Using PE (earth) for signal cable shield?

[T3sl4co1l]

Mains is normally neutral grounded at the panel, give or take local variants.  Or legs of a three-phase circuit, in which case both are hot and we'd cure the error by showing neutral to earth and three voltage sources radiating from it, two of which go to the transformer/PSU.

Either way, I would safely assume that connection is implied to be long -- wiring back to the panel, galvanic purpose only.

Tim

[madires]

It does not matter whatsoever that some current will flow through the shield. Heck, with USB powered devices in most cases there will be more current flowing through the shield than that. USB-C standard demands that cables have GND pins/wires connected to the shield at both ends. The vast majority of micro USB cables also have GND wire and shield connected together although AFAIK it's not explicitly demanded by the standard.

Have you considered the EMI suppression caps in the SMPSU wall warts?

[Siwastaja]

I've got a question about this diagram. As drawn, one side of the mains supply is connected directly to the protective earth. Isn't this a recipe for a big bang? Or at least, to trip the RCD.

I think the drawing depicts a complete equivalent circuit - one side of the mains supply is connected to the protective earth; where exactly depends on the power distribution system, for example here they are typically connected at the main circuit breaker box (in comes PEN wire, out go neutral and PE wires).

[Nominal Animal]

I've got a question about this diagram. As drawn, one side of the mains supply is connected directly to the protective earth. Isn't this a recipe for a big bang? Or at least, to trip the RCD.

The diagram describes the entire system, as drawn in various sources discussing earthing and shielding.  The idea is to show the path to protective earth for any external sources of electricity, rather than allow it into the shielded circuitry.  The supply is not fully drawn either, it's just a simplified diagram.

Another question: what is the red 'Y' capacitor for? I'm trying to imagine why you would couple the secondary back to the primary via a capacitor.

It represents the class Y capacitor in class I isolated supplies, where it is supposed to be between the output and neutral.  However, we have no neutral here; the sockets are not polarized at all.

In a real switchmode supply, we have two full bridge rectifiers.  The first one rectifies the mains AC.  The class Y capacitor is typically connected here, between the output and the rectified mains voltage.  This is what leads to the DC GND having such a big voltage difference compared to "neutral" or PE.  The rectified voltage is then chopped up, transformed, rectified, and filtered.  There is usually an optoisolator or similar to control the duty cycle, to keep the output voltage within the desired range.

These are the four switchmode AC/DC supply types I know of, stolen from Mean Well datasheets, in block diagram form:

^{(click to embiggen)}

First three are class I. The first one has a class Y capacitor on the output, between DC GND and PE/neutral.  The second one has some kind of capacitor between the output and the mains side circuitry, as part of PWM voltage control.  Third one has a class Y capacitor on the mains side only.

Fourth is class II, and has no connection to PE, or even a functional ground ("neutral").

My simplified diagram tried to describe all four in a simplified fashion, similar to how these are shown in the references on grounding and shielding.
To draw the red class Y capacitor right, I'd have to add the first bridge rectifier at minimum, but I wanted to keep it as simple as possible.

[SteveThackery]

Thank you, @Nominal Animal - that's a fantastic explanation.

I noticed a small difference in the symbols:

[Nominal Animal]

Aw crap, I forgot: protective earth has that circle around the symbol.  My fault, I apologise for the confusion!

[SteveThackery]

Don't apologise - it's me who messed up!

So what do those two symbols mean? The one with three parallel lines, or the one with four slanting lines?

[wraper]

It does not matter whatsoever that some current will flow through the shield. Heck, with USB powered devices in most cases there will be more current flowing through the shield than that. USB-C standard demands that cables have GND pins/wires connected to the shield at both ends. The vast majority of micro USB cables also have GND wire and shield connected together although AFAIK it's not explicitly demanded by the standard.

Have you considered the EMI suppression caps in the SMPSU wall warts?

Again, it's not some ultra precision measurement or audio equipment where you should be concerned about current in the shield. As of potential difference, solid shield connection that makes the contact first is important to prevent that voltage getting into data lines. Once you break that connection with voodoo nonsense, you remove that safety net during hot plugging and greatly increase the chance of damage.

