DIY (Mini) ebike/moped -> square tube frame, hoverboard motors - does this look reasonable? Any obvious oversights?

Okay, the rear axle is now clamped in and the wheel spaced properly, just need to cut the left over threadrod sticking out.

Ive also mocked up the motor position.
nNYbkXG.jpeg

Ill weld in a bit of square tube against extant one to which the kickstand is attached, leaving a bit out so there is space for the chain.
Then add another one in front, where the ruler, sawblade and odd wrench are supporting the forward motor mount.

ADDITION: Having looked at it some more i think i will be turning the motor and rear wheel around, putting the sprockets towards the other side as that will let me run the motor cable along the frame a lot nicer.
 
Last edited:
Actually the range for magnet amount goes all the way from 1 to 100, which i think will let me get it to divide down to the correct speed with some calculation/experimentation.
Surprising range, since motors never contain more than six.

I don't understand your plan. Your motor doesn't output RPM data in a form that a geared motor controller can use. it does in the form of hall signals, but controllers for geared hubs don't use these signals for speed calculation, because they have to work with motors with varying gear reductions, and be able to provide speed data even when freewheeling, meaning the RPM signal has to come from the rotating hub shell, which is achieved by embedding (between one and six) magnets in the shell, and conducting the signal on an additional wire (additional to phase and halls I.e. a ninth wire in the motor loom.

Without such a signal from your motor there's no need to trick the controller ... you would input the accurate wheel size, set the magnet number to "1", and attach "1" magnet to your rim. Unless I'm missing something.
Wheelbarrow size, yep :)

Asking about 30cm rims earlier, how big did you think i was building this thing? Wheelbase is about 1m so overall length about 133cm from the front of the front to the rear of the rear wheels.
You mentioned 16", and I thought you meant bicycle size, which are 305mm.

1m wheelbase is same as regular bicycle, so it's not a miniature vehicle, just has smallish wheels.
Most cheap tires like this are only specced for 4kph, less pressure and are usually "2pr", which i gather is the amount of layers of reinforcement, these i picked are 4pr.

I did look up the manufacturer specs, dont recall the name but these are specced for 250kg, 30kph and 25psi max so i can pump them pretty firm for decreasing the rolling resistance somewhat, tho i wont be going rock solid as i want some cushioning of the ride.
They sound good.

I suspect the speed ratings relate to heat. Crappy stiff tires overheat. So, 30km/h rating is encouraging regarding rolling resistance - It can't be too bad for the tires not to overheat at 30km/h.

Got the bolts shortened and the sprocket spaced in place:

Not sure about how bolt spacings are usually measured or what BCD (Bolt center distance is my guess) is but the 42mm was the given measurement from hole center to hole center, not relative to the axle, and there were sprockets listed with those same numbers so i paired them up.
Bolt circle diameter.

If these are typical cheap sprockets sourced from China, they're roughly made. Unlikely eccentric, just that the drilling is imprecise, meaning you should expect the chain to flap.

You've got T8F chain and sprockets do you? I haven't used it, but expect it should be better in all regards to quarter inch #25.
Now working on the "new" mount for the rear axle, re-using the axle clamps form the hoverboard wheels.
Figured that i can use some of the left over aluminium pipe i used to make spacers for the front wheel as shims around the threaded rod axle to make it clamp in firm.
The pipe is only like 1 or maybe 1.5mm wall so it squishes down when the clamp is tightened and as the wheel is on bearings this will be alright. Might make something out of angle iron to bolt in at the back at a later date as it would be more sleek, this is easy for now.
Looks super strong.

Threaded rod lets the team down though, obviously. It's 12mm? Minus thread depth closer to 10mm? Thinner than you'd normally expect for such a long axle. And you'd have something like 50mm each side unsupported between wheel bearings and dropouts? I don't know if that's enough for flex to be an issue. And could threads gouge the bearings? Again I don't know, doubt it since they're hardened whereas the threaded rod is mild steel.
 
Your motor doesn't output RPM data in a form that a geared motor controller can use.
It does output something because i get a speed readout on the display with when revving the motor.
As it doesnt know how many magnets there really are and is probably measuring hall sensor feedback, i think that if i tell it there are more magnets than there really are, then that would make it report a slower speed, and the opposite.

