There are a few other factors here. The big one ne is that no real
tube or transistor swings ALL THE WAY to zero volts at full current!
Another is that for the final to be truly ohmic would require that a
near-Class E or F condition, with the grid swinging from cutoff to
saturation almost instantly(as in a square-wave drive, sometimes used
in broadcast AM Class F setups).

The next condition is that the current drawn through the load at 2X
supply voltage must not cause the tube or transistor's bottoming
voltage to more than double! In the real world, this means that the
current(loading) must be backied off from CW conditions for any
particular device, just as the voltage must be. If you load a final
for maximum output at carrier, guess what-you will be lucky to see 30%
upward modulation with MOSFETS or somewhat better with tubes!

As a result, you want to design the final as though you were building a
CW amp to operate at double the supply votage with so short a duty
cycle that dissipation wouldn't overheat the device first. The active
device must not at any time be an impediment to drawing full AC current
through the load.

OK, here's what my experiments with the IRF 510 MOSFET uncovered at MF:

1: this device can make 55W a part at 17V, but for even 90% modulation
you must cut
loading to 37W a pair. Similar current derating should be expectd
from tubes, bipolars,
etc.

2: MOSFETS tend to lose drive as you raise the supply voltage. Here's
why-in Class C,
You are severely compressing voltage gain. As you add B+ in
modulation, voltage gain
increases. This increases the effect of Miller(reverse transfer)
capacitance). Any non-
neutralized common-cathode device will have this problem-or at the
other extreme
could oscillate on peaks! MOSFETS are lossy in all interelectrode
capacitances, so
even with "unilaterialization" in which both R and C(not just C)
are balanced, drive loss
in the resuloting "bridge" circuit" still rises with the supply
voltage. Since MOSFETS
cannot be driven beyond maximum safe gate voltage, you therfore
will need to modulate
the driver as well, or apply a few volts of gate modulation as an
alternative.

IF YOU DO NOT DO THIS, you will have 100% downward modulation real
easy,
but as little as 30-50% upward modulation, with a real serious
carrier shift problem
and audio that sounds like $%^&.

3: I've heard bipolar transistors also need some base modulation to
follow collector
modulation properly.

4: Tetrodes have a similar problem, but in a different way. Here, the
problem is that plate
voltage and evn bottoming voltage have little effect on plate
current, and real-world
tetrodes and pentodes require some screen modulation to propely
follow the plate
modulation.

5: Triodes with their low plate resistance may well give bottoming
voltage in proportion to
current, and therefore to voltage. This is a desirable condition,
but the only common use
of triodes today is in grounded-grid, where 100% downward
modulation is impossible.

One of the nice things about those Class E and F, true ZVS finals for
AM is that they give a far more ohmic modulation curve!