APL's DIY axial-flux motor

[APL]

This article has some relevance to open and closed gaps. May be worth a look.
https://eprints.lib.hokudai.ac.jp/dspace/bitstream/2115/67264/1/Final_text.pdf

 

[fechter]

This article has some relevance to open and closed gaps. May be worth a look.
https://eprints.lib.hokudai.ac.jp/dspace/bitstream/2115/67264/1/Final_text.pdf

Their design is pretty close to the latest drawings here. Having the semi closed slot makes a huge difference.

 

[APL]

Yes, I was amazed at how close it is in size, so theres a lot of good information here. Not that I understand all of it,..
definitely not. I didn't read it word for word, but what I got out of it so far is that tooth overhang, or 'brim' as they say,
is more efficient at low speeds, which is what we need, and 'no brim' is more efficient at high speeds.
Do to steel loss's at high rpm.

And they were talking really high speeds,.. like 3000 to 5000 rpm. We are only talking several hundred at the most.
So this sheds some pretty good light on the tooth overhang thing. When we see motors with no brims, we can assume they
are high rpm motors.

I couldn't find any reference to what wire size they were using, nor any reference to what voltage was used.

 

[APL]

Just some possible tweeks. If we use a single piece of aluminum for the ring, then I wouldn't mind seeing it a bit thicker
towards the axle. I'm not sure a single piece of 1/4" would be stiff enough. Correct me if I'm wrong.

Anyway, there needs to be some fairly large holes in it as well, to accommodate the coil and phase wires going through it.
Or maybe many small holes.
Also, I thought maybe a taper on the axle, to loose some weight, and maybe bump it up to 2.5" stock?



Something like this anyway... didn't spend much time on it.

 

[damirsky]

I still think where the coils are wound the area should be minimal (based on flux and what material can hande, dont forget 141%), and round !

could you please explain a little why it is so important for coils to be round? in most radial flux motors i've seen the coils have almost rectangular shape and the slots (and the wire inside it) are parallel.

 

[damirsky]

Something like this anyway... didn't spend much time on it.

i think the cores support ring and the shaft shouldn't be that complicate shaped. the ring could be made of 1/4" aluminium sheet by water jet cut or cnc milling machine, the axle could be a bit less beefy too.
i was also confused a little because at the beginning i was thinking you're building middrive motor. now i see you are designing the hub motor.

 

[APL]

It's true the ring should be simple to make, that part bothers me a bit. 1/4" plate might be strong enough by itself too.
Or maybe just turn down some 3/8" where the cores are.

I guess I'm just trying to make sure that there is no side to side movement, and I'm a little concerned with harmonics,
if it can flex at all. Those magnets have a lot of power at full throttle. This motor might be able to do 2000 watts.

Also I'm trying to make as much room on the left side for phase and coil wires as I can. When I got done with mine, there
was a lot of wire in there. And I'm not sure, but I think a 'cone' shape is more rigid. (I guess it's not making a lot of room)
But your point is valid. A lot of machining that may not be necessary.

I'm sure the axle can loose some beef, it's still unclear as to what the dropout arrangement will be, and the method of
contact on the ends, or even the diameter that will mate with them. I guess I figure it's always best to overbuild at first.
At least I know that 26mm will work in my set up. But some kind of universal 'end' mount will need a lot of discussion yet.

Yea, I thought it would be a good idea to start out with a hub motor design, as a lot more people are interested in that,
and it will still work in my cruiser bike. Neptronics has expressed great interest in putting one on his machine, and I'm
trying to head in that direction, so he can give us a good road test, and tell us all what we've done wrong. :lol:


I had to squash the wires in there! (20mm deep)

 

[Lebowski]

I still think where the coils are wound the area should be minimal (based on flux and what material can hande, dont forget 141%), and round !

could you please explain a little why it is so important for coils to be round? in most radial flux motors i've seen the coils have almost rectangular shape and the slots (and the wire inside it) are parallel.

You want minimal wire length, for minimal resistance. For the lowest circumvence of your area it has to be circular.

I would make the coils round, but the area facing the magets can be any shape... the iron inside the coils is a long round stem and at both ends it flares into a rectangular shape facing the magnets

 

[APL]

Lebowski, can you also explain a little more about 141 percent T?

 

[Lebowski]

Lebowski, can you also explain a little more about 141 percent T?

Basically you have the flux from the permanent magenets (100% T) and the flux from the coils. Lets say the flux from the coils at max phase current is also 100% T. These two fluxes are at a 90 degree angle of each other (for max torque), so they add up via Pythagoras, to 141% T. So this 141% T is what the iron inside the coils sees as a max flux, and this is the flux to keep in mind for iron area (saturation)

 

[APL]

Ahh, I see. I didn't see anything about that in my searches,..thanks. I'll have to look for it now that I know. Thats
going straight into the formula and process for finding core saturation. Which I have yet to find,.. but working on it.
We still have to figure that out.

