Idea: Low voltage crowbar

[FrankenPC]

I have a current limited bench supply and I would like to make an external variable voltage crowbar device for it.  There are a ton of crowbar circuits out there based on voltage shunt IC's and SCR/thyristor/triac combos.  But they all have a relatively high Vref demand of at least 2.5V. 

I would like to go down into the < 1V range if possible.  Do I need a smart crowbar?  For instance, how about a battery powered microcontroller with ADC input triggering a SCR?   It seems to me that the floating gnd on the power supply would create a dangerous environment for anything controlling the SCR.

Looking for suggestions/warnings.

A good page with the most common crowbar variants:
http://axotron.se/index_en.php?page=26

[T3sl4co1l]

Those circuits are good, with one mod: I would include hysteresis, at least on the "fancy" transistor'd versions.  10mV back to Vref or the "anode" (whichever is easiest and works correctly) would be enough.  That way you don't risk malfunctions on slowly rising supplies.

TLV431 is 1.24V, so there's that.  There are chips of various flavors with internal references from 100 to 800mV, though they obviously(?) require additional voltage to function.

You need at least 0.8V to guarantee SCR turn-on, and about as much to maintain current through it as well.

A MOSFET can pull down lower, but requires positive gate voltage.  Or if you're lucky enough to find a depletion mode device or JFET that big, but...beats me if such is available.  Even so, you would then need a substantial negative supply to provide reverse bias when it's supposed to be off.

If you can have a bias supply for it, it'll do well.  The purpose of a crowbar is to shunt *excess* power, which implies you have an excess available somewhere..

Tim

[FrankenPC]

Those circuits are good, with one mod: I would include hysteresis, at least on the "fancy" transistor'd versions.  10mV back to Vref or the "anode" (whichever is easiest and works correctly) would be enough.  That way you don't risk malfunctions on slowly rising supplies.

TLV431 is 1.24V, so there's that.  There are chips of various flavors with internal references from 100 to 800mV, though they obviously(?) require additional voltage to function.

You need at least 0.8V to guarantee SCR turn-on, and about as much to maintain current through it as well.

A MOSFET can pull down lower, but requires positive gate voltage.  Or if you're lucky enough to find a depletion mode device or JFET that big, but...beats me if such is available.  Even so, you would then need a substantial negative supply to provide reverse bias when it's supposed to be off.

If you can have a bias supply for it, it'll do well.  The purpose of a crowbar is to shunt *excess* power, which implies you have an excess available somewhere..

Tim

Thanks for the tips.  Actually, in real life scenario's, 1.25V is low enough.

[T3sl4co1l]

It would be nice, in theory at least, to have some sort of solution for protecting 1V core rails, I suppose.  Even just 1.8-3.3 (say, for higher speed / performance embedded) is annoying because there's no such thing as a zener TVS under 5V (or rather, they don't make them, because they're too soft -- too leaky at a useful voltage, or not sharp enough to do a useful job).

Although in practice, I suppose deep inner bits like Vcore (0.9-1.8V at mA to 100A+) are usually fairly well isolated from the great outdoors through one or more stages of SMPS.  Still, it would be hard to protect the rail against a faulty SMPS (suppose the high-side MOSFET fails shorted for some reason).

It's a good question, worthy of good solutions.

Tim