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

madact

Member
Local time
8:44 PM
Joined
Jun 12, 2010
Messages
43
Here's the concept - simple build, motor cost in the ~$500AUD range (plus elbow grease and/or machinists fees). Built on the pattern of the CSIRO motor (http://www.csiro.au/resources/pf11g.htm), but with a somewhat lower operating frequency (about a quarter, in fact), as I can't be bothered mucking about with litz wire...

Rotor:
- Twin 36-pole Halbach array (144 magnets to mount :cry:)
- NIB N42 1/2" x 1/4" x 1/4" magnets
- Standard 6-bolt brake discs as both stator mounting and magnetic backing (Halbach array isn't perfect, there's still a few stray field lines to catch)
- Mounts directly to a disc brake hub
- O.D. of magnets is ~170mm, meaning the magnets may need a touch of grinding into wedginess, to fit nicely... or I just let them hang out a couple of mm...
- Pole to pole gap ~5mm
- Aluminium spacer to keep the poles apart, vented at the periphery for cooling

Stator:
- Double-wave wound (to keep the flux nicely confined) around an inert core about 2mm thick - putting the conductors in the airflow and near the magnets, and leaving the middle of the gap (where flux is weakest) for the core...
- Mounted to hub via a nice big bearing which fits outside the disc brake bolts - I'm thinking a 6812 or thereabouts. Large diameter bearing should make it easier to eliminate play and shorten the air gap :devilish:
- Held stationary to the frame by a bolt-on arm, which also allows for some frame flex.

Magnets (and rotors) are already on the way :whistle: so it's a bit late to change the pole spacing, but I'm interested to know anyone else's experience in building this manner of motor...
 
Last edited:
Some diagrams etc.

A cross section of the rotor, stator and spacer (very much not to scale)
attachment.php


The layout of the Halbach array:
attachment.php


A FEMM flux density simulation (with N40) at the outer edge, where the magnet-magnet gap is greatest:
attachment.php
 

Attachments

  • cross section.png
    cross section.png
    6.3 KB · Views: 9,720
  • outer gap1.jpg
    outer gap1.jpg
    52.2 KB · Views: 23,916
  • rotor 1.png
    rotor 1.png
    10.7 KB · Views: 9,456
Last edited:
attachment.php


You should try the same simulation, but change the backing material from steel (brake disc) to aluminum or anything else that is inert. You will find that steel backing actually doesn't help much and in fact can even slightly reduce the magnetic flux.

Have you done the math on a 36 pole motor?

Will the stator cover 100% of the disc?

My design is to use about 44% coverage and a wider gap and add some core material back in to enhance the performance a little. (at the expense of cogging)

I like the fact that others are experimenting with the Halbach Design.

Also... another thing... I just got done cutting the steel disc out of my project and replaced it with a smaller more rigid base. For a disc motor you really don't want the flex that these discs allow. It's pretty easy to just cut and drill a flat plate and use it instead. I'm doing mine in fiberglass, but carbon fiber would be better.

attachment.php


How much power are you planning to run?

And I'm still not completely clear on how you are going to mount everything. The stator mounts to the frame right? The disc carries the magnets and turns right? Or no?
 
Last edited:
You should try the same simulation, but change the backing material from steel (brake disc) to aluminum or anything else that is inert. You will find that steel backing actually doesn't help much and in fact can even slightly reduce the magnetic flux.

Actually, I've run it both ways, and found the reverse to be true:
attachment.php

attachment.php


It seems what's good for the goose (dual-pole) is good for the gander (Halbach)... though there are a couple of caveats:
- With a Halbach array, too thick a backing will, yes, actually reduce the flux, particularly on the outer edge of the rotor as in this case, where there is a gap between the magnets. With large gaps and thick backing, you lose quite a bit, but the simulations above are for my outer edge worst-case scenario, and I'm still showing a decent gain. I'm basically aiming to use unmodified or minimally modified stock parts wherever possible, so I won't be ultra-optimizing the backing thickness - "better than nothing" is good enough for me here.
- If you put steel in the gaps between adjacent magnets, you kill the air gap flux right quick :sick:.
The academics like to not use any backing because this gives them theoretically better energy density etc. (if the magnets have a steel backing, they have to count it in their 'magnetic materials' mass :rolleyes:), but in the real world we have to make some engineering tradeoffs, and consider that the magnets don't float in space by magic, and have to be held there somehow - a bit of steel here isn't going to break the weight budget.

