Power Distribution in a in a multi-domain weird setup

Hey all,
I’m making lots of progress on my robotic control boards, but I’ve run in to an interesting problem, and I’m wondering if there are best practices or approaches that I should consider using.

In my robot, I have 8 independent and identical motor/sensor control modules that communicate over CAN-FD. These modules each have their own complete power management systems, including separate batteries and a battery management system with powerpath set up for the currents required to drive everything. I want to run power wires alongside the CAN-FD path, so that I can have a single high-power charging port. I’ve designed the hardware to allow the boards to communicate over the CAN-FD bus, decide which boards can draw how many amps, and then they’ll begin charging. The power propagates through terminal blocks at a max of 40V 3A (BMS limit), and the CAN-FD propagates through RJ10 connectors with its own independent 5V power for the isolated side of the transceivers.

The challenge is this: the boards are almost fully isolated from one another right now, thanks to the galvanically isolated CAN-FD transceivers. I want to isolate the long power distribution rails from the boards, to reduce EMI due to the long wire connected to all the grounds, and to protect from destructive transients.

I’ve considered a couple topologies for this power distribution tier:

  • Directly connect the high power to the PCB and include 40V TVS diodes and chokes at each board’s input. (allows for charging when fully discharged and allows for charge sharing via OTG reverse mode from the battery in emergencies)
  • Use a DPST relay with flyback diode protection to physically isolate the power bus (requires some battery power to start charging; OTG reverse mode wastes even more power; physically large)
  • Use 24VAC on the power rails and a transformer for isolation (physically large; cannot use OTG reverse mode power sharing; not sure whether simply rectified AC is compatible with the charging IC’s input)
  • I don’t think I can use MOSFETs instead of a relay–wouldn’t those suffer the same ESD/transient risks, being a CMOS device?

Are there other approaches or pros/cons that I should consider?

This area is of considerable interest to me. I definitely have more to learn here.

The easy, but not so cheap, answer is to use isolated DC-DC converters. I have done this on several projects for CNC machine control. I assume the EMI threats are similar.

Do the motors and drivers themselves need isolated power? In my situation, I don’t worry about that a lot and focus on designing the CNC boards to isolate the digital portions. I am mostly concerned with the controller going crazy (er, rebooting or halting) in the middle of a long CNC job. Nothing pisses off my customers more than a ruined 10 hour job. Your software might need to be aware of potential EMI hiccups in the motor driver section(s).

The motors are actually COTS “smart” servos, which I know little about internally other than they do their own power regulation, want 6-14V, and have a cortex m0 inside. The power used by each of the 2 servos is 1.2A peak each, and then there’s an LED display that uses 2A typical/8A peak. The digital electronics have an independent switching power supply, so I think those parts are isolated enough.

The big challenge for me is this bus that’s only used during charging, since it needs to be able to deliver energy to each of the 8 modules, and the 8 modules need to be protected from one another during normal operation.

I think that the EMI threats are similar to CNC, but a little different: besides servo motors, the LEDs run with a high PWM frequency and can draw a lot of power. Also, the boards are inside of a robotic skeleton that’s covered in faux fur (hello ESD, please destroy these expensive parts) that will be in regular contact with humans & pets (i.e. the faux fur is meant to be touched & petted, which will make the ESD issues even more fun). And, there are capacitive antennas just underneath the fur, to ensure that there’s a clear 1cm gap followed by an ESD path directly into the analog domain.

Is your suggestion to put an isolated DC-DC convertor between the battery charging/power management and the shared power delivery bus? Are there highly integrated convertors? I’m running out of physical space budget.

I would hesitate to say it is my suggestion - just what I am doing.

Traco has a quite a few options: https://www.mouser.com/TRACO-Power/Power/DC-DC-Converters/Isolated-DC-DC-Converters/_/N-brwkv?P=1ytymwz. Not terribly small, though.

I see the difference in your situation but CNC has a couple of significant EMI sources (VFDs and 230VAC 24K RPM spindles, for example) and lots of wire channels back to the controller. Probably the biggest difference is the ESD potential of the fur covering.

