When it comes to polarizing and confusing questions in electronics, wiring up shields is on the top-10 list when sorted by popularity. It’s a question most of us need to figure out at some point – when you place a USB socket symbol on your schematic, where do you wire up the SHIELD and MP pins?
Once you look it up, you will find Eevblog forum threads with dozens of conflicting replies, Stackexchange posts with seven different responses plus a few downvoted ones, none of them accepted, and if you try to consult the literature, the answer will invariably be “it depends”.
I’m not a connector-ground expert, I just do a fair bit of both reading and hacking. Still, I’ve been trying to figure out this debate, for a couple years now, re-reading the forum posts each time I started a new schematic with a yet-unfamiliar connector. Now, of course, coming to this question with my own bias, here’s a summary you can fall back on.
Consumer Ports
Putting HDMI on your board? First of all, good luck. Then, consider – do you have a reason to avoid connecting the shield? If not, certainly connect the shield to ground, use jumpers if that’s what makes you comfortable, though there’s a good argument that you should just connect directly, too. The reason is simple: a fair few HDMI cables omit GND pin connections, fully relying on the shield for return currents. When your HDMI connection misfires, you don’t want to be debugging your HDMI transmitter settings when the actual No Signal problem, as unintuitive as it sounds, will be simply your shield not being grounded – like BeagleBone and Odroid didn’t in the early days. By the way, is a DVI-D to HDMI adapter not working for you? Well, it might just be that it’s built in a cheap way and doesn’t connect the shields of the two sockets together – which is fixable.
Putting an Ethernet port on your board? This is the case where you should probably watch out, to make sure you understand what’s happening. Ethernet doesn’t use or need ground as such — it’s a galvanically isolated interface — but if you ground the shield and then use a shielded cable, your ground will go run along the shield. Do you need it? Likely not. Do you want it? I’d say it’s inconclusive. This only applies if your cable has a shield and metal-plated jacks that connect to it: while that’s not the norm, it’s the case for CAT6 and CAT7 cables, which are now pretty common because they’re pretty much required for 2.5 Gbps and above interfaces. Same goes for magjack shields, though watch out, they have a separate ground pin for the internal transformers and circuitry, you do need to wire that up, it won’t propagate any further than it might need to.
Every coaxial port that has a shield contact, but there’s only two contacts in total? This goes from RCA to BNC – the “shield” is actually often a ground pin, though not everyone might realize this. If it is used as a ground pin, passing the return current through, it has to stay grounded at both ends.
USB Ports
There are a fair few different USB port connectors and roles you could be putting on your board – here’s a quick answer for a somewhat common case. USB-A port that’s a host and therefore a power source? Connect the shield to ground. This is what you are universally expected to do, and something that indeed is done basically everywhere you look, bar devices that are being a bit bizarre.
What about a device port, like microUSB, or USB-B, or, hell, miniUSB? Well, the recommendation is leaving two footprints connected to the shield, so you can stuff them with 0-ohm resistors later, or put, say, a resistor-and-capacitor combo there if you’re having noise issues. Honestly, it is unlikely to hurt you if you ground it, doubly so if it’s a power only port – USB is a seriously sensitive interface when it comes to common mode shifts.
Using USB-C? Just wire the shield to ground. Do not pass go, do not collect 200, just wire it up to ground. Why? The specification requires you to, so that’s an easy call. It’s definitely true that, with dual-role ports being abundant, you can no longer really differentiate between host or device, so figuring out shield grounding based on already shaky roles would be a problem. By the way, this is a gift for all of us reverse-engineers out there – finally, a connector with a shield that is guaranteed to be ground-connected!
Not everyone realizes that the specification explicitly mandates it, and people still think separating shield from ground could be meaningful, including myself as recently as a few months ago. If anyone needs a pointer, tell them to check the USB-C cable and connector specification, revision 2.2. For sockets (receptacles), open page 44 and refer to the note 11, for plugs, open page 78 and refer to the note 6.
Other Concerns
There is some mechanical helpfulness to hard-wiring shield to ground, especially if you’re using copper fills without thermals. In my experience, it’s harder for connectors to break off when they have a plane of solid uninterrupted copper they adhere to! Now, thermals do help soldering a whole bunch, so I don’t want to be gung-ho about it, but if you’re dealing with a fully surface-mount connector that can get torn off, you might as well go thermal-less just this once. Worked for me on a larger-scale product of mine, to the point I completely stopped getting the problem where microUSB shield pads would get torn off the board. Of course, remember, microUSB with through-hole mounting pins is superiour mechanically – in case you do make the very situational decision to use microUSB.
