Mosfet Gain tester?

[Kipper4]

Anybody got a circuit for testing Mosfet gains please?

[R.G.]

That's a tough one.

MOSFETs are best modeled as transconductance devices: Amperes of drain current per volt of gate-source voltage. The voltage gain is then determined by the resistance in the drain multiplied by the transconductance: Amps per volt times volts per amp (i.e. resistance) is gain, a dimensionless ratio.

Like with bipolars, MOSFET transconductance varies in operation. What you can get is either a curve of transconductance over changing conditions, or a single number you'll use as an indicator.

Then there's the variable threshold voltage to be taken into account.

I have some textbooks that have various test setups in them. I took one of these and simplified it to death to get the leak/gain tester for germaniums. I'll see if there is something that can be done for MOSFETs.

I speculate that you could do something like this:

Set up a single-voltage opamp to drive the gate of a MOSFET and take feedback from a resistor in its source lead, giving a voltage-to-current circuit on the drain of the MOSFET. Feed the opamp + input with a voltage that caused some nominal current to flow in the MOSFET, perhaps 1ma, or 10ma. Read the voltage between the gate and source of the MOSFET to get the gate-source voltage at that current, which will include the threshold voltage. Then change the input voltage to the opamp to cause more current to flow, and again read the gate-source voltage. The difference between the two currents divided by the difference between the two voltages gives you the transconductance at that point. The threshold voltage winds up being subtracted out when you take the voltage difference.

Choice of the currents needed is dependent on the MOSFETs to be measured. Small signal MOSFETs like the BS170/2N7000 need smaller currents than power MOSFETs to return useful information.

This could, of course, all be automated. A microcontroller could set the voltage steps for the opamp and read the gate voltages, producing a finished answer by some arithmetic. There are probably tools that do this. But I suspect they will all start with the basics of setting a gate-source voltage, measuring a current, then doing some computation.

[Kipper4]

Thanks RG.
a tester might prove usefull.

[PRR]

The transconductance versus current curve on commercial MOSFETs at amplifier currents is VERY bent.

You need to define THE current you will be working at.

Anyway, all sane transistor design starts with an uncertain amount of excess gain, then sets final gain with resistors. (One problem is that at 1mA and 9V-bias you may not have a lot of gain to work with.)

[Kipper4]

The only reason i asked is because when trialing differant mosfets in a new circuit and i'm talking about the common or garden BS170,2N700 varietes.
The gains to my ears was differant as was the compression too. If thats not the right teminology then excuse me but i know what i mean.
So i wondered about it and then through google discovered that sometimes mosfets need to be matched.
So i got curious.
Thanks guys
Rich

[R.G.]

There's an issue with all matching, and that's - matched to what?

In general simply being matched counts most when two or more of anything are working in tandem, and the differences between them need to be minimized. The prototypical use in electronics is for differential amplifiers, but matched pairs goes all the way back to chariots, when one wanted matched left and right horses to avoid veering left or right.

Matching two of anything is far, far easier than matching more than two. This is why matching JFETs for phasers needs some attention.  To get two "matched" things out of a batch of random values, you generally have to pick one as the standard and try all the others against it, then pick the best one. Even then, you often don't get two identical ones. If you know ahead of time how close you need to be for "matched", it may cut the number of trials a lot. If you don't, you're up for N-1 trials to find a match in N devices.

Getting more than two matched generally means trying every one of N and writing it all down, then picking based on the record keeping. This is how binning and most JFET matching works.

You start to get into trouble rapidly when you want more than one property matched. Remember those chariot horses? What if you not only want them both to run at the same speeds easily, but want them both to be the same *color*? Uh-oh. We're now talking order of N-squared tests. Matching three parameters makes it even tougher.

The tube-matching business suffers from some abuses. Very often you only get tubes "matched" by looking at their emission current on a random tube tester. Matching emission, gain, and distortion characteristics (which one highly-touted company does or did) starts getting very tenuous, and I suspect that "matching" was really binning.

With all that as a preface, what other than diffamps and phasers do you need matched MOSFETs for?

[Kipper4]

I dont need matched Mosfets at all yet.
Its just something i came across when i did a google search for "mosfet gain testing"
btw i wouldnt be bothered if my horses colours didnt match but i'm sure my wife would like them colour coordinated.
I'm colour blind to some colours so it makes no odds to me. A horse is a horse.
I appreciate the input and effort you guys have gone to to explain it to my newbie brain.