There's been some talk here about tube choice for these UltraSonic amps. This thread isn't about tube theory, but I think it might be interesting for people interested in these amps to talk a little bit about how and why tubes are chosen for a specific job.
Now me personally I've tried out a bunch of tubes for these circuits, including 1626, 6V6GT, 71A, 47, 2C22 (output), 6SN7 (output), 6E5P, EL84, EL81, EL41, 6P31, 4P1L, 3C24, 6L6, 6N16P, 8025A, 8012A, 6BQ6, ECL82, 6S45P and some I've forgotten by now. My current everyday amp runs 6E5P finals.
First since not everyone is a DIY guy (you don't have to be to understand this tube selection thing), let's in short go thru the basics.
Aside from "trying out different tubes" how can one know in advance what tube type will sound best in a specific circuit?
Well, what is "best sound"? That's a whole discussion in itself. For the sake of brevity, best sound is lowest possible distortion in an open loop system (Sonic's amps do not employ gNFB at all). I would venture to say if one is specifically fond of coloured sound, perhaps one should look for SE amps with traditional series feed OT to suit their tastes.
How is minimum distortion achieved? Here we come to load lines. Curves and load lines are the heart and soul of any tube system. They tell you so much about the tube and it's operation in a specific circuit I cannot even begin to explain it in full here. The basics are very simple though.
Let's draw a load line. Now in an old fashioned amp, which utilizes the output transformer primary coil as the load for the output tube (primary is in series with the tube), the load is something between 1k to 10k, usually let's say 5k. There were old time rules of thumb for "good results", for example the 3x rule, OT primary impedance should be 3x the output tube anode resistance for "best results" i.e. an acceptable compromise between output power and amount of distortion generated. If you increase OT primary impedance, you lessen output power and also lessen distortion, and vice versa.
Now let's pick a tube everybody is familiar with, the 6V6. To draw a load line for 5k primary load, you take your B+, let's say 300V, and divide it by 5000. Ohm's law says the result is in amperes, converted to mA it's 60mA.
Now look at the curves. The right side endpoint of the load line is 300V / 0mA and the left side endpoint is 0V / 60mA:
All possible op points are on this red line. Op point or operation point is the static condition, the rest condition where the output tube is biased.
The curves denote grid voltage. The control grid is the 'input' of the tube. This is where the input signal is applied. Each curve has a grid voltage attached to it on the top.
Which point should you choose? Now some people like to just try things out and try to find by luck or accident a good sounding op point. That is just shooting in the dark blindfolded and with no ammo.
The 'point' of an op point is that once signal is applied, there is sufficient amount of headroom on BOTH SIDES of the load line (looking out from the op point).
Let's say we chose grid bias point at -25V. So that's the op point now, the resting starting position. Now apply a signal to the grid.
Voltage at grid starts going up, we move left on the load line. Plenty of room to go around.
Now the sine wave reaches it top, and starts going down. Grid voltage goes down, down, past the op point, then what.
There is only a tiny amount of room left to go on the right side of the load line.
Now a sine wave has equal amounts of positive and negative peak to it. If we select this op point, the tube is unlikely to be able to reproduce the sine wave even close to original.
So logically, one should choose an op point that is rightabouts halfway on the load line.
So, about -10V. This is not an uncommon op point for a 6V6 amp running at 300V B+.
Now let's also draw the biggest possible input signal swing to the load line. This is the hard clipping point for the amp, since the grid cannot go positive unless specifically driven by a certain kind of circuit (which was not available / used in traditional circuits). Sonic's amps can drive output tube grids well into the positive part, but we're not there yet.
Now with this load (5k) and chosen bias point (27mA bias current at -10V grid), with the grid going between 0V and -20V, we have the anode voltage varying from 110 to 230 V, at op point it's about 175V.
What does this mean? Well, for one, we have amplification, 20V input signal produces about 120V output signal for a gain of 6. But with regards to signal distortion?
The blue part is the positive input signal portion, the green part the negative. You can easily see with your eyes, that they are not of equal lenght, meaning the equal variance in input signal does not produce equal variance in output signal.
The output signal waveform doesn't look exactly like the input signal waveform. This is distortion.
Now it's mostly second harmonic and third harmonic, but we won't go into that here. We'll stay on the level of avoiding distortion alltogether.
Now you can also see, that if you chose the op point to be any more on the right, the distortion would be significantly higher, because the grid lines start to bunch up closer together on the right.
Similarly, if you chose the op point more on the left, your maximum available anode voltage swing would be lower. So with the design limitations (5k load) you have, you are now at the optimum place for distortion and output power.
But since we are not limited by having to use a 5k load, or the 6V6 tube, what would be the ultimate tube and load line that would produce NO DISTORTION AT ALL? (This is hyperbole, there is always some distortion. But in comparison to the presented example.)
Well, for one, you would want a tube that has curves that are equally spaced apart. Not like those right side curves on that 6V6 load line.
Second, you can see that the curves always tend to have a different slope on different 'heights' (on the Y axis). So closer to bottom they are more level, and closer to top they are more steep.
So, the load line should stay within the same vertical region. The logical ultimate load line is thus horizontal or close to it.
So in summary:
1. Curves should be as evenly spaced in the voltage swing area you plan to utilize as possible
2. The load line should go thru only one "vertical section" of the curves, so as close to horizontal as possible
3. The op point should be about halfway point between B+ and 0V gridcurve.
These are the ingredients that produce maximum fidelity, transparency and liveliness to a tube circuit.