Dumble Push-Pull power amps: what really matters?

[martin manning]

EDIT: comparing them now I would have expected the plate resistors to be more lopsided toward the in-phase side of the PI in the Bluesmaster due to the 9k1 tail as opposed to the 24k tail in the regular OTS above.

But in any case the PI trimmer is there to adjust the balance.

[roberto]

Here's a couple of layouts.
…due to the 9k1 tail as opposed to the 24k tail in the regular OTS above. Also interesting is that they both attempt to eek out a bit more gain than the typical 82k/100k you see in most Fenders and Marshalls.

Thanks! I will search the schematics of those layouts.
I see the 9k vs 24k more of a balance (even harmonics) and bias sp working point. Same for plates, because 100k vs 110k in parallel with 220k doesn’t make that much difference.

Indeed my choice would be to make bias current to be chosen independently, that also brings more dynamics on the ltpi.

[roberto]

I recall numerous comments on TAG saying that ODS with Twin Reverb OT (single 4 ohm secondary) were often operated using an 8 ohm speaker load.

Thanks martin!
Flattish load lines makes the amp intrinsically more linear, and Zout lower before applying any feedback. 8k Raa for a pair of 6L6GC seems way too low IMHO, especially with g2 at around plate voltages. I had in mind something around 6k6 per pair of tubes, with lower g2 to see where the load line hits the knee of the pentode.

[martin manning]

8k Raa for a pair of 6L6GC seems way too low IMHO, especially with g2 at around plate voltages.

I agree. I don't do that with mine.

[roberto]

But in any case the PI trimmer is there to adjust the balance.

but is it generally set to be exactly balanced, or to introduce some asymmetry?

[martin manning]

Since the "Ampeg" procedure as described on the SVT schematic is prescribed, it is balanced, as for Hi-Fi. See in the Dumble files section: https://ampgarage.com/forum/viewtopic.php?f=4&t=38853
There is another long thread stickied, but the above was an attempt to start fresh.

[dbharris]

Has anyone ever tried different load lines (even simply by mismatching a 16 Ohm cab in a 8 Ohm output, with or without modifying the nfb)?

I recall numerous comments on TAG saying that ODS with Twin Reverb OT (single 4 ohm secondary) were often operated using an 8 ohm speaker load.

This was often the case and is well documented in known amps and the Dumble book. But he did not always utilize such a mismatch.

I run my ODS "clone," "replica," "tribute" amps or whatever you want to call them with an 8 ohm load on 4 ohm OT tap. My Wonderland I have run both ways and the mismatch feels a little more compressed in a nice way. Since I play at home for my own pleasure, I have no concerns about fly back voltage at loud volumes.

It would be interesting to simulate the musicman or "surge" power amp alongside the more traditional one.

-Dan

[roberto]

Thank you again martin,

So it seems that the choice I had in mind to use a ccs as the tail of the ltpi can be a good starting point towards symmetry of the ltpi, with the bonus of having the grids of the ltpi grounded, so one cap less and more dynamics in the ltpi.

The flattish load line of the output tubes should help in the other two factors of linearity and control of the speaker (before applying nfb). On top of that, it permits to have lower g2 voltages, so less voltage swing to drive the output tubes to full power.

Indeed I had in mind to apply nfb from the primary side of the output transformer. One option then could be to apply positive current feeeback from secondary to the other grid of the ltpi, or a blend of it with the distortion of the amp.

[roberto]

Hi Dan, thanks for your feedback.
So you oftern play your amp in “jtm45-lamd” in terma of loadline. It would be interesting to try different voltages on g2 (lowering it by 50V each time while modifying bias voltage to reach the same plate dissipation) and target the sound tou prefer with that very flat wa line.

[Fischerman]

Same for plates, because 100k vs 110k in parallel with 220k doesn’t make that much difference.

It's really about double that increase. The 82k is increased to 100k and the 100k is increased to 110k+10k pot. And since the 82k was increased so much you'll almost always have to crank that 10k pot to yield 120k for that plate for it to balance. It still might not be enough with a 9k1 tail.

