I'm doing some EMC work right now, and created something weird. Has anyone seen it before?
Description:
Suppose you have a shielded cable being terminated to a header. So, you tie the shield to a pin, and ground that on the board. The rest of the wires are signal, power, ground, that sort of thing. For purposes of example, say it's CMOS logic level (impedance controlled for signal quality over the cable, but otherwise pretty boring).
Well, it fails EMC susceptibility, because the header and connection isn't shielded. That one inch of unshielded cable is enough to drop more than V_IH(min) - V_IL(max) and screw up the logic thresholds.
Or if it's differential (say, RS-422/485), the difference might be okay, but the common mode either bangs into the supply rails (transmitter stops transmitting), or hits clamp diodes / TVSs, or exceeds the receiver common mode range (which is usually a few times supply due to internal resistors, but still finite).
So that's the background. Follow?
Accepting that we can't perfectly shield the whole connection, suppose we take half the length of the shield / drain / ground wire, and sense the voltage across it. Ideally, this length has half the voltage drop that's causing our problems. So, run that into a 1:2:2 transformer, where all the signal lines go through the '2' windings. Maybe we just use a ferrite bead, one turn for the ground, two turns each for all the signals.
So...you've got a shorted turn around the ferrite bead, that's got shield/ground current flowing across said turn. And the signal lines make two turns through.
And it works. I just tested a 100%+ improvement doing this -- assuming, of course, that "sense" short is very nearly 50% of the total (or whatever ratio the transformer is).
So, good luck cutting all your cables so they have exactly the right measurements. It seems to be a neat trick, but not something you'd be able to put on cable ends reliably. But all I want to know... has anyone even seen this before?
Tim