This mechanism definitely needs a free wheeling second sun gear to prevent exactly this kind of failure. There is an argument for also needing a planet carrier to take up the side to side tilting forces on the planets.
Yeah pretty much what OP said. I saw the failure mode coming way before he even finished assembling it. Though someone sugested brass gears, i dare you fabricate 5 different gears in brass that are not straight lines like normal gears. These gears have the same manufacturing problems double heiringbone gears have. You need to cnc them tooth by tooth on a 4 axis precision cnc mill with a really fine end mill.
lagynas Pffft. La Dee Dah. Conversions are a fact of life, no getting around it by moving decimal points. There’s this really neat technology now that makes miles equal kilometers and cent da grades equal fair-in-hights, it’s a major pain in the neck and requires you to speak out to your google assistant or fav smartphone “convert 6 millimeters to yards!” Baaaam! Get a pen and paper it’s gonna spit that badboy out with no mercy.
I'm using one of these in a robot actuator, quite impressive performance if you can stop the planets from tilting inward. I added an "idler sun" on top that helps with this. It still skips teeth before stalling the motor, but is surprisingly debris and damage tolerant, even in 3D printed PETG!
I stumbled onto something similar few month ago, but their design was backdrivable. The paper is called "Bilateral Drive Gear-A Highly Backdrivable Reduction Gearbox for Robotic Actuators". (commented this already with a DOI link, but comment is gone, I don't think youtube likes me :)
"Why are harmonic drives more popular" - because they have no backlash (0,0°). If a robot wiggles, it is not good. If robot does not wiggle, it is accurate. Anyway thanks for your videos and the very interesting content each time. Best Regards
@@OmegaF77 it's funny because my coworker designed a robotic worm that appeared on a hackaday article and the comments were golden along those lines of "sure, robotic worm, that's what it was for", haha!
What differentiates harmonic drives from this gear system that would make them have any less backlash? The harmonic drive constructed here admittedly has some backlash (at 6:57): ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Emvo3bLT-Z4.html
Harmonic drives can have backlash. Maybe you meant they can't be back-driven? Which depending on the robot, maybe you want a degree of back-drivability
@@bj_ backlash < 0,0°, of course microarcs are always there. back-drivability can work on higher reductions or larger assemplys (or any other) with distortion detection of the flexspline. with load cells.
I would definitely put a planet carrier, with your design the planets can move every which way they want, which puts weird stresses on all the teeth. That would be a big step forward, and I think much better results can be obtained, even with 3d printed parts.
If you do a bit of research this gear design is credited to Ulrich Wolfrom more than a century ago. It is often referred to as a Wolfrom drive. The publication is Wolfrom, U.: Der Wirkungsgrad von Planetenrädergetrieben. Werkstattstechnik, Vol. VI, 1912
The mechanical nerd in me says you need carrier/holder plates for the planet gears, and have bosses coming out of the ends of the gears. Very cool design, and I learned something new today, thanks for the video!!
The planets are indeed loaded in torsion in this GBox, but a carrier does not help in reducing that twist, in my opinion. Although there are many carrierless designs for this GBox, I agree with you: you need a carrier to support the overturning torque produced by the two tangential forces at the meshing contacts with both annular wheels. A sun gearwheel won't solve the problem: it improves radial support, but not overturning, tangential torque. The carrier solves both... but it is heavier! 🤷♂️
The reduction on these is essentially unlimited. The formula is "one minus the ratio of the ratios". Basically, the closer the planet/ring ratio of the top half is to the planet/ring ratio of the bottom half, the higher reduction you'll get... all the way to infinity if they're perfectly matched (i.e. the output won't move at all). Another interesting thing is that if you add a planet carrier, you can drive the carrier directly and eliminate the sun gear. However it does result in very high RPM of the planet gears. Maybe ok with metal, but for 3D printed the sun engagement is usually worth its added efficiency loss.
Fujimoto's Bilateral Drive, indeed one of the most advanced implementations of the 3K/Wolfrom planetary. And the gear ratio was higher, around 100:1 if I recall that correctly!
