9

I have read through many recommendations articles, and many StackExchange questions, but haven't come up with any clear understanding as there are many conflicting information out there. The simple question is, should a USB shield be connected to GND? If yes, should it be connected directly, through a resistor, or through a capacitor?

My case: Computer peripheral with a metal enclosure and USB-C connector. The thing however is, I cannot guarantee that everyone will use shielded USB cables to connect the device. Of course, I can make shielded cable a requirement in the manual, but I'd rather be predicting.

What I've read and tried to take into consideration:

  • Using metal enclosure as EMI shield (both radiated and received)
  • Using metal enclosure as ESD shield (safe ESD path away from the electronics)
  • Apparently, some motherboards do have tied GND to Shield on the host end, and some don't
Marek
  • 143
  • 7
  • All USB cables are shielded I think, so are you saying you cannot guarantee that the users of your USB PC peripheral will not necessarily use USB cables to make the connection? – tobalt Aug 10 '23 at 05:36
  • 1
    @tobalt USB cables are certainly not always shielded. About a third of all cables I tested were not shielded. Users must use a USB cable, but I can't guarantee they will use shielded ones. And even if all USB cables were shielded, I still don't know if the user's motherboard has tied shield to GND on the host end. – Marek Aug 10 '23 at 06:38
  • 1
    Did you look up the shielding requirements in the USB standard? What does it say? – the busybee Aug 10 '23 at 06:42
  • 1
    There's no such thing as an unshielded "USB" cable. If it were, it wouldn't be USB. – Tim Williams Aug 10 '23 at 07:38
  • 6
    @TimWilliams To be fair, there is such thing as unshielded USB cable. They just are used in low-speed devices and are non-detachable. So any detachable USB **cable assembly** with plugs on both ends **should**, by specification, be shielded. Of course, the non-shielded ones, if they exist, cannot be officially called USB cables. – Justme Aug 10 '23 at 08:29
  • 2
    @Justme Precisely! – Tim Williams Aug 10 '23 at 08:30
  • 2
    Could you please check this https://electronics.stackexchange.com/questions/389972/usb-shield-to-ground-or-not-to-ground?noredirect=1&lq=1 – Hari Aug 10 '23 at 09:00
  • Is your device is connecting to an external device beyond the USB host? If so, is a ground loop a potential problem to the functioning of the device (sensitive analog signals, etc)? In that case, if you can't force the user to wire in a particular way you should consider adding a USB isolator and incorporate an isolated power supply into the device. Otherwise people are going to plug it in wrong and complain that it doesn't work. Hell, I've done that and I have design experience. Conversely if a ground loop doesn't matter (and often they're tolerable), let it be. – user1850479 Aug 10 '23 at 15:59

4 Answers4

14

I don't think there's a lot of conflicting information, but rather a lot of information that depends on context or EMC notions.

To understand, it's important not to confuse these symbols. Zero volt (equipotentiality), ground chassis (EMC) and earth (safety).

enter image description here

The general vision is as follows :

In theory, shielding can be floating. If it is hermetically sealed, the Faraday cage effect is obtained and the common-mode currents are drained off by parasitic capacitances.

enter image description here

Sometimes for safety reasons (class I devices requirements, reducing the accumulation of static electricity, ...) the enclosure must be earthed, and the parasitic capacitance (enclosure/earth) disappears, improving the efficiency of common-mode current evacuation.

enter image description here

In your electronic design, you always have to deal with the problem of internal EMC, and the parasitic capacitance between the printed circuit board and the metal enclosure is not ideal. A solution consist to connect directly the zero volt to the ground enclosure.

enter image description here

That's a general view of the problem ! There's a lot to be said for compromises in context. For example, if the two devices are in different buildings and with different earths, this configuration poses a problem because the 0V would be different. There's also a lot to say about one-point grounding or multipoint grounding inside the enclosure versus distributed stray capacitances in HF. The ground loop influence, etc...

