I am thinking about scaling this up in size and stacking motors and rotors to build a lighter and more efficient hub motor for one person electric vehicles.

Very cool..wonder if you could make a via large enough (between the coil via's) to slip a metal nail inside the coil to focus the electromagnetic force (say a thumb tack or rivet? [epoxy in]) very very cool design!

Yes it is possible but its size will need to increase

Good hacking,learning project !

A few comments/questions.

1) it seems that you want to increase the torque. (buy you did not specified if static or dynamic, the way you seem to be testing it is giving you results for static torque.  most real world applications require measuring dynamic torque.  it is indeed more dificult to test in a home-lab, but still affordable and doable.

2) The usage of 'round' magnets, or though cheaper, is not optimal. (Square/rectangular magnets optimizes the 'surface' of the magentic field in an easy'er to capture coil. the coil would need to be shaped as a triangle, but the shape maximizes the 'flux'/Area for any given size.   ... look up axial-flux generator and reverse the power-transfer. the equations  are the same. (*well not exacly, but close enough)

3) given the planar orientation designed, perhaps dissasably of old floppy-drive motors would teach a lot. (their magnets were a single round cylinder with 'weird' magnetization, while preserving the planar/6 coils structure.   not sure the actual static/dynamic torque, but they were able to spin the at over  300rpm... indicative of the Dynamic torque)

4) you did not specify the ferrite material used. (is it ferrite is specified by mixture i.e. mix-43/-68/-71 etc. usually color coded....)  the actual mix has a 'HUGE' influence in the (eddy)losses which of course make it 'heat-up' the material.   such loses  are dependent on both the material and the frequency of operation. (i.e. a single pulse of 1000ms would have a given loss, a same pulse of 2000ms would have a totaly different loss. of course the worst case scenario is continuous current (no pulse) as the 'magentic' field would be constant-steady and  heating the material and would produce no alternating-flux hence no power/torque.

5) in reference to #1 and #4 above, Triaminic drivers specs/explanation pages (all the rage in 3d printer motor drivers) give very clear insight (at a layman's level) of the static/dynamic and totat power (i.e. work) expected from any given setup.

6) I'm a bit confused when you specify your magnets/setup give 1W did you mean 1Tesla?Gauss? .. (magnets by themselves don't give power...) or you meant that the 'motor' consumes 1W to deliver 9gr of static torque?

7) It appears to me, that the easy way to radically improve the torque, without redesing) is to decrease the  magnet to coil distance (if using Mag. above and another below, decrease this distance.) i.e. minimize the distance between the surface of the coil and the magnets.  (dont forget to add the height of you board material in the equation !)    

7a) if using a single plane of magnets (pictures seem to show one above, one below.. ) would almost double your power (torque) .... CORRECTION: mean to say using one above in below would double your power...

Sorry, way too long a comment. :-} 

Good Luck!

Carlos 

Further comments on the ferrite. 

In order to be effective the ferrite must be at the center of the coil and be as deep as the coil itself. Placing  as you did makes it behave as a shield!! (eddy currents in the ferrite sheet create a magnetic field OPPOSING the desired magnetic field ON ALL COILS!.)   ie. is behaves as if a steel plate was between your magnets and your coils, all the while heating it up!.

Also, your coils do not need to go all the way to the center, (that space would be much better used to put a tiny ferrite bead instead)  lowering the coil length will also have the desired effect of reducing the ohm resistance, a therefore allowing a higher current with the same voltage.  since the magnetic flux is proportional to current, your flux is much higher, hence higher extracted power. (more efficient.)  my -personal- Rule of thumb is for the center of the coil to be about 1/3 of the total coil width.

beware that the ferrite bead in the center would be subject to strong axial-push/pull forces. i.e. up down, best  solution is to use 'threaded' beads and thread the PCB accordingly. (adding a touch of crazy glue would not hurt.)

Magnets.

The actual magnets were not specified, but some have WAY more pulling force than others(flux), simply changing the actual magnets can substantially increase their flux/power.      Carefull thought, higher flux magnets de-magnetize at lower temperatures ... i.e. proper cooling may be required for a HIGH-power (i.e. drone motor) application...

Hi! Thank you for your comment and suggestions :)

In the shown video i was testing for static torque which would be more beneficial for servo like applications. As you said dynamic torque is harder to test.

Regarding the triangular shaped coil, this would definitely maximise the flux area of the stator however this shape would have less turns (around 7 on each layer) for this motor-size which would significantly reduce the impedance of the windings and thus heat up. This also applies for the core idea you proposed, its something that is great to have but in practice i'm not so confident it will work due to lack of turns. But this is definitely something that can be considered for larger sized pcb motors though.

The static torque test was done with a constant power supply voltage of 5V. The motor was drawing around 200mA of current so it can be rated to around 1W.

Ok, not really 100%sure static torque is better for modelizing servos, but to each it's own.

