Converting 3-phase to 6-phase by cutting coils at the star point

kubark42

1 W
Joined
May 15, 2020
Messages
61
I have a 3-phase motor where the individual coil leads are very easily accessed at the star point and phase leads. It is a 42 pole, 36 tooth, 20kW, 380A motor which has test data showing 94% efficiency at 20kW.

I am wondering whether it would be practical to cut the leads where the coils join the star point and the phase leads and then to reorganize them to make effectively two electrically independent 3-phase motors, but mechanically 30 degrees apart and in the same casing. While there are already hall sensors, the plane is to drive it in sensorless current control mode, using an ESC on each 3-phase.

The application is a propeller, so there is no need for position control, fast response time, low speed torque, or high-speed field weakening.

The goals are twofold:
  • Higher redundancy than can be achieved with only a single ESC
  • Easier sourcing of ESCs (there's a much wider choice of battle-tested 200A controllers than 400A controllers)

I feel like this ought to work, but I don't fully understand the ramifications of two ESCs acting in the same casing. Will the magnetic fields from one ESC perturb the sensorless feedback of the other? Will efficiency take a hit?
 
I'm not sure about this but i don't think it would work because you'd end up with one ESC on each end of the coil and the effective voltage on the coil would depend on the pwm phase difference which is essentially random.

If you cut the coils in the middle to produce 6 coils (12 wires) then this can almost certainly work, but as you describe,i doubt it.

Try it with a low voltage and current limiter. Would love to know.
 
mxlemming said:
I'm not sure about this but i don't think it would work because you'd end up with one ESC on each end of the coil and the effective voltage on the coil would depend on the pwm phase difference which is essentially random.

I don't understand what you've understood from my question, but I don't think I meant this! I updated it to try to make it clearer.

The idea is that each phase powers 12 coils, and so I can split 12 coils into 2 electrically independent groups of 6 coils each.

mxlemming said:
If you cut the coils in the middle to produce 6 coils (12 wires) then this can almost certainly work, but as you describe,i doubt it.

I don't think I could have access to the middle of a coil. That would require figuring out its midpoint on a stator tooth and then cutting there. That sounds very hard to do and even harder to solder.
 
kubark42 said:
mxlemming said:
I'm not sure about this but i don't think it would work because you'd end up with one ESC on each end of the coil and the effective voltage on the coil would depend on the pwm phase difference which is essentially random.

I don't understand what you've understood from my question, but I don't think I meant this! I updated it to try to make it clearer.

The idea is that each phase powers 12 coils, and so I can split 12 coils into 2 electrically independent groups of 6 coils each.

mxlemming said:
If you cut the coils in the middle to produce 6 coils (12 wires) then this can almost certainly work, but as you describe,i doubt it.

I don't think I could have access to the middle of a coil. That would require figuring out its midpoint on a stator tooth and then cutting there. That sounds very hard to do and even harder to solder.

Your update makes it much easier to understand.

Yes i think that would work, especially for the 36n42p winding scheme which is quite symmetric

The mag fields will interfere with each other a bit, but with that many poles, and assuming you split it at the obvious easy point, the interference will be limited.

This is based on the assumption you have the winding scheme as calculated at https://www.bavaria-direct.co.za/scheme/calculator/ for 36n42p.

In my "terminology" one coil is in your "terminology" 12 coils, so I think we're actually thinking the same just counting the coils differently.

Only issue I see is that if you power only one of 2ESCs there will be asymmetric forces so it might stress it mechanically.

Please post pics and video of you do it. This one's really interesting.

User John in CR seems to have done this but so far I've not found the thread with pics etc.
 
Here's a picture of the back of one of these motors. It's clear how easy it is to access the individual windings. The current configuration is a delta, but with a soldering gun the entire thing can be reconfigured without cutting a single wire.

Screen Shot 2021-11-10 at 09.45.43.png


mxlemming said:
This is based on the assumption you have the winding scheme as calculated at https://www.bavaria-direct.co.za/scheme/calculator/ for 36n42p.

Based on the above picture and the fact that we see 9 soldered connections in 1/4 of the motor, and this implies 36 connections for 36 teeth, that means that two teeth are paired together. I'm guessing it must look like:
Screen Shot 2021-11-10 at 10.01.31.png

only instead of a pair of two teeth running off along the rim of the motor, the wire comes back to the bus bar.

mxlemming said:
Only issue I see is that if you power only one of 2ESCs there will be asymmetric forces so it might stress it mechanically.

With a partial power failure and associated mechanical stress I can live to see another day. With a complete power failure on takeoff, I might not.
 
I'm getting less convinced this (splitting it between 18 and 19) will work because of a subtlety with 36/2 is 18... Fine, still divisible by 3 but 42 divide 2 is 21 which is not even.

