Some time ago I traced the 307a that kindly Vegard Winge sent me from Norway. As suggested by Vegard, I re-traced them with lower screen voltage and increased feedback to improve the linearity of the 307a in this mode.
The modification to the curve tracer is simple. A simple resistor divider (R1, R2) provides the feedback to the grid. The grid driver provides a constant impedance so works fine. With a potentiometer you can adjust the feedback ratio and look at the impact on the curves. Very handy!
With 15.5% feedback and 125V screen voltage I got the best curves as shown below:
Nice to see the anode resistance coming down to 1kΩ.I tried matching a triode model to these curves with a good result
Still a 5K-7K will do best, but let’s have a look at the performance with a 5K load first.
A) 5K Load and OT gapped at 50mA
With 125V screen, an 5K/50mA OT and biased at 308V you can get about 2W @1.12% with 86Vpp.
B) 5K Load and OT gapped at 65/70mA
With 125V screen, an 5K/65mA OT and biased at 307V you can get about 3.6W @1% with 113 Vpp. About 1.6W more of output power by increasing bias current by 10mA and increasing the grid drive from 86 to 113 Vpp.
C) 6.5K Load and OT gapped at 50mA 
If you consider an 6.5K OT instead, then an LL9202/50mA (or even a LL1623/50mA but impedance is 6K instead) would perform nicely to achieve 2.5W@0.85% but with a drive of 102Vpp. This would be enough for a DHT headphone amp, as some were considering in DYIAudio forum not long ago.
Hi Ale-
I’ve been following your build with interest, and Michael Koster’s design. Plate to grid feedback opens up many TX tubes for audio. I have a Sofia curve tracer, I should try this adaption (if I don’t damage the tester!).
Regarding the curves above, is that ’tilt’ at the top of the curves the tetrode ‘knee’ coming into play?
Thank you, Kent
Hi Kent,
Lucky you that you have a Sofia tracer! Nice…
I’m not familiarised with the grid driver circuit but if properly implemented should be fine.
Regarding the tilt of the curves, I haven’t investigated it yet. I’m sensing cathode current and screen current loop is out of the sensing path as the return of the screen supply is on the cathode. What could be is change in current in the feedback path. To be analysed!
Ale