The debate about grounding shields might never end, but if you want a summary, this is what I can say surefire, and I hope it helps. What’s your take?
feel you with seperating the USB-C Shield and have to thank you for pointing it out before i sent my board to the fab
“Do you need it? Likely not. Do you want it? I’d say it’s inconclusive. This only applies if your cable has a shield and metal-plated jacks that connect to it: while that’s not the norm, it’s the case for CAT6 and CAT7 cables, which are now pretty common because they’re pretty much required for 2.5 Gbps and above interfaces. ”
It’s more complicated for Ethernet, though, because when you say “connect shield to ground” you have to be careful since you’ve got literally 3 things to think about:
signal ground
chassis ground (the point where the signal lines are Bob Smith terminated to on the isolated side)
local earth
You never connect shield to signal ground. You should always connect shield to chassis ground – the whole ‘leave the shield float on one side’ thing is fairly nuts, if you do that, it’s not a shield and it’ll resonate.
The earth (safety ground, whatever you want to call it) connection doesn’t have a one-size-fits-all answer because you need to know what you’re doing.
Keep in mind connecting shield to PE is not always a viable choice. In many parts of world (for example in Belarus where I live) most rural only have L and N cables and no PE.
If specific machine needs to be grounded it’s usually done by bridging the ground bolt (one which sticks out of socket) to N cable. But you need to be careful when fixing fuse box because by mistake you can make washing machine or cooker connected to L instead.
You never should connect shield to earth: you connect shield to chassis, and maybe chassis to earth.
The distinction ‘s important, because as you note, there are times you can’t connect chassis to earth, but except for really weird cases, you connect shield to chassis.
oh, I forgot, there’s also the alternative of what Starlink’s Dishy does, which is “shield?”
Wouldn’t an ungrounded shield have way more ESD issues? It’s hard to touch a pin without touching the shield first, which would ground you and prevent the ESD.
Yes, although I’d tend more to think of it as a chassis issue than shield. There are still cases where you would need an ESD grounding strap before you handle something because you can’t earth the chassis easily. Plus Ethernet is fairly ESD resistant because of the isolation anyway.
But more generally than Ethernet, absolutely. Having a ground point contact first always helps ESD. Crate-based systems (VME/VPX/cPCI/etc.) have exposed ground and ESD clips on the slots for this exact reason.
Ooof, yeah. I’ve gone down this rabbithole numerous times. And there are a conflicting-many of loud know-it-alls. Really, it depends on each device and how they’re powered… We get away with quite a lot because of isolated power sources! Try running a 12VDC LCD connected to a 12VDC-powered Pi, and throw in a 12VDC printer, all connected to a car battery, and find out just how much voltage difference 3 extra feet of power cable can make for, and how much current can flow the wrong direction through those “signal” cables. HDMI-goes-poof!
Tangentially-related:
I used to live in a house with a CableTV connector on the wall next to a metal spiral staircase… I got many zaps from that thing, while climing those stairs. Measured 60VAC on its shield to earth ground in the house’s outlets. How could that be? Surely the cable box on the outside of the house was grounded! Did they drive a separate spike? Good thing TVs are isolated… I doubt 60V through that little RF modulator would’ve been tolerated if it was connected to earth through the TV.
In some cases chassis will be connected to hot with one capacitor and to neutral with another capacitor of the same value. This results in the chassis at 60 VAC above neutral for a 120 VAC hot wire. Getting yourself between the chassis and ground can give you a nasty and dangerous but usually not fatal shock.
Ironically, those capacitors are sometimes referred to as safety capacitors.
Hmmm. “Grounding and Shielding Techniques in Instrumentation” by Ralph Morrison. It is a small book. I must have read it three times and every situation is still a new problem. There is a place for driven shields and for single ended shields and for single ended driven shields. Like photon counters and electrometers and gizmos I don’t recall at the moment.
Most of that is actually a question of chassis to earth, not shield. The number of books/articles I’ve read that don’t make distinctions drives me nuts.
A shield connected literally only at one end isn’t a shield: it’s an antenna, and will resonate. Does it still do something? Yes, and it can be useful (think of it like a waveguide), but calling it a shield just confuses everyone, and comparing it to the behavior of proper shielded connectors is totally wrong.
Every time someone starts talking about ground loops or potential differences with shield connections I stop them. That’s not a shield concern, it’s an earthing concern, and if it’s real, you need to think about proper isolation and safety rather than just “should I connect this.”
Robert Feranec has a series about this on youtube interviewing an expert with decades of experience in the field, well worth a watch. Yes it depends, though for small signals there’s the principle of a faradays cage ie the EMI is kept out of it if the cage is tightly sealed. But it shouldn’t be connected to signal ground, it should just float. There’s also issues with EM cavities for high frequency stuff so there are always exceptions and things to be aware of.
A shield that totally floats is not a shield. It absolutely will not have the same propagation properties. It can still do something electrically, but if you want that, you might need to take your shielded cable, connect the shield to the common mode reference on both ends, and wrap it in a second shield that floats.
If you’re doing it for ground loops/earth potential reasons, you need to think more about what isolation is needed and how to set up safety.
