6L6PP 20W amp

[bea]

... this thread is based on that old one: https://music-electronics-forum.com/...ad.php?t=45347



After a lot of spicing i ended up using a pair of 6L6GC in the output stage.
Still using a power supply and OT for an EL84PP stage - with 8 kOhms. Rail voltage as high as possible with the power supply, and fixed bias instead of cathode bias. Idle current relatively small. That means the power stage is closer to B operation than usual.

PI will be a directly coupled cathodyne using an ECC82. Input sensitivity around 2.5-3 Vpp.

Preamp stage is still unclear; i am considering a 2 channel mixer with hiwattish sound characteristics but a lot simpler - two triodes per channel, the 2nd stages possibly forming a mixer through a common/shared anode resistor. Except of a few ideas based on combing experience with older projects there is still no actual design; i first want to get the power stage running.

Spicing indicates a max power of 22W instead of the usual 16W from the EL84. Let's see what the tiny OT will permit in real life.

I'm using the chassis+headshell of a Dynacord amplifier from the 70s which had burnt down.

[bea]

So a screen shot - NFB loop is open.
Yellow: at load resistance
Blue: one leg of the PI through a coupling capacitor.

The resonance is caused by the transformer, isn't it?
I would learn how to compensate for it. Damping on the primary of the PI? Using a a cap in series with the resistor in order not to affect the lower frequencies? How to estimate its values? (i recognize the frequency of the resonance can be read from the screen shots).

And i would like to understand how it is mapped backward into the circuit and whats the reason for the doubled frequency?



and the schematics:

[Chuck H]

I'm confused that the image doubling is in two different colors.?. That implied that the two traces are different pieces of information. Image doubling on a single probe would be all the same color on my scope.

[bea]

These are signals taken at different spots in the circuit. See above, first three lines.

[Chuck H]

Ah. I must have jumped ahead.

The resonance is sometimes called "ringing" and is a common problem. If it's actually a problem. It is a distortion caused by (mild) instability. Believe it or not it's such a common problem that it's considered acceptable in many el84 designs.

I have the same issue in my el84 amps. I've read that it can be from the transformer or due to lead dress. So it's interesting that you're using an el84 transformer with 6l6's and having this issue. More interesting is that it's present at the PI, which (if I haven't misinterpreted things again) is also the 6l6 grid. So...

I'm going to guess that it's actually not present at the PI outputs if you were to remove the 6l6's. I think the reason the blue trace shows the resonance is because it's present on the power tube grid. With the tube clipping it may be attempting to draw a little grid current. So it's output is being mirrored on it's grid.

I might try increasing the value of the grid stops to something like 4.7k or even 10k. It's hard to tell from the image, but it looks like you have the grid stop resistors floating in the middle of the leads to the power tube grids. The grid stop resistors should be snugged up to the grid pin as close as practical with little or no lead between them. The higher value grid stop resistors and more idealized placement of them may be enough to stabilize the output and stop the "ringing".

[Helmholtz]

Blue: one leg of the PI through a coupling capacitor.

Damping on the primary of the PI?

Are you sure that you are speaking of the phase inverter and not the output transformer (OT)? What would be the primary of the PI?

Where do you see frequency doubling?

[bea]

Yes, of corce the iput of the OT, not that fo the PI.


Compare the wavelengths of the oscillation in the blue and the yellow curve. Blue - half wavelength.

And yes, i am aware that the effect is common, not too pronounced, and i know how it is treated on my 4xEL84 amps (Echolette M40). But under certain (avoidable) conditions and with NFB it becomes more pronounced, short before stable oscillation.

And of course i'll re-check my wiring.

[Helmholtz]

Yes, of corce the iput of the OT, not that fo the PI.


Compare the wavelengths of the oscillation in the blue and the yellow curve. Blue - half wavelength.

And yes, i am aware that the effect is common, not too pronounced, and i know how it is treated on my 4xEL84 amps (Echolette M40). But under certain (avoidable) conditions and with NFB it becomes more pronounced, short before stable oscillation.

And of course i'll re-check my wiring.

OK, now I see what you mean (had to take my spex off). There is a ground side ringing at one half of the OT primary of about 3 times the frequency of the other. Obviously it gets compensated by the signal at the other half primary, as it doesn't show at the output.
The output signal shows a ringing (damped oscillation) at about 30kHz. This is a typical leakage resonance of the OT, mainly determined by its primary cpacitance and the leakage inductance between primary and secondary. It will also show as a peak in the frequency response. The peak height is influenced by the OT loading. I.e. by the power tubes' internal plate impedance and the secondary load.
You may study the basic effect by adding a C of around 200pF across the total primary and a small value inductance (start with 100µH) in series with the 8 Ohm load resistor in your simulation.
Without knowing the actual values of your OT a quantitative simulation is not possible.

NFB will even out the frequency peak in the transfer response, but may cause instabilty (because of the gain-phase effects involved). Typical countermeasure is limiting the open loop frequency response by low-pass filtering (many possibilities), dropping the gain above 10Khz or lower.

[Chuck H]

I don't think I've scoped the primary side of my amp that does this so I don't know if the primary/secondary relationship you see is happening on my amp. I hope you get a good explanation for it, but I don't have one.

