3 phase without 6 switches

JackB

10 W
Joined
Apr 28, 2013
Messages
86
Location
Sacramento, CA
Ok, gotta ask this question..I'm building a wheel/hub bldc motor and was thinking can I make a controller that simple cycles through three coils? So instead of delta or star, it is star with a permanent ground at the star. Thus, turn on power to A, then B, then C, then back to A. Now I understand this would power only 1 coil vs 2 coils (or 3 coils like Prius inverter), however, the resistance is half as much so it should make more current/force in the single coil. The big thing is it only needs 3 high side switches vs 6 to switch high and ground for each coil. OK, so tell me why this is dumb...

And while I've got your attention, I'm also thinking of another dumb idea of segmenting my wheel coils into 2 or 3 or 4 groups and have a seperate driver for each one. Then I can use lower power switches rather than 4x more current in a single switch.
Of course I know it would be better to parallel 4 together in one controller, but I have 4 low power controllers so we are not talking about an OEM design, but repurposing existing parts.

Thanks for any insight into my dumb ideas. ;)

Jack
 
This is doable. Years ago, I got an old hard drive motor spinning with this technique. However, its downsides far outweigh the advantages of being able to have three less switches.

Most importantly, since you only have half the copper in the motor active at any given time, you can only get half the rated power from the motor. Also, you will need an additional phase wire to carry current into the center point of the motor's wye winding (3-switch commutation is impossible with delta). That means for the same cross-sectional cable area, you will get less than half the power compared to traditional three-phase. The return wire will be carrying current all the time and will get hotter than the other three wires as well.

On the controller side, you still have to deal with the motor's inductance. That means diodes to catch the free-wheeling current when the FET turns off. Even with schottky diodes, the voltage drop will be several times what a decent FET would drop. You would have to either use several diodes in parallel (waste of space and power)...
Or you could implement synchronous rectification...
But that would take 3 more FETs...
Resulting in a circuit identical to a standard 6-FET controller. If your circuit is like that, you're better off driving the motor like a standard controller (which your circuit is by this point).
 
Regarding parallel FETs / controllers, it's a lot cheaper to parallel FETs than build several controllers. You only have to have 1 MCU, power supply, and gate drive circuit as opposed to several. The only thing you need is to have each FET have its own gate resistor.

Plus, using separate controllers can cause nasty things to happen (blown FETs) if/when they get out of sync.
 
JackB said:
Ok, gotta ask this question..I'm building a wheel/hub bldc motor and was thinking can I make a controller that simple cycles through three coils? So instead of delta or star, it is star with a permanent ground at the star. Thus, turn on power to A, then B, then C, then back to A. Now I understand this would power only 1 coil vs 2 coils (or 3 coils like Prius inverter), however, the resistance is half as much so it should make more current/force in the single coil. The big thing is it only needs 3 high side switches vs 6 to switch high and ground for each coil. OK, so tell me why this is dumb...
It's not. That's how some switched reluctance motors work. For standard motors, though, you are getting only half the power out of them, since they are carrying current for less time (and in one direction.) And of course you have to worry about saturation a lot more, and things like resetting the windings when they are powered off. But if all that's OK, no reason it wouldn't work. (BTW you will probably need 4 switches rather than 3 but you still save a few FETs.)

For an example of an SR inverter check out figure 5 in this app note:

https://www.ti.com/lit/an/slvaen9/slvaen9.pdf?ts=1613364543147&ref_url=https%253A%252F%252Fwww.google.com%252F

But I would point out that FETs have gotten _really_ cheap (about $2 for an IRFB4110.)
 
Thanks, yeah that is pretty much my idea with the "Miller inverter".

I get redundancy too with multiple drivers for a wheel subset, so the truck can keep going if damaged from shrapnel or debris. Also less chance of over heating with the coils are off 2/3 of the time vs ON 2/3.
I am going to build a 50-200hp motor, so the switches are not real cheap..

"This unique yet simple construction and the lack of magnets give SRMs a significant cost benefit as
compared to rare-earth based permanent magnet motors. Furthermore, the lack of magnets allows for a
lighter and more robust rotor which makes them well suited for use in hostile corrosive environments or
extreme temperatures. SRMs also have the benefit of being immune from single point failures in their
winding allowing limp home or degraded operation in automotive applications. Advances in control
methods have also enabled SRMs to operate across a wider range of speeds (0 rpm to > 10,000 rpm in
some designs) with high starting and accelerating torque. These properties have started to make adoption
of SRMs very attractive in automotive pumps, transmission actuators, and traction motors. SRM motors;
however, are more sensitive to mechanical construction and small variations in the air gap between the
stator and the rotor can result in poor vibration and acoustic performance"
 
Also less chance of over heating with the coils are off 2/3 of the time vs ON 2/3.
Heating is proportional to I^2r.
To get the same torque when running with 1 coil at a time instead of 2, you'd have to double the current. That would result in 4x the heating for 1/2 the time, so you'd be running 2x hotter.

However, if you are going with a switched reluctance motor, I don't see any problems with using only low-side switches and flyback diodes. Especially if you are operating at hundreds of volts, the diode losses won't be that much.

