Dear Dick, LF Group,

Maybe there is an error in the schematic, from which I do not know.

I am not aware of one...

The output transformer and some of the surrounding components seem
to play a key role in the results obtained.

Certainly true...

I tried the tip Jim, M0BMU gave, with 2x4t bifilair (made of 3x4x0.3mm
wire)
as primaire coil, and 16 turns secundair.
This coil didn't gave the results I expected.
I found an primairy inductance of 150uH , between the two outher wires and
37uH between the centre tap and each outher wire.
The secundairry inductance around 600uH.

These values are almost identical to what I got for my transformer, and are
quite sensible for this circuit.

But note that, if you use this transformer, it is absolutely essential to do
the other circuit mods I mentioned at the same time (i.e., eliminate C23,
increase choke value, replace 10n drain capacitors with series RC),
otherwise things will be made worse rather than better.

Examining the waveforms with an oscilloscope will reveal whether things are
working properly or not. The gate drive waveform should be a slightly
rounded-off square wave about 12v pk-pk, while the drain voltage should be a
sort of half-wave rectified sine wave with peak amplitude roughly 3 times
the DC supply voltage (although it will also contain some switching spikes,
and if using the original 'MRF output filter, you actually get a bit more
than half a sine wave cycle). If you have a current probe (or make a
suitable current transformer), the source current should be approximately a
square wave (probably with some ringing at the leading edge), with an
amplitude equal to the DC input current. The waveforms for both MOSFETs
should be identical, but 180degrees out of phase of course. The voltage at
the transformer secondary will be roughly sinusoidal, with peak amplitude
close to 4 times the drain waveform amplitude.

In the original circuit, I found the waveforms had multiple peaks for each
cycle of the input drive signal, due to resonances at higher frequencies
involving the transformer leakage inductance, the MOSFET capacitance and the
other capacitors in the circuit. The amplitudes of these higher frequencies
were larger than the 137kHz signal. The filter prevents most of these
harmonics reaching the load, so there was still substantial 137kHz output,
but the higher frequency signal components circulating in the PA led to
greatly increased losses and reduced efficiency.

Cheers, Jim Moritz
73 de M0BMU