Fun little P-P input stage

I got slightly distracted while working on my DD ESL amp, and have been playing around with this input stage; it's a variant of a design I've been using in amps for the past 20 years or so. Main difference is the incorporation of an input transformer (there actually is no center tap on the transformer, that's a mistake on my drawing). In any case, it's good for a gain of about 20, will swing about 150V p-p, and has an f₃ of about 80kHz. The balance is better than I have the capability to measure when the input resistors are well-matched.

Matching of the 2N5462s is a good idea; I used a simple test jig to match I_d with a 1K resistor between source and ground. With FETs thus matched plugged into the circuit, I got plate voltages within 5 volts of one another.

Variations: for feedback, one can add 100 ohm resistors in series with each input resistor to the ground point, then run symmetrical feedback resistors from the 16 and 0 ohm taps of the output transformer secondary, grounding the 4 ohm tap. And you can get more gain and a bit more swing by bypassing the 1K cathode resistors.

This would make a dandy input stage for an ST-70.

BTW, at the expense of esthetic symmetry and galvanic isolation, you can get rid of the transformer, take the input to one of the FET gates, and etiher ground the other one or use it as a feedback point.

And to get enough gain for feedback, you can also direct couple this stage to a 5692-based long tail diff amp- that's pretty much the circuit I've been using in most of my push-pull amps over the years.

It's the bastard child of Dave Berning and Bill Johnson. Cross-coupled circuits are very useful for situations where one may have to use both balanced and unbalanced inputs.

Pardon me for being dumb, but I don't get what you mean. Where would I put the current source? (I know you can draw! )

No, that's the FET's drain (plate), not the source connected to the minus rail; note that the FETs are configured as source followers, where you want the drains to be at signal ground.

Pjotr, no, I suspect it's reflex on Dave's part; you work with tubes too long and your brain always thinks of the minus rail going to cathodes. Easy mistake to make. I suspect he looked at the circuit and thought, "Cascode!" Which it isn't at all.

If we only had positronic tubes...

And all the voltage gain comes from the top devices, none from the FETs. It's really more like a cathode-coupled circuit than like a cascode.

See? I'm not allergic to FETs!

The trannie is optional- what it does is improve common-mode noise rejection, galvanic isolation, and balance. As I mentioned, the trasformer is something new for me; previously, my circuits used one gate for an input and grounded the other- or used it as a feedback point. Doing it that way, the imbalance depends on mu (more mu, more balance). And, of course, you can use both the gates for inputs when driving with a balanced source.

I'm not sure what you're thinking about doing with the drains- feeding each from separate supply rails? If so, that's not necessary, and in fact, a little supply impedance isn't a bad thing there.

Sure. It's exactly a cross-coupled circuit. The twist is, as you observe, that instead of driving with a cathode follower, I'm driving with a source follower, direct coupled.

By cross-coupling the tubes in this manner, as with the classic cross-coupled circuit, you can remove imbalances in the drive. The FET source followers appear as degenerative elements, with curvature opposite that of the tubes -> distortion reduction. And with just a couple of small caps, you can pop the HF response up another octave. If you bypass the cathode resistors, the gain in my circuit will be higher, too- this is an option not available in the all-tube (or tube and n-FET) circuit. Another practical issue I've had with all-tube cross-coupled circuits is that you limit the swing across the second tube because of the need to properly bias up the cathode follower to properly drive it (the classic circuit as you've drawn it has VERY limited swing because of that first stage; normally, the CF is biased up to a higher voltage).

Indeed, you can use triodes in cascode as a diff amp and get excellent performance. That circuit goes back to Hedges (at least), and is the basis for a lot of designs by Curcio and Sonic Frontiers. But the balance is better in my circuit, and there's one less tube to worry about. And because it's push-pull from front to back, it's easy to incorporate symmetrical feedback and the little HF-extending caps.