Multiple controllers on parallel windings

[John in CR]

Like many, I've had it with blown controllers while shooting for higher power. I can't rewind my motors for 6 or 9 phases and simply use 2 or 3 controllers, because the darn stator slot count is wrong and are 3 times a prime number, so it's either 3 phase or that prime number of controllers with one slot per phase and that probably would put the resistance too low anyway.

What about splitting the winding terminations into their separate strands, and then separate them into 2 electrically isolated groups to run 2 controllers? That doubles the winding resistance each controller sees, which I think is a good thing, but doesn't also cut the BEMF by half? My biggest worry is about the resulting parallel coils that are intertwined. Will those coils interfere with each other?

I have a motor with dual halls already installed, but the halls are on slightly different timing, one for torque and one for speed. Wouldn't that really make a mess with coils so parallel?

What do you think, potentially a good idea and worth the risk of messing up a perfectly good motor, or bad idea because the controllers will wig out?

If it has potential, I have another candidate, a dual tap x408-4011. If I separated the windings at the tap and ran them both at the same time with 2 controllers, one would be running 8 turn windings and the other 3 turn, which might be interesting. This is way different than the route discussed above, where the controllers run on identical count windings, just lower strand count copper for each.

John

 

[rhitee05]

Hmmmm, interesting question.

The two controllers would definitely not be independent of each other. The two sets of windings would be very strongly coupled to each other so there would be a lot of mutual inductance between the two sets. I don't see this being a problem if the two controllers are perfectly synchronized with each other. I think bad things would happen if they were not synchronized, though, or at least not-good things. Not 100% sure without carefully thinking it through. I don't think it would be controller-exploding bad things, but they might end up fighting each other a little bit. And it would be pretty hard to ensure that they were perfectly synchronized, even off the same set of Halls.

I would be more worried about the different turn-count version of this. The two sets of windings would act like a transformer, so unless you ran the two controllers off of different voltage ranges I think the lower-count one would explode the higher-count one from over-voltage. Or, at least, it would be a failed attempt to create a perpetual motion machine with one controller trying to regenerate with current drawn from the other.

 

[John in CR]

Thanks Eric,

Ok, I'll drop the case resulting in the transformer, and stick with dividing up the full length strands.

Perfect sync doesn't seem realistic to me, but if there was, would the effect of these closely coupled inductors be something similar to adding inductance to a phase with an external coil, just without the waste heat? If that's the case for perfect sync, then wouldn't out of sync be like having an intermittent bypass switch on the external inductor. With PWM already chopping up each pulse, I don't see some extra inductance being an issue as far as the motor is concerned. What about the controller, won't the extra inductance help slow round off the edges of the pulses when it's there and be like the other coil doesn't exist when it's not? I realize it's probably a gross over-simplification, but doesn't extra inductance in a phase wire "take the edge off" all of the square and triangular stuff a controller tries to do on its own, which kills controllers in the face of low resistance and inductance?

 

[Solcar]

That idea of using parallel windings reminds me of trying that on a switchmode power supply transformer. I seem to recall it behaving very poorly and producing an epic fail. I guess circuit differences between the different magnetically coupled paths led to instability. I didn't persist in continuing to try the idea and it could be that I omitted an important detail in the setup.

Which reminds me, I had read various places that each MOSFET in a parallel group, driving a (single) winding, should have a separate gate resistor. That never worked for me, but a single gate resister per group of parallel MOSFETs did though.

I like your idea of increasing the inductance to slow down transients. Just today, I was thinking about a power supply that I want to build that can test the controller I am building. It occurred to me that i ought to be able to drastically increase the size of the inductor in a power factor correction circuit and then lower the frequency into the hundreds of hertz range.

 

[jdb]

So you mean, wind 2 parallel strands of wire (as described in Thud's 80-100 rewinding guide), and hook up each one to a different controller, right? In this case, the common core of the motor still acts like a transformer between the two windings. Actually, it should make a more effective transformer than the case where whole windings are divided instead of individual strands, because the leakage inductance will be lower.

Controller 1:
Phase A high, Phase B low, Phase C off
Controller 2:
Phase A high, Phase B low, Phase C off

Great! FETs share load with motor resistance to divide power instead of just Rds(on)

Controller 1:
Phase A high, Phase B low, Phase C off
Controller 2:
Phase A low, Phase B low, Phase C off

Plasma! The impedance from controller 1 to controller 2 has 2x winding resistance, and <10% leakage inductance between them (assuming 90% of inductance is effectively coupled**). The motor acts like a transformer with a shorted secondary coil, and the short circuit path goes through the phase A and B low-side fets.

So you don't have to just get the line-frequency output exactly synchronized between the two controllers, you need the PWM frequency exactly synchronized, too. I would take a beer bet that paralleling two Infineon-like controllers this way will destroy them very quickly.

