Push/Pull Pair Removal Theory & Questions

[nickb]

A recovering shredder.... but I've been know to fall off the wagon from time to time

There is a similarity. I've been known to fall off the stage from time to time.

[nickb]

No, it's not.

We all know that single-ended load lines can't be used to model the behavior of a push-pull pair.

It is not a single ended load line. I do show only one of the tubes to save space as the other half would provide no more infomation.

[Malcolm Irving]

... the other half would provide no more infomation.

The composite load line does give a little bit more information. If it is plotted accurately it shows the result of combining the push and pull outputs during the class A phase. Usually that's pretty close to a straight line though.

[bob p]

Admittedly, Class A is not all that common in MI amps.

[Mike Sulzer]

It would seem that you need a whole set of those graphs of plate current versus plate voltage, that is, as a function of screen voltage, which does vary over a sine wave cycle. Taken together they form a three dimensional function. The load line then is a path through the resulting "3D graph". I think you have to use spice, something I have not done. Are good power pentode models available, and if so where do you get them?

[SoulFetish]

Could you not trace curves with a 1K screen stopper? Or is there build in current limiting in the utracer which would affect results? Also how much does screen voltage sag under peak signal conditions? It might stil be possible to plot curves using the formula I posted here:
JJ6V6

[nickb]

It would seem that you need a whole set of those graphs of plate current versus plate voltage, that is, as a function of screen voltage, which does vary over a sine wave cycle. Taken together they form a three dimensional function. The load line then is a path through the resulting "3D graph". I think you have to use spice, something I have not done. Are good power pentode models available, and if so where do you get them?

Ta Da:




Not bad agreement since it use a different tube model and the graphs earlier were just eyeballed.

[Dave H]

So that's saying pulling two tubes without changing the load impedance is neither 25% nor 50% power but actually 35% power? I've just measured around 30% on my 4 x 6V6 cathode biased amp so it's closer to 25% than 50% for that one.

[Chuck H]

So that's saying pulling two tubes without changing the load impedance is neither 25% nor 50% power but actually 35% power? I've just measured around 30% on my 4 x 6V6 cathode biased amp so it's closer to 25% than 50% for that one.

You need to account for other operating conditions in a cathode biased amp because they are even more self regulating as bias is affected by current.

[Dave H]

You need to account for other operating conditions in a cathode biased amp because they are even more self regulating as bias is affected by current.

Yes, true, that's why I added "for that one" at the end. I could probably make it closer to 50% by changing the operating conditions.

[Mike Sulzer]

Ta Da:

[ATTACH=CONFIG]47426[/ATTACH]


Not bad agreement since it use a different tube model and the graph earlier were just eyeballed.

That was fast! Thanks. (I waited until I got into work this morning to comment so I could display the plot on a better monitor; that blue on black is murder on the eyes.) So,yes, the sg swing is substantial. Also, the transformer current waveform is significantly non sinusoidal, a bit flatter at the extremes, but also less curved down from the extremes with just a hint of a==>b transition. I suppose those kinds of distortion have a role in making the loud, nearly clean, sound in an output stage as opposed to the triode sound of the lower level stages.

[Mike Sulzer]

Could you not trace curves with a 1K screen stopper? Or is there build in current limiting in the utracer which would affect results? Also how much does screen voltage sag under peak signal conditions? It might stil be possible to plot curves using the formula I posted here:
JJ6V6

Sorry I did not reply earlier.. I think you can use a tracer for this.The valve wizard does it here: The Valve Wizard for the EF86. I think it has to work for power pentodes as well; a built in current limiter should not affect the results with the resistor if it does not without the resistor. If you have a tracer, great. I was thinking more people have spice than tracers these days.

The thing I do not understand about the VW's text that goes along with the tracer plots is that he says that you use the curves with zero resistance for setting the operating point because the curves with the resistor only apply to dynamic conditions when a signal is present. I do not seen that could be true.

[Malcolm Irving]

...
The thing I do not understand about the VW's text that goes along with the tracer plots is that he says that you use the curves with zero resistance for setting the operating point because the curves with the resistor only apply to dynamic conditions when a signal is present. I do not seen that could be true.

I agree. The screen current as a function of plate voltage and plate current is a 'static' characteristic. If that current passes through a screen stopper resistance it will cause a screen voltage drop and the plate curves will be compressed, whether a signal is present or not.

[nickb]

Could you not trace curves with a 1K screen stopper? Or is there build in current limiting in the utracer which would affect results? Also how much does screen voltage sag under peak signal conditions? It might stil be possible to plot curves using the formula I posted here:
JJ6V6

The current limit on my modified uTracer is 400mA. I don't think it's too much of a problem as we only require accuracy in the region of the the load line. If you then factor in the tolerance of tubes...

As far as estimating the intermediate screen currents and plate curves the approach I use is to take the data and create a spice model from it. In my version of the GUI this is built-in although I still have work to getting the initial estimates better, for some tubes it fails miserably.

Derk Reefman has collated quite decent library of models created this way, I think it's in the ExtractModel (what is what my model extraction is based on) download package The uTracer, a miniature Tube Curve Tracer / Tester.

[Steve A.]

The best advice I heard to reduce the loudness of a Twin Reverb was to pull one pair of 6L6s (inside or outside) while disconnecting one of the speakers.

As with any advice like that you take your chances... if the B+ is so close to a critical level that pulling 2 tubes pushes it over the top you are on your own.

Steve A.