EL84 PP Class B power stage

[bea]

Hi everyone,

meanwhile i am trying to learn a bit how to work with spice.

The project shown is based on a kit which i never got to work correctly (well, meanwhile it should, but since that build i'd like to avoid cathode biased PP stages with a common cathode resistor ... guess what happened ;-)).

Here the initial circuitry:



And here my ideas.
You might imagine something inspired by a Hiwatt preamp stage or the preamp i designed for my G2000, maybe as a low power bass amp for rehearsals and sessions in the final run.





In my eyes that circuitry is ready to be built. The output stage is closely modeled to the output stage of the Echolette M40, which is an ancient PA amp with 4 EL84 in class B.

Independent on that i would like to come a bit closer to a more realistic modeling of the stability of the feedback loop. And that requires better tube models and even more important more realistic models of the output transformer.

Another aspect important to me is the modeling (and limiting...) of the grid currents on the screen grids in the case of tube saturation...

In a further step i would like to extend everything to the case of EL34 PP class B at 750 V (i have the parts lying around and could also measure and more important test the (old) transformer set for sanity...)


Any suggestions, mostly on the transformer measurements? I have the transformers at hand and could do measurements to the degree my limited equipment permits (cheap function generator, oscilloscope, two multimeters).

(BTW: and which tube models would simulate these special aspects at least half way realistic? At present i am using Koren based models modified by someone else with data from µ-Trace measurements, which seem to be pretty good in the LoV regime.

[Dave H]

No takers? I'll give it a bump by asking a question. In the original circuit why have a separate cathode resistor/capacitor for each EL84 and then connect the cathodes together so it looks like a common resistor/capacitor?

[bea]

I do not know. Really.

The two resistors in parallel are obviously necessary to obtain the desired value. But then it would have been no effort to use separate resistors.

The kit is built on a printed circuit and its layout makes it impossible to split the resistors. When i did the 2nd circuit i always kept in mind that it could be implemented by just changing a few parts on the original board and, of course, adding a supply for the G1 voltage.

In the simulation i obtain almost 25% more power from the 2nd approach, even with better linearity.

[J M Fahey]

With due respect, I´m not sure what´s left to design after 50 or 60 years existence of very popular EL84, meaning there have already been hundreds of attempts.

Options are fixed/cathode bias, standard/ultralinear OT, various load impedances, one of 4 Phase Inverters, more/less NFB including no NFB, triode/pentode connection.
Did I miss any variation? .... maybe .... but that gives us a grand total of, what? , 30/40 combinations?

Like on Mc Donald´s , there´s only just so many different sandwiches possible combining the old same 7 ingredients.

Now, if you just want to learn and practice Simulation, perfect, try all combinations, vary parameters, and then test build real glass and metal amplifiers, then measure them and check how close do they behave compared to simulations.

Besides being useful by itself, I bet it will be a real eye opener

[bea]

Actually i do not consider this a "design" - i am fully aware that i can chose between several options. The kit in its original form was already an eye opener. But anyway - if i do not set it up in UL mode (which is possible) i find it problematic that the screen voltage is or at least may become larger than the grid voltage and that in overdriven state the screen currents may become too large.

And i am interested in how the (rarely implemented) directly coupled kathodyn stage behaves.

As far as i can tell, the 2nd version is ready to be implemented in glass, and i will do this in the near future, probably even before i do the 6L6 mod of my Valve Junior (the byproduct of that idea, the optimisation of the existing EL84 amp has already been done).

My intention here in this thread is more about spice:

- which pentode models are "best"? (modified) Koren? Ayumi?
- how do i model at least the onset of clipping?
- can i model the stability? Which directly leads to...
- What about the output transformer? Having its high frequency resonance right is important for stability estimates in the model...
- How can i improve its model which at present is just a hopefully intelligent guess? Which properties would i need to measure for this?



As mentioned already - this is also a pre study to the larger project of doing an amp with 90W from 2 EL34 using the rests of an already existing Amp (a Dynacord HiFi Favorit 2). Again with the main focus on details like the PI stage, the screen grid currents of the power stage and transformer measurements (i need to check if the transformers had survived mistreatment in the past...)