Fast Self-Resetting Fuse

[absentwizard]

I need a self-resetting fuse in my system, up to 40V, 100A. I've been looking through some of the hybrid-metal devices, but they all take ages to trip (5 to 15 seconds, generally). I wanted to bounce a thought that I had off of someone, in case I should just check myself into a madhouse now.

It is possible to make a current monitor output to a Schmitt-trigger comparator feeding into J line of a JK flip-flop with the K line pulled up to positive supply and Q sent to a big FET or IGBT? That way it’ll trip much, much faster and will autoreset once the current is gone, but it may cause serious hunting behavior. How to mitigate the bang-bang?

[Soeren]

Hi,

It is possible to make a current monitor output to a Schmitt-trigger comparator feeding into J line of a JK flip-flop with the K line pulled up to positive supply and Q sent to a big FET or IGBT? That way it’ll trip much, much faster and will autoreset once the current is gone, but it may cause serious hunting behavior. How to mitigate the bang-bang?

Just add a delay on the reset, if you really want it to be auto-reset. I'd think a manual reset might be better, as you'll usually have to check what caused it to blow.
A fuse that just resets itself, doesn't make much sense, as there'll be a reason it blows (or it's set too low and needs to be adjusted).

To switch off a large MOSFET or IGBT fast, you need a current capable driver in between the logic and the switching device.
What voltage will the fuse be handling?

[absentwizard]

Up to about 34V, but I felt that targeting 40V for margin would be suitable.

The reason for the auto-reset fuse is because I expect this unit to be subject to transient shoot-throughs and short-to-chassis. It needs to protect against those, but at the same time, it must not require a fuse replacement every time such an event happens. Those events happen because of vibration and EMI issues that must be tolerated.

[Soeren]

Hi,

Up to about 34V, but I felt that targeting 40V for margin would be suitable.

Sure, that shouldn't be a problem then.

The reason for the auto-reset fuse is because I expect this unit to be subject to transient shoot-throughs and short-to-chassis. It needs to protect against those, but at the same time, it must not require a fuse replacement every time such an event happens. Those events happen because of vibration and EMI issues that must be tolerated.

Not meant to offend you, but that sounds like a very sloppy design; tolerating shorts and not eliminating/dampening high transients - such issues should be dealt with, rather than patched up with an auto fuse IMO.

However, that's your decision of course, but what would you rather have - a manual reset or a delayed reset of ?? seconds?

An SR flip-flop and a monoflop made from a single 4093 could be used for a delayed reset and half that for a manual reset.

[absentwizard]

95% relative humidity at up to +60 C, thermal shock, 5% salt fog for 2 days, up to 0.3 g^2/Hz vibration spectra, 45cm drop impact survival, 2.5 kV voltage spike survival and self-recovery, explosive decompression to 40,000 ft, 20 V/m RMS EME 150 kHz through 2.5 GHz (fortunately, don't need to meet the 50 kV/m pulse requirement)

It's not necessarily the system design. It's the environment that it's operating in. It must not be a manual reset and I think reset in 1 second will work for known spike profiles.

[Soeren]

Hi,

It's not necessarily the system design. It's the environment that it's operating in.

In other words... If the auto fuse fails, you blame the environment rather than me

It must not be a manual reset and I think reset in 1 second will work for known spike profiles.

Given the noisy environment, I don't think a CMOS circuit would b the best idea, but this should work with proper layout and shielding/conformal coating etc.
I've kept the impedances low, to bolster it against noise, but that's not gonna cut it without shielding in the environment that you describe.

Since I don't know how you will be sensing the current, or if that's already in place, I haven't added that. I'll recommend a LEM sensor (Hall current sensor) if you haven't chosen a method already.

The two input resistors should to be calculated based on the output from the sensor, so that's left to you as well.

This is using a low side MOSFET. If your app. needs a high side switch instead, it can be changed, but will probably need 2 (or more) parallell devices then, as I am not aware of any P.ch. MOSFET with a rating of 100+A at an appropriate voltage.

The FQA140N10 I used is a 100V/140A device ("only" 99A at 100°C though) with an R_DSON of 8mOhm (typ) 10mOhm (max). Even though this device is a brute with brass knuckles (560A pulse max), I'll recommend a good snubber network to limit dV/dT and keep large transients out of it's way, if the load is even the slightest inductive - MOSFETs tend to die screaming from even moderate overvoltage kicks (and I probably don't have to tell you that inductive kicks can generate humongous voltages when large currents are paired with fast switching).