Using something like an INA219 as an electronic fuse?

[e100]

Datasheet https://www.ti.com/lit/ds/symlink/ina219.pdf
Would using something like this as the basis for an electronic fuse be considered a good/bad/terrible idea?

The frequency response looks flat to 200 Hz so this would presumably give a response time of about 5ms, or thereabouts?

[ledtester]

How are you planning to use it?

You would need an MCU to continuously monitor what the INA219 was sensing. If your firmware hangs you lose your protection. Also, physical fuses don't act instantaneously -- they only blow when the over-current condition is present for a some duration and this complicates your MCU logic.

[perieanuo]

long story short, it's bad
any problem you have in i2c/smbus or in your code, the protection is gooooone baby gone
the short protections should be quick reliable that bein said, analog

[T3sl4co1l]

What kind of fuse response do you want?  The three questions to ask are always:
1. What clears the fuse -- overcurrent, fault current, energy, time, etc.?
2. How much current shall be permitted?  At some point, any semiconductor switch you use, will be yanked out of saturation.  Is that healthy for it, and will it survive given (1)?
3. How does it trip and reset?  Automatic, time or temperature?  Manual?

Incidentally, a common naive mistake in the analog implementation is to use a comparator, and nothing else: this seems like a consequence of sequential thinking, "if <current exceeds threshold>, then <turn off the switch>" -- sounds convincing, even, but once the switch is off, current immediately falls below threshold and turns back on, right?  Oh...

Not that this would be a problem with a digitally controlled sensor, or any more than usual in a sequential (programming) environment.  1-3 are still fully applicable, however.

Remember, semiconductors die on the order of 100µs.  (Conventional) fuses, 10ms.  If the response rate of that sensor is around 1ms, what pray tell are you going to do for the other nine semiconductor deaths?  Well, you better have an analog solution to keep it alive in that time.

Tim

[Berni]

Yep its a bad idea.

You want an efuse controller chip to directly control the mofset so that it reacts quickly. When a direct short circuit happens the current can get very very high in a very short time.

Typicaly efuse controller chips also work as current limiters. So once the current goes past a value they start closing down the transistor to limit the current. This makes the fault current spike smaller. But this is not to be nice to the load, it is mostly there to avoid too large of a current building up on the switching transistor since at some point the current might become so large that even if you turn it off quickly the avalanche energy rating gets exceeded and blows up the transistor. Since semiconductors typicaly blow into a short circuit state this means that the fuse is now effectively shorted and letting all of the current trough it. By limiting the current it ensures that the current can never get too high for the transistor to be able to interrupt.

But you can typically reuse the same current shunt for the efuse controllers current sensing as well as your own current measurement by simply connecting this I2C chip in parallel with the efuse shunt. The efuse controller can also typically be told to turn off the transistor on command, so you can still turn the power on and off using a GPIO pin on a MCU for example.

[Renate]

I'm a fan of the INA219 & INA226, but this is a case where you want a direct output.
TI makes a bunch of chips in this series with analog outputs.