Andy and All

The parasitic diode is no good in this configuration for catching the
spikes from what I call the transient soak choke, I use separate clamp
diodes up to the rails effectively 'crosing over' the centre tapped
choke - see the circuit in the new LF Experimenters Book.

I see. Yes, the MOSFET parasitic diodes are known to be slow. I will get
a copy of the new LF book and have a look at the circuit.

By my reckoning no extra filtering is then necessary as second harmonic
is practically non existant due to the square wave drive,  and third
harmonic is 26dB down to start with due to the tank and then seriously
reduced further by the antenna loading coil.

True, but watch out for HF, and even VHF, radiation from the wire between
the FETs and the series tuned circuit. I mesaured rise times of a few tens
of nanoseconds on my experimental 4 * IRF640 H-bridge and the harmonics
were strong even on my 144 MHz receiver! Well, the PA was breadboarded
and the "hot wire" was half a meter long. A metal case enclosing the PA and
the series circuit will probably help...

I also have it on good authority from a transmitter designer, that the
classic PI filter with input C is VERY bad news for using at the output
of  MOSFET PAs.

Class D PAs come in two flavours. Current switching and Voltage switching.
Full and half H-bridges (the Decca TX has this configuration I guess?) are
voltage switching amplifiers where the drain voltage is a square wave and
the fundamental frequency energy is sucked out by a series resonant circuit
to the load. The transistor current is sinusoidal for obvious reasons. One half
period per transistor. Yes, a C input filter is really bad news in this case as 
it
would lead to infinite current spikes if all components were ideal...

Current switching PAs, on the other hand, works the other way around. The
current through the transistors is a square wave (a constant DC current that
is alternating between the devices) while the voltage (across the "off" 
transistor)
is sinusoidal. This kind of PA *NEEDS* a C input filter, or a parallel tuned 
tank.
The choke feeding the center tap of the output transformer is very important
since it keeps the DC current constant over the RF period (the voltage at the
center tap is NOT a steady DC!) This circuit is probably more likely to blow
FETs than the voltage switching circuit. I can imagine two reasons for this:
1 - the DC feeding choke (and the leak inductance in the transformer) will kick
up the voltage during the drive gap, 2 - a load mismatch may kill the 
transistors
with excessive voltage (depending on LPF design). Some kind of "spike
eaters" such as Zobel networks (R+C) are necessary to protect the transistors.
The G3YXM design is a well known example of a current switching PA.

73
Johan SM6LKM