High voltage issue of Raspberry Pi 4 PSU with EU mains connector

[radiolistener]

I just catch serious issue with official 15W Raspberry Pi 4 PSU with EU connector.
Some of my equipment was burned out when I connected it to Raspberry Pi, I started investigation how it happens and I found that Raspberry Pi PSU has 80 V AC voltage between its GND pin and mains ground. Since all other devices are powered from good PSU with zero voltage between GND and mains, it leads to 80 V AC voltage between GND pins of devices which has it's own PSU (for example display, sound amplifier, etc). As result it can lead to damage during connect/disconnect.

Usually such issue happens with PSU which is designed for 3-pin mains connector (Ground + Neutral + Phase) when such PSU is connected into mains with no Ground wire. In such case it is recommended to add Ground wire or just remove filter capacitors between N-GND-P pins in the PSU.

But the issue here is that Raspberry Pi 4 PSU with EU connector by design has 2 pins mains connector (Neutral + Phase), there is no Ground pin, so it should not have filter. But it seems that it has filter and it leads to high voltage potential on GND pin.

I'm not sure if this is mistake of Raspberry Pi designers or I just got a broken PSU.
Can someone check if this issue present on all Raspberry Pi PSU?

[Someone]

Check:
PSU 2.5A 5V MicroUSB white and black: 2VAC
PSU 15.3W 5V 3A USBC black: ~80VAC
PSU 27W (PD 5V,9V,12V,15V) USBC white: ~120VAC.

Nasty.

But what is the current/energy available when grounding that? Well designed equipment will withstand those voltages on their I/O connections from compliant (several hundred uA leakage) unearthed power adaptors.

[Jeroen3]

The equalization current between the two floating supplies will still fry any circuit not intended for hot-swap.

[wraper]

Some of my equipment was burned out when I connected it to Raspberry Pi, I started investigation how it happens and I found that Raspberry Pi PSU has 80 V AC voltage between its GND pin and mains ground. Since all other devices are powered from good PSU with zero voltage between GND and mains, it leads to 80 V AC voltage between GND pins of devices which has it's own PSU (for example display, sound amplifier, etc). As result it can lead to damage during connect/disconnect.

It's not about PSUs being good in comparison, they're earthed and most likely having weaker insulation. RPi PSU is double insulated.

there is no Ground pin, so it should not have filter.

Nonsense.

I'm not sure if this is mistake of Raspberry Pi designers or I just got a broken PSU.

Neither.

[Wolfram]

Practically all EMI compliant double-insulated switchmode supplies have a Y capacitor across the isolation barrier, that will leak some amount of AC current if the output is grounded to mains earth. Hot-plugging equipment that uses non-grounded supplies is never a good idea, unless using specific connectors that guarantee earth is connected first.

[Siwastaja]

As others have said, completely normal. It's interesting if this is your first time coming across Y capacitors in switch mode power supplies. This has been normal since 1990's.

You can get lower leakage current "medical grade" supplies to reduce the tingle, or better yet, you can fix your stuff so that it does not break. Signal hotplugging simply does not work without protection. Even in absence of Y capacitors, you still have potential differences e.g. static charges. If you have a hot-pluggable input, you just make it survive ESD (e.g., 8kV at some specific level of energy and duration), and also sustained Y capacitor leakage current. Or you design your circuits and connectors such that undesired grounding path cannot happen.

You can be thankful to Raspberry Pi's power supply as it revealed your broken design.

[Faringdon]

This does sound like some insulation problem as all offline isolated supplies should be ok with connecting their secondaries even if hot-plugging.
And even if some have earthed secondaries, and some are just "floating secondaries".
Saying that, a major TV co used to have (probz still do) 1 MEG resistors across the mains isolation barrier of their offline flybacks.

[uer166]

It's a proper Y1 rated cap and there is very decent creepage in the supply. There is likely nothing wrong with this supply.

The spark and small shock you experience is you discharging that Y-caps, that is expected.

[uer166]

You can be thankful to Raspberry Pi's power supply as it revealed your broken design.

Where or what is it broken when I get a spark or a shock while connecting HDMI ? The monitor is going nuts for a few seconds.
I can understand that the PSU has to pass some compliance test but RaspberryPi has a fully exposed header FCS.

The monitor or its' earth connection back to PE if it has one. It should have zero issues discharging that cap back to PE.

Edit: it seems like the monitor just.. works, so actually nothing wrong there either?

[floobydust]

The guideline is plug in the AC adapter LAST and unplug it FIRST when disconnecting cables. You can damage HDMI ports etc. with the stray voltage.

I have Raspberry Pi AC Adapter, Model KSA-15E-051300HU, 5.1VDC 3.0A, Made in Cambodia Kuantech Co. (same as OP's), North America 2-prong mains connector.
120VAC mains I measure: 50VAC 31uA AC for leakage. But hard to get even 100pF capacitance reading.

[Siwastaja]

1) Protect hot-pluggable data lines properly in design, like everyone else does.
2) Don't randomly modify compliant products. They are fine.
3) Accept the small tingle as a fact of life, as everyone else does.
4) If you still are unsure, instead of modifying compliant power supplies into something potentially dangerous and at very least non-EMI-compliant, get medical grade supply. They cause less tingle.

