# How to design over voltage protection Crowbar Circuit?

I'm designing a Crowbar over voltage protection circuit with following specifications,

Maximum input voltage - 30V

Maximum input current - 5A

Output voltage - 5V

Output current - 5A

So I have chosen zener diode (1n4733A -> Vz = 5.1V,Iz = 49mA) and (2n5060 -> Vgt = 0.8) components for my circuit.

According to the calculation I choose R = 16.3265ohms resistor. But my simulation doesn't work as I expected.

• Vijan, what do you imagine as the current running through the SCR in the first case? (You should be able to rather easily work that out.) Is it rated accordingly? Also, you should be aware that an SCR can be fired with as narrow a pulse as perhaps a microsecond or less. Why haven't you included, at the very least, a capacitor across R2? (Not that I agree with your value for R2 -- but that's another issue.) Finally, even assuming you add a capacitor, your schematic shouldn't ever be used. It's very poor for a variety of reasons.
– jonk
Commented Aug 21, 2019 at 7:10
• I would recommend you to use TL431 instead of zener diode for a crowbar circuit. You can set threshold level much more precisely. Commented Aug 21, 2019 at 9:49

## Explaining the Simulation Results

You don't show your calculations. That's a problem. You compute resistor values to almost insane precision. That's a problem. You don't take into account that an SCR might fire with $$\\lt 1\:\mu\text{s}\$$ runt pulse, so your circuit is also probably likely to be highly sensitive to noise if you ever built one. That's a problem.

But perhaps the most serious problem of all is that you can't figure out why there might be $$\10.2\:\text{V}\$$ at the SCR anode when you expect it to be triggered by an over-voltage condition. A very simple calculation would show you that there is almost $$\1800\:\text{A}\$$ of anode current in the SCR!! Good thing you are just running a simulator!!

Here's my thoughts about the first schematic's simulation results. You have an anode voltage of $$\10.2\:\text{V}\$$ with a supply voltage of $$\12\:\text{V}\$$. This means there is $$\1.8\:\text{V}\$$ across your $$\R_1=1\:\text{m}\Omega\$$ resistor, which implies that $$\I_{R_{_1}}=\frac{12\:\text{V}-10.2\:\text{V}}{1\:\text{m}\Omega}=1800\:\text{A}\$$. Most of that's going through the SCR!!! What do you expect the simulator to do with all that current, given the SCR model it is likely running with? I'm not at all surprised that the anode voltage simulates out at $$\10.2\:\text{V}\$$.

In your second schematic's case, I'm sure it is already operating the way you expect. The output voltage, across $$\R_3\$$, is very close to the input voltage. So no comments from me on that. You already understand that result, I suspect.

## Solving the Problem

I can't tell you how to solve your problem because you've not clearly stated it. I can tell you some things you are not accounting for and therefore some directions to consider further.

1. Don't specify over-precise resistor values which imply something that isn't true. You don't need that kind of precision in a crowbar circuit.
2. Recognize that voltage noise might readily trigger your SCR. Do something to manage that possibility.
3. Recognize that zener diodes not only have wide tolerances in their break-down voltages, but also that they have an especially soft knee at low current levels. Used in your circuit this way is likely to lead to excessively wide tolerances in the trigger voltage for your crowbar circuit.
4. Recognize also that SCRs have wide tolerances in their gate firing current and voltage values.
5. Note also that SCRs do have a limited melting integral, or $$\I^2\,t\$$ specification. Exceed that and they will explode or fuse-over. For example, you cannot expect a 2N5060 to operate with $$\10.2\:\text{V}\$$ between its anode and cathode while also passing nearly $$\1800\:\text{A}\$$ through itself. You must do something to protect the anode. (This goes double if you have lots of output filter capacitance prior to this crowbar circuit.)

Your circuit doesn't even rate as a "cheap" design. Even a cheap design would include a capacitor across $$\R_2\$$ designed to help manage #2 above. But you really should examine what your SCR can handle and insert a current-limiting resistor in series with the anode. At the very least. Also perhaps include at least one transistor (perhaps three, with two in a long-tailed pair configuration, for better precision and temperature compensation) in your crowbar circuit.

I would like to see a current-limiting circuit beforehand. Or, better still, a current foldback circuit. And a fuse of some reasonable value also wouldn't hurt.

Usually, a crowbar circuit doesn't stand alone. It's usually designed as part of a larger system. For example, it is often preceded by a current-limit or current-foldback circuit. And all of this may be designed as part of a voltage regulation system that has these and other sub-sections in it. But if you really want to make this stand-alone, you will need to deal with the points I made above and especially you will need to do something to protect the SCR's anode. (It won't last long the way you've been treating it in your schematic.)

## Some Topologies

I'm going to just assume for now that you have no current-limiting prior to the circuit and that your crowbar circuit is supposed to be designed "stand-alone" for some reason I don't entirely follow. So what follows will include a current-limiting resistor at the anode of the SCR.

simulate this circuit – Schematic created using CircuitLab

There are two other, better, topologies to consider. One with two BJTs instead of one and providing more precision. And another with three BJTs with a relatively cheap temperature compensated reference (zener + diode to compensate for the BJTs.) I won't bother writing those up, here. I'd probably have to include the design steps for each and that's beyond the scope for which I want to go, right now.

## Basic Crowbar Concept and Justification

Once an SCR fires, it remains on until the anode voltage goes below some figure or the anode current falls below some low holding current. Before the SCR-based crowbar circuit can be extinguished, the power supply it is monitoring must be effectively removed sufficiently to meet these requirements for the SCR to turn off; at least momentarily. There is nothing in your example that shows how this might happen. Instead, it appears that the source power supply rail is simply always on. You should have specified the environment within which the crowbar circuit operates and how it might be reset.

That said, an SCR can be fired with either a current or voltage pulse that can be over a range that may span from $$\10\:\mu\text{s}\$$ to less than $$\500\:\text{ns}\$$. Because SCRs are so easily turned on by narrow impulses, you almost always have to parallel the gate to cathode pins with an RC pair designed to integrate out pulses that are narrower than desired (any unwanted impulse is such a "noise spike" that you want to remove.)

SCRs also have very sloppy specs for triggering them. Because of how imprecise all this usually is, an SCR crowbar circuit usually has to include two important sections. The first section is (preferrably) a temperature-stable voltage comparator that can sense the difference between the output (or some fraction of the output) voltage and some kind of temperature-stable voltage reference. The second section is then followed up with a high-gain amplifier designed to drive the SCR gate. You will need to ensure sufficient gain to make sure that it can fire the worst-case SCR conditions (given the datasheet specs on the SCR.)

Sometimes, both the comparison and gain bits are incorporated into the same circuit element -- for example, in the above right-side schematic the BJT's base-emitter junction is performing a voltage comparison as well as gain (in combination with other elements in the schematic.) But a second BJT can be tasked to provide that gain. Also, the comparison portion can be improved using a long-tailed pair and the temperature compensation can be provided by including a diode with the zener, in combination with one of the BJTs in the long-tailed pair. But that's now three BJTs and more elements.