jfet substitutions

[loudthud]

[ATTACH=CONFIG]34843[/ATTACH]

Peer review, please: this is the test rig I put together ( +15/-15 bipolar power supply, 250k linear pot, led and resistor, breadboard). each of the j111 transistors I tried started to dim the LED at about -6.5v GS, fully off at almost -8v GS.

Since each of the 5 or so pseudo-randomly picked samples from my batch of 100 (still on the tape strip) all agreed, do I assume that this is a reasonable result? Do I need a better test method?

That circuit should get you close. To get closer you need another meter to measure the Drain current. The data sheet specifies Vgs(off) at 1uA Drain current.

[eschertron]

That circuit should get you close. To get closer you need another meter to measure the Drain current. The data sheet specifies Vgs(off) at 1uA Drain current.

Heck yeah I'd need another meter. I'm sure the drain current was several mA to light up the LED. Like I said, the LED started to dim a volt or so earlier. Enough to block a uA of current, sure!

Edit: I like the examples you posted, LT. Can you cite the source for that? There's no info on the blueguitar website to suggest where the schem I have came from.

[loudthud]

The problem with most JFET circuits is that people want to bias them like tubes. The commonly available JFETs have too much process variation to work well in those circuits unless you have a trimmer pot or closely selected parts. The current source (or long tail) in the Source circuit with a big bypass cap takes care of that. The Gate bias is set to whatever it needs to be automatically. There are other ways to do this, but this one is the simplest.

Just about any JFET will work in this circuit as long as the Idss is greater than about 10mA.

You do need to protect the Gate on the input. Excessive Voltage will kill the JFET.

I added a diode to protect the 100uF cap. It could see reverse Voltage depending on how the power supplys come up.

The switching type JFETs usually have plenty of transconductance at low currents, although you will see almost a 2 to 1 variation in any batch. The noise is low because it is largely dependent on the channel resistance.

Sunn used a circuit like this in the Concert and Coliseum Series preamps in the early 70's. -15V was used to bias the JFETs and +25V was used in the Drain circuit.

[J M Fahey]

This is not exactly true. Those high pinchoff switching JFETs make fine amplifiers if you bias them right. A long tail resistor in the source circuit will force the current to be whatever we want. A big capacitor will charge to whatever Gate bias is required. You can easily get a gain of 100 or more. You may need to actually limit the gain with unbypassed Source resistors.

Edit: JM won't believe this but I just looked at a PN4391 which is close to a J111. The transconductance at 20V and 1mA is 5000umho. That means in this circuit with a 100K load resistor the gain will be... you do the math but sit down first.

Dear Loudthud, not to disagree, and even less with you, but I have used all kinds of Fets in practical, commercial circuits, not in simulations, and the real world is ... sobering.

To begin with, there's way too much dispersion in parameters, in this case transconductance, to trust the published value you posted.
Yes, it is in the datasheet and you *may* have a few which match it, but in general they will be all over the place.

Buying 100 to have 10-20 usable ones may work for a hobbyist, definitely not for a commercial maker.

Look at the Tayda datasheet mentioned above, it's *official* transconductance rating has 2:1 spread

Forward Transconductance gFS (Max / Min): 0.0035 S to 0.007 S

and it's a "good" FET made by Fairchild; modern ones either are all over the place or are consistent: bad (high Vp, suitable for switches)as Escherton found: consistent Vp (LED full turn off) around 8V ... ouch !!!! .... or harder to find and expensive, as J201 .

As of the PN4393 example: think again

FETs "are pentodes" so it's tempting to do what you did: calculate "what gain would I have with, say, 100k?" and get shocked with the impressive result.
As you said, 100X , way more than a 12AX7 which with the same 100k plate resistor and passing around 1mA will yield around 50X .

So .... where's the problem .... if any?

Well, there is a catch, sadly.

