Crowbar Circuit Acceptable Duration of Overvoltage

Other Hardware General Electronics
1

I am building an extensive, tightly spaced circuit board with both a 5V and 12V rail. I want to avoid frying all the semiconductors on the 5V rail in case of a short between the rails. Therefore I have included a crowbar circuit with a TL431 adjustable voltage reference and an SCR.
crowbar2

When I deliberately apply 12V to the 5V rail the crowbar circuit works, but for about 20 microseconds the voltage goes up to a maximum 9V. Would this short duration be harmful to an Arduino or other general purpose semiconductor chips (MAX7219 LED driver in particular)?

NewFile0 copy

2

Very likely to destroy chips, as the damage can be practically instantaneous.

3

Thanks for the quick reply. Should I use another type of circuit or would adding a TVS in parallel solve the issue?

4

Without detailed information about your circuit and all the components, it is impossible to make predictions.

A better design for the PCB might be more appropriate.

5


Is it not blindingly obvious that C1 is the culprit?

:roll_eyes:

6

Since a crowbar circuit "shorts out" the power supply, there is usually a fuse in series so the fuse blows before anything "important". But the fuse won't blow in microseconds.

Most voltage regulators won't allow the voltage spike through so unless you're expecting voltages high-enough to kill your voltage regulator, a voltage regulator is probably a better solution. (But, you do need some "extra" voltage for a voltage regulator to work.)

7

Thanks for your input. I've tried leaving out C1 or reducing its value but that leads to a lot of false positives. Even the slight voltage spike from plugging the circuit into the power supply will activate it.

8

I would use 5V5 "shunt regulator" instead of a crowbar.

9

Thanks, not very familiar with shunt regulators but I will do some more research. It looks promising.`

10

A small ceramic cap from supply to ground should catch those short spikes.
Leo..

11

Why slow down the response with a capacitor? Why not use a comparator in fact, many of them switch in a fraction of a microsecond.

Crowbarring the 5V to ground might interact with capacitances in the circuit and push some nodes below 0V when the crowbar happens - why not use an amplified zener to clamp the 5V rail to 6V or so and add polyfuse on the 12V rail to limit current on a short?

1 Like
12

Yes, I am beginning to understand that this circuit is too primitive and one with a comparator would be preferable.

What do you mean by amplified zener?

13

To be clear, the cap from VCC to base slows down response.
A (decoupling) cap from VCC to ground slows down the spike.

There should already be decoupling to ground. Maybe it's not enough.
Leo..

15

As I said, that was the mistake. :roll_eyes:

Imagining it was needed or in any way appropriate is simply a misunderstanding in design.

16

Hi, I've tested some more circuits wanted to post my findings to help others who might face the same issues with crowbar circuits. When applying 12V to the 5V rail I've gotten the response down from 20 microseconds to 2 and the peak voltage from 9V to less than 8V. Ignore the circuit at the beginning of this post!

Crowbar circuit with triac: (BT136)
Triac Crowbar
This circuit automatically resets as soon as the voltage drops below the threshold voltage.
BT136Triac

With 100 microFarad cap across power rail I get a lower peak voltage but a longer overvoltage duration. (Cap value not critical, more is better. Mind the voltage rating):
BT136TriacWith100uF

Proper crowbar circuit with SCR (SJ6010DS2RP): (ZD1 = Zener not 9.1 but 5.6V. SD1 not necessary.
This circuit latches until power is removed.

crowbar-protectionSCR
crowbar-protectionSCR2000×1109 57.5 KB

SCR Crowbar according to circuit

With C2 removed I get slightly faster response but the circuit could possibly activate due to noise although I have not experienced that.

17

The one with the capacitor where it should not be?

Sure! :grin:

As you would want.

Nor should you expect to. :roll_eyes:

The only "noise" to which it should trigger, is a transient greater than the threshold you set. if it receives such a transient, you absolutely want it to trip - that was the whole purpose of the exercise. :thinking:

Putting a capacitor across the crowbar input is perfectly correct - insofar as there are minor transients or "noise" on the supply lines, the capacitor will damp them which will benefit the equipment being supplied as well as the crowbar.

18

This topic was automatically closed 180 days after the last reply. New replies are no longer allowed.