DIY Toroidal Axial Flux PM

[HalbachHero]

i dont believe the coils experience the pwm frequency. if they did it would need to be increased and motor would be getting unnecessarily hot.

I would agree with this thought without actually knowing the answer. which if thats the case, and the coils are experiencing a frequency of 2kHz, it seems like 0.2mm diameter is actually ideal

 

[Hummina Shadeeba]

so 10k rpm would be 120,000 electrical frequency per minute with your motor.. and divide by 60 to get Hz right. so 2000Hz.

yea the motor doesnt experience the pwm frequency im sure.

10k rpm sounds insane. maybe with axial flux design it would need more to hold it together at such a speed. to hold the magnets.

 

[Thecoco974]

About copper section, another important thing to take into consideration for correless stators is Eddy losses.
A simple formula to estimate the losses per unit of mass :



Applied to our case : (3.14/6)*(0.0002m^2*0.7T^2*2000hz^2)/(2*17E-9ohm.m*8960kg/m3) = 423.33W/kg

So if you have 1kg of copper it would be 423W dissipated as heat additionnaly to the regular conduction losses. It's quite important to take that into consideration with coreless (not an issue with iron core since the sinusoidal magnetic flux isn't going through the copper).

With e being squared if you double the wire diameter the losses are going up by a factor of 4 .. so 1693 W/kg.

 

[Jrbe]

It's the thought that capacitors, inductors, or something else is filtering / smoothing out the pwm frequency?
I'm trying to understand why the motor wouldn't see this. My understanding of inverter rated motors is that they have higher voltage ratings in the coils to deal with voltage spikes of the carrier / pwm frequency.
The Litz link I posted has links in the body that jump to different parts in the page but also jump to specific info, calculations, etc. They only list a single wire frequency from what I saw, not sure what multiple wires does to their recommended frequency. https://www.elektrisola.com/en-us/Litz-Wire/Info-Details#single-wire-diameter-vs-range-of-frequency

Laser cut parts end up with a bit of slag on the back as the metal is cut. This can keep things from sitting flat. It depends on how well setup the laser is but usually still has some slag. Just something to be aware of if you're thinking of cutting windows for the magnets or other things that might not end up perfectly flat. You can put the slag side in the back. If you need both sides flat they can likely "timesave" it - sending it thorough an accurate belt sander.
Normally you can cut hole diameters down to the thickness of the material.

10krpm is a lot, especially for the size this looks to be. Not sure how you could diy balance the motor to achieve that.
If you can use flanged bearings that pickup your back iron it could help a bunch with concentricity / runout (you can call out the bearing bore ID, circle correction, and min / max diameter.)
How well is the glue holding your magnets? Having a lip for the magnets to seat against would help to resist their centrifugal forces and could help take away some magnet concentricity issues when glueing. You could copy your back iron part and cut out a circle for the magnets to sit in and stack them together.
Might be worth making a scatter shield if something goes wrong.
If you haven't, check that your bearings can handle the rpm you're planning.

 

[HalbachHero]

10k rpm sounds insane. maybe with axial flux design it would need more to hold it together at such a speed. to hold the magnets.

Yes 10k is a lot, but I have successfully ran it at 7k without it breaking, when it was a higher Kv. This did not blow up. There are "guards" that hold the magnets concentrically. These so far, have been about as thick as the magnets, and were epoxied to the rest of the rotor. This has changed a bit in the design, but there is still something there to hold things.

The air core motors benefit from being run at high RPM. So ideally I want it to run fast, but 10k would be the upper limit of what I
would want to build it to handle if possible. Also, even by cutting that in half, you are still running 1kHz, which is still at the bottom range of the 0.2mm diameter recommendation.

About copper section, another important thing to take into consideration for correless stators is Eddy losses.
A simple formula to estimate the losses per unit of mass :

Applied to our case : (3.14/6)*(0.0002m^2*0.7T^2*2000hz^2)/(2*17E-9ohm.m*8960kg/m3) = 423.33W/kg

So if you have 1kg of copper it would be 423W dissipated as heat additionnaly to the regular conduction losses. It's quite important to take that into consideration with coreless (not an issue with iron core since the sinusoidal magnetic flux isn't going through the copper).

With e being squared if you double the wire diameter the losses are going up by a factor of 4 .. so 1693 W/kg.

