Current vs. Gain vs. "correct" harmonics in FETs - FETzer vs. Omega

[midwayfair]

I was breadboarding something that ended up very similar to the Omega's transistor setup as the first two stages, the only change being that the drain is trimmed to half supply (runs on 24V from an LT1054) and the gate goes right to a BJT base (instead of a MOSFET -- this was how they did it in the Britannia). I'm using a lower gain FET and it's 100% clean. Since I'm using the voltage tripler, I'm in a situation where the current draw of a stock FETzer valve is a concern with lower gain FETs.

http://www.runoffgroove.com/fetzervalve.html

In the FETzer valve, the source resistor is chosen specific to the FET for just the right harmonics WITHOUT a source bypass cap.

http://runoffgroove.com/omega.png

In the Omega, the source resistor is trimmed and bypassed by the cap that was dropped from the revised FETzer valve.

Since the resistors are larger in the Omega, two things happen:

1) High output impedance -- needs a buffer.
2) Lower current draw.

And a third thing ... might happen?
3) The value of the source resistor is now only important for bias and not for the harmonics.

One thing I ran into is that if I used FETzer valve resistor values based on the calculator, it pulled down my TL1054's output by a few volts. If I use the Omega's values, my 24V is fine.

Is it possible to get the best of both worlds by use the calculator to figure out the source and drain values, use the source resistor between the source and the source trim/bypass cap, and then use the much larger drain value(s) from the Omega (instead of the one from the FETzer calculator)? Gain goes down a little bit, but I've got more than I need already anyway. Output impedance goes up, but I've got the buffer so the gain loss is trivial.

My ears aren't a scope, so far I can't really tell a difference in the harmonics when I play hard, and both sound just about as clean and full range (it's hard to tell how clean because it's overdriving my preamp tubes a little).

[jubal81]

I have to admit, I can barely tell any difference with the ROG-recommended setup for 'tube-like' harmonics. I've been using 24-25V lately and it seems to make it even less noticeable. I just adjust the Source and Drain resistance ratio for the gain I'm looking for and bias the gate. Sounds great and I can get more than enough headroom, even using high-transconductance FETs. 

I do agree with ditching all source bypass caps. I find they add too much noise. It's a cheap and easy gain boost, but it's just not worth it to me.

Total drain+source resistance set max power consumption, so it makes sense that the lower Fetzer values would draw more Ma.

I guess what I'm trying to say is: I don't Fetzer, and I'm OK with that.

[midwayfair]

I have to admit, I can barely tell any difference with the ROG-recommended setup for 'tube-like' harmonics. I've been using 24-25V lately and it seems to make it even less noticeable. I just adjust the Source and Drain resistance ratio for the gain I'm looking for and bias the gate. Sounds great and I can get more than enough headroom, even using high-transconductance FETs. 

I do agree with ditching all source bypass caps. I find they add too much noise. It's a cheap and easy gain boost, but it's just not worth it to me.

Total drain+source resistance set max power consumption, so it makes sense that the lower Fetzer values would draw more Ma.

I guess what I'm trying to say is: I don't Fetzer, and I'm OK with that.

When you say bias the gate, you just mean elevating it above ground? What are you setting it to?

I never really see anything intentionally bias the FET gate, so an example would be great.

[LightSoundGeometry]

I have to admit, I can barely tell any difference with the ROG-recommended setup for 'tube-like' harmonics. I've been using 24-25V lately and it seems to make it even less noticeable. I just adjust the Source and Drain resistance ratio for the gain I'm looking for and bias the gate. Sounds great and I can get more than enough headroom, even using high-transconductance FETs. 

I do agree with ditching all source bypass caps. I find they add too much noise. It's a cheap and easy gain boost, but it's just not worth it to me.

Total drain+source resistance set max power consumption, so it makes sense that the lower Fetzer values would draw more Ma.

I guess what I'm trying to say is: I don't Fetzer, and I'm OK with that.

When you say bias the gate, you just mean elevating it above ground? What are you setting it to?

