How would I compare a FET to a tube spec wise?

[EricKnabe]

I am trying to build a 2203 preamp with FETs in place of the tubes and I don't know what type of transistor to use. The original circuit, as I'm sure you're familiar with, uses 12ax7 tubes. I'm trying to modify the circuit to use transistors (probably FETs.) but I have no idea how to find the best equivalent as far as characteristics and stats. It's hard to find info comparing FETs to tubes because they typically aren't used in this manner. What I do know is that the voltage gain of a 12ax7 is like 30v and the amplification factor is 100. I have no idea how to calculate the voltage gain or amp factor for a transistor and even if I did, I don't even know if that's even the right number to use for comparison. So I'm at a loss. If any of you could provide some insight into choosing the right parts for my project, please let me know.

[dschwartz]

I know Teemu can make a very elaborate and amazing reply.
But, all i can tell you is that you're expecting the exact same behavior from jfets and tubes, you will be dissaponted.
The only advantage of replacing tubes with Jfets, is that you can keep the same topology and just swap them, lower the V+ to something tolerable by the fets, and tweak the drain resistor to find the sweetspot from each jfet.
With that said, you'll get different gain level and a lot of inconsistency between jfets. We can tell you " use X jfet" but the gains will vary too much on that model, and it's hard to predict the behavior.
Check runoffgroove.com articles. Their marshall project (with opamps) is outstanding and captures the tone of a jcm800 really well, and consistently.

[EricKnabe]

Yeah, I wanted to do this project for a few reasons. I'm aware they won't sound exactly the same, which to me is part of the charm of doing it this way. I was kind of inspired by runoffgroove.com to do it, actually. The other reason is that I figured by going solid state I would save myself the headache of tube maintainence, but choosing the right transistor for replacement seems to be a headache of it's own... Seriously debating going back to just using tubes. Still, I would like to do it for the sake of the project's relative uniqueness.

[dschwartz]

Well, its a great way to learn.
Get a few j201, 2n5457, mpf102, etc.. and socket them. Use your ears to find which sound best

[teemuk]

First off, there is no such thing as "voltage gain of tube" or "voltage gain of FET". Voltage gain is a circuit characteristic; it might be 60x in some circuit, but could just as well be 10x or perhaps 5x in another. Or even less. A circuit like cathode follower introduces no voltage gain at all in typical sense. Its output signal is always attenuated in reference to input signal.

What these devices have is so-called "transconductance", which – factored in to certain kinds of circuits – will result into some specific voltage gain. E.g. 12AX7 with its typical 100K plate load resistance, and 1K5 cathode bias resistance, will provide a voltage gain of approximately 60x. That with cathode capacitively bypassed. Without bypass the degenerative cathode feedback will reduce voltage gain. On the other hand, a "cold cathode clipper" of SLO with its 100K plate load and 39K (!!!) cathode bias resistance will provide a voltage gain of only about 6x.
You see, there is no specific value. It will always depend on circuit you fit the device in.
You can use the transconductance figure to estimate things like voltage or current gain using calculus, though. Do note that the quoted transconductance of a tube is merely a nominal value, even more so with FETs that have tons of variance.

I suggest you take some time and educate yourself about inner workings of tube circuits and FET circuits. The theory how they operate really should be 101 to anyone who wants to start converting tube circuits to FETs. Seems like way too often it isn't.

I don't know where you got the idea that FET is a replacement for a tube. FET characteristic curves are closer to pentodes instead of typical "preamp tubes", triodes. The impedance levels are overall very different, except for the "gate" circuit, which is high impedance similarly to vacuum tube grid. You might have hard time finding FETs that work with high voltages of tube circuits. (LND150?) Conversion to lower power supply voltages will always require that you rethink drain loading and source bias points, gain factors, clipping points and practically EVERYTHING!

The typical FET + 100K drain trimmer –configuration to mimic tube circuits will ensure you achieve performance that is totally unlike the circuit you are mimicking. And not only that, it ensures you will never built a consistently operating circuit, given wide variance of FET parameters.

I'm starting to think this is probably last time I'm going to post about this topic. I don't see the point of wasting my time further to those who are simply too lazy to search. This topic has been discussed previously over and over again and about every thread points out the very same issues why FETs are not equal to vacuum tubes and why there isn't some magic formula for tube-FET-circuit conversion.

[R.G.]

Good point Teemuk. Searching seems to be getting less practiced even as it has become fantastically easier.

