I am currently working with solutions to pass the ESD-requirements for EMC. I have used TVS (ESD)-diodes in the past to good result, but never really understood exactly why they work. Sadly, this is one of those topics where you ask 3 different electronics engineers and get 10 different answers. But I'm going try anyway :)
First, my system: I have a system where I have digital electronics supplied by an AC/DC PSU. The power comes from 230VAC L+N combination (Europe) with a separate PE (Protective Earth, the yellow/green wire) from "the socket outlet". This PE is connected to the chassis of my device. The digital electronics is supplied by the 24VDC with a separate GND connection (0VDC, black cable) from the PSU. I have not connected the the GND/0VDC and the PE myself in any way. I'm fairly certain they are not connected within the PSU, except through maybe a resistor and a Y-capacitor. Also, most signal lines have an RC-filter (100R-10k + 10nF-100nF) for the signal. I imagine these will protect the signal to some extent as well?
As for my questions:
Where does the energy from the ESD event go? I know that the TVS diodes (Uni-dir) will start to conduct in reverse when the voltage gets too high (<Vbr). This creates a low-impedance path for the energy through the diode. However, where does that energy go next? I've seen people claim it's dissipated in the diode as heat, though does this actually happen if the path is low-impedance? Shouldn't it actually continue until it reaches whatever it's referenced to (which in this case would be PE, since that's what the ESD gun is referenced to). If there is no path directly from the signal line to PE through the diode, wouldn't it then just pass through the GND-plane on my PCB, causing all sorts of bad things? I can imagine that it will eventually reach PE through the Y-capacitor in the PSU.
How can I create a shorter path for the event to PE? If we assume that the ESD event is not fully dissipated in the diode, is there any good way to create a shorter return path to PE? I'm of course referring to the broader question "Can/should I connect GND to PE and how?". What safety implications do the various methods have (for example if you rely on PELV/SELV protection against electric shock)? How might it impact radiated emissions (you risk creating ground loops for example). I have a couple of different methods that I'm unsure what impact they would have:
2.1 Connect directly (PE<->GND) My guess is that this method will probably protect well against ESD, but might risk causing ground loops and that noise is highly exported everywhere. Probably not a good idea.
2.2 Connect through R+C
This connects GND and PE through R+C in series. I imagine this will have the effect of only allowing high frequencies to pass as well as limiting the current through the system. This is similar to the connection in a PSU. I imagine the downside here is that it won't react "as fast" due to the resistor creating a higher impedance path. Here you can of course select the capacitor and resistor accordingly. My guess is that they must be able to handle high voltages and that their values should be fairly low (C=1nF-100nF, R<100ohm), but this is just handwaving. If I recall correctly, the ESD event is discharged from the ESD gun with a 330ohm resistor (thereabouts), so anything lower than that should be ok? Maybe someone could shed some light on if this is a good idea, and if so, what values to choose.
2.3 Dual TVS diodes
My colleague suggested this idea. It's basically two TVS diodes connected in reverse towards eachother. This will block any connection between PE and GND until the voltage difference gets sufficiently high, which will only happen during an ESD event (Or something else going horribly wrong). If the Vbr for the diodes is selected sufficiently high, I imagine noise would not get through, though an ESD event will. Since the voltage difference normally would be very low (ideally should be close to 0), the Vbr for the diodes can be selected rather low. My questions regarding this: Is it a good idea? I've never seen anyone else do this before. What Vbr is good to aim for? Is this connection using 2 uni-dir TVS diodes equivalent to using a single bi-dir TVS diode? Most often the symbol is drawn the other way around, but I have seen TVS diode symbols been drawn this way too (compare the SMAJ series with the SD-C series).
To summarize: Project with metal chassis, supplied by AC+PE. AC/DC PSU supplies low-voltage boards. Boards are only connected to GND/0V, while chassis is PE. How to protect the board properly from ESD-strikes? Should GND and PE be connected and if so, how? Where does the ESD energy actually go if you have TVS-diodes?
