Electric motor CVT for experimental unicycle

John Dingley February 2017

I have been building self balancing electric unicycle motorbikes for a few years. One problem is that for going uphill you need a low gear, but to move faster on level ground you ideally need a higher gear. When you see TV programmes on Tesla cars they always make a big deal about how the motor can pull at maximum torque from any speed. However for a cheap brushed motor as I use, this is not the case, it is actually happiest at about 2500rpm and will draw a huge current if stalled or going very slowly against a high resistance. I have been offered much more powerful specialist motors and motor drivers at up to $1500 (without battery pack) but that completely misses the point. I am trying to build machines at a reasonable cost not regardless of cost.

One way to resolve this problem, also seen in 49cc two-stroke scooters, which are also a bit under-powered and deliver their power in a rather all or nothing fashion, is to use a Continuously Variable Transmission. Many of these on petrol scooters come as part of the lower half of the engine casting, but there is one that comes as a stand-alone unit, which is the one I am using here. An example from a China is shown above but mine was $64 from here: http://www.pocketbikesunlimited.com/CVT.html

They were really helpful and shipped it to the UK with no problems. Some places call this versions a CVT2.

I have conducted a sequence of logical experiments to investigate how to modify this to work at lower speeds, with partial success so far and I hope it helps somebody else. These experiments are ongoing.

This is the unicycle project and the machine I am hoping to improve upon. Ideally it will automatically move into a low gear for hills then, speed up into a higher one for going faster on the level:

This video shows how this exact CVT works when coupled to a petrol scooter engine.

The CVT is designed a for a more powerful (petrol) motor than our electric one, and it is designed to change continuously up through the gear ratios when the motor shaft is spinning at around 5-6000rpm. Therefore my CVT will need some experimentation and modification as the electric motor is happiest at around 2500rpm.

There are lots of youtube videos on how to dismantle a scooter CVT. The one I used is designed for "pocket bikes", which are a particular type of micro-motorbike or mini-moto used in the US.

The shaft or spindle that the clutch drum is attached to, that the motor is meant to attach to, needs to be removed from the alloy housing with its two bearings, and a new one made by an engineering shop with a longer 12mm diameter input as shown above (approx $20). This means I can connect it to the 12mm diameter shaft of my electric motor using a flexible coupling.

New spindle in place. The circlip holds it all in place inside the alloy cast housing.

The drive pulley (attached to the motor) has an outer cone that can be unbolted. The inner cone moves in and out, squeezing the drive belt from centre to outer edge to change the effective gear ratios. It does this in response to a ring of weights that are flung out as motor speeds up. This forces the inner cone closer to the outer cone. I may try heavier weights at some point as this might make it "engage" at the lower rpm speeds of my electric motor compared to the two-stroke this CVT was designed for.

The motor needs to be held somehow in a frame that locates it in correct position relative to the CVT gearbox. To begin I made a frame for the motor that allows two 6mm bolts to go into holes on each end of the motor that are already drilled and tapped.

Next I made flat strips of metal that located on available threaded holes (originally for the plastic cover) variously placed around the edges of the CVT casting.

The edges of these flat plates were then welded up to the motor frame, which had now become a kind of L shape.

Here you can see the flexible coupling between the 12mm shaft of the motor and the new 12mm diameter input shaft that was specially made for the CVT gearbox. As it is almost impossible to get the motor and input shaft of the CVT perfectly lined up, it is best to use a coupling with a tiny degree of flexibility in it, else the bearings will prematurely wear out in either the CVT or the motor. On the CVT driven pair of cones there is then a further 1:3 reduction gearbox to the 420 chain sprocket that will drive our wheel via a chain.

On the outermost of the driven pair of cones, there is a spring which pushes the outer one inwards onto the inner one. This means at rest, the drive cones have the belt located on inner part of the cones, while the same belt rests nearer the outer edges of the driven cones. As the motor speeds up from stationary this situation gradually reverses causing the box to run from a low gear to a high gear but it does so continuously and seamlessly.

Here I have fitted a weaker spring to help the CVT in effect change up the gears with a motor running at 2500rpm rather than one running at 5-6000rpm as was originally intended when it was designed.

The end result is a new experimental lower frame for the unicycle. The drive chain can be adjusted as the whole CVT assembly pivots along it rear edge, using an adjuster bolt and nuts to lock it into the correct position.


More experimentation to follow......

Here it is during first bench test mounted on its side.

Closer view of the cones operating and in effect changing gear.

This works fine with the weaker spring fitted onto the driven pulley but at slow speeds the pulley starts to slip so it is not the full solution.

23rd February 2017. Finished well enough for a first test run. I am using upper half of the previous version of the unicycle but if this works I will make a new upper part with larger batteries.

I tried a smaller compression spring on the driven pulley to make the whole CGT change gear at lower motor speeds. This worked but then it slips at low speeds. That might still be OK for a scooter, giving a clutch effect as you move off, but it is no good for a self-balancer as you will fall off! I tried putting fibre washers in as shown to bring the alloy cone nearer to the outer cone of the drive pulley at rest to reduce this slip at slow speeds but it only worked a little bit. I could put a spring in here to tension the alloy cone against the outer cone even at rest, to grip the belt better at slow speeds, but the available gap is very small.

Here is another view of the fibre washers experiment.

So, next option is to keep the standard driven pulley spring and make the weights really heavy on the drive pulley so they are flung out at relatively slow motor speeds. This is not some tweak like you see on moped tuning sites where they add a couple of grams to each weight, I want to add loads of weight to each roller. In the end I used a very heavy low melting point metal alloy sometimes used by product designers. You can cast it straight into a 3D printed mould for example without melting the ABS plastic. It is rather toxic I have been told so be careful if you do this. Also for a petrol motor this would be no good as the ambient heat from the motor would melt it out straight away. However, mine is electric. I melted it into the hollow weights just like melting a candle. See video below. I then sanded the ends smooth and replaced the weights. They are much heavier than standard.

25th February 2017. Did this work?

It worked partially. On a test stand the motor at full power the CVT will now change up to about 2/3 of its maximum range, with the regular driven pulley spring in place (not the weaker one I tried earlier).

However on the ground at regular speeds, i.e. not top speed all the time, it does not change up the gear ratios much.

With the weaker driven pulley spring and these heavy roller weights, it would work great EXCEPT for the fact that the belt then slips at low speeds.

Next experiment: What might work is, heavy weights, weaker driven pulley spring, PLUS tensioning spring that pushed alloy drive cone outwards so it grips the belt even at slow speeds (better then the fibre washer solution).

Watch this space for more updates......