1
\$\begingroup\$

A friend is designing a power meter for 220 VAC mains and is considering Allegro current sensors like the ACS37002 or ACS71240. Those sensors put a wire with a sub-ohm resistance in line with the current and measure the magnetic field with differential Hall-effect sensors. Low resistance means low wattage even with high current, so those components have tiny footprints, like 8-pin SOIC. They also feature isolation of several kV between the measuring and the measured side.

Something feels wrong to me about putting a bare 8-pin SOIC in line with 220 VAC. Every other component on this side of the board is as bulky as you would expect, and then there is this chip. I would feel more confident if I saw a power TVS diode in parallel, but there isn't any. I can't find any relevant Allegro application potes. After reading relevant datasheets extensively, my main concerns are:

  1. If the load becomes a short circuit, the high current will likely make the component explode. Then, any isolation claim will be void and the measuring side may be connected to mains potentially destroying it and anything connected to it.
  2. There's a fuse in line with the load, but then it becomes a race of who will melt first which for me isn't straightforward, especially since between the sensor and the fuse there's a choke. You can replace the fuse, but can you be confident that the sensor is intact? I also don't feel confident that repeated short-time exposure to higher currents than specified won't eventually change the isolation properties of the chip. Motors, for example, can create currents of various sizes and shapes that will not blow a fuse.
  3. Some of those sensors claim MHz-level bandwidth. The PCB designs I often see, including Allegro's Evaluation Bare Board (a 6-layer x 2 oz copper beast) don't seem to try to minimize trace parasitic capacitances. The distance of 8-pin SOIC and the size of the pads necessary to handle large currents make it hard to design a PCB that handles high frequencies well. I guess there are a few concerns about high frequencies: a) the measurement on higher frequencies might be less accurate, b) hall sensor's functionality requires the absence of a Faraday shield, and it might be hard to prevent high frequencies from coupling to the power lines of the measuring side unless you add ferrite beads. Dimers, motor controllers and switching mode power supplies can create spikes with high frequency components.

How can the Allegro current sensors be properly used on circuits connected to mains power, while making minimum assumptions about the load?

Indicative Distances on Allegro's Evaluation Bare Board and pin distances from the ACS37030 and ACS37032 Datasheet.

Indicative Distances on "Allegro's Evaluation Bare Board", and "ACS37030 and ACS37032 Datasheet"

\$\endgroup\$
1

2 Answers 2

0
\$\begingroup\$

I used one of these in a prototype 24 volt system with a 40 amp hour Lithium Ferro Phosphate battery as the energy source.
A short circuit incinerated the current sensor, destroyed the isolation and applied 24V to the low voltage electronics. Fortunately, damage was minimal.

While it is probably possible to design a system so that the risk of catastrophic isolation destruction is minima, if I was using these in future I would provide an isolated low voltage system at the sensor output that was capable of surviving high voltage application. This is relatively easy as you are dealing with very low current at this point.

\$\endgroup\$
5
  • \$\begingroup\$ I couldn't rule out a sensor damaged in ways where higher currents could pass through the low-voltage side. Unfortunately this power meter connects through USB to a laptop leading to potentially expensive mistakes. \$\endgroup\$
    – neverlastn
    Commented May 2 at 3:00
  • \$\begingroup\$ @neverlastn On the low voltage side a voltage clamp (zener or other) and a fuse or similar can be used to isolate the low voltage connection. As you are working at low current this is far easier than on the high current size. In the general case, If this is a digital signal then a series resistance is easily accommodated. If an analog signal you may need an added buffer or a voltage divider or ... . Optoisolation either analog or digital is also doable. \$\endgroup\$
    – Russell McMahon
    Commented May 2 at 3:58
  • \$\begingroup\$ Whatever method chosen adding secondary isolation is relatively easy and in most cases well worth doing. After my experience I would NEVER rely on the current sensor isolation alone EXCEPT where maximum possible energy would still be in spec. \$\endgroup\$
    – Russell McMahon
    Commented May 2 at 3:58
  • \$\begingroup\$ True that. I very much appreciate that your experience ("applied 24V to the low voltage electronics") clearly proves that there's the need for careful design beyond quoting the irrelevant high isolation voltage and calling it a day. \$\endgroup\$
    – neverlastn
    Commented May 12 at 1:24
  • \$\begingroup\$ @neverlastn Yes. As well as careful design I would, as I noted, strongly consider adding low voltage/ signl level output isolation. This is because, while Murphy cn try hard with both low or high energy circuits, on the low energy side it is vastly easier to design an interface that can deal with the maximum voltage and the current is low by design. eg an optocoupler can be designed for voltage only isolation (plus possibly delta V issues). \$\endgroup\$
    – Russell McMahon
    Commented May 12 at 11:18
0
\$\begingroup\$

Some of the ACS37002 parts can handle 133A, and has a 1mΩ resistance through the IP+ channel. So if you were to dissipate the power from 66A in the part, it would be 66mW. The worst case thermal resistance is 8C/W so this would only be ~1C temperature rise, the part won't even flinch at that much current put through it.

The device can only measure 133A, so you must limit the current with whatever margin you feel appropriate and never exceed that current above that value although the part could probably take more abuse since it's just a conductor. So set the fuse to something like 133A with whatever margin you need.

You do need to provide proper creepage and clearance, you design should use a standard like IPC-2221B, whatever NRTL you are passing your design through should be able to comment on what your specific standards are for saftey.

Even if you aren't going to sell this design, you'll need proper creepage to avoid arcing, which means if you have two traces on the PCB at 220V (like a high voltage 220V an low voltage), you'll need 1.25mm of spacing.

enter image description here
Source: https://resources.altium.com/p/using-an-ipc-2221-calculator-for-high-voltage-design
enter image description here

\$\endgroup\$
1
  • \$\begingroup\$ I agree. I usually have .8cm clearance whenever in doubt. I would do that in this case, too. A potentially switched inductive load like a motor, transformer, or fluorescent light can create 2 to 6 times the nominal voltage. On 220 VAC, one can't rule out kV-level inductive spikes in case of insufficient snubbers, flyback diodes, etc. Those sharp spikes have energy in high-frequencies. I updated the question with some distances on the board and between pins. I feel that the small package dictates a particular design that should be provided as a reference for use with 220 VAC mains in the wild. \$\endgroup\$
    – neverlastn
    Commented May 2 at 1:42

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.