The Compound Cycloidal Drive - Something Novel 

Your Contraption ist verry nice! To get around the one big bearing: Replace it with multiple small ones.

A thought for the bearing issue: Instead of buying thin-section bearings, you could treat the inner black piece and outer white piece as bearing races themselves and simply put small balls in a grove. Depending on what you make them out of, it could be as smooth and high-load bearing as needed without being expensive or difficult to source.

Needle roller bearings are a good choice for your second stage in terms of space constraints and mechanical properties, they take high static and dynamic radial loads (vibrations) and as this is your second stage, you don't have to deal so much with their major downside, which is friction at high rotation speeds, the only issue really would be the price tag. Another type of bearing that would push the space and weight constraints to an absolute minimum is self-lubricating bushing (also called sleeve bearing), usually producing more friction, but all comes down to material science. Graphite and other nano-lubes profoundly change - at molecular level - the gliding properties of the contact surfaces, and may therefore do the magic you are looking for.

There are thinner options for bearing, have a look at needle bearings. Skyentific with his robot arm also has some very thin looking bearings and he mentions the part numbers in a few of his videos which might be helpful. For securing the eccentric gears maybe a few axial bearings behind them, between them and the face plate with the pins? Or, if you're 3D printing put a double helix on the gears and they'll hold themselves in position as long as they have clearance to any static parts, or bevels can serve a similar function if you want to machine these on a 3 axis CNC. Finally some flanged bearings (preferably angular contact, two in opposing directions) can be used in a few places to hold the gears and lift them off the face plate so they don't rub. As far as being hard to source, aliexpress is your best friend. You can find every kind of mechanical part on there, nuts, bolts, bearings for fairly reasonable prices. Bearings similar to your large one are less than 5 bucks a pop. Since you're fairly concerned about size, I take it these things are intended to be quite small? At this scale and speed, bushings will do just fine, and save a lot of space. You'll find a lot of small gearboxes with plastic or metal gears use bushings not bearings on the shafts. Gives you more room on the ring gears to add weight/remove and balance them. You're only left trying to balance the wobble of the central gear.

The inside gear design is an amazing idea, it's solving a problem for me. Great job 😀

this could function as an extremly flat and simple clock designt, where you hav three layers, with ratios 60:1, 60:1 and so that every stage has a colck handle as output. and one could easily expand it to acount for days in a week, and the date as well, just by putting more leyers around.

Love the cycloidal drive videos Great job man

Great work Levi, thanks so much

Nice idea. Will work well with slow moving applications. Thanks for sharing 👍🏼

Jansen you did a great innovative idea keep it up.🙂

This is a beauty!

that's an insane gear reduction in such a small space. very clever!

You are a Genius. I am currently playing around with printed gearboxes and was searching for one for a 4th and 5th Axis for my selfmade CNC Mill. This might just be what I need if i get it done with low backleash. Thank you!

I don't know why anyone would down thumb you for any of your videos but I think they are awesome. Please keep up the good work.

I love the inside out concept! It reminds me a lot of what Tom Lipton (channel: Oxtoolco) shows in his 'wabble drive' videos from about 10 years ago. He winds up using the output to drive a scotch yoke to get a reduced reciprocating linear motion, and then fits it to a (from memory) printing press for his wife. Great stuff, keep it coming!

This is great, I will try to integrate it into my project!! :)

I made something very similar to this in March of 2020 except it was vertically oriented and the middle stage did rotate. It had an 812:1 reduction because It had a 28 to 29 lobes then 29 to 30 lobes. The torque output was incredible. I used small NEMA14 motors but I was able to get 100Nm of torque. It is fully fixed and constrained. It uses components from McMaster. The driving shaft is 8mm and I used M3 screws for assembly. The outer diameter was 100mm if I remember correctly.

Awesome video look forward to more

Sintered plastic bearings impregnated with oil have excellent load bearing and very reasonable friction against metal axils. SLS printed nylon bearings could save you lots of space, weight and cost.