There are another couple of other reasons for having a steel backing, too:
- this design calls for a pair of N42 Halbach-enhanced 1/4" magnets to be held North to South within 5mm of each other. IMHO, holding them in place with only epoxy on the 'sides' and 'back' is a recipe for crunch failure in a real-world application - and putting fiberglass or aluminium on the 'front' - in the gap - is somewhat counterproductive. The pull force of the magnets onto a steel backing significantly reduces the tension and shear loads the epoxy has to bear, if not eliminating them completely.
- Catching those stray field lines makes me less likely to wipe my credit card when I check the pressure on the rear wheel.

Have you done the math on a 36 pole motor?

Not a great deal of it :unsure: - but I have worked out I'm going to have to drive the coils at 200Hz (real bandwidth in the region of 1kHz, using pseudo-square excitation) to hit 80 km/h, which is a good reason for ironlessness right there, and means I'll be building my own controller... I'll probably use 3-phase wave-shaped excitation (matched to flux profiles) for the low speed regime to get decent, smooth torque, and single phase pseudo-square for the high-speed regime, so as to only 'hit' the coils when they're in the maximum flux region, to maximise efficiency...

I'll admit to having done a fair bit of "looks about right" in coming up with the basic specs :whistle:, based on experience with model car and plane motors...

Will the stator cover 100% of the disc?

My design is to use about 44% coverage and a wider gap and add some core material back in to enhance the performance a little. (at the expense of cogging)

The coils will cover upwards of 90% of the surface area, but will be <50% of the volume - they will be wound around an electrically inert core, probably fiberglass PCB material. I was thinking about carbon fiber sheet, but the eddy current losses would turn it into an inductive heater...

I like the fact that others are experimenting with the Halbach Design.

Seems like the best-suited tech for the problem, and the axial construction means the number of parts requiring use of a lathe or other specialized equipment is minimal...

Also... another thing... I just got done cutting the steel disc out of my project and replaced it with a smaller more rigid base. For a disc motor you really don't want the flex that these discs allow. It's pretty easy to just cut and drill a flat plate and use it instead. I'm doing mine in fiberglass, but carbon fiber would be better.

Congrats... BTW, carbon fiber isn't as expensive as you might think - take a look on ebay, search for "carbon fibre cloth -vinyl", but be sure to include "-vinyl" or you'll get a thousand stick-on decals :D (this stuff might be suitable - http://cgi.ebay.com/ebaymotors/Carb...temZ390123460979QQptZLHQ5fDefaultDomainQ5f100). it should be fine on the magnet mount, and as long as you don't put it in any gaps with varying flux, it should be fine for eddy currents, too.

How much power are you planning to run?

Only in the 250W region, which is what's legal here in South Australia without a license & rego.... This is why I'm not too concerned about the disc brake rotor flex, and the chances of me exceeding half the rated torque of the disc brake rotor are non-existent - this is only a power-assist motor, after all, if I get 2Nm out of it I'll be quite happy. I will certainly be seeing what the motor as built can do before cooking the epoxy, though, and higher power outputs at higher speeds are certainly on the cards with a suitable controller...

And I'm still not completely clear on how you are going to mount everything. The stator mounts to the frame right? The disc carries the magnets and turns right? Or no?

More-or-less. In the engineering terminology, the stator is static and the rotor rotates, and you can put your coils and magnets wherever you like :D. However, the stator, while bolted to the frame, will be mainly mounted to the rotor through a big ol' bearing, the attachment to the frame will be designed to only transmit torque (and wires, of course) - it will be rigid in the direction of wheel rotation, but able to freely swivel left and right a bit.
 

