Using PE (earth) for signal cable shield?

[David Hess]

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...

It means that a third wire inside the shield has to carry the return currents.  This could be differential signaling like with twisted pair Ethernet, or a signal ground wire like with RS-232 or RS-422/485, or the inside shield of a coaxial transmission line.

I included Analog Devices application note 347 as an attachment.

[EPAIII]

Shield grounds/connections?

I fought ground problems for over 45 years. Came to ONE conclusion: what works, works and what doesn't, doesn't. You can discuss theory all week long but when the shields hit the grounds, if there are problems, you have to work out what works. And if there are no problems, then theory be damned and LEAVE WELL ENOUGH ALONE.

That being said, I have always used solid Earth ground networks in my installations. And I am not talking about the puny little conductors that electricians run to the outlets. Those are safety grounds intended to trip the breaker when an electrical power short occurs. They are not a real signal ground network.

[Nominal Animal]

I fought ground problems for over 45 years. Came to ONE conclusion: what works, works and what doesn't, doesn't. You can discuss theory all week long but when the shields hit the grounds, if there are problems, you have to work out what works.

This is precisely why I started the thread.  I can work with the theory from a physics standpoint all I want, to any level of detail and complexity I want, but that's just the theoretical part.  I need advice from experience.  Including why certain things just won't work.  (I'm just not swayed by authoritative arguments that say "do this, or it won't work".  I need to understand the reasoning behind the advice, because that's why I'm asking; I'm not trying to find out about standard practices.)

I also want it to be legal, but that's just because I am personally not experienced enough to determine when it is the rules and regulations that don't make sense.  However, as I've read the Finnish electric safety laws and such, it seems that this kind of stuff is legal if a sparky with the correct credentials accepts it, and it works in practice.  (Also, I'm working with very low voltages and currents, just a few watts at below 20VDC or so.)

The snubber network between shield and ground is a good example of why experience matters.  In theory, it works well.  Many say it doesn't, but most haven't elaborated exactly why.  I've looked for specifics, and read from others exactly what kind of ground bounce and EMC issues they can lead to in different circumstances in practice, and from those, now have a much better understanding of the difference between theory and practice here.  I cannot stress enough how useful it is to hear/read and understand why, the practical reasons and experience, rather than just what one should do.

[Andy Chee]

I also want it to be legal, but that's just because I am personally not experienced enough to determine when it is the rules and regulations that don't make sense. 

You must obey the law no matter if the rules and regulations make sense or not.  That's how the law works! 

You don't have the luxury to determine which laws you can ignore, you must comply with everything!

[Nominal Animal]

I also want it to be legal, but that's just because I am personally not experienced enough to determine when it is the rules and regulations that don't make sense. 

You must obey the law no matter if the rules and regulations make sense or not.  That's how the law works! 

You don't have the luxury to determine which laws you can ignore, you must comply with everything!

Most legal systems have cruft and silliness because nothing is static and everything changes as time progresses.  Just look at cable color coding rules.  Not all laws and regulations are actually enforced, especially when they are changing because of such cruft.

For example, in the state of Florida (and several others) in the USA, consensual sodomy (anal sex) is against the law.  Nobody can or will be prosecuted, however, because of precedents (Lawrence v. Texas, US Supreme Court 2003).  So, technically, it is against the law, it just isn't punishable, nor will any cops investigate anything related to breaking that (assuming all parties agree it was consensual, of course).

Similar silliness might be regulations allowing non-electricians to connect say ceiling lights to mains using screw terminals (chocky blocks), but not to Wago connectors, because of the wording in the law/regulation hasn't caught up yet.  (The latter are safer.)  Thus, it is about understanding how the relevant laws and regulations are changing in the near to medium future, not about choosing what laws and regulations to obey and what to ignore.  Stuff that capable EEs and electricians should be aware of, in my opinion.

[Siwastaja]

I fought ground problems for over 45 years. Came to ONE conclusion: what works, works and what doesn't, doesn't. You can discuss theory all week long

The key is understanding the theory correctly. Then it will match reality and save from endless problems.

Problems:
* Understanding EMC and SI is not trivial. It really is a complex topic;
* This complex topic is made more difficult by a lot of poor quality pseudo "information", written in good intention but causing more harm than good.
* Lot of unnecessary and wrong "rules of thumb"

"What works, works" is probably more useful than incorrect theories, but correct theoretical knowledge is absolutely necessary because figuring out EMC/SI by experimentation only works in easiest cases and is too expensive. There is no way you could test out 1000 different combinations of wiring arrangements in an EMC lab where every test costs you say $1000. So you need expertise, you need theoretical knowledge. Testing is then verification of that theory. If theory is sound, failures in verification step are rare and small. If the gap between theory and practice becomes wide, you are listening to the wrong experts.