I never really see anything intentionally bias the FET gate, so an example would be great.

haha, I have watched a lot of your videos on the tubes for sound demos etc  ..I didnt know who you were until I clicked your link but recognized the videos and page instantly.

[jubal81]

I have to admit, I can barely tell any difference with the ROG-recommended setup for 'tube-like' harmonics. I've been using 24-25V lately and it seems to make it even less noticeable. I just adjust the Source and Drain resistance ratio for the gain I'm looking for and bias the gate. Sounds great and I can get more than enough headroom, even using high-transconductance FETs. 

I do agree with ditching all source bypass caps. I find they add too much noise. It's a cheap and easy gain boost, but it's just not worth it to me.

Total drain+source resistance set max power consumption, so it makes sense that the lower Fetzer values would draw more Ma.

I guess what I'm trying to say is: I don't Fetzer, and I'm OK with that.

When you say bias the gate, you just mean elevating it above ground? What are you setting it to?

I never really see anything intentionally bias the FET gate, so an example would be great.

Here's the article I go by to get the basic recipe:
http://www.rason.org/Projects/jfetamp/jfetamp.htm

The basic idea is to look at the current handling characteristic of the specific FET you're using, then take the supply voltage and calculate a total resistance to meet the max current you want in the FET (say 70% of IDSS). This total resistance is Rs+Rd. Then you decide the ratio, which influences the gain you're going to get from the stage. They suggest for the Source resistor to be 25% of Rs+Rd, so if Rs+Rd is 10k, the source would be 2.5k and the drain would be 7.5K.

So, with max current and gain profile set, add a small positive voltage to the gate to get the bias voltage you're looking for on the drain. I use the RG noiseless biasing, with a cap-bypassed voltage divider with a 1M or more resistor to the gate. This is only a couple volts, so you don't need to take any Ma from your charge pump. Soulsonic posted about this on another forum showing how he uses 2sk880s.

I've been using SMD JFETs, and with this setup, I've been able to plug in any of the same type and have the drain voltage only vary about 1.5V, which is pretty small when you look at it compared to the supply voltage. At 25V VCC, I've got drain voltages between 13V and 14.5V plugging in different FETs.

One thing I've noticed - and this could be related to the ROG situation (where they run it hot) - comes from using different IDSS variants of the same FET (BL vs. GR). The biasing works just fine regardless, but the lower IDSS does seem to sound a little brighter/more interesting, which leads me to believe that stressing the FET a bit more does add a little extra spice to the sound without obvious distortion.

[midwayfair]

Okay, went off to do some reading.

http://www.linearsystems.com/assets/media/application_notes/AN102FETbiasing.pdf

Figure 6 is described as all three circuits being equivalent.

I'm guessing that the main benefit to C is that the gate can be biased to, say, half a volt without resorting to something like a 22M positive bias resistor. But 22M aren't exactly unobtainium. I have some, in fact. But obviously a trimpot is right out.

So that leads to the gate divider.

What I'm seeing, though, is that using the recipe in the source you provided is that my gain will be about 4x, regardless of the type of FET.

If I run over to the FETzer valve article, I can grab these:

Out of twenty-four (24) MPF102
average Vp: -2.34V (spanning from -1.76 to -3.11)
average Idss: 5.65mA (spanning 3.91mA to 7.15mA)
average Rs: 344 ohms
average Rd: 688 ohms
average Gain: 1 times or 0 dB

The Rason.org's gain is 4x, and the Rd is twice as big, comparatively. I'm guessing in their case they got closer to half supply, whereas the FETzer is made to run on 2/3 supply.

I'm using 2SK30.

And ... hrm.

Out of ten (10) 2SK30AY
average Vp: -1.41V (spanning from -1.26 to -1.52)
average Idss: 2.14mA (spanning 1.78mA to 2.50mA)
average Rs: 505 ohms
average Rd: 2602 ohms
average Gain: 2.4 times or 7.6 dB

Rs is slightly less than 20% of Rd. But gain is only 2.4x! Not 5x. And the drain here would be only at 6V. So is the rest of that gain to be gotten by biasing the gate to 4.5V? Or is something else happening?