Shoot, yeah, thats a lot of losses... but certainly there's an optimization to he made here as well right? balancing the losses between air core/ iron core. This goes back to the idea of the "hybrid" core basically not enough iron to prevent saturation, but also to limit iron losses. I haven't done the math for this yet, but it seems like it could be a worthwhile practice.

It's the thought that capacitors, inductors, or something else is filtering / smoothing out the pwm frequency?
I'm trying to understand why the motor wouldn't see this. My understanding of inverter rated motors is that they have higher voltage ratings in the coils to deal with voltage spikes of the carrier / pwm frequency.
The Litz link I posted has links in the body that jump to different parts in the page but also jump to specific info, calculations, etc. They only list a single wire frequency from what I saw, not sure what multiple wires does to their recommended frequency. https://www.elektrisola.com/en-us/Litz-Wire/Info-Details#single-wire-diameter-vs-range-of-frequency

The thought is basically that all motors would be suffering from these losses if they regularly experienced that sort of frequency. granted generally running less than 10k RPM, but even at a fraction of the PWM freq, it would still have an impact. So really its just a thought. I don't know enough about the commutation to have a good answer.

Laser cut parts end up with a bit of slag on the back as the metal is cut. This can keep things from sitting flat. It depends on how well setup the laser is but usually still has some slag. Just something to be aware of if you're thinking of cutting windows for the magnets or other things that might not end up perfectly flat. You can put the slag side in the back. If you need both sides flat they can likely "timesave" it - sending it thorough an accurate belt sander.
Normally you can cut hole diameters down to the thickness of the material.

Yeah this is something I noticed when I ordered the parts from SendCutSend. They did not deburr anything, so they had a rough side. This was not really an issue. I could easily sand them down flat enough for my means.

10krpm is a lot, especially for the size this looks to be. Not sure how you could diy balance the motor to achieve that.
If you can use flanged bearings that pickup your back iron it could help a bunch with concentricity / runout (you can call out the bearing bore ID, circle correction, and min / max diameter.)
How well is the glue holding your magnets? Having a lip for the magnets to seat against would help to resist their centrifugal forces and could help take away some magnet concentricity issues when glueing. You could copy your back iron part and cut out a circle for the magnets to sit in and stack them together.
Might be worth making a scatter shield if something goes wrong.
If you haven't, check that your bearings can handle the rpm you're planning.

a flanged bearing and ID mounting point on the back iron is one of the things I would like to change about the design, but I am currently working with the parts I have since I already have them. But yeah I would like to have a removable center hub for the rotors that carries the bearings and stator. Eventually it will get there. Also, the bearings I am using I have seen rated for anything from 13k to 20k RPM when greased which they are.


here you can see the part that sits on top of the back iron, holding the magnets in their position and spacing them out evenly

 

[mxlemming]

10krpm/60x12pp is 2000hz electrical frequency... This is the correct answer

Re. The skin effect, you need to consider the current induced in the wires not the voltage on the wires.

For this you need to consider the drive voltage (bus voltage), the back emf and the inductance. From this you can derive the current waveform, which can then be considered for its harmonic content.

The skin effect will be different for the current at each harmonic.

 

[HalbachHero]

For this you need to consider the drive voltage (bus voltage), the back emf and the inductance. From this you can derive the current waveform, which can then be considered for its harmonic content.

Are you saying that the harmonics check that I did was sort of meaningless?


I completed the next set of rotors, and they look pretty good. I ran out of blue PLA, and needs red for another project, so were back to red.





One of them I cleaned up a bit better than the other, and its black epoxy, so it makes it look messy. Going to let the epoxy set a bit more before playing with them, but I cant get anything to stick on the back. there seems to be virtually nothing leaking out the back.

 

[mxlemming]

For this you need to consider the drive voltage (bus voltage), the back emf and the inductance. From this you can derive the current waveform, which can then be considered for its harmonic content.

Are you saying that the harmonics check that I did was sort of meaningless?

Quite the opposite.

The better the voltage field the fewer voltage harmonics there are.

This will induce fewer current harmonics. Therefore less skin effect.