I used to wonder at how the same issues came up over and over here and in other forums. Now I simply accept it. Part of the cause is the fact that there are periodic turnovers in new people, usually around the start and end of school terms, and (from my perspective) a natural progression of new people passing through a gotta-build-guitar-pedals stage. I suspect that surprising number of the forum questions here are from folks who don't remember a time before there was an internet. It's frustrating, but you gotta nurture the growth of the beginners. I suspect that there is a "teachers' fatigue" syndrome that real school teachers get that is similar, because they regularly as part of their profession have new batches of students asking the same questions again.

@OP: what teemuk said. FETs aren't tubes, don't sound like tubes except in on-line claims, and require individual FET tuning to make up for the variations in part to part. The reason most people use JFET replacements is that they self bias similarly to tubes, and the circuits shape the distorted frequency response to be similar to the filtering that the original circuits. It occurs to me that humans do a lot of auditory processing of vocal tones, and we might be confusing tone "voicing" with similar or dissimilar distortions. It's all tightly wound up in human psychoacoustics. In general, JFETs have vastly more transconductance than tubes; the circuit gain without feedback will approach gm (the transconductance) times the load resistor.

In my mind it gets down to a question of objectives. You told dshwartz that you know it won't sound the same, but found that charming; that you wanted to avoid tube maintenance; and that you liked the uniqueness. You are right - picking a JFET to work as inteneded is very difficult, to the point that industry has just about quit using JFETs because of it. But your stated objectives don't seem to require JFETs. Why not use some other amplification entirely? JFETs appear simple, but they're a headache, as you note.

As dschwartz pointed out, this is a great learning opportunity. You will get exposed to how transconductance and voltage gain are related, some few things about feedback, the variation in JFETs, how to bias and choose JFETs, probably more than you want to learn about the variations and distributions of parameters of JFETs, as well as general biasing and circuit theory. So, go for it. But do a lot of searching. You would not believe how much easier searching is now than it was.

[dschwartz]

I had my jfet phase 10 years ago, and the most important lesson was..
Don't use jfets...they are hard to get, hard to make consistent projects, hard to bias, and they sound good, but nothing a good opamp+diodes or even BJT design can't beat.

[amptramp]

Even if you managed to get a FET to bias properly to give you the same DC response as a tube pentode (it could never look like a triode), you have the dynamic problem that tube interelectrode capacitances are on the order of a few pF whereas FET capacitances are orders of magnitude higher.  Higher voltage or current ratings make for higher capacitances and in a semiconductor (FET or BJT), these capacitances vary dramatically with the operating point whereas tube capacitances are more constant and much smaller.  You may find the frequency response changes at different points in the signal waveform with semiconductors.  You may have to use cascode stages to avoid the drain-to-gate capacitance issues but then you are making a wholesale change to the topology.

[J0K3RX]

Yeah, I wanted to do this project for a few reasons. I'm aware they won't sound exactly the same, which to me is part of the charm of doing it this way. I was kind of inspired by runoffgroove.com to do it, actually. The other reason is that I figured by going solid state I would save myself the headache of tube maintainence, but choosing the right transistor for replacement seems to be a headache of it's own... Seriously debating going back to just using tubes. Still, I would like to do it for the sake of the project's relative uniqueness.

This wouldn't be much of a project but, if you find preamp tube maintenance a headache these might be an option for you.
https://amtelectronics.com/new/amt-12ax7ws/

Preamp tubes generally last me quite a while and don't require much maintenance. Power tubes on the other hand can be quite a pain in the ass not to mention a drain on your bank account...

If you're determined to have a go at it these links might help inspire you or discourage you. Either way, KMG is about as good as it gets in my opinion.

http://milas.spb.ru/~kmg/slpfet_en.html
http://milas.spb.ru/~kmg/jcm800fet_en.html
http://milas.spb.ru/~kmg/jcm800fetLnd150_en.html

[PRR]

> FET capacitances are orders of magnitude higher.

Conversely, since there are no 300V JFETs on the market today, for similar current the JFET external resistances will be an OoM smaller than tube circuit. (For same power, to drive similar loads equally well, even lower resistances.)

Tube often works 300V supply, 1mA, 100K load. JFET may work 24V supply, 1mA, 10K-15K load; for same Power we'd use a high Idss device at 12mA and near 1K load (this is almost never done).

We usually "can" make stray C moot. Fender did. Other designers used tubes at 270K even 470K loading and these do show fall-off even within the guitar's limited range, which might be variable with signal. (I don't think it is a major effect.)

Yes FETs are NOT Triodes and not really Pentodes. However the first Yamaha electric (not digital!) piano, and the first Roland Jazz Chorus, used cascaded JFET stages which to my ear have an "organic" sound not found in other transistor circuits. The JFET Jazz Chorus preamp is easy to steal.





Yes, that's a Fender tonestack in the CP-80 electric piano.