[Siwastaja]

You must obey the law no matter if the rules and regulations make sense or not.  That's how the law works! 

You don't have the luxury to determine which laws you can ignore, you must comply with everything!

This is just blatantly incorrect. It is completely normal business practice to even outright ignore laws, or apply interpretation which not everybody agrees with. Of course this requires care as it presents a risk to the business. Finally decisions are made in money, if a decision makes more money than fines paid, then it is doable.

Individuals do the same, for example choose to drive faster than the speed limit and say "sorry" or pay fines if caught.

Besides, in 90% of cases when someone on a forum claims that "law says exactly X", what the law actually says is more complex than that.

[KaneTW]

EMI is inherently a complex topic, so any blanket statement is questionable.

That being said:
* PE is a separate thing than shielding. It's intended to make sure that fault currents or short circuits do not cause dangerous situations.
* PE is almost always tied to chassis ground, if the chassis is metal. Exceptions exist but are rare and not something you'd do without good reason.
* The (metal) chassis functions as a shield for the device, and if feedthroughs/connectors are properly laid out the connector shields will be tied to chassis ground as to not result in gaps in the shield.
* Shields should be tied at both ends unless there's exceptional circumstances. Any ground loop current should be handled by ensuring a sufficiently low-impedance connection between the grounds.
* That low-impedance connection often ends up also functioning as PE in e.g. industrial automation

[snarkysparky]

Yeah the only real way to approach the problem without toooo much guesswork is to draw a complete schematic of the system.

OP has done this mostly but need to add noise sources and locations.

Including

* noise sources at all likely points.
*coupling methods to target circuit points both inductive and capacitive
*add caps and ind for the parasitic elements
*proposed shields with parasitic coupling to them from the noise sources and coupling from the shields to the target circuit.
* try to consider transmission line effects on cables that are long enough to worry about.

what else ??    I suspect there is more.

This is the hardest part,   well maybe,  gotta solve this nasty circuits problem now

[Andy Chee]

You must obey the law no matter if the rules and regulations make sense or not.  That's how the law works! 

You don't have the luxury to determine which laws you can ignore, you must comply with everything!

This is just blatantly incorrect. It is completely normal business practice to even outright ignore laws, or apply interpretation which not everybody agrees with. Of course this requires care as it presents a risk to the business. Finally decisions are made in money, if a decision makes more money than fines paid, then it is doable.

Individuals do the same, for example choose to drive faster than the speed limit and say "sorry" or pay fines if caught.

Besides, in 90% of cases when someone on a forum claims that "law says exactly X", what the law actually says is more complex than that.

Ignoring rules & regulations is what Boeing did.  Two crashed 737 MAXs and a loose door plug later.  Money may temporarily solve the problem, but only if your revenue stream is greater than the cost of the mistakes you make.

Then there is professional liability insurance for engineers.  When a civil engineer designs a bridge, and the bridge collapses, the engineer's insurance will pay for the damage and deaths (unless the court proves that the engineer was untrained and incompetent).

Finally you have entire countries breaking the law, such as China copying products.

In summary, if you break the law, you are being unfair to others.

[T3sl4co1l]

Or financebros short selling various stocks dry, among other shady practices; but hey, if the FTC isn't going to police them, or at worst levies a modest $20M/infraction or whatever, while turning billions in profit -- that's basically saying go ahead, please take their money, give us the tiniest cut and be on your way, right?

(This has been your daily reminder that law is not moral, and the two align only on accident, or in the specific few cases where we have fought for and demanded such.)

Tim

[Nominal Animal]

I'm realizing now that if I ignore the fact that I use a switchmode supply, and instead treat it as a standard mains-powered exposed metal chassis device –– with the caveat that I do not have live and neutral, only two AC inputs ––, I can find literature describing how this is best done and why.  Including when the device is split into multiple enclosing chassises via shielded cables.  Quite standard stuff, really.



(It's a bit humbling realizing how, uh, silly the question must sound to those who work with such things daily, and thoroughly understand these.  All I can say, I've learned a lot from the discussion here.  I truly appreciate the information, everyone.)

There are some details I still need to understand much better, for example how to limit the noise from the switchmode supply to the load when inside the same enclosing chassis –– not just conducted, but radiated and maybe even magnetically coupled ––, but I'll try learning about that from the well-known sources (several mentioned in this thread) on my own first, before bothering others.

One point about the DC GND: when working with USB hosts that have proper grounded power supplies and DC ground voltages at or very close to PE potential, and USB devices that use those wall warts that cause their DC GND to float at couple of hundred volts above or below that, I really like it when I can choose the ground potential for the kind of circuits I'm talking here, by connecting a ground before powering the circuit.  It has been an useful tool.  (Some of you might remember a thread about USB isolators before.) I hope that explains why I have been so reticient at tying the DC GND to PE here: didn't want to discard an useful tool too early/unnecessarily, that's all.