I haven't seen anybody using round coils in trapezoid forms yet, they usually want to have straight sides, but that
last link that I had, used simi-round tooth forms, which I thought was odd, since it's not seen very often, but makes
some sense after what you've said.
Tooth shape and profile needs to be looked at a little more closely yet.

 

[APL]

Also did yet another revision of the axle,.. a little lighter, and a smidge more room for wires. Some 1/2" flat stock
is turned down a little for the ring. And I thought maybe it should be on the other side. After thinking about it, I
guess it doesn't really matter where the stator is held from.

Anyway, just a thought.

 

[APL]

It's amazing how close this core size is to what coleasterling has made,.. basically 40mm long and 15mm high, with 10mm
between 2.5mm faces. Close to the same amount of turns, for the same size motor and a dual coil design.



So I'd say we're on the right track with core metal and saturation, and at least not far off, since I'm sure they have put a
lot of research into it, and have very specific numbers to show.
At some point, it would be nice to have a home brew method for getting core mass correct though, and hit that 141T goal.
Can't really do it until you have something to start with.

Since this motor is going to turn much slower, I'd say keep the full closed slot that we already have, and at most, maybe
round off the inner core a tad, and it should be good to go.
A few tests will show more.

 

[APL]

Hmmm, after reading the article some more, I realized that they are using only .450T ceramic Pm magnets.


So I looked them upend found a little info on them.


Plus I found some info on the SMC they are using, to compare with ours.
Kobe ML35D: https://www.kobelco.co.jp/products/powder/magmel.html

 

[APL]

A new video on the tube. This is a supposedly new type of axial, but really just a new kind of coil.
A lot of flash, and a lot of cash, but they ain't giving anything up on the 'nano' coil tech.
So I guess we have to figure out whats going on with them.
I haven't got a clue.

Axial; https://www.youtube.com/watch?v=hoqlorpHUT8

 

[fechter]

I couldn't completely comprehend how their design works. Seems like it's a skewed winding that uses turns on the sides to add to the main winding face, sort of like a Halbach array. The picture of the "prototype" was easier to understand.

I was thinking about the core shape more too. A circular cross section will give the lowest resistance losses per amount of flux, but a more trapezoidal shape will give more flux per unit area (higher torque density). Trade off between efficiency and compactness. Maybe a trapezoid with very rounded corners might be a good compromise.

 

[APL]

I can't quite figure it either. The prototype doesn't match any of the graphics, and none of the graphics are based on
anything real. The fact that the rotors can spin in two directions is only good for maybe 10 percent of applications,
and since the rotors have to be individually supported,.. that way leads to very large rotor bearing arrays.

If you get rid of all the flash, and boil it down, all I see is a fat ho-hum proto. But of course they're not going to show what
they are actually working on, this is just a taster.

Using some logic, the nano-coil doesn't appear to have any core material, so I assume that it's an air core. And since it
seems to use a lot of fine wires, they must be mostly in parallel, otherwise you can't run much current. So it looks like
they're weaving hundreds of wires between two bus bars to get rid of end turns, and have lots of radial wire.
The coil units are also split diagonaly into two sections, N and S.



I think your right about the core shape, pretty much what coleasterling has done already, and 'maybe' little more rounding
of the center, will be a good place to start anyway. I suppose a simulation like below will be needed to tell the rest.
I'm still doing searches on the 1.41T thing, it's difficult to find the right search title. But what I see so far is that indeed, a
typical core does approach 1.41T, when the tooth face is closer to 1T.



Lots to look at yet.

 

[InterestedinEVs]

Nothing to add to the last few posts other than that a more round geometry is easy to do and probably better for stress concentrations on the cores. External sharp corners are generally bad just like internal sharp corners.

The threading test worked well, but I found another issue in that on tightening, the core will crack right through the thread, axially. After that, I don't feel that we can effectively thread and screw into the cores. Bolting through from the outside face might be the only solution. It may make sense to use shoulder screws during assembly to limit bolt clamping on the core, too.

 

[thepronghorn]

....
I'm still doing searches on the 1.41T thing, it's difficult to find the right search title. But what I see so far is that indeed, a
typical core does approach 1.41T, when the tooth face is closer to 1T.



Lots to look at yet.

