How would I safely design a crowbar for a nominal 20A circuit?

[sparkydog]

Say I have a "thing" that's rated for 20A (@ 120VAC). Say I also have a fault detector that turns on a +12V line in case of a fault, in which case I want to be able to shut this thing down automatically. (Faults are bad. If a fault occurs, turning off is more important than being able to turn the thing back on. Contained damage is completely acceptable.)

One possible solution is to include a 20A fuse and create a dead short in case of a fault. Something like the attached schematic (e.g. connecting the fault line to K1's coil).

This is going to dump an unknown but extremely high current through F1+K1. (Briefly, I hope.)

Assuming all components shown connect with screw or spade terminals using 12ga solid-core wire (i.e. none of this is on a PCB), is this safe? If so, what sort of relay would I need to use for K1? If not, how would I modify it to make it safe? (Note: K1 is not required to survive being closed, but it does need to not light on fire.)

Are there alternatives (preferably <$100) that don't rely on opening a relay to interrupt the circuit?

[rstofer]

Some circuit breakers have what is called a 'shunt trip' mechanism where a solenoid is energized to trip the breaker.

Here is a 20A 1 pole (2 space) CB by Siemens
https://www.homedepot.com/p/Siemens-20-Amp-1-Pole-10-kA-Type-QP-with-Shunt-Trip-Circuit-Breaker-Q12000S01/301895907

The available fault current at most residences is on the order of 10,000 Amps (many are more limited at 5,000 Amps).  Note that the Siemens CB is rated at 10 kA (10,000 Amps).  ETA:  That value isn't chosen by accident!

A crowbar is a really bad idea.

If the load is a motor, the motor won't suddenly stop if you remove power.  It will slow down faster if you short the windings after disconnecting the power.  Or, you can clamp the rotation like they do on the Sawstop table saw.  The blade stops so fast that amputation is highly unlikely.

https://www.sawstop.com/why-sawstop/faqs/   See first query...

[BillyO]

Most crowbar circuits I've seen use a triac (thyristor).  You could use a 100A sloid state relay.  It would respond faster than a mechanical one and would generally be able to take transients of around 600 amps.

[Terry Bites]

A chunky triac or SSR will happily crowbar your rails. But it needs to be able to stay in its SOA until the fuse has blown.
That's very risky on a line powered device.
You can DIY a monster low cost SSR with a zero crossing photo-triac eg MOC3061 and a 40A triac.
You will need a series fuse in series with the triac in case of metdown.
I'd make sure that your control is fail safe and latches off at power down and after an 0ver Volt.
So an old fashioned relay would be a good idea.

[ejeffrey]

I totally get the appeal of a crowbar. Maybe not a bad option given your requirements, but if at all possible shutting off power is preferable to shorting it out.  A crowbar dramatically increases the current that the fuse needs to interrupt, adds stress not only to your device but to the power cord, outlet, and upstream wiring.  In theory it all should survive if rated properly but it's generally unnecessary.

The point of a fault detector + crowbar is specifically to interrupt current when the fault current is too low to blow the fuse directly in an acceptable amount of time.  But this is exactly the situation where a solid state relay can disconnect power with no problem. 

If you want to short out the load after removing power you can do that with any suitable switch (TRIAC, SSR, mechanical relay).  You can power the trigger circuit from a capacitor or from the load depending on your situation.  This circuit can also provide a secondary backup crowbar in case there is a fault that is too low to blow the fuse but coincides with something that damages the MOSFETs in the SSR causing it to fail short-circuit.

[pqass]

FYI: see https://sound-au.com/articles/e-fuses.htm for an article on e-fuses.

Specifically, see the attached Figure 5.2.2.
Q1 can be an ordinary (non logic-level) N-channel MOSFET that can take the continuous current; like TK090N65Z,S1F (650V 30A 90mR 230W).
Same for the diode bridge (D1..D4); like GBJ2506 (600V 25A).  I would suggest at a minimum both be rated at least 400V.
R1 is tuned so that 20A through it will produce between 500mV to 800mV to trigger the SCR; which is going to be 0.025R to 0.04R, respectively.
R2 has to be low enough to maintain the SCR when latched (1K @ 12V gives 12mA which is greater than 5mA holding current).
The [isolated, unregulated, smoothed, DC] power supply has to be greater than 10V to make sure the MOSFET is fully on and can be as high as 24V, although R2 should be raised to lower its power consumption (maintaining 12mA through it).  It could even be a battery with low voltage cut-off.
Between the shunt, MOSFET and diode bridge there will be about a 3.5V drop which will consume about 70W@20A.

Q1 datasheet: https://toshiba.semicon-storage.com/info/TK090N65Z_datasheet_en_20181126.pdf?did=63656&prodName=TK090N65Z
SCR datasheet: https://media.digikey.com/pdf/Data%20Sheets/NXP%20PDFs/BT169_Series.pdf
Diode bridge datasheet: https://www.smc-diodes.com/propdf/GBJ25005-GBJ2510%20N1754%20REV.A.pdf

[sparkydog]

@rstofer, yeah, the bit where I'm trying to not spend an arm and a leg, or wind up with something overly complicated, got lost. How do you activate a STB, anyway? Does it need a low voltage, possibly DC, signal, or just a modestly-rated relay to connect mains-L to mains-N via the trip? (Why are those things so bloody expensive, anyway? European models seem to be a fraction of the cost.) What's really awful about that is I'd need it in addition to the GFCI/CAFCI breaker.

