Cathode Follower Push Pull

[daskew81]

Hi everyone,
Happy holidays! I was messing around on the bench last week trying to come up with a good use for some 4:1, 10k/600 OPT's I have laying around. In a standard push-pull configuration these really aren't the right transformers to be using for any off the shelf triodes, but I had the idea of trying an AC coupled cathode follower in push pull and this is what I ended up with. I didn't include the AC caps in the schematic.

According to the limited information I've found on the internet about cathode follower push pulls, I've read that they are incredibly difficult to implement. However, this sounds and performs fantastically, which is why I think something is really wrong, and this is just a big accident and I'm not seeing a glaringly obvious mistake here.

I don't know why it works, but it does. The numbers all make sense voltage wise. I've included the voltage measurements at different points. Measuring current from my ammeter, from CT to 47k tail resistor, I measure 4mA. From tail resistor to ground I measure 12mA. What doesn't exactly make sense to me is why this is working.

I can't seem to wrap my head around creating a load line for this as it's kind of a mix of a reactive and resistive load...

1. I don't have much experience with push-pull stages, but I understand them and how they work, but bench testing them, I simply don't have the experience. One general question about any push-pull valve stage, if I probe one of the OPT primaries, would I actually see the signal cut off on my scope during the transition from class A to class B if it was operating in class AB? On this circuit, I can send 20v peak-peak signals into the grids, scope the cathodes, and never see a cutoff of any sort even with the grid bias at -0.5v. There isn't any visible crossover distortion on the output making me think this might be running in class A push-pull? I have a scope probe that measures current as well, should I be using that instead?

2. Frequency response is flat with -3db points at 9hz and about 50khz no matter the secondary load. Which is again why I think there is something wrong. There is only .005% THD+n unloaded, and don't get above 1% unless I really load the secondary down with something crazy like 130ohms, which ends up giving me -6db, which I'm also guessing loosely assumes the output impedance? (I know in reality it's more complex than this)

3. If I probe any of the grids with my 10M probe, the output signal completely collapses into an undesirable mess. Almost like the signal on one side is grounded out. What's up with that? I've never had that happen on any other valve stage before. Maybe the input impedance is super out of whack? I haven't tried any other value of grid leak resistors, but that will probably be next on the list.

4. As to the fair frequency response, I had thought maybe my 50ohm signal gen was perhaps doing all the heavy lifting here, so I built a few different PI circuits before this to feed it, with more of a high impedance output, but it didn't change the performance at all. This made me think either A, something is really wrong here, or B, I just kind of whipped out something that works great.

4. No doubt the DCR's are creating the negative grid bias, that math works out. But at such a low bias voltage I'd assume I'd see some sort of signal cut-off on the cathodes?

Anyways, I'd be curious to what everyone thinks is going on here. I'm humbly confused, and humbly not ashamed to admit when I've overlooked something really dumb here.

 

[daskew81]

Quick update, the plot thickens. I'm doing something wrong for sure. If I remove B+ but keep the filament voltage, I still get full signal. If I remove filament voltage, the signal starts going away as the filaments cool down. Reconnecting filament voltage and the full signal comes back even with no B+. These are NOT elevated filaments, which I realize I'm pushing the limits with. Perhaps I should rebuild with a fixed bias supply and elevated AC filaments and test again...

 

[Tubetec]

It may take time for the HT voltage to decay depending on the size of the filter caps .
The figures your putting forward seem reasonable to me,
What does the gain come out as?

Your most likely near the max heater cathode voltage rating on the ECC82 , the elevated heater supply sounds like a good plan.
0.5v drop on the grid of an ECC82 doesnt sound right , try placing an additional 470 ohms in series with the center tap from the TX ,
a 22 or 33 Kohms tail resistor might proportion the voltages slighty better across the anode/cathode , two of the 47k 3w in parallel would do .

The issue with the cathode follower is trying to come up with the drive voltages especially considering the loss of gain in the transformer .
You can also ground out the second grid via cap if you need unbalanced connection .

 

[ruffrecords]

Small bias voltage values do not matter in a cathode follower because the cathode literally follows the grid signal so the bias voltage stays nearly constant. However, looking at the curves you have 80V plate to cathode and 2mA plate current which should put the bias near 2V. I never try to read dc voltages directly on the grid because even very high impedance meters can upset the bias. I always measure it from the cathode bias resistor.

