latching push-button controlled MOSFET switch with IRF7319

[luca0074]

but i still can't turn on the device below VCC of 4.4V

If that is the case, there must be some loading pulling down on the "3.3V_EN" and/or "POWERSENSE" connections. This is possible as the MCU had no supply voltage yet and the above two signal pins could be powering up the MCU via port pin internal diodes. This affect the operation of power On/Off circuit badly.

A good circuit designer should make sure this situation could not happen, i.e. there must be no loading on these two signal pins until the MCU has properly powered up, reset and running.

I tested the circuit without a mcu. i have a micromod connector on my board. i will use the SAMD51 Processor board https://www.sparkfun.com/products/16791

If VBUS is, say, 4V, and you press the button, the first thing that has to happen is the P-fet turns on.  Is that happening?  If you put your meter on Vcc, do you get any kind of blip when you push the button?

"A scope!  A scope!  My kingdom for a scope!"  - Richard III

I will have access to a scope tomorrow. then i can check.

[luca0074]

i made some measurements with a scope.
These were my measuring points:

With the scope the circuit worked down to 3.75V. Without the scope as before to 4.4V

[luca0074]

more screenshots:

[luca0074]

in the first 5us there is some contact bouncing (or another phenomena)
but i dont think this should affect my circuit.

[Peabody]

No, it's oscillating.  I don't know if that matters in this case.

Looking at the scope displays, the P-fet gate (the blue line) just drops part of the way to ground when the switch is closed.  Then it drops all the way when the N-fet turns on.  But as the input voltage is reduced, the GS voltage during the first knee gets closer to the threshold voltage, and the response is delayed because the P-fet just isn't turning on as fully.  Then finally, when the input voltage is low enough, the threshold voltage isn't met at all, and the P-fet never turns on.  Which means the N-fet never turns on.

Also, it seems odd that the level of the first knee on that blue line doesn't vary more with the input voltage.  In any case, increasing the capacitor value should lower the knee level.  But I don't know why the circuit doesn't behave more like the simulation, or only does so when the scope probes are attached.

[Peabody]

Ok, I think the knee on the blue line occurs because the P-fet gate is in the middle of a capacitive voltage divider made up of  C3 and C4.  So when you close the On switch, the voltage only goes half way to ground.  In theory, this could be improved by increasing the value of C3, or decreasing the value of C4, or both.  But I would first try reducing C4 to a very small value.  I think it might work for its intended purpose on as little as 20pF.  You could test what effect this would have by temporarily removing C4 altogether, and see what your scope displays look like then.

[eblc1388]

i made some measurements with a scope.

There is something not quite right about the traces. With a normal circuit, when the whole circuit is in the OFF state, the red and blue trace should have the same potential(voltage). I even perform a simulation and indeed the two voltages are found to be the same.

You would need to resolve this situation (or find out why) before your circuit will operate correctly. It is possible that the scope probes has too low an impedance but this type of measurement should only be carried out with a normal X10 probe to avoid the probe's impedance affecting the circuit performance and measurement results. What kinds of scope probes are you using for the measurement?
 

[Peabody]

Assuming the actual circuit is the same as the schematic, then the difference in the red and blue voltages when the power is off could also be caused by a leak of a few uAs through the N-fet to ground.  In that case, he would get the same difference when measuring the voltage with a multimeter, and there would be a voltage differential across the 100K resistor, indicating that current is flowing through it.

[Peabody]

Looks like the OP has lost interest.  I was hoping we would get to a solution on this circuit, but I guess not.

[luca0074]

Looks like the OP has lost interest.  I was hoping we would get to a solution on this circuit, but I guess not.

I have allowed myself a few days holiday
of course i want to find out what is wrong with my circuit.
Also I am incredibly grateful to you and also to eblc1388 for the thought, time and effort you put into troubleshooting!
so... Thank you!

What kinds of scope probes are you using for the measurement?

I use a PicoScope 4824A. https://www.picotech.com/download/datasheets/picoscope-4824-data-sheet-de.pdf

It is possible that the scope probes has too low an impedance but this type of measurement should only be carried out with a normal X10 probe to avoid the probe's impedance affecting the circuit performance and measurement results.

Maybe thats also a problem. I just have very long long BNC to banana plug cables.

You could test what effect this would have by temporarily removing C4 altogether, and see what your scope displays look like then.

I cut open AUTOON, so C4 isn't connected to anything.
The knee is gone, and the circuit works as intended!  (even without the scope attached)

But I would first try reducing C4 to a very small value.  I think it might work for its intended purpose on as little as 20pF.

I don't have any pf caps lying around to test with such a small capacity. Is there a need for a capacitor at C4?

[eblc1388]

What kinds of scope probes are you using for the measurement?

I use a PicoScope 4824A. https://www.picotech.com/download/datasheets/picoscope-4824-data-sheet-de.pdf

It is possible that the scope probes has too low an impedance but this type of measurement should only be carried out with a normal X10 probe to avoid the probe's impedance affecting the circuit performance and measurement results.

Maybe thats also a problem. I just have very long long BNC to banana plug cables.

I'm glad you've finally get the circuit running OK. If you remove C4 (AutoON), then the operation of your circuit is unpredictable. It might turns ON upon power Up without user pressing any button.

The Picoscope has an input impedance of 1Mohm, when used directly as an input. This is way too low for measuring this circuit. You need to use a proper probe which allows x1 and x10 selection. When the X10 range is selected, your probe will have an impedance of 10Mohm, which only slightly loaded the signal your are measuring. You'll need some probes like the following:

[Peabody]

C4 only matters when you first connect power to the circuit.  It's there to make sure the P-fet doesn't turn on automatically when power is applied.  But it may not do that even without C4.  You would have to test it a few times after leaving power disconnected for several minutes to make sure all the caps have discharged.  And in the worst case, if it turns on when power is applied, you can always just turn it off by holding down the button long enough.

But it doesn't have to be 20pF.  If it's 0.1uF now, you might try 0.01uF if you have one of those. That would be cutting the capacitance by 90%, which may give you more than enough low-voltage performance.