Fuse speed and sizing for overvoltage protection circuit

[Sigmoid]

Hey,

I'm quite familiar with using Zener diodes to eliminate overvoltage spikes and to get a regulated, stable power supply, but I've also seen quite a lot of designs using them as a crowbar against heavy overvoltage, combined with a fuse.

The question is, how do I size that fuse? Also, can someone explain fuse speeds (delayed, normal, fast)? Don't get me wrong, I can read datasheets just fine, I'm just confused when you might want to use each - and how exactly can a delayed blow fuse provide protection.

(On the same note, how do I size the Zener? Should it be the exact value of the maximum voltage allowed, or higher? Or a tad lower?)

[KJDS]

All fuses, in electronic terms are slow.

From memory, and it's been 20 years since I looked at a fuse data sheet, but run a fuse at 5% over it's rating and the blow time is measured in hours. Slow blow fuses are designed not to react to the spike in current when a motor is started which is many times the normal run current until the motor is moving.

A fuse is a device to prevent a fire starting when a fault occurs and they shouldn't be used for much else.

[Sigmoid]

A fuse is a device to prevent a fire starting when a fault occurs and they shouldn't be used for much else.

So basically what we want to avoid with it is the Zener diode burning out, ie. catching fire and/or becoming an open circuit, right?
The question is, how should the fuse be sized to fulfill this role? XD

[richard.cs]

Two things are important here, the let through energy of the fuse which is generally given on datasheets in terms of I²t, and the energy the zener will shunt. Some manufacturers will give you pulsed power ratings from which you can extrapolate a damage energy but this is often lacking for zeners. Zeners will usually fail short-circuit (unless you let enough energy through to vaporise the bond wires) so you can design a circuit where both the fuse and zener are sacrificial and thereby use a much smaller zener than you otherwise would need. The downside for that one is obvious.

It's common to use a small zener and either a power transistor (when you want linear shunting) or a thyristor (when you just want to crowbar the supply and blow the fuse quickly). There power devices tend to be cheaper than large zeners, but also are better specified for pulsed power. The thyristor approach also works better for mild overvoltage when a zener or power transistor might just catch fire without blowing the fuse unless they were hugely oversized. With thyristor designs it's common to use some RC filtering on the gate to prevent it triggering on short duration spikes (unless of course that's desired behaviour).

Generally when used with semiconductors you want the fastest fuse you can get as for a given fusing current as these will have the lowest let-through energy. When using this type of crowbar circuit don't use a slower fuse unless you have loads on that power supply which require substantial inrush current. Having chosen a fuse you can then pick a shunting device you think will survive, or fail comfortably short-circuit. That is based partly on datasheets, and partly on some educated guesswork.

Sizing the zener - with only a zener it should be rated such that its leakage current during normal operation is acceptable (base this on the I-V curves in the zener datasheet but remember to allow for temperature). With a zener and power transistor it is similar except you have to account for the base-emitter voltage drop. With a thyristor crowbar you need to allow for the gate-kathode voltage but you must also ensure that zener leakage current will never trigger the thyristor except in fault conditions. That usually requires a higher voltage zener than the other designs, and usually a resistor from the gate to ground.

If you have a specific application in mind we could make some suggestions as to the best approach.

[AG6QR]

I think you're looking at it backwards.  You size the fuse to supply the current which your load will need, with some reasonable margin, knowing that fuse ratings aren't very precise (see the fuse data sheet).

After you've got the fuse sized right for normal operation, you design the crowbar so that it will reliably and safely pull enough current to blow the fuse with the desired speed.  Again, see the data sheets, but you probably want your crowbar to pull at least a few times the nominal rating of the fuse.  The idea is that you want to make sure the crowbar will hold together and keep shorting the output until the fuse blows.  It would do no good if your crowbar failed open before your fuse failed open.

A typical schematic, with some explanation, is here http://en.wikibooks.org/wiki/Practical_Electronics/Crowbar_circuit

[ElectroIrradiator]

Note that there may be situations, where using a zener diode as a crowbar can cause oscillations due to current fold back in a regulator. This may prevent the fuse from blowing.

If using a 'proper' crowbar, like a Triac or SCR, it may be a good idea to put the crowbar on the input side of the regulator, fuse located *after* the filter cap, and put a reverse biased diode across the regulator. The diode prevents (further) damage to the regulator semiconductors, in case the crowbar fires. Doing things this roundabout way overall helps to prevent further damage to the active devices in the regulator, if the overvoltage condition is due to something simple, like a faulty passive component.

