Plated mounting holes: grounded or not?

[Solenoid]

Greetings. I'm wondering when it's recommended to ground plated mounting holes - attach them to the ground plane - and when not. I haven't found any conclusive answers on Google and I haven't studied high frequency signals to understand all the EMI problems.

I have a particular project in mind, I made an Arduino clone for my robotics club for various technical reasons and it works well with grounded mounting holes, but still I'd like to know if in case of a 16MHz MCU it's recommended to do it. As I understand ground loops should be avoided and there might be a situation where external sensors, also having grounded mounting holes, are connected to the main board with wires and both are attached on a metallic chassis effectively making a huge ground loop, might this pose a problem? Also any additional suggestions are welcome

[cellularmitosis]

You could always just run the traces to the plated mounting holes, and then allow the end user to choose between brass standoffs or nylon standoffs.

[c4757p]

I'll typically ground one so it can anchor the chassis and board to the same potential, and leave the rest floating to avoid loops. Sometimes, I don't even directly ground any, but instead connect them through large (1M-ish) resistors, to add the benefit of increased ESD protection if the metal is touched (depends on whether the chassis itself will be earthed or not).

For a robotics-oriented board like you have, I'd probably tie one through 1M and float the other. Or even tie both through 1M, if you're worried about muppets properly anchoring one and not the other (as you don't get to choose which!)

On a similar note, the shield of the USB connector is not supposed to be directly grounded. Typically you see it grounded through a 1M-ish resistor parallel with a 100pF-1nF-ish capacitor, depending on whose app note you read. The direct shield grounding goes only on the host side, to ensure return current goes through the ground wire instead.

[MagicSmoker]

This is, indeed, a dilemma. For devices that run on DC supplied by batteries or an isolated power supply I usually connect a 10nF ceramic capacitor from each incoming power supply rail to chassis ground. This allows the chassis/enclosure to act as an AC-shield, but prevents any DC ground loops.

[T3sl4co1l]

Often it is better to connect one or more to ground, via a cap, so there is RF bypass and DC (or galvanic) isolation.

Tim

[SirNick]

On a similar note, the shield of the USB connector is not supposed to be directly grounded. Typically you see it grounded through a 1M-ish resistor parallel with a 100pF-1nF-ish capacitor, depending on whose app note you read. The direct shield grounding goes only on the host side, to ensure return current goes through the ground wire instead.

This is a good piece of info.  (Hope this is useful enough to the OP not to be thread-jacking...)  What if you had, say, USB and Ethernet connectors?  Would you ground these two to the same "chassis ground" plane, and then tie that one plane to PSU ground via R // C?  Or just a single ground plane with each connector separately decoupled?

[c4757p]

I'd do each connector separately.

[T3sl4co1l]

Ethernet has its own standard, and methodology:

- The signaling is transformer isolated.  Differential pairs.
- ESD on the pins is shunted to ground (chassis ground) through a 1nF 1kV capacitor.  The isolation transformer provides center taps from the pairs, so they can be tied together and then to the cap.
- ESD on the shield is chassis grounded.
- EMC (RF noise in/out) is protected by common mode chokes (usually) included in the magnetics, the ESD cap, and the isolation barrier (fairly low capacitance -- 10s of pF range I think).
- What your circuit is grounded to (chassis or remaining isolated) doesn't matter very much, thanks to the isolation.  It would be prudent to add >= 1nF between grounds at that point, however.

USB, you can't count on the shield and power ground being different potentials, so you pretty much have to bond both to circuit ground and chassis ground.  This brings EMC right into/out of circuit ground, which may be noisy, so it can help to apply ferrite beads to the cable.  Hence, in part, why many USB cables have molded-in beads on them.

It also stands that, you don't want a huge cable running from a USB connector (the signaling isn't made to hold up through long cables, anyway), and you certainly don't want it routed through a building (and therefore subject to various possible accidental or "act of God" electrical events, besides the long-wire antenna you get).  Ethernet was invented as the all-around, do-it-all, heavy weight standard, and serves well for those purposes (routing in buildings and all).  It does fall short of general outdoor routing (like telecoms), which has to be rated for much larger surges (hence, in addition to the isolation transformer, MOVs or other TVSs).

