What do VGS off and IDSS in JFETs mean for audio amplification?

[mzy12]

Hello!

I'm trying to understand how JFETs operate in audio amplification and what characteristics are important when considering which ones to use in guitar pedals. I am building a Tillman preamp with a 'tone suck mod' as part of a fuzz I am designing and doing some research on JFETs, I came across this post on freestompboxes.

In it, the OP quotes a passage from a book called Solid State Guitar Amplifiers by one Teemu Kyttala. You can see it on the link I inserted above. The gist of it is that the author says the J201 kinda stinks if you want to use as a buffer/other applications because of it's relatively low VGS(off) and IDSS values (by low I mean less negative, because we are talking about N-Channel JFETs here), saying that a J201 will clip at a peak to peak signal of a couple hundred mV. This is obviously not ideal, as any guitar signal's transients, even among the weakest pickups, will be at leas 1V peak to peak.

Maybe my question here is dumb, I don't know. I couldn't really find anything online about this. But I want to know what exactly we should be looking for in JFETs for small signal audio amplification? What VGS(off) and what IDSS are ideal? What do they even mean in this application? What maths would you have to do to say that the J201 can't handle guitar signals without clipping?

Thanks in advance for any help.

[Clint Eastwood]

When using 'autobias', by means of a source resistor, you want a jfet with a VGS(off) that is higher than the peak voltage of your input signal, otherwise it will turn off when you strum hard. In practice at least 1 volt.
If VGS(off) is high, let's say more than 2 volts, you run into another problem when using a low voltage supply like 9 volts: the low part of the output wave cannot go much lower than VGS(off). So you have less clean headroom, wich is already limited with 9 volts.
Today there are very few jfets with legs still being made. One of those few is the j113, wich works fine in many different applications. Its VGS(off) is about 1.5V, and IDSS around 18 mA, and a pretty high amplification factor.

[R.G.]

Idss is the maximum current the JFET will let through in normal use. A JFET's current range for normal uses is zero current up to Idss.
Vgsoff is the voltage that forces the channel to turn off. The range of gate-to-source voltages useful in normal amplification is Vgsoff up to zero volts, at which the channel current is limited to Idss.
The ratio of a tiny change in channel current to the tiny change in Vgs that caused it is the transconductance (often called Yfs), effectively the gain of the JFET. For a setup with a drain resistor, a change in Vgs causes a change in channel current, and that is reflected in a change in the voltage across the drain resistor. So dVout = Rdrain * Yfs * dVgs, where the "d" just means "a small change in...". High Yfs means high gain.

As for what to look for - like everything else, it depends on what you're trying to do. Idss and Vgs off are limits, where the device will quit responding linearly. If you have to accommodate big input voltages, you need a big Vgsoff or the JFET will turn off or "saturate" on big signals. If you want high currents, pick a big Idss. If you need big amplification, small Vgsoff and high transconductance is the ticket.

The biggest voltage swing you can get out of a JFET drain is with the source grounded and a resistor in the drain that equals the power supply voltage divided by Idss, because the JFET current can only go from zero to Idss, whatever you do on the gate. The biggest general-audio signal you can use on the gate input is from Vgs=0 to Vgsoff, peak to peak. At Vgs=0, the channel current is Idss, and at Vgsoff, the channel current is zero. It is possible to force the gate towards forward conduction, but exactly what happens there depends on exactly what the manufacturer did on the substrate silicon the JFET is made on.

The J201 is an anomalous JFET in that it's got weirdly low Vgs, high Yfs, low Idss. It's intended for high gain, small signal work with high impedances in the industry. Guitar pedal types like it because it distorts easily in ways we happen to like. It's not great for general amplification work or for signal switching because of the low Vgsoff and small Idss.

If you care, here's some "how it works" stuff.
JFETs are a bar of conductive silicon (the channel, from drain to source) with a gate conductor region of the opposite beside most of the middle of the channel. For N-channel devices, the channel is ... drum roll... N-type silicon. The gate is P-type. This forms a diode from the gate lead to the channel. I'll use the N-channel as a model; P-channels work the same way, just the semiconductor type is reversed, as are the voltages and currents. So for N-channels:

If you leave the gate open, the channel looks like resistor for small currents. The value of this resistance is quoted in the datasheet as Rdson.

