bias vs gain flatness

[PRR]

In another thread we were talking about bias, small-signal smoothness, idle dissipation, and "the 70% rule".

I worked out how to get SPICE to plot dynamic Gm of a push-pull tube stage. (PSpice hint: the "D()" operator takes the derivative of the Vg/Ip curve.)

Attached are plots for idle bias way-hot, near nominal and near optimum, and very cool.

The Yellow is the gain plot. Ideally this is a horizontal straight line.

[PRR]

Do not take the Voltages too literally. And I have some doubt about the computed Gm. The simulated 6L6 is not-quite like any real 6L6. But the trends are plausible. 

The test case is 400V supply with 900r plate load and 400V on G2. This was picked to be similar to 5F6a, however the 1K/plate load showed odd plate-bottoming and I decreased Rl to stay out of the funny-zone. The DC Sweep setup is flawed because the plate voltage does not kick-up above 400V on the off side of the wave. I ignored this because Pentode plate current does not change much at higher voltages.

"Pdiss %" is based on 6L6GC 30W rating.

The 130% Pdiss case has a severely lumpy Gm curve. This gives more gain for small signals, less gain on large peaks, large 3rd-order distortion.

The 20% Pdiss case has way-low gain for small signals, relatively exaggerating the peaks. "Rough" in the transition from small to medium signal and large 3rd all over.

I picked the "67% Pdiss" condition by reading the peak Gm, then finding the current which gave half of that Gm. This makes the both-tubes and cutoff points about the same. The 67% Pdiss case is about as smooth as this condition gets. Raising idle current just 10% gives a bump in the middle. Does this confirm "the 70% rule"?

I think not. Note that we could be running Original 6L6. The Gm curve "should be the same". Now 50mA 400V is 95% of 21W, or 105% of 19W. In fact I am wondering if RCA/MO, in designing a tube for push-pull, fitted it out so-that the plate was just big enough to handle the idle current needed to flatten the Gm curve.

Look another way. Gm is about cathode-grid dimensions. Pdiss is about plate dimensions. There's no reason they naturally match. RCA could have put a tiny cathode with a huge plate, so even 50% Pdiss would be over-biased; or a big cathode with a small plate so 200% was still not smooth. Of course they thought about it and found a happy proportion. If there is a "70% rule" it may come from idling 6L6-m near 100%, then the 6L6GC having plate 150% bigger.