It kind of doesn't even look like balance is the goal with the BM PI anyway.

[dbharris]

Hi Dan, thanks for your feedback.
So you oftern play your amp in “jtm45-lamd” in terma of loadline. It would be interesting to try different voltages on g2 (lowering it by 50V each time while modifying bias voltage to reach the same plate dissipation) and target the sound tou prefer with that very flat wa line.

I believe you are correct, but my EE knowledge is very superficial compared to most here. That sounds like an interesting experiment and perhaps one day I will do that. For now, I am on a building hiatus while we have a newborn at home. If I find time before the end of the year to swap a few pots and tweak the trimmers in my latest build, I will be happy. But it has been great seeing more activity on the forum lately. And reading all of the threads is keeping me engaged. Very curious to see and hear what circuit you end up implementing.

-Dan

[roberto]

It's really about double that increase. The 82k is increased to 100k and the 100k is increased to 110k+10k pot.

Thanks, in any case with the 220k in parallel the gain should be 49x vs 54x, that’s less than 1 dB. It’s more a working point imho.

[roberto]

Very curious to see and hear what circuit you end up implementing.

Hi Dan, I have something in mind I will propose here in the next days. There are few options I need to check.

[nuke]

So towards hi-fi land, without going that far. The experience I did on LPTI and power amps in that field can come handy, as with the SE version.


I would like to try two approaches, one with 12ax7 and one with ecc88 or similar (so andriver with more or less capanility to drive the output tubes).

Has anyone ever tried different load lines (even simply by mismatching a 16 Ohm cab in a 8 Ohm output, with or without modifying the nfb)?

Thanks,
Roberto

In the conventional capacitor-coupled PI-to-output grids circuit topology, such as the ODS and most Fenders, what is really reaching the grid of the 6L6GC or whatever output tube is mostly voltage. So, you can just ignore the current, because there really isn't much there and we're not driving the grids into grid-current, so it is class AB1. You get into the Steel String Singer, or SVT and some others, where the output grids are being driven with a DC coupled stage that is capable of driving the grids into grid current, then it gets to class AB2 operation.

The PI stage drives into a load that consists of the coupling capacitors, the resistors, that do grid-leak or apply fixed bias to the output tubes, and the load presented by the grids themselves, and all the parasitic capacitance inductance, Miller effect and so on.

The tube selected for the PI is sort of immaterial. As long as it can make the grid voltage swing necessary to drive the output pairs to full swing, or to some desired level of clipping, and it can deal with the effects of the secondary components and parasitcs, all you gotta do is operate the the PI tube within its operating range and they'll pretty much sound the same. That's not saying you just swap the tube itself, but adjust the PI's bias resistor, tail resistor, the plate load resistors and supply voltage for the selected tube. They don't sound a lot different actually when you do that, provided they're capable of supply the necessary drive to the next stage.

Add in the negative feedback and the overall PI gain, and the feedback is adjusted by the resistor/caps in the loop and what tap on the OPT secondary, set the feedback to match the gain, they're going sound more or less similar to each other.

[Roe]

Here's a couple of layouts.
…due to the 9k1 tail as opposed to the 24k tail in the regular OTS above. Also interesting is that they both attempt to eek out a bit more gain than the typical 82k/100k you see in most Fenders and Marshalls.

Thanks! I will search the schematics of those layouts.
I see the 9k vs 24k more of a balance (even harmonics) and bias sp working point. Same for plates, because 100k vs 110k in parallel with 220k doesn’t make that much difference.

Indeed my choice would be to make bias current to be chosen independently, that also brings more dynamics on the ltpi.

The tail resistance affect balance yes as well as operating voltages. Due to the voltage drop across the tail resistor, 24k effectively lower the voltages that the PI tubes sees (compared to 9k1), unless you raise the supply voltage to compensate. Lower voltages gives a softer attack with less punch.