Tim

[ajb]

Controlling return currents can be a serious problem when you've got a bunch of low voltage high current loads switching on and off--like, say in a robot  

If you've got all of your electronics on one PCB, then it likely doesn't matter a bit (for the purposes of internal loops anyway) whether you have it solidly bonded to the chassis via the mounting holes.  But if you've got several boards spread around the machine and any of them are pulling a substantial amount of current, then you need to be very careful about the return paths you create.  Choosing if, where, and how you bond your circuitry to the chassis is a part of that.  For instance, say you've got a control board connected via a cable to your motor driver, and both are connected back to the battery pack.  As the motor driver draws current from the battery, the wires that connect it to the battery pack will wind up dropping voltage (yes even the big thick wires), which has the effect of raising the motor driver's "ground" to a positive voltage relative to the battery's negative terminal.  If both the motor driver and the control board are solidly bonded to the metal chassis, then there's a low impedance path between them and the control board's ground potential will also be raised relative to the battery, which means you've now got a voltage across the ground wire from the control board back to the battery--yes, that little skinny wire you used because it was going to be carrying hardly any current, only now it's carrying some of the current that's supposed to go through the big thick wires between the driver board and the battery.  Whoops!

Of course you can also have problems with the signal wiring between different boards, which is where isolation comes in handy.

[Solenoid]

Ideally I wouldn't want to add more components, I would like to know if it's sensible to remove the ground connection from the USB shield and mounting holes. Do these things even matter at MHz speeds? I've heard EMI is really going to be problematic in GHz ranges...

[T3sl4co1l]

Ideally I wouldn't want to add more components, I would like to know if it's sensible to remove the ground connection from the USB shield and mounting holes. Do these things even matter at MHz speeds? I've heard EMI is really going to be problematic in GHz ranges...

Sure they do -- even if you're doing the lowest USB (HID class, can be bit-banged by most MCUs I believe), the bitrate might be paltry, but the transitions still contribute harmonics up to, say, 100MHz or so.

And anyway, you're concerned as much with keeping outside stuff out, as inside stuff in.  Unshielded cables invite all sorts of nasties into your signals.  And if someone touches that unshielded connector on a cold, dry winter day, after walking across the carpet...

Tim

[jakeypoo]

... for various technical reasons and it works well with grounded mounting holes ...
... there might be a situation where external sensors, also having grounded mounting holes, are connected to the main board with wires and both are attached on a metallic chassis effectively making a huge ground loop, might this pose a problem?

Can you be more specific about why it works better with grounded mounting holes?

If your sensors are analog, then that would cause more of a problem than if they are digital. If return currents from any sensor signals start flowing through the chassis you are creating a loop antenna that will pick up chatter from digital. Generally, if you have multiple boards you want to make sure that the current is flowing through the wires between the boards and not the chassis, so that should be your guiding reasoning about whether to ground mounting points.   

Electrically, you are better off leaving everything floating. You might end up blasting emissions out, but who cares it's not going to be FCC certified. Once you ground the board to chassis you have to worry about current loops and ESD getting discharged through the board.

:wq

[Solenoid]

Can you be more specific about why it works better with grounded mounting holes?

I didn't say it works better, I have no idea what it does.

So I guess as a rule of thumb there's no real reason to ground the plated mounting holes unless one knows what one's doing.

Next question: plated mounting holes or not? I guess the plating can add some rigidity and protect the PCB, does it have any disadvantages? The Raspberry Pi for example has plated surfaces, but not a via.

[Richard Crowley]

As can be seen from the detailed discussions, shielding and grounding is a complex topic. The original question is only a small part of the overall system scheme and cannot be answered in a vacuum.  If it were me, and I wanted flexibility, I would isolate the mounting holes, but connect them to the board ground-plane with a single trace that can be easily broken where necessary.

[jakeypoo]

Next question: plated mounting holes or not? I guess the plating can add some rigidity and protect the PCB, does it have any disadvantages? The Raspberry Pi for example has plated surfaces, but not a via.

It really doesn't matter. Pro is that it adds protection from any mounting pads de-laminating from the substrate. Con is that hole dimensions are easier to control using a unplated hole (tolerances on an unfinished hole are tighter than a plated one).

You can always pop vias through any annular pads to keep them from de-laminating though. Like so:

[Phaedrus]

As an example of the issues that can occur with improper grounding...

We had a 3rd party board with two inputs: +12VDC and USB. The +12VDC went to a buck regulator to power the bulk of the electronics, while the USB ran straight to the micro. The USB ground was connected to the power ground by a ferrite bead.

Unfortunately too many of these boards were connected to one power supply cable, which caused the cable to burn. Two of the ground wires lost connection, and the third darkened and went high-impedance.

This meant that the current return path for the main buck regulator was suddenly high-impedance. So the current took the lowest impedance path to ground: through that 5 ohm ferrite bead. Somewhere around 5-8 amps shunted through that 0805 1/4W part and nuked it, actually blowing craters in a couple boards. Fortunately no damage was done to the USB hub, but...



Grounding is complex.