1m wheelbase is same as regular bicycle
Huh, well i though bikes were a bit longer, its low anyways, lower than the small bicycle i took the front forks off.

just that the drilling is imprecise
Now that its on there and i can spin it on the axle and look at the sprocket, it seems to spin true, time will tell when its all together.

You've got T8F chain and sprockets do you? I haven't used it, but expect it should be better in all regards to quarter inch #25.
Yeah, T8F, no idea how thats measured either, is quarter inch #25 regular bike chain?

Threaded rod lets the team down though, obviously. It's 12mm?
M12 yeah, i guess it could wear but that stuff is cheap and if i get too unhappy ill figure something out, there are hardened axles commonly used for minibikes around, no idea if they do M12 or just hamburger units.
 
I've had controllers where they expect you to multiply the motor magnets by the gearing in a setting somewhere. So 20 motor magnets times gearing that spins the motor 5 times per one turn of the wheel = "100" magnets, 3 motor spins per wheel spin= "60" magnets, etc.. Combined with wheel size setting, allows determining speed without a separate speed sensor. No need to tell the controller the motor's gearing.
 
Got the motor mount welded in and drilled so i could install the motor, it looks pretty much exactly like the last pic, minus the ruler/wrench, plus 4 bolts.

I switched out the bullet connectors on the controllers for abiko crimp on rings, like what are on the motor cable and used the small junction box with screw posts to connect the two togeter.

Then i tried to ride it.

But as soon as there was some load the motor started stuttering.

So i guess the controller might not have survived the previous motor burnup after all.
No load, spins fine, some load, even just turning the wheel via the chain, without it touching ground, stuttering starts.

The power/stutters were enough to snap two of the three bolts holding the rear sprocket on.
Welded a couple of nuts to the broken off bolts still stuck in the mount holes and was luckily able to get them out.

So i guess its time to order some quality bolts, make a more substantial spacer and order another controller.

On controllers: Can i check if the one i have really dead?
If it is dead: Whats the thing to get? Ive seen vesc based ones written about and what i read is interesting, but are there options i should be looking at?
 
Last edited:
Bit of excitement.

You only used 5mm of packing washers behind the sprocket I think you said. I would've thought that would be acceptable. Obviously not though. Still, might pay to use 12.9. I've never broken an 8.8 (although haven't tested their strength with full stall torque, and use smaller wheels than you).

Regarding your controller, you're sure you got the hall phase combos correct?

It's great that controller can do your speed calculations. To keep it legal (<20km/h) it'll need to. If you get around to setting that up, you're motor has three pole pairs, so six magnets. Multiplied by your gearing 76:11 or whatever it was, you'll come up with a number <100.

There's plenty of good controllers. VESC are tempting. I think they are the future. But currently expensive, and a bit delicate. Having already blown up a motor, and potentially a controller, I think it might be one kick in the nuts too many for you to then cop a blown up VESC on top of that.

Best value controllers I've found are Votol. They're ridiculously cheap for the quality of hardware. All you'd need is an EM-30, but might as well spend an extra ten dollars on an EM-50, because they're almost identical form factor. They're easy to program when you understand them.
 
But as soon as there was some load the motor started stuttering.

So i guess the controller might not have survived the previous motor burnup after all.
No load, spins fine, some load, even just turning the wheel via the chain, without it touching ground, stuttering starts.
Is it using the correct phase/hall combination for that motor? If not, you could have a "false positive" where no load appears to spin normally, but is probably faster than it should be (with less torque).

If the controller has a self-learn function, you could try that. Otherwise, manual testing of the various combinations (there are six, and six "rotated" versions of each of those, for a total of 36...but you only need to test the primary six. (there are various threads to show you testing methods, some wiht a flowchart and the combinations of UVW phase vs ABC halls (or using colors like BGY-BYG, etc.).

You want the combination with the correct motor RPM for that system votlage and motor kV at full throttle, with the lowest current draw, minimum (and "normal") noise. But start each test with just a tiny amount of throttle, so you don't break things (motor, controller, etc) with wrong combos.


If it doesn't work regardless, then the controller could have blown FETs, or damaged gate drivers, or even more likely, damaged hall inputs to the controller MCU (if phase voltages shorted to halls).


