Cable shielding, connect at one point, two points and ground loops

[trevortjes]

Hi all,

I think this is a known discussion in electronics, and because of that I can't find a straight answer.

We got two devices to connect together via RS232 (RX, TX, GND). The documentation tells us to only connect the shield braiding on one DB9 connector which in turns connects to the GND of that device. The braiding runs along the cable, but gets cut off before reaching the device on the other side. So the shielding is connected at only one point, and doesn't even cover the whole cable to the next device.

Ofcourse! Team A says, this way you elevate ground loop problems!

Oh no! Team B says, your shielding is now near useless!

Which team is right, or is there a team C?

[Someone]

This was pretty comprehensively covered 2 weeks ago:
https://www.eevblog.com/forum/chat/using-pe-(earth)-for-signal-cable-shield/

[Siwastaja]

TLDR: ground loops are normal. They are usually less harmful for digital single ended signal integrity than trying to avoid them. For precision analog, differential signalling should be used. Basically the #1 exception (i.e., where ground loops are actually bad) is insistence on single-ended audio.

RS232 connect on both ends, unless proven otherwise for some very specific reasons.

[radiolistener]

There is no straight answer.
It's the same as to ask which antenna is correct - half wavelength dipole or magnetic loop 

But single point grounding for the shield is preferred, because it reduce interference due to common mode currents.

Oh no! Team B says, your shielding is now near useless!

This is incorrect, the shield works even if it connected at one point. And even if its not connected at all, it still works. But each case may lead to another issues. It all depends on the case.

[Phil1977]

It´s of little use to discuss where to connect the shielding if you don't talk about the GND-line inside.

If you shield a differential pair then things are quite clear: Connect the shield only at one side to avoid ground loops and magnetic induction loops.

But RS232 is not a differential interface. It´s signalling via voltage pulses that are referenced to a common GND line. That means, if you don't connect the shield on both sides of the cable, then the ground loop or magnetic induction loop is established purely by the GND-wire. In this case it can be better to connect the shield on both sides to create a low impedant connection that takes a big part of the parasitic currents.

The best solution definitely is to use RS422/485 instead of RS232. Or Ethernet or CAN or anything else differential.

[trevortjes]

So in an hour I got connect to one side from 2 replies and two sides from 2 other replies.

Is there literature I can dig into to find the right answer?

[radiolistener]

Is there literature I can dig into to find the right answer?

As said before there is no "right answer", but preferred solution is to ground shield at one point.

You can read more in transmission line theory.
Here is a short overview: https://innovationspace.ansys.com/courses/wp-content/uploads/sites/5/2020/12/DesignTools_TransmissionLineTheory_Rebranded.pdf

[tooki]

So in an hour I got connect to one side from 2 replies and two sides from 2 other replies.

Is there literature I can dig into to find the right answer?

That’s literally the point: there is no One Right Answer, and if you look into existing discussions, you’ll see that that is the only consistent answer.

[Phil1977]

I´m not sure if there is an obligatory definition for it, but companies producing RS232 cables specify to connect it at both sides.

https://ftp1.digi.com/support/cabling/cabling_specifications.pdf

[Siwastaja]

Then again, if we constraint the problem, then there will be right and wrong answers. I.e., let's not overgeneralize the OP's problem.

Am I correct that these apply for the OP's case:
* Single ended digital communication, referenced against internal GND node at both ends;
* Both ends are either unearthed, or tie internal GND to the earth solidly, like personal computers for example

(Internal GND means the reference level against which the TX/RX signals are referenced to decide their state. I.e., the transceiver's GND pin.)

Under these constraints, I think that my advice "both ends of shields connected to said internal GND" is the correct one. There will be a ground loop of course, then what? Ground loop will be there anyway - you need that ground to communicate at all. What is the other option, route the ground from some long parasitic route? The shield in parallel with any other ground wire (yet undiscussed) yields minimum possible impedance for that net, and therefore lowest possible voltage from ground loop currents.

[trevortjes]

Am I correct that these apply for the OP's case:
* Single ended digital communication, referenced against internal GND node at both ends;
* Both ends are either unearthed, or tie internal GND to the earth solidly, like personal computers for example

Yes and not quite, this is on a vehicle. So we got chassis ground and internal ground. These are not always connected with eachother.

The cable is RX, TX and a GND wire with a shielding braiding around them.

[Siwastaja]

Yes and not quite, this is on a vehicle. So we got chassis ground and internal ground. These are not always connected with eachother.

Maybe they should?

[trevortjes]

In this specific automotive usecase we have reasons to not have all modules connect to chassis earth.

But it seems like to get a good answer for this problem, a lot of context of the system is needed.
Sadly I only have details on the source side of the cable but not where it is being plugged into.

I want to touch upon another part of the question tho.
This particular cable ends with a few centimeters of unshielded cable. It also goes to an automotive connector with no shielding or metal parts that can act like one.
I guess we can call the end of this cable a "pigtail".

Don't the shielding capabilities of this braid decimate when a cable is executed this way?

[Phil1977]

I suppose it´s only a debugging interface or something similar, because for automotive things that need to be reliable other protocols like CAN are chosen.

How many people die if you have biterrors? What bitrate are you using? Do you use parity or any other checksum?

RS232 with a "standard" bitrate usually survives to have many centimeters without shielding. Ground loops are usually more annoying in that context, but you need the GND-connection for this interface.