Hi Bill,

inspired by your project and the answer from Ted Yapo, I did build the following on a breadboard:

http://www.enscope.nl/rrca/ideas/single_npn_latch.png

This is a working design, but it will most likely need further improvement, especially when you want high speed.

How does it work ?

A 5 volt (TTL) clock is needed, that is in common for all latches, and present all the time. I used 3 kHz.

The output has two states:

1) in the OFF state, the 5V clock signal is strongly attenuated by the 6K8 resistor and the 3K3 - C2 combination. The signal on the base of the transistor is below 0.6 Volt, so the transistor stays off (output high, LED off).

2) The latch can be put in the ON state by shortly applying a 5V pulse to C2 via a series resistor. This will change the operating point of the transistor. The DC voltage on C2 will provide a bias voltage to the base, and the transistor will now amplify the clock signal. So, in the ON state, the output has the clock signal, the LED will be ON. The output signal will be rectified by the C1 - D1 combination, and the resulting DC voltage will appear on C2. This closes the positive feedback loop, and the latch will stay in the ON state. However, the DC on C2 could be so high that the transistor is ON all the time, and that would remove the positive feedback signal. To prevent this effect, D2 is added. Each time the output of the transistor is low, it will discharge C2. When the circuit is properly dimensioned, D2 could also act as a Baker clamp to prevent saturation of the transistor.

You can bring the latch back to the OFF state by discharging C2, or by shortly removing the clock signal.

If you want the latch to be fast, you should make C1 and C2 smaller, and use a clock that is a few times faster than the maximum operating speed of the latch.

Yes, I just simulated this in LTspice, and it works!  I used the initial condition at the node between C1 and D2 to set the  state of the latch at the beginning of simulation. If it's on, it stays on, and if it's off, it stays off.  Very nice!

Hi Bill, Ted,

I added my design in the 2018 competition. See https://hackaday.io/project/112126-one-transistor-flipflop

What about using both AC and DC supplies?  This is how #The Diode Clock works, for example - it uses an RF power supply with DC bias controlling PIN diodes (1N4007s).  Maybe you could make a one-transistor latch along the same lines as the old one-transistor reflex radios, where the same transistor amplified both RF and audio?

Maybe the transistor can switch RF that it also rectifies to produce its own bias, and work like the two-diode latch I added this morning? https://hackaday.io/project/11860-diodes-only/log/71315-two-diode-latch

I'd like to only use one silicon component. Once I get an idea that simulates ok, I'll have to see how fast I can make it too....

Hi Bill,

Two transistors, most definitely.

One transistor? Yes but messy!

I came across something a worth a look a few day ago.

It was a neon latch or memory circuit.

This was not it but similar, see ge69.jpg in http://www.tuberadios.com/ge.zip.

Found it: http://hackaday.com/2016/12/12/the-many-uses-of-the-neon-lamp/.

So swap the Neon with a transistor in reverse CE breakdown mode (about 8v) as shown in this relaxation oscillator:

You will find similar circuits using tunnel diodes.

Regards AlanX

Thanks for the info! 

That is kinda strange, I also just recently stumbled across an old article about a neon-based latch/memory design.

I guess that it might have been a link from inside the pages from one of Yann's postings.

A few kilobytes of neon memory would look awesome, I really would like to see one....

hmmm...  Saw the "Found" note one your post. :-)   So not posted by Yann then... But it easily could have been - it's straight up his alley.