I read somewhere about carpeting that was anti static - had carbon fiber to dissipate the charges. You might be able to built a faraday cage underneath the covering.

Can you elaborate on why the daisy chained design is bad from the get-go?

It’s quite normal til have parallel power converters and a distributed bus also. I do not see which case you would have an EMC problem. For sure you need to protect them from surges etc, but that is standard on power inputs.

A structure to divert surges to is good, otherwise if you isolate them, the barrier might be in the motor instead, and depending on the motor technology, that could cause wear in bearings if you have a shaft connected to chassis

I think daisy chaining power is fine, but the thing I’m thinking about is that the simple daisy chained power system design results in long wires between modules (including boards, servos, LED displays, and sensor) that connect all of the grounds together. I worry about those wires acting as antennas or catching & propagating ESD, and I worry about designing the chassis and system ground. Without the power wires, each module is completed isolated, which is simpler to reason about.

I haven’t totally grasped your situation, to be honest, but I want to respond to the “long wires” comment. The long wires are only an issue when:

  • The load draws “spikey” currents, resulting in the wire acting as an antenna. If you have enough decoupling at the remote end the wire carries DC and should not act as an antenna. As I understand each of your “loads” has it’s own power management so this should not be an issue if you have capacitance on the remote end. DC-DC converters will not help with this as far as I understand.
  • The steady state DC current will indeed create a “ground bounce” situation, but again your control signals are all isolated so this should not be an issue. The small voltage drop due to IR drop in the long wires should quite small and so should not be a problem.

Like @kabhijit wrote, there should be no problem with long wires. You have a small EMI filter for each station and that includes surge protection

@kabhijit @kvk
The load drawn from the main power bus will be consistent, as it’ll be for battery charging. This long power bus will actually be unpowered except when charging; otherwise, each of the chargers has its own local battery power system. Could there be issues with the long wires when they’re unpowered? And could there be an issue that the long wires create a common ground between all the boards?

Thanks for the tip on DC/DC converters! If I absolutely must, I’ll include them, but I’m starting to wonder about cost, since they would be the 4th switcher on each board.

In general, the DC-DC converters would be in place of (some of) your other voltage regulators. There are 2, 3 and 4 output models. That implies a redesign that you probably aren’t willing to do.

[Edit] I think it is great that you are spending time looking at these issues. Too many makers simply ignore the EMI issue and then wonder why their creations have weird problems. EMI is notoriously hard to track down and fix afterwards.

I am open to redesigning with an isolated DC-DC converter, however, I don’t think I can replace the current set of regulators: the only thing connected to the main power bus is a highly integrated power management IC, which muxes the power bus & the batteries to feed the system. This integrated IC takes 3-40V input, connects to a 3S lipo battery, and outputs 9-14V. It uses a single inductor as a buck-boost on the power input and the same inductor for battery charging. I think I’d need the isolated DC-DC converter upstream of this module, since that’s where the power bus connection lies. Downstream, I have a 3.3v and 5v regulators for the logic. One of the parts I really don’t want to redesign is this charger/mux IC, since integrating 3S lipo charging is difficult, and it’s nice to be able to power the system without a battery inserted.

Maybe the unpowered wires and shared ground could work without introducing EMI/ESD danger, if I put a choke on the power input of each board.

I do not see a problem with unpowered wires, nor with the common ground

1 Like


  • Long wires when unpowered: no issued except they may pick up RF energy at that frequency if they are disconnected at the remote end. Since they are power lines this should not be an issue, but you should give it some thought - depending on what else is connected to those lines.

  • Another thing with the long wires is that they may act like transmission lines, so make sure your dV/dt is low enough that you don’t get unintended reflections - especially important if you add L as a choke. And you will have parasitic C, so you might get some ringing as well. Isn’t this the gift that just keeps on giving? :smiley:

*Common Ground: Like I mentioned before the wires will indeed give you a galvanically connected ground at DC.