Please I'm trying to work out how to calculate reduction for this setup. Could you please point me in the right direction? A textbook maybe or a website? I've been trying to figure this out for weeks
@@boluwarin do you still need it? I've just finished making spreadsheet on determining teeth configurations (teeth number, pitch diameter, and gear ratio) for this type of planetary gear train (split-ring compound)
Glad I know what this is called now. What i never understood was how the one ring can fit the extra tooth in 13:17 i also wondered how this worked out but you explained that really well
Thank you for naming this for me as it has also rattled around in my head quite a bit. I was sure it wouldn’t be an original thought and what ya know here it is.
Maybe I am missing something buy couldn't you put a sun gear on the top set isolated from the drive with a bearing. That should force those planets out on the top.
That is one idea, though I might have to get clever with the ratios to get this to work. I forgot to mention in the video, but the top gearset has a really weird fractional module, so a sun might not fit.
or even a planet carrier on the top reduction would hold the planet axial centers in place no? You'd have the extra friction of the planet gears against the carrier axles to worry about but given that you're not taking a drive force off of the carrier not sure how impacting that might be? you might even be able to build a cage structure if you use a carrier plate on the top and bottom?
I’ve had lots of success with similar split ring compound mechanisms with a cycloidal gears. Conjoining the center gear on a 2 slightly different ratiod cycloidal drives allows for far greater gear reduction in a smaller space then planetary can offer.
Using the same gear tooth numbers in the same volume you can get yet another order of magnitude by allowing the planet tooth count to differ on each layer. The highest ratio is when the product of opposite planet and ring teeth differs by only 1 (ie. P1R2=P2R1+1). Such high ratios are of limited use, but it does give you much more flexibility when choosing other constraints.
With stepper motors, it may seem like they have a low max speed, but if you slowly accelerate them you can go faster. If you're starting from a stop and you just try a high speed it will just stall because there is no momentum, but if you slowly increase the speed in code to match the momentum you can make them go much faster.
yeah, totally. drop a ring on top with 5x bearings that support the centers of each planet gear. easy, and will remove that failure mode. the next failure mode would be the teeth shearing off.
Idler gear at the center shat for the top part of the planetary gears and a carrier/cage for them all. I think that would suffice. Although I'm not an engineer.
excellent! i am adding you to my 'youth that will shape a better future' list. keep at it(and explore quantum science while you are at it) hats off to you sir
Look into hydraulic torque hubs made by Fairfield Mfg Company. Used quite a lot in mobile and industrial hydraulic applications. The one I worked on used in a recycling plant, driving a large crusher/shredder. The torque hub was driven by a very small displacement Charlynn-Eaton geroler type hydraulic motor. It was a pain getting the planet gear timing set correctly
The two places I've seen this mechanism used before is in a patent for car mirror electric actuators where they didn't need fast actuation but needed to be able to fight forces on the mirror like wind traveling at highway speeds. The second was in the patent for a helicopter gear reduction, where they invert the mechanism and drive the two ring gears separately with two turbines so there doesn't need to be any kind of linkage between the turbines otherwise. In that case one turbine ends up operating at a slightly higher RPM than the other, but I guess this is perfectly acceptable. I believe the turbine drove the ring gear through hypoid gears for a further reduction.
16:52 - That's a "no load current". A stall torque is when your motor transitions from full-load (still dynamic) to stall (output shaft no movey-movey). The phrase you're looking for is "idle current draw". 