If you short the gate lead to the source, the resistive voltage drop in the channel interacts with the gate-to-channel diode and causes the channel current to limit at some point, and be effectively constant as more voltage is put across the channel. The current it limits to is Idss.  You can't normally get more current through a JFET than Idss without killing it first.

If you make the gate-to-channel diode reverse biased, it causes a non-conducting region to form inside the channel. This is much like squeezing a garden hose progressively more closed. The reverse bias literally squeezes down the conducting area of the channel and eventually pinches off the current flow. The voltage at which the channel is pinched off is called Vdsoff.

The ratio of how much the channel current is changed per a small change in gate reverse voltage is the trans-conductance and this is listed in the data sheet as either transconductance, or often something like "Yfs" or "Gfs". It amounts to the gain of the JFET. For JFETs this is often milliamperes per volt.

[mzy12]

Thanks for the response, RG. Very detailed and very helpful.

In practice, passive guitar signals can and will reach upwards of 2V peak to peak, so would that imply that for a clean buffer, the JFET would at least want that much in Vgsoff? Or is it something else, as a guitar signal is, of course, AC and will often swing around a reference bias? And is it different in a common drain configuration like a buffer vs a common source configuration, like the Tillman preamp which has no input cap and no V bias applied to the input signal?

[antonis]

Tillman JFET preamp has no input cap 'cause it imagines of a completely zero DC offset..
Which also imaginary stands for bias..
(no Gate leakage current, irrelevant to working temperature, hence bias resistor "infinite" value..)

[R.G.]

In practice, passive guitar signals can and will reach upwards of 2V peak to peak, so would that imply that for a clean buffer, the JFET would at least want that much in Vgsoff? Or is it something else, as a guitar signal is, of course, AC and will often swing around a reference bias?

And is it different in a common drain configuration like a buffer vs a common source configuration, like the Tillman preamp which has no input cap and no V bias applied to the input signal?

Kudos on your analytic thinking. It is different in a common drain setup (also known as a source follower) because the source voltage follows the gate voltage, only having enough error for the change in gate voltage to make changes in channel current to get the source to follow the gate. In this case, the limits on the gate to source voltage depend primarily on the Gfs/Yfs being large enough and the source resistance large enough so the change in channel current can make it all come true before the gate-source error runs out of Vgsoff.

For a common source circuit, generally you need some source resistance to keep the total Vgs swing within bounds, and the smaller Vgs is, the touchier the balancing of drain resistor, source resistor, and Yfs  gets. JFETs in common source are notoriously difficult to bias predictably because the Vgsoff and Gfs variations from device to device are large, even within a single type number.

For Tillman enthusiasts: I haven't looked at the Tillman setup in a long time. I'll have to go refresh myself before I can say much about it.

[PRR]

You can select random/sorted devices to suit your problem, OR you can design the circuit to fit your needs.

Negative Feedback was a game-changer in tube telephony and can be for you (if your priest/imam/rabbi allows).

Vgs(off) is rather moot in cathode (source) follower b/c your load resistor drop may be far larger than a Vgs(off).

Unless you are running 9V supply as many folk here do.

TL072 is more accurate and much cheaper than any JFET sold today. There are other chips with better audio and/or power parameters for a few cents more.

[Rob Strand]

JFET's aren't really "clean" buffers from a HiFi perspective.   When the output swings Vds must change and when Vds is a small the JFET operates in the "Triode Region".