 

[HalbachHero]

After letting the epoxy set a bit more and making a spacer. I put it together. It really snaps together now. its a little concerning. The fins on the spacer dont seem strong enough to keep it apart. I checked to see if it was bending at all. This is definitely the most rigid one yet.


You can see theres a slight bend to it, and that will obviously be even more exaggerated when its in operation, but its still much better than where it was a few versions ago.


I tried to put the most recent stator in it, but I spacer was not tall enough because of how big the airgap is on one side. But heres some photos of it assembled.



You can see the issues im having with the printer. Some parts are under extruded and the back face is only 0.8mm thick, which is 4 layers. So you can see the epoxy literally leaking through the layers. Its all pretty stiff though. I printed quite a few before I got one that I was happy enough with, so hopefully the rest of it is strong enough to hold up for a bit at least. If the issues are purely aesthetic, I can paint it or something.

Im making a new spacer now to give it a test spin and will report back soon.

The better the voltage field the fewer voltage harmonics there are.

This will induce fewer current harmonics. Therefore less skin effect.

I need to read more, I still don't completely understand what the "voltage field" is, and if you are saying that this is experiencing fewer harmonics, and therefore less skin effect, does that imply that a larger diameter conductor might be suitable?
is there a measurable metric for skin effect. Could I play with larger diameter wire and measure this to find an optimal diameter wire? I worry about anything high current in this 32 gauge wire, its very thin. If I could go larger, while still achieving similar fill factor, and being able to carry higher current while also mitigating skin effects/ harmonics/ and other losses, I would like to do so

 

[HalbachHero]

Okay so I have some results and they are looking pretty nice.

I used the Mk5 stator on the new rotors (going to call that Mk7) and the Mk5/6 rotors. This was the last 3d printed stator I made and had 6 turns of 20/32 litz wire. I used a drill to spin the motors at a constant speed and recorded ~25 samples for each test.

Mk5 stator with Mk5 rotors

Voltage: 1.84V
Freq: 97.89Hz
(97.89 / 12) * 60 = 489.45RPM
489.45 / 1.84 = ~266Kv

I originally estimated this to be 240, but showing the number helps get an accurate number.


Mk5 stator with Mk7 rotors (I actually did this test first, and the drill battery was dying, so the freq is lower)

Voltage: 1.96V
Freq: 74.831Hz
(74.831 / 12 ) * 60 = 374.155RPM
374.155 / 1.96 = ~191Kv

That's a 28% decrease, which I feel is quite substantial given the subtle differences with the back iron.

I then tried the Mk6 stator (7 Turns of 20/32 litz wire) with the Mk7 rotors (There is a large air gap on one side that I cant fix without making a new stator. So the numbers are lower given more turns, but still promising)

Voltage: 2.52V
Freq: 97.94Hz
(97.94 / 12 ) * 60 = 489.7RPM
489.7 / 2.52 = ~194Kv



Also, check out those harmonics, or rather, lack thereof

 

[mxlemming]

The better the voltage field the fewer voltage harmonics there are.

This will induce fewer current harmonics. Therefore less skin effect.

I need to read more, I still don't completely understand what the "voltage field" is, and if you are saying that this is experiencing fewer harmonics, and therefore less skin effect, does that imply that a larger diameter conductor might be suitable?
is there a measurable metric for skin effect. Could I play with larger diameter wire and measure this to find an optimal diameter wire? I worry about anything high current in this 32 gauge wire, its very thin. If I could go larger, while still achieving similar fill factor, and being able to carry higher current while also mitigating skin effects/ harmonics/ and other losses, I would like to do so

The controller generates a voltage. (Approximately via pwm)
The current is then equal to (V controller-Vbemf)/complex impedance (resistance and inductance)

The skin effect is caused by the magnetic field caused by the current flow, therefore in calculating the skin effect you need to consider the current not the voltage.

But the back EMF of the motor determines the current in the windings indirectly... Less BEMF error, less current error.

If your BEMF has lower harmonics, your current will have lower harmonics. If the current has lower harmonics the losses to IR of those harmonics (which don't generate torque) will be lower.

Those harmonics have a shallower skin depth than the fundamental (motor erpm) but don't contribute to the fundamentals skin depth (can apply the principle of linear superposition I believe).

Yes I believe you could use thicker wires from a skin depth perspective. As someone else pointed out though, you'll need to consider Eddie losses in the wire which you wouldn't see with an iron core.