17:09 - That 2 lb water bottle drew 290 mA, as you transitioned from 270 degrees, static and proceeded to decrease from there. (Which makes sense, since you've overcome the static friction and also abs(sin(270)) =1, w/r/t gravity. Which is why if you're reefing on a bolt with a ratchet and cheater bar it's always easiest around perpendicular+the amount of downwards angular momentum you can develop). Then it reverts back to an idle current of 110mA. (Which is a lot. Was the motor still drawing that much current after you removed the assembly?) Your failure analysis is spot on re: folding itself in. It's typically bad form to say you "designed it", even if you cite the Wikipedia page. Even more so when you say "of anything I've designed", implying this is the result of multiple iteration cycles of you and/or your team. If you want to see real power density, take a look at The Torque Test channel. Handheld impact hammer drills are putting out > 1,000Nm running on stock Samsung battery packs in messy environments (ever seen a contractor work in a class 100 ISO Clean Room? Yeah me neither.) The general technique is 3 stages of planetary gearsets, each stage sequentially moving slower/made out of a denser material. By the time you hit stage 3, everything is forged steel and the lost kinetic energy (those BLDCs get up to around 22k or so) is made up for by the mass. Their setup is pretty good, but not AvE good, sadly. Developing 1,000Nm of torque is impressive, but we've been able to do that for ages with diesel generators and airlines (railroads, oil rigs, etc all operated with Ingersoll compressors). The issue is those setups have literally < 5% output efficiency. It blows my mind that these *handheld drills* can drive exploratory drill string with comparatively no losses
That's the thing about this channel, it's extremely armature. Nothing more than experiments, sometimes unrealistic designs, and poorly educated guesses. There's been issues with past projects, there's issues with this project as well, especially when it comes to design; a lot of these designs also seem to be initial prototypes with no real refinement. These same designs are also unrealistic, such as this design having gears of eight different profiles, which would extend production time and drastically increase cost _if_ this were to become a production product, with a refined design of course. Then there's the issue brought up in this comment, falsely equivalating two different gearboxes based on the idea that they have a 'similar operation,' which is nothing more than an uneducated guess, which leads to poorly informed claims. Then things like questioning why harmonic drives are favored over a dual stage planetary gearbox, as if the host has forgotten that backlash exists, and every implementation of harmonic drives specifically uses them to minimize backlash while also having a deep gear ratio; while stacked gearboxes have a multiplication of backlash, I think the lack of backlash here comes down to the printer's tolerances, a properly machined set of gears more than likely would not have this this tight of a fit, _plus_ printed gear mechanisms _do_ need to be broken in before they're tested, something that isn't really seen on this channel at all. I like this channel for it's interesting concepts, but there really needs to be another channel hosted by someone with an actual engineering background that would analyze these designs from a proper prospective. Like the last video, the differential drive, it had numerous flaws that an actual engineer can fully improve the design of, the idea has potential but the implementation was subpar. Then there's designs from a couple videos ago, the compliant harmonic drive, which still does have it's flaws but was an actual educated guess that could actually lead to improvements in harmonic drives, both the idea and implementation were good but it needs refinement. Then there's videos like this, that're DOA, as the concept itself already exists and has already been refined throughout it's existence, for example this exact mechanism was in use in WWII planes for driving propellers, where this mechanism itself is just one of many dual-stage planetary gearboxes; the poor design here, uneducated guesses, and falsely equivalating the mechanism to other mechanisms is 100% the fault of the host, and there's not fixing that unless they better educate themselves, or at least do better research, and the video is almost entirely a misrepresentation of the mechanism. There's similar issues with other channels, notably when mechanisms are either entirely or almost entirely 3D printed. Herringbone gears, for example, are simple to 3D print, yet machining one is an absolute pipe dream. That's not even the entire issue, every time a herringbone gearbox is assembled from all of these 3D printed parts, especially planetary sets, the housing either has to be split or the housing has to be flexed, and this doesn't work with real-world machined parts. It's a disconnection between what rapid prototyping can provide and how a long-lasting real-world part can be machined, something seen with almost every channel that primarily uses 3D printing for designs. A similar issue can be seen with every one-off project, where the design may work fine for a singular production, but doesn't make sense if production is scaled to anything larger.
I've only seen one example of such a double ring compound harmonic drive is Oskar Puzzle's "Harmonic Hyperdrive", which is a 300:1 gear ratio and extremely small (and delicate).