Fact is that 100k @ 1mA (which is a sensible value, both for impedance and for current) mean:
* a 100V voltage drop across the resistor.
* so for a symmetrical sinewave we need a 200V supply, dropping 100V across the resistor, 100V across the FET .
* so you need a 200V FET ... good luck finding one.

I have used FETs since forever (think early 70s , starting with 5:1 spread TIS 58) , most regularly available ones stand around 25V , so far only found one standing 40V , so a realistic +V rail for a standard TO92 FET is +25V .

So forget the classic 100k load resistor
I have settled long ago on 25V supplies and 22k load resistors as the best compromise on many areas, including that I have a high yield of usable FETs under those conditions, and measured gain is consistently around 20X

Not bad but not headline news either.

I also have tried all kinds of biasing systems, from "tube style" self bias source resistors to, say, 15k resistors connected to -15V rail which guarantee around 1mA no matter what I plug there (similar to what you suggest) , to LM358 based DC servos (one Lab Series amp does that) to constant current sources made with a bipolar TO92 transistor .... after many trials settled on selecting them by Vp and ussing the proper source resistor to get around 1 mA or 0.5mA .

As said before, I *consistently* get around 20X gain which is fine if I design around that, and, most important, with self bias gates clip like grids, coupling caps get charged and wreck bias like tubes, transconductance follows that bias shifting and gain reduces when overdriven (what tubes do) so I get "natural compression" and in general "behave like tubes" which is the main reason to use FETs in MI amplifiers.
Forced bias such as relatively large resistors going to -15V or servos, do not allow that, they are rigid masters, and kill the "tubeyness" so I avoid them now.

Otherwise I wouldn't even consider them , they are a very poor choice compared to a TL072 which can really give me 100X gain, is not fussy at all with supplies, is as consistent as can be and I get two gain stages for 40 cents.
Of course they do not clip like FETs or Tubes, nobody claims that.

I have also made FET long tail PI and still have 20X gain .... not surprising if you consider that one FET behaves as a cathode bypass for the other, while the large tail resistor keeps average current constant.

The limiting factor in FET designs is max Vds they can stand, available gain is directly proportional to supply voltage, that's one reason why all those horrible "I believe I'm a Tube" ROG circuits are so abysmal: they pull out of the blue cathode bias resistors, straight values from the original tube circuits, with absolutely no measuring, no selection, nothing, pure Magic thinking (hey!!! if Leo used 1k5 and Jim 2k7 and 820 r then I will also) and have 9V supplies .

Of course, they "bias" them (NOT !!!) with drain trimmers, and *think* they are doing their homework because they see "100k" printed on them ... fact is that once they bias drains to +4.5V , the actual trimmer value can be anywhere ... usually 10k or less.
Imagine what that does to gain.

The only commercial FET which consistently performs well there is J201 ... not casually an around 0.5V Vp part.

But other Fets? ... not even close.
J11x class? even less.

Yes, they will "bias" (after all those multiturn trimmers can go down to 0 ohm, if needed )
Yes, they will have *some* gain .
No, they will be nowhere close to the original Tube circuit they claim to be emulating.

So in a nutshell, regular FETS are fine to excellent in +25V fed preamps, think Randall RG or old Yamaha, for pedals only J201 (that I know of) perform acceptably, as switches or in compressor/limiters all the cheap ones are fine.

[bsco]

This is exactly why I like this forum....There is so much information that is freely given.....and everybody here freely helps everybody.....Where else can one find this kind of support....Fantastic.....

[Pdavis68]

I mentioned Tayda Electronics' coupon codes on Facebook and just thought I'd follow up and mention that they put out a new coupon code yesterday and it expires on the 20th, so if you have stuff to order from them, now's the time.

[Pdavis68]

* so you need a 200V FET ... good luck finding one.
Juan,

I may not understand completely what you mean here, but can't you cascode a JFET with a power MOSFET on top and let the MOSFET handle the voltage? I've seen stuff about doing that. Would that not deal with the voltage issue? Like I said, I may not be understanding the issue.

