I thought some hams would appreciate the fruits of this research
by WSQT Guerrilla Radio in DC. We have managed to make a 55W power amp
for 88MHZ from a random pair of RF transistors, one a VHF device and
one a UHF device! The amp works clean and stable despite its
unconventional design, forced by components available to dumpster-
diving pirates with little funds. Although broadcast amps are single-
frequency, the matching networks used in VHF transistor amps are so
low-Q that communications amps can often span their bandwidth with
only a little further development, treating the matching networks as
filters for purposes of design.

The results of this research should be relevant to Amateur FM mode
operation in 2 meters(and 6 too for that matter if anyone is so
inclined), as well as to the FM broadcast pirate work this particular
amp was made for. Never done a linear(not used in FM nor AM broadcast
in these power ranges), so have no idea if it would work or if one
side would clip first and generate splatter! This is a 55W amp, but
if you are sitting on two medium size or big VHF or UHF transistors
that don't match, consider trying them together!

The amp works like this: Each half is just an ordinary VHF power amp
with one transistor, with the junctions of the input and output L
networks ties together in parallel. Each L-network matches to twice
the usual value at the junction point, as there are two in parallel.
The output matching/loading capacitor and lowpass filter are in
common, as is a pi network to allow a 600 ohm input impedance for each
side(to keep coils realistic). Nothing else too it, except that
stabilizing it works quite differently than any of those matched-
transistor parallel designs that are referenced elsewhere and provided
background for this project.

The key to the whole design is this: You CANNOT stabilize a
dissimilar parallel amp with the usual low-value resistors running
base-base and collector to collector. That was tried first-and the
collector side one ran red hot! Strange, as the amp sounded clean
with no big spurs, though it loaded awful heavy. Removing that
resistor brought the expected storm of hash, while using 500 ohms
instead of 10 collector-collector brought near stability, but not
quite. At that point I was able to load and tune the amp close to
correctly, but power was short of expectations and stability was
marginal. Then I tried removing the base-base resistor-Presto! Far
less, even after removing the collector resistor too! Each amp needed
to be damped separately for low frequencies just as though it was
running alone, and that was the end of all the hash. At that point,
the amp could finally be loaded to full power.

Here's the rest of the stabilizing tricks, used on each side: 5 ohm
resistors to ground from each base(10 ohms base-base center tapped
results, not 2ohms floating), connect the base RFC's to a pad with a
150PF cap to ground and a 15 ohm resistor paralleled by a bead choke
instead. On the collector side, each RFC simply needs the same 150PF
cap from its pad, and a bead choke shunted by a 10 ohm resistor to
another pad with a .047 capacitor and a single 100uf electrolytic for
the pair, and you've got stability!

Except for the full on pi network matching the input to the base
networks, it is otherwise just a dissimilar pair of 25-30W amplifiers,
but they work,and a Pentium 4 heatsink and fan is enough to keep
things cool. Expect the PA to draw about 6A at 12 volts, and expect to
use a 5W driver to run it.

As for why the collector resistors ran so hot with the amp running
clean, here's what I suspect: Since the two devices are so dissimilar,
the current has different rise and fall times on each side. This
amounts to different even-harmonic amplitudes and phases on each side-
which amounts to a push-pull harmonic current superimposed on the
common-mode, even in the absence of oscillation! This isn't coupled
out, but appears across any stabilizing resistor used collector to
collector. With single-ended transistor amps, the lowpass filter does
most of the work on stopping harmonics anyway, and in the absence of
oscillation or a parametric (addition) situation, no new frequencies
not in either amps separate output appear. Only relative phases and
amplitudes of harmonics need differ side-side to create a push-pull
current.

Anyway, the amp works, dummy load testing sounded clean, and on-air
testing gave no nasty suprises and no overheating with a Pentium 4
heatsink and fan.