FET Input Tube Stage Attempt

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

tk@halmi

Well-known member
Joined
Jun 3, 2004
Messages
999
Location
Oregon, USA
Hi,

Recently, I put my grubbies on a dozen NOS 6C4WA tubes for about a buck each. It is a single triode tube, like a half 12AU7.
First I thought about making a cascode, but I don't like the non-existent PSRR that comes with it and I wanted to make the gain easily adjustable.
So I came up with this, for now just a simulation, feel free to pee on it.

FET Input Tube Stage

It is set to 29dB with the 7.5k resistor, an 8.2k in the feedback network resistor would give a bit over 30dB gain. My intent is to slap it in front of New York Dave's line output stage for a lower noise input.

Tamas
 
In production with real (not simulated) FETs, R5 will have to be trimmed to keep the FET voltage from being kinda low or way too high. As shown, the 12AU7 grid will almost never be at the voltage the simulator says.

In general, the resistances seem small, especially R8 loading the ~6K of the 12AU7 plate. The neat trick with tubes is to work them at fairly high impedances so their thin vacuum isn't strained.

Altho, if this goes in front of a high-gain stage, its output level may never be high enough to matter.
 
[quote author="PRR"]In production with real (not simulated) FETs, R5 will have to be trimmed to keep the FET voltage from being kinda low or way too high.[/quote]
Yes, I was half expecting it to be a trim pot due to variations of FET parameters.


[quote author="PRR"]In general, the resistances seem small, especially R8 loading the ~6K of the 12AU7 plate. The neat trick with tubes is to work them at fairly high impedances so their thin vacuum isn't strained.
[/quote]
Hmm, I did not think of that. Am I a typical vacuum tube beginner or what? Let me see if rescaling resistor around the FET allows using a higher feedback network. That FET likes at least a mA current to work well.

Thank You! :grin:
 
I forgot to say that the 220uFd output cap seems excessive.

I suspect that, in the end, AC coupling between stages is more practical, though you will get into subsonic stability issues.

> see if rescaling resistor around the FET allows using a higher feedback network. That FET likes at least a mA current to work well.

Look at it this way. Assume the first stage runs at 1mA, and the second stage runs at 1mA. The feedback network AC current must be a small fraction of the second stage DC current, and the feedback network current should be much larger than the first stage AC current.

With BJTs, cascaded common-emitter, the gain available for feedback is only the Beta of the second transistor. While the 12AU7 appears to have infinite Beta, it is working with about 3K on its grid. Even though the grid takes no current, that 3K resistor must be fed. So you really don't have much current gain.

You could add a second FET as a LTP so you could use a hi-Z feedback network. This could also stabilize your DC operating point. Of course you double your FET costs and voltage noise, and reduce voltage gain. You can get back your noise by running each FET at 4X the current of the one FET, though this becomes a power-filtering issue.

> cascode, but I don't like the non-existent PSRR

What's the problem? You have 100uFd and 220uFd caps scattered around, surely you can clean-up any old DC voltage, certainly any voltage suitable for the NYD line stage. We did it with pentodes (no PSRR) with much smaller caps.
 
[quote author="PRR"]I forgot to say that the 220uFd output cap seems excessive.[/quote]
Mostly a leftover habit from the low impedance BJT world. I like to pick coupling caps large in simulations. It helps quickly track down what else causes roll off in the circuit.

[quote author="PRR"]I suspect that, in the end, AC coupling between stages is more practical, though you will get into subsonic stability issues. [/quote]
Now I remember you pointing this out in other discussions. I will change to AC coupling.

[quote author="PRR"] Assume the first stage runs at 1mA, and the second stage runs at 1mA. The feedback network AC current must be a small fraction of the second stage DC current, and the feedback network current should be much larger than the first stage AC current. [/quote]
That is an easy to follow guideline even for me. Thanks!

[quote author="PRR"]You could add a second FET as a LTP so you could use a hi-Z feedback network. [/quote]
I think I will stick with the single FET and play around with the circuit a bit more. Maybe increase the current to 10 to 12ma through the tube. As a mic input stage it might work OK.

The cascode I tried gives way to much gain. It would work well as a phono stage.

I might dork with a FET-Tube SRPP too. :green:

Thanks,
Tamas
 
That's where we're different. The first thing I'd think of is to use it as a common-cathode stage. It wouldn't even occur to me to stick transistors in there--but then again, I'm not big on hybrid designs.

You say you're using a line stage from me. Can you refresh my memory? I assume you're talking about one of my 12AV7/12BH7 concoctions.
 
> Maybe increase the current to 10 to 12ma through the tube.

That's one way to go. It does make power cleanliness more difficult. And the 12AU7 is near its linear limits at 10mA. It may be fine for small outputs.

If the next stage is an NYD open-grid amp input, throwing 10+mA at the problem is swatting gnats with a carpenter's hammer.

> As a mic input stage it might work OK.

I've been wondering what the use was. 30dB gain in front of another 30dB gain suggests a very small input signal. In "good" audio, there is a strong tradition of running weak signals balanced. Your amp is unbalanced. How are you going to connect a balanced mike to this gate?

