Is it viable to connect multiple ESC units in series?

Question

I am designing a system that requires a large number of low-power motors in a weight and space constrained application.

We have a high voltage power source of 400 V DC, that needs to run these motors at a total power output of ~20 kW, or 200 W each (96 motors). Using RC aircraft brushless motors and ESC units is attractive, but they are all in the 8-30 V range, so connecting them in parallel is not possible without a step-down DC/DC converter (inefficient for weight power and cost), or custom ESC units.

However, if I were to connect them in series, the voltage drop across each would be right in the correct range, and the total current would likewise be relatively low. The biggest advantage is that we get to use standard components.

Potential issues i'm already aware of: resonance between the ESC's. In this application single failure bringing down the whole system is beneficial, as partial failure is more dangerous than total failure.

Are there obvious or non-obvious problems with this architecture?

Answer 1

As someone who spends all my time at work connecting things in series: Don't!

Moreover, don't connect anything but resistors in series unless you absolutely have to.

What will happen is that one ESC will draw ever so slightly more current than the other despite your best effort to keep them equal. Said ESC will subsequently receive less voltage over it due to the series connection. To compensate, the ESC will try to draw more current, and this creates a positive feedback loop with it hitting undervoltage lockout. This is where the problem starts and is true for just about any regulated power supply. When it does, it stops drawing current but the other ones don't, pushing current though the one restarting and reversing the voltage across it (a bit depending on how much inductance you have in your circuit, but if the inductive spike won't kill it immediately, the new DC voltage across it will anyway), hopefully shorting it out. Once shorted, all the others will now have n/(n-1) higher voltage than before. This will continue until all have failed in a spectacular manner. If it does not short, you will have 400 V DC across it and possibly a flashover. Both are very unfavorable.

You can have bypass diodes and voltage clamps across each and try your damdest to make them equalize, but this is a loosing battle, uphill and in deep snow.

Please don't. Find another solution.

Answer 2

I can see 2 problems:

• If all motors don't draw the same current at the same time, the voltage will not be spread equally, and you will end up with too much voltage on some units and to little on others. I see no simple solution to avoid this at first glace (excepted very inefficient ones, that would defeat the whole purpose)
• How do you send commands to those ESC's, as they will not share the same ground than your controller? This one can easily be solved with optocouplers or digital isolators (it adds a bit of cost, quite some PCB space, not a lot of weight excepted the one induced by bigger PCBs).

Answer 3

It wouldn’t be possible by just paralleling the ESCs. The DC source would need an inverter on the output to generate AC somewhere between 1kHz and 50kHz. Then each ESC would be tapping power off an autotransformer. The transformer could be fairly small. Using the highest possible operating voltage for the ESCs and motors would help a lot with keeping the windings physically small. This would require a fair bit of optimization to make it competitive with the more usual solution of front-end converters that generate a couple of low voltage DC buses.

Answer 4

It would probably be possible to design a "balancing" switch-mode power supply that allow a 360-440VDC power supply to be used to power e.g. eight 40-60V devices in series, using the inductors only to handle differential currents among the devices. If the loads are reasonably balanced, and the devices didn't care about how well their inputs were regulated, this might in theory allow a higher level of efficiency than would be possible with a 400VDC to 50VDC buck converter.

If e.g. seven devices each needed 10 amps and one only needed nine, the total load would be 3,950 watts, of which the coil of a buck mode switcher would need to handle about 3,500 watts. The coils in a balancing switcher would only need to handle about 50 watts, drawing it from the bottom load connection and sending it back to the supply so it could assist in powering the other loads. Since the act of storing energy into an inductor and later extracting it will necessarily have some losses, reducing the amount of energy the inductor has to handle could in theory improve efficiency.

In practice, making such a thing work would require passing current through so many switching devices that the power losses there would probably outweigh the power losses in a simple-but-decent buck mode switcher. If one wants to have many motor controllers run of 50 volts, but one has a 400 volt supply, using a buck converter to convert the 400 volts one has into the 50 volts one wants is apt to be simpler and more effective than trying to jump through hoops to avoid such conversion.