[Nominal Animal]

This is the corrected block diagram, with the correct protective earth symbol:

^{(click to embiggen)}

The symbol I erroneously drew into the diagram in the question is normally used for chassis or enclosure.  I intended to draw the entire ground as a PE, as is common in grounding and shielding references, using a line with slanted hashes underneath to indicate solid ground, but decided it looked silly so simplified it –– and although I definitely know it, I just didn't notice I ended up using the confusing chassis symbol for the protective earth! 

My sincire apologies to anyone who looked at the original block diagram and considered the bottom left symbol as a chassis ground; it was supposed to be protective earth.

(These symbols are standardized in IEC 60417.  The standard ground symbol is #5018.  Noiseless earth, #5019, has a downwards-facing half circle above the ground symbol, cupping the symbol, or acting like an umbrella above the ground symbol.  The above protective ground, a ground symbol with a circle around it, is #5019.  The wrong symbol in the initial post is #5020, indicating frame or chassis.)

[m k]

by TDK.

[Nominal Animal]

by TDK.

Yes: my use of class I and class II do not actually refer to the class requirements, but how those requirements are implemented in currently available low-voltage low-current switchmode power supplies.  Apologies if that annoys or confuses, but I don't know a better way to identify the four typical ones I've seen as I described above.  If there are better/more correct terms I should use, please do let me know; I'm just a hobbyist, after all!

[m k]

Just a supplementary.

[Nominal Animal]

Let me restate my thinking on this a bit further, in case it helps.

Even if not connected to anything, the continuous shield – including cable shielding not electrically connected to the electronics – acts as a Faraday cage, as explained in the video JohanH linked to in reply #2.

As the shield shape is complex, it is possible for external electromagnetic fields to generate a current in the continuous shielding; for example, if it has a loop in it.  (Such loops typically involve a human directly or indirectly touching two different parts of the shielding, or putting two metal enclosures connected with a shielded cable on top of each other.  So not by design, but by practical situation.)

However, voltage can still electrostatically couple to the shielding, and in the case of AC sources, cause current flows in the shielding.  ESD events can charge the shielding to rather high potentials.  The shield having varying charge distribution in/on it will generate varying electromagnetic fields, including some inside the shielding, thus allow some unwanted noise "through" the shielding.

Fixing the shielding to a specific potential, preferably close but not necessarily exactly local true earth (to minimize human discomfort from ESD zaps), keeps the charge distribution as flat and static as possible, minimising any kind of electromagnetic or electrostatic noise passing "through" the shielding.

With respect to USB and others, I'm not interested in how standard cables are required to be constructed, I was only referring to the types of low-level, low-current signals used inside the enclosure.  I could not use typical USB or Ethernet connectors anyway.

Most importantly, I do not intend having multiple grounds in the electronics inside the shielding.  The shielding is not connected to the innards at all, so the only ground within the shielding is the supply ground.  (In that sense, it is similar to doing your electronics projects inside a shipping container, which happens to be grounded.)  In the case of low-leakage class II supplies, the supply ground is not tied to any specific potential, so it might be sensible to tie it to PE or something somehow, perhaps class Y capacitor or some kind of safety circuit, I'm not sure.  There may be electrostatic fields inside the shield (between the shield and the electronics inside), but that's fine, as long as it doesn't oscillate or vary too much.  We already have one everywhere around us, typically on the order of 100 V/m vertically in open areas, because of atmospheric electricity.

It seems that to verify whether it is safe or not, I'll have to make it first and then have a qualified electrician check it over.

I don't like the blanket statements "you need to do it this way" without explaining why, because if I just wanted a solution I'd buy one and be done.  I'm doing this at least as much to learn and advance my own understanding than achieving a specific outcome.

[Andy Chee]

This is the corrected block diagram, with the correct protective earth symbol:

^{(click to embiggen)}

Isn't this fundamentally the same question that often comes up;

"my DVD player only has a two pin power cord.  Can/should I connect the metal case of the DVD player to protective earth?"

[Nominal Animal]

Isn't this fundamentally the same question [...] "Can/should I connect the metal case of the DVD player to protective earth?"

