In my Viking II, I used just the series dropping resistor that had supplied screen voltage from modulated B+ before. I simply moved the resistor to unmodulated B+.
It was a significant improvement. With the resistor on modulated B+, the screen was overmodulated, and this caused a kink at around -85% modulation. This kink can be seen in the classic trapezoid pictures from old ARRL handbooks.
However, full self-modulation of the screen produces slightly less than the absolute optimum screen modulation level. WA1KNX Dino found that splitting this resistor into a larger value to unmodulated B+, and an even larger value to modulated B+, about a 40/60 ratio, resulted in the very lowest distortion modulation.
"Class C Optimization for Ultra Low Distortion", AM Press/Exchange issue #71, May 1989, Dean, WA1KNX,
www.amfone.net/AMPX/71.htm"Screen Modulation Optimization for the Heathkit DX-100", Dean, WA1KNX,
http://www.amwindow.org/tech/htm/scrnmod.htm"The Self-Modulated Screen Grid", AM Press/Exchange issue #98, Jan/Feb 1992, Bacon, WA3WDR,
www.amfone.net/AMPX/98.htmAnd now, AB2EZ Stu has brought up another idea: super-quality screen modulation with regular quality plate modulation. This isn't screen self-modulation, but it is of interest in the general topic of screen and plate modulation.
http://amfone.net/Amforum/index.php?topic=12117.0