I am going to build a 50-200hp motor, so the switches are not real cheap..
Controllers at these power levels are not cheap and the last thing you want to do is blow FETs during development. At these voltages and currents, inductive turn on, the Miller effect, and ringing will be quite problematic and deadly to FETs. Unless you have experience designing a controller of at least 10 kw peak, I'd start with something a lot smaller. Designing a 10kw powerstage for a vEsc or Lebowski (or your own) controller is a great place to start.

Also you'll need a good oscilloscope. Something with at least 100 MHz bandwidth. I've seen places where a cheaper scope showed that everything was fine, but a more expensive scope showed terrible ringing at tens of megahertz. This was with GaN FETs and switching times of 10s of ns (very fast) though.
 
I'll chime in and point out the possibly not obvious:

The extra 3 FETs in a six FET controller do half the power handling. By losing them, rather than saving yourself 50% of the cost, you're moving the cost into more expensive remaining 3fets.

Gate drivers cost sod all. A 4 amp high and low side gate drive can be had for like 1$. E.g. ncp5183, ncv5183, even the ones thor is using are only 1.5$ and they're fully isolated.

Fools errand making a big controller with 3FETs.

SRM is another issue entirely... That's a sensible endeavor. But you'll still end up with 6FETs, or waste your life trying otherwise.
 
I have one of these washer motor stators picked up at the junkyard to prototype with. I know all about blowing up fets..
s-l1600.jpg
 
well well it seems switched reluctance motors are replacing inductive motors now since they can be more efficient and controllers are now cheap, no need to run on 60hz main power only.
anyway, found another example of a controller, it looks like the diodes are hooked up differently?
 

Attachments

  • singleswitchctrl.png
    singleswitchctrl.png
    121 KB · Views: 358
I see what that circuit does.
At all but the lowest speeds, the inductance in the coils makes it such that you need to reverse the voltage across them so that the current turns off. To do that, you need to make the switched side of the coil higher than Vdc. The circuit shown uses a boost converter arrangement with the motor and FETs to generate this high voltage. To prevent the voltage from rising too high, Tr, Lr, and Dr act as a buck converter to recycle the energy back to the main supply.

Looks like a pretty good design, but the transistors must be rated for more than Vdc (Twice?). An alternative circuit only requires transistors rated for Vdc (https://www.microchip.com/images/de...d-drive/switched_reluctance_block_640x480.jpg) but requires more components.

Personally, I think the design without the buck-boost in it would be easier to control but I have zero experience with switched reluctance motors.
 
hmm, seems to me would just be simpler put the diodes to go back to the positive rail just like a dc motor controller does,
the coil collapse would raise its voltage enough to flow by itself,
and the freewheel current then routes into the next coil, but I'm sure I'm confused.
So maybe better using high-side switches, which already have the built-in diodes to route back to the positive rail,
instead of the low-side switches.

m0kX2.png
 
So more thoughts, maybe this thread is just my personal notes..
The switched reluctance motor (SRM) with only 3 switches does not reverse the polarity of the coils, and thus it doesn't change the magnet polarity, which in a normal bldc it will push away the magnet as well as pull in one magnet. This really makes uses magnets not work as the moving magnet will want to stick to the steel in the coil as it doesn't get pushed away. Now one wonders if you can have what I think is a 'coreless' motor without steel in the coils, and then can use magnets, but not needing magnets is a big plus to the design.

And next I was thinking of putting the switches on the high side to allow the coil energy to reverse direction and transfer into the next coil through the diode, but after thinking about, if the high side is joined together, the coil energy will xfer over the joined connection and not need to go through a diode, so this would seem an even better way. And finally, the SRM done this way does not need any 'dead time', in fact it wants to have overlap, so coil A low side is on, then turn on coil B low side and then turn of A, so there is a transition of current from A to B. Seems to me this should make a more robust controller less prone to crashing and burning.

Finally, I got some super cheap 3-phase bridges off ebay, looks like they have plenty of them, anyone else try using them?
https://www.ebay.com/itm/FUJI-ELECT...e=STRK:MEBIDX:IT&_trksid=p2057872.m2749.l2649
Very little info available, no real datasheet. With 10 at basically $5 each, I have no concern to destroy a few.

I am waiting on a 3d printer to make the wheel for the washer motor coils, it is stuck in Texas, along with the igbt driver chips from digikey, so continuing to work on the other 53 projects in progress..
 
The coils in a motor have significant inductance.
If you put the diodes like that, the transistors would be able to get a current flowing in the coils, but the current would take a long time to stop flowing with no way to speed it up. Trying to spin a motor wired like this would be a disaster. All of the energy you pumped into the coil to energize it would end up as heat in either the diode or coil. Not to mention the reverse torque caused by the magnet still being energized after you stopped powering it.

With 6-switch controllers, the leftover energy in the coil flows back onto the power rail and into the next phase when the controller switches. The leftover energy keeps getting recycled until it has been used up producing useful (mechanical) power.

With the good deal you got on IGBT's, it never hurts to experiment, but with your current schematic you WILL either blow components or have horrible efficiency because the leftover energy HAS to go somewhere.
Also see Mxlemming's previous post about power handling. With half the FETs handling twice the current, you'll have 4x the losses per FET. To get the losses back down, you will need.... 6 FETs.

If three-FET controllers were a good idea, all the low-power ebikes would be using them.
 
Back
Top