2

12 VDC automotive environment, protecting the logic-level silicon.

Instead of, say, the crowbar's thyristor, one could actuate a mechanical relay and cut no more than 10 A.

Will that be enough to protect the circuitry downstream or is the bouncing / slowness to react going to fry it anyway? How about adding a TVS across?

kellogs
  • 1,044
  • 9
  • 18

1 Answers1

1

Nope. The fastest relays we got - reed relays - switch in about 100us. That's eternity as far as overvoltage on digital logic goes.

Now, you can coerce them to go a few times faster, perhaps as fast as 5-10us, but there's no guarantee they'll survive too many cycles of it, as what you essentially do is discharge 100-200V from a capacitor via a diode into a coil of a 12V reed relay...* In some types, the coil just rips itself apart from electromagnetic forces, without even having a chance to melt down the insulation.

Don't get hung up on a thyristor - you can use it, but you can use any other equivalent latching circuit made from silicon devices.

The circuit below uses an SCR formed by Q1-Q2, followed by a switch Q3 that turns on when the SCR has latched, causing the crowbar M1 to conduct.

Once the SCR is latched, VCC can fall back to zero and the crowbar will stay active - until 12V is removed.

schematic

simulate this circuit – Schematic created using CircuitLab

As shown, the circuit is configured to trigger at 3.6V - as if it was protecting a 3.3V logic supply. The triggering action is clean and decisive. Assuming 1 ohm power supply source impedance, the output voltage is low enough to keep most of the 3.3V ICs off and safe:

The VCC voltage waveform, showing triggering at 3.6V

The VCC current wave form, showing the action of the clamping MOSFET

*I've been playing with reed relays to see ultimately how fast they could switch in a digital computer application. Turns out they can switch much faster than anyone would think when the word "relay" is heard.

Kuba hasn't forgotten Monica
  • 32,734
  • 1
  • 38
  • 103
  • Well, Q1 and Q2 are still a latching SCR, as you say. So why not use a thyristor instead? Also I don't quite understand how will these transistors survive the usual automotive spikes. – kellogs Aug 31 '22 at 04:14
  • @kellogs Sure you can use an SCR, but it may be easier or cheaper to get transistors instead. And at the power levels you got, the crowbar needs to dissipate much less than a power SCR would. As for automotive spikes: the low voltage transistors can be high voltage types, and you can add base protection diodes. But I thought that you wanted a circuit that protects from a low voltage regulator failure, so it would not be exposed directly to the 12V rail. The 12V it would get would have already gone through a layer of protection in front of the LDO- unless the LDO protects itself. – Kuba hasn't forgotten Monica Sep 02 '22 at 12:51
  • If the LDO protects itself, then just use another LDO set for higher voltage to protect the crowbar control circuitry. It doesn’t really need 12V, just something stable and present even when the 3.3V rail is crowbarred. – Kuba hasn't forgotten Monica Sep 02 '22 at 12:53
  • I am actually looking to protect everything; both logic-level silicon and some 4 x 2.5A @ 12V P-MOSFETs + their BJT drivers (forgot to mention these). And I would highly prefer if it did not latch. I would not go beyond a 5kW TVS diode but that load dump pulse from ISO 16750-2:2010(E) surely looks nasty - rise / fall time = 10/400 ms, U_max = 101V, R_i_min = 0.5 ohm. – kellogs Sep 02 '22 at 14:43
  • Those reed switches look just fine at their ~ 1ms switching speed, they could work in tandem with a TVS diode, but they are quite limited in their current ratings. I would need 4 of them to cover the 4 P-MOSFETs + lots of PCB space and $ spent on them :-/ – kellogs Sep 02 '22 at 14:45