Isolated AC-DC Wall power supply -> 118V AC on barrel connector

[JKSniper]

Hello!

I will try to explain this and why does it matter to me - I need your help.

So, I live in Slovenia, Europe, where we got 230VAC 50Hz. Socket is properly earthed. All AC voltage readings are done between Earth PE (main socket) and barrel connectors.

- For project I have Meanwell GST60A12-P1J 60W 12V AC-DC power supply via "3-prong cable", which has L, N and PE. On output I get 12V DC on the barrel connector, but also I also get 118VAC on both + and - of the barrel connector in reference to Earth. This seems to be exactly 1/2 of VAC of the socket (236V).

- When I compare this with 10 other wall plug power supplies (I've tested all I have on hand), they mostly are 80-90VAC, some are 35VAC, some around 17VAC.
- DELL laptop power supply brick has 0VAC, as it seems it has DC GND internally connected to Earth PE.
- HP laptop power supply brick also has 0VAC. Same as DELL.


Now why this matters to me? I've found out that my PCB that I power with 12V through this PSU has problems. It has long wires to sensors and sensors are really sensitive. I've found correlation with this VAC reading and problems with sensors. PCB works with ANY other PSU, except the one I've bought for it (Meanwell GST60A12-P1J), it also starts working if I touch the PCB with my finger, decreasing the VAC to around 70-80VAC. It also works if I have USB connected to it, as it's connected to PE ground via PC, it also works if I try to probe it, as scope has GND connected to ground PE, ofcourse it's also working with lab power supply etc.

Obivousely, choosing different PSU solves the problem instantly for me, but I want to know what is happening, why and what I can do on the PCB itself to reduce this. I've replaced the cable to the sensor with shielded one, which has shown a little improvement, but the sensor is still twitching (i've tried connecting shield to DC GND or float and it's the same). So shielded cable doesn't help with this issue. Is there anything else I could do to the PCB or the sensor, apart from the obvious (choosing different PSU or connecting PCBs DC GND to Earth PE)? I mean like some chokes, ferrites, capacitors...?

Please, if possible, explain why this is happening and why I'm having probelms because of it. This is first time seeing this phenomena and I want to understand it. I'm also "too short" to know what exactly to look for, but I've found this youtube and I think this is actually what I'm seeing here.


I'm also curious on how to know when buying, if barrel GND is tied to PE Earth or not.

Best regards,
Jure

[tooki]

This question gets asked every few months.

The cause is the EMI suppression capacitors. They shunt high frequency noise from output to input, but also allow approximately 1/2 mains voltage to appear on the output, albeit at such tiny current that it normally is totally irrelevant.

This normally disappears in grounded power supplies, or by bridging one side of the output to ground. (This is almost certainly what your Dell and HP power supplies are doing.) So your annotation on the Mean Well datasheet of “is that it?” is exactly correct.

As for how you tell when buying… well you look at the datasheet.

When buying power supplies, you can also look at “medical” power supplies, as they are designed to have extremely low leakage currents. They only cost a little bit more than regular versions.

[JKSniper]

That is outstanding and fast response.

May I ask - is there anything a person could do, with such PSU, to a PCB or sensor itself, to minimize the problem with it? Where exactly lies the problem of having 118VAC on the PCB itself, when everything on PCB is DC? The sensor is very low consumption, it takes less then 1mA on 3.3V. Is there a problem? Does cable with GND and 3.3V which has 118VAC acts as an antenna in regards to Earth and do something?

Sorry if I might sound stupid asking this. I'm just trying to squash this bug, as I feel just replacing with different PSU is like sticking a head in the ground and hoping the problem doesn't emerge again.

[Siwastaja]

You can simply connect the DC output (your project's DC GND) to protective earth.

Or if you can't / don't want to do that but need to keep your DC and PE separate, then pick a power supply which leaks less current (e.g. so called "medical grade"; they are not much more expensive, e.g. Meanwell makes those). There is no such a thing as absolute zero leakage. There is a reason why many products are supplied with a power supply (even if it's from a generic manufacturer like Meanwell) and manual saying not to use any other power supply. Sometimes things like leakage currents matter as you have noticed.

And of course, you can redesign your circuitry to operate without problems even in presence of this current-limited half mains voltage respective to protective earth.

[Someone]

You can simply connect the DC output (your project's DC GND) to protective earth.

Or if you can't / don't want to do that but need to keep your DC and PE separate, then pick a power supply which leaks less current (e.g. so called "medical grade"; they are not much more expensive, e.g. Meanwell makes those). There is no such a thing as absolute zero leakage. There is a reason why many products are supplied with a power supply (even if it's from a generic manufacturer like Meanwell) and manual saying not to use any other power supply. Sometimes things like leakage currents matter as you have noticed.

And of course, you can redesign your circuitry to operate without problems even in presence of this current-limited half mains voltage respective to protective earth.

Another option being to buy a power supply that explicitly connects the output to PE for you already (and typically has larger leakage but at lower voltage).

[Zero999]

Here's a simplified schematic of a switched mode power supply. The leakage current is passing through the Y1 capacitor.


In addition to the above suggestions, you can connect the output to protective earth via a suitable non-polarised capacitor, say 100nF to 1µF. This will shunt the AC leakage to earth, whilst ensuring it's still isolated at DC. The capacitor doesn't have to be rated to the full 120VAC, because the voltage dropped across it will be tiny.

