Question about class AB operation

[Ptron]

I have this fundamental mis-intuition about the nature of class AB operation that I know is wrong, and I've had it as long as I've known what class AB is. I assumed one day I would know enough to get it. That day never came...or it came and I've already forgotten (happens way too much)

This question applies to all class AB amps but it is a little easier for me to picture if I imagine a very hot biased amp. With low enough input signals, neither tube will reach cutoff during a cycle and the amp is operating effectively at class A. Increase the input and each tube is cut off for some portion of the cycle. Intuitively, I feel like the output should drop for this portion of the waveform, and you'd have some sort of class-breakover distortion. Why doesn't this happen? Where does the energy come from to keep the signal going during those portions of the waveform when only one tube is doing the work, same as it was when there were two?

I can accept this much easier in solid state amps, where open loop gain is a bazillion and tons of feedback can force the working transistors to keep up. But that can't be the reason since tube PAs don't have nearly as much gain or feedback, and in some cases, no feedback. What am I missing?

[mooreamps]

You're not missing anything.
People don't take into account their is a varying amplitude of signal that can be seen into the input of the power amp stage. Yes, if the signal is a low enough value, it may not have enough amplitude to drive either tube into cutoff.

Also, bear in mind, those classes of amplifiers, were defined in the early days of radio, where signal levels were {more or less} of a fixed amplitude, not guitar amps.

-g

[Ptron]

I'm not sure I follow you. Are you saying the that the drive to the working tube increases when the other goes into cutoff?

[BackwardsBoB]

I don't think he said that exactly, but yes, in most cases the drive to the power tubes goes up when the push tube goes to cutoff. Class AB amps that can do clean and loud at the same time will have substantial negative feedback from the output to the input of the phase inverter (or maybe a bit earlier), and the phase inverter will increase its drive as a tube goes into cutoff. Cutoff isn't all that sharp, and you can get away with or even desire less than perfection in a tube amp, so you'll see various amounts and filtering on the negative feedback, but it's almost always there on a big head. If it's not enough to compensate for the distortion, you get compression and harmonics that will include some components unrelated to the note being played. It sounds like rock.

A power tube's gain goes up as you move toward saturation (until you get close) and down as you approach cutoff, so there is a bit of a handoff between the tubes even without feedback. The bandwidth of the output transformer helps too, reducing really high-frequency distortion from the handoff.

[Ptron]

Thanks guys. All this time I've been wondering why there wouldn't be any distortion and it turns out there is. I'm really pretty surprised to hear this just because I don't recall ever hearing anyone talk about it, here or otherwise. Does this type of distortion have a name, BTW?

So I guess the gain and feedback of your typical tube power amp, along with the other things Bob mentioned, is enough to fight this. Then again it can't be that bad or AC30s would sound like ass.

[Shea]

Thanks guys. All this time I've been wondering why there wouldn't be any distortion and it turns out there is. I'm really pretty surprised to hear this just because I don't recall ever hearing anyone talk about it, here or otherwise. Does this type of distortion have a name, BTW?

So I guess the gain and feedback of your typical tube power amp, along with the other things Bob mentioned, is enough to fight this. Then again it can't be that bad or AC30s would sound like ass.

The KT in KT66, KT88, and KT90 stands for "kinkless tetrode," because these tubes were designed to (supposedly) eliminate that distortion you're talking about when used in push-pull amps. The distortion looks like a kink in a curve when you plot the power amp's output on a graph. The KT tubes are supposed to eliminate that kink.

So, people are aware of that kind of distortion. I'm not sure what it's called. Maybe it's crossover distortion. I generally think of crossover distortion as something that happens when the bias is too cold, but maybe there's always a little bit there at normal bias conditions too.

Shea

[loudthud]

BackwardsBob nailed it pretty close but there are a couple of things I'd like to add. In the region where both tubes are conducting, the gain is twice as high as it would be with a single tube. As you transistion to the region where only one tube is conducting, the tube that is cutting off is slowly losing gain and the tube that is conducting has increasing gain. Distortion can be minimized by setting the bias to optimize the overlap. You can see this by observing a sine wave and adjusting the bias. You can go from severe crossover distortion to sort of a buldging sine wave. This is the method used when adjusting bias on HiFi amps. Some techs even use a distortion analyzer to adjust bias. In guitar amps the bias is usually adjusted hotter than it would be in a HiFi amp. That's why you see all the plate dissipation calculations. If you just adjusted a guitar amps bias for best sound, it would probably be red plating when you were done.

The scope photo below show a slightly underbiased power amp. You can see the sine wave is just a little too straight in the crossover region. The wave in the middle of the picture is an X-Y plot of input vs. ouput. The horizontal is the input and the vertical is output. The gain is low near the center (more horizontal) and gets higher (more vertical) near the ends. Picure was taken with feedback disconnected.

[Old Tele man]

...Loudthud -- excellent statement and pictorial illustration!

[jrfrond]

BackwardsBob nailed it pretty close but there are a couple of things I'd like to add. In the region where both tubes are conducting, the gain is twice as high as it would be with a single tube. As you transistion to the region where only one tube is conducting, the tube that is cutting off is slowly losing gain and the tube that is conducting has increasing gain. Distortion can be minimized by setting the bias to optimize the overlap. You can see this by observing a sine wave and adjusting the bias. You can go from severe crossover distortion to sort of a buldging sine wave. This is the method used when adjusting bias on HiFi amps. Some techs even use a distortion analyzer to adjust bias. In guitar amps the bias is usually adjusted hotter than it would be in a HiFi amp. That's why you see all the plate dissipation calculations. If you just adjusted a guitar amps bias for best sound, it would probably be red plating when you were done.

The scope photo below show a slightly underbiased power amp. You can see the sine wave is just a little too straight in the crossover region. The wave in the middle of the picture is an X-Y plot of input vs. ouput. The horizontal is the input and the vertical is output. The gain is low near the center (more horizontal) and gets higher (more vertical) near the ends. Picure was taken with feedback disconnected.

At what percentage of output power did you sample the scope trace? If the "10V" seen on the display is your vertical deflection, then I am just guessing that the amp under test is app. 40WRMS into 8 ohms and you are just before the onset of clipping, but like I said, it's a guess.

[loudthud]

As I remember the amp was an early 70's (small logo) 4 input (5F6a derivative) 50W Marshall converted to EL34's. The tubes were nipple top Yugoslavian. The Y-T display is 10V/div about 1kHz. The input was connected to the treble pot wiper with the pot somewhere near center. Scope was a Tektronix 7904. I love those old lab scopes.