Ucurrent surge protection

[qwaarjet]

Currently I am using 16 Ucurrents in a project. The nature of the project occasionally subjects the Ucurrents to a couple thousand volt transients which typically results in blowing the sense resistors and (surprising less) often the max4239. Right now I don't mind replacing the bad parts in individual meters, but I'm looking at going to more channels in the 50-1000 range for practical reasons I would spin a new PCB with all the channels integrated onto it instead of hundreds of individual units. With this substantially more complex system servicing the channels become more of a problem.
I'm looking for an approach to add protection but most of the tricks in know don't work at such low currents (Zeners) and burden voltage is to low for MOVs or GDTs to be able to do any good. Suggestions?

[alm]

If input protection was trivial Dave would probably have included it in the design. You could consider putting some beefy diodes anti-parallel across the sense resistor, but leakage might be an issue. I'm quite sure Dave discussed this arrangement in one of his videos (the input protection tutorial maybe?). In DMMs these diodes are generally only to handle the overload until the fuses blows. In your case they would have to handle the overload on their own.

[qwaarjet]

The anti diodes was the first thing I thought of but that math showed they would skew the signal badly. I didn't think this was a trival problem that why I was asking in what non trival direction should I look.

[Harvs]

occasionally subjects the Ucurrents to a couple thousand volt transients

Given this is a current measurement device with no ground connection, talking about voltage on it's own doesn't make sense.  What voltage are you talking about?  What current levels do you need to protect it from?

What range on the uCurrent are you using when subjecting it to said transients?

Keithley made some meters that did low current measurements with "zero" burden voltage using the standard opamp configuration, maybe their schematics might be worth a look if you can find them.

[qwaarjet]

The device is at ground as referenced to the transient. normally  only a few hundred uA  are passing through it and thus only a dV of a few mV but during the transients large currents of amps pass through if for a few uS. The range is uA

[Marco]

Is the current to be measured only one polarity by any chance?

PS. there are very low voltage TVS's like the ESDARF01-1BM2, the max leakage in the datasheet is probably too high ... but you never really know how conservative the max is till you measure it.

[Harvs]

I'd be looking at doing something like this as a starting point, you'd need to test it on the bench though.  The opamp must have very low input current, as well as offset, for this to work.  That'll cope with transients up to a hundred or so volts with no problems.  If you're going to experience transients beyond that, then having a MOV across the input would keep it within bounds.

[EEVblog]

Take the mA range for example with it's 10miliohm shunt resistor. Even using parallel diodes you still have 0.6V^2 / 0.01 = 36W if the user puts directly across a low impedance voltage source.
The only way to protect that resistor is with a polyswitch or fuse that has too large a resistance, so defeats the low burden voltage purpose of the unit.

[c4757p]

Is the current you are measuring DC? What about a choke after the clamp diodes and before the resistor? That way the transient hits the diodes first and is hopefully suppressed before it hits the resistors.

[qwaarjet]

The current is DC. I'm basically measuring the cathode current on what effectively is a vacuum tube diode however the anode which is at several KV will occasionally arc to the cathode damaging the Ucurrents.

[alm]

Some ideas: How about a parallel relay/FET triggered by a comperator? The shunt should be sized to be able to survive the peak power until the switch closes. You could put a low leakage cap across the shunt and/or an inductor in series to limit the rise time if you don't care about fast response. This only works from relatively high-impedance sources, obviously. You could also use a series FET/relay if the on resistance is low enough, in that case you wouldn't have to worry about leakage.

[Marco]

Using two 1 mOhm range power MOSFETs in series with the shunt (or it's equivalent for the feedback ammeter) could give you the necessary voltage to make the zener or diode clamps kick in even in the mA measurement range ... would add ~3 bucks per uCurrent.

[SeanB]

inductor in the line and a fast recovery power diode across the sense resistors to limit both spike current and induced noise might help. You do not need much inductance, just a ferrite bead or three in the line to reduce the first spike of the flash over, and a capacitor to bypass the input ahead of the ferrite's, or a low voltage TVS diode.