@@FeignJurai Oskar has some amazing designs, but it's a shame that he's only really into puzzles. So many of his designs and concepts could have actual real-world applications if they were refined, improved upon, made of proper materials and had proper construction (e.g. using bearings on the various mechanisms that have rotating pieces, proper housings, etc.), and tested under real-world conditions. Oskar also has some interesting concepts that I don't think would have any real-world applications outside of puzzles. For example, he had a spherical gear design years ago, Spherigear I think it was called, that could've worked in two rotational axes, but the housing only hinged in one axis, and the tooth profile on the spherical gear is not optimal. This eventually evolved into his Flip Gear, which better suits the singular hinged axis design. Then the Flip Gear eventually evolved, by request, into a design that has the coaxial configuration decoupled, called Flip Gear Disconnect, where the original Flip Gears were coupled in that orientation. Real world use of all three of these would require a CV joint boot around the gears and housing, and of course bearings between the gears' shafts and the hinge housing; then of course a way to actuate the hinge, but that's on a per- use case basis, not something that's inherent to the core design of the mechanism. The Sphereigear would also need a bi-axis hinge mechanism of it's own to operate correctly, which has yet to be designed to my knowledge. I've also noticed that the Spherigear has some variations, five total that I know of, four can be found by searching 'Spherigear' on his page, and the fifth is in the form of his Tracker Ball 3D; though spherical gear designs go past this, such as the ABENICS Active ball Joint Mechanism, or theoretical designs that use a stellated (positive and negative stellations as a gear pair) geodesic or Goldberg (dual of a geodesic) polyhedron, I believe there's also a peg-based design that uses dimples and nubs, and there's a few other designs floating around out there. Biggest downfall is that spherical gears would ideally need teeth separated by equidistant points along the sphere's surface, something that's a concept that, as far as I know, has not been mathematically been proven to exist outside of using the vertices of the platonic solids; and even then, there's no telling if we can make spherical gears have the same performance as 2D involute helical gears once we do find optimal tooth placement. Another example would be Oskar's Magic Gears, the bi-lobed friction gears, and his Illegal Gears, the tri-lobed friction gears, though not his original design, I believe these lobed friction gears largely originate from Jacques Maurel in the form of 'paradoxical gear.'' Jacques made a publication(s) in '04-'05, though some of his papers on relevant topics go back to '02, but I don't know if Jacques _invented_ the type of 'gear' or not, it's just the oldest publication of the gear that I've come across so far. I also don't believe they've been tested outside of proof-of-concept scenarios, I'd imagine the best real-world use of them would be in the form of a modified traction drive. Oskar's variations seems to be the most popular source for these types of gears, though. But again, efficiency is a concern with this kind of 'gear,' though further research into it, and further improvements, should be far easier than the spherical gear issue. Unfortunately, his Grinder Gears and Harmonic Hyperdrive are victims of the same shortcomings, designs and concepts that could have real-world uses, or at least potential of being more than just a toy, yet nobody has really taken the ideas and put them into real-world scenarios with actual engineering behind them.
I had the same idea starting from the same assumptions about the harmonic gear. but your use of asymmetrical gears is really smart. As for the gear slip an unconnected sun gear should suffice, but I believe you have already gotten there.
Thanks for explanation. I'm reading "the great detective" by Wesley Stout. About Chrysler development of WWII SCR 584 radar pedestal that mentions high ratio compound planetary but your details helped me understand much clearer.
This design concept is called Wolfrom drive. I have been looking up for similar high Reduction ratio speed reducers on the internet and discovered things I never knew existed Example: Archimedes Drive Traction Drive Wolfrom drive harmonic drive cycloidal drive cycloidal ball planetary transmission Planetary roller screws Its almost crazy to imagine, there is nothing anyone can think of today that someone somewhere 50-100 or more years ago hasn't thought about, we just got to find the right key word. Consider "Planetary roller screws" I first sow that design concept on one of the "Tesla Bot | Actuators Team". At the time I thought it was a new thing but until recently I discovered that same design existed some decades ago. Crazy right?
Made a clock that uses this mechanism it was for an engineering design challenge with restricted material, had to be water jet cut and minimum tooth size this allowed for good nesting of gears etc. Was only one of two teams that was able to get a functional clock out of 70 teams. Can say I have a soft spot for this gear type!
Harmonic drives are used in automotive seat recliners for example. They are fantastic in that usecase, since they are really small stamped steel cheap parts.
Insert a exta gear in the middle on top which is loose to prevent skipping. Also integrate a Ballbearing made out of airsoft bbs. I got mine up to 50Nm
If you put a sun gear between the planet gears that doesn't interface with the shaft, on the second ring layer, it will help support the planet gears. Oh, and now you're talking about that exact thing... :P
This looks pretty cool. Have you heard about Archimedes drive? The design is based on a planetary gear system but it uses friction (toothless gears aka cylinders) instead of gears. I'm not sure if you can 3d print it but you could definitely give it a shot.