I don't think so, because I'm asking about shielding against EM noise, both electromagnetic (radiated) and electrostatic (capacitively coupled), and mitigating the effect of ESD to the electronics inside.  Situation is similar, but the question is different.

[Nominal Animal]

I created a new block diagram of the various switchmode supplies, with (simplified) differences between typical implementations and my musings in color.  Since the load within the shielding doesn't really matter, I simplified that too.  The switcher and any power factor correction is hidden in the SW box.  Although there usually is optoisolated feedback from the DC side to the mains side, and perhaps even regulation on the DC side, I omitted those because they're not relevant here.

^{(click to embiggen)}

The four groups of differences and options, from left to right:

Leftmost (green): Typical EMI suppression class Y capacitor on the mains side.  I don't care whether this is used or not, as long as the supply is safe and legal.

Second from left (red/pink): EMI suppression capacitor between DC GND and "neutral", except here in Finland (and in most of Europe using CEE 7/3 "Schuko" grounded sockets and CEE 7/4 plugs) mains sockets are unpolarized, and there is no guarantee one of them even is "neutral"; they're both considered live.  This is the one that causes DC GND to not be anywhere near PE or true ground potential, and may cause "tingles" when touching e.g. USB cable shield.

Third from left (brownish), below the second full bridge rectifier: Tying DC GND to PE potential.  In this case, the shield is connected to DC GND and PE (short in the fourth group).  Above, I've described my issues with this (ground bounce at ESD events), when the transmission line to PE is long.

Rightmost (purplish blue): Connecting the shield to PE only, either directly (topmost; not my preferred option), or (my preference), via a capacitor and/or resistor for smoothing out ESD spikes.  As I understand it, a snubber network shouldn't affect the operation of the shield much, but it would mean that in case of an electrostatically charged human touching the shield, they wouldn't feel a zap.


Some of the members in this thread have stated that a snubber network (resistor and capacitor in parallel) between the shield and whatever potential it is tied to is categorically wrong and would not work.  From a physics perspective, there is no reason why it should be wrong or why it would not work, so if there is such an engineering reason, I'd like to know.  And I do mean other than "this is how these cables must be done to comply with the standards" or "everybody does it this way" or "because it is standard practice to do it this way", because I'm not making a product or installing this for others, just wanting to learn and find out if this would work (better than current solutions), and at most build and use it for myself for no real reason except this is a hobby.  I do have a physics background, and I'm familiar with EM fields and related theory.

[Andy Chee]

Isn't this fundamentally the same question [...] "Can/should I connect the metal case of the DVD player to protective earth?"

I don't think so, because I'm asking about shielding against EM noise, both electromagnetic (radiated) and electrostatic (capacitively coupled), and mitigating the effect of ESD to the electronics inside.  Situation is similar, but the question is different.

What about Japan?  By far the majority of household mains outlets are 2 pin only (over 95% of outlets). 

Does Japan have any increased issues or problems with EM & ESD, compared to countries with protective earth?  I've never heard of any issues.

[T3sl4co1l]

I created a new block diagram of the various switchmode supplies, with (simplified) differences between typical implementations and my musings in color.  Since the load within the shielding doesn't really matter, I simplified that too.  The switcher and any power factor correction is hidden in the SW box.  Although there usually is optoisolated feedback from the DC side to the mains side, and perhaps even regulation on the DC side, I omitted those because they're not relevant here.

^{(click to embiggen)}

Ok, so you've established pretty well I think that these circuits are with respect to DC/mains/galvanic grounding connections.

Then for EMC purposes, we are safe to assume all such connections outside the shield are spaghetti, and thus subject to unlimited external fields, and will offer potentially unlimited impedance (albeit only at resonant frequencies) between nodes in the circuit.

Is that correct?

When do we get to the EMC part then--?  Or am I jumping the shark (well, "have I been"), and you (and/or others) have just been nailing down the galvanic equivalent circuit precisely here?  (Which seems confident now, so, perhaps "next" is the answer?)

Tim

[David Hess]

And if that's done, we collapse one connection:

(Image attachment broken in quotes, see below)

now the signals are always referenced to the ground plane / cable shield, and no ground loop flows so it's good at all frequencies.