Attachments

  • aluminium backing1.jpg
    aluminium backing1.jpg
    34.2 KB · Views: 7,351
  • steel backing1.jpg
    steel backing1.jpg
    35.4 KB · Views: 7,922
Last edited:
Links to eBay may include affiliate code. If you click on an eBay link and make a purchase, this forum may earn a small commission.
However, the stator, while bolted to the frame, will be mainly mounted to the rotor through a big ol' bearing, the attachment to the frame will be designed to only transmit torque (and wires, of course) - it will be rigid in the direction of wheel rotation, but able to freely swivel left and right a bit.

I'm having a hard time imagining how the bearing interacts here... I'm still not getting it.

Maybe by comparing what I have in mind you can tell me the different way you are doing it by contrast.

attachment.php


What I'm doing is making the rotor/disc be made of that steel base part and then fiberglass on top of it. Out on the ends of the fiberglass will be a "U" shaped groove that faces OUTWARD and in that groove will go the Halbach Arrays in parallel with each other. (same as yours) My stator will be attached to the frame and penetrate the groove from the outside facing inwards in the radial direction. You can see the stator mount I've already prepared on the frame. The stator would "insert into" about 44% of the disc's "U" groove. I don't require any added bearings this way.

From the drawing it looks like you are doing a reversed "U" facing inwards and then I sort of lose where you go next.

Explain more... I've read all the CSIRO stuff and they place the magnets on the wheel and then hold the stator stationary inside the wheel... my guess is that you are more closely imitating the CSIRO somehow? But where does the bearing go?

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

And I agree on the steel plate issue... it's normally not a big difference either way unless you push iron in between the magnets.

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

Another question...

How do you plan to get the level of precision required to achieve such close tolerances?

Part of my design effort was to admit to myself that I do not have the high tech tools to create a narrow gap and that's in part why I went with the extra cores because it relaxes the tolerences needed a great deal. Unless you have professional shop equipment it might be hard to achieve such high precision.

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

I ran a quick simulation on a spreadsheet I've developed using my best guess about what you are up to and arrived at something that looks like this:

attachment.php


...I'm sure that with more tweeking you could get more accurate predictions and better results, but it does seem to suggest that you are in the ballpark. To get those last few percentage points you really need to run a lot of simulations because getting a configuration that is just slightly off means that you are in effect running the wrong effective gearing. The odds of just "getting lucky" with this stuff is pretty low I think. No doubt you know that thickness of wire and number of turns defines the effective gearing.

I didn't realize at first you are using the 1/2" x 1/4" x 1/4" magnets. I got the 1/4" cubes so what you got is a better starting point. It takes twice as many of the smaller cubes to equal your "doubles". If I were to start over I'd go for the "doubles".
 

Attachments

  • Forum Axial Flux Idea.jpg
    Forum Axial Flux Idea.jpg
    143.7 KB · Views: 7,172
Last edited:
There are another couple of other reasons for having a steel backing, too:
- this design calls for a pair of N42 Halbach-enhanced 1/4" magnets to be held North to South within 5mm of each other. IMHO, holding them in place with only epoxy on the 'sides' and 'back' is a recipe for crunch failure in a real-world application - and putting fiberglass or aluminium on the 'front' - in the gap - is somewhat counterproductive. The pull force of the magnets onto a steel backing significantly reduces the tension and shear loads the epoxy has to bear, if not eliminating them completely.

Didn't manage to follow up on this previously...

Yes, I see your point on this. I found that I could setup a Halbach Array on my brake discs and they wanted to stay in place without any need for adhesive. It has it's advantages that's for sure.

