Since you already have a variable power supply, an option would be to control its voltage to increase it when the robot is drawing a lot of current. So how is it controlled and is it able to react quickly? Due to the length and inductance of the cable this risks overshooting voltage on the robot side when current goes down quickly, so you will still need a voltage clamp, but at least it won't be dissipating power constantly. Is this possible?

Does the power supply have separate sense connections, and are there any spare wires in the cable to the robot? Perhaps the power supply sense connections could be at the robot end of the cable, to allow the power supply to adjust it's output to accommodate for the voltage drop down the cable.

nother option would be to change your motor controllers for ones that allow a higher input voltage. But you need to consider the insulation rating on motors, wires, cables, etc. Have you tried calling the motor controller manufacturer? Maybe you can convince them to make a higher voltage version in small quantity.

@bobflux : for now, the power supply is controlled manually (by buttons). It can also be controlled through ethernet, which we are not using yet. I haven't found information about reaction time yet (but it might be rather slow, for some read commands it's up to 10ms). But it might be an interesting option to combine controlling the supply real time with a voltage clamp on the robot side to protect it when current decreases quickly.

@ChesterGillon : in the main cable, we do have a few unused signal wires (there is a rather expensive slip ring which we would need to change to wire them through), but it should be doable. The PSU has remote sensing input, but can only compensate for +-5% of its voltage range (ie 0.05*750 = 37V), so it will not be enough

@bobflux : changing the controllers of the motors seems not very likely (it's already hard enough to get the standard version)

@JonathanS. : I agree that at 360V at the supply, there is no way I can get enough power on the robot. So I will need to increase it. Increasing it a lot (to 700-750V) and using a buck converter is a solution. Increasing it less (like 450V) and clamping it is "good enough" : I get pull power of my motors when I use only a few of them, and still decent power when I use all of them (which is not very frequent)

Out of interest, what is the highest DC voltage the cable is rated for?

1000V for the power wires (and just 300V for the signal wires, so to little for direct use for remote sensing, but it could be tricked buy using amplifiers at both ends

Clamping continuously is never going to work. If you want 350V 30A over 8 ohms cable you need 350+30*8 = 590V DC at the power supply. And when the motors are off you will get 590V at the robot. So you'd need a voltage clamp... but it will have to waste 30 Amps (ie, dissipate 10kW) to keep the voltage at 350V all the time. Not possible. Clamping can only be done on short transients which don't involve that much energy. You need to control voltage, either source side or robot side. Easier on robot side if it fits.

That's why I was asking about reaction time of the power supply -- if your clamp needs to dissipate 10kW for 1ms, it's okay. But it sounds like the power supply will be a lot slower than that.

Usually, a buck converter is never used in front of a motor controller, because it is much simpler to use a motor controller with higher voltage parts instead, and adjust the PWM according to supply voltage to get desired motor power. Adding a buck effectively duplicates all the power semiconductors (and losses) so it should only be considered if you're absolutely sure you can't get higher voltage motor controller.

What's the standard AC three phase supply voltage on a boat, btw?

@bobflux I was thinking about clamping the voltage like in a linear voltage regulator (ie adding a variable resistor in series). If I say that I want 320V @ 30A on the robot, but that 340V are OK at lower currents, then the maximum power at the robot is 1.5kW (if current is high, most of the voltage drop happens on the cable, so the power on the resistor is not that high). It's a lot, but I think still feasible. And if I allow voltage to drop bellow 320V for high currents (for example, I ask for 320V at 15A, then I only have to dicipate about 350W in the resistors)

I agree that having a higher voltage controler for the motors would be a nice solution (it would require also changing the regulator supplying the low voltages to the control electronics). It might be worth pushing a bit further (but there might also be a limitation due to the breakdown voltage of the motor itself). Standard 3 phase AC (for the types of boats we work with) is 380V AC. But I would be more than happy to buy a transformer to get it to another voltage if this means I can get rid of the expensive lab supplies we use right now)

Here's a buck that will do the job but you're not going to like the size tame-power.com/en/c-products/hv-hv-dc-dc-converters/…

Indeed : there are plenty of big solutions on the market, but they are out of specs for us. Size is probably going to be a big part of the challenge. On the other hand, there are many factors that can make things easier (low importance of efficiency, possibility to choose the supply voltage, possibility to drop voltage in the cable, water cooling for the enclosure (excepted at <3.5A)). That's why I think a custom solution might be needed

If you want to pay for a custom job, a higher voltage motor controller would be a better investment XD

Have you looked into cable voltage drop compensation by measuring the current and supplying Unom+I*Rcable?

Excellent idea : I will check if the supply can do this directly. If not, according to the manual, there is an analog interface that can be used instead if the digital one : it should be easy enough to adjust the voltage setpoint according to the measured current. I might still need a clamp (or maybe just a few capacitors) to avoid overshooting when the current decreases quickly, but now it's only transients I have to care about