I’m looking at designing a system which has two connectors and attached to the chassis which completely surrounds the circuit board. The connectors are a shielded 8P8C connector and the other is a USB C connector. Ethernet is used with discrete magnetics to connect to a PHY onboard. Due to the application, the USB and shielded 8P8C connectors need to be physically close (within an inch between each other).

When it comes to grounding, is it better to have the chassis and the shields of the connectors connected to system ground with a low impedance ground pour or isolate both of these and attach to the system ground using a low value capacitor?

Whilst I have seen examples of a single 8P8C connector, I was looking for some guidance on where to look for existing designs that do this successfully.

  • \$\begingroup\$ How much does it matter whether the chassis and shields get tied together (at DC / galvanically)? Will that ever create a ground loop of significance? \$\endgroup\$ Commented Dec 11, 2022 at 19:14
  • \$\begingroup\$ @TimWilliams The chassis and connector shields shouldn’t create a ground loop of significance and they have a fairly low impedance connection to chassis. I’m more interested in the grounding scheme between the chassis and the board ground. I’ve asked a separate question on this forum about the Bob Smith termination implications of this specifically too. \$\endgroup\$ Commented Dec 12, 2022 at 17:56

1 Answer 1


Sounds like it would be fine to just toss everything to ground early and often: screws or gaskets to chassis GND, pours around connectors for their shields, and gaskets around the connectors themselves if applicable (chassis to connector grounding). BS termination would then be to the same GND, which will be right near the connector and its shield, no problem.

Isolation and small capacitors and such, really only matters when there's, well, isolation required, or enough other stuff going on inside the box that it could interfere with itself.

For isolation, take for example if you added an isolated RS-485 port. There will be a DC-DC converter, logic isolators, and transciever (and maybe filtering and other associated EMC protection, but that's not important here, that's with respect to the interface chip). You might have the situation that, although the isolation capacitance is small (10s of pF?), it resonates quite strongly with the attached cable, which, maybe it's hard-grounded at the far end so acts as an inductor against your isolation capacitance. Given a source of internal interference (maybe a low impedance source at the grounded end, like a ground loop including a switching converter?), it could radiate quite strongly, or conversely, receive external interference could develop quite a large AC voltage across the isolation gap, corrupting data (exceeding isolators' dV/dt rating).

In that case, it might be desirable to introduce loss to that capacitance: a small capacitor (100s pF?) with significant ESR (as from a ferrite bead or resistor -- in either case, 10s-100s of ohms resistance at radio frequencies), acts in parallel with the isolation capacitance, in turn acting as a termination for the energy on the cable/shield. Instead of resonating strongly (standing waves), the cable develops much less gain. Note you're not filtering out the noise here, you're just knocking down the peaks due to a poorly damped element in the system.

It would do equally well to place a ferrite bead near the low-impedance (grounded) end of that cable, both terminating its CM impedance, and increasing the ground impedance so that less current can flow on it in the first place (and thus, less voltage develops at the 1/4λ points away from that node, etc.). Perhaps for some reason, you can't ensure a ferrite bead there: then the R+C across the isolation serves a similar purpose.

The other case, with "other stuff going on inside the box", would be more like the motherboard and expansion cards of a PC. Here, while grounding may be generally okay, there's just so much stuff going on, on both sides of all PCBs inside there, there can be some noise generated -- you've got the logic levels (or whatever), maybe it's on the order of a volt of noise, but it's also over a broad spectrum (100s of MHz to 10s of GHz). That's a whole lot less to deal with than the wild outside world -- it's not some 3, 10 or more V/m of radiated fields, or ESD or EFT. But you still need to be careful that, say you're designing an expansion card, and it has this kind of CM noise coupled through it, that noise isn't transmitted through the slot guard and connectors. So you might isolate the main PCB area away from the slot and connector area, and use common mode chokes to carry signals over to the connectors.

Since your project apparently doesn't have much inside it, it seems the latter case wouldn't really apply, and common mode currents/voltages will be easily resolved, so that little comes out from the connectors. Just a solid ground plane on the PCB should suffice.

  • \$\begingroup\$ good idea, I understand that for Ethernet specifically, connecting the shield to the system/board ground means we no longer have isolation. Would this completely nullify the need for Ethernet magnetics (and BS termination) or would they still be useful for minimizing DC and common mode noise? \$\endgroup\$ Commented Dec 13, 2022 at 14:10
  • \$\begingroup\$ If the Ethernet shield should be isolated, then it can be returned to the Bob Smith common point (the cap to GND) instead. \$\endgroup\$ Commented Dec 13, 2022 at 14:29
  • \$\begingroup\$ If it does need to be isolated, should I also connect the USB shield to the Bob Smith common point or just use the chassis as the connection there? And if it doesn’t need to be isolated (or at least, this was not the main concern) is that when we’d connect it to system/board ground? I guess my question is whether or not those components (magnetics/ BS termination) would still be useful in that instance? \$\endgroup\$ Commented Dec 13, 2022 at 15:40
  • \$\begingroup\$ USB is not isolated, keep it on system/board ground (which again, it sounds like can be the same thing) \$\endgroup\$ Commented Dec 13, 2022 at 23:09
  • \$\begingroup\$ Shielded Ethernet may still benefit from the magnetics. I don't know much about it, but I assume it's used in extremely noisy environments (>30V/m?) where the CMRR of the transformers alone won't do enough. \$\endgroup\$ Commented Dec 13, 2022 at 23:10

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.