In any case, I don't think litz wire is necessary but you probably want to stay below about 0.25mm diameter from the Eddie current perspective.

 

[APL]

Probably a dumb question, but since the motor is being driven by another motor, could it be acting like a sensor and picking up
'some' fluctuations from the drill?

 

[mxlemming]

Probably a dumb question, but since the motor is being driven by another motor, could it be acting like a sensor and picking up
'some' fluctuations from the drill?

Yes it absolutely could, but in this case, the FFT shows it really isn't...

 

[HalbachHero]

I collected a bit more data using the new rotor.

This is once again using the drill to spin the rotor on the test coils I made. Each coil is 2 turns of 20/32 litz wire. I did three tests with each group of channels. I also noticed the average frequency dropping slowly over time, so I incorporated that change as an approximation based on the order of the tests. Remember, the goal is to minimize Kv, the lower the better

The first test incrementally adds 1mm to the outer diameter


The second incrementally subtracts 1mm from the ID


I feel like this is a bit more conclusive data, and now I want to test making the coils smaller and seeing how rapid the decline is. I found that once you to go to +-4mm with this test, the voltage starts to drop off, which gives me a outer/inner limit to the size of the stator I think.

The inner diameter seems to also be more tolerant to the change in length as well, which I think makes sense because the magnets are closer together.

The issues I still have with this test, are that it only includes one rotor, which I think might affect the flux path quite a bit. Also the gap between the test coils and the magnets, its not analogous to the stator.

Maybe those issues are enough to make this bad data. I don't know. Either way, it's interesting.

 

[HalbachHero]

So over the last week or so I made some decent progress on a side project that I have been working on. The Vi Engineering MPCNC. I have printed and purchased the parts I needed a while ago, but finally decided to move forward with the build.



I think it came out alright. I am hoping I can use this to make my own back irons, we'll see, maybe cutting steel is a little too ambitious



On the motor front, I have made a 2 new test stator things (ive been calling them octopi), each one with 6 tests and one "control". I think these ones will solve some of the problems I had with the previous one. These will all sit very close to the magnets like the stator would, and I will be able to put both rotors on. One test has a static ID and a varying OD (1mm increments), and the other has a static OD with varying ID, same increments. These also dont have the rounded edges which I think created some variability in the lengths. Once I think I have nailed down an optimum, I'll make the new winding jig and test with the nylon




Oh also I did win the fight against the printer. LFG!

 

[APL]

Good work. :thumb: Also, very interested in how the MPCNC turns out! I've contemplated setting up something like that for some
time now. It will be nice to see how this one works for your various motor parts.

 

[HalbachHero]

Good work. :thumb: Also, very interested in how the MPCNC turns out! I've contemplated setting up something like that for some
time now. It will be nice to see how this one works for your various motor parts.

Hey thanks, yeah funny enough. I head read about the MPCNC project before, but ultimately it was your thread that really made me consider it, after someone else had posted about it. I am also really looking forward to seeing how it works, and if it can eventually cut metal.

if you or anyone else is curious. I am using a SKR 1.4 Turbo and a TFT35 v2.0 running marlin, I have a Meanwell 450, but I haven't wired it up yet, just using some power supply the SKR board came with to do some plotting with a marker.
I have a raspberry pi with a flavor of octopi installed. So I'm using octoprint and cncjs to control it. (Most of this I found through V1 Engineering's forums). I printed some cases for the electronics. I have some more I want to do to tidy it up too, but ill get there.
Still need to learn the software a bit, but I was able to draw some pictures, next step is foam, but I don't have a tool yet




I finished up with the test stators, and I think this is even more reliable data. I used the two stators, one had a static ID, one with a static OD. two turns of 32/20 litz wire. this time with both rotors (12 poles with halbach and back iron)



after staring at this a bit and triple checking that things were correct... I started thinking. Maybe this is micro-optimization, and you know what they say about that....
But I think what I can glean from it, is that its more tolerant to change on the inner diameter, beyond the magnet edge. But thats not the case on the outer diameter. Therefore, I think I can save myself some copper on the OD and move it to the ID. This hopefully will optimize the resistive losses and at the same time vitally shrinking the OD a bit. reducing inertia, the strength needed from the back iron and the cantilevered magnets, weight, etc.