** 10% is probably generously high. You'd have to run an FEMM simulation to be sure, but I would not be surprised by low single-digit "parasitic" inductance on an 80-100 w/ Standard Thud Rewind.

 

[John in CR]

We can't have the plasma case though. The magnet position (same for both controllers) for any given slot means the high/low/off HAS to be the same, doesn't it? I'm only seeing a difference being in the PWM chopping points for a same direction pulse. I hope I'm wording this stuff correctly.

 

[Doctorbass]

The motor acts like a transformer with a shorted secondary coil, and the short circuit path goes through the phase A and B low-side fets.

So you don't have to just get the line-frequency output exactly synchronized between the two controllers, you need the PWM frequency exactly synchronized, too. I would take a beer bet that paralleling two Infineon-like controllers this way will destroy them very quickly.

** 10% is probably generously high. You'd have to run an FEMM simulation to be sure, but I would not be surprised by low single-digit "parasitic" inductance on an 80-100 w/ Standard Thud Rewind.

That's what i thought.

Anyway, John.... let's try it 8) if it work you'll be the first to confirm :wink:

Doc

 

[John in CR]

So you mean, wind 2 parallel strands of wire (as described in Thud's 80-100 rewinding guide), and hook up each one to a different controller, right?

Yes, though call me lazy, I want to avoid the rewind. I'll just separate the strands at the terminations, identify both ends of each strand, and separate the strands into 2 separate groups for each phase and terminate them separately.

 

[kenkad]

John,
Since there is a great deal of discussion of not only the motor but controller design on the Forum, try something even more complicated (or fun), a totally different motor design, AF PM BLDC, with six 3-phase sets of windings using six controllers implemented with the Apex 306A bridge IC (17A max at 60 V) and PIC18F25K22 micros (3 independent PWMs) on a PCB about 1.37x2.00 inches. Now you have six slave 18Fs and one 18F master sequencing the sinusoidal drive. Add a magnetic rotation sensor (off to the side with a timing belt) to get ultra acurate rotor position and you have the cat's meow of integrated motor/controller, including the ability to advance/retard the PWM sequence and a CAN interface to boot.
kenkad

 

[Lebowski]

I think you can make an interesting (ant true) analogy with an internal combustion engine.
You're trying to build a 1 cilinder 470 cubic inch engine.... a capacity which is normally spread over 8 cilinders.

 

[John in CR]

Kenkad,

If I understand your post, that's something I've wanted to do for quite a while. My best motor seems to be perfectly tailored for it, since it's the only one with a slot count that works at 63 slots. 3 separate controllers, giving me a 9 phase motor with 7 slots per phase should work for a great smooth running motor. The problem it doesn't solve though is the high current I need going across each coil, because I need to maintain the existing Kv and torque.

That's where splitting the existing windings would have the big payoff if it works. Not only do the controllers get to see an easier load due to the higher resistance, but they split the current, and it's the current that's killing those of us chasing higher power. Once you pass 100A (battery side), the controllers cost more than the motors, but higher cost doesn't stop them from blowing. OTOH, I can pick up a 40A controller good for up to 100V for $30. Split the windings 4 ways and I have a controller bank capable of pumping over 20hp into the motor for only $120, enabling plenty of spares. If it works, I can even go to the tiny 6 fet controllers, ventilate the motor and house the controllers inside the motor.

 

[John in CR]

I think you can make an interesting (ant true) analogy with an internal combustion engine.
You're trying to build a 1 cilinder 470 cubic inch engine.... a capacity which is normally spread over 8 cilinders.

Which part is the cylinder in your analogy? Big cylinders work just fine anyway. Have you seen those that someone can climb inside for servicing like on big cruise ship engines?

 

[jdb]

We can't have the plasma case though. The magnet position (same for both controllers) for any given slot means the high/low/off HAS to be the same, doesn't it? I'm only seeing a difference being in the PWM chopping points for a same direction pulse. I hope I'm wording this stuff correctly.

The magnet position and closely matched hall effect sensor timing will synchronize the line frequency between controllers, the commutation frequency. But it doesn't synchronize the high-frequency PWM pulses themselves.

Recall the case where someone made the division of windings by the whole tooth. This still makes a transformer, but its a weak transformer. The quality of the transformer depends on how well the flux from one winding couples with the other winding. The flux linkage from one tooth to another is fairly poor in a BLDC motor, which I think is part of why the inductance is so low. The reason is that the magnets also count as an air gap when you are going from tooth-to-tooth. So while the magnetic circuit is pretty good for the magnets, with < 1mm air gaps, the circuit is pretty poor for the windings themselves, with 2-4mm air gap depending on how thick the magnets are.

