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..
Here's how I understand it, but I am hoping that someone will correct anything I get wrong or perhaps add more detail if needed.
0V
A point on the circuit designated by the designer against which all other voltages are measured unless specified otherwise. By definition there cannot be any noise at the 0V point.
Ground
Often used synonymously with 0V and will be connected to 0V. Ground is more than just 0V, it is the common return point for the circuit and is also implemented on a PCB as a 'plane', meaning it is filled in with copper wherever there is space for it. Ideally a complete layer would be the ground plane with nothing breaking it up. By covering a complete (or as complete as possible) layer it provides shielding against noise and ensures that the return path for current can exactly follow the trace that provided the current to whatever device it is going to. This is increasingly important for higher frequencies, and probably unimportant for DC. I don't really know at what frequency it really starts to matter, buy maybe low Mhz. Remember that a square wave (logical signals) have frequency components far beyond their fundamental frequency. To be clear about the difference between 0V and ground, 0V really should be a single point in the circuit, ground is everything else connected to that point.
Earth
Connected to the muddy stuff outside. Used for providing safety so everything is at ground potential, ensuring there is no voltage difference between what one part of your body is touching and another. May be connected to ground / 0V, but does not have to be. Will be connected to a metal case or similar for safety.
Shield
Provides protection against electromagnetic interference into a cable. Can be connected to ground at one end of the cable, but not both. If connected to both ends it then carries circuit current and provided a path into the circuit for noise. So, to the question, I think the shield of a USB cable should be connected at the A end (PC for example) but not at the peripheral end (Uno perhaps?).
That's as much as I know, it's incomplete and possibly wrong in places, happy to be corrected.
In my experience working on PC motherboards (repairs, finding issues, replacing components, repairing stuff, etc) the USB shields are generally connected through various protection systems to the earth plug of the PC. Therefore, I believe the Arduino/shield end should be connected to GND, again, through protection systems/circuits.
Is there, or can there be, a difference between the 0-V supplied to a board like the Uno depending on which power source you use?
Referencing everything to the the general earth, i.e. a large metal rod or plate buried in the soil, can there be a difference between the zero volts with a meter connected to earth on one side and 0-V when powering the board through the USB socket or from an unearthed plug top (Wall-wart) supply through the 2.1-mm socket?
Often, these plug in power supplies don't have a functional earth pin, just a plastic one to operate the sprung shield in UK type mains sockets.
My property has an electricity supply (240-V AC 50-Hz) that just has two wires, live and neutral.
The neutral doubles up as the earth. It's called PME or Protective Multiple Earthing where the supply company earths the neutral at points along the supply cable route. Sometimes also called TNCS (Terre, Neutral Combined, Separate). Separate means the house earth cable is connected to the neutral before entering the property.
All USB shields I have seen were firmly grounded to the ground pour
what is this noun : a pour ?
What is getting poured?
i.e. One end of the shield and one end only "directly" to earth ground.
I recently had comms problem and found the USB C cable I was using did not even have a shield braid. Just the data and power wires. The plug body was not connected to anything in the cable.
Is the 0V of the USB power floating or connected to system chasis at PC host end?
" A ground pour (or copper fill) in electronics and printed circuit board (PCB) design is a technique where unused empty spaces on a circuit board layer are filled with a solid sheet of copper connected to the system ground
Sometimes Ground plane. It's just a term, nothing gets poured. Exception could be conformal coating.
If you were etching a PCB, it would be a large area covered with etch resist.
Sometimes connected through multi-layer boards to other grounds with via pins or holes.
Thanks for the explanation. I suspected it meant ground plane but using verbs as nouns is always confusing especially when the verb is totally irrelevant.
PCBs , as you say, are done by applying a coat of photo-resist, UV curing and etching not pouring. It is a totally inappropriate term.
The term probably originates from the way it looks like you are pouring into an area on the computer screen in the program that you use to create the PCB art.
"Copper pour" makes no sense since nothing is poured anywhere, whether it is connected to ground or not.
Copper fill is more correct since it is an area filled with copper. Not in the sense that anyone "fills it up" but that it is full because the copper was not removed.
Ground plane was relevant where I used that term, because that was the original context in which it was called "a pour".
Once again pour is not a noun , it is a verb. Degrading language like that creates ambiguity and impedes comprehension. I had to read it three times because I thought I was misreading. Then I realised "a pour" was being used as a noun and had to work out what was allegedly being "poured".
According to the Oxford English Dictionary, the noun pour is defined as an act of pouring a liquid, a heavy fall of rain (a downpour), or an amount of molten metal or liquid dispensed at one time. It entered the English language around the late 1700s.
The key distinction is that USB shield and signal GND serve different purposes. Host and device GNDs are connected through the USB cable, but they are not guaranteed to be at exactly the same potential because the cable and return paths have finite impedance, so transient or DC currents can still flow. For EMI/ESD, the usual goal is to give shield currents a low-impedance path that keeps them away from sensitive signal ground, often by bonding the shield to chassis near the connector. Whether that is direct, capacitive, or through an impedance depends heavily on the enclosure, grounding scheme, and EMC requirements, so there isn't really a universal "always float" or "always connect" rule.
- may be transitory state ??)
