Class AB2 Power Amp: Bringing 2nd Harmonics Back to Push-Pull

Jerry garrcia

Tele-Holic
Joined
Nov 14, 2021
Posts
563
Age
48
Location
Sweden
What's the deal with the 6N7s? Obviously I've missed something. Isn't that the tube Fender used for the paraphase in the late 40s tweeds before switching to the 6SC7? As recall, they can be a bit microphonic.
Gibson did also use them in one of the EH-150/185 models as the PI.
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
@andrewRneumann it seems that the voltage of a capacitor coupled supply would vary significantly under load, so a long as the load doesn't change much, it might work.

Playing around in LTspice, I came up with a full wave version using two 470n caps, each with a 33k resistor to ground and a 1N4007 diode. I wasn't sure that a full wave version would work, but so far I don't see any reason that it wouldn't. One potential pitfall is the charge time for all of the caps in the circuit.

You know way more about that bias circuit than me. What kind of voltage does it put out and can you adjust the voltage without wasting a lot of power? How much current that circuit needs to supply depends where you bias the cathode followers. There will be some AC signal appearing because AC does flow through the cathode follower cathode resistors. I took pains to make sure no feedback signal made its way back to the grids of the CFs. There is the filter cap there, plus each bias for the CFs is filtered again. I also had to use a choke because half-wave rectification was just too noisy.

Getting the voltages and resistances just right on the CFs is the trick to this circuit.

Was this these little beauties that started this amazing thread?

No. That tube has its own thread here https://www.tdpri.com/threads/5w-class-a2-concept.1105833/

What's the deal with the 6N7s? Obviously I've missed something. Isn't that the tube Fender used for the paraphase in the late 40s tweeds before switching to the 6SC7? As recall, they can be a bit microphonic.

It seemed like a cheap, odd-ball tube no one knew what to use for. People kept talking about it and it became sort of a joke about using it. I just thought it would be fun to design an amp around them for the fun of it. I have never seen or heard one personally myself. See link above.
 

Phrygian77

Poster Extraordinaire
Gold Supporter
Joined
Apr 30, 2016
Posts
6,188
Location
Crawfordville, FL
@andrewRneumann I hadn't either until I worked a 1947 Dual Professional a few years ago. Someone had stuck a Sovtek in there that was labeled 6N7GT, but my research on that tube and the Sovtek lead me to believe that it wasn't like a real 6N7 (trying to weed out all of the idiotic internet HiFi discussions about 6N7s vs 6SN7s). I went ahead and ordered a NOS RCA 6N7GT for the amp. The owner and I did some A/B testing and he concluded that he liked the Sovtek better. So, it was pretty much wasted time and effort.
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
@Phrygian77 I just looked at Merlin’s full-wave bias supply for center tapped PT’s. My main concern with this particular supply in my design is stability of bias in preventing runaway on the CF’s. Merlin’s design needs a fairly large and robust dropping resistor to get the voltage down in the range I would need (my target is around -65VDC at 4mA). I estimated I would need a 25K/5W resistor burning off about 2W at idle just to produce about -65V. This dropping resistor means that if current increases in the CF’s, the bias supply voltage will also drop (less negative), which will lead to more current which will lead to more bias voltage loss and so on—basically run away conditions. So I’m very cautious about trying to implement one of Merlin’s modern designs. I could be wrong though—this supplies bias voltage and cathode voltage for the CF’s, so if they both drop simultaneously, maybe there’s no risk of run away. Dang… now maybe I will have to try it…

Would the capacitor coupled version suffer from the same potential problem?

It just so happens that the 50V winding on these transformers gives me a nice low impedance supply for negative voltage right in the range I needed it. Yes I am stuck with half-wave rectification, a small choke, and some DC in the PT, but all I can say is that it’s working and seems to be working well.

And before anyone ( @King Fan ) thinks I am somehow responsible for something new, this AB2 design is simply an implementation of what Aiken wrote about here in his article on blocking distortion:

  • Add a DC-coupled cathode follower between the phase inverter and the grid of the output tubes, with the cathode follower cathode resistor returned to a high negative voltage, and the grid bias applied to the grid of the cathode follower. This effectively isolates the output tube grid circuit from the phase inverter and its associated AC coupling, and provides a very low impedance source for the output stage. This will prevent the output stage from going into grid clamp, and will eliminate the long time constant of the AC coupling. This method has the unfortunate side effect of requiring an extra tube and completely ruining the value of your vintage amp, so it is best used only on new designs, but is highly recommended. You will also get more power out of the output stage because it is now running in class AB2 or class A2 (the "2" suffix indicates grid current flows for a portion of the cycle).

I assume he writes this and highly recommends it because someone else has done it. Maybe I’m just the dude who put pots in series with the grids to control power tube headroom and distortion character.
 

Phrygian77

Poster Extraordinaire
Gold Supporter
Joined
Apr 30, 2016
Posts
6,188
Location
Crawfordville, FL
@andrewRneumann try it out in LTspice. I think it would take a lot of tweaking to get a cap coupled bias supply right. I didn't even get around to modeling the choke, the rest of the voltage dividers, and filtering. It could take several seconds, from power on, for it to get up to speed. The charging path includes the main HT load, so before the tubes warm up, the first filter has to charge up through a combination of (on your schematic) RV7, R33, R34, and R35. That's a total of about 150k to ground (although I'm not sure what the regulator may contribute).
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
They should both therefore avoid blocking due to grid excursion. Does this also mean they can operate in AB2?

