The outputs have about 100uF ceramic capacitance each (the peripherals have large in-rush current, and we can't change this; the capacitance supports this.) HBM ESD is 100pF + 1.5kohm series resistance at several kV. If I assume the loading is negligible (true for open circuit connector) then worst case at HBM 8kV is the 100uF could charge up by 8mV. Negligible, right? Even if the 100uF goes open circuit somehow, there are still 2x100nF caps, which would give worst case 4V peak. That is still 'safe' for the output load switch, within the abs rating of about 18V, although some energy would probably be conducted downstream (switch does not block reverse current), there is further capacitance there which would adequately absorb this... So it does not immediately appear to be a risk. I already know our buck-boost converter tolerates the output being driven above its set point (provided this is within the maximum switch voltage), it just stops switching, so that also seems OK.
Right, the impedance absolutely stomps anything like that.
But, key word "impedance". It's assumed to be capacitance, but it has ESR and ESL too, and there may be other ESL or transmission line effects in the surrounding circuit, too.
If the layout is careless (i.e., not a solid ground plane), the common mode surge could disturb other things connected to it, like, imagine there's a current sense shunt resistor in there, it has nonzero inductance so both CM and DM on its Kelvin sense leads get a fraction of that incoming ESD. And if that's going to a current sense amp IC, maybe that will cause problems.
Note this is also what makes supplies useful to absorb ESD. You can use clamp diodes because the supply has some ~uF on it, and the peak overshoot in the local area isn't going to be more than a few volts, easily handled by a small TVS.
Note also, the low impedance means, when it does go over voltage and cause something to break down (preferably a TVS, but failing that?), it's not 1.5kohms anymore, it's... milliohms or whatever. Which makes piddly little clamp diodes useless.
So it's easy to see why >= 10 ohms is often seen in series with a signal path, between transceiver and TVS diode. (Whatever the defacto transceiver happens to be in the case: a bare MCU or logic pin, an RS-232 interface, an op-amp output, etc.)
So, given good layout, a power supply can have quite low enough impedance not to care. But given poor layout, well... it can always be worse!
Are spark gaps really effective and dependable for ESD protection? As you said, "the initial edge of the ESD pulse may have a rise time less than 1ns". But spark gaps have a turn-on time on the microsecond scale. The rise time is fast (subnanosecond), but not the response time (microseconds, from my experience with 70-volt gas-discharge tubes, I assume free-air spark gaps are even worse). The spark gap may not even have the chance of turning on long after the ESD has ended...
[/quote]
Maybe. But it needs to be done in a place where the voltage drop will be high enough that breakdown is ~instant.
Note that it might even make things worse. The last project I tested (personally, ESD gun in hand), I found I could upset the equipment at a much lower level with a certain technique. That technique turned out to be, hovering the nose a ~mm or so above the target: presumably, the nose charges up to near rated voltage, sparks to the target, and then discharges. This, to a certain extent, bypasses the impedance / networks in the gun -- the nose and surrounding material discharges ~instantaneously (fraction of a ns) into the target. In other words: I made a pulse-sharpening network.
So it won't help you much in terms of required filtering, to keep that EMP away -- indeed the EMP could be worse in terms of dV/dt, dI/dt or bandwidth, if not in terms of total energy or peak amplitude. You'll still need good filtering or shielding to keep that out. (But, obviously that's a lot easier to do when you have the spark gap somewhere local in circuit, so you can plan all the filtering needed around it.)
Anyway, that's the primary difficulty, the breakdown: if you don't have a high voltage drop, it won't spark, or not right away, and most of the energy continues down the line. It simply won't fire at all if the wire/trace is short and loaded with a TVS or bypass cap. And notice you can't simply put a chip resistor or ferrite bead in series to increase the impedance locally (or for short periods of time): the resistor most likely sparks over, and the ferrite bead saturates almost instantaneously (a typical 0603 chip saturates at ~50mA, it's only a few ns before the ESD is dumping multiple amperes).
You can potentially get chip resistors that are rated for high voltages -- they're utterly useless without potting, of course, but it's interesting that it's an option at all. That would do when you can afford say a kohm here or there, like on medium bandwidth digital or analog inputs.
Other than that (medium to high Z inputs), I think you'd have a hard time finding use-cases for spark gaps. Especially the PCB kind; voltage is just so completely unreliable. Even the component kind are pretty gross (50% tolerance?). GDT are alright (and available in much lower breakdown voltages than air gaps, even accounting for the short time scale).
I, too, have definitely seen sparking in ESD testing; besides the above example (which pretty obviously will spark, it's acting in series with the gun), I've seen it across (or within(!)) common mode chokes, for one. Which often are made with spark fingers on the PCB, to prevent this damaging the wire. (The spark gaps act in parallel with each winding.)
And yes, while a ferrite bead doesn't, a CMC does have enough inductance and saturation flux to manage to drop at least some of an ESD pulse. Still not enough to filter the whole thing, you'll need a much bigger than average one do that -- but certainly enough for spark gaps to be relevant around them.
Tim
Home
Help
Search
Login
Register
