I skip the 32KHz generation for a while and I'm moving to the main piece : does my 10T flip-flop gate work ?
I looked at one "leg" of the circuit : the complementary FETs that load and enable the pull-down transistor seem to work. With the scope I can observe significant Miller effect / capacitive coupling through the gate but the BS250 and BS170 work well together to hold the charge.
Even a single inverter works well at 2V so a pair of 1.5V alkaline batteries is ideal.
However the full circuit doesn't work. I'm still wondering why. I'll have to rebuild the flip-flop in a different order, that might give a clue... Reversing the CLK and Data pass transistors could be another solution (to be tested).
Well it's a half-success anyway but I still can't sleep.
I can sleep now !
I made another version with the CLK input at the low side, considering the capacitive coupling. I also spotted a bad transistor. And now it works :-)
My system has been validated ! It works at >= 2.5V though.
Now, I have to find why there is this weird noise during one half of the cycle. Why does it pick up parasites ? There is a pair of 100K feedback/latch resistors, are they too weak ? Are the probe leads too long ? Did it pick up EMF from my lamp's transformer ?
It seems to work well at 1KHz down to 10Hz, I try to slow it down more but my scope's remanence is too low. It *should* work down to 0Hz, right ?
For the "fastest" input signal, I see it working up to about 8KHz (120µs cycle time). The below scopeshot is at 20µs/square:
I'll have to add special "high speed" features for the 4 first stages of the 32KHz divider. Some Schottky diodes might help. Oh and a nice, well laid-out PCB would help ;-)