Interesting idea. Thanks for sharing.

I can't wait to see this kind of gearbox avalible for equatorial mounts around or below 500 EUR for each gearbox. You would not believe how expensive equatorial mounts with high accuracy are.

Keep designing your micro servo! I designed a robotic triple drive years ago using 3 matched tubes, each inside the other for space constraints (1/8" OD was the largest.) The surfaces were teflon impregnated after machining. ( several materials can be fused to tge metal surfaces) Very low friction and good life on slow (

The single stage version of this would make a pretty good final drive for a model tank, one to each tread.

Great job!

Quick tip: While you are giving long explanations, have the animation running instead of stopped. That makes it way more interesting visually, and a bit easier to understand.

If you were 3d printing this cycloidal drive you can include custom bearings in that design. You could actually 3d print the whole thing in one go so that would actually make it really easy to manufacture that way.

Very good design.👍👌

Really clever mechanism! Much simpler (cheaper?) in some applications compared to an epicyclic gearset when trying to get big ratios from each stage.

Hi Levi, nice design. I have manufactured few planetary gearboxes for hydrostatic wheel drive transmission, they have high torque capability and are very compact. So cannot relay on the size of ball bearings and the avilable needle bearing won't withstand the very high radial load. The solution it's to design the planet gears and bearing as one piece, so basically you use the gear as the outer race of a custom cylindrical roller bearing (these rollers are readily available in lots of sizes and they are cheap) and the inner race would be the driving pin (in your case this would be the inner eccentric ring piece), both surfaces must be heat treated to same hardness to ballbearing races and precision grind (in high precision or high speed you may need to lap these surface too). The I.D. and O.D. of the outter cycloidal disk and the inner eccentric ring have to be designed to match an "almost" integer amount (around 1% - 2% bigger) of the rollers to be inserted (cage not used). This approach works great and if properly built, it will have the same service life as an off the shelf bearing. This same approach is also used for the high power/high speed gear stages in the trucks transmission boxes.

Good work

oW baybe! New upload!

1:299 is such a cheeky reduction.

Very nice design and clever. As for the bearings you might try making the parts out of material with natural lubricity and eliminate the ball bearings. Combining that with some grease might be all that is necessary particularly for low speed applications. If you want to stick with 3D printed materials you might try reinforcing the mating surfaces with brass. You could design so that the edges of the mating surfaces are close to the diameter of standard tubing and cut rings to fit. Since brass tubing can be expanded to larger diameters you could use the same tubing for both the inner and outer surfaces. You could build a cone shaped tool to expand the outer ring to the proper diameter. A little grease between the inner and outer brass might be all you need.

Gives me a neat idea for a 4th axis on my mill. imagine this in metal and driving a rotary table. 1st stage counter mass could be done by inserting metal pins into the black eccentric lobes. You could loose the bearing and machine the gearbox from cast iron with precision ground and polished "bearing" surfaces. Cast iron is already a self-lubricating metal, and friction is almost as good as a bearing when oil is used. Machined Teflon Delrin or nylon are also self lubricating but also can have long life when an oil is used. The crank in your cars engine is cast iron and uses tight tolerances between bearing surfaces and pressurized oil as the bearing itself. The metal parts never touch even though there's thousands of PSI pushing down on the pistons when the fuel air mixture is burned.

Great design. I’ve been thinking about this type of gearbox for a while but couldn't get the design to work. I haven’t seen this in any patients before. What do think is the smallest diameter of this type of gearbox?

Have you thought abiut using 3d printed IGUS drylin linear bearings?

I know this was posted long ago but i recently stumbled on your account and am loving the "straight up" or "straight forward" delivery. That being said I would love to see you look into the 3d printed bearing models on Thingiverse, some of them use actual ball bearings to further reduce friction. Great design nonetheless!

What about using bushings in combination with a tephlon sleaves ?

Also, would love to see you make something on your cnc

You can make custom bearings -SLS nylon, impregnated with oil (immerse and thermal cycle) on a polished metal shaft will excellent load capacity and reasonable friction.

