Automatic Spherical Continuously Variable Transmission (CVT)

[Hrafn]

I really like the idea of continuously variable transmissions, where the transmission can shift smoothly and continuously along a range of ratios instead of moving from one fixed ratio to another. Inspired by Zblj's differential-based CVT and its development by Nico71, I wondered if it would be possible to build a spherical CVT similar to the NuVinci CVT, using Technic ball joints as the spheres. For those unfamiliar with spherical CVTs, the NuVinci site has a decent explanation. The goal is the same as for other CVTs, but the mechanism is different.

The answer is: yes, such a thing can be built, but it is so limited in the torque it can transmit that it's of little practical value. In theory, one could get over the torque problem by gearing up between the motor and CVT's input, and then gearing down again between the CVT's output shaft and the wheels. In practice, I had limited success with this approach because of frictional losses in these two additional geartrains. However, I'm new to Technic and maybe one of you can solve this problem. If you do, please let me know!

There are plenty of other features of the CVT that could probably stand improvement. For one, the spheres contact the side of the tires, pushing the tires away from the sphere. Ideally, the spheres would contact the edge of the tires, keeping the wheels from dislodging sideways. However, in practice there are two problems. The first is that this is hard to do while building using parts with standard stud- and half-stud distances (8mm and 4mm). The second is that if the two wheels are too close to one another, their faces tend to rub together. I tried using Bionicle Zamor/Chima spheres, which are much larger and therefore might allow a better wheel/sphere contact geometry without having the wheels too close to each other, but didn't find a good way to get them to work.

I also tried a wide selection of small tires and wheels; none of them worked better than the wedge belt wheel and its tire, though maybe someone else can find a better solution.

An early version of the CVT was also an Infinitely Variable Transmission (CVT-IVT), at least in theory. The sphere could rotate enough that the radius of rotation of the sphere where the output wheel contacted it was zero; this gave, in theory, infinite speed reduction and infinite torque multiplication. In practice, the output sphere stopped rotating - and stopped transmitting any torque - well before the sphere reached the theoretical infinite-torque angle.

One happy accident I encountered while building the CVT is that while I planned for the shifting to be manually controlled, the CVT as built is actually automatic. As you apply a resistive torque to the output shaft, the spheres rotate of their own accord to slow the output and increase its torque. Note that if the tan axle pins in the pictures were not present, the spheres would keep rotating until the flat face of the ball joints (where the axle pin attaches) faced one of the wheels, which would then cease to be in contact with the sphere and cease to transmit rotation.

OK, on to the pictures. Since I don't have a Flickr or Brickshelf account to host photos, I can only upload 3 small photos in this post; hopefully they'll be adequate. For some reason they upload out of order, so the first one is of the CVT partially assembled; the second shows the parts needed; and the third shows the completed assembly. Note that 8 of the lime connectors are connected to the 5x7 frames using black friction pins; the central 4 lime connectors use 3/4 pins instead, with the 1/2 stud portion inside the lime connector. This is to prevent the pins from rubbing against the ball joints.

[aol000xw]

So the balls are prisioner between the lime connectors and rotating like ball bearings?

Edit:

Ok now I see how it works, it's genius.

Just an idea....

All rubber for increased friction.

The inner wheel needs free movement so something with round hole instead of that red hub. Don't have anything at hand at the moment.

Two of these one infront of each other, with some frame putting some presure on the big wheels so they keep 6 of the small "spherical" ones, prisioner.

I bet one stud of separation between the big wheels, so a liftarm can provide half stud of support for the axle of each half.

The small gray axle may need a bush on each side, to account for the wheel not being an sphere. Movement range and so ratio are limited.as it isn't a complete sphere.

Needs some development but I think it might work.

Edited August 17, 2013 by aol000xw

[Nazgarot]

I can't see how this would be a variable speed transmission. It seems to me like a fixed speed transmission (i.e. a linear clutch). A video would be very helpful.

-ED-

Edited August 17, 2013 by Nazgarot

[aol000xw]

Think of it as a disc than spins at a constant rpm. The angular speed is constant, but your lineal speed varies depending on the track you are. Slower when you are closer to the center, faster when you are far,

Now look at the sphere and think of it as a disc,, The sphere moves to compensate differences of speed between both surfaces, but it also transmits force (speed) from one to the other, trying to find the equilibrium, when its axis is perpendicular.