2mA bias and 80V plate to cathode seems rather low for a 12AU7/ECC82 and will limit your headroom if drving 600 ohm loads.

Cheers

Ian

 

[merlin]

Reminds me of this old snippet:

 

[Walter66]

I've studied this theme of classical tube amp schematics since some years.
Sometimes, one comes about such a piece of strange circuit that someone has put effort in to create, beside all the standard circuits.

Those are like black sheeps on the pasture, you spot them, think about them, they may work but nobody uses them, they can't be spotted in any serious textbook. So how to deal with them, when they may have no real favors, otherwise they would be part of the relevant literature in this field?
PP tube output stages have always been done with audio transformers, as it is the case here.
The only favor I see is the fact, that with a cathode output, the transformer could be driven more ease and the amplification of the circuit is much lower.

 

[solkatten]

I have not tried PP cathode follower, but SE cathode follower with 6n6p tube and LL1671SE trafo as OPT and cathode resistance Its 2:1 great freq response and low distorsion

 

[abbey road d enfer]

I didn't include the AC caps in the schematic.

Do you mean the grid caps?

According to the limited information I've found on the internet about cathode follower push pulls, I've read that they are incredibly difficult to implement.

Probably because people expect too much. The fact that a cathode follower has a low output impedance is taken by many as an increase drive capability. This is not true. The drive capabilty at the cathode is the same as that at the anode (for identical B+ and quiescent current. Using a low impedance xfmr for a cathode load often results in less output because of the current limitation of the tube.
And actually, the output capability is limited by the low quiescent current.

However it seems, according to your other posts, that something else is fishy.

 

[abbey road d enfer]

PP tube output stages have always been done with audio transformers, as it is the case here.

No. Check Series Regulated Push Pull.

The only favor I see is the fact, that with a cathode output, the transformer could be driven more ease

This is very debatable. The only notable advantage is that, since it's driven from a low impedance, the transformer is more damped, so its response is flatter.

and the amplification of the circuit is much lower.

Which may or may not be an advantage.

 

[Walter66]

The symbol indicates for a PP output audio trans. But what is the extra winding for, is this the serial regulation? What purpose has it and why it's needed in this circuit? Or is it just the field coil of the speaker?

 

[emrr]

Well we should maybe call it 'classic' PP, as in merlin's post above, as that's what we seem to be talking about here.

More on SRPP:

https://www.tubecad.com/may2000/

 

[solkatten]

I think in a preamp context, a cathode follower is not that hard to drive. In context of power amps this can be a different story.

In a mic preamp you pretty much trade gain for low THD and freq response.

And as Abbey pointed, if you really want big output it could better to connect a high ratio transformer to the anode and rely on the tubes great voltage swinging capabilitys, than rather poor current capabilitys.

The biggest benefit here is you take a pretty mediocer transformer and make it perform like a very expansive (anode loaded) Lundahl or Jensen. Due to lower impedance and/or lower turns ratio.

 

[merlin]

But what is the extra winding for?

It's just an old fashioned way of drawing the speaker voice coil.

 

[abbey road d enfer]

The symbol indicates for a PP output audio trans. But what is the extra winding for, is this the serial regulation? What purpose has it and why it's needed in this circuit? Or is it just the field coil of the speaker?

It seems there is an ambiguity there.Both windings are on a common core.
Field coil, if any, is on a separate core.

 

[solkatten]

Maybe I will try the PP CF in my 4in1 pre it could be arranged like this:

 

[abbey road d enfer]

Should work...

 

[merlin]

Maybe I will try the PP CF in my 4in1 pre it could be arranged like this:

What's the purpose of the shared 5k resistor?

 

[solkatten]

I was under the impression that tail resistor provided better balance in a PP stage, but maybe thats just true for smith inverter stage with unbalanced input?

Otherwise I go for a dc-balance pot instead

 

[merlin]

I was under the impression that tail resistor provided better balance in a PP stage

It does, but I doubt you need better balance here (your PI is perfectly balanced already). The CF output stage is very limited in power, and that resistor is stealing some power, but I haven't done the sums so maybe you're not losing anything important.

 

[solkatten]

I get your point, balance is a non- issue with CF.