The signal loss from a JFET buffer can be reduced using a current source load instead of a source resistor.   That's actually what the Boss Waza pedals do.   (Search for Waza buffer on this forum.
https://www.diystompboxes.com/smfforum/index.php?topic=122929.0
https://www.diystompboxes.com/smfforum/index.php?topic=122929.msg1161023#msg1161023
)

[mzy12]

Kudos on your analytic thinking. It is different in a common drain setup (also known as a source follower) because the source voltage follows the gate voltage, only having enough error for the change in gate voltage to make changes in channel current to get the source to follow the gate. In this case, the limits on the gate to source voltage depend primarily on the Gfs/Yfs being large enough and the source resistance large enough so the change in channel current can make it all come true before the gate-source error runs out of Vgsoff.

For a common source circuit, generally you need some source resistance to keep the total Vgs swing within bounds, and the smaller Vgs is, the touchier the balancing of drain resistor, source resistor, and Yfs  gets. JFETs in common source are notoriously difficult to bias predictably because the Vgsoff and Gfs variations from device to device are large, even within a single type number.

This is starting to form a much clearer picture in my head now, thanks. So VGS(off) isn't a huge deal in source follower designs?

Vgs(off) is rather moot in cathode (source) follower b/c your load resistor drop may be far larger than a Vgs(off).

Unless you are running 9V supply as many folk here do. 

How much headroom would you be leaving on the table with 9V? Assuming that's the right question of course ahaha.

JFET's aren't really "clean" buffers from a HiFi perspective.   When the output swings Vds must change and when Vds is a small the JFET operates in the "Triode Region".

Thanks for pointing that out. I suppose I should have prefaced all this by saying I'm not looking for an ultra clean buffer, but trying to understand JFET choices in already existing designs with characteristics I want - like a modified Tillman preamp - or in cases where I'm kind of stuck with a design and want to know the best choices available to me. For example, over on the PedalPCB forums, a brilliant chap by the name of Dan Schumaker has graciously released build docs of his copy of the Boss DM-3 and the first stage has a source follower buffer using the J201, of which I am skeptical of its choice again to its uncharacteristically low IDSS and VGS(off) values. Perhaps, however, my skepticism is unfounded and I don't have the knowledge to offer a better alternative? That's where all these questions come in . On that note - is a J201 okay as input buffer that you want to be as clean as is reasonably possible?

For Tillman enthusiasts: I haven't looked at the Tillman setup in a long time. I'll have to go refresh myself before I can say much about it.

Haha I'm no enthusiast, it just got suggested to me that I could use it in front of a two transistor fuzz and tweak the gate resistor and add a voltage divider (potentiometer) to the output to get the right 'tone suck' desired. I don't think I need to go into more detail on that lest I get extremely off topic! I'm probably overthinking the transistor choice here if I'm going to intentionally degrade the tone, but I at least want to understand what I am doing.

The signal loss from a JFET buffer can be reduced using a current source load instead of a source resistor.   That's actually what the Boss Waza pedals do.   (Search for Waza buffer on this forum.
https://www.diystompboxes.com/smfforum/index.php?topic=122929.0
https://www.diystompboxes.com/smfforum/index.php?topic=122929.msg1161023#msg1161023
)

Thank you, I will check those out.

[Ben N]

Idss is the maximum current the JFET will let through in normal use...

Bookmarked!

[R.G.]

So VGS(off) isn't a huge deal in source follower designs?