One possiblity would be to wind (well reconfigure) it as two single layer motors like:
Screenshot_20211110-195841.png
This would mean either motor on it's own would perform perfectly, and both together... Well why wouldn't they work?

The winding factor for this configuration I believe is the best possible.
 
mxlemming said:
I'm getting less convinced this (splitting it between 18 and 19) will work because of a subtlety with 36/2 is 18... Fine, still divisible by 3 but 42 divide 2 is 21 which is not even.

I don't understand the importance of that. It seems like there would still be symmetry about 120 and 240 degrees?

When I look at the picture of the winding, I realize that the teeth are not done sequentially, but rather each tooth pair is a stand-alone unit. I think what might be happening is what is in this picture:

Screen Shot 2021-11-10 at 18.54.43.png

I'm not sure what to make of that. Is each tooth pair seeing the same voltage, but a different current?

mxlemming said:
One possiblity would be to wind (well reconfigure) it as two single layer motors like:

I think rewinding is not a realistic option. That's significantly more invasive than reconfiguring the copper ring.
 
I meant using the original coils but resoldering them in 2 single layer windings. Might have to snip between the coil pairs though. The windings in that motor look so tidy and nice out would be a shame to ever undo them.

Why not try it? One, any, all of the options? Your motor looks so easy to reconfigure. Just take really good pics so you can put it back to how it started.
 
Nah scrap that. It probably is wired exactly as you say, so just cut the big ring like busbars so there's 3 coils on each half from each phase. Surely that's got to work.
 
It might run into controller / phase synchronization issues if you run 2x 3 phase setups in one motor. It might work itself out and find the sync on its own or might get into a resonance of chasing each other. I think this really depends on the code.
If they do have a hard time syncing I see a few options:
- You might be able to run 1 controller & 3 phase group from the hall and let the other go sensorless. Should be a bit of a delay on the sensorless to allow controller 2 to find it's sync.
- you could setup a degree shift of the hall alignment in the second controller that equals the gap of coil sets so they can share the same hall sensors and know where each group is.
(Both of the above should be easy to try.)
- last one I can think of is making a second hall sensors setup that is shifted the number of degrees the coil groups are offset. I don't think this one is worth pursuing besides as a last attempt.
--This might be just adding a 4th hall sensor and you'd use halls1,2,3 for controller 1 and halls 2,3,4 for controller 2.
 
Jrbe said:
It might run into controller / phase synchronization issues if you run 2x 3 phase setups in one motor. It might work itself out and find the sync on its own or might get into a resonance of chasing each other.

Do you mean while starting up, or a subtle inefficiency which persists as the motor is running?

For startup, I would have only one ESC active. Once the motor is going fast enough for closed loop mode, the other ESC would come online.

Jrbe said:
I think this really depends on the code.

Could you go into this concept a little more? I'm using a VESC, and I've contributed to the VESC codebase a little so I would feel comfortable adding a feature if it were clear what needed to change.

Jrbe said:
If they do have a hard time syncing I see a few options:
- You might be able to run 1 controller & 3 phase group from the hall and let the other go sensorless. Should be a bit of a delay on the sensorless to allow controller 2 to find it's sync.
- you could setup a degree shift of the hall alignment in the second controller that equals the gap of coil sets so they can share the same hall sensors and know where each group is.

While this motor is equipped with hall sensors, my understanding is that it's not good to use them for high power, high current drive. The concern is that the hall fails, and the result is instantaneous and violent as current goes through the wrong phases and FETs. If this is correct (is it?), then I think I have to run both sensorless.

If hall sensors were an option, then I could easily set up the mechanical shift in software. It might as easy as renumbering the hall sensors in the shifted controller.
 
I have no idea what the 2 esc's will do. My reply was logic specific, not esc specific. If it doesn't sync the motor won't run well. It may sync 9/10 times you try, never, always, or may drop out mid run. Or some weird situation where it lets the magic smoke out.
Shifting the timing of the second group of coils could change the efficiency for better or worse. Could add in some interference too.

The code may be fine or may freak out. Using code / logic outside it's normal usage can have strange or completely unexpected results. This being in the air you don't have the safest place to just try it without proving it out. Don't go winning yourself a Darwin award, be safe testing this.

Also, having a strange or multi step startup procedure sounds horrifying to me if for some reason you have to restart while flying. I personally would want this simple and well proven before using it.
 
Jrbe said:
This being in the air you don't have the safest place to just try it without proving it out. Don't go winning yourself a Darwin award, be safe testing this.

Being a teensy bit of a pendant, in the air is actually a really safe place to test this. It's on takeoff when things are riskiest, as there are few options if the glider is 50 meters up and experiences a motor failure.

