Well... I was able to get my DX-20 to screen modulate (see the circuit in my prior post, above)!
The solution:
I used a 2:1 toroid transformer (a single Bytemark SB-1020-43 core, with 4 turns on the primary and 2 turns on the secondary) to turn my 50 ohm antenna load into a 200 ohm load facing toward the DX-20. This higher value of load resistance, after re-adjusting the existing pi network, results in a lower rf impedance on the output of the 6DQ6A. [Somewhat non-intuitive, but that's how the math comes out
]
As a result, I ended up with enough headroom to modulate the plate current by modulating the screen voltage (with no modulation of the plate voltage). Finding the right setting of the loading control, and the associated setting of the tuning capacitor, gave me the impedance (looking into the pi network) that I wanted (~1250 ohms)... and produced excellent linearity... i.e., a nice correspondence between the audio waveform applied to the audio amplifier, and the output of the off air monitor... with close to 100% modulation. However, finding the right setting is critical with this approach. I had the input audio sine wave and the output of the off air monitor on a dual trace scope... and I carefully adjusted the loading and the tuning to get the largest output that I could from the off-air monitor... while avoiding distortion on modulation peaks.
Update (added 1/22/08): The above approach works, but there is a remaining issue. By increasing the antenna load resistance to 200 ohms (using the 2:1 r.f. toroidal transformer), and after adjusting the loading capacitor and the tuning capacitor of the existing pi network, I ended up with my desired impedance (1250 ohms) looking into the pi network at the fundamental frequency... but, I also ended up reducing the ratio of: the impedance looking into the pi network at the fundamental frequency (e.g., 3.8 MHz) vs. the impedance looking into the pi network at harmonics of the fundamental frequency.
Again, my target was 1250 ohms of impedance looking into the pi network at 3.8 MHz. With the existing pi-network inductor (roughly 12.5 uH), I ended up with the following (for a 200 ohm resistive antenna load)
Tuning capacitance: 184 pF (no problem with the existing DX-20 tuning capacitor on 80 meters)
Loading capacitance: 418 pF (ditto for the existing DX-20 loading capacitor)
However, the impedance looking into the pi network at the frequency of the second harmonic (7.6 MHz) ... which is approximately equal to the impedance of the tuning capacitor at 7.6 MHz... is roughly 113 ohms. This is 9% of the impedance looking into the pi network at the fundamental frequency.
Therefore:
Since the Class C current waveform contains lots of harmonics, this produces at least one problem. The harmonics in the plate current waveform develop a larger than desired voltage across the impedance of the pi network... which uses up voltage headroom. It takes some careful tweaking of the tuning and loading controls (while watching the output of my off air monitor on a scope) to find the combination that produces good modulation linearity, a high modulation index, and maximum peak output power.
Playing around with a pi network design program... it appeared that by using a 450 ohm antenna load on the DX-20, and the existing pi network, I could produce in the plate load on the tube that I am looking for (1250 ohms)... while obtaining a somewhat larger ratio of: the impedance looking into the pi network at the fundamental frequency vs. the impedance looking into the pi network at the harmonics of the fundamental frequency. So... I changed the rf transformer from 2:1 to 3:1 to convert my 50 ohm antenna feed to a 450 ohm load on the DX-20. This was helpful... but it still takes some careful tweaking of the loading and tuning capacitors (while watching the output of my off air monitor output on a scope) to obtain good modulation linearity, a high modulation index, and to maximize the peak output power.
A much better way (if you don't mind changing the component values in the pi network) to reduce the impedance of the pi network by a factor of 2 at the fundamental frequency, while still maintaining a reasonably high ratio of the impedance looking into the tank circuit at the fundamental frequency vs. the harmonics of the fundamental frequency, would be as follows:
a) Keep the antenna load resistance at 50 ohms (i.e., don't use the 2:1 or 3:1 rf transformer)
b) Reduce the inductance of the tank coil by roughly a factor of 2 (e.g., use the 40 meter tap on the tank circuit when you are working on 80 meters)
c) Increase the tuning capacitance by roughly a factor of 2. Add extra fixed tuning capacitance, if necessary.
d) Use the loading control to load the transmitter to 1/2 of the output power it would normally produce (e.g., 20 watts for a DX-20). Add extra fixed loading capacitance, if necessary.
Using this approach, you would have more tuning capacitance at resonance (about twice as much), and therefore a 2x (roughly) lower impedance looking into the pi network at the second harmonic frequency.
Using the pi network design program, you would have (for example):
L = 6 uH
Tuning capacitance = 342 uF
Loading capacitance = 1500 uF
Q= 12
The impedance of the tuning capacitor at 7.6 MHz would be 61 ohms, i.e. 5% of the impedance looking into the pi network at 3.8 MHz.
Now the real test: switch between plate/screen modulation (a.k.a. plate modulation) and pure screen modulation on the air... and see if anyone notices.
Best regards
Stu