There are also some controllers that just don't do low-pole-count motors very well.
 
Regarding your controller, you're sure you got the hall phase combos correct?
That im not actually sure about, with the hoverboard wheel i had to colormatch wires taken out of the hoverboard wiring harness and stick them individually into the different connector on the controller i have.
The new motor had the right connector so i just assumed it would be some sort of standardization going on.
Had a look and the hall wire cables line up across the motor/controller connector.

I dont really mind having blow the hoverboard hub motor, even if it had held up, the torque it was putting out wasnt enough for going up any sort of gradient, those motors are 6 inches in diameter with the solid tires so expecting them to perform with 13 inch tires was quite optimistic.

Best value controllers I've found are Votol.
The votols look nice but the batteries i have are only 36v 11.6ah and probably wont do more than 1C output as they were made for 250w ebikes.
I could run them in series for 72v but the motor i have is 48v and id prefer not to be overvolting.
 
Fair enough.

The standardisation for motor wiring is use blue yellow green ... in no particular order.

You're right, I think 48V is the listed minimum for Votol. Although I wouldn't be surprised if it still worked on 36V. You can program the LVC to any number you like.

1C could be a little pessimistic for your cells. If it's 4P, using 2900mAh cells you could likely push them harder.

Also, since you have 80 cells (10S8P), you could rebuild the battery as 48V (13S6P). 20A for a 6P pack of 2900mAh cells barely exceeds 1C.

It's not that big a job, since you're only butchering one pack. It's four cuts, and eight joins. If you keep enough nickel strip in tact when bisecting the pack you can get away with soldering instead of spot welding. I've done it with tool batteries.
 
It's not that big a job, since you're only butchering
Thats one part im still staying away from as ive accidentally shorted RC lipos, not to the point of venting/fire but, they turned into really quite spicy pillows.

I guess they might do more than 1C but the display was showing a noticeable drop in voltage, the stack of dots went from five down to two or three, as soon as i tried to go up an incline when i was test driving with the frankenhubmotor.

Havnt gotten better bolts yet but ive made a more substantial spacer which gets sandwiched between the 64t sprocket, which is mounted against the hub, and the 74t sprocket, should be way more stable than just standoffs as its circular and goes some way outside where the bolts are, i suspect i was getting some flex or sideways motion with just the three points of contact.

If the controller has a self-learn function, you could try that.
Ive got it back together with the above described spacer and its able to run the self learning function, also spins up quite well but it seems the drilled holes in the sprocket are ever so slightly off center, as someone mentioned was likely, im getting some chain slap as i increase the throttle.

Gonna see if i can whip up some sort of chain tensioner.
 
That im not actually sure about, with the hoverboard wheel i had to colormatch wires taken out of the hoverboard wiring harness and stick them individually into the different connector on the controller i have.
The new motor had the right connector so i just assumed it would be some sort of standardization going on.
Had a look and the hall wire cables line up across the motor/controller connector.
There's no real standardization, generally.

Certain controllers and motors that use the "same" connector, such as the fully-sealed waterproof types, will at least use the same pin set for phases, and the same pin set for halls, and the same pin set for other sensors (if any).

However, the motor could have the other end of that cable wired so that the blue phase goes to the "first" winding set, the blue hall to the second, and the controller could have it wired so the yellow phase goes to the first winding set, and the green hall to the second, so that nothing on them matches up. (or any other combination).

That's where the self-learn functions in the controller are pretty much required, as there isn't any way to manually swap the wires around without opening up either motor or controller and cutting or desoldering wires, etc. Many controllers with the WP types of plugs have this function, probably for that reason. Not all do....


Color matching at the connector doesn't work, for the above reason, as well. You can't know what order the wires are actually in within the motor or the controller

(controller is relatively easy to check, but you'd have to figure out what order the windings are done in the motor, and which order the halls are in relative to that. It's usually easier to just manually swap the wires in the connector (for non-WP types) or use the self-learn function if it has one. ;) )
 
Yes, self-learn for the win.

Pity your batteries are weak. That sounds like serious voltage sag. Might be weak parallel groups. Have you checked the balance? Termination voltage around 42?

Sorry to jinx you on the chain slap thing. All the sprockets I've come across are like that. I felt it worth mentioning at the time, while you were still mounting the motor.

