Hi Paul,

>> It's a physically small FET and maybe was not screwed down tight enough with the Sil-Pad interface,
 
Same thing happened here a few years ago during an attempt to melt off a thick layer of hoar frost from the antenna. It almost worked and I could see the resonance creeping up but then the PA died. At the time, the four BUZ111 devices in my 300 W class-D were bolted to sil-pads using M3 nylon screws. Due to increased dissipation with reactive load, the plastic became soft enough to release thermal contact. After FET replacement, I went to metal bolts and insulation washers, with no such problem since. This was actually the one and only FET that I lost in my LF career (knock on wood).

As capacitance is not a big issue at LF you could also solder your FET to a copper heatspreader plate and use a larger pad to a grounded heatsink. Or use a configuration with grounded drains and negative supply and take off the output from the source pin. You'll need to use an input transformer with a gate-source secondary but that's a good idea anyway.

Regarding dummy loads for LF: I've been using a coffee maker or a toaster, with 1 kW at 230 V being close to 50 ohms, and negligible reactance. Just make sure to use an older one containing no stupid and unnecessary microprocessor. Albeit American 120 V devices may have less favourable impedances. 

Best 73,
Markus

-----Ursprüngliche Mitteilung-----
Von: N1BUG <[email protected]>
An: rsgb_lf_group <[email protected]>
Verschickt: Sa, 24. Nov. 2018 12:48
Betreff: Re: LF: No 2200m TX tonight

Hi Markus,

That is very interesting. It changes about 25 % during the first 2-3
minutes, then it seems to settle down and not change any more. It
could be moisture somewhere, but any moisture here is solid ice or
frost now. I did not find any ice or frost in the transformer box or
the variometer. Could be insulators or something with the antenna
itself. I don't see any "fuzz" on the scomematch voltage trace so I
think (hope) nothing is arcing.

What worried me is this did not happen last winter so something has
changed. Everything accumulates some dirt here because of blowing
dust, smoke, etc. I wonder if a small amount of dirt on insulators
plus moisture can combine to make funny things happen.

Anyway I examined the little PA and it seems to have died due to
poor thermal interface between FET and heatsink. It's a physically
small FET and maybe was not screwed down tight enough with the
Sil-Pad interface, which was also a previously used one.

Normally I do not like to put drain voltage on the heatsink but as
an experiment for this little PA (which is totally an experiment
itself, but served me very well last winter) I will isolate the heat
sink from the chassis/PCB and mount the little FET directly to it.
The thermal resistance would be much lower! I think this is fine so
long as nothing shorts the heat sink to ground. In that case some
fuses die. ;-) This would add some pf of capacitance between drain
and ground but it would be in parallel with the quite large C of the
Class E tank, probably not much difference at 137 kHz!

Parts to repair the big PA should arrive Wednesday.

73,
Paul