Like the article. The planetary gears do not have a cage holding them. This was not pointed out. The only issue is that their is a torque force with the axis between the two gears with one tooth difference of the planets thus there is also a torque in each gear keeping it from typing and for the gears without a sun there is a thrust also towards the center that need a matching torque on the sun gear. I do not know if you can make a floating sun gear for the top planetary. My Atlas lathe has a 14.5 degree gear tooth pitch angle. The most common is 20 deg, however; special gear have higher angles that will be stiffer and carry more torque at the expense of higher thrust towards the axis of the gears. Which means that the ring gear needs to be beefed up. Same thing happens for cyclonic gear because the physics or geometry is the same at the line of contract angle. It will produce the same forces independent of the type. Friction drive can use the same approach. Also there is not cage, but without teeth slippage will cause a problem. So at least two sets of rollers are needed to set an even spacing. Then there needs to be a free wheeling sun for the second. Now at least one of the roller could be an elastic item such as a tube. This will reduce the tight tolerance at the cost of the peak torque.
MKS and BTT make those NEMA 17 and 23 closed loop stepper drivers for the backs of stepper motors. They're called SERVO42 and SERVO57. They're fairely cheap and work well.
Nice work. Automatic car side mirrors are another example of their application. Other layouts can include an idler gear (by dropping same number of teeth as planet gears) or even an idler cylinder to prevent the gears bending inwards. I've also done some math to fuse the idler to the driven gear which really helps distribute the force to the planets. Would like to see some efficiency measurements - I believe the limiting factor is friction of the meshing gears under output load transmitted to the input and stalling the motor.
Do you know who is manufacturing those mirrors...? I heard of this application and even saw a picture of one, but never managed to get the name of the manufacturer... Thanks!
@@pablolopez-garcia853 I probably saw the same picture. Might be less common than I thought, looking around most appear to be worm gears. Probably a case of one manufacturer avoiding patent royalties! Thought it was a good example of using a very low gear ratio for positional accuracy but not torque transmission (due to very low efficiency). Gears easily skipping could also be a protection feature where a worm gear would otherwise strip teeth.
@@TroyMackay Efficiency is indeed the main limitation for these devices, althought it can be substantially improved with some tricks including a sturdy carrier and optimized tooth geometry. Thanks for your useful comments!!
Boring and facing heads often use a similar mechanism. Mostly one or two pinions running against two internal or spur gears with one or two tooth difference. For most applications it's probably easier to use a backlash free mechanism with lower reduction and just use a bigger servo though.
Alexandre Chappel did a video where he used 3D printed herringbone gears to drive his bench vise. Apparently the shape of the gears allows for a larger surface area for the gear teeth to mesh, making them much stronger.
If you want a HUGE differential, I would design a gear that has a 3D stepped star shape pattern, there would be a small gear that has a diagonal stepping patter that goes from either edge to center or center to edge, and then it raises a level and repeats the stepping in the opposite direction, and you would design it so that it keeps raising till you hit the 180 degree mark of the gear, at which point it starts to fall back off. Lets say you have 10 steps per star arm level. That means traversing what would would normally be 1 step on your small gear actually traversed 10 steps in the same amount of rotation, and then the pattern raises a level and reverses itself with 10 steps back in the opposite direction.
There was a 5 part article in Elektor Magazine in 2001 that used this mechanism for a Polar co-ordinate PCB drilling machine. Links aren't working in comments so search - Elektor Magazine PCB Drilling Machine March 2001
you could simply put a star shaped cap connecting all of the planetary gears together on the 2nd stage that holds them at a fixed distance. doing this would also cap off the exposed planetary gears for saftety
Torque and ratio are decoupled in a compound planetary. I found this when I compared a 48:1 to a 705:1 (same size) and found the torque to be about the same.
If that is enough torque for your application, then you can use a weaker motor. When the torque is limited by the output gears, then the only way to increase the torque is to improve those output gears. The pessimist, optimist, and engineer see the same partially full glass. The engineer sees a glass that is twice as big as it needs to be, assuming no safety margin. The engineer then takes into account that the liquid is red wine and there is a white carpet. The engineer then realizes the amount of fluid isn't the issue. They are simply at the wrong kind of party. They make an excuse about a problem at work. Then leave the pessimist behind and go off to a backyard party hosted by another engineer.
@@hamjudo I didn't have enough torque, he motor stalled and nothing broke. The torque was good, just not what I expected. I had to change the geometry not the ratio.
Torque Ratio is Speed Ratio multiplied by efficiency in a planetary. You efficiency in the 705:1 GBox must have been around 15x lower than for the 48:1. It sounds like a lot, but it is actually possible if your planet wheels were large enough, compared to the sun gearwheel...
