Dual rotor axial flux motor design

[Miles]

I can see that your motor would be more versatile (more efficient over a wider range of torque outputs) but mine is being designed specifically for this: viewtopic.php?f=30&t=14484 In practice, I'll be using mine with a 2 speed gearbox - they are designed as a pair.

[Miles]

This looks to be a useful resource: http://www.emetor.com/

As well as the revamped: http://www.powerditto.de/bewicklungsrechner.html

[markobetti]

Dear Miles, the gearbox isnt published at ES ? Its okay if its a secret , but if not please post link since i havent been able to locate it. i remember you and hal discussion about retro direct drive ....but thats not it right ?

[markobetti]

Maybe you can try your calculations in trial version if they have it :

http://www.infolytica.com/en/products/

http://www.infolytica.com/en/applications/ex0072/

[Miles]

Dear Miles, the gearbox isnt published at ES ? Its okay if its a secret , but if not please post link since i havent been able to locate it. i remember you and hal discussion about retro direct drive ....but thats not it right ?

Hi Marko,

No I haven't published anything about it here. I will but not just yet. It is a retro-direct as Hal and I were discussing then but I've spent ages exploring all the possible options and going through numerous iterations of design.

[Miles]

Maybe you can try your calculations in trial version if they have it :

http://www.infolytica.com/en/products/

http://www.infolytica.com/en/applications/ex0072/

Thanks! They didn't do a trial version before - now they do.

[kenkad]

miles
This is what I meant by vertical laminations, bent into a 'C', stacked, to produce a trapezoidal end. It seems to me that the flux path out the end is now distributed. I used different colors for the lam layers to better see the result. I have never seen a sim of this, but have always wondered why this is not done. My preference would have been to do this as a powdered core so that the center could also be trapezoidal, providing a better motor fill with a split copper strip winding like Shane did in one of his motor experiments. I hope this explains my question to everyone. As to why the more than one 3-phase group question. That is because starting torque requirements to get a vehicle moving are greater than running torque requirements at a speed, so I would op to turn off a 3-phase group to reduce motor power consumption since I do not believe the cogging would be noticeable at speed, momemtum effects.
kenkad

[markobetti]

Glad to be a service to you. You helped me alot, for example with freewheel . Finally made indestructible freewheel with two bearings inside al 7005 mat. and more pawls than any freewheel on the market i can think of. http://img217.imageshack.us/img217/326/heavymetalx.jpg . Logo is covered with white color due to my project that is in state not intended to be shown right now under the name. Anyways , thank you MILES

[Miles]

miles
This is what I meant by vertical laminations, bent into a 'C', stacked, to produce a trapezoidal end. It seems to me that the flux path out the end is now distributed. I used different colors for the lam layers to better see the result. I have never seen a sim of this, but have always wondered why this is not done.

Ken,

I don't think this would work. You would get serious eddy currents on the end face. The laminations need to restrict flow perpendicular to the direction of the rotor. Also, there will be a strong reluctance between the face laminations, I think.

[Miles]

Glad to be a service to you. You helped me alot, for example with freewheel . Finally made indestructible freewheel with two bearings inside al 7005 mat. and more pawls than any freewheel on the market i can think of. http://img217.imageshack.us/img217/326/heavymetalx.jpg . Logo is covered with white color due to my project that is in state not intended to be shown right now under the name. Anyways , thank you MILES

Wow! That's nice Marko. Look forward to seeing what it's intended for....

[rhitee05]

This is what I meant by vertical laminations, bent into a 'C', stacked, to produce a trapezoidal end. It seems to me that the flux path out the end is now distributed. I used different colors for the lam layers to better see the result. I have never seen a sim of this, but have always wondered why this is not done. My preference would have been to do this as a powdered core so that the center could also be trapezoidal, providing a better motor fill with a split copper strip winding like Shane did in one of his motor experiments. I hope this explains my question to everyone. As to why the more than one 3-phase group question. That is because starting torque requirements to get a vehicle moving are greater than running torque requirements at a speed, so I would op to turn off a 3-phase group to reduce motor power consumption since I do not believe the cogging would be noticeable at speed, momemtum effects.

There's a few separate issues we can break down. First, lamination direction: vertical/radial vs. horizontal/tangential. Eddy currents are due to the induced EMF, whose direction is given by v x B (vector cross product). For an axial-flux motor, this gives an induced EMF in the radial direction (v is tangential, B is axial, thus EMF is radial). Thus, eddy currents in the radial direction will be created in any conductive material present. By laminating the cores in a horizontal fashion, we keep the conductive distance in the radial direction short and thus limit the eddy currents which can flow.