Also, as long as we're on the topic, can you tell me how to determine Vp from the datasheet? It always seemed to me to be dependent on Vds, but then i see it referred to as a single number here and other places. I confuse easily.

[loudthud]

This thread is about a JFET to MOSFET cascode. Look at the link in post #4.

Vp (V pinchoff or Vgs(off)) will be specified with min max range at specified Vds. There is usually a Drain current specified like 1uA. Usually in a production environment tests are done on a pass-fail basis. For the J111 Vgs(off) is specified -3V min, -10V max at a Vds of 5V amd Id of 1uA. The test might be something like this:

Source in grounded
Apply 5V to Drain
Apply -3V to Gate. If Drain current is greater than 1uA PASS else FAIL.
Apply -10V to Gate. If Drain current is less than 1uA PASS else FAIL.

Other tests are run to insure that the part is operational and meets specs on the data sheet.

There are 300V JFETs made by InterFet. They are quite expensive.

[Pdavis68]

Ah yeah. So far up on the message I had completely forgotten that it was a cascode.

So doesn't that address the voltage issue Juan raised? Or is he talking about something else? Because the IRF820 has a Vds of 500V.

[loudthud]

To begin with, there's way too much dispersion in parameters, in this case transconductance, to trust the published value you posted.
Yes, it is in the datasheet and you *may* have a few which match it, but in general they will be all over the place.

JFETs are square law devices. The transconductance varies with drain current. When you look at the Drain curves, the higher the current, the farther apart the lines are. Because transconductance is often not specified on the data sheet of switching JFETs, and if it is, it's not anywhere near where I intend to operate the parts, I just measure it but I rarely need to know it. Most of the time I'm just drawing load lines on the Drain curves.
　
I have a Tektronix 576 Curve Tracer. This instrument has a very usefull feature that I use all the time on JFETs. Since JFETs are depletion mode devices, first you must INVERT the step generator. A knob adjusts the Volts per Step from .05V to 2V. Another knob adjusts the number of steps from 1 to 10. In addition there is an offset facility (10 turn pot with vernier) that can add an offset (aide or oppose) to the output of the step generator. Finally there is a switch that allows you to attenuate the Volts per Step by 10X, but it does not attenuate the offset. In the curves posted below the Volts per Step is set to 1V, the 10X attenuator is activated so the steps are 0.1Volt. The offset is read from the dial and noted as Vos on each pic.

Edit: Because the offset can vary quite a bit between parts even from the same batch, I had to come up with a circuit that would automatically adjust because I don't like trimmer pots. The circuit I posted on SSGuitar (Look Ma, no pots) with the servo was an effort to do that but I abandoned that circuit because it does have a problem with bass as somebody noted. Actually, the way Sunn and the Lab amps avoided that was to have a cap on the input rolling off the bass.

[J M Fahey]

* so you need a 200V FET ... good luck finding one.
Juan I may not understand completely what you mean here, but can't you cascode a JFET with a power MOSFET on top and let the MOSFET handle the voltage? I've seen stuff about doing that. Would that not deal with the voltage issue? Like I said, I may not be understanding the issue.
Also, as long as we're on the topic, can you tell me how to determine Vp from the datasheet? It always seemed to me to be dependent on Vds, but then i see it referred to as a single number here and other places. I confuse easily.

Yes, I've also seen that and experimented .
Cascodes are excellent, flat, linear, wideband, high gain, thermally stable, you-name-it , the best choice ... for a scope vertical amp, a TV tuner front end, etc. ... the opposite of what we want in a guitar amp.

If I use a FET I want to use it as close to a tube as possible, because I want it to react to the input signal and modify it, not just linearly amplify it.

FWIW I have made and sold tube replacements, based on a selected FET and a high voltage *bipolar* transistor per "triode" , a 2N5401 for up to 160V supplies or an MPSA42 for higher voltages , although these have lower Hfe and complicate biasing a little.