It's much less important, although it's always an issue to consider. JFETs will need their gates biased to somewhere between 0V relative to the source and -Vgsoff. If Vgsoff is several volts, this gets big and you wind up spending more of your available power supply voltage elevating the source.
That's not too bad with lots of supply voltage, but with 9V you don't have many volts to play with. For a source follower, you want something like half your supply voltage on the source anyway, so the gate bias voltage is built in and doesn't subtract so much from the available power supply.
For common source circuits, spending half your power supply for biasing is much more severe.
The real tricky part with JFETs is biasing in the face of variation in Vgsoff. From InterFET's data sheet for the J201, Vgsoff is between -0.3 and -1.5V. A perfectly good J201 might be anywhere between those two. You have to design your circuit to work with anything in that range. That gets tricky, and always involves tinkering negative feedback into it to correct for the variation.
Again, not too bad in a 9V circuit for a J201. But other JFETs make it harder. Looking at InterFET's master table (https://www.interfet.com/jfet-master-table/) is eye-opening. Many of them have Vgsoff maximums over 6V. Some don't list a minimum at all. It's difficult to do old-school designs to bias an amplifier with a huge Vgsoff range with resistors. Servo circuits work well, but then you have to use an opamp to do it. JFET-bipolar pair circuits work very well, but again, you're up for some heavier design work, and the feedback can wipe out any nice distortion you're trying to get from the FET.
There are really two classes of JFETs - amplifiers and switches. For an amplifier JFET you want low and small-range Vgsoff, the J201 being an exceptional case even for amplifier JFETs. For a switching JFET you usually try to get big Vgsoff, because that gives you a bigger range of switched signals before the signal itself forces the JFET between conduction and off.  I really, really like the J175 and J176 p-channel devices for 9V signal switching. In this application, you bias the drain and source to 0V/ground and hook the gate up to a diode/resistor/capacitor slow-down circuit that lets the gate drift to "open" when it's not pulled up. If you do nothing to the gate, it conducts. If you pull it high, it softly turns on with speed determined by the slow-down circuit. The on resistance is 125-250 ohms. Vgs for the J175 is 3 to 6 volts, so in a 9V circuit, you can switch signals of 3V p-p solidly.

How much headroom would you be leaving on the table with 9V? Assuming that's the right question of course ahaha.

It's close. The right question is how much of your power supply is lost to biasing. If you do simple resistor-current biasing, it can be up to nearly Vgsoff maximum, and is usually something like half Vgsoff. Not too bad with a 1.5V J201, but things get trickier with something like a -8V 2N3819. The wild variation in Vgsoff, Idss, and Gfs really start to bite when the power supply is only a couple of times bigger than Vgsoff. This is a fundamental problem which makes nearly all JFET circuits in 9V pedals need bias tinkering to get them biased properly.

is a J201 okay as input buffer that you want to be as clean as is reasonably possible?

It's a definite maybe. The J201, as noted, is good-ish for an input source follower, but won't be "as clean as is reasonably possible" because JFETs are just not that in a hifi sense. Not bad for guitar work, but not low distortion by today's terms. The current source sub for a source resistor is one way, as Rob notes, but circuit complexity goes up.

I would be terribly tempted to use an N-JFET/PNP transistor pair. 

The JFET gives you 0-volt bias through a high resistance for input impedance. The JFET source connects to a resistor to ground, providing the source feedback and the bias voltage. The JFET drain pulls current from a resistor to supply voltage and the base of the PNP. The PNP has its emitter on the power supply through a 10 ohm and its collector through a resistor to the source of the JFET. The 470p cap compensates it so it doesn't accidentally oscillate at several MHz on funny loading.
The circuit is essentially the input preamp for the last models of the Thomas Organ Vox Beatles, adapted for the J201 and a 9V supply. Funny thing – when its overdriven, it has a nice soft overdrive, not hard clipping, and it gets progressively more asymmetrical as you push it harder. Not bad on its own. With signals under a volt or so peak, it's not distorted and with the caps shown it is flat from 2Hz to 24kHz. The output at the collector of the PNP goes to nearly 7V p-p, at a gain from the input of about 8. The output can be taken from the junction of the 8.2K and 1K for a lower output signal. Or you can put a pot on it for variable output.

I'm probably overthinking the transistor choice here if I'm going to intentionally degrade the tone, but I at least want to understand what I am doing.

Again, kudos for wanting to understand before launching!

[mzy12]

I would be terribly tempted to use an N-JFET/PNP transistor pair. 

The JFET gives you 0-volt bias through a high resistance for input impedance. The JFET source connects to a resistor to ground, providing the source feedback and the bias voltage. The JFET drain pulls current from a resistor to supply voltage and the base of the PNP. The PNP has its emitter on the power supply through a 10 ohm and its collector through a resistor to the source of the JFET. The 470p cap compensates it so it doesn't accidentally oscillate at several MHz on funny loading.
The circuit is essentially the input preamp for the last models of the Thomas Organ Vox Beatles, adapted for the J201 and a 9V supply. Funny thing – when its overdriven, it has a nice soft overdrive, not hard clipping, and it gets progressively more asymmetrical as you push it harder. Not bad on its own. With signals under a volt or so peak, it's not distorted and with the caps shown it is flat from 2Hz to 24kHz. The output at the collector of the PNP goes to nearly 7V p-p, at a gain from the input of about 8. The output can be taken from the junction of the 8.2K and 1K for a lower output signal. Or you can put a pot on it for variable output.

