开云体育

Tom, if you study some of the old Keithley electrometers, you may find the two E-B junction clamp. I believe it was sort of a trade secret, developed when they started using MOSFETs (the type was probably a trade secret too), instead of electrometer tubes. In nearly every one (a lot) that I studied, the clamp circuit is shown as a black box. Later, the MOSFETs and clamps were combined on a small board, that could probably be figured out by inspection, by going through the trouble and risk to remove it. The sub-parts were never identified - the whole module was to be replaced if bad. I finally stumbled upon it about a year ago in a schematic of something they made, actually showing the details. I'm pretty sure I remember it right, because I had always assumed it was just some low-leakage diodes or JFETs - it only needs to protect the summing node MOSFET gate which operates right close to zero, and keep it from breakdown, which is fairly high - and I was surprised that it was a little different.

It actually makes sense though, and is pretty slick when you think about it. If you use a single diode junction for each direction, it's forward biased with any excursion of the summing node away from zero (usually up to a mV or so in operation), so even though the diode may have a very small saturation current (Is), it's still forward biased, and can have a big effect in a fA circuit. If you use the diodes in reverse bias, against some DC clamp voltages, then you have to deal with diode reverse leakage. By using two transistors, putting the forward clamp in series with the reverse breakdown junction, you get that extra blocking (the breakdown junction will see only very small reverse voltage due to the diode's forward current, so will have tiny leakage), and there's virtually no leakage until the excursion is well past the operating level, where it doesn't matter anyway. The end result is a clamp with extremely tiny leakage near zero voltage in normal operation, which starts to leak a little more as it leaves the normal range (don't care), then clamps solidly as the Zener level (say 7 V total) is approached, where you do care, to keep it well below the MOSFET gate breakdown, typically 20-30 V.

If you have a circuit where you need to hard-clamp at one junction, which is very common, you wouldn't use this kind of circuit. But, for these MOSFET inputs, it's awesome. Also, the whole thing is a passive, two-terminal device, so easy to apply. No wonder they kept it a secret. I'd guess that the surface body leakages on the transistors and MOSFETs is a much bigger problem than the clamp leakage itself - that's how good I think it can be.

I have seen the same or close used in some DMM circuits, so others have used it too. If I can re-find any of the definitive info, I'll post about it. If it turns that out I've imagined all this, I'll post about that too.

Ed