[T3sl4co1l]

Yep, isolation is a useful tool -- not just in galvanic terms (opto/logic isolator, transformer, etc.) but it works for EMI too, within reason.

Suppose you have two boxes, joined by a shielded cable; if the shield is solid, that's all that matters, signal quality is maintained across them.

Suppose the shield is cut in the middle, going for some length unshielded, with the cut joined by a wire; a voltage drops across this wire, depending on frequency relative to its length.  Some of this voltage couples into other wires in the cable, but only partially; the coupling factor for wires in a multiconductor unshielded cable might be... 0.5 to 0.9 say, ballpark, but depends obviously on geometry.  Noise on one reference plane, with respect to the other, couples only partially into the signals; as seen from the other, there's a noise source in series with the signals.

If we put a ferrite bead around the cable, we improve coupling between shield wire link and the other signals.  The shield currents likely also decrease (min(|Z_CM|) (as measured at EMI frequencies) has increased), further improving CMRR.

Or, on a PCB, we might have a slot between two reference planes.  It is acceptable to route signals between these planes if we have a CMC which couples the plane-to-plane voltage into the signals -- thus decoupling the signal currents between planes, and maintaining noise immunity.  I showed an example of that here,
https://electronics.stackexchange.com/questions/686009/what-is-the-best-practice-when-replacing-a-ribbon-cable-ferrite-clamp-with-chip/686238#686238

Tim

[Siwastaja]

live and neutral, only two AC inputs

A side note (because I know you are interested in correctness!), line and neutral are both wires that have attribute live attached to them. So the two AC inputs are two live wires, but you don't know which one is line and which one neutral; and even when you think you know (e.g. in countries with polarized plugs), you are still not allowed to assume. Therefore neutral is always live.

The main distinction between PE and N is that with PE, you are kinda guaranteed it is not live. (There are exceptions to even this such as those super scary neutral faults.)

[m k]

Take a machine of few tens on meters long or wide.
It's big enough that it usually have different independent parts.
Usually metal framed also.

If you go circling around that machine you'll find that all around are thick Y/G wires connecting metal frames together.
Then its many 3-phase supply lines are nothing special, just normal rubber cables here and there.
The installation has made its own earth.

Maybe there are also few different thick walled cabinets for different electronics.
From there you may see leaving some thick plastic tubes.
If those tubes have a flat surface it's likely that inside of that surface is a shielding, then single wires for normal logic connections.
It's also possible that they are just cable collecting wrinkle tubes, but then there are no single wires, only shielded cables, except one, that thick Y/G wire.
So own earth is maintained and external world is shielded out.

Black magic starts if own earth is not present and distances are long enough, that's "earth long" not meters long, then the "long earth" must be cut.
Some current here and there may not be a problem, maybe it's even disturbing all signals equally, but single ended driver may feel it differently, not the current, but the potential.
Actually it's not even the potential that is a problem, its balancing route is.

Flat ribbon cable with flat external shield, it's clearly shielding against external world.
But it also has another function, it's damping noise between individual ribbon wires, that can also be its more important property.

One more thing for that black magic.
Nowadays we don't see open wire telephone lines anymore, so spark gaps may be out of the thought process.
Back in the day when dumb terminals were the norm the rule was that wall side, or server side of the shield is grounded.
Nowadays it can be different when both ends of the shield are floating.

[Nominal Animal]

Take a machine of few tens on meters long or wide. [...]

Yep, exactly how I was thinking.  Just because my voltages are lower and currents much smaller and in a sense I'm inverting inside and outside compared to say a CNC mill with 3-phase power, does not make the situation fundamentally different.  And that is well documented as to what is necessary and/or useful and why.

Flat ribbon cable with flat external shield, it's clearly shielding against external world.
But it also has another function, it's damping noise between individual ribbon wires, that can also be its more important property.

I'm old enough to remember the time when IDE (parallel ATA nowadays) became fast enough to switch from 40-wire to 80-wire cabling, with every second one a ground, to reduce crosstalk at the higher frequencies used.  Soon, flat cables were bunched in the middle, and even shielding started to be used, even though the insides of a desktop computer enclosure aren't that noisy electromagnetically speaking.

(I did use SCSI cabling too, especially with servers and Macs (pre-OS X), but I don't think I ever split an external one open to examine the construction; they were both too robust and expensive to waste that way, and I don't think I had a single one fail on me.  On the servers prior to the turn of the millenium, I used flat ribbon SCSI cables.)

Looking at history, and examining the reasons for the solutions they arrived at, is always useful in my opinion.  The solutions are usually not the optimal ones, though, because we're human, and technical superiority has traditionally not been a main market driver.

I find it helps predict what one doesn't yet know, like the use and purpose of guard traces for example.