ESD protection in battery powered devices

[stmc]

A friend loaned me their ESD gun to have a play with and I was testing a battery powered signal generator (in a plastic enclosure) I designed. Everything was fine up to about 8kV but above that the unit would completely turn off.

I haven't included much ESD protection on the metallic output connector other than some series resistors. I tried adding through hole TVS diodes from the connector to the battery but I had no luck so I'd like to use some ESD diodes on the PCB itself. I'm a bit unsure how to apply them effectively though so this leads me to my question:

How does ESD protection work in a battery powered device and what should be the discharge path?

If I place ESD diodes at the output connector should I be trying to route the discharge towards the battery? Do I need to connect them to both terminals of the battery compartment or am I completely wrong in assuming that the static is trying to discharge through the battery compartment. Or do I need to split the ground plane and route the ESD current directly to the batteries because at the moment I just have one continuous plane across the whole PCB. 

Thanks in advance 

[T3sl4co1l]

Basic questions first: does your circuit have a ground plane?  If not, it's not meaningful to even discuss signals entering and exiting the board -- the whole board itself is the antenna, and you can expect variable amounts of induction between any pair of traces in the entire design!

How are the connectors bonded to the ground plane?  Same idea -- if loosely, say on wires, or with lanky traces, not big fat grounds -- you'll have a problem right there.

What good is the battery compartment?  That is, are the contacts part of a good ground plane?  Usually the battery holder is wired on leads or terminals, so I would think not.

How did you place the ESD diodes?  If on long lead lengths, well, see above.  Length is your enemy.

ESD is very fast.  It is not simply some charge bumped into the circuit -- rather, it is very meaningful to speak of it as a propagating wave.  When ESD strikes a connector, the connector becomes energized, following the wavefront of the pulse.  Everything carrying current along that path, develops a voltage drop.  This is why solid ground planes, and contiguous shields, are necessary: the voltage drop occurs on the outside of that shield, keeping the signals inside clean.

Say the BNC is mounted in a plastic enclosure and wired to the board with a coax cable.  Which sounds good enough, right?  But it's hard to make that actually coaxial the whole way.  Say the cable splits open, and shield and signal separate, to connect to solder loops on the BNC.  This leaves a cm or two unshielded at the connector.  Introduce the ESD pulse, with a voltage gradient on the order of 10kV in 3ns, or about 10kV/m at light speed, or 100V/cm.  That's 100-200V peak dropped across that short unshielded length, dropping across the ground connection specifically, and much less voltage dropped across the length of exposed signal line; the circuit sees the difference, or -100 to -200V peak, on the signal line.

If the enclosure were metal instead, and the PCB and connector are grounded to it, then the ESD washes over the enclosure, like a wave crashing around a lighthouse, while the interior remains quiet.

If you don't have a ground-plane-based circuit, it's not a fatal mistake -- but it does take a somewhat heroic effort to address.  You can get metallized (sprayed, evaporated, or metal-painted) plastic enclosures, and proper metal enclosures, and use bulkhead connectors which are grounded to the enclosure where they pass through it.  As long as the enclosure is contiguous, ESD will simply splash over it, no problem.  Or line a plastic enclosure with metal tape.

Tim