Have similar output power on QCX-30. (1.6W)
Am looking into the Class-E output stage. Have done a careful simulation, using Hans' nominal values for the low-pass filter toroid inductors:
Class-E drain efficiency is calculated: compare DC power coming in through L4 with RMS power delivered to a 50 ohm load resistor.
As a sanity check, a lossless case was tried, where the three BS170 MOSfets were replaced with an infinitely fast switch, having ON
resistance of very near zero ohms, and OFF resistance of many MEGohms. All the inductors, capacitors were ideal (no parasitic resistance).
Gratifyingly, efficiency was 100%.
Estimates of inductor losses, and capacitor losses were added to each component.
An estimate of switch ON resistance, rise and fall times were added.
Another run included three BS170 models.
Efficiency did drop some, but still produced good output. It is a good design.
A 'scope check on all the waveform shapes around those BS170's looked excellent, just a hair low.
I did notice that a simulation sweep of the low-pass filter put the cut-off frequency very, very close to the top of the 30M band.
Suppose capacitors were lossy....
They're tiny, and if any were soaking up the power we're missing, they'd get warm-to-the touch.
So try a key-down and see if you can notice heating.
If they're running cool, then suspicion moves to inductors L1, L2, L3. They would likely benefit from
some turns taken off. But that's risky, because filter attenuation of harmonics suffers.
I had intended to try measure those capacitor Q's, but test equipment must be built for measurements,
and that darn roundtoit seems to be missing.