sub mV level shunt

How about actively *driving* it to 0.000000 volts and using the FET to isolate *that* from the circuit when its at high potential?

Don't use a MOSFET, use a mercury-wetted reed relay. You'd have to mount it with 15 degrees of being vertical, and it will only last about 10^7 operations.

You can get thermocouple voltages between the (largely iron) reeds and the copper on the board, but at uV levels.

In the past I've used low thermal EMF latching relays for applications like this. Art

Mosfets are ohmic at low Vds.

Mess with bipolar transistors, not FETs; "inverted" configuration (ie: use collector as emitter and vice-versa).

due to the environment and switching speed requirements a solid state device is required. Switching speed in the 10s of uS range from digital I/O signal to shunt on or off.

Do you think getting extreme low RDS fets (like 15mohm) would result in ohmic values less than 1 ohm near 0VDS?

I was wondering of there was a biasing trick I could use to make the mosfets see voltage across them so they look like a low impedance shunt at AC frequencies.

I'm looking at muting frequencies on the 100K - 500KHz range.

Would IGBT or BJT be better. I wouldn't think so.

Are you sure that your measurements are not being corrupted by thermal emfs or some other source of error?

Surely the transconductance doesn't matter very much. You are turning the FET on as hard as you can so you don't want the conductance to vary with small changes in gate drive.

John

Rds(on) is flat and continuous from some positive voltage (usually on the order of Vgs; depends on transistor construction) until a negative voltage where the body diode starts to kick in (say, -0.5V at room temperature).

If you're measuring differently... try looking for errors in your measurement? Use Kelvin style connections, null out bias and offsets, use an accurate current source (perhaps calibrated against another known pure-resistor shunt?), etc. Make sure it's not silly things like meter auto-ranging or loose connections?

Mind that Rds(on) has a horrific tempco, so it makes a pretty shitty shunt... I hope you have a way to compensate that out! (It's more or less a linear PTC, so an RTD part should at least be helpful. You can actually get Si RTDs, which should be pretty damn close to the same response.)

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website:

I would look at a servo of some sort. Make a few volts -ve for the MOSFET source and make an op-amp driver with the inverting input at 0V and the non-inverting connected to the drain through a resistor and with a zener to ground for protection when the drain is at high voltage.

That's over-simple, of course, but might be a spicey start.

Cheers

A 15 mohm fet should be about 15 mohms near 0 volts. It's ohmic.

I think he's trying to kill an audio signal by shorting it to ground. The Rds-on tempco shouldn't matter as long as Rds-on is low ehough.

I hope the signal is small enough that the body diode never conducts.

Personally, I'd use an analog switch chip, series-shunt or a tee switch. Or double series-shunt for extreme attenuation.

That'll move the problem to the offset of the OpAmp.

Wonder how a depletion-mode FET would act at low VDS?? ...Jim Thompson

Does it work better if you switch it for long times? Maybe you are seeing some other effect... charge injection? Capacitance or inductance somewhere else in the path? How big are the signals you are shunting?

George H.

Correct.

Your measurements are in error. As John said, Rds(on) is basically ohmic. I just measured a FQP33N10, a 100V TO-220 33 amp MOSFET. It measured 45 milliohms at 1mA, 10mA, 100mA and 1.0 amp. Moreover, as expected, I got the exact same value for the reversed current direction.

If I had extended my measurements to 60 to 100A, I would expect to see a small increase in Rds, see Fig 3 of the datasheet. First this comes from higher current density, and then from junction heating, see Fig 8.

I'd get similar measurements with a 400V or 600V part.

Irrelevant, even if true.

George, I think you spotted the issue.

A quick curve trace of a MOSFET (high quality LEVEL=7 model), shows Ron continuous thru the origin. ...Jim Thompson

Den onsdag den 2. november 2016 kl. 16.14.26 UTC+1 skrev John Larkin:

it is shorted by the milliOhm FET so until the current is high enough to generate diode voltage over that that shouldn't happen

-Lasse

Low-value resistance measurements should be made by the four-terminal (Kelvin) method.

See adjustable Ron plots in AoE III, Figure 3.47.

As I mentioned, a power MOSFET, when turned on, is bi-directional for current, so it has the same low resistance for current flowing in either direction. However, when turned off, it's an open circuit in one direction, but a body diode in the other. So if you want to make an AC switch, you need to use two MOSFETs back-to-back (sources and gates tied together) and drive Vgs with a floating voltage. Figure 3.107, page 205, has the complete circuit. See Table 3.5 for some suggested MOSFET choices, with switch resistances from 0.003 to 16 ohms. The latter has low open capacitance, only 52 pF.