Lebowski said 141%T not 1.41T. He meant that you should design your core and magnets such that the flux density (one unit of which is T) without any stator/windings flux is 1.41x less than the flex density when the core is saturated. This way, if the flux from your magnets and stator/windings is equal, since they are 90 degrees out of phase, they will sum to the max desired flux density.

I'm not sure how big an IF it is that the flux from magnets and flux from windings would be equal. I guess it makes sense that you would want the fluxes to be pretty matched at the normal operating point of the motor so that one or the other doesn't dominate and push core or copper losses away from being matched as well.

 

[APL]

Thats to bad coleasterling, one step back. Perhaps a shoulder bolt/screw then. Maybe with a fiber washer or something,
to get just the right tension. The stuff is like cement, like you said. Thanks for testing it out. :thumb:

I've been thinking again,.. and making two back irons on the cores just seems like unnecessary weight and material.
Plus, we have a 1/4" piece of aluminum between them.
All this makes the motor 68mm wide when done.
The spokes won't clear the rotors, when they come off of the bearing cap flanges.

If it goes back to the iron center ring, then you don't need the bottom irons, and the cores can bolt directly to it,
making the total width much less, and saving weight, and no worries about cracking the lower lips.
They could be bonded along with bolting, but I'd rather not.
Any excess steel anywhere can be drilled out.

If the same thing is done with the rotors, rivet the back irons to the insides, then that can be narrowed too. The
whole thing would be 48mm wide, enough to clear the spokes,(?), and also give more room for a freewheel, which will
be needed as well.

I'll have to draw it into a wheel,.. to see if it will fit,.. something like this.



I haven't really thought about a freewheel yet, and I suppose it will need a disc brake mount as well. Dang. More
problems coming. This is why I like mid drives.

 

[fechter]

The back iron for the cores needs to be laminated or powdered iron. Otherwise it will act like a shorted winding. If the backing for the cores was fiberglass or other non-conductive stuff, it would be better but you really don't want a big gap between the two core pieces. If the gap is fairly small compared to the width of the core, it might be OK. It would be good to see how it looks on a MFEA model. If you use something like 1/4" fiberglass, it should be strong enough. You'd want to wind the cores on both sides in series so it behaves like a single core. Since the middle of the core is smaller than the 'brim', you might be able to make some shallow recesses in the fiberglass to get the ends of the cores closer together without sacrificing too much strength. It would be much stronger if the cores were glued to the fiberglass, but then you couldn't take it apart.

 

[APL]

Of course,.. we've been through that, I've forgotten more than I can ever remember. :lol: Well, we certainly don't want to
use lamination's, although I have seen it used in disc layers for this purpose, so I guess it can't be ruled completely out.



One thought was to make a 1/8" deep circular protrusion on the bottom of the cores, that could go half way through the
1/4" fiberglass disc, and connect with the other core. Allowing the two cores to pinch the disc with the rest of the
trapezoid shape. Would have to drill eighteen 3/4" (aprox.) holes in the disc though, which makes it weaker. So I'm not
real hot on that idea yet, although I do like the idea of having separate coils again. It is the lightest.

Another thought is to revise cleasterling's finger design, and have separate 1/8" thick iron brackets connecting the cores
to the disc, which also separates the coils again. the fingers could be permanently riveted and bonded to the disc.

Not exactly like this, but similar; (the two different designs)

 

[InterestedinEVs]

I have an idea for this, but it would require some testing.

What about a single traditional lamination on either side of the powdered core? Say 0.2-0.3mm, laser-cut. Machine the powdered iron into whatever shape we need to be able to use the surplus material(probably the same trapezoids we're using), then form a composite using the powdered iron as the core. The laminations take most of the stress, so it doesn't really matter that the core is brittle. We could either epoxy or use a bonding sheet.

Would we need to make sure that the core powdered iron is electrically conductive with the face lams?

Edit: got two posts in while I was typing on my phone. I'm talking about a back iron ring to bolt the cores to as in your previous sketch fetcher replied to last.

In general, I like the previous idea a lot. No undercut and easy to make a large root radius on the brim by using an end mill with a corner radius

 

[APL]

I'm not sure what you mean in the first part of your post, I'll have to re-read it.

But I drew a full size sketch of a 8" glass board plate, and put in 18 short 5/8" slots. (ovals were easier to draw) The slots
go all the way through, and will keep the cores from twisting, and are easy to mill.

It looks better than I had thought, seems to be a good amount of distance between them. Maybe this idea does have some
merit.



Your right coleasterling, the cores could be milled from one side, and radiuses could be aded to the steps.

 

[InterestedinEVs]

Sorry, just a sandwich panel using the smc as the core and a set of traditional lamps as the face sheets.

https://en.m.wikipedia.org/wiki/Sandwich_panel