So let's flesh out the details for the output section first, and find out what it implies we need from the phase inverter and preamp.
We'll work again from the assumption that the Hammond 269EX is our power transformer, and will not be substituted. We're also using solid-state diodes for the rectifier, so the B+ available will be 268v. This will be rounded up to 270v for calculation.
We've already found that the screen voltage will have to be lower than plate voltage to manage plate and screen dissipation. Since there are handy plate curves for G2 = 150v, we'll use that as our planned screen voltage. Right now we know plate = 270v, G2 = 150v.
If we look at the 6AU6 plate curves (see attachment below), we can draw a line from the knee of the 0v gridline, which occurs at Vplate = 40v and 15.5mA of plate current. That will be one end of the line, and the other end will be at 270v and 0mA plate current. From this line, we can estimate power output and the OT load. Note that we selected our screen voltage to be 150v, and that is the stipulated G2 voltage (Ec2) of these curves.
The equation for power output is Po = 1/2 * Imax * (Eb - Emin).
Imax = peak plate current, which is 15.5mA in our case.
Eb = supply voltage, 270v
Emin = plate voltage at peak plate current (voltage at knee), which is 40v.
The 1/2 factor converts from peak power to RMS power.
Po = 1/2 * 0.0155A * (270v-40v) = 1.78w.
The line itself implies a resistance of (Eb - Emin)/I = (270v - 40v) / 0.0155A = ~14.8k ohms. Ordinarily, we think of the load for one tube as being plate-to-plate primary impedance divided by 4. If the output stage runs class A (never cuts off either tube), each tube sees half the primary impedance. Given that, our estimated primary impedance required is 14.8k * 2, or about 30k plate-to-plate.
A suitable transformer is the Hammond 125A, which handles 3w of output, and has many possible primary impedances. Assuming no problems with this approach, we can get a 30k primary impedance by placing an 8 ohm load on the 4 ohm secondary tap, with the transformer wired to provide 15k when a 4 ohm load is placed on the 4 ohm tap.
Great, but how do we bias this thing? There are two places to find the answer, but the "Characteristics and Typical Operation" on page 2 of the
6AU6 data sheet tells us for G2 = 150v (shown in the column with Vp=250, Vg2 = 150) that the approximate G1 voltage for plate current cutoff occurs at -6.5v. We want to bias halfway between cutoff and 0v on G1, or -3.25v.
Now we need to know the plate and screen current at idle, and figure biasing. We see from the plate curves that plate current at -3.25v is ~2.6mA. Looking at the bottom graph on page 5, screen current when G2 = 150v and G1 = -3.25v is 1.05mA. Cathode current is then 3.65mA for a single tube, or 7.3mA for the pair, so the cathode resistor should be 3.25v/0.0073A = 445 ohms. 470 ohms is the closest standard value. 3.25v * 0.0073A = ~1/4 watt, so a 1w resistor will run cool. I'd suggest a bypass cap for the shared cathode resistor, and even a 50-100uF 16v cap is small and cheap.
The output stage at this point only needs a 3.25v peak input signal to drive it to full power, so we don't need anything bigger than a split-load inverter.
We know we have ~270v B+ and 150v for the screen. Since we don't need much signal from the phase inverter (or the rest of the preamp), there's no reason we can't drop the screen node of the power supply to 150v. We'll have the screen current for both output tubes running through this resistor, so we'll have at least 1.05mA * 2 = 2.1mA through this resistor. (270v-150v)/0.0021A = 57k ohms, and 56k is the closest standard value. We'll have to revisit this value as we determine the current requirements of the phase inverter and preamp. It is probably advisable to add a 100 ohm or so screen stopper in series from the screen supply node to each 6AU6 screen.
At this point, we need to breadboard just the output stage and feed it test signals to verify that the estimates so far agree with reality. There's no point proceeding with a preamp design until we know this initial draft works.