just a rough guess from knowing the diameter of things, I might be able to get it down from 122mm diameter to ~107-110, its not a ton, but if there's any where to lose weight and size, its there. might have to lose the cool impeller though. shoot

 

[HalbachHero]

I realize that the scale I set things too is a bit misleading when comparing the same test stators together. I was using it to compare the difference in change inside or outside of the magnet edge.
This is a more direct comparison of the same data which makes my conclusion a bit more obvious I think.
and again, lower kV is better here

 

[HalbachHero]

I changed the model for the motor once again, incorporating the findings from the previous experiment, and I was really able to bring in the diameter. its went from 122 to 106. The number of poles has not changed at all, so I should expect to get about the same amount of power out of a smaller package.
Also there are a few other changes to the design I made. the back iron for the two sides are no longer the same. one of them has a mounting point in the center so that I can change the center of the motor out if needed. Also this hub for the rotor, now has a much larger diameter. This is to hopefully minimize the amount of flex it has. Before by hand, you could manipulate it in such a way that it would drag. This will beef it up a bit. Oh and I did go back to not using an impeller, though there might still be room for a small one.... otherwise I just shrunk things everywhere I could, and made the model a bit more parametric.



I also modeled the new winding jig given the new dimensions, so I will hopefully have that soon, and will be able to start making nylon stators.

I also would like to do a test to compare the fiberglass sheets vs the fibers mixed with the epoxy to see which one will deflect the least. Also I will be getting some materials together to make silicone molds for the press, as was suggested.

happy with the progress so far

 

[HalbachHero]

Got the winding jig re-printed with the new dimensions. Im loving how small this version will be. I wound up a 4 turn nylon stator which I will tie together.

I have all of the parts for the mold printed. and the stuff to make the silicone mold showed up in the mail. So assuming the mold making goes well, I should have some test "stators" made of nylon soon. If the mold and press work out well, I will go back to doing copper, but I imagine there will be a few iterations of mold and presses before Im happy with it.

New bearings and the custom rotor parts will be here this evening too, I already have everything I need for those, so I may have a new set of rotors even sooner.

 

[TorontoBuilder]

Got the winding jig re-printed with the new dimensions. Im loving how small this version will be. I wound up a 4 turn nylon stator which I will tie together.

I have all of the parts for the mold printed. and the stuff to make the silicone mold showed up in the mail. So assuming the mold making goes well, I should have some test "stators" made of nylon soon. If the mold and press work out well, I will go back to doing copper, but I imagine there will be a few iterations of mold and presses before Im happy with it.

New bearings and the custom rotor parts will be here this evening too, I already have everything I need for those, so I may have a new set of rotors even sooner.

resized-image-Promo - 2021-12-16T125108.587.jpeg

Very nice. I'm just about ready to start printing some test motors before tackling my own build. I'm going to copy your design since it is such a good one.

 

[HalbachHero]

Very nice. I'm just about ready to start printing some test motors before tackling my own build. I'm going to copy your design since it is such a good one.

Nice! cant wait to see it.



I finished up tying the nylon stator. ill use this as the first test stator, and start winding up another one. Also the custom rotor parts showed up. They are a similar area to the last ones, but obviously a decent size smaller, but they are also made of a slightly different material, called Chromoly. its 0.1mm thicker at 1.6mm. Supposedly its tougher, so im hoping for less deformation in the rotors once they are both on, but we will see!

 

[Vbruun]

Beware that magnetic properties change between steels. I don't know if it matters in this case though.

 

[HalbachHero]

Yeah its certainly magnetic still, and some brief reading didn't lead me to believe I have anything to be concerned about, but I should be able to tell if its leaking flux by doing the paper clip test on the back of the rotors once they are made.

The last ones seemed to have nearly 0 leakage. Hopefully I see the same results with these.

 

[Jrbe]

Great job optimizing all of the parts, these look excellent!

Do you have metal collars for the bolts between the 2 back irons? Just having plastic there is likely adding to the flexibility. Ideally the bolt hole collars would hold the shape of the 2 back irons with just the irons, bolts, and spacers bolted together. You can experiment without the plastic parts in the metal stack then mod the plastic parts to fit or reprint if necessary.

You could likely print an impeller that bolts onto the side.