FET as variable resistor

[fryingpan]

Hi, everyone,

I'm looking for a FET that is able to be used as a VCR in the 500-2kohm range. I don't need it to be terribly linear across all the range (I only need the resistance to be locally linear for each Vgs), anyway the Vds range would be 0-0.1V (with a 0.3V offset). An N-JFET or a P-JFET are equally OK (an N-JFET would be more convenient, I suppose, but it's no big deal). I suppose also a MOSFET could do (the biasing is a bit different but no biggie, again).

Thanks!

[PRR]

That's not a tough spec. Have you tried it?

[fryingpan]

At the moment I'm only simulating it in LTSpice, of course I'll have to measure stuff but I'd like to start somewhere. Ideally the 500-2000 (or 2500 too, or higher still - well, that's not too hard to do) range should be its outer bounds, as I'd like to exploit the fact that the resistance does not vary linearly with Vgs but is actually somewhat exponential.

Let's say that it should ideally follow the black curve within 0.1 or 0.2 and 1 (x axis: 1 - normalised Vgs), the red curve is its ideal behaviour in parallel with a 50k resistor, I think.

[PRR]

The "outer bounds" are unavoidably 1/Gm @ Idss, and Infinity (for all practical purpose).

Gm is very often in the area of 2,000uMho which is 500 Ohms.

You know how to shim-down "infinity".

Adding a series resistor sets a bound but screws-up the shape of the curve. Now you would hope to select JFETs for 1/Gm in the desired range. But in production Gm varies about 10:1, with 3:1 sort-outs as purchasable part numbers. Note that you can conversely trim your resistances to suit the Gm which you have.

But beware of getting too clever. None of this is precisely specified and it all drifts with temperature. Even the "Square Law" is only a concept, real JFET structures have end-effects and other tweaks.

[fryingpan]

The "outer bounds" are unavoidably 1/Gm @ Idss, and Infinity (for all practical purpose).

Gm is very often in the area of 2,000uMho which is 500 Ohms.

You know how to shim-down "infinity".

Adding a series resistor sets a bound but screws-up the shape of the curve. Now you would hope to select JFETs for 1/Gm in the desired range. But in production Gm varies about 10:1, with 3:1 sort-outs as purchasable part numbers. Note that you can conversely trim your resistances to suit the Gm which you have.

But beware of getting too clever. None of this is precisely specified and it all drifts with temperature. Even the "Square Law" is only a concept, real JFET structures have end-effects and other tweaks.

Luckily, I don't really need the variable resistor to be that precise (I mean, the more precise it is, the better). And being this for a self-build, I can "afford" (not really, heh) to discard FETs that drift too far from my requirements. Of course, if I had something more precise, I'd use it, but are there any other variable resistors that are not digital potentiometers?

[fryingpan]

The "outer bounds" are unavoidably 1/Gm @ Idss, and Infinity (for all practical purpose).

Gm is very often in the area of 2,000uMho which is 500 Ohms.

You know how to shim-down "infinity".

Adding a series resistor sets a bound but screws-up the shape of the curve. Now you would hope to select JFETs for 1/Gm in the desired range. But in production Gm varies about 10:1, with 3:1 sort-outs as purchasable part numbers. Note that you can conversely trim your resistances to suit the Gm which you have.

But beware of getting too clever. None of this is precisely specified and it all drifts with temperature. Even the "Square Law" is only a concept, real JFET structures have end-effects and other tweaks.

Anyway, I've been looking through available JFETs and generally Gm/Yfs is much higher (thus leading to a very low Rds for my purposes; having an minimum Rds in the area of 400-500 ohms would be much better because it would allow greater dynamic range). Do you have any specific suggestions? The only part that looks like it could do, with some hand-picking, is the SST5486 (or its TO counterpart, which is very hard to find anyway).

[iainpunk]

[OT]
made me chuckle...

2,000uMho

> using ohm in reverse
i think you might want to use Si for Conductance in Siemens, named after the German engineer
[/OT]

you can look in to LED/LDR sets
or just use a Jfet with a parallel resistance to lower the total max resistance.
if you feel cheeky, you might try an LED or bipolar transistor as variable resistor, if you don't mind distortions.

cheers

[fryingpan]

[OT]
made me chuckle...

