Ironless axial flux Halbach project - 170mm, 36 pole, 44mm bcd disc brake mounting

[safe]

I think you are imagining my design in a different way.

The cores go directly from one array across to the other. There is no looping or anything like that. The cores simply increase the field strength in the middle of the gap where it's the weakest. You could get rid of the cores in my design and just use air and the only difference would be lowered field strength. The standard Halbach design is nothing but copper, something inert or air in the gap, so all I'm doing is adding a minor enhancement to allow for wider gaps. The silicon steel in effect "conducts" the magnetic field across a wider gap giving more room for copper. But you can see that this design forces me to abandon the idea of wave winding because I can't have things overlapping... it's a digital sort of wiring pattern where there is only one phase.

...in the middle of the cores the field strength is about 1 Tesla and it's on the edges where the higher 2 Tesla peaks occur. The magnets themselves are only rated as 0.5 Tesla, so being able to get this much field strength is not easy to do. The simulation shows only a few extra watts heating at 200Hz. Keep in mind I'm shooting for peak power of 2000 watts in certain configurations... you can't achieve that without some sacrifices in efficiency. (or bigger magnets)

Mechanically the edges of the cores will also serve as the attachment point to an aluminum plate(s). It would all be glued together with fiberglass resin to make a solid mass connected to the aluminum plate(s). The aluminum plate(s) becomes the "frame" for the stator and helps to make things more rigid. As of yet I haven't decided on using one or two aluminum plates, one on each side might be overkill. However, as the power increases you have to be more worried about mechanical breakage as opposed to heat problems... too much power and the stator could fly apart.

 

[madact]

I think you are imagining my design in a different way.

No, I'm just slicing it on a different axis (if your picture is in the x-y plane, mine is in the y-z plane...)

 

[safe]

We're not on the same page yet I think...

My take on what the image above is showing is that the magnets are in the center on both sides and the iron surrounds the magnets but the whole thing has a sort if "H" shape that is rotated onto it's side. This is not at all what I'm doing. (actually I don't really know what's going with that)

As the saying goes:

"The shortest distance between two points is a straight line."

...so if you imagine a disc which on it's ends forks open to have two lips to it. On each lip goes a Halbach Array. Now the gap between the two lips of this disc has a distance of 0.75" which means when you add the width of the lip (magnet and fiberglass) you would get:

0.125"' (fiberglass) +
0.25" (magnet) +
0.75" (total gap which includes two air spaces, the core and the copper) +
0.25" (magnet) +
0.125 (fiberglass)

= 1.5" Total distance from outside of one lip to outside of the other lip, so this is the total width of the disc overall. Think of the image as if you were viewing the disc "from above" looking downwards onto the edge much like you might look at your tire on your bike looking at it on it's edge while riding.

Inside the gap there are cores spaced at exactly one pole distance so that when the stator is in alignment with the poles everything lines up and when the disc rotates then everything is changing in a perfectly simple sinusoidal way.

What you get is a pure Single Phase motor.

However, since there are 72 poles (36 pole pairs) the degree of granularity in the wheel movement should be such that the controller will get up to higher frequency earlier and that assists in reducing electrical surges. (it becomes more like a servo motor)

Hopefully that makes sense... the cores never do any serious bending of the magnetic field...

This is an earlier attempt that ended up being too small, but you can clearly see the two lips and the groove in the middle which is the gap where the stator would go.

 

[madact]

We're not on the same page yet I think...

My take on what the image above is showing is that the magnets are in the center on both sides and the iron surrounds the magnets but the whole thing has a sort if "H" shape that is rotated onto it's side. This is not at all what I'm doing. (actually I don't really know what's going with that)

Sorry, sometimes I tend to use 'engineering shorthand' with people who aren't familiar with the conventions of a field, my bad... I'll explain what's going on in my picture a little better: the 'loopy bit' is just there to complete the magnetic circuit, and give a better idea of how much flux the core can generate over an airgap without any other fields in the system. It has two sides simply because I wanted to compare two approaches on the one page (tapered vs. non-tapered)

Just look at this bit (from the RHS), which is the 'business end' if you will:

the blocks above and below represent the positioning of your magnets, the not-quite-rectangular laminated steel section is a slice though the 'I' section from your diagram (like I said, rotated 90 degrees - the 'wide top of the I' would be along the top edge of the central steel section in this diagram, and the 'wide bottom' along the bottom edge), and the coils shown are where the wires wrap around the ends of your 'I' sections.

