13
\$\begingroup\$

I need to measure current going through my PCB, which can be anything between -1200A to +1200A. Accuracy within 1-2% with bandwidth of anything higher than 50kHz.

The conductor is decided by now - I will use a copper bar.

The original plan was to use eight parallel 200uR resistors and measure the voltage drop across one of them. The problem is that bringing such current from the copper bar to several resistors seems to not be possible - the thin copper of the PCB might behave as a fuse.

\$\endgroup\$
2
  • \$\begingroup\$ We're getting flags about the comments having turned into a discussion, which is not an allowed use of comments. Therefore the comments have been moved to chat (see link below), where discussion (as long as it complies with the Code of Conduct) is allowed. || This bulk moving of comments to chat can only be done once per question, so any further comments posted here may be deleted without notice. Edits & new answers are still allowed. Keep comments & discussion on the question in chat from now on, please! \$\endgroup\$
    – SamGibson
    Jan 30, 2023 at 23:14
  • \$\begingroup\$ Comments have been moved to chat; please do not continue the discussion here. \$\endgroup\$
    – SamGibson
    Jan 30, 2023 at 23:14

6 Answers 6

1
\$\begingroup\$

Use a shunt meant for busbar / screw mounting. There is really no point in bringing that amount of current back onto a PCB, if you don't need to.

We had an automotive application where we measured some quite extreme battery currents. We used something similar to this shunt, just packaged up and potted. For just 24,40 € this is shunt is practically free.

https://www.bourns.com/docs/product-datasheets/csm2f-7036-0.pdf

I would add a half decent current sense amplifier or an ADC with PGA on a small pcb and be done with it, something like this but a little bigger. This, you can then screw this module to the ends of a busbar. something like the bigger automotive fuses

\$\endgroup\$
8
  • \$\begingroup\$ Is it simply pressed to the bus bar? I can't understand the direction of pressure. Do i mount them as SMT on PCB, then fasten to the bus bar on top of it? \$\endgroup\$
    – TQQQ
    Jan 30, 2023 at 15:41
  • 1
    \$\begingroup\$ I'd probably use a daughter board connected with a cable. You can screw that onto the busbars wherever needed, mounted something like this but with a current sense amplifier. \$\endgroup\$
    – tbuerg
    Jan 31, 2023 at 8:38
  • \$\begingroup\$ Yes, good idea. Frankly, your answer is my favorite so far :) When no new ideas will come up, i will tag it as the final answer \$\endgroup\$
    – TQQQ
    Jan 31, 2023 at 17:25
  • \$\begingroup\$ what do you think about soldering this resistor on the bus bars? \$\endgroup\$
    – TQQQ
    Jan 31, 2023 at 17:36
  • 1
    \$\begingroup\$ For low voltages you can find complete current measuring units, mostly for solar charging applications, like these, by victron. \$\endgroup\$
    – tbuerg
    Feb 1, 2023 at 13:42
20
\$\begingroup\$

Use several Hall effect sensors, arranged symmetrically round the bar, so that they add for the bar field, but cancel for external fields. Much the same principle as how a Rogowski Coil rejects external fields, but going down to DC as well.

Of course if you don't have the complexity of external fields, then only one is needed.

\$\endgroup\$
4
  • 4
    \$\begingroup\$ 2 Sensors should be enough to solve the two unknowns: 1) bar field 2) homogeneous background field. this already goes a long way. 2 Sensors at opposite sides, 1 flipped, could even be averaged in analog, to compensate for the background field influence. \$\endgroup\$
    – tobalt
    Jan 28, 2023 at 20:25
  • \$\begingroup\$ You can reduce the requirement for symmetry around the bar if you arrange them symmetrically around a magnetic toroid that is then itself around the bar, instead, but that requires cutting up a toroid to put sensors in--perhaps a bit difficult of a process. Ferrite's hard to cut without shattering. \$\endgroup\$
    – Hearth
    Jan 28, 2023 at 20:39
  • 1
    \$\begingroup\$ Magnetic field concentrators suffer from hysteresis, the magnitude of which depends on the characteristics of the ferrite (or powdered iron), and saturation can occur as well. They are mostly needed for low currents. \$\endgroup\$
    – PStechPaul
    Jan 29, 2023 at 0:54
  • 2
    \$\begingroup\$ @PStechPaul your criticisms are correct but they are used for both high and low current measurements. So called closed-loop sensors have a coil around the concentrator to keep it at zero flux continuously. The amount of current required to maintain zero flux is proportional to the current being measured. This avoids offset, hysteresis and saturation issues and makes degaussing unnecessary. \$\endgroup\$
    – user57037
    Jan 29, 2023 at 18:53
13
\$\begingroup\$

How about field sensors? If the busbar is rectangular it shouldn't be that difficult to place the sensor.