What I'm planning to do is have 1/8" gaps between the magnets and you would think that it would harm the magnetic field strength (and in fact it does a little) but it also INCREASES the disc radius at the same time because of the added spacing. The advantage of the larger radius is when you get to the torque calculation it increases and so it pays for itself. Plus, as an added bonus, the extra space allows for more copper and that's the path to improving the efficiency. In a "perfect world" you would have an unlimited budget or selection as far as magnets, but I was more or less designing around 1/4" cubes because you can get 100 of them for something like $25 on ebay. I got 300 for $75 so I can do 72 poles and the number comes to 288.

http://cgi.ebay.com/100-Neodymium-M...tem&pt=LH_DefaultDomain_0&hash=item5ad69509d0

B444-100pic2.jpg


...as for assembly I plan to epoxy all the north / south poles in place first, then epoxy the halbach cubes in the gaps afterwards. That way the Halbach cubes are trapped in a box and the 1/8" of spacing means that there is more epoxy in place to keep things tight. Also, it's probably a good idea to use some sandpaper on the magnets to dull the coating a little. Epoxy should glue better if it has more to grip.
 
Last edited:
Links to eBay may include affiliate code. If you click on an eBay link and make a purchase, this forum may earn a small commission.
I'm having a hard time imagining how the bearing interacts here... I'm still not getting it.

Maybe by comparing what I have in mind you can tell me the different way you are doing it by contrast.

From the drawing it looks like you are doing a reversed "U" facing inwards and then I sort of lose where you go next.

Explain more... I've read all the CSIRO stuff and they place the magnets on the wheel and then hold the stator stationary inside the wheel... my guess is that you are more closely imitating the CSIRO somehow? But where does the bearing go?

So yours would have the stator in red here (if half the rotor were cutaway)?
attachment.php


Got it in one - whereas your stator pushes in from the outside, mine will push out from the inside. Here's a more complete schematic (once again, very much not to scale - I need to get me some 3D CAD software working...)

attachment.php


The bearing mount (lime green) and the rotor (dark green) bolt onto the hub to the right. The fiberglass stator core (black) bolts onto stator mounting (orange). The stator mounting (orange) then has an arm attached to the left, which attaches to the frame, much like the arm on a coaster brake.

Another question...

How do you plan to get the level of precision required to achieve such close tolerances?

Part of my design effort was to admit to myself that I do not have the high tech tools to create a narrow gap and that's in part why I went with the extra cores because it relaxes the tolerences needed a great deal. Unless you have professional shop equipment it might be hard to achieve such high precision.

The beauty of axial flux is that all the really close tolerances are in the axial direction - in theory, I could make everything I need by plasma cutting it from sheet steel and rounding it off with a file with the parts in a drill press, given a certain amount of laissez-faire on the subject of limb retention :geek:. One could centre the bearings using bolts with plastic spacers - a small amount (i.e. a couple of mm) of radial wobble is fine, it's axial wobble which is the killer here, and squeezing a bearing race between two bits of sheet steel gives a pretty flat rotation in the plane of the sheet steel...

It would be nicer to use the right tools for the job, though, and machine the bearing mounts in aluminium... I have the machining skills, the trick is getting access to a lathe. I have at least one friend who has a suitable machine - I'm keeping the machining requirements to an absolute minimum, though. Also, getting the magnets all lined up nicely may require milling the spacer on a rotary table - I may have to renew some old ties with machinists at the local uni for that... otherwise, it's a dead simple job as milling goes, there are plenty of small engineering shops that could do it without costing an arm and a leg...
 

Attachments

  • Safe Halbach 001.jpg
    Safe Halbach 001.jpg
    96.3 KB · Views: 7,325
  • Rotor full 2.png
    Rotor full 2.png
    8.9 KB · Views: 7,240
Last edited:
I ran a quick simulation on a spreadsheet I've developed using my best guess about what you are up to and arrived at something that looks like this:

I'd be interested to grab that spreadsheet off you - I presume that's using sinusoidal voltage 3-phase excitation? I'm planning on doing something much fancier with power delivery (I iz mode-D-amplifying with a buck convertor in u'r statorz :D) - the waveform used will depend very much on the simulated and measured EMF curves of the system, but will probably end up being close to a 30-50% duty cycle square wave - say, for 40%, 20 +ve / 30 neutral / 20 -ve / 30 neutral. No point pumping current through wires with no back-EMF, right? Also, it's a touch unrealistic to feed the thing >250W all the way from a standing start, that's just asking to turn the wheel into spaghetti and/or burn out the coils, whichever comes first... I'd be feeding it a curve more like 40W at 5mph, 80W at 10mph, 120W at 20mph, 260W at 40mph.