If what you're asking about is a zobel across the OT primary then I do have some advice. Use a 2kV cap and a 10W resistor that's spec'd for higher voltages. I've roasted a couple of these circuits and the one I'm using now is holding up fine. The usual formula is resistance at 1.4x primary impedance and the capacitance is actually up for grabs. IME the resistance doesn't need to be 1.4x though. It can be anything you want. Whatever gives the most desirable wave form. Figuring the value of the cap for the target frequency should be easy enough, but of course that will change with the resistance too. And the high quality components for this circuit aren't cheap and they're not the sort of stuff you can get at a brick and mortar place, so they have to be ordered. I might start with a 10k, a 12k and a 15k resistor and I'd get caps valued at 1n, 1.5n and 2.2n. Then you'll have some tuning options relative to what you see on the scope.

EDIT: I would still try the larger grid stop resistors right on the sockets first. It's not the fundamental cause of the issue, but will have an affect on it because it trims the HF going in.

[bea]

The resonance is sometimes called "ringing" and is a common problem. If it's actually a problem. It is a distortion caused by (mild) instability. Believe it or not it's such a common problem that it's considered acceptable in many el84 designs.

My motivation for this thread actually was the following picture of a resonance i obtained yesterday - this oscillation occured at a certain and pretty large output level.



With an experimental Zobel at the primary of the OT (10k, 1.1nF = two caps of 2.2nF/630V + 10kOhm/1W in the center) the -3db cutoff occured at 13 kHz. Which is more than i can hear, but a bit low for just the power stage + PI.

So i experimented a bit with the size of the cap across the NFB resistor and found a value where the 1st resonance (~28.5 kHz) was damped out, and the frequency response was nearly perfectly flat (small signal). Still the ringing, but it remains damped well across the full amplitude range --- especially that nasty resonance shown above has gone.

What do You think? Ok, at least for now.


Different topic, although not quite unrelated: the power supply. The two caps of the filter section are a bit large, aren't they?

What about the following:

- estimate "optimal" values from formulae in the literature, e.g. Schröder or Blencowe's power supply book. Which might, after an order of magnitude estimate, be roughly half the size.

- split into branches at the reservoir cap: one for the screens of the 6L6, the other branch for PI and the preamp stages.

- put the bleeder at the end of the filter for the 6L6 screens in order to stabilize the voltage.


Something like this:

[bea]

Just finalizing the project:

i did a fairly complete redesign of the whole amp. Using separate rectifier branches for G2 voltage and preamp / pi voltage in oder to decouple them a bit and to keep the preamp voltages larger. Bleeder is now attached to the G2-stoppers and therefore also acts as a simplistic shunt stabilizer of the G2 voltage. Which results in much less variation of the G2-voltage until short before maximum power is reached.

The bias supply significantly reduced the available rail voltage. Therefore i tried to use a 6 fold cascade after one of the two heater rails. It was a bit tricky to get this hum free, but anyway...

The available power ... actually reduced to 12W. But these 12 W remain clean till the onset of clipping. Cleaner than before.

Here the schematics (still with the old bias supply):



and here the preamp (the last triode is the input stage of the PI above):



The mid control needs a really large input impedance; the 1Meg or 2.2Meg of the typical valves is to small. Fortunately it was possibly to create a grid leak path through the tone filter by bypassing the cap at its potentiometer. This mid control is able to create a mid low very similar to that of a Hiwatt tonestack. It was also necessary to give the 2nd stage some local NFB in order to reduce its output impedance and to avoid bandwidth limitations due to the Miller effect.

Advantage: very clean tone. Disadvantage: even with gain fully on it is not possible to overdrive the stage. It'll just sound "a bit fatter".



Output signal of the power stage at the onset of clipping, no NFB:



And here with the head shell cleaned an a new front plate (plywood with cardboard):



Summary: i did not reach the goal of obtaining more power than with a pair of EL84. But i doubt that even with those tubes it was possible to obtain more than those 12W from these transformers. This simply looks like the limitations imposed by the 72VA mains transformer. But the sound is great and absolutely worth all this effort. And of course i learned a lot.

[Helmholtz]

Your PA schematic shows identical voltages at both sides of R3, meaning zero screen current. How this?
I would expect a difference of 5V or so.

[bea]

That must be an artefact of the simulation: the .op values seem to be computed separately, apparently by some kind of a static approximation.
Possibly this also indicates incompleteness of the tube model - which, however, does predict screen currents (there are a few more of such discrepancies regarding things like modeling the voltage sources, the unavoidable assumption of multiple independent sources without any coupling which is unrealistic for that little transformer close to its limits).

Regarding the transformer: actually i would have needed more information on the PT - measurements under heavy load and translate that into an expression of its internal resistance.

To me that is a situation where either my knowledge of LTSpice is insufficient or there are limitations of the modeling approach. I would guess, both.

But anyway, it is very useful to learn about such limitations.

[Helmholtz]

Possibly this also indicates incompleteness of the tube model - which, however, does predict screen currents (there are a few more of such discrepancies regarding things like modeling the voltage sources, the unavoidable assumption of multiple independent sources without any coupling which is unrealistic for that little transformer close to its limits)

Sorry can'thelp with tube models in LTSpice for lack of experience.

But as the simulation obviuosly doesn't take account of screen currents, the voltage across C2 may be around 50V higher than in the real amp.

Edit: Increasing screen voltage in the real amp to around 300V (by omitting R26 and decreasing R2) should give more output power.

[bea]

It is not "the simulation". It is the .op analysis, which regards idle state. And indeed, i am also frustrated not to see any voltage drop across the G2 stoppers - while i do see the effect of the G2 currents in the voltage drop of R2.

Consistently, not only in this simulation.