Thanks for all this RG, and sorry for the delay in getting back to you. There's certainly a lot of information there to chew on. I think I understand what I'm looking at now. Perhaps I should think about redesigning that DM-3's buffering sections with an opamp like an OPA1678. That would be much easier than trying to find the right JFET. For my adventures in the Tillman preamp weirdery, perhaps I should just settle for a J201. It is quite frustrating that there are not really any JFETs that have a generous VGSoff that also have an upper limit that doesn't come scraping up against the 9V of a guitar pedal power supply. Just the way the world works huh 

I am curious about that Thomas Organ Superbeatle now. I've looked at a few superbeatle schematics online but I can't find one with an input similar to the one you mentioned above. Would you happen to have a model number/schematic for that?

[diffeq]

I can almost feel Antonis' urge to bring bootstrapped BJT buffer options to the table.

[R.G.]

For my adventures in the Tillman preamp weirdery, perhaps I should just settle for a J201. It is quite frustrating that there are not really any JFETs that have a generous VGSoff that also have an upper limit that doesn't come scraping up against the 9V of a guitar pedal power supply. Just the way the world works huh 

The J201 has a VGSoff of 0.3 to 1.5, likely about 0.8 to 0.9V, so it will be clipping the guitar signal on peaks. This may be OK. A lot depends on where the 6.8K/2.2K biases with the exact device you get.  The Idss is 0.2ma to 1ma, so a raw guess says to expect something in the middle, about 0.6ma. If you get a "typical" device, with a Vgs of about 0.45V for middle-of-the-road biasing, that would put about 0.45V on the 2.2K source resistor, giving a desired current of 0.45V/2200 = 2ma. Ooops. Busted. The J201 probably won't go there, not enough Idss. I'll go look at some JFETs that might.
The ramifications of JFET variability in Vgssoff, Idss, and  Gfs led the industry to sorting JFETs, and then largely abandoning them. It's great to think that you can just design circuits smarter/faster/better to use that nice high input impedance, but given the variability in characteristics, that road eventually leads to sorting/selecting, and then finding another way. On one-off circuits, you can always tinker the resistor values and get something that works. For instance, in the Tillman preamp, you might change the 6.8K/2.2K resistors up to 12K/4.7K and get better results, as the current needed in the JFET then goes down while still giving the same voltage swing.

I am curious about that Thomas Organ Superbeatle now. I've looked at a few superbeatle schematics online but I can't find one with an input similar to the one you mentioned above. Would you happen to have a model number/schematic for that?

The preamp I mention is from the last series of Beatle/Guardsman/Buckingham/Viscount, the V11*3 series. The earlier series used bootstrapped bipolars. It is possible to get over 1M effective input impedance by using bootstrapped bipolars. JFETs just make it easier.

@diffeq: Yep, I expected to see something like that earlier!

[antonis]

JFETs, V_GS, I_DSS, pinch-off, g_m spread, etc..    

Too much mind trouble for a bit quieter and a bit (OK, more than a bit..) higher impedance device..

P.S.
I DO like R.G.'s pseudo-CFP above but I'd use a lower value Drain resistor (1k to 3k3 say, instead of 10k)..
(always depending on particular JFET V_GS & g_fs)

>The output can be taken from the junction of the 8.2K and 1K for a lower output signal<

Correct me if I'm wrong but I presume in such a case the whole configuration should work as a voltage follower..
(almost 100% feedback, hence no signal gain at all..)


@diffeq: OP's queries concern JFET parameters.. I think there is no room for personal aspects about impedance, linearity etc..
(I do know that you know that I know you were joking..)