Not really sure why you say a triangle-shape would have less turns... (mind you, even the circular-shaped coils don't need to coil all the way to the center, the extra loop's contribution to flux is quite minimal and the extra ohmic resistance far outweights  its benefits, specially since they become 'heating' elements). Regarding the impedance, a lower impedance means more current , which means a more intense flux, which means a more 'power-full' (holding) torque  which is what you wanted!    If heating is a problem (perhaps your 'wire'/board traces's cross-section, is too small (hence their resistance is too great, and act more as a 'resistor' instead of a coil)    

One way to decrease the current is to lower the voltage (ohms law), but an easier way (if using drivers such as the 4988/drv8825/triminic 2100 etc. etc etc. ) is to Increase the frequency of the pwm signals. (mind you, doing so would require-you to calculate the actual reactance (impendace) of the coils for the new frequency. (actually easier to do trial-error, than to acurately model it via maths)

Another way to achieve the same result is to Lower the voltage, instead of driving it with a 5 volt source, try a 3.3 v. inversly, you can drive it with a 12 volt source (watch your driver's specs.) and LOWER the max current limit of the driver. either way you'll limit the input power, hence the torque. (the input-power to torque curve is NOT linear, it is indeed a log. curve !  (log/antilog depends how you plot it).

Also, it occurs to me, given the size of the motor,  that instead of using ferrites it would be easier to use steel screws. (make sure they can be magnetized, not all steel can be ) iron screws would be better, but they rust quite fast ..

Gotcha for the 1W calc. makes sense.  However, (haven't actually done the math but, gut feeling,  it appears to me quite to be very IN-efficient. 

ANOTHER type of motor (see axial-flux generator's coil/magnet arrangement) in which 2 coils face 1 magnet can be designed, it would allow for the same current (at 2times the voltage) to 'repel'(or attract)  the magnet with the same force. HOWEVER, the 'cooling' surface is twice as big, and therefore the accumulated heat is much lower (not quite half, but thereabouts) at the same "power-input to torque" point. .... (perhaps I don't make myself clear, I'm at a loss to how to explain it better.... :-|

Alternatively, a *much* more complicated 'geared' motor could be done. (specially for servos, as they don't really need to move very fast.) the added advantage of a geared fast moving rotor, is that a simple fan-blade can be attached to cool it down.  and the static-torque is multiplied by the gear-ratio. 

PPS: the ferrite material you used is primarily used for shielding, (very high loss) basically it is designed to convert the Rf/Em to heat !. exactly what you do not want in this instance. 

Sorry I know it was a very long reply :-/

For that prototype, my goal was to make it as small as possible, and the spiral shaped coils have been designed to fit in the smallest area possible with the minmal clearance. For the triangular shapes coils, the stator diameter would result in a slightly larger. However, I am currently designing another board to test slightly larger motors (less than 20mm diameter) with different coil shapes.

There is a balance between the amount of current that you pass through the coils and heat. From what i have tested with these coreless printed coils the limit is around 40-turns for 4/4mil traces and 60-turns for 5/5mil. 

Using iron screws i the middle is something that I have considered. These would act as the core of the stator. But my original goal for this project was to try and make the smallest and easiest to assemble motor, and adding these screws would defiantly increase the size and complexity of the build. 

I need a motor driver details

just google "uln2003 stepper driver schematic"

What is the driving circuitry for this motor? Can you please share the schematic?

Fantastic project - Perfect for solving a headache I had in my project - Thank you. I have ordered a few off OSHPark.. :) ill upload project when further down the road.

I am currently using them in a prototype that will be a fair bit larger than the final version of my project. I am unfamiliar with the limits of PCB manufacture and wonder if you can enlighten me? Do you think it could be made any smaller? In terms of Diameter? Less windings, thinner traces? Looking to get it down to 10mm so about 35% smaller. Obvious huge reduction in power etc.. but still functional? I thought to ask you as you may have already experimented with further miniaturisation and come across some limitations. Be great to get your thoughts! Thank you again for a great concept!

Hi! It is very difficult to make it smaller than 16mm. In my prototypye I am using 4/4mil traces on a 4 layer board. I have around 10 turns on each layer adding up to a total of 40 turns.  I don't recommend going much lower than 40turns with 4/4mil, as these would reduce the phase resistance and increase the overall temperature of the pcb. The only remaining options to reduce the area would be either use smaller manufacturing traces and clearance (which would be much more expensive for a supplier to manufacturer) or increasing the number of layers. However, you need to be careful as increasing the layer would mean you have to leave a larger area in the middle of the coil for more vias.

That is really useful to know, thank you Carl for your time and considered response. Very appreciated. I may look at 2/2 mil traces in the future if the prototype has any promise. In the meantime though I am very happy with the boards, that have just arrived!

Awesome project, it'll be great to see people using this!

Thanks!

I was curious, where did you get the semicircular magnets from?