I saw a post a while ago where RG or PRR mentioned that the actual gain isn't really Rd/Rs. So maybe that's what's going on here?

Those 2SK30 values are given for a 9V supply. If I plug them in with 24V, I get 8.1x, with 8.9K for the drain and 547R for the source. Not too far off from my breadboard guess of 10K with a 2K source trim.

2SK30 can run on 50V, though. I have a 48V supply on order. It's overkill, but let's see what happens there ...

Okay, 17.1x gain, and 25dB. And the source resistor barely changed at all, and the drain is 19K. If I plug that into a calculator, it says that the gain "should" be 34x.

I'm seeing a pattern here. Gain is actually HALF of Rd/Rs, at least when the drain is biased to 6V, using the FETzer method. The FETzer method also leaves the source resistor untouched, whereas Jason's link would increase the source resistor based on the supply voltage.

So ... anyone knowledgeable enough to tell me if I can get the other half of that gain by biasing the gate to drop the drain voltage to half supply? Or will that damage the FET because the source resistor MUST go up to avoid exceeding the JFET's gate-source voltage limits?

Jason, that's a typical Vr for you when you do the voltage divider noiseless biasing?

[PRR]

> Gain is actually HALF of Rd/Rs

I have not followed your full essay.

Rd/Rs would be correct if the FET had INfinite transconductance (Gm) and infinite internal drain impedance, and infinite load impedance.

No device has infinite Gm.

An unprovable assumption, good for first-pencil design (and expressed in tube terms), is that your cathode bias resistor will be very similar to 1/Gm (at the actual operating point).

Book-value of Gm for 12AX7 is 1,600uMho. 1/Gm is then 625 Ohms. General designers would use 680 Ohm cathode resistor. If un-bypassed, the total (external+internal) resistance is 1,305 Ohms, roughly twice the tube's own 625 Ohms, and un-bypassed Voltage Gain is about half of what you first thought.

(FWIW: being gitar-guys we all know to use 1.5K under a 12AX7. This was Leo Fender's preference, taken from an old GE amp-table. Always bypassed. This works fine too.)

FETs will work about the same; just change the pin-names and the likely values of Gm.

Another way to look at it: tubes have less curvature than JFETs. The Fetzer et al technique "swamps down" the JFET's excess Gm and curvature to a more tube-like level.

Yet another peek: there is an absolute limit of Gm, the voltage needed to control a current. BJTs hit this limit over their practical operating range, because the control charge is intimately mixed with the current controlled. FETs do a little worse because their control charge is only "near" the current. Tubes do considerably worse (for all practical currents) because their control charge is on wide-spaced grid wires.

> if I can get the other half of that gain by...

You get it by using ALL the Gm of the FET and do not mind that it is more-curved than a tube.

If you want tube-like, you cut-down the Gm for a somewhat better approximation of tube action.

Why are you whining about gain? Straight clean gain is cheap. You can always add a stage and get as much gain as you can stand. The whole point of Fetzer etc is to get a *bent* curvature which appears to be more-similar to tube curvature. Flavor, not quantity.

[midwayfair]

Okay, did some experimentation and answered at least a couple questions.

I used a 1K source resistor most of the time to keep the ratios simple.

1) With 24V to 30V, elevating the gate about 1V will put a 2SK30, 2N5457, or 2SK170 at roughly half supply with a 10:1 ratio of gate and source resistors. Closer to ground easily hits 2/3 supply. That's 10x gain, at least hypothetically. Probably not in real life. I can stretch it to a 20:1 ratio by upping the drain resistor, and then with the gate grounded I can get just a smidge above half supply, but nowhere near 2/3 supply. Still, the FETzer valve calculations hold up in this situation, with minor minor changes in the gate being good enough. So that's useful because I can keep the sound I like.

