MOSFET output not stable

"Ben Vaughan"

But how do you know it has NO variation or noise ?

By 10s of mVs only !!

...... Phil

What is the "power supply" you're using to drive the gate? You're building an amplifier, you should expect to be amplifying small amounts of noise at the gate into large amounts of noise at the drain.

Could you have a 60Hz hum on the gate? This would cause a big 60Hz hum on the drain; if you're measuring the thing with a digital voltmeter you'll see that as the numbers jumping around.

Well, that power supply driving the gate is not all that stable or quiet, and the power that the FET dissipates does warm it up - which you see as a dropping drain voltage. So....trash those incorrect assumptions and re-start.....

Hi guys,

Thanks for your responses. I agree that there is noise in my supply and that is part of what I am seeing. However at the beginning when I switch the circuit on there is a significant time (30 seconds +) before the output voltage of the FET has settled. It seems to drop around 2 volts in this time. Once it has settled then yes I am just looking at noise in the supply, which I suppose I should be using some bypass caps to alleviate.

Is this settling time normal?

Cheers, Ben.

Adding a small resistor from source to ground will help a lot. Something like 75 ohms will essentially make this an inverting amplifier with a gain of 10 and an offset equal to the gate threshold voltage. There will still be some drift due to temperature, but it will be more stable and less noisy. You can add a capacitor across the resistor to increase the AC gain at higher frequencies.

Paul

"Ben Vaughan" "Phil Allison"

....... Phil

Hi Paul,

I added the 75ohm resistor and the settling disappeared. Now apart from noise it seems stable. I understand that adding the Rs is a form of source degeneration and it helps reduce the effect that changes in the transconductance have but I'm not sure how this has eliminated the settling? Is it because the settling I was seeing was in fact an overshoot correction? ie. I switch on the gate voltage, it shoots up to it's required value but in this short time period the transconductance of the FET changes due to the changing gate voltage, causing an overshoot? The Rs value stops the transconductance having such a large effect, therefore no overshoot? Does this sound right? Also if this is the case what sets the settling time when I have no Rs resistor?

Thanks everyone for your help so far!

Cheers, Ben.

The source resistor essentially provides negative feedback of a fairly exact amount. The transconductance is determined by a number of factors that may vary with time, temperature and noise. A very small change in Vgd causes a rather large change in Id, and the larger the drain resistor, the more this results in a fluctuating voltage. With the source resistor, as the current through the device increases, the voltage on the source resistor acts to decrease Vgd, which stabilizes the voltage. The several millivolts change in the gate threshold voltage during warmup are swamped out by the much larger voltage on the source resistor, so it stabilizes almost instantly.

If you look at a SPICE model for a MOSFET, you can see some of the internal factors that affect its operation. A Fairchild HUF75321 has an internal source resistance (apparently) of 0.016 ohms, about half of its full ON resistance of 0.028 to 0.034. Thus a change of 10 mV may cause an output current change of 10/16 or 625 mA. The transfer characteristic shows 45 amps change for 1.5 volts gate change. 16 milliohms in the first case (at very low current), 33 milliohms in the second (at much higher current). I don't use transconductance in my designs, but the similarity between these two shows some validity. But as you can see, just 10 mV causes a large change in conduction, and the larger the drain resistor, the more voltage change will occur. Similarly, the internal drain resistance need not change much to make a large change in output at a fixed gate voltage.

The data sheet for a similar MOSFET, STP20NF06, shows a transconductance of

Linear circuits almost always need some sort of negative feedback for stability. Switching circuits are so much easier, like TTL or CMOS logic.

Paul

One problem we occasionally had with MOSFETs at Cambridge Instrument was their tendency to oscillate at a few hundred MHz. This wasn't always visible on regular oscilliscopes, and could lead to very odd DC measurements.

We ended up putting 10R resistors close to and in series with the gate of pretty much every MOSFET we used. The seemed to be enough to kill the tendency to high frequency oscillations in most cases, and you could always change it to 100R or something larger if it didn't provide enough damping in a particular situation.

-- Bill Sloman, Nijmegen

Thanks guys for the responses.

Paul your certainly right about switching circuits being easier, this is really my first time dealing with FET's in a linear sense and there's a lot more involved than I had first imagined! Thanks for your explanation of why the feedback kills the overshoot, I understand a little more now :-)

Bill I'll try adding a low series resistor to the gate tomorrow and see if it further improves things.

Cheers, Ben.

It is not "settling"; it is the FET warming up as i mentioned.