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

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
Hello All,

I’ve been slowly working on a design of my own based on some discussions I had when I finished my last amp, the MC10. I became intrigued with Class 2 operation and the possibilities opened by grid current. I wanted to design a Class AB2 power amp that utilized 6V6s and had output similar to a Deluxe Reverb. I also wanted to avoid the use of transistors if at all possible. I dug around in old books and magazines and found that there were Class AB2 designs before the advent of solid state devices. Could I figure out how to adapt them to the guitar world? The answer is YES!

The general specifications for this amp are:
2x 6V6S in Push-Pull
1x 12BH7 driver
1x 12AT7 long-tailed pair phase inverter

B+ 425V
Maximum Clean Output: 23.8W

What’s the big deal? That’s about the same as a Deluxe Reverb. What is gained by pushing into Class AB2? Because grid current is flowing to get those 24 watts, we can adjust the series resistance of the grid circuit and control the amount grid current. I accomplished this by a single 25KL pot in series with each 6V6 control grid. When the pots are at 0, the tube is permitted to run up into Class AB2 as far as it can. (I estimate that I can maintain voltage on the grid all the way up to Vgk +20V! That’s PLUS 20V.) When the pots are set at 25K, the control grid loses voltage at 0Vgk and operation is limited to Class AB1. The best part is that the player gets to choose the output of each power tube individually with a pot. Crank one up and one down for plenty of second harmonics. Crank both down for limited headroom and lots of third harmonics. Or, choose anything in between. The player is in harmonic control of the power amp. In full Class AB2, the amp puts out about 23.8W of clean power. Crank both “class” controls down and the output power is limited to only 4.8W of clean power. Yes, that’s an extra 19W of power achieved by driving into Class AB2.

How is this all achieved? I’m attaching the schematic as built, so you can take a look. Some differences from the standard push-pull configuration:
  • There is a 12BH7 cathode follower between the LTP and the power tube. These 12BH7s provide the grid current necessary to drive into Class AB2. The 12BH7s are direct-coupled to the 6V6s—providing another benefit of no bias shifts and blocking distortion in that stage.
  • Individual bias controls for each side of the power amp. The bias controls control the driver current which in turn control the bias of the power tube. Essentially the driver tube and power tube become one unit with one bias control.
  • 50V bias winding used for 12BH7s but also for setting up the correct cathode voltage on the 12BH7s.
  • MOSFET voltage follower used to achieve lower than normal screen voltage and cathode follower plate voltage. I tried to find a way to avoid solid state devices in this design, but there was no feasible way to drop enough voltage, avoid excessive sag, and keep from wasting tons of power into heat. The screens and cathode follower plates run at about 225V. (B+ is 425V, so that’s 200V of drop that needs to happen.) At least the signal chain stays completely tube.
  • Type-II PPIMV
Here are some scope images.
First image, maximum clean power into an 8-Ohm dummy load. 13.8Vrms = 23.8W.
1660425654400.jpeg


Second image, one power tube limited to Class AB1. Other tube still in Class AB2.
IMG_0866(1).JPEG


Third image, both power tubes limited to Class AB1.
1660425911237.jpeg


Forth image, maximum clean power limited to Class AB1. 6.2Vrms = 4.8W.
IMG_0868(3).JPEG


Finally, some images of the different wave-forms achievable by the player. These were all created with the same input signal and gain levels. The only thing I'm changing in these are the "class" controls which are just adjusting the control grid resistance on the power tubes.
IMG_0870(3).JPEG
IMG_0871(1).JPEG

IMG_0872(2).JPEG
IMG_0873(3).JPEG


This thing is a real screamer and bone shaker when hooked up to my Screaming Poseidon pre-amp. But with 24W of clean power available, beautiful, rich, loud clean tones also available. Just add a touch of uneven grid resistance with the “class” pots and the complexity just deepens. I really like this amp. I will attach some audio once I get a descent mic up and can shut down the A/C in my house for awhile.

So what’s next? This amp is built on an open test rig. It needs to be married up permanently with a pre-amp and put into a spacious head cabinet. I’ve documented everything as built, so it may be time to dismantle the test subject and start layout on a real chassis. In the mean time, that will free up my test bed for the 6N7 Class A2 amp that I will be building and testing. It follows the same general principles, but the power supply will be quite a bit different.
 

