cable shielding -- connect to ground on either or both ends?

[guenthert]

When a cable is shielded, is one to connect either (which?) end to ground or both?  Seemingly there are proponents for either.  It has been stated that connecting the shield to ground only on one end creates an (clearly undesired) antenna.  But if galvanic separation is desired (either due to elevated potential on one side or in order to avoid earth loops), connecting both to ground is not possible.  How about connecting it (at at least one end) using a (low-capacity) capacitor?  That allows for separate potentials (DC and - perhaps generally more important - at power line frequency), while avoiding the "open end" antenna scenario for high frequencies.

With non-ideal capacitors, there won't be perfect galvanic separation (so perhaps not suitable for electrometers), but even not-terrible elcos have leakage current in the nA range at say 10V potential difference, so >1GOhm 'effective' DC resistance.  PP or ceramic capacitors should be much, much better and quasi-ideal in this context for all, but the most demanding applications.

Does this makes sense?  If not, why not or what are the shortcomings?

[TimFox]

Are you asking about a coaxial cable (one wire inside shield), or a shielded twisted pair (two wires inside shield)?

[Whales]

But if galvanic separation is desired (either due to elevated potential on one side or in order to avoid earth loops), connecting both to ground is not possible

N.B. disconnecting the shield doesn't give you galvanic isolation.  The centre wire is still not isolated.  You need something like a transformer or optical isolator for that (which also incidentally fixes the ground loop).

How about connecting it (at at least one end) using a (low-capacity) capacitor?  That allows for separate potentials (DC and - perhaps generally more important - at power line frequency), while avoiding the "open end" antenna scenario for high frequencies.

Yes that could work.  Unfortunately in modern environments 50/60Hz is only one component of the interference you need to deal with, so it would only be an incremental improvement over directly connecting the shields.

[Whales]

Are you asking about a coaxial cable (one wire inside shield), or a shielded twisted pair (two wires inside shield)?

One of the best solutions of course: use a balanced signalling pair inside a shield, rather than a single unbalanced wire with (semi)-return shield.

[guenthert]

Are you asking about a coaxial cable (one wire inside shield), or a shielded twisted pair (two wires inside shield)?

     I was thinking of multi-conductor shielded cables, but I fail to see how that is relevant.

[guenthert]

But if galvanic separation is desired (either due to elevated potential on one side or in order to avoid earth loops), connecting both to ground is not possible

N.B. disconnecting the shield doesn't give you galvanic isolation.  The centre wire is still not isolated.  You need something like a transformer or optical isolator for that (which also incidentally fixes the ground loop).

Well, yes, of course, the other conductors musn't defeat the galvanic separation.  I might have clarified that.

How about connecting it (at at least one end) using a (low-capacity) capacitor?  That allows for separate potentials (DC and - perhaps generally more important - at power line frequency), while avoiding the "open end" antenna scenario for high frequencies.

Yes that could work.  Unfortunately in modern environments 50/60Hz is only one component of the interference you need to deal with, so it would only be an incremental improvement over directly connecting the shields.

   I don't quite follow.  Here, 50/60Hz was considered 'DC' (well, sufficiently low frequent that the capacitor would effectively block such currents).

[TimFox]

Are you asking about a coaxial cable (one wire inside shield), or a shielded twisted pair (two wires inside shield)?

     I was thinking of multi-conductor shielded cables, but I fail to see how that is relevant.

See reply above by Whales.
Normally, you need to connect two conductors to get a voltage from one point to the other.
If the shield is added to a pair, then you have options on where to connect the shield to the chassis ground at one or both ends.
If a coax, the shield must be grounded at both ends, or you need another conductor to connect the "low", "common", "return", or "ground" terminals to each other.