[radiolistener]

As others have said, completely normal.

This is not normal, all my PSU have 0 VAC between their output GND pin and ground (they all have ground pin on mains connector). Our country don't have Ground pin on mains connector but I implemented it in order to avoid such issues.

I also have branded USB chargers with no Ground pin, they have some moderate AC voltage between their output GND pin and Ground, it's about 5-6 VAC.

But RPI PSU has 80 VAC, and this is very bad, because RPI is intended to connect other devices such as HDMI display or USB devices which may have it's own PSU or just connected to Ground. When you power on it when all cables is connected, their GND is connected together and may work ok. But if you try hot plug/unplug such device it can lead to damage because high voltage AC can flow through sensitive signal inputs.

You can be thankful to Raspberry Pi's power supply as it revealed your broken design.

It revealed that official RPI PSU has broken design and can damage your equipment with usual use cases.
It damage my USB sound card when I connected audio cable to amplifier which has 0 VAC on it's GND pin.


Now it's pretty clear how it happened. I think it happens because I disconnected HDMI cable from display which also has 0 VAC on it's GND pin.

When display was connected it shortened RPI GND to the Ground, so RPI GND voltage was 0 VAC and all worked ok.
Technically when RPI is connected to display which is powered with proper PSU, it connects RPI PSU GND with Ground through HDMI cable and display PSU. So when there is connected display with proper PSU all works ok.

But bad thing happens when you disconnect HDMI cable. When I disconnected HDMI connector RPI GND pin voltage raise to 80 VAC due to by design issue in official RPI PSU.
As result it damage my sound card...

I don't have such issue with all my PSU which has properly implemented mains ground.

So, now I'm looking for replacement with proper PSU...

I think it can also damage display or HDMI port if you power your RPI from official RPI PSU and do hot HDMI plug/unplug.

Be careful with it...

The bad thing is that official PSU don't have Ground mains pin and its enclosure cannot be disassembled so there is no way to remove capacitors to fix it.

[Siwastaja]

Voltage does not matter. In a dry winter day, you can easily have 20 freaking KILOvolts going into the connectors. So 80V is next to nothing.

Current and energy matter. When you read 80V, you are measuring with your multimeter with some input impedance, maybe 1Mohm, forming a resistor divider from the actual 115VACrms. Apply Ohm's law to find the current. It is quite small because the very large input impedance of your multimeter already brings it down this much.

[ejeffrey]

But RPI PSU has 80 VAC, and this is very bad, because RPI is intended to connect other devices such as HDMI display or USB devices which may have it's own PSU or just connected to Ground. When you power on it when all cables is connected, their GND is connected together and may work ok. But if you try hot plug/unplug such device it can lead to damage because high voltage AC can flow through sensitive signal inputs.

Only if the products are defective.  If a plug is allowed to be hot plugged, among other things the shield ground shield *must* make positive contact before any data pins.  That will shunt the few hundred microamps  leakage safely to ground and equalize the grounds of the two devices.  This will also dissipate static electricity.  All standard connectors like USB and HDMI are supposed to be designed this way although there is plenty of crappily made products and cables, or products designed by idiots that doesn't think they need to connect the shell ground to their circuit.

There is also a bit of a problem using equipment designed to be operated with grounded power of that ground isn't present. They may have much higher  ground leakage than a double insulated device is allowed.  Normally this only shows up when people use ground defeat adapters but can also be a problem in countries where grounded outlets aren't mandatory.

The bad thing is that official PSU don't have Ground mains pin and its enclosure cannot be disassembled so there is no way to remove capacitors to fix it.

Please don't modify any mains power supplies.  The raspberry pi power supply is not broken, but now I think every other power supply you own has been recklessly modified to be dangerous and/or non compliant.

[Siwastaja]

Well, it should be possible to transform the PSU into a class 1 SMPS. I suppose that 1nF Y1 should be replaced by 2x2nF Y1 caps to earth, right ?

Do not modify products that are compliant, legal and work as they should. Any modification is not only illegal, it's also likely to destroy the EMI performance. EMC design is not easy. Because there is no problem that needs to be solved, you don't need to do anything. Just keep using the product.

[Karel]

That powersupply is perfectly fine. If you want to get rid of the leakage current, you can consider to connect the
negative output to ground/earth but that can potentially cause other problems like ground loops.
Another but expensive solution is to connect the powersupply to mains via a medical grade isolation transformer
with no overlapping primary and secundary windings and where the iron core is connected to ground/earth.

[radiolistener]

Please don't modify any mains power supplies.  The raspberry pi power supply is not broken, but now I think every other power supply you own has been recklessly modified to be dangerous and/or non compliant.

No, almost all my PSU has proper ground on the mains connector and I prefer to provide ground terminal on the mains connector rather than remove Y capacitors from PSU. For example, display and amplifier are powered with expensive branded PSU from Samsung. Both have 0 VAC on their GND output. I didn't modified it, they works in such way because they are properly designed. They have Y capacitors with proper Ground pin on the mains. This is very contrast with official RPI PSU, which has 80 VAC on it's GND output.