Have you made one? What kind of efficiency do you get? Keep up the great work!

If you need small bearings, start by thinking "Japan". They make the mentioned bearings but also some that are 'microscopic'. Maybe you don't need ball-bearings, maybe you need roller-bearings. POM is often used in machines instead of aluminum or steel. This allow you to avoid using lubricants. One thing you should have in mind for simple bearings, is that you should not use the same material for the bearing as the axle. For instance: Use a bronze bearing and a steel axis to rotate inside the bronze bearing. Do not use a bronze axis, because the bronze will fuse to itself (absolutely insane example). Same thing about PLA. You can place a PLA axle in a PLA bearing, but rotate this axle at 3600 RPM and you'll find that the heat fuses the two PLAs together and the device stops working. If the bearing is a material that won't bind to the PLA, then the device will have a much longer life.

I saw a cross between a cycloidal and a split ring where two cycloidal disks are stuck together and the top on has 1 less tooth and the top can spin. Doesn't need those big carry holes through the disks and can get stupid high reduction with 3 parts.

you might consider carving a channel on the outside of the friction surface and placing roller bearings in there. Barrel style rollers with your very low friction. And you might be able to find thin steel bands to limit wear. it looks like the ideal elbow joint. It looks like the ideal elbow joint

Really interesting work! Do you think this could be translated into a transmission where multiple gears exist and could be selected as needed during operation.

could you instead embed individual roller bearings in your eccentric cam that would replace the single big ring bearing?

How big would the friction be in reality, if you skipped the bearing and used greased surfaces? If you used it for an articulated arm/leg, maybe it would still work?

Great video! Awesome design. The cycloidal drives look so cool running. If you can adjust the ratios and motor speed I think it would make a very cool wall clock with the different speed rings driving the clock hands. Toss in some color changeable LED lights and a glass front to show it all off. Respect to you. Good job.

Hi , you could turn a plain bearing from brass,bronze or some self lubricating stuff. Can you reverse drive it? Great work!!

I am trying to design a cycloidal gear setup in an application (just one stage in my case). I'm still confused as to why the four posts rolling in the larger holes are necessary. I'm also not clear on how to get the cycloidal gear tooth shape.

Very cool! I'd call it a fractal cycloidal. Although you could also use a planetary gear set in ring-output configuration for the inner stage to solve the vibration problem and perhaps make it smaller. Search "planetary gearbox micro stepper" on ebay for a potential source of tiny gears.

Hi, any chance you'll release the CAD files for this? Very interested in incorporating this into a robot I'm working on. Thanks!

Very interesting! What I didn’t get is how you would drive the input. I assume an asymmetric shaft on the motor?

In terms of space constraints, would our thin bearings or precision cross roller bearings give some help?

While theoretically feasible and very, very clever, I think this design would be a nightmare to implement. From what I understand, cycloidal disks' eccentric motion and thin cross-section makes them vulnerable to tilting and flexion, which jams them between pins and lobes, with forces concentrated near the faces. That means that the wear rate will accelerate as time goes on and leave the drive vulnerable to locking up. Without offset disks to counteract tilt and less-than-stellar fixation options, keeping the disks flat and level to prevent this would be tricky, at best. This design is also harder to free from sliding contact. The conventional design only requires rotating elements for the ring and output pins, whereas this design would requires that the ring pins, both sets of output pins, and the teeth of the inner or central cycloidal disks be free-rotating ... which, by the way, would make fixation of the inner/central disk even more painful, as you'd need to accommodate eccentric end plates.

I wonder if you could use a drive like this for the arm drives on a Delta Robot?

you can get custom laser cut 932 bronze bushing cut from oshcut if you think it can handle the extra friction.

