When does a PCB require a chassis ground connection?

Question

Do all PCBs require a chassis ground connection point?

I am designing a circuit board which will be used in many different types of assemblies. It is a small mixed-signal, low-voltage (24 VDC power, 5/3.3V logic) device, low power (<0.5W). In some cases, it will installed in an all-metal enclosure, and in others it will have only a plastic enclosure with a few metal parts. The board is not directly exposed once assembled in the application. The plastic enclosure applications mean we can't rely on the enclosure as a Faraday cage for EMI purposes (and that's not my approach for EMI anyway)

Each application calls for different cabling, in some cases there will be shield that could be used for CHGND, in other cases there may be a designated wire in the cable for CHGND, or there may be no CHGND connection in the cable at all. This is out of our design control.

So I'm trying to determine if I should include a CHGND connection point on my PCB, or if it's better to simply attach all the exposed metal parts externally to any CHGND conductor in the cable with no direct connection to the board. If a CHGND conductor exists, it should be tied to PS GND back at the power source.

My thought is that the CHGND metal should act as a lightning rod for ESD events. As long as the metal has that low-impedance path back to the CHGND conductor in the external cabling, and as long as the spark-gap between the CHGND parts and my circuit is large enough, then the ESD event discharges through the metal and never comes near my circuit.

In a recent EMI compliance test, I had a board with a 1 Meg resistor in parallel with a 10 nF cap to connect CHGND to circuit GND near the power supply. Unit passed all tests until we hit it with ESD charges. I surmised that the ESD coupled thru the cap into the circuit ground and caused ground bounce, even though the metal parts were grounded through the wiring harness to CHGND. Removing the cap solved the ESD issue, and we re-ran all the EMI tests again with similar results as before (passed again).

This makes me think that the CHGND was not really active in any EMI suppression, but caused an issue with ground bounce during an ESD event. So it seems to me that I should try to just keep CHGND as far away from my board as possible. I even question if the 1 Meg resistor is really a good idea. Should I just remove the CHGND connection from my PCB designs? Obviously, I'd have to do something with a CHGND conductor that is part of a customer-specified connector, but even in that case I'm thinking I should just keep a wide spacing for that net, keep it as small as possible, and just put the 1 Meg resistor on the board in that case.

This answer encourages a super-low impedance connection between a metal chassis enclosure and signal ground via the mounting screws. Given that CHGND and DGND are not the same thing, why would we tie them together like that? Seems to me that allows any energy coupled onto the CHGND net to go directly into the sensitive digital ground and screw everything up. Why would one ever want a low impedance path between CHGND and DGND? Seems to me we want to keep the current from any ESD events in CHGND and not in DGND. The bleed-off resistor is there just to keep from getting a large static buildup between CHGND and DGND that could cause a spark. When plugged in and grounded at the PS, this isn't an issue, but when disconnected there could be a build up of charge which the 1 Meg resistor would bleed off. What other connection would one want?

Answer 1

The short answer is, the PCB should always have a chassis ground connection somewhere, having it float is not a great idea. If there is no ‘chassis’, there should be some kind of shield ground that is split from power such that not much noise can flow between them.

I’ve done products (DVRs) using both metal chassis and plastic enclosures. I used different grounding approaches for each.

• metal box: ground everything, multiple studs to mount the PCB, screws to connect shield grounds to rear panel, lock washer for RF coax F-connector. Has internal PSU.

This ‘ground everything’ approach is much like that used for desktop PCs and other IT gear. It’s tried-and-true and very robust: the low impedance of the chassis shorts out any potential ground loops.

• plastic box: hidden ground plate in rear panel for shields and F connector. Ferrites between internal power ground and shield ground to block ground noise, HV bypass caps in parallel to catch ESD and fast transient events. Has external PSU.

The ground plate and on-board shield plane are treated as a separate domain from power. This ground takes the brunt of ESD abuse and dissipates it before it can get to the power ground.

In theory I could have used the metal box with plastic standoffs and achieved good results with the latter ferrites-and-caps arrangement.