[EricKnabe]

Well, thanks to all who replied. I actually read some of Teemu's book on SS amp design, but as you've all probably inferred, I'm very new at this. What I've somewhat gathered from your advice is that maybe I should simply build the amp using tubes rather than trying to turn it into something it isn't. However, I'll keep reading some of the resources you guys linked me.

[teemuk]

^ Good points above.

Yes, a FET is not exactly a pentode (even less a triode), and likewise a bipolar junction transistor isn't a FET (or any vacuum tube device).

I think "tubeness" is not the right term but probably indicates better what I try to imply when saying that certain device architectures can - to some extent - be somewhat "analogous" to each other. A FET common source amp operates somewhat similarly to a tube common cathode amp, and likewise even a BJT common emitter amp operates somewhat like the two aforementioned. At least there's more similarity with the two aforementioned than with a BJT emitter follower amp.

Try this just for kicks: Build a basic BJT common emitter gain stage. Basic voltage divider bias at base circuit, resistors at emitter and collector. No additional feedback, no cascaded gain stages with global feedback and all of that stuff how transistor stages are typically built today. Just the "schoolbook" example circuit of a BJT common emitter amp. At low input signal levels one can extract pretty "nice" VERY asymmetrically and softly clipped output from one. None of that "hard clipping" BS folks try to associate to transistors. In comparison to a typical distorting triode common cathode stage (with sane bias values) that one will, however, introduce much more asymmetric distortion and much "softer" gain compression characteristics. FET common source amp is sort of an intermediate of those two (BJT vs. triode circuit).

Of course, BJT-based circuits are hardly ever built like that today. Such circuits will have more issues with overall device beta variance, temperature coefficiencies and all that sort of "inconsistency" stuff. Not to mention, that soft clipping behaviour requires VERY small input signal range, something in the order of only few millivolts. THAT is very unpractical considering overall "gain staging" and while input can always be attenuated the downside of that practice is that signal-to-noise ratio quickly goes through the roof. Plus, the distortion figures are pretty appalling, which isn't too good considering that 99,9% of applications for "amps" (as in circuit that amplifies) is something other than effect processing and generating "musical distortion".
But yes, one could build a nice distortion effect that way. Some people might even regard it as "tube sounding". 

[Jeema]

Putting aside all the arguments about FET distortion vs. tube distortion... and the fact that it's going to sound different anyway due to the lower headroom... and the giant pain in the ass that JFETs are... and assuming you don't want to go through and individually select a bunch of parts... let me try to answer the original question anyway.

First, the 2203 preamp has 3 common cathode gain stages (followed by a cathode follower before the tone stack) with the following AC voltage gain by my calculations:

Stage 1: 61.5x
Stage 2: 8.5x
Stage 3: 41x

So you would probably want to try to approximate that, but with FET common source stages.

The voltage gain of a common source stage is given by the following equation according to Wikipedia:

Vgain = (gm * Rd) / (1 + gm * Rs)

As you can see, if you're using a trimmer on either the Rd or Rs side to try to bias a JFET, you're going to also alter the gain.

If you bypass Rs with a capacitor, though, then the equation simplifies to:

Vgain = gm * Rd

Now you can put a bias trimmer on the Rs side and not affect the AC gain of the stage. Then it just becomes a matter of choosing the right Rd resistor to achieve a similar gain as the original circuit.

However this approach means that the gain is now completely dependent on the FET's transconductance, which can vary quite a bit.

But you can at least guarantee X amount of gain with this approach and not have to pick through parts to find one that will bias well using fixed resistors...

P.S. Here's the page I used to try to calculate the voltage gain of the tube stages: http://www.rru.com/~meo/Guitar/Tubes/t3.pl

[PRR]

> it's going to sound different anyway due to the lower headroom... Stage 1: 61.5x... .....

Well, gain and voltage headroom are related. If you drop supply from 300V to 30V, you want 1/10th the gain for the same input sensitivity before overload.

FWIW: that "tube gain calculator" gives high answers because it does not account for loading.

[EricKnabe]

This is all super informative. I've learned a lot. Also, I said FET in my original post, not necessarily JFETs. Would MOSFETS be more consistent? Could I make the circuit work with a high voltage MOSFET and make little to no changes to the rest of the circuit? If not, should I bother learning how to build a similar solid state amplifier to the 2203, or should I just cut my losses and build the circuit with tubes as it was originally intended to be?

[teemuk]

Yes, indeed the gain ratios are designed for that specific circuit with its specific headroom limits introduced largely by the supply voltages and grid bias levels. They would have to be scaled down with circuits that are operated with lower power supply voltages and consequently lower thresholds to clipping overall. However, note that in the case of FETs, for instance, supply voltages and bias voltages (both important as clipping thresholds) will not "scale" in similar fashion. A tube circuit may have 300VDC B+ and 1.5VDC cathode bias voltage (which is also approximately the grid clipping threshold). A scaled down FET circuit may have 30V supply voltage but may still require 3VDC source bias voltage.

