Single pulse generator using discrete components

[Yaroooo]

Hi all!

I've a problem with a schematic and I'd like to have some support. I've a commercial board like this:



What I want to do is to drive LEDs with OUTPUT pin. This LEDs are used like flash source, so I turn them ON for a very small amount and then off for a relative long time.

I have a button to manually drive the flash, so I must limit ON time by external hardware to avoid LED damage.

What I've done is this schematic:



What this circuit should do is:
- Circuit is 24V powered -> Button is randomly pressed -> Flash ON for RC time (about 2ms) -> Time finished Flash goes OFF -> Button is released -> C de-charge -> Button is randomly pressed..etc..

I've tested on breadboard and circuit actually works. But I've found some problems:

1) When button is released, OUTPUT take some time to go HIGH again. Actually not a real problem, but if there's a way to solve I'll able to control better ON time length;
2) Very rarely circuit freezes with OUTPUT to LOW, happened 1 or 2 times after pressing Button and some more times when circuit is connected to power for the first time;
3) ON time sometimes doubles or triples, sometimes OUTPUT stay LOW for 4ms or 6ms instead of 2ms;

I attach an oscilloscope image about problem 1) .



Any help is appreciated!

[Kleinstein]

The part with the slow ramp looks a little like there is some oscillation. It looks like the circuit does no go cleanly from on to off, but kind of oscallates for some time.
The case with stuck on could be a kind if sustained oscillation.

One point that could cause troulbe is that not many MOSFETs like some 24 V at the gate. Some added resistance at the drain sides could reduce the voltage and als redue the tendency for osciallation by reducing gain.

[T3sl4co1l]

If you insist on discrete circuits, you might do something like the high side here.



The gate has a ton of capacitance, so you need to pull it up with something stronger than a 10k resistor.

This uses a 680 ohm into an emitter follower, which effectively reduces its value by hFE.  So, the pull-up resistance is close to 5 ohms at the gate, for typical transistor choices.  This will swing the gate in the ballpark of 100ns.  (Instead of 2N4401/3, also good are higher current duals like ZXTC2045E6TA or ZXGD3003E6TA, note that the latter is crudely spec'd as a "gate driver" but yeah it's just two damn transistors in a pack, treat it as such.)

Tim