The Effect of Idle Anode Dissipation On Max Power Output

[pdf64]

This topic crops up every so often. My view is that idle dissipation doesn't change power output per se, rather it just affects linearity. And as we see with solid state amps, with enough NFB, even cold bias can be made linear.
I made some tests on my Sound City 50Plus, which I've previously rearranged with a JTM50 circuit (HT and bias supplies excepted, given the FWB rectified HT and voltage doubler rectifier for the slightly inadequate bias winding).
The HT reservoir cap is 100uF, choke is a Hammond 157J (10H, 205ohms), screen grid HT node is 64uF.
The thing with the JTM50 is the high degree of NFB, 27k series resistor sampling the 16ohm output, 5k shunt resistor at the LTP tail.
Output valves are an old, well used pair of white shield Mullard EL34 that still seem to be working very well. BOH2 code, whatever that indicates.
The 8ohm output and cathode current (ie cathode voltage on a 1ohm cathode resistor) were scoped at full power output at the onset of clipping, at idle anode dissipations of 2.5W, 12.5W and 17.5W.
The HT supply is fairly stiff but sagged from 400V for the 2.5W test to 386V for the 17.5W test.
Anode current was assessed by measuring the voltage dropped across OT half primary resistance.
Here's a screenshot of the 17.5W test

[pdf64]

Here's the 12.5W test

[pdf64]

Here's the 2.5W test

[pdf64]

Here's the 2.5W test, but repeated with the NFB removed (mains voltage dropped a fraction unfortunately)

[g1]

So approx. 2W reduction in power (for 50W amp) going from 70% down to 10% idle dissipation?
I'm surprised how slight the crossover notch is even down as low as 10%.

[Helmholtz]

I can confirm Pete's results.
Power output of my original JTM50 slightly increases by maybe 2W when increasing plate dissipation from 7W to 14W.
As I'm calculating power from RMS voltage, I think that's mainly a shape effect (crossover and bias shift distortion lower RMS voltage).

So I tend to agree that bias has no significant influence on power output (at least within reasonable limits).

[trobbins]

I would have thought that any such influence would relate to how well the loadline (given you are using a resistor load for testing) approaches the knee of the pentode curve. The loadline for class B (assuming operation gets in to class B on signal peaks) could change with droop in screen voltage and droop in B+ (which could be different if average load on those supply rails were different due to different idle setup) - so would be good to know if the output stage was aimed for that knee or noticeably to one side.

[Helmholtz]

My JTM50 OT reflects an effective Zaa of 3.5k. I guess that means at or just below the knee with EL34s at idle (EL34s don't really have a sharply pronounced knee). Vp=450V, Vg2= 435V.

With significantly lower Zaa, i.e. above the knee, where plate current saturates, I'd expect ouput power to decrease at hotter bias.

[Helmholtz]

I was surprized to see almost no change of p-p cathode currents with different bias in your measurements.
Actually bias current should add to the peaks.

So I did the same measurements on my JTM50 using a Tek 2430A scope.
In fact I'm seing cathode current peaks rise with higher idle current.
Increasing idle current by 20mA increased p-p cathode current by 15mA.
The jump in cathode peak current is obvious when connecting/disconnecting the parallel resistor which changes the bias.
(I used DC coupling, of course.)