Attachments

  • As Built 0.1.pdf
    156.2 KB · Views: 118
  • IMG_0869(3).JPEG
    IMG_0869(3).JPEG
    215.8 KB · Views: 15

Bendyha

Friend of Leo's
Silver Supporter
Joined
Mar 26, 2014
Posts
3,419
Location
Northern Germany
Wow! Nice work. Fascinating stuff, I'm going to have to look into this more. I was aware of its possibility, but would not have considered building a guitar amp that way. I got some rereading to do.
Thanks for the impulse.
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
Wow! Nice work. Fascinating stuff, I'm going to have to look into this more. I was aware of its possibility, but would not have considered building a guitar amp that way. I got some rereading to do.
Thanks for the impulse.

Thank you!

I struggled the most with the screen supply on this. I actually ran the amp down at really low screen voltages—like less than 100V. That was interesting because the amp is almost all Class 2 all the time. I just felt bad not using the power available. It’s got a PPIMV so it’s not like it can’t get quiet.

I also had the voltage regulator for the screen working with an 0A2 tube for awhile. That was fun, but ultimately wasteful and I decided to go with the MOSFET.

Lot’s of learning. And still lots to improve on this design, but I’m pretty happy with where it’s sitting.
 
Last edited:

printer2

Poster Extraordinaire
Joined
May 24, 2010
Posts
8,780
Location
Canada
Class AB2 helps drive the tube when there is limited voltage on the plates. But that is thrown out the window with guitar amps running at voltages way over spec. I am jealous at your playing around.
 

NTC

Tele-Afflicted
Joined
Mar 17, 2003
Posts
1,374
What happens if you raise the screen voltage? How high can the screen voltage be set before tubes want to burn up? I assume this also would require a lower bias voltage as well.

There is a very early 2x 6L6 (not GC) class AB2 design in RCA RC-14 that is supposed to be 55 watts, though it uses a coupling transformer driven by a 6F6. 360V on the 6L6 plates, 270V on the screen grids, and -22.5V on the control grids at rest.
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
Class AB2 helps drive the tube when there is limited voltage on the plates. But that is thrown out the window with guitar amps running at voltages way over spec. I am jealous at your playing around.

Yes, we are not limited in suitable tubes or voltage. No doubt, the 6V6 works great in AB1. Triodes benefit the most from AB2, because they have no screen to keep the current flowing when the plate voltage drops. That said, I created the conditions I thought would be a good balance between AB1 and AB2 with 6V6s. I don't know how long they will last but I have been playing it loud on and off for a few months with no problems yet. Also running it at 24W through a resistive load for long periods at a time. Here is the loading:

1660501185394.png


Everything above the Vg=0V is AB2. I'm using JJ 14W tubes, so the max diss. line is a little deceiving (it is showing 12W).

What happens if you raise the screen voltage? How high can the screen voltage be set before tubes want to burn up? I assume this also would require a lower bias voltage as well.

A couple things happen. One is that the bias voltage has to be set more negative. So that's more voltage headroom in AB1, which I actually didn't want to have. I purposely wanted this to operate mostly in AB2. The second thing is that the loadline will hit below the knee. I tried this and felt unsure the screen and grid electrodes could handle the high amounts of current this would generate. So I'm stuck hitting above the knee in a fairly safe area. (I found that I couldn't drive more than 20V positive grid voltage before the clipping the power tubes.)

Load line above the knee keeps voltages pretty stable, which might feel a little "stiff" for some players liking. A dynamic sag resistor is included in the power supply which supplies the screen AND the cathode followers (just after the MOSFET follower). So even though the screen current doesn't spike (because I'm hitting above the knee), I can still sag the screen because the cathode follower current DOES spike in AB2. So the cathode follower current spikes, the sag resistor sags the voltage on the CF and the screen simultaneously. Right now I'm only using 200 Ohms of sag resistance, but it could be increased to suit anyone's fancy. (I learned this when I initially tried to set this circuit up with a standard dropping resistor supplying the screens, and then another dropper supplying the CFs. I thought all the sag was from the screen, but then I realized it was the CF's that were sucking down all the current and sagging the voltage on the screens.)

I haven't burned up the 6V6s yet! I assume the control grids will be the first to go since they are taking a beating. I installed fuses in each 6V6 cathode circuit to protect in case of massive short--something I saw people doing in the HiFi world.
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
There is a very early 2x 6L6 (not GC) class AB2 design in RCA RC-14 that is supposed to be 55 watts, though it uses a coupling transformer driven by a 6F6. 360V on the 6L6 plates, 270V on the screen grids, and -22.5V on the control grids at rest.

