Peter Dodd wrote:
snip snip ...

Most LF PA designs employ FETs in class D push-pull, with the output
taken via a transformer. A Pi section filter is then used to remove
the harmonics.

The documentation on the Decca transmitter modules shows a different
method, with the the output FETs driven in a bridge configuration.
Furthermore the output is fed via a large (airspaced coil) high C
tank circuit

The main problem with a bridge circuit for a radio transmitter is that
it is very difficult to apply bias if a linear final is wanted (such as
for PSK-31, or SSB where an LF band is wide enough) as two of the four
FETs (or bipolars) are "floating".  With suitable insulation it is
possible to transformer couple the RF drive, but it is the DC bias that
is not easy to apply to the floating transistors.

Thus I believe there are advantages in the "single ended push pull"
approach for LF transmitters, and could use 2+2 if four transistors were
needed for power reasons, or n+n for more power still.  Bias can easily
be applied for linear operation, or removed for "switch mode" higher
efficiency of Class C or D or whatever.

For output coupling of any transmitter, the main points are to transform
the impedance (resistance) so it is suitable for connection to a 50 ohm
load (antenna system), and to incorporate a prudent amount of low pass
filtering (more needed for Class D than Class B).  Class B can use
broadband coupling, but as I understand it, Class D is necessarily
"tuned" so a Class D PA is never a broadband stage.  There are no doubt
several ways of getting an equivalent outcome for coupling any
transmitter to any antenna.

73, Bob ZL2CA