[Rob Strand]

I DO like R.G.'s pseudo-CFP above but I'd use a lower value Drain resistor (1k to 3k3 say, instead of 10k)..
(always depending on particular JFET V_GS & g_fs)

In the waza thread I went through some of the buffer options and discussed some pro's and con's.    The feedback type does do a good job.   If you used it as an output buffer it would be wise to add a resistor in series with the output line (100R to 1k - value TBD) in order to stop capacitive loading making the buffer oscillate (since it's a feedback amp).

Maybe somewhere around here,
https://www.diystompboxes.com/smfforum/index.php?topic=122929.msg1161238#msg1161238

[mzy12]

I can almost feel Antonis' urge to bring bootstrapped BJT buffer options to the table.

"Bootstrapping BJTs is a pathway to many abilities some consider to be... unnatural."

Seriously though, bootstrapping BJTs is another slightly mind-boggling topic. It honestly deserves a thread of it's own, but I imagine many-a-one on the subject matter has been made before. I dunno if you've ever taken a gander at Douglas Self's book, Small Signal Audio Design, it contains some truly insane configurations that can get up to 61M ohm input impedance with just five MPSA42 transistors. I don't think that's ever something someone in the guitar pedal world would need, but even the three transistor bootstrapping configuration gets you to a 6.1M Zin. I feel my neurons buckling under the weight of the information in that book every time I open it up.

[Rob Strand]

Seriously though, bootstrapping BJTs is another slightly mind-boggling topic. It honestly deserves a thread of it's own, but I imagine many-a-one on the subject matter has been made before. I dunno if you've ever taken a gander at Douglas Self's book, Small Signal Audio Design, it contains some truly insane configurations that can get up to 61M ohm input impedance with just five MPSA42 transistors. I don't think that's ever something someone in the guitar pedal world would need, but even the three transistor bootstrapping configuration gets you to a 6.1M Zin. I feel my neurons buckling under the weight of the information in that book every time I open it up.

Probably so.

As a first approximation Collector bootstrapping is conceptually simple as it make the collector resistor look like current source; which will be in Self's book.

Booststrapping the input stage for higher input impedance needs a lot more work.  Ideally it's an infinite input impedance but because BJT buffers have less than unity gain you have to take that into account. That complicates things quite a bit and the input impedance is never infinite.   A great job for spice.   Input bootstrapping only works if the amplifier isn't clipping.  As soon at the stage clips your input impedance will drop to a low value, often lower than a non-bootstrapped stage.

[R.G.]

I DO like R.G.'s pseudo-CFP above but I'd use a lower value Drain resistor (1k to 3k3 say, instead of 10k)..
(always depending on particular JFET V_GS & g_fs)

>The output can be taken from the junction of the 8.2K and 1K for a lower output signal<

Correct me if I'm wrong but I presume in such a case the whole configuration should work as a voltage follower..
(almost 100% feedback, hence no signal gain at all..)

Yeah, the original circuit had a 1.2K resistor in the drain of the JFET. In reality, that resistor sets the portion of the JFET drain current that goes through the PNP base versus being bypasses. It's effectively a transconductance-lowerer. In the sim runs I had to kick it up to 10K to get better results with the low-Idss J201. At 1.2K, the thing wouldn't bias the PNP collector much above ground. The 10K forced more of the JFET drain current through the PNP base-emitter.
Yes, taking signal from the lower resistor makes it a follower.

The whole circuit can be viewed as an odd opamp, with a (very!) high input impedance noninverting input, a low impedance inverting input (JFET source), low overall gain, and higher frequency response than the "real" opamps of its day.

Re: CFP buffers, see http://www.geofex.com/FX_images/Onboard_Preamp.pdf

[PRR]

Douglas Self's book, Small Signal Audio Design, it contains some truly insane configurations that can get up to 61M ohm input impedance with just five MPSA42 transistors

In the late 1950s, a Darlington Triple was reported at 300Meg. Details in IRE and I let my subscription lapse.