I make spring loaded jockey wheels out of roller blade (inline skate) wheels. They need a good high speed bearing. ZZ, not RS. Anyway, you get heaps of spare bearings out of a five dollar set of skates from the thrift shop. I pop rthe bearings out, clamp the compound wheel in a drill, cable tie the trigger, then hold a sharp chisel against the spinning wheel to carve out a deep channel the width of the chain. Takes a little while, but it's the best I can come up with. You want a goldilocks compound, not too hard, not too soft, to provide a little damping, which the skate wheels provide. Rock hard plastic is too loud.

As to your spacer for the sprocket ... sounds excellent. Probably looks excellent too, but alas no photo :-(

I use old worn sprockets for spacers. Not as good as what you'll have. Heavy too.
 
Pity your batteries are weak. That sounds like serious voltage sag. Might be weak parallel groups. Have you checked the balance? Termination voltage around 42
Something like that at full, i havnt taken them apart to check, just opened the ends with the connectors to replace them with xt60s.

Sorry to jinx you on the chain slap thing.
Not your fault :)

I make spring loaded jockey wheels out of roller blade (inline skate) wheels. They need a good high speed bearing. ZZ, not RS.
Ive only got some RS 608 bearings laying about so ill give those a try.
Remixed an idler, changed it from press fit to have a groove that fits the bearing in the center and printed it out of A95 TPU.

image.png


Dont have any reasonable springs around either so first try will be with just PLA springiness, if its not enough im going to print TPU "springs" to install.
 
That's where the self-learn functions in the controller
It seems to spin up well with the chain tensioner so i guess it might just have been the slappiness playing tricks when i tried to test ride on friday.

Still refining the design to position it better and make it a bit stronger.
 
Heres how the chain tensioner turned out:

The springiness of the material keeps it pretty taut, its mounted by a longish bolt in one of the motor mount holes with a nut counter tightened against the face of the motor so it can pivot freely on the bolt itself.

Its made from two of these that are bolted together with M3s through the holes in the centers of the bearing posts.

SxBNNTv.png
 
I have concern on the long-term strength of the print material, while simultaneously applauding your clever tensioner design. It looks like even under the slow speeds shown in the video there is a stress cycle of movement on the PLA. Id suppose that will eventually weaken and cause the arms to loose their rebound and integrity.

Maybe you could adapt the print with a tunnel or bore through each tensioner arm for inserting and gluing thick gage steel wire strength? Maybe you could also add an off the shelf or salvaged spring connected between the ends of the two tensioner pulley/bearing bolts to maintain the 'pinching' tension force?
 
Last edited:
Maybe you could adapt the print with a tunnel or bore through each tensioner arm for inserting and gluing thick gage steel wire strength? Maybe you could also add an off the shelf or salvaged spring connected between the ends of the two tensioner pulley/bearing bolts to maintain the 'pinching' tension force?
Both are very possible, the latter is what i had in mind if the PLA itself didnt have enough spring, printing TPU strips with mounting holes in each side and clamping the two under a washer on the M3s on each side.

I will have to see how it behaves when ive gotten to riding it a bit.

Now that the chain slap is sorted and it can go to higher RPMs i decided to sit on it again to see if it will take off.
It doesnt, as soon as i give it any throttle the motor "cogs", feels like chain skip but is likely the magnetic fields moving faster than the rotor.

Hall/sensor damage on the controller from the burnup as it seems to be ignoring the motor position/feedback?
Seems like a possibility as the phase and hall wires were shorted on the hub motor.

Same with just holding it upright and trying to have it drive just itself.

I tried taking it outside too and it can propel itself if given a push start, as long as too much throttle isnt given.

ADDITION: Ive thought some more about the tensioner, i have an idea for a different design using a single roller, with a hinge and using a spring instead of relying on the printed material for rebound, i might have a couple of springs left over from RC kits and one or two of them might be beefy enough, perhaps when put together.

Either way i think having the roller a bit furher back, towards the middle of the bottom chain run, would be better support and require less tension to do the same job.


Started mocking it up in 3d but doesnt feel like i have enough space for it, will perhaps revisit it at a later date.
 
