Inherent distortion in class AB

[Malcolm Irving]

... it results from crossing over from one tube to the other in a push pull amplifier.....

That's interesting. I see what you mean. I've always thought of cross-over distortion as 'crossing over the zero axis'. (Although, strictly speaking, in this case we are crossing over from both tubes to one tube, rather than one tube to the other.)

[Mike Sulzer]

That's interesting. I see what you mean. I've always thought of cross-over distortion as 'crossing over the zero axis'. (Although, strictly speaking, in this case we are crossing over from both tubes to one tube, rather than one tube to the other.)

True, but if you view the process as going from the one current limit to the other, you just pass through a region with both on. Even if you are trying for class B, I think this is still the case. At what value of grid voltage does the plate current really go to zero? And if you want a practical means of accomplishing the goals of push-pull class B, would you not want to have just enough plate current at idle so that you could set the currents in the two tubes to be the same while meeting the condition that the plate dissipation at idle is very small?

[pdf64]

Yes, it's crossover distortion Jim, but not as we know it.
So there really is a theoretical sweet spot for static current, below OR ABOVE which crossover distortion will increase?

[nickb]

Yes, it's crossover distortion Jim, but not as we know it.
So there really is a theoretical sweet spot for static current, below OR ABOVE which crossover distortion will increase?

I don't know!

Here are the results for a PP 6L6 stage generating about 20 watts. Captured with 40V (58ma), 48V(31mA) and 58V (13mA).

Waveforms - green 40V, blue 48V, red 58V



FFT green 40V, blue 48V, red 58V



Attached is (Is it possible to put attachments inline?)


Audio 40V, then 48v then 58V


Fourier components table

[Malcolm Irving]

...
Here are the results for a PP 6L6 stage generating about 20 watts. Captured with 40V (58ma), 48V(31mA) and 58V (13mA). ...

Sorry, I'm a bit slow this morning. Is the 58mA the idle current, and the 40V the bias voltage?

[nickb]

Sorry, I'm a bit slow this morning. Is the 58mA the idle current, and the 40V the bias voltage?

Exactly. Not as slow as you thought, eh?

[Malcolm Irving]

From the results presented by @nickb in posts #5 and #19, it seems that for a typical class AB amp the 'Gm doubling' is actually reducing the distortion that you would otherwise get from the nonlinearity of the pentode Gm characteristic.
Gm doubling in an ideal pentode would increase the slope of the output voltage wave as it crosses the zero axis. The naturally reduced Gm in a real pentode as it approaches cut-off, reduces the slope. The two effects together can compensate for each other (as pointed out in various posts above).
I'm still curious about what it would look like in an AC30 style amp, which has much warmer bias, relatively speaking.

[nickb]

...

I'm still curious about what it would look like in an AC30 style amp, which has much warmer bias, relatively speaking.

Well, it all depends...

The thing in an AC30 with cathode bias 47 ohm + cap 220uf is the time constant of the bias. Hot and clean at low volumes but the bias cap charges up at higher outputs so the whole thing slides from class A to class AB in a dynamic and, subjectively, ear-amusing fashion.

I'll try to give it a sim tonight - busy right now on the mega tube tester project.

[nickb]

Well, it all depends...

I'll try to give it a sim tonight ...

Using Vs=320 (not 230!!! see post #26) , Vp=370 Zpp=8k:

The bias slides from about 10.5V (==50mA) to 12.7 volts (==26mA) as the output goes from zero to max. The tubes run hot dissipating 18.4W(!) at idle and a much more sensible 6.8W at full blast.

Eye-balling the large signal linearity curve suggests that you wouldn't want to go colder than 14V before crossover becomes a problem.

Incidentally, upping the bias resistor from 47 ohms to 62 ohms gives a using useful reduction in idle power dissipation without affecting the performance, it would seem.

[pdf64]

Using Vs=230, Vp=370

Vs seems a tiny bit low, Vp a lot high?

[nickb]

Vs seems a tiny bit low, Vp a lot high?

Sorry Pete. The Vs is a typo. It should be 320V. Will edit the post.

Vp is 370 at idle according to the schematic. I read somewhere that it's 345 is under load. But, I don't have one to confirm, If you have different numbers, do spill the beanz...

I agree that the Vp seems high based on the max power output & dissipation but I can't just randomly choose a different Vp without good justification.

[R.G.]

The audio math guys have been through this a few times. Fundamentally, the current per input voltage change on a triode or pentode is a power law polynomial involving more than just the zero or first power. You can't do any addition of two of these that sum to a straight line. Close, yes, but not perfect.

The closer the two cutoff points are, the more the inaccuracies in how well the combined characteristic matches a straight line. The further apart they are, the easier it is to hide, as human hearing seems to be especially sensitive to grack right at the zero crossing. Full class A or partial class A hides it pretty well, Class AB makes it not very noticeable, Class B doesn't hide it well, and Class C features it.

It's worse with bipolar transistors as their current/voltage is an exponential law, even further from summing to a straight line.

There isn't any pair of devices that are perfectly complementary for distortion free summing in pushpull that I know of. However, triodes, then pentodes and beam tubes come close.

So (1) yes, push-pull has inherent distortion (2) it varies with bias and (3) different amplifying devices have different amounts of it. Advanced circuits techniques and feedback can make it vanishingly small. Mother Nature seems to insist that it be this way.

[ThermionicScott]

Sorry Pete. The Vs is a typo. It should be 320V. Will edit the post.

Vp is 370 at idle according to the schematic. I read somewhere that it's 345 is under load. But, I don't have one to confirm, If you have different numbers, do spill the beanz...

I agree that the Vp seems high based on the max power output & dissipation but I can't just randomly choose a different Vp without good justification.

NB that's for the late-model AC30 with a solid-state rectifier (I'd hate to be the EL84s in that amp!) The earlier GZ34-rectified AC30 shows 320V at the screen node, but maddeningly doesn't show the plate voltage for any original schematic I've seen. I'd estimate that the original measured for the schematic obviously ran a little more than 320V on the plates, but probably not more than 350 or so.

There's also the possibility that the voltages reflect the earlier 285-0-285 power transformer and weren't updated with the switch to a 280-0-280 PT noted in the sidebar. (Fender was also guilty of this on several occasions.)

[pdf64]

I can't find my old notebook, so have just checked mine (JMI AC30/6 treble with Wodens with AV / MV date codes, so ~early 65?).
The HT is 313 at idle, 306 at max sine (13.8 Vrms@8ohms), 290 at max square (16.8Vrms@8ohms).
The corresponding screen grid node voltages are 301, 285, 257
And cathodes are ~10.5V, 12V, 15.5 (240 ohms per EL84)
@240-241Vac mains
Unfortunately the Vib/trem channel was sacrificed to various mods in my impetuous youth, so it's only running 3 x 12AX7, rather than the full compliment of pre-amp tubes.

[nickb]

TS + pdf64, Thank you both

It's a mess! The '65 shows 280VAC into GZ34. Allow a 20V drop means 260*1.4 =364V
The SS Rect version specifies 370V.
I found this chart which says 387V.



The moral of the story - specify which version we are taking about.

Pete, your measurements are best on the table. Let's go with that since the EL84s will survive at those levels.

(I too, curse your impetuous youth.)