Mysterious N-MOSFET behavior

[1337ralfy]

Hi, long story short im trying to build electronic load - pretty simple design Arduino Uno, RC Low-pass filter(440uf and 400ohm res) and n-mosfet(irf3205). Mcu gives out 16bit pwm signal that goes into low-pass filter and transforms it into linear voltage to drive mosfet so it can be used as "variable resistor"(pwm adjusted by pot).

Here is the video of setup without anything connected:
Yellow line: mosfets gate, blue line: output form mcu

https://youtube.com/shorts/ABAT-hBQKj4?feature=share

This is when power supply is attached (power supply directly attached: + to mosfets drain, - to source):
https://youtube.com/shorts/zvgtWQUTWrQ?feature=share

If voltage on power supply is set to about 6volts, it can go up very lineary to 6 amps (power supply limit).
But if voltage on power supply is set at 8 volts or above, it at around 2 amps it starts to act up, isnt liner anymore and short after mosfet just turns fully on, sometimes killing itself.

When this happens it still works ok, but why it start doing that, is mystery for me (starts when gate voltage is about 3v, and i think it starts to conduct there).


From about 3.5v to 3.73, there is no noise

But when gate voltage reaches about above 3.73 than this happens and mosfet just makes a dead short (this happens at around 2-3amps, not at 6 as i would expect) :


What could cause those both noises?

[Andy Watson]

A schematic would be useful! (But, taking a flying guess:  ) Your video seems to indicate that the mosfet is unstable - try adding about 500\$\Omega\$ imediately before teh gate connection. If you want to know why, look-up "grid stopper resistor".

[1337ralfy]

A schematic would be useful! (But, taking a flying guess:  ) Your video seems to indicate that the mosfet is unstable - try adding about 500\$\Omega\$ imediately before teh gate connection. If you want to know why, look-up "grid stopper resistor".

Not much happening hw wise, fore now there is also no shunt for current feedback:


Okay, will try to add that 

[Doctorandus_P]

It is an instability in the control loop. The MOS fet itself is just a part of that. Often the opamp between the filtered DC, the shunt resistor feedback and the mostfet gate is turned into an (limited) integrator to get the loop stable under all conditions. It can be quite difficult to do this properly, so don't be surprised if you spend quite some time on this.

[Doctorandus_P]

Not much happening hw wise, fore now there is also no shunt for current feedback:

Euhm, oops. Without a shunt for current measurement and a loop with feedback it is never going to work properly. It is not even needed to post a schematic. It will just NEVER work reliably this way. Do some research into how to build such a circuit with feedback. There are plenty of examples available.

[1337ralfy]

try adding about 500Ω imediately before teh gate connection. If you want to know why, look-up "grid stopper resistor".

Adding 500ohm after Low pass filter helped, there is no more noise from 0 to about 3.73v.

It seems to me that after that 3.73v its near its saturation range, and that noise seen in second photo can be causes by power supply trying to limit current?

Without a shunt for current measurement and a loop with feedback it is never going to work properly.

And yes now it looks like i need a feedback from shunt and opamp so it can do its control loop.
Btw, ive built this simple circuit just to test the theory about mosfet acting as "adjustable resistor".

[Doctorandus_P]

From what I remember: J-fets work quite reasonably as adjustable resistor, but also only for low currents and voltages. A MOS-fet is even more like a switch then a Jfet. Below it's threshold voltage (usually 3 to 5V) it is "off", then there is a small trajectory with a "significant resistance change" and resistance still about halves when the gate voltage changes from 8V to 16V, but then you are already in the milli ohms with most power MOS fets.

The threshold voltage and resistance characteristic also change with temperature. That both explains why it goes wrong with higher input voltages (The Fet gets hotter) and why you need feedback to stabilize the loop to keep the current stable.

[jbb]

A few years ago someone did a work up about MOSFET current sinks on the forum here. They showed that MOSFETs often experience an increase in small signal gain (ie transconductance) as the drain current goes up. A circuit would be stable at low load and then eventually become unstable as the current went up. This could be tamed by tweaking the control loop.

[PCB.Wiz]

Adding 500ohm after Low pass filter helped, there is no more noise from 0 to about 3.73v.
It seems to me that after that 3.73v its near its saturation range, and that noise seen in second photo can be causes by power supply trying to limit current?

The mosfet is going to change rapidly near the threshold.
See the attached plot - notice the drastic changes from 2.4/2.6/2.8V, and that is before temperature changes kick in.

And yes now it looks like i need a feedback from shunt and opamp so it can do its control loop.
Btw, ive built this simple circuit just to test the theory about mosfet acting as "adjustable resistor".

Feedback on a multi loop system, is very hard to get stable, and you have a very large filter capacitance there.
A MOSFET will work more as a current sink than a resistor, and will need mV steps on the gate.
It will also change quickly as it heats up, so you need very good cooling.

Those are not unsurmountable, and you may be better to work on a design that tracks thermal changes and sets a target load current at a calibrate step.
Such a design does not then care nearly as much about the load behaviour.

You can use NPN transistors or MOSFETs as the load.
A benefit of NPN transistors, is an accurately defined base current, gets you a good starting point.

See the 2SC5200 plots below, that's a 'good' NPN part, with high and flat HFE.
If the MCU measures temperature and sets precise base current, you can knob-set a test current, capture the Base current, and then apply a small temp correction, as it warms up.
No multiloop contentions to worry about.