shielded cables in enclosure

[MBK]

Hi all,

what is the current scholarly thought about cabling within an enclosure? I mean, for wiring to/from I/O and pots or swtches. I see problems with all shielding options:

- leave shield disconnected at both ends: ineffective shield
- leave shield disconnected at one end, ground the other on signal ground: shield may act as antenna and inject noise to ground
- as above but use enclosure to ground one end: you create an enclosure protrusion, some say this is not good
- connect shield at both ends to ground or enclosure: ground loop.

Incidentally, if using unshielded wiring inside an eclosure, I also see dilemmas:
- twist tightly (as often recommended): create high capacitance, which the usually unprotected outputs of, say, op amps *within* a circuit won't like. Plus, may pick up crosstalk
- don't twist: create a loop antenna

Err. There has got to be a compromise that works. opinions?

MBK

[lucpes]

The best 'compromise' would be using the enclosure to ground the 'source' end (use microphone cable, the 2 cables inside for ground & signal and connect the outer shield to the case).

[MBK]

Thanks Lucpes for your input. Though Microphone cable is usually too stiff to use it for connecting, say, volume or balance pots in an enclosure, and we still break some rules as of my original post.

This thread didn't generate much interest, I wonder why. It's a situation everyone has in DIY and the recommendations I found are mutually conflicting.

[maylar]

Actually, I'm very interested in this subject myself. I've seen a wide variety of different methods used in commercial audio equipment and I wonder what works best.

My current thinking is to use isolated RCA jacks and differential amplifiers for input circuits. Output wires would ground the shields at the source end. At switches I'll ground the source end and leave the other side unconnected. No matter what scheme I think of though there's always the chance for ground loops when connecting to external equipment like tape decks.

[MBK]

Maylar,

I think that's a good idea, to use signal ground, at least you don't get the "outside world" into the enclosure.

As for me, my compromises are evolving... For DIY stuff I now do everything balanced or pseudo-balanced (equal impedance on signal ground output). That makes the shield grounding *much* easier (shield to enclosure and hot and cold to diff input - no shield penetration, no ground loop, no infestation of signal ground). For RCA I use insulated jacks, then all RCA shield/grounds go to the same diff amp as for balanced inputs. Problem here is that the shield should have a zero impedance connection to enclosure at entry point and not penetrate the enclosure... in theory... but then you mix shield currents and signal ground currents. I guess with RCA you can't get it "theoretically" right. I may change the scheme to connect all unbal grounds to enclosure at a single point, then 50R to diff amp input. Then at least the shield does what it's supposed to, and the 50R should be protection enough for the signal ground. All this however leaves the question as to the internal shielding - because here you basically have to connect the shield at receiving end (diff input) or else you get outside shield currents inside your enclosure and it's going to radiate inside.

For switches and pots I now use mini-coaxial, very flexible, but I suspect high capacitance. In some cases the source has an R impedance so that's ok, best would probably be to use 50R or so at the source in case the sourc is a "naked" op amp output, to prevent ringing/oscillation. Outputs, I usually don't bother but try to get the output close to the connector, and twist the wires.

Do you have any idea if shields connected at just one end will really act as significant antennas? This pops up every once in a while up to the point that ppl recommend plastic XLR jacks just so that no portion of the enclosure shield, even the 1/2 inch of an XLR jack, protrudes inside the enclosure. Because if this really matters, then even say, the shield of a cable to a switch, connected at source = signal ground, will amplify the problem rather than help it...

MBK

[SY]

Do you have any idea if shields connected at just one end will really act as significant antennas?

Seriously doubtful, especially if they're dressed along the chassis. This grounding method is standard in instrumentation, low noise, and other precision applications.

If you want to be a real maniac, use a buffer amp to drive the shield so that the inner conductor and outer conductor track one another. That gets rid of any issues with respect to shielded cable capacitance, and if the buffer has the requisite low-source Z, the shield will still provide shielding. This is wildly more complicated than needed for this application, but so is worrying about a shield injecting stuff into your ground. If you limit the bandwidth of your circuit to any sane value, it's an absolute non-issue.

[roadkill]

A Crystal Semiconductor appnote I once read mentioned returning all shield grounds to the star point separately. This was in a different context (DAC output driving coax out of the enclosure) but this will solve the problem of RF pickup from the shield injecting itself into the local ground (which will be isolated from star point by interconnect impedance). Then again, the shield is *inside* a metal box, which is unlikely to have a strong enough RF component to be significant when it's being coupled in by a poorly matched antenna into a very un-cooperative load (the inductance of the wire from the star point to the circuit under considertion).

SY,
RF signals so injected can get rectified and introduce interference (ghost voices from AM radio stations, TV frame buzz, etc)

[maylar]

I think that level of 'precaution' is nonsense.

I'm sure the best solution is to use balanced lines throughout, and only use the shields for RFI protection. As long as no signal passes through them there's no chance of ground currents becomming part of the signal. But that doubles the amount of switch contacts required and makes buffers more complicated.

Single ended signals that use ground as a return are more complicated to evaluate. Especially if signals travel between multiple PC boards. The question I have is how to wire preamp, filter, and output boards together and make sure that only the appropriate return signals travel down interconnect shields (or even separate wires) and not power supply currents.

I need to think about this...

[SY]

RF signals so injected can get rectified and introduce interference (ghost voices from AM radio stations, TV frame buzz, etc)

Sorry, but a few inches of shield inside a chassis is highly unlikely to pick up RF at anything less than UHF. Take a look inside your oscilloscope, for example.

RF is FAR more likely to originate from stuff outside the box, and is best dealt with using standard methods (e.g., input resistors and caps, ferrite beads, external shields...).

[SY]

The question I have is how to wire preamp, filter, and output boards together and make sure that only the appropriate return signals travel down interconnect shields (or even separate wires) and not power supply currents.

That is a VERY pertinent question. There's some good discussion in Horowitz and Hill, and a very illuminating set of app notes on the Jensen Transformers website. You may not want to implement transformers at every signal junction as they'd like you to, but the basics of grounding problems and solutions are very well treated.