The earth connection isn't needed at all (probably) if a double-insulated supply is used, but the circuit-GND can still be connected to shield in the same way.

The problem with that is that it makes a ground loop, and now the shield current couples into the signals that it is protecting.  Differential signals in things like USB should be able to ignore this, but not all differential signals can.

There are situations where EMC compliance requires the shield to be grounded at both ends, while signal integrity requires the shield to be connected at only one point.

Does anybody remember Ethernet Thicknet and Thinnet?  Grounding twisted pair Ethernet at both ends may lead to dangerous hilarity if there is a difference in ground potential, which is possible over Ethernet distances.

[T3sl4co1l]

I wonder how much the usage of the word "ground" has changed over time.  Maybe we disparage old references (like "only ground one end") because they specifically meant safety earth, not local circuit ground; and we've reinterpreted it to mean local circuit ground, and much wailing and grinding of teeth ensues.

There are situations where EMC compliance requires the shield to be grounded at both ends, while signal integrity requires the shield to be connected at only one point.

You'll have to provide a reference for that, though. Signal integrity demands the reference plane be extended continuously from end to end.  Which one is meant, is ambiguous: is the shield not actually the reference plane? is there another reference conductor in the cable? is the cable multi-shielded? is the signal differential? etc.  That makes a big difference, whether this statement is simply acceptable, or utterly reckless...

Tim

[Nominal Animal]

As I understand it, EMC (electromagnetic compatibility) is the desired outcome: not being affected by electromagnetic noise in the local environment.  Sometimes EMC is meant only in the physical sense, sometimes it includes the local regulations.  It is not clear to me when.

EMI (electromagnetic interference) is the umbrella term for the unwanted effects of electromagnetic noise on an electrical circuit.  There are four types of electromagnetic interference, or four types how noise can couple to the the electrical circuit:

• conductive, directly via wires
• inductive, via varying magnetic fields to conductive surfaces
• capacitive, via varying electric fields to conductive surfaces
• radiative, via EM radiation to antenna-like structures

Then for EMC purposes, we are safe to assume all such connections outside the shield are spaghetti, and thus subject to unlimited external fields, and will offer potentially unlimited impedance (albeit only at resonant frequencies) between nodes in the circuit.

Well, no, because that is kinda-sorta part of the question: where should the shielding start? Should the (entire?) supply side be covered by the same/continuous shield?  Would a break in the shielding after DC filtering before the "sensitive" circuits help?

You see, I can freely choose between different types of switchmode supplies, and decide whether they're within the shield or not, or perhaps separately shielded, but I don't want to build it from scratch, because I don't like dealing with voltages in excess of 20 volts or so.  I want to add a filtering and post-regulation stage, too, dropping a couple of volts (to say 5V at say max. 1A load), getting a nice clean DC output.

The secondary questions – after the initial shielding connection question – involve whether the SMPS should it be within the same shielding as the switchmode supply, or unshielded, or within the device shielding?  I believe the answer is dictated by the practical behaviour of noise currents in SMPSes, depending on they're configured wrt. EMI suppression capacitors, so I wanted to show all my practical easy options above.  It would be pretty funny to make an enclosure which generates a lot of EMI and traps it inside itself, blasting the circuits with more EMI inside than outside.. but not practical.

We all know how standard and compliant circuits and wiring is done, I'm not looking for that.  This is something I haven't seen before in practice, so I want to know why, specifically the reasons why it is not done in practice.  The reasons could vary from "it's not how we do stuff" to highly specific engineering or manufacturing reasons to legal/code limitations.

For now, we can leave the load as a simple black box.  For wiring between "modules", I do intend to use connectors and cabling where the shield is continuous and not easily accessed inside the shielded environment.  While this does mean transmission line behaviour in the shield, it is that equally for the DC supply rails, so I'm ignoring that part for now as unavoidable.

]When do we get to the EMC part then--?

Other than the switchmode supply, although the load is digital circuits, it is low voltage and low current, and doesn't radiate much (although I have no equipment to measure that), as it is just your typical run of the mill microcontroller stuff.  So, I'm not worried about emissions inside the shielding.  I might even stick the MCU inside a secondary metal can, connecting the can and related shielding to the DC GND, if it is a problem.