You might say that my diagram here is the view 'looking along the groove' rather than 'looking at the bottom of the groove'.

 

[safe]

Okay, now I think we are on the same page.

Have I tested tapered cores as a concept?

Yes... at first I was excited because it increases the peak magnetic force passing through the core when things are at zero degrees rotation. But then later I wondered what happens to the cogging as the core moves from pole to pole. Seems that the tapered design doesn't improve the average power at all.

The reason I found this to be true (based on observing the field lines and magnetic strength) is that as the core passes from pole to pole the lines begin to bend. The more the lines have to bend the more you lose strength. So what happens is the center of the core is functioning with barely any magnetic load at all and the edges fail to be able to keep up.

To me this explains why most every motor uses an "I" beam sort of shape for the core ends.

In my case I simply recycled another motors core and cut it up into little pieces that were "I" beam shaped and then fiberglassed them together. Cogging goes UP with an "I" beam shaped core, but average power also goes up... it's sort of a trap that as you increase cogging you also increase strength. You cannot use static analysis alone... you need to do the complete analysis including comparing "Apparent Force" verses "True Force". (this involved subtracting out the "Cogging Force") In my spreadsheet there is a full section where I grapple with the difference. To be honest I took the static case as "fact" for about a month before I realized that I'd better test it. (and was shocked at what I learned)

In the Avanti motor case they might be dealing with such large losses on the outer loop that the inside edge doesn't matter... or... they were insistant that cogging was kept to an absolute minimum at the expense of power. My guess is that they wanted to satisfy the 200 watt limit without creating a motor that cogged too much and they could afford to make the tradeoff. The Avanti might have been designed just focusing on low cogging.

Do you know if it's known for low cogging?

 

[madact]

Okay, now I think we are on the same page.

Have I tested tapered cores as a concept?

Yes... at first I was excited because it increases the peak magnetic force passing through the core when things are at zero degrees rotation. But then later I wondered what happens to the cogging as the core moves from pole to pole. Seems that the tapered design doesn't improve the average power at all.

The reason I found this to be true (based on observing the field lines and magnetic strength) is that as the core passes from pole to pole the lines begin to bend. The more the lines have to bend the more you lose strength. So what happens is the center of the core is functioning with barely any magnetic load at all and the edges fail to be able to keep up.

To me this explains why most every motor uses an "I" beam sort of shape for the core ends.

Nope, still not on the same page I'm saying keep the 'I' profile, but stack em like this, to get a taper in the other axis:

Thus avoiding nasty iron saturation in this region. I only drew the taper on the closer end, you can use your imagination for the other end...

 

[safe]

Wow... amazing that we still aren't on the same page yet.

Getting back to this image:

...this would be a "top" view of the disc.

So if you were sitting on the bike and looked "down" onto the disc on the back wheel you would see literally what the image shows. The larger boxes between the cores represent the copper coils which in real life would be 64 turns of 18AWG magnet wire. Since this is Single Phase each bundle of copper is alternating from going "in" to going "out". The alternating "in" and "out" occurs at exactly one pole intervals so there is no "skew" from having different numbers of poles and coils.

As for the idea of tapering in the most recent image... the place your image shows taper is where the core runs off to the edge of the magnet in the "in" and "out" direction. I just don't see any benefit to doing that. The taper is not facing the direction of the magnets but is going off into free space.

----------------------------

Let's start over.

My hunch is that you have the orientation I'm using rotated along a different plane. Using X-Y-Z langauge let's try it all over again.

The disc attaches to the rear wheel.

The rear wheel has radius along the X-Y plane.

The rear wheel has depth through it's axle along the Z axis.

The magnet arrays are lined up along the X-Y plane.

The "I" beam cores face along the Z axis with the long part of the "I" beam being like the rear wheel axle and the curved parts on the ends diverging from the Z axis and going into the X-Y plane.

The copper coils make circles in the X-Y plane.

