2
\$\begingroup\$

Update

It was pointed out the PE wire has to carry large current in fault condition so using a 3-phase CMC is problematic. In IEC62368-1 (2nd edition) chapters 5.6.3 and 5.6.4 it is implied that the protectice current rating for EU is 16A and US 20A, this is based on common fusing in the regular mains indoor setup. So that translates to 1.25mm^2 wire for 16A and 1.5mm^2 wire for 25A. That's one chunky CMC. Possibly you could get away with that thick wire for PE and lighter wiring for L and N, if you're procuring a custom CMC.

An alternative / additional measure could be to use a heavy-duty ferrite on the PE line with regular CMC attached to the L and N lines.


Like the title says. On IEC61000-4-6 conducted immunity test, a coupling device is used to inject 150kHz - 80MHz interference to the external cabling. As this noise is common mode in nature, it's not filtered by ferrites, pi-filters, 3-terminal caps etc.

The device here has a touch screen, which has a sensitive front end and these are notorious to be nasty with regards to immunity testing as you're dealing with something designed to pick up very small changes in capacitances.

See here for example: How do I Design Capacitive Touch Interfaces with EMC in Mind?

Standard response for this is to include a common mode choke that'll take the noise right out, at least if your part frequency range and characteristics are OK.

Now if you have a device with earthing, you have a separate PE wire. This will happily carry the noise right past your CMC and you will have your chassis acting as an antenna.. Which messes with our sensors. I have verified this is indeed the vector as disconnecting the PE makes the problem go away, however this is not acceptable to the client.

Now there are three-phase CMC inductors which are obviously designed for high-er power 3-phase devices. My immediate instinct would be to find a suitable 3-phase CMC and connect the extra winding to the PE.

Is this going to work? Initial testing indicates not so well. Why not? Shouldn't the inducted RF be exactly the same on L + N + PE as it would be for L1 L2 L3?

2-winding choke

This is what I expect to see:

3-winding choke

3-winding CMC connection: enter image description here

NB: We're trained professionals, whatever the euphemism is in EN60950 that allows using life-threatening contraptions like this. Do NOT try this at home.

\$\endgroup\$
  • \$\begingroup\$ I'm reasonably certain your protective earth should avoid being coupled to live in any way. And putting it through a transformer winding next to live and neutral is just going to make your noise worse! \$\endgroup\$ – Hearth May 10 '18 at 15:12
  • \$\begingroup\$ @Felthry I think the coupling @ 50Hz (or 60Hz) is going to be pretty neglible. In any case it's a different conversation, it might zap our system at fast transient but right now I have touch screen showing garbage at this test. It's a known problem actually with touch screens which are very sensitive devices designed to pick up small differences in capacitance. \$\endgroup\$ – Barleyman May 10 '18 at 15:20
  • \$\begingroup\$ Edited OP to specify this is a touch screen device and put in a link which shows what's going on. \$\endgroup\$ – Barleyman May 10 '18 at 15:23
  • \$\begingroup\$ If you do, last time I checked for 60950 you had to pass 25 A continuously with limited temperature rise on any such ground choke. \$\endgroup\$ – winny May 10 '18 at 15:52
  • \$\begingroup\$ @winny care to dig out the chapter for that? \$\endgroup\$ – Barleyman May 10 '18 at 17:25
1
\$\begingroup\$

I would first attempt to improve the integrity of the PE connection path, either with a shorter/lower inductance path to ground, or a dedicated "functional" PE. If that doesn't suffice, then I would consider galvanic isolation in your sensor path.

Common mode voltage becomes a problem only when a secondary potential exists to turn it into differential mode voltage. Consider the lineman working on a 110kV transmission line from a helicopter. He's at a common mode potential of 110kV; his bigger safety risk is rotor failure.

\$\endgroup\$
  • \$\begingroup\$ PE connection path is quite simply from 3-terminal mains plug directly to the chassis. I added an image of the ugly 3-winding test we did. Galvanic isolation for touch sensor is no go and wouldn't likely work anyways as it's liable to pick up the noise by capacitive coupling or near field RF. \$\endgroup\$ – Barleyman May 10 '18 at 15:42
  • \$\begingroup\$ If you build a common mode choke, I recommend you keep the wires that pass through the core twisted, same as in the power cord. This will cut down on interference. Have a look at the pictures shown on this site for ideas: palomar-engineers.com/rfi-kits/acdc-power-line-chokes \$\endgroup\$ – AlmostDone May 10 '18 at 18:27
  • \$\begingroup\$ It's a commercial product so I'm not going to build anything :-) Well, perhaps I would design something and ask a contractor to build a couple of thousands of them. Anyways there was a good point in comments re: 25A for PE current withstand. 25A 3-winding chokes definitely do exist but they're pretty darn chunky. \$\endgroup\$ – Barleyman May 10 '18 at 21:23
1
\$\begingroup\$

Based on further feedback from the customer I can draw some conclusions.

  1. The 3-phase CMC does in fact work as advertised, it's just not good enough. The large size touch screen is really great at picking up noise so you need very robust filtering.
  2. Adding a choke to the PE wire is indeed problematic. If you check requirements of IEC62368-1, if the AC input has only one fuse, it's presumed your plug is going to be the other way (live not fused) and the fuse is bypassed. You then have to use building fuses as the benchmark for sizing the CMC wiring on the PE line. This translates to 1.25mm^2 for EU 16A fuses and 1.5mm^2 for US 20A fuses.

The way around both problems is to add fuse(s) to cover both L and N wires. This way in short circuit situation the current is limited to whatever your fuse protects against and the wire size can be reduced accordingly. For example, 3A current limit implies 0.3mm^2 wire is enough. If you use the limited short circuit test on standard Annex R to demonstrate the overcurrent device trips reliably without PE becoming HAZARDOUS LIVE, you can get away with less.

Since we now can use reasonably thin CMC wire, we can commission a 3-phase CMC using a lot more turns that will in turn suppress the common mode noise carried in the PE wire better and will hopefully eradicate the problem with the touch panel. Well, it will, if you keep throwing inductance at it. This may need two separate CMCs, one for low frequencies (150kHz-1.5MHz) and one for "high" >1.5MHz frequency range.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.