Start with 1V is adequate as the dive is Current not Voltage for transistors.
Altering the resistors make the previous stage a bit happier and lower the loss
to the resistors but the basic issue is crappy transistors, transformers, and layout.
For L8 and L9 the common practice is bifiler wind on a common core
so the current balance out to zero flux in the core. I used 2t bufilar #26
on FB43-202 to get about the same result. Then again anything
did help compared to existing. The driver and predriver transformers are
not so good either. While your at it the q90 transformer is not as good
as could be. T11 rewind it as something reasonable for 10mhz and up.
try 2:3 or even 2:2. For better IMD run about 150 to 200ma per transistor
on the finals.
Trying to get more than 1.8W on 10 with basically 10W at lower end is repeating
the work I did 2.5 months ago. Ditch the 3904s put in 2222s tell me that is not
remarkable difference. Or go straight to 3866s or 5109s. FYI everyone found
the MPSH10s to be rather marginal improvement.
Then add RLC feedback to the driver and Predriver to keep the gain fade
with frequency a bit flatter. With that you should be to 12W on 80 and 40
and about 5-6W on 10M. The spur at 17mhz (10M) will now plague you.
Reset the bias to Q90, its a low level stage and even 10ma is high.
some of them are running at over 25ma for 1 milliwatt or RF. BEst way
1K is series with a 2K with a parallel .01uf chip. The idea is 3K DC R
and 1K AC R. Use a real transistor there too MMBR2369 is better
(flat to about 22mhz) BFR106 (flat pas 30mhz) work very well.
A mmbt5179 should work if kept under 10ma as well.
You will have to reduce the extremely high standing current of the
predriver and driver. The 22ohm emitter resistors tend to cause voltage
limiting on the pre-driver and driver as well as much to great emitter
degeneration. Using 8.2 to 10 ohms improved that. Bypassing them with
220/330/470 pf helps get some upper frequency peaking.
Flattening the power was more or less doable.but spurs and harmonics made
it worthless. Also if the gain gets to about 60db at 20mhz and up the amp is then
subject to oscillation from ground loops extending back to the 33mhz low pass filter.
the fix is slice the board and separate. There are secondary loops with DC power
lines having RF (TX-DC+) due to parallel paths under the relays and near the
power amp. Look at Kees picture of the board without relays.