Twist the gears 1/tooth arc for the smoothing out of contact faces. The output sun gear can be magnetically cushioned away from the motor shaft. What about making both sides beveled and increasing the face to face compression to try to mitigate even more tooth slippage. You could build a very powerful hinge if you had a stack of thin compressed gears like that.
Gears sets, when properly produced, only have rolling friction. The involute profiles for the teeth are such that when the teeth touch, there is no sliding friction. Did you change the pitch diameter for the top teeth so that the reduction from 50 teeth to 49 teeth did not result in sliding friction on the top gear sets? It’s basically a differential gear set at that point.
I don't know if you're aware of skyientific's channel. He made this same gear reducer with herringbone gears without bearings and it was quite af. Great job never the less!
As far as this gear set goes you add more planets to a planetary gear set and the torque goes up there's a reason why cars and semi trucks use these in their automatic transmissions. I would like to see you print another sun gear to go in the center of that second planetary without a keyed input to the shaft and then see what your torque limits are. But I will agree that the backlash would be relatively unacceptable for robotics though
It is interesting to see the sun gear slipping as the weakest link. If you use a different number of teeth (say 60:56) then you can use 4 normal planetary gears without needing the offset between the top and bottom halves. The sun gear can then extend further, reducing the bending you saw here.
The pcd's of tye outer gear set does not allign properly. For 50T ring gear, the radius is 50T*module which is equal to (20T/2+15T)*module=25*module. For the 2nd ring gear the radius is only 24.5*module. The appropreite planet gear is 24.5-20T/2=14.5T. This means that standard gear will not mesh properly. You may get awsy with this by increasing backlash which is done by scaling down the gear
I made a few of these compound planetary gear with 62:1 and 100:1 reduction. First saw them in "gear down for what" channel. they are real powerful but confusing when calculate the gear ratio. An additional sun gear is definitely needed to keep the planetary gear from curling inwards.
Great video and project! thank you. Actually, I think that the greatest benefit of harmonic drive is the extremely low back-slash provided by the pressure that the ellipse generate between the flexible gear and the static outside gear. This is much more important for robotic or CNCs (rotating table) than having to add more torque to the motor: Just remember how much imprecision a few minute-arc of back-slash is already adding to an industrial 2m robot arm.
i'm using a harmonic drive for a turntable system for a photogrammetry rig. i needed a gear ratio that would give me a number of steps around the table that was highly divisible by many other numbers, so that i would have many options for the amount of rotation between each photo. 7200 steps per rotation ended up being that number and i used a 1:36 harmonic drive reducer to give me that number. i would have gone higher, but the teeth would have been far too small. i definitely want to give this a try.
It is impossible to install a second sun gear (it should have 19 teeth, but it is not divisible by 5), just install the axes of the planetary gears on the carrier in the form of a ring. If the planets have 15 teeth, and the ring gears are different, then the planets cannot be on common axes, with the same module. The module of one half of the gearbox needs to be slightly changed so that everything engages correctly and the module becomes non-standard.
Put small bearings at the top of each red gear, and one around the shaft that they engage with. That way they push outward always and maybe get higher weight/torque?
If your printer has an extra pin, you can get a plugin for Cura that will trigger a camera whenever it finishes a layer. If you don't, there's other options. For example, my printer triggers the camera by running the bed into a switch.
Nice! I suppose if you'd reduce the length of the arm it would be much easier to lift. In naming the weight you should name the length of the arm too. I like how you made the planetwheels longer, that must increase the strengt of it. The idea of this makes me smile. I subscribe.
@@nemoskull2262 What would be the benefit of that approach? Do you think you can vacuum cast gears with a higher yield strength than what you could cut using EDM? How are you going to produce the molds?
I don't understand why the bottom planetary gears are all different. I made something similar with them all the same size. I also had a free spinning sun gear on the top, which you discovered helps things.
Great video. The problem with the design is it only has one tooth fully engaged at any one time. The planetary gears should all be linked top and bottom
All you needed to add was a star with bearing to go in the outboard end of the planet gears. That should about double the usable torque, or will at least reach the strength of the material. I don't think a free second sun gear would match the offset teeth of the planets.
These are also produced with hollow cylinders as gears, and they provide even greater reductions and zero backlash since there are no teeth and there is always contact. Quite sure the company calls them Archimedes drive