Jrbe said:
Also, having a strange or multi step startup procedure sounds horrifying to me if for some reason you have to restart while flying.

Hehe, not disagreeing, but here's an example of how some self-launching gliders deploy their propulsion unit: https://youtu.be/0v8Af8O4NeE

In my case, an AC-5M, it's done manually and the existing engine is too heavy to actuate from the cockpit. I actually have to dive down to pick up airspeed, pull back sharply, and then push the plane forward into a zero-gee maneuver so that the weight is off the mechanism. (This is described in full detail, in Russian, in the glider's flight manual.) So there is a multi-step process just to get the motor deployed, and that's before we even start it up.

But, getting to your real point, yes, safety. Safety, safety, safety. The propulsion unit has first to pass a 25 hour ground-run before it gets deployed. And from the pilot's perspective, a staggered startup is completely invisible. The pilot moves one throttle lever, the motor controller code does the rest.
 
If it's on fire mid air what happens? Is it burning / melting the wings or frame?

Glad you have safety in mind.
 
Jrbe said:
If it's on fire mid air what happens?

I step out.

100' later I pull the D-ring on my parachute and watch the fire from a safe distance above.
 
Thought about this more, I can't see any reason why this wouldn't work other than mechanical.

The way the 36 coils are in pairs means the magnetic flux between the phases is as close to zero as possible and you should get any interference between the two ESCs atall. The only problems might be mechanical but that'll probably be ok for as long as it takes to restart/whatever. Unlikely to be terminal.

Have a go! Still really interested.
 
mxlemming said:
Thought about this more, I can't see any reason why this wouldn't work other than mechanical.

The way the 36 coils are in pairs means the magnetic flux between the phases is as close to zero as possible and you should get any interference between the two ESCs atall. The only problems might be mechanical but that'll probably be ok for as long as it takes to restart/whatever. Unlikely to be terminal.

Have a go! Still really interested.

We're headed to the airport in a few hours, wire cutters in hand and fingers crossed. Let's see how this works!
 
Didn't work. Using an almost 500W heat gun, we were able to easily unsolder half the coils, removing three groups of three. However, when we went to run it, the motor simply shook agitatedly. We tried a variety of ways to get the VESC to run it, but it was like it wouldn't advance even in open loop mode.

I'll do an in-depth analysis tomorrow of the motor and see if anything jumps out at me.
 
The pink lines are where we unsoldered the motor windings:

Screen_Shot_2021-11-10_at_18.54.43.png

I wonder if this isn't how we should have done it?

Screen Shot 2021-11-10 at 18.54.43.png
 
kubark42 said:
Didn't work. Using an almost 500W heat gun, we were able to easily unsolder half the coils, removing three groups of three. However, when we went to run it, the motor simply shook agitatedly. We tried a variety of ways to get the VESC to run it, but it was like it wouldn't advance even in open loop mode.

I'll do an in-depth analysis tomorrow of the motor and see if anything jumps out at me.

What exactly did you unsolder? Your pic above of how you think it's wired is clearly wrong but I'd assumed it was the drawing equivalent of a typo...

Scope on the phases to see the BEMF?

I can't see how this could not work given the coil pairs all seem to be wired in parallel.
 
Motor wound like this.jpg

Surely the wires must be attached something like this.
 
mxlemming said:
Surely the wires must be attached something like this.

You are, of course, very correct.

20211113_105103.jpg

Here's the motor as it stands. It's a little hard to see the unsoldered leads in this picture, but you can get the idea:

20211114_023612.jpg

The blue tape delineates subgroups of +-AB/+-AC/+-BC windings. You can see how we unsoldered all leads in three of the subgroups.

I'm also sure that none of the leads were accidentally shorting to the phase rings when we were testing. And while I'll check that none of the remaining windings could be harmed, I don't see how since the only mechanism would be heat and we'd expect the first leads damaged to be the ones we were unsoldering, as that's where the heat was concentrated the most.
 
Works for one half! Must have been a settings issue, I upgraded to the latest firmware and restarted the setup. Now it seems to run just as it should.

Now need to get a second controller and give 2x3 a try.
 
I did think it was really weird that it didn't work. I spent ages looking at your pics. Far longer than I should have.

The joys of VESC... I have had to reflash the firmware using STLink before to get it to stop being confused.

New firmware so much better though.
 
mxlemming said:
I spent ages looking at your pics. Far longer than I should have.

I am really sorry for that, but I wanted to say a sincere thanks for encouraging me to try the modification. I think I'd still be trying to wrap my head around it if it hadn't been for your help.

VESC has matured a lot, and we're doing what we can to move it even further along. There's an exciting future for VESC, I think it has cracked the code for a viable open-source high-power ESC.
 
Back
Top