Alex - PA1FOX
This may be a futile attempt if the intermediate gain stages (pre-driver and drivers) fall off with frequency too. I have not done this myself yet… still on my list… to actually publish a table showing the output at each stage with frequency to determine what needs attention BEFORE the PA. Then a change to the PA makes full sense.
William, I made some measurements on the TX chain without the PA. In my case this is up to the 2N2219's as I use the original schematic. The overall gain dropped 6dB from 3.5 up to 30Mhz. This should have some influence but is not a huge problem. With the power drive input which the 2N2219's generate for the IRF's it was clear that the IRF's were hardly doing anything at the high frequencies. I've played around with bias inductors and different impedance levels at the gates of the IRF's but apart from increasing or decreasing the overall gain for all frequencies, there was not much change in frequency gain curve.
What influences most is the output transformer of the IRF's. The trifilair setup is a good starting point but I could optimize the PA efficiency by adding and removing a winding at the primary side of the transformer. (So one winding ends up bifilair.) It's not ideal though.
With the change to RD16HHF1 (not the HVF1) I have kept the FT43-37 output coil. It could be a bit small for this, so I ordered binocular coils with a slightly larger size (BN43-202) to use as an output transformer. Yes, it also helps that the RD16 series is desigined for 12 volts. The IRF's lose their lazyness at 24 volts.
I'll post the results of the experiments for those who said to be interested. I find it very interesting myself as well.