Outside the shielding is either a normal household environment, or a nasty workshop environment with brushed motors and sparky welding equipment within a hands reach.  The latter is enough to occasionally glitch a microcontroller running in free air.  This, because workshop shenanigans and DIY microcontroller gadgetry as soon as I have my own workshop.

So, we can consider the shielding to exist solely for the outside nasty environment, plus an occasional ESD event due to a highly charged human touching the shielding, which I don't want to generate a ground bounce within the shielded circuitry.

[Nominal Animal]

[In a ground loop] the shield current couples into the signals that it is protecting.  Differential signals in things like USB should be able to ignore this, but not all differential signals can.

It is a problem here when connecting desktop computers via USB to devices powered by doubly-insulated supplies (wall warts), because the computer supplies do have a PE connection and the host USB GND is typically at (or close enough to) local ground potential, but the doubly-insulated wall warts' common voltage differs by a couple of hundred volts because of the unpolarized mains connectors and EMI caps between DC output and mains.

I even started a thread about USB isolators in hobby designs because it is such a common issue here and I deal with it often, but most members considered my solution wrong/improper/utterly silly.  The question posed was why would I even use a wall wart and isolators, and not a proper lab supply which would be safer too.

Sometimes it is difficult getting into the meat of a question about what is feasible and how and why, rather than how it is normally done.

[m k]

I think we need at least 5 shields.

1st is the one against a sword, so a general shield, like a casing that has nothing to do with a signal.
3th shield is an inner shield of a waveguide, so pretty close to signal but still just pretty close.
2rd is a shield not so close to a signal of a waveguide.
4th is a 1st shield of a coaxial cable.
5nd is a plate, a part of a signal that has also shielding properties.

Hmm..the above is not defining the direction of shield.

[Nominal Animal]

Which one is meant, is ambiguous: is the shield not actually the reference plane? is there another reference conductor in the cable? is the cable multi-shielded? is the signal differential? etc.

While that was a rhetorical/descriptive question about the standard instructions related to this situation, I'll point out anyway that in my case, the shield is not the reference plane, there is a dedicated reference conductor within the shielding, I can use multi-shielded cabling if useful (with secondary shielding "grounded" to local reference plane according to common practice), and generally the signaling is single-ended.  There will be some cables that are partially multi-shielded; I can choose between individual inner shielding, inner shielding for each twisted pair, and inner shielding for a set of non-differential but related signals, depending on what is needed.

While I am the only one asking about this currently, this is a relatively common problem, so exploring the possible solutions ought to be useful for a wider audience, especially if it can be distilled into "you can do this if you also do that but make sure you don't do the other thing" for hobbyists to apply.

[nctnico]

And if that's done, we collapse one connection:

(Image attachment broken in quotes, see below)

now the signals are always referenced to the ground plane / cable shield, and no ground loop flows so it's good at all frequencies.

The earth connection isn't needed at all (probably) if a double-insulated supply is used, but the circuit-GND can still be connected to shield in the same way.

The problem with that is that it makes a ground loop, and now the shield current couples into the signals that it is protecting.  Differential signals in things like USB should be able to ignore this, but not all differential signals can.

Yes and no. If you have HF coupling into your shield, then you'll likely won't pass EMC testing (radiated and/or conducted). So you'll need to fix your circuit so it doesn't couple HF currents into the shield. In most cases a power supply is the primary source of HF currents into a shield and / or ground. So typically I design the power supply to have filtering on at least plus but ideally on both plus / minus lines using common mode and differential mode filtering. Then the secondary side can be connected to shield + circuit ground.

But to circle back to the question through a practical example: years ago I designed a device for use in HV labs. In order to protect it from the outside, I choose to leave the ground floating in respect to the chassis for some parts of the circuit which connect to the outside. Much like Nominal Animal sketched in reply #9. That went well until they placed the unit on top of a pole at 65kVAC and started drawing arcs. It would misbehave every now and then. Connecting the circuit ground firmly to the chassis ground fixed that problem. Somehow drawing the arcs produced enough energy to create a potential across to board. With the board connected to the chassis, everything is equalised (even though the entire unit isn't connected to ground obviously).