Nice. Design

There is a need for a easy shift in and out, inline, 20 to 1 or higher, hollow shaft, 1to1 feed through reduction unit for manual input on milling machines. Machines where one turn equals .2", it takes a lot of tapping of handle to get the number you want on a .0002" digital read out. I will keep looking

I think instead of "counter mass" they add a second stack on lobes timed 180° from it's twin

would it be possible to replace the bearings with journal bearings? that should make it thinner

Couldn't you use rollers (like in a roller bearing) between the two plastic parts instead of a separate bearing or pure sliding friction? Would be low profile while also greatly smoothing the motion, essentially replacing the inner and outer race with your plastic parts

I know it's messy but what about a greased shaft instead of the center bearing? Lithium grease has great thermal operating ranges and decent longevity.

Nice design. I am looking for a 2500 reduction in as little stages as possible. This might be the solution. Maybe changing the colors to something brighter could help visualizing the whole thing. Thanks

It is indeed a compound gearset, in my opinion.

3d resin print graphite impregnated bushings. As they wear, more graphite particles are exposed, and they get even more efficient. Some experimentation is called for, lol.

What if you printed the parts connect by the bearing as 1 peice?

Oil pump gears can be use for the first stage . Some motorcycles as the old honda cg 125 has very small oil pump gears , 8 tooth outside and 7 inside , some honda cars also have rotary oil pumps in larger diameters . I m thinking on a very small two stage reducer , with very precise , very cheap metal gears.

Could the bearing be a slip bushing?

这个结构的摆线针轮变速器最大的特点有2个：1、同步针是固定的，因此同步针及其底部结构可以设计得非常的坚固，这是最大优点；2、外齿轮作为输出动力部分，直径可以做到很大，可以直接设计为法兰盘使用，输出极限扭矩非常大。这两个特点结合起来就成为了最坚固的摆线针轮变速器的结构！但是缺点也有2个：1、 内小齿轮与内大齿轮之间的空间紧凑，他们之间的轴承比较难选择；2、同步针及其底部结构，与外齿轮之间的轴承会过于庞大，可选轴承型号有限，这会大幅增加方案成本。

You could print a ball race into the eccentric inner and the wobbling second stage and feed the loose bearings into the space. Lazy Susans use this technique. They use a grub screw to close the hole. This would remove your large bearing and give you complete freedom to size this to suit need. The largest lazy susan I know of that uses this method is 1000mm o/d.

I don't know how I missed this video for so long. This flows with exactly what would help my current project. Am I correct in understanding that the final output ring won't be rotating around a fixed axis. If that's the case, I'm not currently smart enough to figure out how to have it match the perfect circle that is the platform base.

The only place you'll find bearings on a real cycloidal drive is on the very middle eccentric part and the output shaft. All of the pins and surfaces of the plate are polished very precisely and just ride on a film of oil. You don't need to make things complex with needle bearings. Just very precise holes and pins with a very smooth finish, and a nice heavy oil. And the only thing novel about this is that the two stages sit in the same 2d plane. They make multi-stage drives that are simply stacked along the shaft axis.

Did you ever try making this? I've seen a great deal of your videos lately and was just wondering. BTW: If you see this comment, could you comment on why no one making these are using bushings instead of bearings. Just thinking that the output of these devices are slow moving high torque. Traditionally that is where a bushing shines. And they can be much thinner than bearings. Just and idea. TY for sharing.

You could replace that big bearing with small steel balls held in place with a custom 3D printed cage. The inside and outside ring could be custom manufactured on a lathe. Well, not exactly easy to implement...

Use pin Bearings?

Проекты ваши с волновым редуктором мне нравятся. Отличная работа, однако не помешало бы сделать их открытыми. Это даст возможность совершенствования другими людьми и делится в сети

What is the equation of the outside(ring) gear?

i love this and it looks really fascinating, but if i understand it correctly it looks like the central shaft is oscillating. so that is not the motor shaft? if you are adding a pinion to the motor shaft to oscillate that central shaft that we can see, then there is actually a second layer below the assembly you're showing here, right? effectively a 3rd, 1:1 stage i suppose. it kind of negates the flat construction since you have to offset the whole thing from the motor housing by the motor shaft length. or am i missing something?