Shielded cables nowadays are mostly for signal propagation purposes more than EMI, which is why USB-C says “shield to signal ground, period.” Earthing questions are separate.
Sadly, I still don’t know now what to do. But this is how I intepretted it.
Always connect shield to ground via thermal relief, except:
When mechanical strength is required for smd parts. Then ommit thermal relief.
Galvanically isolated connections, such as ethernet. There it depends. In such cases, connect shield via 2 components, so you can do an rc filter, an inductor, or a zero ohm part/blob to give you options.
But what about rs485 for exaple. Its differential, but not galvanically isolated. what if it is an usb2485 adapter? What if its not 2m but 200m cable there? I have seen 485 links not wotk (reliably) without gnd.
What about usb2 usb-c links? Usb-c piwer-only? If the spec says always ground it, why is there even a difference to usb-b?
Ill redo by boards now with gnd only :) i did the double components thibg and figured i just zero ohm it. But generally the answer is to just gnd it.
Thermal relief for production part is not an issue: selective/wave soldering will have preheater and good wetting.
For prototype, with a good cartridge 100W+ solder iron (JBC and like mentioned in the previous article), it a walk in the park.
So I don’t bother much with thermal relief anymore as it’s inductively and mechanically inferior. Of course if you do 2oz+ boards with large MOS, that’s another game, but I’ve a small 35€ hot air heating table doing wonder to bring the whole board to 150°C
About the article mentioning a jumper to ground the (connector) shield, definitely not a good way (inductance and mechanically), for prototype purpose simply place 2 concentric copper ring on the pads and if needed, a large blob of solder will joint thoses with identical performance than the plain ground plane in future production.
For professionally made PCB’s that are ‘done’ and perfect sure, solid fill all the way. But for hobbiests, that make small projects, that sometimes need to desolder connectors …
Soldering is not my concern, I can solder a solid shield pretty easily with my Pinecil. Desoldering is where the problems start, and I don’t have a proper hotplate etc to do that. Just some hot air and a soldering iron :)
“Desoldering is where the problems start, and I don’t have a proper hotplate etc to do that.”
Buy a container of bismuth and mix it in to form low-melt. Cheap and you can recover it anyway. Concerns over low-strength afterwards are nuts, once you get it off easily, you can clean up the pads beautifully (just don’t use the low-melt to solder!). Practice a bit and it’ll become second nature.
It is 100% not worth sinking so much heat into a board to desolder a non-relieved connector. You’re almost certain to weaken the pad/via anyway if you try.
What is it with people and hating on thermal relief?
Inductively it makes virtually zero difference. The signal and its return already have to split due to the connector, the tiny relief voids don’t do anything.
Mechanically it also doesn’t do anything. You think the bond to super thin copper makes a difference? If something yanks on it strong enough to rip out the via and pad, it’d just tear off the connection to the plane as well. If you’re worried about mechanics, relief or secure the connector for real.
No thermal relief on vias or power connectors is one thing. Not putting them on signal connectors drives me nuts.
“Galvanically isolated connections, such as ethernet. There it depends. In such cases, connect shield via 2 components,”
No, you do what the connector tells you. You connect it to the point that you Bob Smith terminate to. If you don’t do that, what the heck are you using STP for?
The question of how to connect that point is totally separate, and that’s where the “ferrite? resistor?” etc. question comes in, not the shield. You connect shield to the common mode reference, otherwise, it’s not a shield.
Well, yeah. Differential doesn’t imply zero common mode bias with respect to earth. Ethernet, PCIe, and many other diff standards are designed to not only be galvanically isolated to remove common mode bias over long runs, but they use coding like 8b10b that maintains a zero DC bias at the receiver.
RS485 does neither of these, but it’s thresholds are high enough that some DC offset can be tolerated.
All nice and fun, except: how the heck do they earth anything in space?
A common voltage level… that’s all that really matters.
It’s way more than that. The issue is 1) the physical demands of a spacecraft means you can’t distribute a common voltage easily, and 2) spacecraft charge up constantly (and not uniformly), and so you need to make sure that anything that’s ESD sensitive isn’t going to be in a discharge path and also minimize ground differentials between points on the spacecraft.
For large spacecraft because it’s really hard to do that with a chassis-grounded-to-return setup, you often deliver power isolated and balanced (so you have a ‘high’ and ‘return’) and have each subsystem tie high/return to chassis equally through some high resistance, run it through an isolated DC/DC and tie the isolated side to chassis. So now the return from that subsystem can’t flow along chassis because that’s not part of the return path (sometimes you need to isolate the subsystem too).
I don’t know if that’s got a common name elsewhere but NASA docs often refer to that as a “balanced floating ground system.” It’s not a one-size-fits-all scheme though, NASA’s got long documents on spacecraft grounding and spacecraft charging testing.