I know very experienced people, who repairs, teaches students and develops various interesting electronics devices for sale on the Internet and he like just to remove these capacitors from all their new devices to avoid such issues... But I prefer to use PSU with no modification, but with grounding terminal on the mains connector.

The issue with official RPI PSU is that it don't have Ground pin on mains connector and has Y capacitor between mains lines and output GND.  So there is no way to avoid dangerous high voltage current from mains through sensitive devices. Even despite the fact that this leakage current is limited with capacitor, it still dangerous and can damage devices and even can kill people.

As you can see, I already have losses due to this issue in RPI PSU. This leakage current leads to damage for my USB sound card. And an unpleasant feature is that this issue is hidden, because RPI usually is connected to display with HDMI, which provide ground through GND pin of HDMI cable so you won't notice this issue, but once you disconnect HDMI cable from RPI, there is high voltage mains potential appears on the RPI GND and can damage some device...

There is needs to be very careful if you use official RPI PSU with RPI in headless mode...

[Someone]

Even despite the fact that this leakage current is limited with capacitor, it still dangerous and can damage devices and even can kill people.

So what is the current? I'm betting it's under the limits in the international safety standards and considered entirely safe to humans. You're making up unsubstantiated exaggerations.

Safe for badly designed electronics? not so much. That small current is fully expected by properly designed equipment and diverted safely without damage.

[radiolistener]

So what is the current?

just tested it with DMM (EU version 15W RPI PSU): 89.7 VAC x 65 uA

I'm betting it's under the limits in the international safety standards and considered entirely safe to humans.

As I know, according to IEC 60601-1, there are three category of contact for leakage current safety limit:
- Type B (body): contact with an earthed patient
- Type BF (body floating): contact with a non-earthed patient
- Type CF (cardiac floating): contact with the heart or bloodstream of a non-earthed patient

Per the standard, the limits are:
- type B and BF: 100 uA under normal conditions, 500µA under single-fault conditions
- type CF: 10 uA under normal conditions, 50 uA under single-fault conditions

As you can see it fits with B and BF limit, but exceed CF safety limit for at least 6.5 times.

This is for human. But for electronics such current at high voltage can easily damage sensitive components like amplifiers, ADC, etc.

In my case it leads to damage for output amplifier of USB sound DAC. It happens when I plug 3.5 mm audio jack from amplifier powered from separate power supply with proper grounding and 0 VAC on it's GND output. Previously I do that many times, but it was done while HDMI cable from display was connected to RPI, it provide ground from display power supply. But when I do that with disconnected HDMI, the bad thing happens... 

[Siwastaja]

IEC 60601-1 is a medical device standard. Of course non-medical grade devices exceeds these limits. Get a medical grade supply if you so wish. They are not even that expensive, I remember Meanwell having a few.

[ejeffrey]

Yes, don't use the raspberry pi as for implantation inside a human body abdominal cavity or for performing open heart surgery.  I'm glad you have figured that out.  Since I'm guessing you are not a trained surgeon, that shouldn't be an issue.

You area also correct that headphone jacks specifically are poorly designed from an EMC perspective.  You can chalk it up to being an almost 150 years old design, and hey it's simple and cheap and audio isn't particularly demanding.  It doesn't follow the rule of making ground contact first, and therefore circuits using it need better input protection on the signals.  It's not hard -- audio outputs should be low impedance anyway and most audio inputs have at least a resistor to ground that can take the touch current.  A 10 kohm resistor to ground would clamp your 65 microamp touch current to <1 V.  That shouldn't damage anything.  It will provide an annoying sound, and that's one reason audio inputs sometimes buzz when partially inserted.  For handling ESD a current limiting resistor and clamp diodes would generally be required although the ESD protection inside an IC might be good enough.  All of this is pretty easy for audio signals, such a design would be a lot more problematic for high speed digital connections.

[Karel]

For non-medical devices with double insulation the limit is 250 uA according to IEC 950.

[Siwastaja]

People here just say that's fine, that's the way it is, and so on but this leakage current is a cause for concern. Period..

Not "people here", but all international standards, followed in legislation, everywhere. You are in tinfoil hat territory. Adding "periods" does not make your opinion more valid. You are free to be concerned, but 99.9999% of people are not.

[f4eru]

This is not normal, all my PSU have 0 VAC between their output GND pin and ground (they all have ground pin on mains connector).

It is absolutely normal for Class II supplies and devices with this sign to float typically at half mains voltage, with the current limited by the Y-Cap(s)


What can destroy equipment is not directly the voltage or the current, but the discharge of the ~170V-charged approx 1nF cap into the data pin of the earthed equipment you are hotplugging it to.
I recently damaged a measuring equipment this way.
If this is a problem, simply add an external ground wire to your class II equipment.

[radiolistener]

If this is a problem, simply add an external ground wire to your class II equipment.

There is no way to connect ground to this PSU. It don't allows disassemble without enclosure damage and there is no ground pin on its mains connector...