How can a motor drive it, when Center shaft (it is the motors shaft,isn’t it?) is wobbling. Doesn’t it need to be eccentric?

You din't need counter weights to stop any wobble. You just have to dobble the hight of the gear and rotate the middle 2/4 at 180°. I know because of the berings you efectively would have to increase the hight up to 4 layers so that you can use the 2 middle ones as counters. I mean to just eliminate the original wobble you only need to add a second layer and turn it to be 180° offset to the first one but that would create another wobble by itself because the second layer is in the third dymension also offset to the first one. But if you acheave that you wouldn't have to add any counter weights. And if the mechanism works how I understand it and the outer output ring just turns but doesn't wobble around (I mean isn't that the reason for this design?) you also could connect the outer rings to one big ring. Same goes for the gray inner inputs so that they somewhat look like the input shaft of a piston engine just that the counter weights for the offset inputs themselfes are another inputs for the other layers of the gear. In term of conecting the layers you can also combine the back inner output gears/rings to one part. Just 180° ofset because of them being the inputs fo the outer gears. The only parts not being connected to one part are the bearings and the white gears. And yes. this doesn't solve any other design issue with this concept. Tbh I think it just makes them more complex even though it technically doesn't add any. Instead of one floating and pressfit self supporting layer, you would have the issue that 2 (aka 4) layers would support and rub against each other in this simplistic design. And you are still limited by the berings you can get in terms on how big you have to make the gear. But I don't count them because they are already issues of the one layer design.

Sliding bearing instead of a ball bearing maybe? Or put the bearing on a plane behind the gear set?

You could replace that bearing by making the OD of the stage 1 output and the ID of the second stage input into a race and then run a frame and spacer (aka “caged”) supported bearing with 4, 6, 8 ball bearings in the space between them… and with the right filament and lube (think graphite)… you could potentially 3D print the ball bearings themselves

If the parts are cut in Teflon, you may be able to slide them against each other without the bearing

You might call it jank - but when I make custom bearings I use airsoft BBs and they do just fine for me (lightweight too) :P

smart guy!

Are you still designing the cycloidal disks the same way? I saw your video and I'm working on a way to make it parametric.

Hi Levi, Have you thought about decreasing the eccentric travel of the gears, thereby decreasing the height of each tooth? This might make the reduction a bit worse, but I think it would be a good way to make everything smaller, since it would decrease the size of the pin holes, the intermediate output ring, and the teeth.

I'd be most concerned about frictional losses. For your "outer" brg I'd think a bush would be better, cheaper and thinner only not as ridged.

there are not so much movement space?

I love the idea of reverting the rotating elements on the inside unit! By the way, this gearbox driving a second gearbox reminds me of another oldie and goldie: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-79OnADoQp8g.html. It's like "Yo, I heard you liked gearboxes so I put a gearbox inside a gearbox so you can gear down while you gear down" ;)

use it to build a single layer clock milli seconds, seconds, minutes, hours, days

What cad/Design software are you using?

I realize this is two years old, but could you use a bushing instead of a bearing between stages?

Multi stage, single plane cycloidal drive?

For the size of your assembly, why not just make the teeth / lobes smaller? Your problems seem to be that you want it to have a small diameter, made out of plastic and for fast input rotation. If you machine this out of metal with an oil flooded case, you can almost ditch the bearings, fix the vibrations by making parts bigger and adding holes on the heavy side. It is a very interesting concept.

why not make your output ring gears from the 1st stage out of delrin or other type of machinable bearing material. That would reduce the size by eliminating the bearing completely. You could also consider a thrust bearings on the output rings to improve axial load support and a low friction teflon shim underneath both wobbling components since they are not rotating.

But it's a fantastic design, very compact!

LoL !! AKA Radial Motor . Yes. Ok . TYVM for 19th Century help.

Great job! Become your patreon with no hesitation ;)