Second, the shape of the teeth at the end of the cores. I agree with you that it would probably be beneficial to make the ends wider than the center - the proper tradeoff between iron and copper can be determined through simulation later. Looking in towards the center, this would give the cores a sort of dog-bone shape. You're also talking about making face of the tooth wedge-shaped, rather than rectangular (looking axially). I brought that up a little while ago as well. That would be relatively hard to do with this lamination direction, and I'm not certain it would really give significant benefit. On the other hand, leaving the ends rectangular will I think give the BEMF more of a sinusoidal shape, which might actually be a plus.

Third, Miles is correct that electrical steel is a better material than a powdered core. Harder to work with and more expensive, but better magnetic properties.

I looked at it from a different angle... In my opinion you need to convert
electrical power efficiently into mechanical power, independent of whether
you want to race uphill or just ride around on the flat. Core laminations help
with efficiency but only in a certain rpm range (too low rpm -> cogging,
too high rpm -> eddy current losses). No iron core gives a lower specific
torque but this can be solved by chosing a different gear ratio so I wouldn't
say this is a disadvantage of having no iron core.
For me in the end the efficiency with no core was good enough, the disadvantages
of the iron core outweighed its advantages.

Again, a couple of separate issues. Miles is correct that an iron core gives more torque. Specifically, because you can achieve higher flux densities, the motor has a higher power/torque density (obviously you could make a large ironless motor with more torque than a small iron-core motor).

Second, efficiency. I disagree that cogging represents an efficiency loss. Cogging is both positive and negative torque, so the net energy contribution is basically zero. It increases the torque ripple, but since Miles is gearing his motor down I think torque ripple will be much less noticeable than with direct-drive (higher frequency, lower amplitude).

There's probably a study out there to prove this one way or the other, but I would tend to think that ironless is not inherently more efficient than an iron core. You do avoid most of the eddy losses, but in exchange you need more current to achieve the same torque with lower flux density, so the copper losses will be higher. I suspect that eddy vs copper losses would be a pretty even tradeoff in a motor with a high-quality core (thin lams of high-grade electrical steel). For our practical purposes, however, I think ironless is attractive because it's easier to build a good DIY motor this way even if there is a torque density tradeoff. You could also probably say that ironless has more advantages at high RPMs and iron core has more at low RPMs.

[kenkad]

rhitee05, miles
I am not a motor designer. If a wedge shape core end is not beneficial, why should wedge shaped magnets be used on the rotor instead of rectangular shaped magnets. It has been quite a while since I spoke with people involved in powdered cores, but, it seems I recall that there is an orientation process in making cores like this. I dropped the whole design issue because in talking with companies and universities about sim software, I was told the software was not sophisticated enough to deal wtih these questions. That was more than two years ago and I have not checked if there are any improvements. This is why I am now asking the question.
kenkad

[rhitee05]

I am not a motor designer. If a wedge shape core end is not beneficial, why should wedge shaped magnets be used on the rotor instead of rectangular shaped magnets. It has been quite a while since I spoke with people involved in powdered cores, but, it seems I recall that there is an orientation process in making cores like this. I dropped the whole design issue because in talking with companies and universities about sim software, I was told the software was not sophisticated enough to deal wtih these questions. That was more than two years ago and I have not checked if there are any improvements. This is why I am now asking the question.

I can't really answer your questions, which are getting much further down into the details of motor design than I'm familiar with. A lot of these parameters are design trade-offs, though, rather than distinct A is better than B scenarios. For example, you might get a little more power out of wedge-shaped magnets (I think), but the rectangular magnets would I think give more sinusoidal BEMF and probably lower cogging. That's more of a design choice.

If it were me in Miles' shoes, I would be most interested in the easier-to-build tradeoffs even if the resulting motor was a little sub-optimal.

[Miles]

As far as my decision process goes:

A completely wedge shaped core is out because 30 different sized laminations would be a PIA.

This leaves the possibility of trimming the teeth (top and bottom of the I shape) of an oversized core module after assembly.

So, the choice is between increasing the overall area of the core at the gap, even though it all ends up in a uniform section in the middle or, sacrificing the extra area for a BEMF which is closer to sinusoidal and is likely to behave better with a sine controller.

With an SMC core, a wedge shape makes more sense, because the flux is automatically distributed uniformly.

[Miles]

I guess I could try machining the complete wedge from each block of bonded laminations.... That way I could start with simple square lamination blanks and avoid the punching step altogether.... I'd have to be careful to avoid introducing any bridging between the laminations, though.

[Miles]

Next decisions:

How many slots? 18, as modelled or I could possibly reduce it to 15? without increasing the lead in/out angle across the laminations too greatly...

How many poles?

[Lebowski]

Next decisions:

How many slots? 18, as modelled or I could possibly reduce it to 15?

How many poles?

These choises have no effect on efficiency... just make sure the motor is physically big and has lots of copper in it.

