SCR crowbar circuit design

[tom66]

I'm powering a long string of delicate WS2812B LEDs, and just in case my buck converter decides to go bad, I'd like to have some protection for £100 worth of RGB LEDs... 

I think the best solution will be to stick a crowbar circuit on board using an SCR and zener. For this to work I need to have sufficient trigger current flow through the SCR to latch it on at the overvoltage threshold, which should  not exceed 6V. The maximum operational supply voltage will be 4.8V and this may vary under load.

The SCR I am considering (TS820-600B-TR) has a threshold voltage of 0.8V and threshold current of 200uA. So the zener diode needs to drop 6-0.8 = 5.2V at 200uA.  So, I should pick a 5.1V zener diode which would give ~5.9V threshold for OVP, give or take about 5%.

The most common failure I anticipate is shorted FET leading to output voltage = 12V which will trigger the circuit rapidly, no problem.

The shunt current should be sufficient to either blow the input fuse (3A rated time delay) or cycle the power adapter (12V, 5A)

I do worry about the possibility of marginal trigger currents causing excessive power dissipation in the SCR. To overt this I need to select the correct pull-down resistor.  The pull-down resistor can be small, low enough to keep the SCR off and high enough to increase the threshold voltage too much. A quick guess suggests 47k, because at 0.8V this gives 20uA, 1/10th turn on current.

In my case two separate strips share the same OVP circuit, any one goes it will take out the other supply too. So I need to adjust the zener voltage to 4.7V to account for about 0.5V drop of the diodes.

Is my reasoning correct?

[David Hess]

Is my reasoning correct?

It is but I would use a much lower value of shunt resistor so the zener diode operates closer to its test current.  I would also add a shunt capacitor to prevent false triggering.

[T3sl4co1l]

Ugghhh, shitty datasheet doesn't even know what it's talking about.

Absolute Maximum Ratings
VDD   3.5-5.3V
...
Electrical Characteristics (VDD = 4.5-5.5V)

If you didn't already notice their characteristic measurement spec is performed under conditions exceeding the absolute, you might've noticed they're unsafe when powered down.  Yep, can't ever have VDD less than 3.5V.  It'll explode if you drop below 55... or something.  (Bad movie reference.)

Also like how "typical" is abbreviated "tpy".

I would wager a guess that these are HCMOS compatible, and as such, fine to 7V.  Beyond that, they may or may not fail gracefully.  If not, well, if I were you, I'd feel rather ashamed busting 100 pounds on crap like this...

Anyway,

That's a good way to do it, though I've been warned about slowly rising conditions (not just marginal voltage, but slow dV/dt as well).  Zeners aren't terrifically accurate, and lower voltage devices are leaky.  You can compensate that some with the G-K resistor, but leakage remains temp dependent.

A more accurate way is to use a TL431 as a ref + comparator (or a TLV431 would work at this voltage -- save about a mA), and add positive feedback (using a resistor divider into the "anode" pin to fake the + input, or an outboard transistor to invert the signal and feed back a little into REF -- which is also better for finally driving the SCR) so it snaps on at a fixed level, regardless of dV/dt.

Tim

[T3sl4co1l]

Ah, I do have an example laying around:



Size R1-R2 and the 100 ohm resistors to get the desired threshold, hysteresis, quiescent current and SCR drive.  The rails are shown separately, but they can be common (one example might be, if you want to crowbar the primary supply in case the secondary supply's buck lets loose).

Tim