Dear J, LF group,

I assume you mean the voltage waveform across the dummy load is the one that
has the humps - which would indicate a significant odd harmonic content at
the output of the low-pass filter. Assuming the low-pass has the correct
component values, and so low attenuation at 137k and reasonably high
attenuation at harmonics of 137k, this implies an enormous harmonic content
at the input side of the filter - probably more 3rd and 5th harmonic than
there is fundamental..

So the next thing to look at is the waveform on the input side of the
filter. In a well-behaved current-switching class D stage, the waveform at
the input of the low pass filter should be a reasonable aproximation to a
sine wave, probably with some kinks and wiggles due to the "ringing"
produced by the switching operation. On the drain of each MOSFET, there
should be a waveform that looks roughly like a half-wave rectified sine
wave, i.e. near 0V during 1/2 of  the period, and the positive half of a
sine wave for the other half. In practice, the real waveform is somewhat
distorted and usually has various spikes and wiggles, especially around the
switching points. At the centre tap of the transformer, there should be a
waveform something like a full-wave rectified sine wave, with about half the
amplitude of the drain waveform. again, there will probably be various
switching transients present. The gates of the MOSFETs should have
reasonable antiphase square waves, with levels of around 0V, 10 - 15V.

If this is the case, then the problem lies with the low-pass filter, but it
seems more likely you will not see these waveforms. Let us know what you do
see...

BTW, to reduce the casualty rate among the MOSFETs, you can reduce  the DC
drain supply voltage to the PA. Because of the switching nature of the
circuit, all the voltage and current waveforms will be virtually the same
shape, just scaled down in amplitude. If you run it off less than 5V, it is
unlikely that the MOSFETs will blow up even under fault conditions.
Obviously, the gate drive voltage must stay at the original level to achieve
proper switching of the MOSFETs. This saves a lot of time and money when
experimenting with switching PAs.

You wrote:
> The only variable which I know of is the output transformer, final choke
and
> RF bypass capacitor.  Since I suspected the output transformer I have
taken
> it apart and rewound it with several windings, on two different cores, all
> suggested by different friends who are technically sound.  There are NO
> insulator disks between the core halves, and at 56 Watts of input the core
> cannot be saturating.  The latest winding matches the Jim Moritz, M0BMU
> alternative for the G0MRF amp on page 77 of LF Today. with taps on the
> output winding to set power output to lower values

Are you using the same circuit as the "M0BMU alternative"? Specifically,
there should only be a bypass capacitor on the DC supply side of the feed
choke, there should be no bypass capacitor connected to the centre tap of
the output transformer primary. If there is one, it very likely will result
in huge amounts of harmonics being present, and the odd waveform you are
getting.

Cheers, Jim Moritz
73 de M0BMU