  • Example 1 : ACS70310
  • Example 2 : two or three sensors with integrated conductor like ACS772

or other devices from Allegro https://www.allegromicro.com/en/products/sense/current-sensor-ics

\$\endgroup\$
1
  • \$\begingroup\$ it's an option, thank you. The field concentrators are a headache though. But i will definitely explore this option \$\endgroup\$
    – TQQQ
    Jan 28, 2023 at 19:20
10
\$\begingroup\$

Use a linear Hall effect sensor attached to or close to the bus bar. They are less than two dollars. However, they measure the magnetic field, so the output will vary due to external influences, although at 1000A the effect should be insignificant.

This is available as a flattened TO-92 package or a SOT-23. The device must be mounted such that the magnetic field passes through the flat surface orthagonally.

DRV5053 TO-92

\$\endgroup\$
0
9
\$\begingroup\$

Super easy. Have a conductor of known characteristics somewhere in the system. This can (and arguably, should) be any conductor that has another job and is fit for the task. Having a dedicated ammeter shunt is unnecessary; any wire can be the "shunt" if its characteristics are known. And why wouldn't they be? :)

Measure the voltage difference between opposite ends of it.

schematic

simulate this circuit – Schematic created using CircuitLab

It is measuring down the voltage drop of the thick black line, because that line has known characteristics so it's a simple matter of plugging values into Ohm's Law.

It is vital that the voltage sense wire (fine black) have voltage drop on itself kept to functionally zero. That means the millivolt measurement device must itself be high impedance, and there cannot be ANY other load placed on these sense wires. You must resist the urge to use the sense wire as a supply wire for some auxiliary load.

Interesting note: Presuming that your supply is oriented with one leg of the source being nominally GND, the voltage seen by the voltmeter will be very close to 0 volts even though system voltage is hundreds of volts. However if the thick black wire suffers a casualty, all bets are off.

\$\endgroup\$
13
  • 3
    \$\begingroup\$ For better accuracy, the temperature of the copper conductor should be monitored and compensation applied accordingly. \$\endgroup\$
    – PStechPaul
    Jan 29, 2023 at 0:30
  • 8
    \$\begingroup\$ For 1% accuracy, the temperature of the conductor must be known to within +- 3°C. \$\endgroup\$
    – jpa
    Jan 29, 2023 at 7:23
  • 3
    \$\begingroup\$ 1% or even 5% accuracy with a copper conductor is rather not possible. Deformation, aging, oxidation - factors build up. \$\endgroup\$
    – fraxinus
    Jan 29, 2023 at 18:35
  • \$\begingroup\$ Copper is a terrible shunt material because of the high temperature coefficient. \$\endgroup\$
    – user57037
    Jan 29, 2023 at 18:46
  • 3
    \$\begingroup\$ @Harper-ReinstateMonica I am chemist, in fact, and not a real EE. Metallic copper is found sometimes, but it is pretty much known to oxidize in air. 3-4 decades old bare copper conductors do get black (CuO). Not enough to increase resistance to impractical level, pretty much enough to ruin your measurement efforts in an year or two. \$\endgroup\$
    – fraxinus
    Jan 29, 2023 at 20:03
5
\$\begingroup\$

To measure currents as high as this, magnetic flux sensors are preferable to resistive, as they give you a galvanically isolated signal.

If you don't have galvanic isolation, you have to be extremely careful with the wiring, since a stray connection on the sensing circuitry can cause very high currents to flow, with considerable damage and the potential for injury.

Regardless of the measurement method, the big problem you will face with any self-made solution is calibration; it is all very fine making your own resistor, but the value will be very low and difficult to measure accurately; even with an (expensive) low-ohm meter, you will struggle to get within your 2% accuracy requirement.

Similarly, how are you going to calibrate the signal from a hall sensor? I'm guessing you don't have a calibrated source of 1000A to hand, in which case you'll have to use the highest calibrated current you can muster, and multiple turns of wire (e.g. 100 turns of 10A to make 1000A) and hope that the magnetic field isn't too distorted by the approximation.

When measuring DC railway traction current (up to 600A) we did use inline current shunts (somewhat hairy on a moving train) but ended up using LEM magnetic flux current-sensors which were much easier, and not too expensive, considering what they were measuring; I recently found one on an auction site for minimal cost, since there isn't much demand for them.

\$\endgroup\$

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.