Tube specs were quoting theoretical power and not the actual power to the load. They also did not include distortion power.
QST articles regularly stated to run substantially more than half the RF input and even quoted mod iron losses as around 20%.
Heck, Im running 304TL's to modulate 250TH's as well as a four pack of 813's. We wont discuss what reaches the antenna
but be certain its fully modulated.
Clipping the highs isnt going to do anything for the peak power of male human voice as its already falling off at 3 kc. The true power is in the lows. Cheap iron already kills the lows so if you have limited power put it into "communications" audio until you can score the right iron. Dont get carried away with lots of processing either with cheap iron, it doesnt like the high average power. If people tell you the audio sounds restricted just tell them its a Collins
Im using a wet 7500W mod xfmr with the 304TL's, no worry about saturating
but Id still like to find something that doesnt need a cherry picker to move. Dont laugh...it was cheap!
All the shaping is done at low level ( 150-3500cps) and with just a pi-net splatter filter at high level. I can easily go 150% on the positives and see nothing nasty on the spectrum analyzer.
Before that I was using 810's and a 600W xfmr to modulate the 250TH's and even at 900W RF input I couldnt get the clean processed levels I wanted. Thats now with the 2 x HK-354 RF deck which is running at 600W and works fine.
In the late 50's I had the classic PP 250TH's with 810's and didnt know then what I was missing until I tried duplicating it a few years ago. And went thru a big awakening and actually read the old articles with understanding this time around.
Carl
KM1H