I have a baggie (a few tubes, really) of silly-low-resistance silly-fast-switching trench MOSFETs lying around (P50N06L - 50A, 60V, 0.022 Ohm), which are nice enough, but not ideal for voltage handling - this transistor choice also requires smaller turn count / lower voltages / higher currents - a limit of about 13S (with over-engineered transistor protection), probably 10S, realistically. Smaller turn count gives more room for fast switching, though. The geometry constraints mean I have to pack all my conductors into 8.5mm per phase, too. That graph is for 22 turns per pole, yes? Not going to fit, I don't think...

My current plan is for a paired wave-winding of around 4 turns per winding (8 turns per phase), probably something like two strands of 22AWG in parallel, and relying on the transistors to step-down the voltage and step up the current - I'm going to need a decent inductor on each phase, too... (I've seen someone try this without an inductor - HF thermal shock and skin effect powdered the insulation of the magnet wire in the motor in no time flat - not pretty). As you say, the maths has a ways to go yet... I may just need to widen my gap to fit some more copper in there :unsure:.

Yes, I am familiar with turn count & effective gearing - I was actually considering two windings run through a circuit breaker to switch from series to parallel (while suspending the drive current momentarily) - an electric gear change with a cha-chunk :devilish:.

I didn't realize at first you are using the 1/2" x 1/4" x 1/4" magnets. I got the 1/4" cubes so what you got is a better starting point. It takes twice as many of the smaller cubes to equal your "doubles". If I were to start over I'd go for the "doubles".

Mainly 'cos I really don't like what the magnetic fields of the square blocks do on the inside and outside edges in this configuration. Still, they do cost a bit more - I paid almost $120 AUD for 150 of em, including shipping.
 
Last edited:
So yours would have the stator in red here (if half the rotor were cutaway)?

Yes.

And now I understand what you are doing. By adding an extra bearing you do slightly increase friction, but it's mostly being used as a sort of alignment tool... as long as the bearing is in alignment the stator will be too.

The trick on mine is that since the stator is external and the wheel needs to be able to be moved around to set the chain tension that I'll need the ability to adjust things in three dimensions in order to get it all to work. So it's not going to allow for a quick change if I get a flat. (would have to spend ten minutes getting everything realigned)

Less drag though...
 
I have a baggie (a few tubes, really) of silly-low-resistance silly-fast-switching trench MOSFETs lying around (P50N06L - 50A, 60V, 0.022 Ohm), which are nice enough, but not ideal for voltage handling - this transistor choice also requires smaller turn count / lower voltages / higher currents - a limit of about 13S (with over-engineered transistor protection), probably 10S, realistically. Smaller turn count gives more room for fast switching, though. The geometry constraints mean I have to pack all my conductors into 8.5mm per phase, too. That graph is for 22 turns per pole, yes?

Sounds like you have more experience with designing the controller part. I'm probably going to try to pick your brain on that when I get to doing mine. Since Single Phase produces a tighter copper packing (lowers the kv) all I need to do is make a single H bridge to implement a chip that I've found for doing such a thing. The chip has current limiting built in as a feature and that's something I felt was important. But I'll get to that after I finish the disc fabricating and the stator winding and mounting.

Take a look at the spreadsheet. Keep in mind I've made certain assumptions based on the parts I'm using and the Single Phase design. In the upper right hand corner there is a section where I do FEMM compliance. (lower left are the results) The idea is that you enter a known current value into your FEMM simulation then get the "Force Via Weighted Stress Tensor" value and run that through the spreadsheet to arrive at the "actual" average magnetic flux. As you were mentioning some controllers use full "on" or full "off" current and others use sine waves. When I reintroduced cogging with small cores it skews the data so that it was harder to be confident of the results. It's nice to be able to go from the FEMM simulation directly to a graph and see the results. (it's a bummer that FEMM is not able to simulate a motors motion very well)
 

Attachments

  • Ironless Axial Flux.zip
    71.8 KB · Views: 697
Back
Top