I found my inspiration for a design with no coupling transformer in an old magazine article. My mind is foggy--I should have saved it somewhere. I think it was referenced in RDH4 and then I was able to find it on worldradiohistory.com . That is an AMAZING site... they must have a million magazine over 100 years... all SEARCHABLE.
 

printer2

Poster Extraordinaire
Joined
May 24, 2010
Posts
8,780
Location
Canada
I found my inspiration for a design with no coupling transformer in an old magazine article. My mind is foggy--I should have saved it somewhere. I think it was referenced in RDH4 and then I was able to find it on worldradiohistory.com . That is an AMAZING site... they must have a million magazine over 100 years... all SEARCHABLE.
And down the rabbit hole we go...
 

peteb

Poster Extraordinaire
Joined
Apr 25, 2003
Posts
5,286
Location
Cascadia
Interesting project!

I Find this interesting because I have considered the bias as a practical upper limit to signal level. At least fender amps appear to be designed this way. I have done a lot of signal measuring and the measuring bears that out.

I drove a champ into class A2. my impression was that nothing really happened when the signal surpassed the bias, except it had to be really cranked and the pick attack is over the top.

your recording is a nice recording with nice playing. It sounds low volume. At that volume I would expect the amp to still be in class AB1.


have you measured the signal to confirm that the signal is greater than the bias?


I would be interested in seeing measured voltages.

great thread.


please fill us in.

I am sure many would be interested in


plate voltage
screen voltage
bias voltage
signal voltage
plate current
plate dissipation

maybe it is possible to measure current thru the control grid.
 
Last edited:

printer2

Poster Extraordinaire
Joined
May 24, 2010
Posts
8,780
Location
Canada
Interesting project!

I Find this interesting because I have considered the bias as a practical upper limit to signal level. At least fender amps appear to be designed this way. I have done a lot of signal measuring and the measuring bears that out.

I drove a champ into class A2. my impression was that nothing really happened when the signal surpassed the bias, except it had to be really cranked and the pick attack is over the top.

your recording is a nice recording with nice playing. It sounds low volume. At that volume I would expect the amp to still be in class AB1.


have you measured the signal to confirm that the signal is greater than the bias?


I would be interested in seeing measured voltages.

great thread.


please fill us in.

I am sure many would be interested in


plate voltage
screen voltage
bias voltage
signal voltage
plate current
plate dissipation

maybe it is possible to measure current thru the control grid.
What circuit did you use to drive the grid positive?
 

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
Here is a scope shot of the one of the 6V6 control grids at max clean signal.

1660527330179.jpeg


Max positive swing is 20V, the negative swing is obviously cutting the tube off well before the grid reaches -63.2V. The flat top on the bottom is because the cathode follower doing the driving is cutting off. You don’t hear this because the power tube itself cuts off before the cathode follower cuts off. I guess you could say that cathode followers are operating in Class AB1.

Did some more measuring. I estimate the peak control grid current at +20Vgk is about 9mA and 3mArms over the cycle. I believe if I raised the screen voltage these numbers would get higher. (Control grid current spikes when plate voltage drops—just like the screen.) I do have about 780 ohms of built in grid resistance too, so that’s limiting it somewhat.
 
Last edited:

printer2

Poster Extraordinaire
Joined
May 24, 2010
Posts
8,780
Location
Canada

Volcanicash01

Tele-Meister
Joined
Apr 8, 2022
Posts
130
Age
79
Location
Mayenne France.
Hello All,

I’ve been slowly working on a design of my own based on some discussions I had when I finished my last amp, the MC10. I became intrigued with Class 2 operation and the possibilities opened by grid current. I wanted to design a Class AB2 power amp that utilized 6V6s and had output similar to a Deluxe Reverb. I also wanted to avoid the use of transistors if at all possible. I dug around in old books and magazines and found that there were Class AB2 designs before the advent of solid state devices. Could I figure out how to adapt them to the guitar world? The answer is YES!

The general specifications for this amp are:
2x 6V6S in Push-Pull
1x 12BH7 driver
1x 12AT7 long-tailed pair phase inverter

B+ 425V
Maximum Clean Output: 23.8W

What’s the big deal? That’s about the same as a Deluxe Reverb. What is gained by pushing into Class AB2? Because grid current is flowing to get those 24 watts, we can adjust the series resistance of the grid circuit and control the amount grid current. I accomplished this by a single 25KL pot in series with each 6V6 control grid. When the pots are at 0, the tube is permitted to run up into Class AB2 as far as it can. (I estimate that I can maintain voltage on the grid all the way up to Vgk +20V! That’s PLUS 20V.) When the pots are set at 25K, the control grid loses voltage at 0Vgk and operation is limited to Class AB1. The best part is that the player gets to choose the output of each power tube individually with a pot. Crank one up and one down for plenty of second harmonics. Crank both down for limited headroom and lots of third harmonics. Or, choose anything in between. The player is in harmonic control of the power amp. In full Class AB2, the amp puts out about 23.8W of clean power. Crank both “class” controls down and the output power is limited to only 4.8W of clean power. Yes, that’s an extra 19W of power achieved by driving into Class AB2.