If its equatorial axis that in neutral is horizontal tilts, on one side forces a faster speed and a slower in the other

Well, if I understood the principles correctly...

[Hrafn]

So the balls are prisioner between the lime connectors and rotating like ball bearings?

....

Just an idea....

All rubber for increased friction.

The inner wheel needs free movement so something with round hole instead of that red hub. Don't have anything at hand at the moment.

Two of these one infront of each other, with some frame putting some presure on the big wheels so they keep 6 of the small "spherical" ones, prisioner.

I bet one stud of separation between the big wheels, so a liftarm can provide half stud of support for the axle of each half.

The small gray axle may need a bush on each side, to account for the wheel not being an sphere. Movement range and so ratio are limited.as it isn't a complete sphere.

Needs some development but I think it might work.

aol000xw,

Yes, the balls are kept prisoner between the lime connectors; they hold the balls firmly in place but allow them to spin.

Something like what you described should work. I wasn't able to get it to work myself, but the principle is sound. The key is increasing the coefficient of friction between the spheres and the discs. ABS and rubber simply don't stick well enough to each other. I tried increasing the contact friction by increasing the normal force (jamming the discs up against the spheres) but that just added friction that wasted energy.

I did try using some of those small wheels you show, but the fact that they're not spherical stymied me. The center of curvature of each side of those wheels does not go through the axis of rotation, so as soon as you tilt the axis, one side disengages from the wheel driving it.

I also was trying to get the CVT as small as possible so I could use it in a 1:10 scale vehicle I'm working on, but it may not be possible to get it to work at that scale and have it transmit a useful amount of torque.

Nazgarot,

Yes, a video would help. I'll have to sign up for a Flickr account and upload one.

[EyesOnly]

I found some videos.

[07jkearney]

Wow! You have done amazingly well to translate this concept into bricks so compactly

[aol000xw]

The problem is that in most CVTs friction is key for torque transmision, Friction directly of the components or indirectly througth inmersion in a fluid and in this case, the concept itself its mecanically complicated, way more than two cones....

[Kronos]

.....

[Hrafn]

Wow! You have done amazingly well to translate this concept into bricks so compactly

Thanks! It took quite a while to go from the original prototype (about 9x7x11 studs) to this. I just wish it could transfer more torque, though aol000xw is probably right - getting the right level of friction with Lego is difficult at best. The only spherical or hemispherical rubber Lego pieces I could find were the old-style Technic competition arrow heads, but I can't find a way to remove the shaft without damaging them, and the shaft is so long it would make the CVT enormous.

I've uploaded some construction photos and a short video to Flickr:

and here is the movie (I'm not sure how to embed it)

Edited August 17, 2013 by Hrafn

[Nazgarot]

Hi again,

I do understand the Principe, but I still see no gearing. Only a reduction in speed due to friction. There is no change in the radius in any way as the red "balls" are locked in place, and no way for the "drive rings" to change contact point with the balls. That makes this a clutch and not a CVT. A cvt is capable of changing the speed and torque of the output compared with input. This only slips (friction) when the output is held. There is no increase in torque.

I also tried to embedded the video. To embed a video you normally put ["media"] before a link and ["/media"] after it (without these " signs). But I can't seem to find a direct link to the video... Maybe upload it to youtube?

-ED-

Edited August 18, 2013 by Nazgarot

[cgg199]

I can only agree more. It does just appear to be a clutch.

[07jkearney]

Granted, in the video the mechanism does appear to function as a clutch, but in reality the gear ratio between the input and output is changed by variation in the angle of the red technic ball joints, and, as Hrafn stated above:

One happy accident I encountered while building the CVT is that while I planned for the shifting to be manually controlled, the CVT as built is actually automatic. As you apply a resistive torque to the output shaft, the spheres rotate of their own accord to slow the output and increase its torque.

[EyesOnly]

The fact that real thing exist that use this suggests that it does indeed work. Yet i too fail to see any actual gearing taking place.

[Nazgarot]

The fact that real thing exist that use this suggests that it does indeed work. Yet i too fail to see any actual gearing taking place.

But it doesn't, because it lacks a very important factor. There is no change in the contact point of the red "balls". if you hold a ball between two disks it doesn't help rotating the ball in any way as long as you don't change contact point on the disks. The ball has the same radius no matter what angle it is rotating at. This can only work as a gear if one disk changes contact point with the balls relative to the other (As can be seen quite clearly in the video of the bike CVT, look at how the contact points changes relativ to output speed). And with this construction it doesn't. Sorry...