Not only that, if you want to nail operating characteristics similar to circuit you mimick you also have consider overall symmetry or asymmetry of that clipping. In the aforementioned Marshall 2203 for instance...

First gain stage: Very prominent asymmetry -if clipping- BUT typically receives signals of very low magnitude so doesn't clip under typical operating conditions. Grid will start conducting and clips the signal around 2V (peak) input (note that it is asymmetric process), Plate will start to clip with approximately 1.3V (peak) input signals (again an asymmetric process). Plate clipping will be slightly harder than grid clipping. Note that typical input signal levels of this stage are in the ballpark of 10 - 100 millivolts.

Second gain stage: Highly asymmetric clipping. Grid begins to clip with approximately 20V (peak input) but you only need approximately 3.5V (peak) input to produce clipping at the plate.

Third gain stage that is effectively a cascade of common cathode "input" and cathode follower: This cascades three clipping mechanism. Grid clipping of the input tube, plate clipping of the input tube and cathode follower clipping. Cathode follower clipping has extremely "hard" characteristics compared to all others. Plate clipping is also "hard" but not as hard as clipping of the cathode follower.
Cathode follower clipping is caused by capacitive coupling and does not have the typical "flat top" characteristic of clipping. (It has a "rising" characteristics, difficult to replicate with usual tube emulation circuits).
All in all, the stage has moderately "symmetric" clipping characteristics but there is some asymmetry present. Note that the driving signal of this stage (if distorted) will be highly asymmetric under typical circumstances as well.
Cathode follower stage will start (asymmetric and hard) clipping at approximately 1.5V (peak) input. Grid clipping of the input begins simultaneously as the overdriven cathode follower stage starts to "load" the input stage. Plate clipping will begin at approximately 2V (peak) input. As you may note, this difference (1.5V at one lobe vs. 2V at the other) results to considerably less asymmetry than in the case of stage #2 (20V vs. 3.5V).

The generic "drain trimmer" bias of FETs usually suggest to bias to a drain voltage that is approximately half of the supply voltage. Biased this way, the FET stage will introduce very little overall asymmetry and it is evident that it doesn't really perform like tube gain stages that are, for preferred effect, very often biased for deliberate asymmetry.
Not only that, the drain trimmer will alter stage gain (as noted) and its output impedance, which in turn will alter corner frequencies of subsequent RC filters. Source biasing with trimmer is not technically any more consistent. Yes, you can minimize gain variation of AC signals by capacitive bypassing but the trimmer resistance will affect ALL DC bias conditions of the circuit (e.g. idle drain voltage) and therefore clipping characteristics.
You perhaps start to understand now why the most recent "Runoffgroove" project abandoned using FETs and began to exploit much more predictable operational amplifiers and clipping diodes with predictable thresholds...

As for the question should you use tubes instead I can't really provide any other answer than that if you look for the "easiest" method to build a circuit that achieves what some kind of tube circuit does, then most certainly the easiest method is to build that specific tube circuit as is and using tubes. Any emulation attempt will introduce fair bit of difficulty to such project as you really have to –design- things by yourself. I leave it up to you to judge the difficulty of that approach.

[EricKnabe]

If you drop supply from 300V to 30V, you want 1/10th the gain for the same input sensitivity before overload.

I'm reviving this thread because I need help with this very thing. I restarted this project, after learning A LOT (of things I probably should've known before. But hey, electronics is daunting, I didn't know where to start.) but anyway, I'm using enhancement MOSFETS.
With a Vgs(th) of roughly two, they'll get about the same gain in the same circuit as per the common source Av equation as the 12ax7 does using the tube gain equation.

In any case, my problem starts when it comes to dropping the gain by 1/10th. It'a too clean.
If it was say, crunchy at 250 volts, then it will be edge of breakup at 25 volts with one tenth the gain. I've tried almost everything at this point to figure out some sort of formula or ratio for how much gain I have to add back for it to clip the same. I just don't understand why this is at all.

Now, I am using Falstad's circuit simulator. Maybe it could be chalked up simply to that fact alone? I'm not sure, but I'm at a loss at this point.

[PRR]

> It'a too clean.

ANY circuit will go from clean to dirty to hammered. Just raise the input level.

[EricKnabe]

I understand that but I want the amount of clipping I'm each stage I design to be as exact to the original amp as possible.

[antonis]

I understand that but I want the amount of clipping I'm each stage I design to be as exact to the original amp as possible.

What about making your design as close as possible to the original..??