What are the most desirable characteristics for a FET in an input buffer?

[midwayfair]

I did some searching but couldn't find a question that asks this, though I'm sure it's been covered often.

I'm trying to pick "the best" common FET for an input buffer. Jack Orman's site notes that

if the input voltage exceeds the gate-source forward voltage plus the bias voltage at the source, the signal will be clipped

The datasheet for, say, the 2n5457 has Vgs noted at the top as 25V (!) and later as 2.5V @ 100uA. They also have a Vgs "cutoff" voltage, which sounds more like the sort of thing I should be concerned about as far as clipping.

So I guess my questions are:
1) Which of these characteristics should I be looking at when I dig through datasheets?
2) Is there a particular range that is best in a 9V or 18V single supply circuit?

[jubal81]

It's expensive and looks like hard to get, but I just saw this great article about the LSK489. There a ton of great info in the article, though. The first thing the author points out is a low gate-to-drain capacitance.

http://www.edn.com/design/analog/4419861/Product-How-to--LSK489-Application-Note

[R.G.]

You want a JFET that will put the source at about half the power supply, possibly as much as one diode drop lower.

A JFET is a resistor which the gate can squeeze off to some higher resistance by the electrical equivalent of the garden-hose effect. Literally, the gate reverse voltage causes field effects that make the channel material be pinched off, like squeezing a garden hose's diameter. When the gate is at the same voltage as the source, the channel is able to act as though the gate is not there.

You actually can raise the gate as much as a diode drop - which starts really, really low, and current in the diode increases exponentially as the voltage increases. So things get nonlinear when you try to raise the gate above the source on the theory that it will make the channel bigger and conduct more. Probably what's happening is that the gate is inserting more charge into the channel. But then,

blah, blah, blah, blah. If you're doing a buffer, don't go there. Assume Vgs=0 is all you can use.

Ideally, the source will sit nearly at half the power supply voltage. This gives you the maximum range for the gate voltage to pull the source up towards the power supply and let it drop towards ground by turning the JFET off completely.

If you're trying to do the self-bias trick where the gate is grounded by a resistor and the source floats up a bit, held up by feedback and transconductance, you want VGSoff to be 4.5V (or otherwise half the power supply) and then to tinker the source resistor to make the source sit at this same 4.5V up from ground when the gate is at ground - except that there are feedback effects that stabilize it there, and there's the Idss maximum current to worry about, and the transconductance, and variations of maybe 5:1 on each parameter... ARGGGHHHH!!

For the self-bias setup, pick a Vgsoff of about 3-6V, and diddle the source resistance to make it come in.

The JFET has to have enough transconductance so that when the source is up at nearly the power supply, the device has enough current capability left over to drive the source resistor. JFETs simply won't conduct more than Idss, so your source resistor HAS to be greater than the power supply divided by Idss; probably quite a bit bigger.

All of this is getting you nowhere, I bet. Sorry.

This used to be done graphically, using "typical" curves of Id versus Vds with Vgs as a parameter. Draw the load line (i.e., source resistor) in, note the variations in ID from min to max, note the Vgs variations, shift over to the Id-versus-Vgs chart with those currents, make a guess at a middle of the road Vgs, and check that against Vgsoff.

I would pick a device noted as a low noise amplifier with a Vgsoff of -3 to -6V, then go diddle the resistor to make it the best I could.

No, first I'd go look and see if I could actually **buy** them. Quite a lot of JFETs are disappearing off the market, as you probably know. Then diddle the resistor on the ones you can actually buy.

Do you begin to see why JFETs never made it big with guys who had to feed families based on the qualities of their designs?

[midwayfair]

Thanks, RG. I'll measure some of the stuff I have here and see what happens with the source resistors. Probably better to experiment than just digging through datasheets that just show me a 6v swing for Vgs. 2N5458 does look a little better than some others, I'll probably start with that.

[brett]

Hi
the J201 is widely available, and is useful for lots of things but not buffers (Vgs too close to zero)
cheers

[midwayfair]

Hi
the J201 is widely available, and is useful for lots of things but not buffers (Vgs too close to zero)
cheers

Yup. That's actually what prompted me to post this. The pedal I'm futzing around with used it as an input buffer, and I knew it was basically crap for that (but wasn't sure why).

[mremic01]

Hi
the J201 is widely available, and is useful for lots of things but not buffers (Vgs too close to zero)
cheers

I've actually used J201s in my AMZ buffers, and the first two I put together were nice and transparent. I put one in the front of a multifx pedal a few weeks ago and it had a tiny bit of tone loss. Not much, but enough to make me rewire it for true bypass. Is there any reason why Jack Orman calls for the J201?

[PRR]

For source-followers with low supply voltage (9V is low) and odd-lot JFETs:

Pick a source resistor 2X to 10X smaller than your load. Load in guitar-chain work is rarely less than 50K; FX loops have been seen as low as 10K. 5K (4k7) is probably a good bet.

Put JETs in one at a time and note the source voltage. It must be significantly higher than your peak signal voltage. Peak output of a guitar is a slippery thing; however I would want at-least 1V, happier with 2V, 3V even 4V is not too high (8V is much too high but also unlikely).

If you only get 0.5V, it will work for much playing but will distort for hotter pups or stronger pick-arms. That's part of why some well-regarded plans work for some folks and not for others; pick-power and pickup sensitivity varies a LOT.

Drain-output stages get MUCH harder. The right answer is to go to 24V supply, then jack-up gates so you have a bunch of voltage drop on the source resistor (by design; not by trim). Then you *know* the bias-point and the current and can select a drain resistor.  In general, unless you "need" the euphonic coloration of a FET and are prepared to tweek, there are better ways ('072) to do gain.

