A fellow amateur (PA0SIM) told me that he worked often with a simulation package called Simetrix. You can freely download it and work with a lot of components without any charge. I downloaded it from the site, and got luckily some help to do the basic things.

I completely put the amplifier (not the electronic switch) in the package and did a lot of simulations.

If you want the complete .wxsch file, give me a mail and I send it to you via mail (this website doesnt support wxsch-files).

 

Important parameters:

V+ = 20V

Input 14 MHz, 20V peak-peak (4watt)

Output after the filter = bit more than 50W (72V)

I also copied a graph to this site so you can see what signals I have in my amplifier.

Lets give it a look:

The red wave is the voltage on the 50 ohm dummy load. (Probe-1-NODE)  It gives a top of a bit over 70 Volt, so 72V. Power is 72e2/2*50 = 50 Watt

The yellow wave is the bias voltage, it is very close to 3.9 Volt. The Q1 and Q2-G probes.

The blue wave gives you the dissipated power in the upper FET. The average is quite high ! About 40 Watt. This goes into heat and has to be sunk into the cooling block. Two FETs give 80 Watt, which is not a problem at all, for my big over-dimensioned cooling block.

The green waves are the voltages on the FET drains. This is really an ideal situation when the parasitic inductances are very low. In my case the voltages have some serious harmonics in it.

The brown waves are the input voltages on the gates of the FETs. They are a few volts around the bias of 3.9V.The purple wave is the current flow trough L7. This is an average of about 7Amperes.

In the scheme you see that I have put a few L's in the source leads of the FETs. This is to simulate the leads of the device. Two cm lead can easily be 5nH !

The longer the leads of the FETs and the power input in the middle of the primary of the output trafo, the more higher harmonics you get in the drain voltages (which is waste).

So, KEEP THE LEADS SHORT. 1mm is better than 2mm,.......

 

It is fun to simulate the amplifier with Simetrix. I can tell you, a lot of issues are better understood after a few evenings playing with it.

Well, this is it for today,....