Power Stage Transient Simulations - Fixed and Cathode Bias

[martin manning]

There is a lot of discussion about what goes on in fixed vs. cathode bias, and what the differences between shared and separate Rk, bypassed and un-bypassed are. Here are the results of some simulations for a pair of EL34's in push-pull.

[roberto]

Thank you for sharing, Martin.

[Bob S]

Thanks for posting Martin.
Interesting to see the huge difference that my ears tell me.
The math behind this is scary though - to my fried brain anyway.
Good stuff.

[LeftyStrat]

Thanks for this! Very interesting.

[alvarezh]

Thank you Martin for your time and effort in helping us expand our knowledge.

Really appreciate it!

Good weekend.

[Reeltarded]

You have been less active here lately, so I obviously imagine you are out saving the real World one circuit at a time.

Always, thank you..

[brewdude]

Very cool.

Forgive my ignorance, but how do you interpret these curves. That is, what sonic characteristics (descriptions) are assigned to which graphical relationships?

[martin manning]

You can Google or search TAG for descriptions of the sound of fixed vs. cathode bias. From the plots posted above i think you can see that the electrical behavior of a fixed bias arrangement is much more stable than that of cathode bias, at least until you overdrive it and the bias shifts. All of that dynamic content in the cathode biased case, even at moderate signal levels, is more than likely to be the source of the difference in sound.

The other interesting detail is the effect of separate vs. shared and bypassed vs. unbypassed cathode resistors. For the same drive signal level the output is severely restricted with separate, unbypassed resistors. Shared unbypassed is better, but does not produce the power of a shared or separated bypassed arrangement, and those last two are essentially equal.

Lastly, while the static load lines are shown to be generally representative of the anode current vs. voltage relationship, there is much more going on.

[katopan]

Great stuff Martin. Thanks for posting!

[overtone]

Yes, this is really very much appreciated Martin.
Especially the comparison of cathode bias bypassed shared / separate.

I know this is not a request show, but can one model in Spice a shared 470R screen resistor supply (Mullard method for better screen regulation) compared to individual 1K stoppers?
If Spice could do that, then I would start looking into investing time to learn how to use it.

Best, tony

[pdf64]

I'm sure there's some really signification information there which is going over my head, not least because I'm not familiar with seeing it charted like that; how about a bit of discussion and analysis of some, or even just the first chart, to help us dummys get our heads around it?
To me, it looks a bit like a BH chart in magnetics, used to demonstrate hysterysis loss / magnetic saturation; am I on the right track?
What, for instance, should be interpreted from say the first chart?
Pete

[martin manning]

These are plots of anode current vs. Va-k, just like in the tube data sheets but without the Vg1 lines. I drew the static load lines, class A and class B for a 3k4 Za-a, which is what the transformer model simulates. The trace is the "actual" current vs. voltage when a 400 Hz sinusoidal signal is applied for 50 mS. It's as if the time axis is perpendicular to the page, with each loop being one cycle of the input waveform. The trace does not follow the load lines because the load is reactive. Notice also that when the current goes to zero the voltage goes up dramatically due to the inductive kickback. In the cathode biased cases where the Rk is bypassed, there is the effect of the cathode voltage rise with time, which causes the bias point to shift colder. A plot showing cathode voltage vs. time is shown at the bottom of page 2.

Tony, I made some plots showing the effect of shared and separate screen resistors, which I'll post later.

[Colossal]

Killer post Martin!

[jazbo8]

Martin,

This is one of the best posts I have seen in a long while! Lots of good information in those charts... Which version of the SPICE model did you use?

Thanks,
Jaz

[martin manning]

Thanks guys. This was done in LTSpice. One additional comment is that the static load lines adequately indicate the position of the operating point and give a good indication of the area of the anode curves over which the stage will operate. However, getting any more detailed information or accuracy through graphical methods built around static operating conditions would seem to be unlikely.

I've put up a revised .pdf with traces showing the anode current, screen current, and screen voltage, first for the base fixed bias case, and then for a shared screen resistor of 500 ohms (matching the static condition) and also with a shared 1000 ohm resistor to be similar to the case where one tube is in cut-off. The screen voltage is shown to be better regulated in both cases, and the output is increased in the shared 500 ohm case. The sonic effect of a shared screen resistor of that type could easily be tested in practice by shorting the screen pins together.