======================

> 2n5457 has Vgs noted at the top as 25V (!)

Datasheet has sections for "Maximum Rating" and "Typical Parameters". KNOW which section you are in.

If you back-bias a JFET Gate with 40V, you punch-through the Gate. It breaks down. We never need to get anywhere near that (a few Volts is typical in circuit) so the sheet-writer does a CYA and tells you not to abuse it with more than 25V.

The Vgd(max) is a similar rating, and in high-voltage work it may matter. If I run a 60V supply and bias-up the Drain to 30V, the Gate-Drain junction feels 30V. The 2N5457 is only rated 25V this way. I bet you a dollar that 99 out of 100 2N5457 will stand 30V for a week. Actually I'd bet on a decade but I won't leave a bet hanging that long. If you do need >25V breakdown, there are other devices with more-bold numbers (may be the same actual parts). If you need >40V breakdown, tough. (Reinfeller used to work with 300V JFETs but they never became common.) As this forum runs on 9V, maybe 24V, I won't go into the workarounds here.

[midwayfair]

Thanks folks. I know that a TL072 or 2072 are the way to go if I'm concerned about having it right the first time every time. But what prompted this was that I went back to fix up an old build (a Madbean Fatpants), which is already designed to use a FET. I ended up having to solder up a completely new board due to some very bizarre issues and a (probably) fried 1044 (I use the 1054 now). I guess I could have squeezed a TL072 on the board but at the time I was anticipating just having a socket for a transistor.

Anyway, I got some time to do some experimentation with the FETs I have here. I used a 10K for the source resistor for the tests just to get a sense of what they did. The supply voltage was 18V and the gate bias resistor was 2.2M. No positive bias on the gate.
I tested:
2n5457 (fairchild from Smallbear and the mystery ones from Tayda)
2n5458
2n5485
2n5952
J201 (fairchild from Smallbear and the mystery ones from Tayda)
MPF102 (fairchild from Smallbear and the ones from Tayda that seem to be the real thing)
2SK30
2SK170

MPF102 ended up with the highest source voltage, about 3.5V (all were above 3V out of the ones I had). I was a little surprised by this based on the datasheets.

The second best was 2N5458, about 2.5V.

2n5457, 5485, and 5952 all ended up around 1.5V-1.7v, despite the differences in Vgs on the datasheets. 2SK30/170 were a little closer to 1.5. The supplier didn't seem to matter in this usage even though I know that many of the Tayda ones have less output when used in a booster.

J201 was all the way down at .7V! And I could indeed hear it distort with some humbuckers.

MPF102 is fairly common, but the 5458 is a little cheaper.

I went back to an MPF102 on the breadboard to mess with the source resistor some, and no matter how big I make the source resistor it never got above a certain level (about 3.6V). None of the other FETs I have on hand get anywhere close to that.

Okay, I know that 3.5V is VERY BIG for a guitar signal, and anything in front of the buffer that might be creating a signal that big is probably a dirt pedal, so any buffer distortion is not likely to be too noticeable or objectionable. But it's certainly nothing like the half supply you guys have mentioned. I don't exactly have a huge collection of FETs, but these are a fairly wide range. Is there something I'm missing?

[PRR]

> down at .7V! And I could indeed hear it distort with some humbuckers.

Thanks for the data-point.

[B Tremblay]

Do you have any 2N3819? That was available at Radio Shack at one point, but maybe not anymore. I remember testing some when we did the revamped Fetzer article and the specs were higher than the MPF102.  I believe that I still have some and I could remeasure them, if that would be helpful.

[midwayfair]

Do you have any 2N3819? That was available at Radio Shack at one point, but maybe not anymore. I remember testing some when we did the revamped Fetzer article and the specs were higher than the MPF102.  I believe that I still have some and I could remeasure them, if that would be helpful.

That would be helpful! Thanks. I'll see if I can locate them for sale somewhere, too. I feel a little safer buying less common FETs since they're less likely to be faked.

[brett]

Hi
if you want a bit more than 3.5V on the source, you could try a 15k or 22k source resistor. As long as the current doesn't drop away too much, the voltage will be higher. Because you increased the R in V=IR.
I love a formula that always works.
I'm sure that a few of us will be interested to hear what the voltage is with more resistance.
thanks

[midwayfair]

Hi
if you want a bit more than 3.5V on the source, you could try a 15k or 22k source resistor. As long as the current doesn't drop away too much, the voltage will be higher. Because you increased the R in V=IR.
I love a formula that always works.
I'm sure that a few of us will be interested to hear what the voltage is with more resistance.
thanks

Like I said, it doesn't increase beyond a certain point no matter how bit the resistor is. Even a 100k or 1M is still only 3.6v.

[B Tremblay]

Using the Fetzer Vp/Idss measurement jig, I tested the two 2N3819 I was able to find.  One had Vp of 3.17V and Idss of 6.57mA while the other had Vp of 4.81V and Idss of 14.7mA

[PRR]

> try a 15k or 22k source resistor

We need voltage AND current. Current to drive the load, also to bring out the Transconductance that R.G. mentioned (in this case, low output impedance).

That's why I'm suggesting more like 5K.

[brett]

Hi
Should have read and thought more before I posted.
Upwards of 10k will be moving out of the zone, not into it. Agree that 5k probably makes sense as a default value.
A couple of years ago Jack Orman wrote an article saying that the 'default' source resistor of 2.4k (sometimes 2.2k) doesn't make sense in source follower buffers and common source boosters.
Plenty of current but too little voltage drop.
cheers