I'm quite interested in this as well. Can't find anything of the same size. Any help is highly appreciated.

Hi :) Those were both custom from a manufacture I found on alibaba.com 

How much for how many units?

Does this DC motor need a hall effect sensor to control it? What controls the sequencing of the windings like a usual brushless DC motor?

My plan was to use a sensorless back-emf speed controller, however the measured back-emf was too weak to implement this. I'm now working on a prototype with hall sensor for speed and position control.

Yeah, that makes sense. Would be cool to get the PCB coils strong enough to go sensorless. I bet you could do it if you made the coil strips longer.

great project, I am trying stepper motor as direct drive wheel for robot https://youtu.be/78K-oa2GHU4 instead of printed rotor I am using PCB rotor (3 layers of 1mm PCB and 5x3mm magnets) https://oshpark.com/shared_projects/V5EU4Lhf

cool :) just make sure you have enough torque to rotate the wheels and move the robot

This is really cool. Thank you for sharing the Gerber files!

I'm interested in using this approach for building an axial flux generator. I can't find any existing designs that fit my needs and making PCBs seems easier than hand windng coils and potting them.  For those of us new to the PCB design world, would you mind providing an outline of the tools you used to design your PCB?

Hi Peter! I used a CAD tool to draw the spiral coil windings for each layer. These layers were saved as a dxf file and then imported into a PCB design software as different copper layers.

maybe ferrite foil in the back would be beneficial?

http://katalog.we-online.de/en/pbs/WE-FSFS and the like.

Do you have a power requirement for your drone?

I'm curious how you are driving the motor? Is it just from the Digital IO pins on the Arduino? Or are there transistors, FETs, etc...

Hi I'm currently using the STSPIN230 3-phase driver (availble on the X-NUCLEO-IHM11M1 dev board). I am then driving it with a dPic microcontroller.

Hello , can you please release the pcb design files , or just draw a schematic of the coils and all the details?

Hi! The files are now available for download

great idea!

Thanks :)

I that HP MultiJet material or your own print for black rotor ?

Hi i have 3d printed from shapeways

This is such a great idea. I agree it would work better with a different stator/magnet configuration. It would be interesting to see how torque and kv would be affected with different thickness traces. And how much more output and balance you could achieve with two magnet sets sandwiching the stator. 

I can see a revolution in quad copters coming on. It would be amazing to see this on the crazyflie (open source)  quad

Sure :) a rotor combined with propeller is coming soon

This is neat! I would like to try to build my own, and also have some things I'd like to try to improve the design:
Most axial flux motors have magnets on both sides of the stator. The purpose of this is to not just to fit in more magnet, but to make sure all of the magnetic flux passes perpendicularly through the stator - with your current design, some of it wraps around the magnets and doesn't even make it through the PCB. Likewise, on the rotors behind the magnets is iron, to complete the magnetic circuit and get better utilization of the magnets. Iron-filled plastic could substitute here, though it's less important and just using beefier magnets might also be OK.

The downside of facing rotors is there is a huge amount of force between them. You'll either need strong 3D printed parts, or attach them on the outer diameter, running the wires through a non-rotating axle.
For bigger motors, you can get a better winding factor with an 8/9 or 10/12 arrangement. If you don't already know about it, there's a nice winding calculator here: https://www.emetor.com/edit/windings/

Note that coreless axial flux motors like this can be quite efficient. You might find some inspiration from the CSIRO design often used by solar cars. http://www.ata.org.au/wp-content/uploads/marand_high_efficiency_motor.pdf

Hi Thomas! Thanks for the tips! For the second prototype I have order semicircular magnets so that all magnetic flux pass through the stator, just like you suggested :) stay tuned for the update

Excellent feedback. Btw, I tried the iron-filled plastic myself  some time ago when I wanted to make a linear actuator with 3d-printed parts and the results were quite disappointing. The effective permeability one gets ends up being around 2, which is quite disappointing (https://hackaday.io/project/11082-measuring-blackmagic3ds-ferromagnetic-filament).

Thin out your rotor and add a second stator above.  Gives you your second bearing mount, doubles your output torque, and will probably be more stable as both sides will push/pull the rotor at the same time.

Hi Daren cool idea! I would consider adding another rotor to have magnets on both sides rather than adding another stator. This will definitely help increase the torque,  

Torque comes from flux density.  There are a couple ways to increase that. 

Simplest way is to use a core material.  This compresses your field lines for your electromagnet yielding higher flux density, at the expense of additional inductance.  But that doesn't really fit here.

Next simplest would be to compress/redirect the field lines from your permanent magnets back towards your electromagnets.  This is usually accomplished by backing them with iron. A washer or the bell from a fridge magnet might do it.

So where I was going with the second stator is related to the latter.  Besides doubling the flux density of the electromagnets, you are also confining the stray back field lines from your rotor magnets.  Of course I could be full of it, but it works in my head..