6
\$\begingroup\$

Consider a PoE solution that uses shielded CAT cables and where the shield is connected to signal ground on both the PSE and the PD side. In this case, the shield is most probably presenting a lower resistance path than the twisted pairs in the cable, so the DC return current will flow through the shield. Are there any disadvantages with this setup? Should you AC couple the shield to the signal ground to prevent this? Are there any standards or best design practices that address this?

\$\endgroup\$
11
  • 1
    \$\begingroup\$ To emphasize @Justme's good answer. In general (There are always exceptions), shields, chassis, structures, etc should not carry any power or return currents. They should only carry leakage, stray, unintended coupling currents, that sort of thing. \$\endgroup\$
    – SteveSh
    Mar 4 at 12:00
  • \$\begingroup\$ In case of an instrument of some kind with interconnected (power and data) modules, a common ground philosophy is to connect the signal ground to the chassis ground. In this case, the DC return current will inevitable flow through the chassis. In what way is this a bad thing? \$\endgroup\$
    – FreddaH
    Mar 4 at 13:23
  • 1
    \$\begingroup\$ For a couple of reasons. First, the chassis ground connections are usually not rigorously designed & built. This means the resistance in that path may not be well defined, Dissimilar metals and moisture can lead to corrosion. And using the chassis and shields for return currents can exacerbate radiated emissions, causing you to fail EMC testing, \$\endgroup\$
    – SteveSh
    Mar 4 at 15:21
  • \$\begingroup\$ Also note that especially in industrial situations, pretty solid differences in local chassis potential are hard to rule out. Do you really want to risk helping the neutral connector on a 63 A device carry current by offering a second path via the ethernet cable? \$\endgroup\$ Mar 4 at 15:24
  • \$\begingroup\$ Agree when it comes to the chassis. And if you intend to use it for shielding the construction and design is really important. However, my experience is that connecting the 0 V reference on all PCBs in an instrument of some kind (medical, industrial etc.) to the chassis ground yields lower radiated emissions. And my point is that this will lead to DC return currents in the chassis in case the chassis provides a lower resistance path than the 0 V reference wire in a signal or power cable. \$\endgroup\$
    – FreddaH
    Mar 5 at 8:28

2 Answers 2

18
\$\begingroup\$

Yes, it is a bad idea.

The cable shield should not be used as a current conductor.

And as Ethernet is supposed to be a transformer isolated interface, including the PoE part, there should not be any possibility of return current of PoE flowing in the cable shield.

If there is a possibility of return current in the shield, there is likely a design error and device is not compliant with Ethernet or PoE standards. Standard requires 1500 Vrms isolation between PoE voltages and frame ground.

Also, if you require shield as a return path, it will not work with a standard but just non-shielded cable. Also not all cables or devices provide a connection to shield, and in buildings, inside the infrastructure cabling, the shield will be bonded to PE earth/ground, and wall sockets in the room may not provide a connection to the cable shield at all in order not to cause ground loops.

\$\endgroup\$
2
  • \$\begingroup\$ Isn't the isolation requirement specified like "provide isolation between all accessible external conductors, including frame ground, and all MDI leads"? I cannot see that it says anything about isolation between PoE voltages and frame ground, but I may have misunderstood something. Regarding the return path, I was just pointing out that the DC return current will flow in the shield whether you want it or not unless the shield is left floating or AC coupled to signal ground. Also, I often read that it is a bad idea to use the shield for the return current, but why exactly? \$\endgroup\$
    – FreddaH
    Mar 4 at 12:45
  • 3
    \$\begingroup\$ @FreddaH But the point is on a properly made device there should be no DC return current at all in the shield, not even if you want it. If there is, the PoE device is not made properly. Why? PoE supply and return must go in the MDI data wires, and MDI data wires should have no reference to any exposed circuit, connector, or metal chassis/frame, as the MDI wires must be isolated. Running return currents in the shield is bad because it creates a ground loop and might act as an antenna for all the electrical noise. Also unsafe if you accidentally ground a device only through cable shield. \$\endgroup\$
    – Justme
    Mar 4 at 13:39
-2
\$\begingroup\$

As I understand, ground loops are not really a problem in small ethernet installations and even less of a problem within a single instrument.

Large industrial ethernet installations on the other hand, where for example two machines are communicating can be a problem if the shield and the PE-ground are connected to the chassis. If the PE-ground differs at the two machines, this will drive a current through the shield. Leaving the shield floating at one end solves this, but effectively creates an antenna instead.

Running current in the shield per se is not a problem. All coaxial cables obviously has the return current in the shield.

\$\endgroup\$
6
  • 1
    \$\begingroup\$ Comparing coaxial to shielded twisted pair makes no sense as they are different to begin with. And in the case of good old 10Base2 coaxial Ethernet, the coax shield is isolated from your PC metal case so it anyway does not apply here. \$\endgroup\$
    – Justme
    Mar 4 at 16:38
  • \$\begingroup\$ The shield is an antenna whether connected at one end or two. It just changes the frequencies of the set of standing waves that are possible. Ethernet should never have ground loop issues. It is transformer isolated at each end. \$\endgroup\$ Mar 4 at 23:07
  • \$\begingroup\$ @RobinIddon The shield of shielded cable is not transformer isolated. If you do connect two desktop PCs together with shielded Ethernet cable, it is very likely that the cable connects the metal chassis grounds together. Which should both be anyway earthed through mains plugs. So there is a ground loop. Ethernet data will still be transformer isolated. \$\endgroup\$
    – Justme
    Mar 4 at 23:59
  • \$\begingroup\$ Running current in the shield is a problem because there should be no DC connection at all between the shield and the PoE supply circuits. So if you got a current flowing through the shield, something is seriously wrong. \$\endgroup\$ Mar 5 at 17:02
  • \$\begingroup\$ All coaxial cables obviously has the return current in the shield. Yeah, but in industrial scenarios the shield must be insulated from the frame just as it is in PoE Ethernet. It must be insulated even if it's just a signal connection. Otherwise you get instant ground loops and radiated emissions and it's a mess. SDI folks have messed up big time on that, thus the "SDI isolator dongle" nonsense. SDI should have been isolated from the get go. Ethernet people have learned that lesson well, and I'm glad, having dealt with a setup full of SDI ground loops. It was awful. \$\endgroup\$ Mar 5 at 17:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.