Hybird Headphone Amp: Help with Valves!

[b1o2r3i4s5]

Hi, I am designing a Hybird Headphone Amp with Triode input stage and BJT Output stage. I am experienced in designing for BJT's but completely new to valves, I learned about Anode followers and Cathode followers but don't really know how to put them together. Can anyone give me some pointers?

I am planning to use a Anode follower with Vgain of 60 as input stage then a Cathode follower as current driver and finally a BJT class-AB output stage which will drive a 22:1 transformer before outputting to the headphones.
I currently have a +/-80V (160V) supply, 2x 12AT7 dual Triode and planning to use MJE15034(5) BJT pair as output stage. The output transformer I have is Oxford electrical B18A001F.

[sgrossklass]

I'm not exactly a hollow state guy myself, but here's my 2 cents:

You don't need to relearn everything when dealing with tubes/valves. A cathode follower behaves much the same as a npn emitter follower, and a common cathode circuit compares to a common emitter circuit. Of course you equations will be different, you need the Barkhausen eq'n etc. With the commonly high impedances involved, capacitive coupling is everywhere, too, not to mention microphonics, so shielding and sturdy mechanics seem like a good idea.

You want to do a gain of 60 (36 dB) in one stage? Would you do that with BJTs? Probably not. What you need is an input transformer, which also helps getting noise figure down. Unfortunately there doesn't seem to be a lot of literature around that deals with noise densities in vacuum tubes, though here's one book I found. The noise resistances in the high 100s and low/mid 1000s of ohms suggest that at least an 1:3 turns ratio would be desirable (up to 1:10 for low-noise applications). As the volume pot is transformed back with a factor of turns ratio squared, you'd want a 100k pot in this case.

That still leaves us with a relatively healthy gain of 20 (26 dB). I suggest you try running the input stage at rather high gain and then liberally apply negative feedback from the output to set closed-loop gain. (Make sure it's AC feedback.)

Using BJTs with an output transformer seems a little odd. They can, after all, easily drive headphones without one, and the output transformer tends to be a major (!) source of nonlinearity, too. Not to mention that a 22:1 turns ratio seems a little extreme, leaving you with no more than about 2.5 Vrms of maximum output (and that's abs max). Any +/-12 V solid-state amp easily does >6 Vrms.

I'd rather look at supplying an extra set of +/- 15 V rails for the o/p and placing a transformer before that. I think 5:1 would do fine. (That would make our necessary gain shrink to about 14 dB, which seems like a relatively easily attainable value with good linearity.) Not sure what's available that would be suited to audio use. I might even look into suitable toroids (again, not sure what materials you need in the audio range) and winding one myself if necessary.

The only thing I'm not sure of is handling of output DC offset. I wouldn't expect too much in a center-fed high-bias A/AB buffer (2+2 bias diodes, 10R emitter Rs or so), but still. Maybe a few 10 millivolts.

[b1o2r3i4s5]

Thanks for your advice, the reason that i need a voltage gain of 60 is because the output transformer has a ratio of 22:1 and i want a net gain of around 3, the output transformer is there to reduce output noise levels and protect the headphones from DC currents.

I still don't really understand how to put the anode stage and cathode stage together, or do i have to couple the 2 stages with a capacitor?

[b1o2r3i4s5]

I was also wondering if the valve cathode stage could drive the output transformer directly, the primary winding has a DC resistance of 300Ω and inductance of 4.2H.

[sgrossklass]

Originally Posted by b1o2r3i4s5

Thanks for your advice, the reason that i need a voltage gain of 60 is because the output transformer has a ratio of 22:1 and i want a net gain of around 3,

I did get that part. That doesn't make it much less of a problem though. You're dealing with almost as much gain as in a phono MM stage but much higher signal levels. Two inverting gain stages give a basic but by no means high-performance phono preamp. With just one, you'd be looking at no-fi.

Originally Posted by b1o2r3i4s5

the output transformer is there to reduce output noise levels and protect the headphones from DC currents.

The main motivation for using an output transformer actually is transformation of impedances. When using yours on the output, a 32 ohm load would be transformed to a comfortable 15.5 kOhms.

In terms of hiss, it probably is a better idea to improve input-side SNR. The most convenient way is an input transformer, as tubes tend to be voltage noise limited (like FETs) while having large supply voltages. Line-level transformers with a 2:1 turns ratio (for balanced/unbalanced conversion) don't seem to be that uncommon, maybe you can find other ratios as well.

(Of course you could also cheat and use a boring opamp circuit with 3x gain instead of the input transformer. Especially if you already have +/-15 V supplies for the buffer...)

Originally Posted by b1o2r3i4s5

I still don't really understand how to put the anode stage and cathode stage together, or do i have to couple the 2 stages with a capacitor?

This is commonly done, yes.

Originally Posted by b1o2r3i4s5

I was also wondering if the valve cathode stage could drive the output transformer directly, the primary winding has a DC resistance of 300Ω and inductance of 4.2H.

You need to consider equivalent AC impedance of the transformer + load within the signal range.

In order to compute that, I had to brush up my somewhat dusty knowledge of transformer theory.

It seems that for an ideal transformer with only copper losses and neglected capacities the amplifier would see a load of
Zl = Ra + jwLa + ( jwLm || n² * (Rb + jwLb + Rload) )
with
n = turns ratio primary: secondary
w = 2 pi f
j = ye olde imaginary number, sqrt(-1)
Lm = +/-M
La = L1 -/+ M
Lb = L2 +/+ M
Ra = resistance primary-side
Rb = resistance secondary-side
L1 = inductance primary-side
L2 = inductance secondary-side
M = k sqrt(L1*L2) = mutual inductance
k = coupling factor, ideally k = 1

For a no-load case (secondary open, not a bad approximation for a buffer on the secondary either), this reduces to
Zl = Ra + jwL1
|Zl| = sqrt( Ra² + (wL1)²)

With our values of 300 ohms and 4.2 H, that gives about:
f/Hz | |Zl|/ohms
020 | 607
050 | 1k35
100 | 2k65
200 | 5k27
500 | 13k2
999 | 26k4

The output impedance of your average cathode follower tends to be in the low 100s of ohms, sometimes under 100 (further reduced by feedback), so frequency response should be fine. Linearity towards the lower end would suffer somewhat though, as Zl is in parallel to the cathode (degeneration) resistor of typically about 1k.

(Properly) Designing with tubes and transformers is at least as painful as it is with "modern" stuff. (Not to mention that the result tends to be much heavier and more power-hungry...) I tend to look at hollow state as maximum effort for a given outcome - which happens to line up with the definition of "hobby".