As of determining Vp, it's the gate bias needed to completely cut it off.

Since leakage currents ara a fact of life, they usually specify, not "0" mA which might be impossible to match, but, say, 10nA or a similarly low value.
It should be in the datasheet, either as "a number" or shown in some curve.

But your Led turning off value is close enough.

My matcher, which I use to find preamp friendly FETs (because that's what I use them for) ,finds "voltage necessary to pass 1mA" .

[eschertron]

So I cobbled together a test rig tonight based on the drawing I had (the PDF in post#4), along with some component values based on preliminary 'bar napkin' calculations. Using the J111 fet, I got some decent results.
Vs = 7.23vdc (right in line with the LED testing I did)
Vd = 19.57vdc
Rs was 4.7k for this test, so the jfet is dissipating about 20mw. No sweat.

With Rs unbypassed, the gain was ~20. Not too hair-raising. Fully bypassed, though I got ~1300! I tried a couple different resistors in series with the bypass cap:
no cap (inf resistance) = 20gain
10k + cap = 28gain
3.3k + cap = 46gain
1k + cap = 96gain
I'll draw up a table for this.

[edit] of course, in my limited understanding, I assume the next jfet I plug in will have a completely different gain structure [/edit]

The input head room for this jet was ~14vdc. Unless it's a coincidence, it doesn't surprise me that it's about +/- Vgs_off.

I'm thinking of using this as a clean input stage, enough input HR so that a 9v pedal can't clip it, and just enough gain to simulate a first tube stage. I say let the rest of the gain stages do the work. This is assuming that I'm using enough triodes to obliterate any audible effect of stage one, or I need to reduce the tube count!

[J M Fahey]

Cool
So you are using a cascode with a high voltage supply?
In that case yes, you will have tons of gain, definitely less hum (and maybe less hiss) than a regular triode stage and it will work well as a first stage.

I gave up using them because I wanted full tubes replacement in a preamp, and I found a problem.

First stage? ... perfect.

Later stages?
Big (or small) trouble (pick one) :
a next stage grid, maybe the second but definitely the third , in a typical high gain preamp, will easily receive some 250Vpp .
Maybe you cut that doen with an attenuator to 100 or even 50Vpp .
A typical grid will take that all day long, no big deal, and to boot will rectify audio, shift bias, clip unsymmetrically, the works.

Try that with a FET gate

To begin with it will breakdown with 5 to 15 V peak and, even in the best case, "clip like a Zener" .
To boot, if I add a protective Zener so it carries the load, it gets even more SS sounding (i.e. buzzy) .
And it does not pump/compress like a tube.

Mind you, KMG, a distinguished SSGuitar Forum member, who's done an incredible job of turning a Fet+MOS stage into a very close match of a real triode ... but he's had to add a lot of tweaks (even a germanium diode) to make it bend and kink signals the same way.
In fact, he skipped the FET completely because of lack of consistency and straight "un" linearized a MosFetb to act like a HV triode:
LND150 tube emulator attempt.

the actual schematics:



fully protected (against what I warned):


unprotected (if it's not driven that hard):


then he proceeded to build and test side by side, a "glass" JCM800 preamp

and his MosFet version,

including some samples.
In a single word: amazing.
http://milas.spb.ru/~kmg/files/proje...et_vs_tube.zip

EDIT: to be fair, the thread was started by Steve Conner , but KMG took it WAY beyond what any of us would have expected.

[tedmich]

I believe the AMT Stonehead SS amp uses a variation of KMGs Jfet and doesn't sound half bad

[eschertron]

Cool
So you are using a cascode with a high voltage supply?
In that case yes, you will have tons of gain, definitely less hum (and maybe less hiss) than a regular triode stage and it will work well as a first stage.

Yes. I am happy with small victories at this point