I've considered using three relays to implement a wye-delta transformer, so I can turn the motor to wye or delta to change its Kv at any time, high Kv for high power and low Kv for high efficiency. But I need the transformer not to bring in much resistance to the system, while most relays have a contact resistance of >50mΩ. 50mΩ relays will bring in 100mΩ resistance at the wye state, creating more energy loss compared to the energy loss of diodes in the ESC when using a high Kv motor and low PWM duty cycle.

Does low resistance relay resistance with contact resistance less than 10mΩ ever existed? Maybe relay is not a good component to achieve that, what is the common solution to build a low resistance wye-delta transformer? -----update-----

I want to use wye-delta transformation to increase efficiency at low throttle. To get same power output, high duty cycle with low Kv has higher efficiency than low duty cycle with high Kv. A test in an anwser of this question has showed that.

High Kv motors have less inductance, creating more ripple current in the coils, causing higher crest factor thus more heat generation. Higher PWM frequency is required to decrease the coil current crest factor, putting more streess on the ESC.

At low duty cycle, freewheeling current will take a larger partion of total current in coils. More freewhelling current means more current pass through diodes with higher voltage drop.


An answer said that this can be done with MOSFET, I come up with an image of using 12 MOSFETs, or 6 pairs to build 6 relays. 3 connect the coil ends to a COM point, and 3 connect each coil end to the next coil head.

enter image description here

Each relay has its pair of MOS with each drain connected to the other one's source. enter image description here

In this image, there will be 7 individual points with sources connected, each having a different changing voltage to GND. So we need 7 some kind of isolated drivers to apply voltages between gates and sources. Is this image a proper solution?

  • 2
    \$\begingroup\$ Can you explain why you would need a wye delta transformer for a BLDC motor. \$\endgroup\$
    – Andy aka
    Dec 19, 2022 at 10:31
  • \$\begingroup\$ @Andyaka What I think he is saying is he wants the KV / RPM per volt to be variable and Wye will have lower KV than delta. WHY he wants switchable KV is TBD :-) \$\endgroup\$
    – Russell McMahon
    Dec 19, 2022 at 10:51
  • \$\begingroup\$ Search for adjustable field for ideas for variable "k"V motors. \$\endgroup\$
    – greybeard
    Dec 19, 2022 at 13:31
  • \$\begingroup\$ (I guess IPM SynRMs don't come that small.) \$\endgroup\$
    – greybeard
    Dec 19, 2022 at 14:37
  • 1
    \$\begingroup\$ Interesting relay in rev. 8. With body connected to source, there won't even be any voltage breakdown. For something that works, the MOSFETs need to be anti-serial. When you detangle a schematic close to a \$Y\$ or \$\Delta\$, you end up with the three full bridges of mkeith's answer. You can, for the exercise, try the same with a Dahlander pole-changing motor. \$\endgroup\$
    – greybeard
    Dec 22, 2022 at 5:47

2 Answers 2


It could be done with MOSFETs. But each motor winding needs to have both ends connected to the switch, so it needs twice as many wires and connectors as normal. The switch needs a lot of MOSFETs in it to get the required bi-directional connections at low resistance. This will increase the weight of the system, reduce reliability, cost more and be more difficult to maintain than a standard setup.

Now consider what advantage the extra efficiency has in a drone, compared to the extra weight. Let's say you manage to improve efficiency by 10%. It would be much simpler to just add 10% capacity to the battery, and the reduced performance from the increased battery weight would be barely noticeable.

Considering all the other areas that should be optimized for best system efficiency (airframe, propeller, motor, esc, flight controller, battery etc.) this is probably the last thing you should be looking at - if at all.