But when you co-wind the two windings on the same teeth, then the flux linkage between them should be pretty good. When you separate the windings on individual teeth, the linkage is poor. The total inductance of the motor is the same in both cases, but the division into mutual and "parasitic" inductance is different. Higher mutual inductance means that voltage changes on one winding are reflected very well into the other winding - it makes a strong transformer. Higher parasitic inductance will limit the fault current flow between them. When the teeth are co-wound, the mutual part is maximized.

In the separate-teeth case, you still generate the short-circuited-secondary events at the PWM frequency. But because the flux linkage is so poor (and therefore, the mutual inductance is so low), the parasitic inductance is still very high and keeps the dI/dt low. Maybe it increases your switching losses and makes the control somewhat screwy. But you only kill the controller when the commutation frequency gets out-of-sync.

I think that the co-wound tooth case makes the flux linkage so much better (and therefore, mutual inductance is high, and parasitic inductance is low) that shoot-through becomes a problem at the PWM frequency.

 

[rhitee05]

I put a little more thought into exactly what might happen with this. I'll assume that both sets of windings are very closely matched, and since they are very tightly coupled the mutual inductance will be only slightly smaller than the self-inductance.

I don't actually think it would be so bad anymore if the timing was not synchronized, or if they were operating off two different sets of Halls. Let's say one controller is currently running A+ C- and the other is at A+ B-. This is the only reasonable scenario, with the two controllers in adjacent commutation periods with one phase in common and one different. In this case the mutual inductance will cause voltage to be coupled onto the non-operating phase of each controller (for example, from phase C of the first controller onto C of the second). Since the mutual inductance is a little smaller than the self-inductance, the induced voltage will not be high enough to force current to flow through the diodes. So there doesn't seem to be any big down-side to this scenario. Possibly some slightly higher losses but no explosions. Note that because the FETs stay off there will be no regeneration effect that would cause current to flow.

However, I think things will get a bit crazy when both controllers are on the same pair of phases. If the PWM is not synchronized, I think that will cause the current ripple to increase. The mutual inductance will alternately either augment or partially cancel the self-inductance of the phase. That's not catastrophic, but would be stressful on the controller. Even if the PWM is perfectly synchronized, I think there would be an issue with current sharing between the two controllers. I'm curious to try a SPICE simulation to see if it agrees with my intuition, but here's what I think would happen. No matter how hard you try, there will be some imbalance between the two sets. That means the current will be rising a little faster in one than the other. That higher di/dt will induce a slightly larger voltage through the mutual inductance onto the other winding, which will in turn cause it's di/dt to decrease. That decrease will reduce the coupled voltage onto the first winding, which will allow its di/dt to increase... It appears to me that this will set up a positive feedback that will result in one set of windings carrying more current than the other. This is the common-mode choke effect. I'm not entirely certain how that would interact with slightly de-synchronized PWM. I still don't think that's necessarily catastrophic, since each set of windings will have 2x the resistance and inductance of the standard wind.

I'm not 100% certain that my intuition is right about exactly what would happen. But even if I'm not entirely right it seems to me that this probably would not work in the way you want it to. Which is unfortunate, because it is an innovative idea! Using alternate sets of slots for multiple controllers seems like a safer bet.

 

[jdb]

I put a little more thought into exactly what might happen with this. I'll assume that both sets of windings are very closely matched, and since they are very tightly coupled the mutual inductance will be only slightly smaller than the self-inductance.

I honestly think that individual teeth are poorly coupled to each other, and the permanent magnets relative permeability of ~1 makes them act as an air gap with respect to the motor windings. That's why thicker magnets, and more teeth make for a lower inductance motor.

I don't actually think it would be so bad anymore if the timing was not synchronized, or if they were operating off two different sets of Halls. Let's say one controller is currently running A+ C- and the other is at A+ B-. This is the only reasonable scenario, with the two controllers in adjacent commutation periods with one phase in common and one different. In this case the mutual inductance will cause voltage to be coupled onto the non-operating phase of each controller (for example, from phase C of the first controller onto C of the second). Since the mutual inductance is a little smaller than the self-inductance, the induced voltage will not be high enough to force current to flow through the diodes. So there doesn't seem to be any big down-side to this scenario. Possibly some slightly higher losses but no explosions. Note that because the FETs stay off there will be no regeneration effect that would cause current to flow.

That actually makes pretty good sense to me, but it isn't what John is proposing. He wants to divide the windings from a single tooth into separate controllers. In this case, the timing must be identical and each controller will have the same commutation state.

 

[rhitee05]

That actually makes pretty good sense to me, but it isn't what John is proposing. He wants to divide the windings from a single tooth into separate controllers.

That's the case I was describing. There will be relatively little coupling between windings on adjacent teeth.

In this case, the timing must be identical and each controller will have the same commutation state.

That's not necessarily true. John mentioned a possibility using two sets of Halls at different timing angles, so I was considering what would happen. Not surprisingly, it doesn't seem advisable.