[RJSV]

I've read claims, saying a twisted pair (inside a shield) would help 'isolate' the pair against lower frequency cross-talk, while shield is more effective at blocking higher frequencies.  I would (guesstimate) that means 200 hz cross-talk is better reduced by twisted pair, while 3000 hz electrical noise is better reduced by the shield, and so the combo gives all-around block.
That could apply, somewhat, like mentioned above, to something like having six twisted pairs, all inside a single outer ground shield.  That would be your local phone company, using those 3/4 inch diameter weather resistant multi-line cable.
   I've seen, (owned) a jumbo thick cable having 14 pairs, each with individual shields, PLUS that jumbo outer ground shield, and thick thick vinyl jacketing.
   Audio stuff generally I'd only ground a shield, at input end (guitar pedal), but that's with return path normally via the 'grounded' wire in each twisted pair.
   NOT ANALOG EXPERT, however. Thankyou

[CaptDon]

It was often said with one twisted pair inside a shield and used for 'balanced' signalling like a microphone that the shield should be connected only at one end. The truth is undeniably "What ever works in a given situation is best". I have had cases where connecting the shield at both ends creates a ground loop and more often where NOT connecting shield at both ends is so horrible I can hear local AM radio stations being detected in the audio chain. And don't forget about the buzz those old light dimmer packs caused!!!! Then there is the ongoing debate with XLR connectors, does the shield go only to pin 1 or also to the shell? Many pieces of audio gear have a 'ground lift' switch to divorce local ground from the pin 1 (shield) but that becomes a problem if the shield is also wired to the shell!!! I don't connect to the shell (normally). My worst nightmare as an audio engineer out in a venue is when the mixer/fx rack is powered from somewhere mid venue and the amp racks are powered from the stage area and there is up to 1 volt a.c. of horrible harmonics and noise when measured between the ground of the mixer rack and the ground of the amplifier rack. In that case I run a 3KVA isolation transformer to supply the mixer rack with the PE/Ground NOT carried through the isolation transformer connection. As we are moving away from snake cables ground problems are also going away!!!

[David Hess]

It can be a tough problem.

Analog Devices application note AN-347 - Shielding and Guarding

Analog Devices mini tutorial MT-095 - EMI, RFI, and Shielding Concepts

[Whales]

How about connecting it (at at least one end) using a (low-capacity) capacitor?  That allows for separate potentials (DC and - perhaps generally more important - at power line frequency), while avoiding the "open end" antenna scenario for high frequencies.

Yes that could work.  Unfortunately in modern environments 50/60Hz is only one component of the interference you need to deal with, so it would only be an incremental improvement over directly connecting the shields.

   I don't quite follow.  Here, 50/60Hz was considered 'DC' (well, sufficiently low frequent that the capacitor would effectively block such currents).

Loop-coupled interference at lower frequencies will be blocked by your capacitor but loop-coupled interference at higher frequencies will be let through by your capacitor.  Admittedly higher freqs will couple much more poorly into big human-scale coax loops, but your exact numbers will depend on a lot of factors and you might find the performance still unacceptable in certain locations/environments.

It's better to fix the problem properly unless you absolutely can't afford to do so.  Balanced signalling or transformers are the best options.  Disconnecting shields and/or adding complex impedances between shields are less predictable solutions that can stop working if you eg change cable lengths, grounding points or outside environment.

Are you asking about a coaxial cable (one wire inside shield), or a shielded twisted pair (two wires inside shield)?

 I was thinking of multi-conductor shielded cables, but I fail to see how that is relevant.

This is very relevant to your problem.

"Balanced signalling" (as is often done on twisted pair cables, but they don't necessarily need to be twisted) is an effective way of avoiding ground loops and other interference entering your signal.  You can use a pair of wires from your shielded bundle for this.

Converting signals between balanced and unbalanced forms can either be done by dedicated ICs or by "balun" transformers.

[EPAIII]

In my experience coaxial cables are always grounded at both ends. Audio cables using a twisted pair are usually grounded at both ends. An overall shield on a multiple conductor cable is usually grounded at both ends.

BUT, if there are problems any of the above may need to be changed. If only one end of a shield is grounded, it is usually the source end.

