I'm not at all a hardware expert—I'm a software engineer—but a recent discussion piqued my curiosity about whether grounding a USB shield is useful, necessary, or something to avoid entirely.
The Universal Serial Bus Specification Revision 2.0 states:
6.8 USB Grounding The shield must be terminated to the connector plug for completed assemblies. The shield and chassis are bonded together. The user-selected grounding scheme for USB devices and cables must be consistent with accepted industry practices and regulatory agency standards for safety and EMI/ESD/RF.
Whatever that means!
From what I've gathered, the literature suggests that connecting a shield directly to signal ground on the device side can create unwanted ground currents or interfere with EMI/ESD behavior. Some recommendations include placing a resistor or ferrite bead between the shield and ground to limit high-frequency noise conducted onto the shield (e.g., from Texas Instruments), rather than making a direct connection.
On the other hand, there are recommendations to leave the shield floating on the peripheral side to avoid ground loop currents between host and device grounds. The reasoning seems to be that if the host and device grounds are at slightly different potentials, a direct DC connection could allow current to flow through the shield, potentially injecting noise or causing interference. (which I don't get since the GND are linked they are all at the same potential by definition... ?? - may be transitory state ??)
So what do experienced hardware engineers actually do when faced with this question? What's the decision tree for deciding whether to connect the shield (and through what), or to leave it floating?
Any wisdom, experience, or perspectives would be greatly appreciated!
My decision tree normally works something like this:
Do I have a chassis or earth?
If yes, the USB shield ties to that directly. This is how some of my industrial PCBs are designed, as they are made for applications such as government security etc (One of my clients is a security professional who subcontracts to the Australian Government). These devices are installed in buildings/applications that have a true earth or chassis system already in place.
If no, I have to create a on-board chassis. This is where my user-friendly boards are located. The USB shield ties to an isolated copper pour around the connecter. This pour is then tied to the main GND pour via a 1M resistor, and a 1nF/4.7nF capacitor. This gives a bleed point but blocks DC.
Is this a cable/shield termination problem or ESD problem?
Usually it’s both. Hence the first point (1) is first.
If I leave the shield floating, it can make ESD and RF behaviour worse, as it basically turns the board into an antenna with undefined potential!!
Is the device supposed to be isolated from the host?
Almost always, this is a definitive no. An exception is my latest industrial custom order requires galvanic isolation which is taking me down a brand new learning curve!!
For my normal boards thqt anyone can buy, I generally use this standard:
Connector shell/shield to a local “chassis ring” copper at the connector.
That chassis ring → digital ground via a parallel network, containing a capacitor with 1 MΩ.
ESD TVS diodes for D+/D- (and VBUS) placed close to the connector, with a very clean return to ground.
I RARELY leave the shield floating.
NOTE: Layout matters!
Keep the shield pour/traces as close to the connecter as physically possible. Keep it tight!!
Stitch the chassis pour heavily with vias.
Don’t route D+/D- under a noisy split or across gaps.
Keep the D+/D- pair tightly coupled, short, and referenced to a solid plane.
Make the ESD path short and not through any core ground return bottlenecks.
It's common practice in the audio world to ground the socket shield, but not the plug shield.
This prevents a ground loop.
All USB shields I have seen were firmly grounded to the ground pour, also for mechanical reasons.
Leo..
- may be transitory state ??)