Building a flip flop with MOSFETs

[marcelomd]

Soooo... yeah...

I tried both the MOSFET and the CD4049 version.

CD4049 wins because:
- I tried it before and know it works, but, mostly, because
- It's half the PCB space of the discrete version. One third if using MOSFETs to drive the optos.



Thanks a lot for solving this puzzle, even if I didn't use the solution!

[Rob Strand]

CD4049 wins because:
- I tried it before and know it works, but, mostly, because
- It's half the PCB space of the discrete version. One third if using MOSFETs to drive the optos.

The CMOS version has a lot of miles clocked up in the DOD pedals so you know it will work.

You should ground the input on the unused CMOS gate.

[R.G.]

I did get the MOSFET version to work in a sim finally. I was lazy and didn't write up the math, just subbed parts. It turns out that the MOSFET version is sensitive to the time constants of the input switch cap, the pulse injection caps and the feedback caps. Once I tinkered them, watching the drain and gate waveforms until I got triggering. 
It was weird. The trigger switch capacitor was a surprise. This one had to be small, under 100pF to get regular triggering. I suspect that this would have been a problem if I had modeled the effect of switch bounce, but I didn't go that far. The trigger caps and feedback caps had to be within 2:1 of each other to get flipping.
I think that the internal capacitance of the MOSFET and the plateau-ing that MOSFETs do when switching is mucking with the transients of flipping.

[Rob Strand]

I did get the MOSFET version to work in a sim finally. I was lazy and didn't write up the math, just subbed parts. It turns out that the MOSFET version is sensitive to the time constants of the input switch cap, the pulse injection caps and the feedback caps. Once I tinkered them, watching the drain and gate waveforms until I got triggering. 
It was weird. The trigger switch capacitor was a surprise. This one had to be small, under 100pF to get regular triggering. I suspect that this would have been a problem if I had modeled the effect of switch bounce, but I didn't go that far. The trigger caps and feedback caps had to be within 2:1 of each other to get flipping.
I think that the internal capacitance of the MOSFET and the plateau-ing that MOSFETs do when switching is mucking with the transients of flipping.

Was that the circuit you gave earlier with the positive pulse?

For the MOSFET Boss circuit I got working had the added gate-drain diodes and low MOSFET threshold: It used negative trigger pulse the same as the Boss circuit (1M, 100R, 10n).   For that circuit the only limitation was the feedback caps had to be at least 120pF.  For 4.5V operation they needed to be at least 220pF.   Obviously something to do with the MOSFET capacitances.  Feedback caps in the order of 470pF to 1nF looked good.   Once the feedback cap was chosen the trigger caps needed to be small to stop glitches on the falling edge, say less than 220pF.   To my surprise I could drop the trigger caps down as low as 22pF.  So the circuit worked with fairly large and small feedback to trigger cap ratios.  In the end I decided the feedback caps 470pF to 1nF and trigger caps 100pF.

With those circuit values the circuit also worked with the higher threshold MOSFETs.  However, the higher threshold MOSFETs didn't work at 4.5V supply.  The most likely cause is the drain-gate + gate-ground resistor is dividing the gate voltage too much.  I stopped there.

I suppose the key mod for the circuit I played with was the addition of the gate-drain diodes.

Once you pull the diodes some delicate balancing is required.
FWIW,  it's possible to get the Boss type circuit to work with higher threshold MOSFETS (ie. 2N700x's).   However, what we then see is the drain-gate diodes have an undesirable effect: namely  the drains cannot swing less than one diode drop below the MOSFET threshold, which would be about 2.0V-0.6 = 1.4V.   The diodes were the key to making things work nicely but they clearly have a very undesirable side effect!!