I'll try to keep your other point under advisement if I ever hook a motor to this, but those safeties would presumably need to be on the motor-powered thing. (Strictly speaking, what this is powering is unspecified. Practically speaking, though, probably not anything motorized, or at least not anything where fast stopping of something involving significant kinetic energy is a safety requirement.)

@ejeffrey, I guess I should specify, the main potential fault I'm trying to guard against here is that I tried to shut off power, but power is still flowing. What do you do, for example, when you need to guard against your SSR failing short?

[ejeffrey]

I see.  I assume you don't want a mechanical switch because you always want it to turn off when power is removed?

You said you didn't want to use a relay.  Why is that?  Safety switches on heavy equipment and the like is often made using a relay with a 120 VAC coil and a pair of momentary on/off buttons.  The relay uses an auxillary set of contacts to energize it's own coil, and the momentary buttons force engage/disengage it.  You can use simple circuits to give a time delay as well, where you have to hold the "on" button for 2 seconds before it powers up if you want.  As long as the relay contacts are adequately sized, this is quite reliable and dead simple.

A similar approach is a magnetically latching mechanical switch.  This is basically the same mechanism on most modern toasters: the bread lever is not mechanically latched.  It is retained by a small electromagnet that is turned on when the circuit is energized.  It can be turned off normally by the timer, but if you remove power the magnet will turn off and a spring will push the switch will open.  This approach works only if the load current will always be large enough under normal operation to keep the latch engaged. 

If neither of these options are suitable, a crowbar should work as long as everything is properly (over)-sized for the fault current.  Keep in mind the failure mechanisms here, including "the crowbar (or other conditon) blows the fuse and some dumbass replaces it with the wrong rating"

Whatever option you use, an indicator lamp that is directly in parallel with the load will tell you that the load is on, even if the switch fails, or if the user fails to operate it properly.

[rstofer]

@rstofer, yeah, the bit where I'm trying to not spend an arm and a leg, or wind up with something overly complicated, got lost. How do you activate a STB, anyway? Does it need a low voltage, possibly DC, signal, or just a modestly-rated relay to connect mains-L to mains-N via the trip? (Why are those things so bloody expensive, anyway? European models seem to be a fraction of the cost.) What's really awful about that is I'd need it in addition to the GFCI/CAFCI breaker.

The shunt trip coil voltage is usually 120V - there are usually 2 connections for the trip coil and you just hit them with 120V.  Those that I have used don't depend on the output voltage of the CB, it can be dead due to some upstream device but if the trip coil is hit with 120V, the breaker trips. There are variations when it comes to big breakers (48V DC is popular, so is 48V AC)

There are also under-voltage release breakers that trip when incoming power to the trip coil is removed.  They can be a nuisance.

How and where you get the trip signal is a separate topic.

[rstofer]

It certainly doesn't reduce the total cost but there are ground fault sensing relays intended for use with external shunt trip breakers.  On the 5 mA setting it will perform as an ordinary GFCI device.

https://www.automationdirect.com/adc/shopping/catalog/sensors_-z-_encoders/ground_fault_sensors/gfs30-d1c-120a-f?

Either the sensing relay or the external circuitry can energize the shunt trip coil.

[sparkydog]

The shunt trip coil voltage is usually 120V - there are usually 2 connections for the trip coil and you just hit them with 120V.

Okay. I was somehow under the impression they were mostly 12V/24V DC, which would make triggering the thing... complicated, especially if I want more than one means of tripping it. (Basically, I'd need a separate PSU dedicated to the STB.) If all I have to do is close a relay to dump live from that breaker back to the trip, that's much easier, and also trivial to have parallel paths. (Still needs separate relays, but that's hardly the end of the world...)

I assume you don't want a mechanical switch because you always want it to turn off when power is removed?

You said you didn't want to use a relay.  Why is that?

Did I say something about mechanical switches? I don't need "turns off when power removed"; in fact, since the fault detection line is normally-open, I have a mild preference for not adding another continuous power draw. (But "latching" behavior is also desirable, i.e. when the crowbar / "fault trip" / whatever is activated, it shouldn't automatically reset.)

The reason I don't want to rely on a relay is because I have relays already. They're the normal mean to turn the "thing" off. I'm looking for a backup, and particularly, one following the principle of not relying on a single class of device. (BTW, you're aware that the "magnetically latching mechanical switch" you described is... an EMR? 🙂)

[rstofer]

Most shunt trip breakers have an internal coil clearing contact such that there is no damage if the trip signal remains present long after the breaker has cleared.  The mechanism is such that once the trip coil is energized, the coil voltage can be removed by the breaker itself.

I don't know anything about the Siemens CB in particular but I would want to know whether the breaker can be reclosed momentarily, with a standing trip signal.  I couldn't do anything about it but I would want to know if the breaker was going to close for a cycle before re-tripping or whether it wouldn't close at all.