A Very Accurate Current Probe

There’s many different ways of measuring current. If it’s DC, the easiest way is to use a shunt resistor and measure the voltage across it, and for AC you could use a current transformer. But the advent of the Hall-effect sensor has provided us a much better way of measuring currents. Hall sensors offers several advantages over shunts and CT’s – accuracy, linearity, low temperature drift, wider frequency bandwidth, and low insertion loss (burden) being some of them. On the flip side, they usually require a (dual) power supply, an amplification circuit, and the ability to be “zero adjusted” to null output voltage offsets.

[Daniel Mendes] needed to measure some fairly high currents, and borrowed a clip-on style AC-DC current probe to do some initial measurements for his project. Such clip on current probes are usually lower in accuracy and require output DC offset adjustments. To overcome these limitations, he then built himself an invasive hall sensor current probe to obtain better measurement accuracy (Google Translated from Portugese). His device can measure current up to 50 A with a bandwidth stretching from DC to 200 kHz. The heart of his probe is the LAH-50P hall effect current transducer from LEM – which specialises in just such devices. The 25 mV/A signal from the transducer is buffered by an OPA188 op-amp which provides a low output impedance to allow interfacing it with an oscilloscope. The op-amp also adds a x2 gain to provide an output of 50 mV/A. The other critical part of the circuit are the high tolerance shunt resistors connected across the output of the LAH-50P transducer.

The rest of his design is what appears to be a pretty convoluted power supply section. [Daniel] wanted to power his current probe with a 5V input derived from the USB socket on his oscilloscope. This required the use of a 5 V to 24 V boost switching regulator – with two modules being used in parallel to provide the desired output power. A pair of linear regulators then drop down this voltage to +15 / -15 V required for the trasducer and op-amp. His blog post does not have the board layout, but the pictures of the PCB should be enough for someone wanting to build their own version of this current sensor.

13 thoughts on “A Very Accurate Current Probe

  1. “If it’s DC, the easiest way is to use a shunt resistor and measure the voltage across it, and for AC you could use a current transformer.”
    The shunt resistor works great for AC too, despite the above implying that it doesn’t. There may be an (unstated) assumption that the DC being measured is a fairly low voltage and/or current while the AC is a high voltage and/or current, and while it is fair to say that the best measuring technique for each of those situations might be different, it doesn’t really relate to it being AC or DC.

      1. Absolutely. But that applies to DC just as much as AC, although in that situation AC would be the only one relevant.

        In case my original post wasn’t clear, the quote is the second sentence in the article.

  2. For stuff like this I just use a prepackaged DC-DC converter that outputs +/-15v. They are pretty cheap and you dont need much current for the LEM. I have a bunch of these LEMs around, some up to at least 4-500A. They are pretty nice.

  3. I expected something a bit more sophisticate, such as a custom made hall-effect-based sensor. The setup here is pretty common and I’m not even sure the power supply is well designed, many of these packaged DC/DC converters are not meant to be connected in parallel and have all sorts of issues when you do (if the voltage regulation is tight, the current sharing might not be even, the switching ripples may interfere and cause issues). In some cases these supplies don’t even work well if you chain them (for example if you need to generate 12 V from a 24V supply and 5V from the 12V supply).

  4. Hi, project author here…
    About the options to address this (current measurement), my views are as following:
    Resistor – Highest bandwidth, high power waste, no isolation, DC measurement capable, very good precision
    CT – Good bandwidth, low power waste, isolated, no DC measurement capable, ok precision
    Hall – Good bandwidth, low power waste, isolated, DC measurement capable, ok precision
    Closed loop Hall (the one I chose) – Good bandwidth, low power waste, isolated, DC measurement capable, very good precision (albeit expensive)
    Rogowski_coil – Very high bandwidth, low power waste, isolated, no DC measurement capable, ok precision, hard to make it work
    Fluxgate – Good bandwidth, low power waste, isolated, DC measurement capable, very good precision (albeit expensive and hard to make work)

  5. Author again…
    About the power supply:
    I wanted to power the probe from my scope´s USB port, so I had only 5V. Sensor needed +15/-15V. I used 2 DC-DC converters paralleled (not perfet but it works) to generate +24/-24V (with poor regularion and ripple) and then 2 linear regulators to make +15/-15V with good regulation and no ripple. A bit convoluted but worked… I got a very clean output.

    1. Thanks for replying. No objection with the combination of dc/dc plus regulator, I’ve seen it before. The parallel connection is just something I planned to do but I couldn’t find much information from treco power, which gave me the impression it was discouraged. Still, if the load regulation of the supply is bad, parallel operation works better, which in a way resonates with your need to use an additional regulator to get a good voltage. Good job anyway, keep your dick in a vice (as He says).

  6. I have a bunch of LEM current sensors I pulled from some circuit boards over a decade ago. They are ones with a hole in the center through which you put a buss bar. Since I have many to sacrifice I’ve often thought about trying to cut the ferrite core in half and make some sort of clamp on current probe out of one.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.