  • \$\begingroup\$ If we use MOSFETS, this is the image in my mind. There should be 12 MOSFETs of 6 pairs. Each pair makes up a "switch". 3 switches connect the coil ends to a COM point, and 3 connect each coil end to the next coil head. In this image, there will be 7 individual points with sources connected, each having a different changing voltage to GND. So we need 7 some kind of isolated drivers to apply voltages between gates and sources. Is this image a proper solution? \$\endgroup\$
    – lotsof one
    Dec 19, 2022 at 13:46
  • \$\begingroup\$ As for what advantage the extra efficiency has in a drone, yes you are right, this is the last thing to look at. I didn't know that. I just found active free-wheeling is complicated and thought, is there a simple solution that can change a motor from delta to wye? And now I got there isn't, I should look at other areas first. \$\endgroup\$
    – lotsof one
    Dec 19, 2022 at 13:46
  • \$\begingroup\$ What is a push-pull drive \$\endgroup\$
    – lotsof one
    Dec 24, 2022 at 10:55
  • \$\begingroup\$ You have the circuit right (except FETs should be in series, not parallel). I wouldn't call it a 'proper' solution though, rather a complex solution for little benefit and several potential issues. I wouldn't do it. Active freewheeling is not complicated at all - just buy an ESC which has it, simple! \$\endgroup\$ Dec 25, 2022 at 7:18

The first thing I want to say is that I see no advantage to converting the motor from delta to wye on the fly. I don't think there will be any important operational advantage to doing this.

The second thing I want to say is that motor losses, not controller losses tend to dominate except when the motor is under utilized (very low torque output).

If someone tasked me with developing a delta wye motor controller, this is what I would propose.


simulate this circuit – Schematic created using CircuitLab

How does this accomplish the desired result?

  • First, delta. In order to drive the motor in delta, M3 and M5 would be driven identically, and M4 and M7 would also be driven identically. In this fashion, A2 and B1 would be always connected together. Similarly, M6 and M9 would be driven together, and M8 and M11 driven together. In this fashion, B2 and C1 would be always connected. Finally, M10 and M1 would be driven together as well as M2 and M12. This would connect C2 to A1. With A2 connected to B1, B2 connected to C1 and C2 connected to A1, the motor would be driven in delta configuration. This is pretty much how all ESCs normally operate already (except with extra transistors).
  • Now, wye. In order to drive in wye, M3, M6 and M10 would always be driven together. Likewise, M4, M8 and M12 would always be driven together. In this fashion, A2, B2 and C2 are always connected together. This is the wye configuration. If the duty cycles of bridges A2, B2 and C2 are kept at 50 percent, the effect is the same as grounding the common node of the wye configuration to the mid rail.

    So this configuration allows both delta and wye drive. However, I don't see any advantage to using the wye configuration. So I would just use the standard ESC configuration with six transistors, and a conventional delta wound bldc motor. Switching the motor to wye will not bring any benefits that I can see.
  • \$\begingroup\$ I think so. It's better to just to use active free-wheeling. Your solution need total 24 MOSFETs to allow free-wheeling. Both of them are better than actual wye-delta transformation. \$\endgroup\$
    – lotsof one
    Dec 24, 2022 at 11:12
  • \$\begingroup\$ In order to achieve the equivalent of a Y (or Delta) connection you need to be able to power any 2 windings in series. This scheme doesn't do it. To get the required speed reduction with this scheme you still need to apply PWM, with the losses that implies. To fix it you could try putting a bidirectional switch (back to back FETs) between each H bridge (3 switches total). That's half as many extra FETs as the 'conventional' method of switching from Delta to Y, but requires a custom commutation scheme. \$\endgroup\$ Dec 24, 2022 at 18:36
  • \$\begingroup\$ I added a lot more detail, @BruceAbbott. \$\endgroup\$
    – user57037
    Dec 25, 2022 at 0:24
  • \$\begingroup\$ There is nothing about this configuration that prevents active free-wheeling. If space vector modulation (SVM) is used to drive the motor, then active free-wheeling becomes the default and is always enabled. I don't see why you wouldn't use space vector modulation. Unless maybe you are super cost constrained. \$\endgroup\$
    – user57037
    Dec 25, 2022 at 0:31
  • \$\begingroup\$ @BruceAbbott how would you implement speed control of a BLDC without PWM? Every drone ESC uses PWM (as far as I know). \$\endgroup\$
    – user57037
    Dec 25, 2022 at 1:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.