[Miles]

These choises have no effect on efficiency... just make sure the motor is physically big and has lots of copper in it.

But they will affect the amount of copper in the links between coils... If I used 15 slots and 14 or 16 poles, most of the coils could be linked to their adjacent partner, I think. As I'm intending to use rectangular copper strip and the connections are outside of the case this would be an advantage...

Need to find out the trade-offs WRT the winding factor?

[bearing]

I don't think a stator with 15 teeth is going to work well. With 14 or 16 poles, windings will, as you say, be put in groups of 1/3 of the diameter. Every time a phase pair is energized it will "bend" the rotor(s) and shaft. See the summary at the bottom of this page: http://powerditto.de/15N14P.html
It might work better with an axial design with small diameter, but I would personally avoid it.

From a mathematical perspective, it looks like you end up with a solution like this (windings in groups of 1/3 of diameter) every time the number of stator teeth is 3 * prime number, and at the same time use a pole count as close as possible to number of teeth. (3*5 and 14/16 in this case.) It will, however, give a very good winding factor. To solve this "imbalance", use twice as many teeth, which will put windings in two groups, on opposite sides of the centre. 18t (2*3*3) 18p/20p are twice as many teeth as the "imbalanced" 9t (3*3)p 8p/10p. Following this logic, 30t (2*3*5) with 28p/32p would be a good solution, or 42t (2*3*7) with 40p/44p. This can be confirmed with http://powerditto.de/bewicklungsrechner.html
The famous 12t (2*3*2) with 10p/14p also follows this logic/design of 2*3*prime number and then the closes possible number of poles.

[Lebowski]

These choises have no effect on efficiency... just make sure the motor is physically big and has lots of copper in it.

But they will affect the amount of copper in the links between coils... If I used 15 slots and 14 or 16 poles, most of the coils could be linked to their adjacent partner, I think. As I'm intending to use rectangular copper strip and the connections are outside of the case this would be an advantage...

Need to find out the trade-offs WRT the winding factor?

What is a winding factor ?

I thought about using copper strips as well, an idea which I had with this respect that you may find usefull:
at farnell they sell (quite cheaply) heatshrink tube in all diameters and up to 5 meters in length (5 meter for 4 euro
for the thinner stuff). You can sheeth your copper strip with heatshrink tube before winding it to isolate it....

Another thing I thought about at the time was a tripple stator design with one phase per stator. You can
wind the strip in a cloverleaf design... disadvantage is the big piece of iron (a.k.a. axle) going through
the middle of your coils

Also usefull, think Lorentz force, not magnetic attraction / repulsion.

[Miles]

I don't think a stator with 15 teeth is going to work well. With 14 or 16 poles, windings will, as you say, be put in groups of 1/3 of the diameter. Every time a phase pair is energized it will "bend" the rotor(s) and shaft. See the summary at the bottom of this page: http://powerditto.de/15N14P.html
It might work better with an axial design with small diameter, but I would personally avoid it.

Thanks bearing. That's a very good point. I guess the "unbalanced" rotor/pole combination would be noisy, too...

[bearing]

Winding factor has got to do with the amplitude of EMF compared to the maximum possible EMF based on peak magnetic flux. If the number of teeth is the same as number of poles, the winding factor is 1, since all flux will generate EMF in windings. But you can't use a teeth/poles ratio of 1:1 with an iron core (probably with ironless core though) because of the super high cogging. If the number of teeth is different from the number of poles, the winding factor will go down. With 12t/2p, the winding factor is 0.25. With 12t/4p, WF is 0.5. With 12t/8p or 12t/16p, winding factor is 0.866 (=SQRT(3)/2). With 12t/14p or 12t/10p WF=0.93.

[Miles]

What is a winding factor ?

I'm hoping someone else will explain that....

[Ed. Thanks bearing!]

Ref:

[Miles]

To solve this "imbalance", use twice as many teeth, which will put windings in two groups, on opposite sides of the centre.

18t and 20p (winding factor 0.945) seems the best solution for me, I think. I'll see what a 12t layout looks like, though.

[Lebowski]

My axial flux motor made me think...

The magnetic force from the iron-less windings is very small. The torque comes from Lorentz-force which
is B*I*L. I can't rememer B at the moment but the torque matched the force I calculated with this Lorentz
force equation. For Lorentz force it's necessary that there is a current carrying conductor in a magnetic field.

But if you have a motor with an iron core all the magnetic field channels through the iron, there will be no
Lorentz force. It works base on magnetic attraction and repulsion.

I found it very interesting that apparently there are more than one way of generating electro magnetic force
and that different motors use different principles. My main conclusion however was that adding iron to my motor
would allow the flux to go through a smaller diameter area which would lead to a shorter bit of wire necessary
for 50 the turns I use -> less resistance -> more efficiency.