine have just graduated with mechanical engineering degrees. In our free time, we have undertaken the task of trying to build human exoskeletons. I want to use cyclical drives for joints that have one degree of motion. I want to know if you think the drives you've made would be applicable.
Thank you for the videos. I've been following you since your hulk buster build. Keep up the amazing work.
They are pretty strong, you could use bigger motors (the 9235) and also more cycloidal discs stacked to get more torque.
idk if you did already but you need to use evolutionary algorithms on the large functional pieces and some of the gears too to reduce weight and increase strength instead of geometrical infill which is wasting weight and not adding much benifits compared to a solid or hollow part
also, have you considered using coiled nylon fishing wire and a healing element for artificial muscles instead of motors?
Have you seen the Ebay listings for "Cam Follower Bolt-type Needle Roller Bearing"? I think they would be ideal for these cycloidal drives.
Hey James, I'm really struggling to actually discover a source of brushless motors like what you've used here. A lot of is due to not knowing the exact size / kv that are available making it hard to search for what I need. Could you give me some pointers?
Im wondering myself!
If im not wrong (and im totally New to this topic...), you need a motor able to handle enough V, so you can lower A and heat for the same Watts.
Kv should be as low as possible i think.
Since i dont know where to Look, i just scroll all shops 😆
Correct me if im wrong pls
@@cycust "Kv should be as low as possible i think" makes sense. I came here wondering why a multi-rotor motor is used since it is designed to give high speed but low torque. A dog trying to walk needs the exact opposite. Seems its compact-ness is the reason.
8308 may refer to the main bearing size not the motor physical dimensions.
I believe it's the stator size.