2,000uMho

> using ohm in reverse
i think you might want to use Si for Conductance in Siemens, named after the German engineer
[/OT]

you can look in to LED/LDR sets
or just use a Jfet with a parallel resistance to lower the total max resistance.
if you feel cheeky, you might try an LED or bipolar transistor as variable resistor, if you don't mind distortions.

cheers

My problem is not with maximum resistance; it's with minimum resistance. Having an minimum Rds at around 400 ohms would make it much easier to control.

[iainpunk]

[OT]
made me chuckle...

2,000uMho

> using ohm in reverse
i think you might want to use Si for Conductance in Siemens, named after the German engineer
[/OT]

you can look in to LED/LDR sets
or just use a Jfet with a parallel resistance to lower the total max resistance.
if you feel cheeky, you might try an LED or bipolar transistor as variable resistor, if you don't mind distortions.

cheers

My problem is not with maximum resistance; it's with minimum resistance. Having an minimum Rds at around 400 ohms would make it much easier to control.

maybe look at the BF245, its kind of common, and has good Rds-on rating, its Yfs is between 3 and 6.5 miliSiemens or between 150 and 350 ohms at Vgs=0v
check the datasheet, if you want.

cheers

[Rob Strand]

Anyway, I've been looking through available JFETs and generally Gm/Yfs is much higher (thus leading to a very low Rds for my purposes; having an minimum Rds in the area of 400-500 ohms would be much better because it would allow greater dynamic range). Do you have any specific suggestions? The only part that looks like it could do, with some hand-picking, is the SST5486 (or its TO counterpart, which is very hard to find anyway).

If you limit how close to zero Vgs gets it limits Rds.  The other way is to just add a series resistor.
If those aren't options you might need to explain a few more details of your requirements.

[Eb7+9]

 

ohmic region equation for a jFET is given in 11.6.1 here:

https://eng.libretexts.org/Bookshelves/Electrical_Engineering/Electronics/Book%3A_Semiconductor_Devices_-_Theory_and_Application_(Fiore)/11%3A_JFET_Small_Signal_Amplfiers/11.6%3A_Ohmic_Region_Operation

take partial derivative of Id against Vds and you get (the slope equation)

del(Id)/del(Vds) = 2 Idss / V^2

so Rds(on) calculated "at the origin" is given by Vgs(off)^2/2Idss

data sheet for BF245 provices corresponding characterizing range limits for both variables

VGSoff gate-source cut-off voltage ID = 10 nA; VDS = 15 V 0.25 8.0 V

IDSS drain current VDS = 15 V; VGS = 0;
BF245A 2 6.5 mA
BF245B 6 15 mA
BF245C 12 25 mA

at the 0.25v/2mA and 8.00v/25mA outer corners we get
15.6ohms AC and 1.28kohm AC Rds(on) limits respectively

since these two data points form the opposite corners of a rectangle
it's an easy algebra problem to find the exact Vp-Idss combination that gives you
your 400ohm target ... and the same would go for any other jFET

take my 2n5457 data for example ... see a device in there that comes close ??!

https://viva-analog.com/characterizing-and-matching-2n5457-jfet-transistors/

[fryingpan]

Anyway, I've been looking through available JFETs and generally Gm/Yfs is much higher (thus leading to a very low Rds for my purposes; having an minimum Rds in the area of 400-500 ohms would be much better because it would allow greater dynamic range). Do you have any specific suggestions? The only part that looks like it could do, with some hand-picking, is the SST5486 (or its TO counterpart, which is very hard to find anyway).

If you limit how close to zero Vgs gets it limits Rds.  The other way is to just add a series resistor.
If those aren't options you might need to explain a few more details of your requirements.

The idea is basically to take a Fetzer Valve (which basically simply "freezes" a JFET to an approximation of one specific area of operation of a triode) and swap the source resistor with a voltage controlled resistor, and have it vary its resistance from low values at low input to high values at high input. The black curve in the picture I posted above shows the ideal source resistance for 1 - normalised input voltage (where 1 equals Vp basically) for a "typical" J201 (one with 0.93 |Vp| and 0.96mA Idss). I know I shouldn't get too hung up on these numbers but I first want to simulate it well (and choose a range of desired values), and just later go through the process of acquiring and sorting the JFETs.