Touch sensitive transistor latch; erratic behaviour and other problems

[QwertyTSecond]

I've currently got a small project underway to latch an LED on with no moving parts. So far, I've run a 2n2222 darlington pair into a pnp-npn pairing. The pnp-npn setup is based off an approximation for a thyristor.

Using a 2n5401 and a 2n5551 it works, although it's got a bit of a personality, ie when connecting the 9V battery it sometimes just turns on anyway, sometimes it won't latch, won't turn on at all etc. although disconnecting and reconnecting the power a few times usually fixes this.

My first question is: what causes this behaviour, and how can I fix it?

My second question: I've tried this using a 2n3904 and 2n3906, but upon connecting the battery the LED simply turns on. I'm thinking this is just because the voltage bull dozes its way through the transistors, but if anyone knows why I'd love to know.

 

[duke37]

The 'thyristor' has a very high gain and once it starts to turn on, it switches fully on rapidly. A rapid rise of input voltage will set it off due to capacitive effects. Thyristors are specified for maximum speed of voltage rise.
To make the pair less sensitive, place a resistor across the base/emitter of one transistor, this holds it off until the trigger voltage rises above about 0.6V. Try 1k.

 

[QwertyTSecond]

Whay hey, it worked! You're a genius!

I've found, with further experimentation, that adding a darlington pair causes it to remember its previous state. It seems I've accidentally invented non-voltatile solid-state memory.