2) If I bypass part or all of the source resistance, I still get quite a bit more gain. So clearly the FET is capable of producing MORE gain. If I drop the drain voltage, it will produce that gain without the bypass cap ... and eventually sound like crap and stop working. (To be expected.) It sounded best to me with the source partially bypassed. This presents a bit of an issue because the gain is no longer predictable and goes right back to being device specific! But my current draw is more predictable because I can use larger resistors. (Though the value of the source resistor may be more important than the drain voltage for getting the best mix of harmonics ... I think Transmogrifx mentioned that.

3) One of my 48V power supplies arrived today. I plugged it in with some 2SK30s and with the extra voltage, I can get a bigger ratio, up to 33:1 (maybe a little higher depending on the FET, but I didn't try every 2SK30 I had), with 30:1 only requiring a few hundred millivolts of elevation, and it's VERY touchy (multiturn trim to the rescue!). That's awfully close to the ratio the FETzer valve quoted me, except that was putting the drain a little higher. Maybe it has something to do with the smaller source resistor, or maybe I'm missing something else. 48V is overkill, obviously. It diminishing returns above 30V, and there aren't any other FETs that can take voltage that high anyway. (2SK170 is 40V max, but higher gain and clips easier.)

4) I noticed that the drain voltage took a very long time to settle while I was dialing it in at the gate. It might be my multimeter.

5) The only benefit I'm really seeing to biasing the gate in a FET circuit at the moment is that my gain is, indeed, fairly predictable. If I add the source bypass cap, the gain actually becomes LESS predictable and it's back to being device-specific. Gain doesn't seem to vary hugely among different FETs of the same part type running on voltage this high. But if I wanted predictable gain, I'd use an op amp (ideally), or a BJT. (Worth noting that the gate bias lets me use a BJT on the same layout. I've done the opposite in other circuits.)

The drawback to this biasing is basically just the addition of the filtering cap. If I need full gain, I can apparently just bypass the source caps to get it all back: The 4:1 resistor ratio is also pretty close to what RG suggested in his Fooling With Fets article. http://www.geofex.com/article_folders/foolwfets/foolwfets.htm That ratio would normally suggest 4x gain, but it's actually 17.5 with the FET he used.

6) PRR replied while I was typing this ... I get that the FETzer is meant to do one thing in particular. I was curious in part whether there was a way for me to get the harmonics of the FETzer while getting the low current draw (and parhaps some of the gain) of the Omega. The current draw is the main thing, but Jason's note was a useful tangent.

7) I think the place I'm at right now is: If I bypass all but the FETzer's recommended source value for the FET I'm using (560R for the 2SK30), then use the gate bias to dial in the drain (to 6V), will I end up with a similar harmonic profile as a FETzer? In my head this would would allow me to use more resistance overall (dropping the current draw), maintain the 2/3 drain bias point (for color), but fool the transistor into thinking its source resistance was much smaller (for the right amount of gain for the FETzer). I lose the predictable gain, but if I need that on the same layout, I just ditch the source cap and use the appropriate resistors.

Does that sound accurate? I don't really have a way to test this except with my ears.

[PRR]

> I don't really have a way to test this except with my ears.

I believe this path will confuse you no end, and any "result" will be blind luck.

The "right" path would be signal generator, Audio voltmeter, and 'scope (and probably FFT display), and a large notebook (or spreadsheet database). That might simply take a long time to get a good notion.

There is NOT a JFET "equivalent" to a vacuum triode because the triode has internal voltage feedback plate to grid. Any simple way to include this significant effect becomes non-simple very quickly.

But tubes are not "ideal guitar tone" devices, they are just very-very familiar and tend to avoid the worst tones of hotter devices (FETs, BJTs, amplified diodes).

[Gus]

Something to read
http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat.pdf

Note BJTs with emitter feedback compare to JFETs and tubes. 
I often like to use BJTs more than fets.

An example I posted in the past. Move the jumper to the cap for a bootstrapped input, note the collector and emitter resistors on the first stage. A SPDT can make this a LPB like input resistance to a bootstrapped input resistance.