That is a keen question. The answer is "no, not necessarily." Just because there is a CF and DC coupling, does not mean it operates in AB2. Aiken is a little misleading in this regard. Take the SVT as an example (I chose that one because I've looked at it before... haven't looked at the Dumble yet... thank you for the examples). The SVT has a 47K resistor in the grid circuit of each power tube. This is going to kill most grid current right in its tracks. So the point of CFs in the SVT was not to operate in AB2, but to avoid coupling capacitors and the attendant blocking distortion. That is my knee-jerk analysis just by looking at a schematic--no math was done.

The Fender super twin also uses cathode followers to drive its 6x 6L6's.

It does, but the 6L6s are still AC coupled with 220nF caps.

In this case, I would think the cathode follower is required because with 3x 6L6s per side the grid leak resistor is only 33K! You can't drive a 33K load with a typical LTP without a huge reduction in signal level. So the CF is able to drive the full signal across a 33K load... but I would think this amp would suffer horribly from blocking distortion once grid current limiting was reached. This amp, like the SVT, does not operate in AB2. (The first few signal cycles that reach the grid current limit would push it into AB2, but the coupling cap quickly discharges after a few cycles, the bias shifts, and the circuit stays in AB1--with a huge bias shift and most likely major blocking distortion.)

So, on to the Dumble Steel String Singer to see if it is truly operating in AB2! Does anyone have a schematic of that? I'm having trouble finding one.
 

sds1

Friend of Leo's
Joined
May 4, 2017
Posts
2,229
Location
Orlando, FL, USA
So, on to the Dumble Steel String Singer to see if it is truly operating in AB2! Does anyone have a schematic of that? I'm having trouble finding one.
Here's what I have on file (relevant bits only):

1661257150947.png
 

FenderLover

Poster Extraordinaire
Joined
Jun 11, 2009
Posts
6,351
Location
Minnesota
I think I would use a separate bipolar supply on the followers and set the bias on the follower grids (?) A separate bipolar supply is another chunk of iron, but small, and easy to rectify and filter on standoffs.
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
Here's what I have on file (relevant bits only):

So that's interesting. I feel somewhat vindicated in my design choices by someone of his stature arriving at a similar solution. For instance, he uses the same supply voltage for the CFs as the screens. That's exactly what I ended up with.

Still, there are questions... why does he still include the 470K grid leak resistors? I did not feel they were necessary and it makes the analysis of the CF circuit difficult. The grids can leak through the CF cathode resistors and bias supply. But maybe I'm wrong here... either way, I have had no problem with tubes running away because I don't use the standard grid leak path.

Secondly, he uses a 12AX7 for the driving tube. These can't supply the necessary current to drive deep into AB2. They reach grid current limiting at only a few mA. To truly determine the current limitations of the CF, I would need to know the voltages.

Do you know what the "E", "F" and "H" voltages are? Anyone know what the power tube bias voltage was supposed to be after all the dust settles?
 

sds1

Friend of Leo's
Joined
May 4, 2017
Posts
2,229
Location
Orlando, FL, USA
Do you know what the "E", "F" and "H" voltages are? Anyone know what the power tube bias voltage was supposed to be after all the dust settles?
I don't have any info on F/bias, but this design uses a Twin PT and H is just on the other side of the rectifier so say 450VDC, and E is just a choke away from that so say 445VDC. A layout I have here says H and E are 454 and 453, respectively.
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
According to the ClassicTone and Hammond datasheets: 2kΩ

Ok, looks like we have enough data for me to make a little analysis of the Dumble SSS. Here's the load line:

1661287691390.png


We can estimate the bias to be -48V, but that's just a guess. Somewhere between -45V and -51V is likely.

In the screenshot, I've left my mouse over the Vgk = 0 characteristic--the point at which Class AB2 is assumed to begin. You can see that in this case, the load line is striking just below the knee, which is the classic setup and really no surprise. Does the amp operate in class AB2? It will certainly suck down as much grid current as the poor little 12AX7 can supply, but the whole time the 6L6 will be in saturation. In other words, the grid current does not contribute to any more audio power. The cathode can provide no more electrons to the plate at plate voltages that low, regardless of how positive the grid goes.

And this is why Dumble doesn't need a 12BH7 driving the power tubes... it doesn't rely on grid current to do its thing (i.e. it's not deep into Class AB2). Like the SVT, the CFs and DC coupling eliminate the coupling capacitor and blocking distortion. There might be a tiny bit of AB2 power being generated if the screen sags--but I doubt it would be much with 12AX7s doing the driving. Either the 6L6s reach saturation or the 12AX7s hit grid current limits before too much AB2 action occurs.
 
Last edited:

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
I don't have any info on F/bias

Looks like the supply is a half-wave rectifier attached to HT. Would you guess F is -450V?

Man, if driver plates are at 450V and the 6L6 bias is -50V, that’s 500Vak on the 12AX7s. I have never seen anything like that before. In cut-off, Vak would be 900V! Is this amp known to blow up the driver tube? Yikes! Maybe my assumption about F is way off.
 
Last edited:

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
And here’s an analysis of the Dumble SSS V6 12AX7 driver tubes.

B045B450-3ED8-4B02-AC13-2B348F9C99A1.jpeg


1. V+ is 860V. That’s -410V (the bias voltage “F”) pulling down the cathode and 450V pushing down on the plate.
2. Bias point is 500Vak.
3. In normal operation the voltage swings from 450 to 550Vak. Power tubes in Class AB1.

You can see that the tube cannot supply much grid current to the 6L6s without overdissipation. My initial analysis assumed the current limit would be set by the grid current in the driver tubes, but now that I see the huge voltages Dumble used, the main limitation is plate dissipation.
 




Top