I mean, "not going bankrupt" is probably very different for someone having a steady job as EE in central Europe and someone still at university on his own money in an emerging job market. So, maybe an impression of what would make you bankrupt would help. Also, do you have a way of calibration, i.e. can you conduct at least 2 (better 3) different currents that you know through the copper which you know / can still measure that are not "super close" compared to your overall range? What did you look into already? Why are the standard solutions out?

@MarcusMüller standard solutions are in. The budget constraints stem from the customer's pressure on volume price and availability of parts in stock. LEM for example are expensive and mostly out of stock. Calibration- i am all in. Not sure i can properly account for temperature though. Copper has 1% error on 2.5 degrees. But this is absolutely an option.

It is not clear if you need to pass the current through PCB anyway for some reason or if you want to use PCB only for measurement purposes. If the second, why not use a "screw mounted" shunt resistor which can be attached directly to bars/structure? There are some down to 25 uOhm available. 1 % total accuracy can be quite a challenge (or expensive), but with callibration and low TCR maybe acceptable? (But note that paralleling 8 resistors, it will be almost impossible to get exactly even share of current, so you can not have easily <1% measuring only one of them ...)

What is the frequency range of the current? Lowest to highest? What do you mean by "with bandwidth of anything higher than 50kHz" -- do you mean you want to know the current of any component with frequency from 50kHz up? Do you mean all the way up to visible light?

1200 A through a PCB is sheer madness. I suggest you rethink your approach.

@DavideAndrea there will be a bus bar. PCB is to host MOSFETs and stuff around them.

@TimWescott it is a motor drive. So the current loop will have bandwidth of somewhere around 2-3kHz, maybe 5kHz, then the sampling should be higher. 50kHz works well. Higher bandwidth is welcome.

Please show schematic including where you are sensing the current as it goes through the driver to the motor.

1000A 50mV or 100mV shunts are around $50 and have extremely good accuracy (0.25%). Is that too much?

@SpehroPefhany 1000A * 0.05 V = 50 Watts. Maybe it is the heatsink that will bankrupt you.

It is a bit expensive, but the other problem is that this resistor alone will dissipate 50W. That becomes an issue... I mean, i prefer measuring on a resistor, but then i simply take 8 of 200uR and measure on one of them. In terms of cost it is probably 8-10 dollars per all shunts.

One motor controller design I looked at used a 31 Amp current sensor even though it is an 80 Amp controller. They used copper trace to divide the current into two different paths and measured the current in one of the paths. Even though copper has a high tempco, in this case, both paths are mostly copper hand have the same tempco. So you could use copper busbar material with precise dimensions to divide the current into several pathways, and use a magnetic sensor to measure only one of them.

@TQQQ You could buy a 2000A shunt and run it at 1000A but that would be more expensive. Accuracy may be an issue- smaller signals and the resistors not sharing current equally. But maybe you don't need high accuracy. I expect this will be a challenging environment EMI-wise no matter how it is approached (and something like the LEM sensors would help with that).

Too little information to run the numbers but using the copper bar itself as a shunt might be an option

Well, a copper bar for 10uR is 25mm long, 2x2mm in cross section. It way thinner than any wire rated for 1000A. I am not sure what to do about it

@SpehroPefhany I don't expect EMI problems, usually measuring the shunt works great. Uneven current distribution is a problem though. Except calibration, do you have ideas?

Average the voltages across the series resistors with 8 pcs 1% resistors on each side.

That would be 8 adc 😱

Does your accuracy requirement hold for the entire range? E.g. if currently only 1mA are flowing, you want your accuracy to be 0.01mA?

Copper is a terrible shunt material due to the high tempco. Around 4 percent for every 10 degrees. So we could call that 0.04/10 deg = 0.004/degree. That is 4 parts per thousand or 4000 ppm tempco.

I think divide and conquer. You let the current split up and flow equally through 6 busbars (so that is 200 amps per busbar). One of the 6 has an ACS 722 on it. The ACS772 is 100 uOhms and can measure up to 400 Amps. If necessary, you can add a restriction to the other busbars to mimic the 100 uOhms of the ACS722, but that might not be needed.

Just so we know, where are you getting 1kA from? Maybe you don't actually need to be measuring the current here.

@Graham it's a motor drive

@laolux no.. it will be measured with a 14-bit ADC. Hard to tell the exact requirements, but generally the more accurately the better.

When working with 1000A, my first priority would be not to die.

Tamura L06P800S05 will get you to 800 Amps for 25 USD per channel. Tamura S29S1T0D24Z will get you to 1000 A but it is 5x the price. I know one guy who is basically re-creating a Tamura type hall-effect sensor using laser cut electric steel for the concentrator and off-the-shelf linear hall sensors PCB mounted in the gap of the concentrator. He does have to fight drift problems and temperature coefficients, etc. Whereas Tamura seems to already have that problem beat. But he is getting a 1000 Amp sensor for low unit cost (but not zero R and D cost).

@TQQQ No, that's where the current is going. Where is the current coming from? Battery or mains? Is it going through a step-down transformer to get that current? And on the subject of possible X-Y problems which have a better solution elsewhere, what are you wanting to do with this current measurement?

@Graham I am pretty sure the OP is measuring the phase current outputs from a battery powered ESC. So it is correct to say the current is coming from the motor controller.

@Graham let's assume the current is coming from a battery. What i need to achieve is the best current sensing i can on the given budget. It is good for a lot of things, all around algorithms.

@mkeith thank you for Tamura, i somehow missed it. I am preparing an option for a concentrator and an Alegro hall sensor, but it scares me. Even the good old ACS712 was noisy, this thing will probably, like you said, drift and change.

@EricDuminil i have 100kohm on my fingers, so i am not in danger by current :) Next step is going to be raising voltage to something serious, hopefully well past 400V. That will take some courage.

You could hook up a torque sensor to a prybar and mount it parallell to whatever you are doing.

@TQQQ At that current, you're in a LOT of danger! The best case (and this is best!) is that things gradually gets hot and melt or set on fire. The worst case with that kind of current is that they literally explode and shower you with burning fragments. If you live on your own then perhaps you're the only person at risk, but if you live with other people then you are a direct risk to their lives, as surely as if you were firing bullets through walls at random. You literally don't know how dangerous what you're doing is. And now you're talking about 400V?! Stop. Now.

@TQQQ FWIW, I have worked on equipment with those kinds of voltages and currents, when I worked on the electronics for HVDC links. We had inch-thick aluminium bus bars, we had serious water cooling on all the electronics, and any testing where we were unsure happened with the equipment several metres away, in a high-ceilinged industrial space, with proper fire extinguishers on hand, and with us behind a perspex window. And that's for equipment designed by very serious engineers with 20-30 years experience. You don't take risks with high-power stuff if you want to stay alive.

@Graham thank you for your concern. I promise I've been in this business for 15 years and i have certain idea about what this is about. You are right on all accounts. While 1000A is a bit new for me, 400V is not, and hundreds of amperes is not new as well.

If you are going to 400 V why do you need 1000 phase amps? Is this motor running a big machine in a factory or something?

400V will not go higher than 400A. Both are traction motors