Scott Gemmell writes:
I have a question for you regarding Digital and Chassis grounding.
I've been designing 6 layer micro-controller boards running 3.3V processors with 90-133MHz system clock frequencies. All these boards contain a variety of 10 and 100 Base T Ethernet controllers, hub circuits, RS232 driver circuitry and other peripheral add-on's like PCMCIA cardholders, 56k modem modules and encryption processors.
We've had these products certified for EMC and are regularly obtaining Class B compliance the first time, however, one question always pops up, and I haven't been able to answer it yet...
The question is:
For devices powered by DC plug packs (either switchmode or linear), where no earth reference is present, what is the best way to employ a digital and chassis ground regime? Or, is one board reference only required ie: digital ground plane?
We have had problems in the past with separate chassis grounds on DB9 and shielded Ethernet connectors, and have been using decoupling caps and high voltage suppression caps to increase ESD immunity... what we've discovered is that the best solution from an radiated emissions perspective is to have no chassis ground... but does this effect the board from an immunity perspective???
Thanks for your interest in High-Speed Digital Design.
The electrostatic discharge (ESD) susceptibility problem works like this: huge current spikes come into your product through one path and leave by another path. If the pathway through your product traverses the digital ground plane, it will disrupt (or destroy) your digital circuits. That's the point (from an ESD perspective) of separating the digital ground plane from the chassis plane. All the cables and cable shields and external ground connections through which ESD currents might be expected to flow should be connected to a common piece of metal in such a way that currents circulating between them don't traverse the digital logic ground.
Now, if you can arrange ALL the connections on one side of your product, then your common piece of metal could be really small, and you might think you could get away with no other shielding of the digital logic ground. Let me tell you though, if I were performing an ESD test on such a product here's what I would do. First I'd set the product on a metal table or desk. Then I'd attach the ground terminal of the ESD generator to the desk. This arrangement establishes a capacitive connection from the digital logic ground of your unshielded product directly to the surface of the desk, and from there to the ESD generator.
Then I'd wrap a piece of foil around each cable in turn and zap it with the ESD generator. This procedure forces a spike of current into the cable, from which point it flows into your product. Once inside your product the current moves through the digital logic plane, through the parasitic capacitance of that plane to the desk, and back to the ESD generator. Many products tested in this manner fail.
Lest you think this is too unusual a test to take into serious consideration, I will point out that when ROLM corporation was working on the first generation of their digital telephones back in about 1980 they found that if a user on a metal desk draped their shiny silver-satin telephone cord over a metal in-box tray, and if the metal in-box tray had cork feet (so it wasn't grounded to the desk), then when the user touched the metal in-box tray the phone sometimes failed (it would go catatonic and lock up). The ROLM-phone scenario works just like the more repeatable experiment I just described. ESD current goes into the tray, couples through the capacitance of the tray to the phone cord, then moves into the phone, traverses the digital logic ground, and then couples through the parasitic capacitance of the digital logic plane straight the desk, and thence back to the poor, shocked user.
A good ESD plan considers ALL the parts of a system. If you leave the digital logic plane exposed (with no chassis-shield protection), then it comprises an unprotected port for the ingress (or egress) of ESD currents. Extending a solid metal plate underneath the digital ground plane and connecting that metal plate to all the wires coming in and out of the product prevents the digital logic ground plane from coupling directly to a metal desk. ESD currents will flow from the wires through the metal plate to the desk instead of through the digital logic ground. That's the purpose of a metallic chassis. Another common approach is to center your digital logic within a thick, insulating box so it can't get near anything else.
A thin, plastic package sitting on a metal desk, with wires hanging out the back of the package will prove embarrassingly susceptible to ESD.
Similar comments apply to the susceptibility of your product to nearby RF fields from cell phones and handheld walkie-talkies.
Dr. Howard Johnson