But the idea is nice, though very hard to complete in this scale.

-ED-

Edited August 18, 2013 by Nazgarot

[EyesOnly]

Oh sorry you meant in the scale model. I thought you meant that even the real thing doesn't work although it clearly does.

[Hrafn]

But it doesn't, because it lacks a very important factor. There is no change in the contact point of the red "balls". if you hold a ball between two disks it doesn't help rotating the ball in any way as long as you don't change contact point on the disks. The ball has the same radius no matter what angle it is rotating at. This can only work as a gear if one disk changes contact point with the balls relative to the other (As can be seen quite clearly in the video of the bike CVT, look at how the contact points changes relativ to output speed). And with this construction it doesn't. Sorry...

But the idea is nice, though very hard to complete in this scale.

-ED-

If nothing else, posting this has made it clear to me that I am not cut out to be a technical writer, since I haven't done a good job of explaining the CVT.

Nazgarot, there is a change in the relative contact points, but it's very hard to see in the photos. If the wedge belt wheel tires contacted the sphere on exactly opposite sides, that is to say if a line drawn between the contact points went through the center of the ball, then you're right that the speeds of the two wheels would always be identical. However, because of the way the tires do contact the spheres, there is some difference in the speeds of the wheels. It's difficult to explain with words, so here's another video; this one is of a modified, cut-away version of the CVT with manual control of the sphere's tilting. Hopefully this video is viewable. Unfortunately, by cutting away much of the structure, I also made the CVT even less reliable than it already was, so it stalls a few times in the video. As you can (hopefully) see, the speed does vary. Torque also increases some as the speed drops, but I'm not sure how to show that, especially since the transmitted torque is so low anyway.

If you can't see the video by clicking on the link above, try going to my photostream: http://www.flickr.com/photos/100476839@N08/with/9541362712/. The video is titled "CVT cutaway in operation".

Edited August 18, 2013 by Hrafn

[Sardo]

seems that it works,as the pin changes the angle of the balls rotation and distance of the 2 disks

[EyesOnly]

Nice to see it in action. Well it looks to me like you need to scale it up a bit. Not sure what you would use as sphere but getting larger tires would give more torque through the gearbox. Maybe larger tires but the same spheres is all that's needed. How about Tire 43.2 x 22 ZR. They are just slightly larger but has a nice grip and are pretty common.

[Nazgarot]

If nothing else, posting this has made it clear to me that I am not cut out to be a technical writer, since I haven't done a good job of explaining the CVT.

Nazgarot, there is a change in the relative contact points, but it's very hard to see in the photos. If the wedge belt wheel tires contacted the sphere on exactly opposite sides, that is to say if a line drawn between the contact points went through the center of the ball, then you're right that the speeds of the two wheels would always be identical. However, because of the way the tires do contact the spheres, there is some difference in the speeds of the wheels. It's difficult to explain with words, so here's another video; this one is of a modified, cut-away version of the CVT with manual control of the sphere's tilting. Hopefully this video is viewable. Unfortunately, by cutting away much of the structure, I also made the CVT even less reliable than it already was, so it stalls a few times in the video. As you can (hopefully) see, the speed does vary. Torque also increases some as the speed drops, but I'm not sure how to show that, especially since the transmitted torque is so low anyway.

That was a much better video! And yes, I admit it works!

The reason it works seems to be the same as your problem with torque. The ball is not aligned on it's center. This gives the contact points on the ball a relative difference in speed. Very clever! But it also gives you a problem with torque as the ball is pushed out from the rings. In a real CVT of this kind the torque transmitted can be a lot higher because the ball always is contacting the "friction surfaces" through it's center, so you can apply a lot more force from both sides of the ball without losing much power to friction. I'm afraid that you are at a limit of what is possible with lego at this scale the way you have done it. I would however very much like to participate in a community build to make a working CVT on a slightly larger scale.

Your work is a great inspiration as I have never before attempted a CVT of this kind in Lego before. Thanks!

-ED-

EDIT: It might be possible to realize this without using balls. I need to draw a little... No, it wasn't. But I might have a different idea. It needs a lot of work, and will be bigger, but it should be able to transfer more torque.