Last edited:
The only thing I would definitely change in the tensioner for longterm strength is to remove all the stress riser points in the arms, by sculpting all of these sharp edges out, turning them into smooth curves that go all the way from the cylinder full diameter down to the thickness of the arm itself.
proxy.png
 
Now that the chain slap is sorted and it can go to higher RPMs i decided to sit on it again to see if it will take off.
It doesnt, as soon as i give it any throttle the motor "cogs", feels like chain skip but is likely the magnetic fields moving faster than the rotor.

Hall/sensor damage on the controller from the burnup as it seems to be ignoring the motor position/feedback?
Seems like a possibility as the phase and hall wires were shorted on the hub motor.
Have you verified it's the correct phase/hall combo yet, as previously described?

If not, you *must* do that first, as if it's not, you will not get the right motor behavior even if everything is operating correctly. :(

It's even possible for wrong combos to overheat motors, wiring, controllers, to the point of damage (burning out windings, melting phase wires/etc together, etc).
 
by sculpting all of these sharp edges out
I guess it could do with some fillets but it doesnt feel like a priority for now, with the way the parts were laid down there are continuous lines of filament going all the way through, ill redo it when i need another due to breakage or "pinch" gets weak.
iXHnRKE.png


Have you verified it's the correct phase/hall combo yet, as previously described?
(controller is relatively easy to check, but you'd have to figure out what order the windings are done in the motor, and which order the halls are in relative to that. It's usually easier to just manually swap the wires in the connector (for non-WP types) or use the self-learn function if it has one. ;) )

I hooked up the learning line with the wheel suspended, once hooked up it begins turning smoothly, if disconnected and connected again it turns the other direction, did this prior to trying to run it last night.

So i think the phase/hall combo ought to be as correct as self learning can get it, and that the cogging is happening despite it.

That it spins up under no load and cogs when there is some seems to me like the controller ISNT actually using the sensor data to modulate the pulses its sending, allowing the phase shifts to move faster than the, slowed by load, rotor.
 
Last edited:
I guess it could do with some fillets but it doesnt feel like a priority for now, with the way the parts were laid down there are continuous lines of filament going all the way through, ill redo it when i need another due to breakage or "pinch" gets weak.
Hopefully it won't break, but that's usually where things like this do break, when they do. It'd be more important at the bolt-on center frame mount area than the roller mounts, as that's where most of the cyclic bending load stress occurs.


I hooked up the learning line with the wheel suspended, once hooked up it begins turning smoothly, if disconnected and connected again it turns the other direction, did this prior to trying to run it last night.

So i think the phase/hall combo ought to be as correct as self learning can get it, and that the cogging is happening despite it.

That it spins up under no load and cogs when there is some seems to me like the controller ISNT actually using the sensor data to modulate the pulses its sending, allowing the phase shifts to move faster than the, slowed by load, rotor.
That sounds likely. It could also be the halls in this motor are not neutrally timed (so it's made to spin better in one direction than the other, by accident or design), or that something is wrong with one of it's signals....but we've seen controllers damaged from the phase/hall short issue such that they can't read the halls anymore. :(
 
It'd be more important at the bolt-on center frame mount area
Out be the rollers the arms is 4x4mm, widening a little towards the middle of the arm, as can be seen in the slicer preview.
Then as we get closer towards the bolt-on the arms are wider, as can be seen in the 3d file view.
I did include a bit of thought in the making of the file, just not enough to do fillets ;)

It could also be the halls in this motor are not neutrally timed
After work today ill try self learning it to run backwards and then sitting on it and giving it some throttle to check if it has power, or cogs, in that direction.
 
Out be the rollers the arms is 4x4mm, widening a little towards the middle of the arm, as can be seen in the slicer preview.
Then as we get closer towards the bolt-on the arms are wider, as can be seen in the 3d file view.
I did include a bit of thought in the making of the file, just not enough to do fillets ;)
The main problem isn't the thickness of the arms, it's stress risers which exist just about anywhere a sharp change in angle exists in a structure that flexes.

That's what breaks shafts and axles with circlip grooves, for instance. Even a tiny cyclic flexure induces stress at those points because taht's where the thing stops flexing, instead transferring the energy change (terminology?) to the material at that point.
If it doesn't break within the useful lifetime of the device it's part of, it doesn't. matter... :)
 
Back
Top