latching circuit - not latching off - Bloody diode!

[Scott D]

I need some help with this latching circuit; it’s really doing my head in. I’m not an electronics engineer, so I would appreciate more conceptual/explanative rather than mathematical answers as the math is often over my head, but any help is welcome and appreciated.
The issue is it will latch ON without issue, but latching OFF fails/is inconsistent between components. This was more prominent after going to surface mount on a PCB.

Here is my circuit:


(Button = P-Switch header)

•   Input voltage is 9V, intended for a 9V battery, but ideally would work between 4-9V.
•   The circuit works on a breadboard.
•   The load on the circuit to the right of the diode consists of 2x Opamp IC’s. 1x single package configured as non-inverting for voltage divider, and a quad package configured with multiple-feedback bandpass filters. 3x 100nf caps to ground on each respective V+ for noise filtering.
•   The circuit works as expected without the load connected to the circuit, in all cases.
•   The circuit overall works as expected when latched on (correct expected voltages/current etc).

My current understanding of the circuit in simple terms is;

•   In order for the P-MOSFET to turn on, the gate is maintained low (within -2 to -4 VGS(th)) by the NPN transistor being saturated. It is pulled low by toggling the switch, resulting in current to flow to ground through the capacitor.
•   When on, in order to latch off, the NPN must be switched off, this is done by again toggling the button which should cause voltage to the gate of the MOSFET and in turn, causing the latch off, resulting in the NPN being turned off simultaneously.
•   The positive voltage needs to be well less than -2V as per the -2 to -4 VGS(th) as per data sheet for the MOSFET, in order to be outside the threshold and resulting in the MOSTFET ramping back up and staying on. In this case ~8.4V - 8.7V = ~-0.3v well clear of -2V threshold.
•   The capacitor needs to (ideally) charge up to maximum voltage, which is governed by the RC time constant determined by the resistor value (1M in the diagram) and the capacitor value. With 1M and 330nF cap that’s 330ms.
•   I think: the 1M resistor, and the 10k resistor on the MOSFET base form a voltage divider when the button is activated whilst latched on, causing the voltage to be expected at the gate to be the product of the divider, in this case ~8.4V. This is likely to be a wrong assumption? If using the https://www.digikey.com/en/resources/conversion-calculators/conversion-calculator-voltage-divider calculator including 10k load, it would be ~4.4V instead, which could be close/within VGS(th).

I have also added two NPN transistors to this circuit to act as touch on/off respectively, these are tied into the gate / base. For these I have added biasing resistors so the bases are not floating. Again this worked fantastically on the breadboard, but the issue exists without this additional inclusion, with/without the biasing resistors.
It would seem that when attempting to turn off, the MOSFET does not fully latch off (linear mode?), or depending on resistor values, will drop between 0.3-4.5v and ramp back up. I expect that either the NPN is not turning off, or the voltage at the gate of the MOSFET is still within VGS(th). OR aliens.

I have tried isolating the circuit to just the original, changing resistor values, trying RC time constant <330ms and up to 3.3 Seconds. All have yielded the same result. The only certainty is that when removing the rest of the circuit, it does work as expected. It has worked before with some transistors, but I think I just got lucky with the specific transistor in question perhaps having a higher VGS(th)?

Point being I am seriously lost and unsure why it doesn’t work in these conditions and what I need to do to fix it and make it work consistently between parts.
I have simulated the circuit on falstad, http://tinyurl.com/qr8dnve with a fake load to the right of the diode. Of course everything works fine in the simulator.

Resistor values, component choice, Incorrect understanding of the circuit? Perhaps it is obvious but I am not seeing it.

Scott

[Scott D]

Through extreme persistence and a touch of lunacy I was able to determine the cause of the issues.
For what its wroth and since this post cant be removed - The cause ended up being very simple.

The circuit and values are sound (but could perhaps be changed). The issue was caused by the BAT60 diode.
This was a poor choice of diode for this application, as its reverse current is actually quite high, around 2.6mA @ 8V according to the datasheet https://docs.rs-online.com/ba5b/0900766b81073ee3.pdf

This was enough for current to flow back to the base, through the 100k resistor and keep it in saturation.

Replacing the diode to a BAT760, or really any other with a much much lower reverse current, resolved the issue and restored my sense of self worth (for the day at least).