[JKSniper]

And of course, you can redesign your circuitry to operate without problems even in presence of this current-limited half mains voltage respective to protective earth.

Is there any trick to eliminate this "problem" at the very entry point of the PCB with a filter, choke or something? Since having DC barrel connector at the input and whole PSU as stand-alone component, it's very hard to reach the "PE" at all. At this point to continue with project I'll just buy medical grade PSU, as it has this solved, but I would really like to know why is that affecting my PCB and what I could do about it as preventive.

Thanks!
Jure

[TimNJ]

Your issue is common-mode in nature, which indicates your device has some coupling to earth (well, everything does). At mains frequency, common-mode filtering is usually large and expensive. There could be a CMRR issue with an amplifier in your sensor circuit. If your sensors have high impedance/low level outputs, they can be easily disturbed. You may want to amplify directly at the sensor node, if applicable.

If you can post a schematic and a photo of how everything is physically laid out in reality, I think people can better estimate exactly what is going on.

[Haenk]

I'd replace the switchmode PSU with a linear PSU. (That of course could be a problem when you need tens of Amps, but for anything up to 5 Amps or so, it should be easy to build or obtain...)

[Zero999]

And of course, you can redesign your circuitry to operate without problems even in presence of this current-limited half mains voltage respective to protective earth.

Is there any trick to eliminate this "problem" at the very entry point of the PCB with a filter, choke or something? Since having DC barrel connector at the input and whole PSU as stand-alone component, it's very hard to reach the "PE" at all. At this point to continue with project I'll just buy medical grade PSU, as it has this solved, but I would really like to know why is that affecting my PCB and what I could do about it as preventive.

Thanks!
Jure

You haven't posted enough information to go anywhere near solving the problem.

What does the PCB do?

What sensors?

What is the problem with the sensors?

Something will need to be redesigned. At the very least try screened cable to the connectors. Failing that I would consider using an isolated DC:DC converter on the board to power all of the sensitive, low level analogue circuitary and switch the high power loads i.e. motors, lights etc. with relays, which are isolated.

[Siwastaja]

I'd replace the switchmode PSU with a linear PSU. (That of course could be a problem when you need tens of Amps, but for anything up to 5 Amps or so, it should be easy to build or obtain...)

A medical grade low leakage SMPS might be almost as good, and they are not special or expensive even if "medical" sounds like it. And this only matters if earth connection is not available at the plug. If it is, you can choose an earthed SMPS which behaves like most earthed SMPS do (showing very low leakage voltage, because leakage current, even if large, is directed into the earth connection of the plug.

[JKSniper]

I have changed to the other PSU and device happily works now. I will do more testing to see if I can pinpoint the component, cable, sensor or something that is misbehaving and why. Thanks for all support and good recommendation to get different PSU.

[Nominal Animal]

You might want to look at medical grade class II supplies, double insulated ones without any protective earth connection, and no EMI suppression capacitors between the output and the input.

Medical grade is just a simple term for Means of Patient Protection leakage current limit specified in IEC60601-1: 500µA for AC and 50µA for DC in normal operation.  I prefer them because they give no tingles when touching one of the DC outputs.

For example, the Mean Well GSM60B12-P1J (21.50€ at TME.eu, 22.80€ at Mouser, and so on) is equivalent to the one you already have except with class II isolation, 2×MOPP (leakage current < 50µA), and < 0.1W no-load current.
MPM-65-12ST is a 65W class II 2×MOPP encapsulated supply (109mm × 52mm × 33.5mm) with 2-pin screw terminals for AC input and 12V outputs (and four 3.5mm mounting holes at 33mm × 98mm center-to-center spacing, 6.7mm thick).


The GST60A12-P1J you already have is specified to have up to 0.75mA (750µA) leakage current, and as it is class I, it has a EMI suppression class Y capacitor from the output to the input.  The datasheet explicitly says that the negative output is not connected to functional ground.

Slovenia and Finland both use either 2-pin non-grounded europlugs, or CEE 7/3 "Schuko" grounded sockets.

Now, I'm only a hobbyist, but it seems to me most class I supplies sold internationally are designed to have the EMI suppression capacitor between the 0V output and neutral, not the separate functional ground (on the springy side contacts in Schuko sockets).  Unfortunately, the CEE 7/3 sockets are not polarized –– there is no "live" or "neutral" –– and depending on how the mains AC is generated (especially when derived from two legs of incoming three-phase power), this can lead to the output to be coupled to midpoint of the AC input.  (In other cases, it could be coupled to the most negative potential; in which case rotating the plug 180° in that particular socket should yield coupling to 0V.)

If so, for us using non-polarized AC sockets, using double-insulated class II medical/low-leakage supplies without a ground connection is one alternative.
Another is to use class I medical/low-leakage supplies with the EMI suppression capacitor between output and functional ground, not output and neutral; but because most of the world seems to think in terms of "live"/"neutral", I don't think those are common in wall-wart/desktop adapter forms.

[JKSniper]

I've bought GSM60A, so I use the same 3-prong AC cables.  There was no differnece in old PSU in regards to how I turn the power plug in. "normal" or 180°.

Thanks for your input!

[Phil1977]

There are also SMPS-modules with PE-connection that use two Y-capacitors to get rid of this issue:

Primary      PE     Secondary
     o---||--o--||---o
         C1     C2


This way the capacitors shortcut the EMI-relevant AC, but all mains frequency components are sunken in PE without directly connecting the secondary side to PE.