How is this all achieved? I’m attaching the schematic as built, so you can take a look. Some differences from the standard push-pull configuration:
  • There is a 12BH7 cathode follower between the LTP and the power tube. These 12BH7s provide the grid current necessary to drive into Class AB2. The 12BH7s are direct-coupled to the 6V6s—providing another benefit of no bias shifts and blocking distortion in that stage.
  • Individual bias controls for each side of the power amp. The bias controls control the driver current which in turn control the bias of the power tube. Essentially the driver tube and power tube become one unit with one bias control.
  • 50V bias winding used for 12BH7s but also for setting up the correct cathode voltage on the 12BH7s.
  • MOSFET voltage follower used to achieve lower than normal screen voltage and cathode follower plate voltage. I tried to find a way to avoid solid state devices in this design, but there was no feasible way to drop enough voltage, avoid excessive sag, and keep from wasting tons of power into heat. The screens and cathode follower plates run at about 225V. (B+ is 425V, so that’s 200V of drop that needs to happen.) At least the signal chain stays completely tube.
  • Type-II PPIMV
Here are some scope images.
First image, maximum clean power into an 8-Ohm dummy load. 13.8Vrms = 23.8W.
View attachment 1016647

Second image, one power tube limited to Class AB1. Other tube still in Class AB2.
View attachment 1016650

Third image, both power tubes limited to Class AB1.
View attachment 1016661

Forth image, maximum clean power limited to Class AB1. 6.2Vrms = 4.8W.
View attachment 1016667

Finally, some images of the different wave-forms achievable by the player. These were all created with the same input signal and gain levels. The only thing I'm changing in these are the "class" controls which are just adjusting the control grid resistance on the power tubes.
View attachment 1016664 View attachment 1016668
View attachment 1016669 View attachment 1016670

This thing is a real screamer and bone shaker when hooked up to my Screaming Poseidon pre-amp. But with 24W of clean power available, beautiful, rich, loud clean tones also available. Just add a touch of uneven grid resistance with the “class” pots and the complexity just deepens. I really like this amp. I will attach some audio once I get a descent mic up and can shut down the A/C in my house for awhile.

So what’s next? This amp is built on an open test rig. It needs to be married up permanently with a pre-amp and put into a spacious head cabinet. I’ve documented everything as built, so it may be time to dismantle the test subject and start layout on a real chassis. In the mean time, that will free up my test bed for the 6N7 Class A2 amp that I will be building and testing. It follows the same general principles, but the power supply will be quite a bit different.
Fascinating stuff mate! When I did my electronics apprenticeship in the Royal Navy many years ago, it was all thermionic emission. I loved building amps and using a scope in the workshops Tube theory is fascinating. Glad to know that guys like you are out there! Mind you, if an amp distorted, that was an exam fail!
 

Attachments

  • 1A5E184B-F8DE-4E33-B7A6-42A31890222D.jpeg
    1A5E184B-F8DE-4E33-B7A6-42A31890222D.jpeg
    138.3 KB · Views: 18
  • 95A74ACE-66BE-402B-B6FE-040F8B83ED8F.jpeg
    95A74ACE-66BE-402B-B6FE-040F8B83ED8F.jpeg
    145.3 KB · Views: 20
  • 788DBE39-987D-42F5-8169-07AA5C8BBF91.jpeg
    788DBE39-987D-42F5-8169-07AA5C8BBF91.jpeg
    158 KB · Views: 20
  • 9106B194-5216-4205-A341-54BD4CFB974A.jpeg
    9106B194-5216-4205-A341-54BD4CFB974A.jpeg
    162.4 KB · Views: 21
  • 20939AD5-5FAE-42EF-8FFE-0403457230DD.jpeg
    20939AD5-5FAE-42EF-8FFE-0403457230DD.jpeg
    148.8 KB · Views: 18

andrewRneumann

Friend of Leo's
Joined
Mar 22, 2020
Posts
2,153
Location
Cincinnati, OH, USA
Sorry, I meant with the library of magazines to dig into, finding other material taking you off on a tangent, four hours later remembering what it was you were looking at in the first place. The schematic looks wrong with the cathode resistor on the grid of the output tube.

I knew what you meant. I just was excited that I found the original schematic that inspired me. And yes, I still don't understand how they got the power tubes to bias correctly. But it did show me there was a way to do this without solid state (in the signal chain at least) and without a coupling transformer. The key was having a enough negative voltage to pull some of the current through the cathode follower instead of push it all through plate.
 




Top