Some also use a (very small) current limiting resistor in series with the SCR to guarantee it and the PSU diode bridge survives intact, yet the current is still very much high enough to quickly blow the fuse.

There is a special IC for controlling crowbar circuits, if you want to get really fancy: MC3423 aka. NTE7172.

A crude rule of thumb is to use quick blow fuses for overvoltage crowbars, where the current limit is twice the normal maximum load current. Minimum current limit on the crowbar, if used, should be at least a further 10 times higher still, 20 times load current, though this does depend on the fuses used and the particular schematic.

Forgot to mention: The voltage drop across a fuse isn't quite negligible. Having the fuse on the input side of a regulator prevents it from ruining the voltage regulation of a precision lab supply.

[richard.cs]

If using a 'proper' crowbar, like a Triac or SCR, it may be a good idea to put the crowbar on the input side of the regulator, fuse located *after* the filter cap, and put a reverse biased diode across the regulator.
...

I like this idea and would recommend it. Triggers the same way, brings the supply down to within a diode drop of 0V, much kinder on the regulator.

[Sigmoid]

Hey thanks everyone, awesome info.

I think you're looking at it backwards.  You size the fuse to supply the current which your load will need, with some reasonable margin, knowing that fuse ratings aren't very precise (see the fuse data sheet).

I'm mainly thinking "audio input / output" right now (think guitar amps and effect racks, not living room hi-fi stuff).. Normally no problem, but if someone plugs something in the wrong place, or something blows its mind, then the shit can really hit the fan.

(And now imagine a vintage computer with legendarily vulnerable outputs plugged into that high-voltage crapstorm. Goal: the computer, and the sound chip, should survive.)

[ElectroIrradiator]

I'm mainly thinking "audio input / output" right now (think guitar amps and effect racks, not living room hi-fi stuff).. Normally no problem, but if someone plugs something in the wrong place, or something blows its mind, then the shit can really hit the fan.

That is not a crowbar circuit you are describing, that is AC overvoltage protection...

Audio inputs are relatively easy to protect: Use a pair of suitably high voltage zener diodes back-to-back, chosen so they are not conducting for any reasonable, expected audio input signal. The audio input impedance of most preamp/line-in circuits is very high, thousands of Ohm, and line-in typically has an input impedance of about 10K or more. So a relatively large, series resistance usually doesn't influence the signal much at all.

You could use a combination of a series current limiting resistor plus a precision 50mA meter protection fuse, and a symmetric TVS diode (actually a pair of 'super-zener' diodes connected back-to-back in a single package) connected to ground to protect each input. Use two of these combinations for balanced inputs, TVS diode from each input leg to ground.

The series resistor limits the current to something halfway reasonable, to protect the TVS diode until the fuse has time to blow. Bonus feature is that if you wire this properly, using a Kelvin (4 point) connection scheme for the TVS diode, the whole thing will also make the input virtually bullet proof. Even against rather ridiculous static and common mode voltage/current spikes.

The actual voltage limit has to be selected for each input individually. You should be able to check the datasheet of the active devices in the input circuits, combined with the schematic, to get an idea of how high an overvoltage can be allowed to get. For audio inputs I'd hazard a guess and say that in most cases you will have a fairly wide margin of error. Meaning it would take several times the normally expected peak input voltage to cause any damage.

So if your absolutely maximally expected line in signal is, say, 3V peak, about twice normal line in, then a voltage limit of +/- 5-10V is probably unlikely to cause any damage.

Special input protection circuits have been devised for mic inputs, not the least due to the combination of low input impedance and the potential presence of +48V phantom power. Check THAT Corp's homepage for some app notes on this. Search for "The Phantom Menace", part I to ?. 

Audio outputs: Suspect you may have to design the protection circuits individually. I don't recall any one-size-fits-all solution.

[Zero999]

I agree with the above.

You could also use both a thyristor crowbar and a zener so you get the best of both worlds.

Add a capacitor to the crowbar circuit so it only trips is the voltage spike carries on for a long time and let the zener protect against shorter transients.

Say you need to protect a 5V line. Use a 6.8V zener which will protect against short transients and a 6V crowbar which will trip and protect the zener if the surge continues for longer than a half a second or so.