Note: The image does not follow this X-Y-Z orientation in the same way. In the image the wheel radius would be only along the X axis, the "I" beams are on the Y axis and the copper coils would go "in" and "out" in the Z axis. I have taken orthographic design classes and have two science related degrees, so I know how the language is supposed to work. Ideally the FEMM program would permit orthographic projections and 3D rotation... and dynamic simulation... but we have what we have. (it needs to be more like a CAD program)

--------------------------

Follow up... tapering the cores in the direction you suggest might not hurt.

Since it's nearly impossible to get a sharp 90 degree bend with copper magnet wire it does make some sense to fill that area with a little extra silicon steel. In my case it would be too much work to do it because the shape of the steel is more or less predetermined by another motor, but I can see some value in it. I've pretty much just thrown a little extra fiberglass resin on the edges to round things out.

I think I DO get your point... and it is of value.

 

[madact]

As for the idea of tapering in the most recent image... the place your image shows taper is where the core runs off to the edge of the magnet in the "in" and "out" direction. I just don't see any benefit to doing that. The taper is not facing the direction of the magnets but is going off into free space.

That is, in fact, exactly what I'm suggesting. Why? Because it will heat the iron up less, when you're running at high power. Why would that be a good thing? (a) efficiency and in extreme cases (b) avoiding damage from localised heating.

Sure, it's only a couple of percent advantage in field strength, but it gives you more headroom for pumping lots of current through the thing before you start experiencing substantial losses...

Let's start over.

My hunch is that you have the orientation I'm using rotated along a different plane. Using X-Y-Z langauge let's try it all over again.

The disc attaches to the rear wheel.

The rear wheel has radius along the X-Y plane.

The rear wheel has depth through it's axle along the Z axis.

The magnet arrays are lined up along the X-Y plane.

The "I" beam cores face along the Z axis with the long part of the "I" beam being like the rear wheel axle and the curved parts on the ends diverging from the Z axis and going into the X-Y plane.

The copper coils make circles in the X-Y plane.

Note: The image does not follow this X-Y-Z orientation in the same way. In the image the wheel radius would be only along the X axis, the "I" beams are on the Y axis and the copper coils would go "in" and "out" in the Z axis. I have taken orthographic design classes and have two science related degrees, so I know how the language is supposed to work. Ideally the FEMM program would permit orthographic projections and 3D rotation... and dynamic simulation... but we have what we have. (it needs to be more like a CAD program)

Don't sweat it, I did read your thread

Incidentally, if you're doing single phase, have you thought of doing wave-windings with flat conductors? Big fat wide 'wires' like this guy http://scolton.blogspot.com/2010/01/epic-axial-motor-iaprogress.html is using - with single-phase you could just loop the conductors in and out through the core (if you don't mind an odd number of poles before you bring it back...)

 

[safe]

I might need a microscope.

...these cores are very small and I did make them just slightly larger than the 0.25" width. (it's shown laying on it's side) There's a point where you accept that you are making a homemade project and give up on some of the more exotic things.

As for the flat copper bars... yes I did think about those. They have the same effect as using thicker wire and fewer turns and that means you get a lower inductance and higher kv value. Since I don't really want to deal with adding inductance back into the motor to make it work I'd rather leave it as a "standard wind" where the inductance is built in already. I really don't want to push more than 20 amps through the motor at 48 volts and the simulations show that as being very moderate in the heating department.

Electric Go Kart race machines are not given the self imposed 1000 watt input limit I'm designing around, so they are free to build with unlimited torque in mind.

More current means more stress on the controller. (I don't want that)

Many of the racing electric motorcycles are using controllers that allow well over 100 amps of current. Generally speaking when you load up on the torque at low rpm you use more current and that creates more heat because it's less efficient. The design concept for this bike is to pedal from 0-20 mph to get up to speed and then the powerband really comes on strongest from 20-40 mph. So there are actual reasons for wanting more inductance and less current.

For Australia with the 200 watt input limit and the speed limits you are going to want to get as much current to flow as possible, so your design criteria is probably different.

Though... in America for the Federal Ebike Law the restriction is 20 mph, so I'm going to have to have some sort of "variable winding" needed to be able to switch between optimizing for top speed and optimizing for low end.

So you expose a legal and design issue with anything using fixed gearing.

 

[safe]

Any luck finding someone to build the stator?