EDIT 2: There is a old rubber "balloon" tire from the old Space line that might come in handy, Link It might let you increase torque in your setup if you can modify it to fit the tire.

Edited August 18, 2013 by Nazgarot

[EyesOnly]

I'm stumped at the moment for solutions but it would be nice to have someone else building it. Balloon tires, even larger ones seem like a good idea.

[Hrafn]

That was a much better video! And yes, I admit it works!

The reason it works seems to be the same as your problem with torque. The ball is not aligned on it's center. This gives the contact points on the ball a relative difference in speed. Very clever! But it also gives you a problem with torque as the ball is pushed out from the rings. In a real CVT of this kind the torque transmitted can be a lot higher because the ball always is contacting the "friction surfaces" through it's center, so you can apply a lot more force from both sides of the ball without losing much power to friction. I'm afraid that you are at a limit of what is possible with lego at this scale the way you have done it. I would however very much like to participate in a community build to make a working CVT on a slightly larger scale.

Your work is a great inspiration as I have never before attempted a CVT of this kind in Lego before. Thanks!

Thank you! I'm glad the second video worked. I agree that the geometry of where the tires contact the sphere is an issue. In the NuVinci design, they sandwich the spheres between outer wheels and an inner idler wheel; that geometry probably will not work with Lego. Barring that solution, I'd like to see the wheels much closer to one another, so that the force on them from the spheres is mostly perpendicular to the axes of rotation of the wheels. That would keep the force pushing the wheels outward to a minimum. It would actually increase the force pushing the balls outward, but in practice I've found it's easier to keep the balls in place than to keep the wheels in place, at least with these particular wheels. Unfortunately, so far I've found that if the wheels get too close they tend to rub against each other and the whole mechanism reverts to what you originally thought it was - a kind of linear clutch.

EDIT 2: There is a old rubber "balloon" tire from the old Space line that might come in handy, Link It might let you increase torque in your setup if you can modify it to fit the tire.

Yes, aol000xw mentioned the same part previously, and I did get some from Bricklink while searching for alternate solutions for the CVT. Those tires have a bit more grip than pure ABS, but much less than the current, soft tires do. They also have some disadvantageous characteristics. For one, they are not truncated spheres, so when you tilt them around their center point, both sides don't stay in contact with the wheels. For another, they are odd sizes - not quite 2 studs tall - which makes them difficult to keep in position using standard stud- and half-stud-length pieces. Still, they're probably the best bet I can think of.

Nice to see it in action. Well it looks to me like you need to scale it up a bit. Not sure what you would use as sphere but getting larger tires would give more torque through the gearbox. Maybe larger tires but the same spheres is all that's needed. How about Tire 43.2 x 22 ZR. They are just slightly larger but has a nice grip and are pretty common.

Thanks! I'd actually prefer to keep it as small as possible, so I can use it in a vehicle; but since the small version doesn't transmit enough torque, maybe I do need to make it bigger just to get a version that works better.

I do have some of the 43.2 x 22 ZR tires, and have been experimenting a bit with them them. No real successes yet, but they're promising, and while they're wider than the wedge belt wheel tires, they are sturdier.

The existing spheres are pretty good, though they vary to a shocking degree - some are noticeably wider or rougher than others, and some have prominent or misaligned mold lines that make them unsuitable. I tried using tow balls, which are even smaller - but they're too small, and the mold release divot on the end makes them lose contact with the wheels too easily. I also got a few Zamor/Chima spheres (which are about 2 studs in diameter), hoping that they'd be softer and "grippier", but sadly they seem to be made of the same hard, slick ABS. Maybe some of the sports balls / GBC balls would work? I don't have any and don't know what they're made of.

Also, if anyone knows a "purist" way to get the rubber end off of old Technic competition arrows, let me know! Those things are pretty close to being hemispheres, and they're very "grippy."

[Doc_Brown]

WOW that is very impressive, first as the real thing and then as a Lego recreation!

I hope a larger version can be made, seems like we need bigger balls to maintain the torque.

Thanks for sharing!

[EyesOnly]

How did you tilt the zamor balls? I though that whatever used as a sphere had to have an hole in it.

[Hrafn]

How did you tilt the zamor balls? I though that whatever used as a sphere had to have an hole in it.

That is a big problem with them, yes. There is a very narrow hole in them but I haven't found a part that would engage that hole - it's much narrower than a bar. The sports balls would have the same problem.