The first and probably the best way of eliminating noise in a cable is to use a low impedance connection at both ends. A simple termination resistor can be used. Another technique often used in modern audio transmission is to use an output amplifier with a very low output impedance. Many circuits will have just a few Ohms, less than 5 or 10. That provides the low impedance on the line and a high input impedance can then be used at the far end. There is no reason why this can not be used for control circuits. But not high speed digital ones.

Basically it can be cut and try.

[jonpaul]

Entire subject very well covered, 

Text: Noise  Reduction Techniques in Electronic Systems by Henry OTT,

 https://www.amazon.com/Noise-Reduction-Techniques-Electronic-Systems/dp/0471850683

SYNAUDCOM, Fine work by Jim Brown and Bill Whitlock decades ago.

https://www.prosoundtraining.com/

Jon

[David Hess]

In my experience coaxial cables are always grounded at both ends. Audio cables using a twisted pair are usually grounded at both ends. An overall shield on a multiple conductor cable is usually grounded at both ends.

I can think of three examples were the shield of a coaxial cable is not grounded at both ends.

The Ethernet coaxial cable standards do not ground the coaxial at more than one spot because of the potentially large group loop this could produce.  Instead each transceiver is galvanically isolated, and the coaxial cable is suppose to be grounded at one point.

Single ended audio inputs and outputs can have major problems with noise from a ground loop.  One solution is to ground the coaxial cable at the output side, and then use the shield at the input side for remote sense to remove the common mode voltage.  This can also work in reverse with the output side remote sensing the input side, but it is not standardized.

RF connections sometimes use capacitive coupling on the shield connection to remove low frequency ground currents.

The first and probably the best way of eliminating noise in a cable is to use a low impedance connection at both ends.

If there is a ground loop then grounding the shield at both ends will make a transformer coupling into the signal wire.

[radiolistener]

you're needs coax cable with two shield layers with isolation between them. The outer shield protects inner coax shield from interference. Also pair of coax cables placed in a common shield also used for such purposes.

[gf]

you're needs coax cable with two shield layers with isolation between them. The outer shield protects inner coax shield from interference. Also pair of coax cables placed in a common shield also used for such purposes.

Still does not prevent the induction of a voltage if the shields are connected to ground at both ends, and if they are part of a ground loop.

One solution is to ground the coaxial cable at the output side, and then use the shield at the input side for remote sense to remove the common mode voltage.

You mean differential sensing between center conductor and shield, at the receiver side, right?

[David Hess]

One solution is to ground the coaxial cable at the output side, and then use the shield at the input side for remote sense to remove the common mode voltage.

You mean differential sensing between center conductor and shield, at the receiver side, right?

That is right.

[Terry Bites]

The answer is. It depends on the signal source and load the signal size, system impedance and the power level. And ground loops and the level of ac or dc leakage in or out of your cable.
Small signals:
A shield can be either a return condcutor or an electrostatic shield. Or both.
Lets say you have a shielded signal generator. The metal box is a Faraday cage, so for maximum signal integrity you extend that cage along the cable.
If you have another similarly shielded unit at the other end, you have effectively put the souce and load in the same cage. You can ground either or both ends.
If the source isn't "caged" then you want to connect your shiled to the load or signal input's "cage". Then you only ground at the receiving side. Again you are extending out your cage. You still have to provide a return path/ reference for the signal.
Here a shilded twisted pair can be used to carry the signal inside in a cable that is grounded at the recieving end.
AC magnetic fields can induce probelm currents in shields if the ends are differing grounds. Read up on ground loops. The solution here is to get rid of diifering grounds buy gibving some thought to the wiring.
You lift one end if its safe to do so.
You'll see guard shilds used to keep the "cage" ground separate from the signal return ground. A driven shield is sometimes used to eliminate cable capacitance and hence electrostatic coupling to the centre condctor.
A separate shield goes over the top of that.

With big signal, power and low impedance circuits. It usually doesnt matter that much.

[JustMeHere]

Here's an article from the best person on the planet when it comes to high speed signals: Howard Johnson

Cable Shield Grounding
http://www.sigcon.com/Pubs/news/2_2.htm