Alright, we'll step through a class AB example. NOTE: Like any
real design, I had to go through a couple iterations, as well as make choices about trade-offs at every step. So this is not a mechanical process, but the following should help you see what "right looks like."
Class AB is mostly about making more power than is possible in Class A, and generally uses a higher supply voltage, lower load impedance and larger bias voltage than Class A. I'm going to stick with the 12AU7 as our output tube (no fair trying to change to the 6V6 in your last run-though; you won't see how things should change if you switch tubes).
... the PT I want to use for this project. Its an AnTek toroidal transformer model AN-05T240. It has a 220V tap that I plan to use.
220v * 1.414 = 311vdc, and subtracting diode drops will put us around 310vdc. That seems like a usefully-high B+ to build a Class AB amp with, so we'll use that for our new amp.
Now it's time to look at a
12AU7 datasheet, and the maximum ratings on the first page. Our constraints are a plate dissipation of 2.75w and a peak cathode current of 60mA. We'll bend the rules a bit and use 310v instead of the data sheet maximum of 300v.
Now it's time to look at the plate characteristics on page 4. Look at 310v and 0mA plate current. If the -24v gridline continued all the way to the axis, it would land somewhere near 310v, so let's estimate the
cutoff voltage at B+ as being -24v.
I said earlier that the limit of class A operation is half the cutoff bias for your plate voltage. If you run into no other limit first, the grid is able to swing up to 0v and down to 2x bias, which is just enough to momentarily cut off plate current. Since plate current cutoff is a requirement for class AB, we must bias
more than half the cutoff voltage.
We need something between -12v and -24v... I'll pick -17v. It's far enough toward cutoff that when the class A loadline extends above B+, we're still likely to hit cutoff not far above -24v.
Now for our Rule #2:
The loadline for maximum power intersects the 0v gridline at 0.6*B+. That happens at 185v and ~27mA. You can calculate the loadline impedance manually (like I did at first) or with:
Rl = 1.6 * B+ / Imax = 1.6*310v / 0.027A = ~18.4kΩ plate-to-plate
Class A load = 18.4kΩ / 2 = 9.2kΩ
Class B load = 18.4kΩ / 4 = 4.6kΩ
Check: (310v - 185v) / 0.027A = 4.629kΩ, so our math makes sense. The attached graph shows the proposed operating point, the class A loadline (green) and the class B loadline (red).
When does the tube transition from class A to class B? When the
other tube tuns off. We see the class A loadline intersect the axis between the -24v and -26v gridlines. Let's then say cutoff happens at about -26v. That is our idle bias minus 9v of signal. Since each tube is receiving equal and opposite signals, the opposing tube is at idle plus 9v at the same moment. Therefore, our tube switches from the class A to class B loadline at -17v + 9v = -8v.
In reality, there is a transition from the class A line to the class B line that forms a smooth curve between them, but it fully shifts to class B by -8v on the grid.
Both these lines look like they're well above the dissipation limit. What's up with that?
Well, the transition from class A to class B loadlines happens at a point on the class A loadline below the dissipation limit on the class A line. Our headache now is we need to calculate average plate current, power input, power output and see what the resulting plate dissipation is to see if this choice of operating condition will work.