Vincent
  • 675
  • 4
  • 11
  • 1
    …and this awesome book by Henry Ott covers your "etc...": [Electromagnetic Compatibility Engineering](https://www.wiley.com/en-be/Electromagnetic+Compatibility+Engineering-p-9780470189306) – pipe Aug 10 '23 at 11:42
  • "For safety reasons, the enclosure must be earthed" says who? Or are you saying to design a device that will work in a circuit having earthed enclosures, even if not all such enclosures will be grounded to earth? – Ben Voigt Aug 10 '23 at 14:44
  • @BenVoigt I think that sentence works better with a "sometimes" at the beginning :) – hobbs Aug 10 '23 at 15:43
  • @hobbs, you're right. I will edit my message and add "sometimes (class I devices requirements or reducing the accumulation of static electricity, ...)" – Vincent Aug 10 '23 at 18:33
  • @BenVoigt, I modified my message, but it was an example. The earth is optional for good EMC as in aircraft. – Vincent Aug 10 '23 at 19:33
  • The different symbols are not always used in the way you want them to be used. – user253751 Aug 11 '23 at 01:42
  • 2
    Also, the English language has a clear distinction between "security" (protection from bad people) and "safety" (protection from accidents); earth ground is obviously for safety, not security – user253751 Aug 11 '23 at 01:43
  • It's a good post, but I am afraid it doesn't answer the question, because you present both choices of the OP as legitimate. IMO: with a proper USB cable, there is no need to connect 0V to the chassis in their USB device. – tobalt Aug 11 '23 at 04:41
  • @user253751, you're right, I'm sorry for the confusion. I'm french and in my language is a same. I have modified my original message. – Vincent Aug 11 '23 at 10:49
  • I still don't think it answers the question. There's many ways how it could be connected, based on a lot of things. But USB Type-C specification mandates how to connect it and there is only one way to connect it. – Justme Aug 11 '23 at 10:58
  • @Justme, In this document https://www.usb.org/sites/default/files/USB%20Type-C%20Spec%20R2.0%20-%20August%202019.pdf : "The receptacle connectors should be connected to metal chassis or enclosures through grounding fingers, screws, or any other way to manage EMC." Like any cable shield, in fact, otherwise the shield is no longer effective at HF. However, as I said at the end of my message, it's the connection of the 0V to the chassie that will be different depending on the context (directly, by ferrite bead or capacitor and parallel resistor) – Vincent Aug 11 '23 at 12:40
  • @Vincent It does say that, but the question asks, should the receptacle shell be connected to ground, and for that question, the answer is simply yes, directly. – Justme Aug 11 '23 at 13:01
7

There isn't a simple answer in generic case so that is why you read conflicting information because the correct answer how to connect it depends on many things.

However, since your device has Type-C receptacle, the Type-C standard mandates that receptacle shell shall be connected to the PCB ground plane.

Justme
  • 127,425
  • 3
  • 97
  • 261
3

I always connect USB port shields to GND. Even when designing PCBs, you can see on their footprints, that shield pins are connected to the ground plane. This is common practise for handling USB port shields, I even use USB ports as a ground source when testing something on microcontrollers like ESP32 or Arduino. It is much easier to place the GND probe on the USB port rather than on some pin, which you might accidentally short with +5V. Connect shielding directly to GND no need for resistor or capacitor.

Dominykas
  • 183
  • 9
3

You'll hear plenty of people arguing about this online, but only in outdated discussions. "USB Type-C Cable and Connector Specification" Release 2.0 says in the notes to Table 3-11:

Shield and GND grounds shall be connected within the USB Type-C plug on both ends of the cable assembly.

I've tested the continuity on a few different real USB-C cables so I've also observed this first-hand. Even if you don't connect your shield and ground on your device's connector, the USB-C cable will do it for you.

mjt
  • 251
  • 1
  • 3
  • 1
    The standard also says that receptacle shell must be connected to ground. – Justme Aug 11 '23 at 11:04
  • 1
    @mjt I don't think it's outdated, because technically that spec only applies to USB-C, whereas a great many people (especially on this forum) still use other USB connectors, because they don't need USB C features. – BeB00 Aug 12 '23 at 00:40