In electrically quiet rooms, with true floating mikes, you can often ignore tradition, wire a dynamic (even a ribbon) unbalanced into the input. But not always.

For single-ended tube grids and FET gates, the most obvious way to couple is with a transformer. That not only does the bal-un conversion, it can also transform voltage and current.

Note that both tubes and FETs have near-zero nouse current. You can approach theoretical noise levels with any clean tube or FET, just by transforming the source noise voltage up above the tube/FET's noise voltage. In theory we can transform to infinite impedance and infinitely low noise figure; in practice tube input capacitance limits the useful transformation, and the transformer adds C and L of its own. Still, it is perfectly possible to get noise figure below 2dB with a 12AU7. It may be a little easier with an FET, because lower noise voltage and lower input capacitance, but we don't need no silly FET.

Or if you like the FET, it is perfectly possible to design a hot little 2-FET amp, and run it on two 9V batteries, eliminating some supply-buzz headaches and a LOT of supply cost.

Sure, you can design an FET/tube combo, but I wonder what the combo brings to the party that is better than some single-minded all-FET or all-Tube idea. We have a high-current FET in a place where its marginally lower noise can be lost in transformation. Its pentode-like output plus the large supply voltage and loosely specified bias needs lead to huge variation in DC operating point with different devices from the same bag, and probably with voltage and temperature. This high-current device feeds a tube with very-low input current, but not a lot of output current. The output current would be ample, except you need overall feedback and the handy place to stick it is a high-current pin on the FET.

If you put the tube first, after the transformer: say 1:4 transformer and 12AU7 with unbypassed cathode gives gain around 8, you have overall gain of 30. The FET might buffer the output, but the output impedance is 10K-15K already. If you are driving a 470K grid resistor, or even a 100K gain pot, I think the design is nearly done without reaching for the FET. You need to check maximum output voltage: peak is about 20% ofB+ or say 45Vpk, 30V RMS. If the line-amp stage is gain of 30 and slightly higher output swing, then it won't need more than 1V or 1.5V RMS to hit its limits, so our 30V RMS output is plenty, even with some gain-drop between amps. The 1:4 input transformer transforms 200Ω sources to 3,200Ω. The noise resistance of a triode is about 1/Gm, which for 12AU7 is around 250Ω, but this is a very hot resistor so triple it: 750Ω. This is enough less than the transformed 2,400Ω that we can hope for a good noise figure. In practice, we will have to select tubes without excess noise (dirt in the cathode coating).

There are many ways to skin cats. You really should try a lot of ways. Some will be obviously flawed. Others may be more subtle. Spoiler: You are not the first to pass this way. If you look at existing plans, you find a few popular ones. It isn't that other ways don't work, but that a few ways work better than most others ("better" is somewhat subjective).
 
I hear ya! Thanks, Paul. I am glad I posted this and didn't spend days wondering why it wouldn't work on the breadboard.

Did check out the FET-Tube SRPP also and it has a lot of oompf leaving no headroom for a decent range of input signals.

I think I will use these tubes as simple common cathode amplifiers as suggested by Dave.

Cheers,
Tamas
 
What are triodes good for?

Gain with low distortion even when run open loop, moderate bandwidth, high input resistance, swinging a lot of volts compared to most sand state. Also because of the long thermal time constants the shift in characteristics with signal drive tends to produce subsonic effects.

To get the low distortion you need a high external impedance at the plate and a low impedance at the cathode. So with a big choke load or current source and light loading afterwards you get a gain of -mu with very low distortion, and mostly low order distortion at that. If that is the amount of gain you need then you are in luck.

Depending on the transconductance the noise can be pretty low, too, although the flicker noise comes up and spoils things at low frequencies.

If you have a lot of gain then the Miller effect will raise the input capacitance and load down the source at high frequencies.

What are triodes not particularly good for?

Moving a lot of current for one. Efficiency overall is poor. And there are always microphonics at some level if there is mechanical vibration.

If you are looking for lowest input noise and using a FET to get it then you need to run the FET common-source and get some voltage gain at least, for that noise to dominate. If you follow this stage with a common-cathode stage with a gain of nearly -mu, your FET stage, even though more distortion-prone, will be operating at lower voltage swings given the gain that follows (otherwise you'll be overloading downstream eventually). Thus, you won't be getting that much distortion from the FET. So, this might be one optimal configuration, if fed then to a white or regular follower in either hollow or sand state, or a high-Z load.

But this may be more gain than you need.

There is also the possibility that if you run with some unbypassed cathode resistance, and less than a very high Z plate load, you can set up the lower gain structure so the second harmonic distortion of the FET is cancelled by the inverted second of the tube. Still no overall negative feedback, but potentially low second harmonic distortion. Some of the lower order distortion from the first stage will be doubled in order by the second stage, but that's going to happen in any system.

EDIT: Beware of simulator lies when looking at low distortion numbers using Fourier techniques. A reality check can be to contrive a cascaded notch filter or filters and look at distortion residuals.
 
Back
Top