Class B amplifier with SINAD of 120, how is that possible?

[DonH56]

Thanks Don, I kind of knew much what you posted. The thread and many posts seem to point to a true high power low noise/distortion class B is a unicorn, and Topping has mischaracterized what they are selling. So my current vision of class B is that without some initial class A pure class B mode is rarely occurs and historically so are stand alone class B circuits. I'm a ME and audio enthusiast and not an EE, so a lot of this is over my head.

I have no idea what Topping is doing. There are tricks that can be played to achieve low distortion with class B circuits, like certain using additional feedforward compensation or ancillary paths ("little amplifiers") to "bootstrap" the main output stage during the transition from one side to the other, having some lag in the circuit so the feedback stays continuous during the transition, and so forth. RF circuits, where I see most class B designs, are generally tuned thus have reactive "tank" circuits to keep everything happy during transitions and often the high-power outputs need limited dynamic range so higher distortion is not a problem. The huge dynamic range issue for RF is usually the receiver; the transmitters are often big brutes to belt out power with 40~60 dB THD being just fine.

 

[pma]

There is another issue with class B definition. Douglas Self in his books on power amplifier distortion defines class B as optimally biased class AB. Not underbiased and not overbiased, just biased to get lowest distortion in the class AB output stage, depending on Re resistor value an voltage drop across this resistor. This makes even more confusion in semantic.
And it is nothing more than semantic. Semantic to be used by marketing departments and other product promoters.

 

[LKA]

With heavily under biased OPS (1-2 mA per output pair, no emitter resistors) it's no problem to get THD1 -120dB. For low N you have to have low gain.

 

[pma]

With heavily under biased OPS (1-2 mA per output pair, no emitter resistors) it's no problem to get THD1 -120dB. For low N you have to have low gain.

That is correct. So, 1-2mA you are saying. Then NFB can correct it easily. I was under impression that you use much lower idle current, far below 100uA. Or is it just Danhard who used zero bias?

 

[dguillor]

Class D is like crossover distortion to the max? With feedback it behaves like "oh i need to raise the output node voltage BRGHUAGHAGH ok I've done enough imma go sleep now"

My train of thought may be weird but I'm always thinking about the possibility of switched linear-region or PAM output stages for either performance or efficiency improvements.

You don’t understand how class d works.

 

[Cbdb2]

I have an interesting Cambridge Azur 840A v2 integrated here. Given to me for parts.

It too has a "patented" amplifier innovation and they call Class XD. This was Cambridge's first amplifier to incorporate this circuit. A Douglas Self design element I believe.

View attachment 392899

Class XD Explained | Cambridge Audio US

It sure sounds fancy, but what exactly is Class XD?

www.cambridgeaudio.com

Patent appears to have lapsed, so I've attached some details for general interest. Including their white paper below (PDF)

View attachment 392896

Power amp stage schematic attached below. Notice the XD circuit can be switched on and off, like a VW defeat device. LOL.

View attachment 392900

Similar to the old opamp trick of putting a resistor between the output and a rail to force it into class A.

 

[Cbdb2]

It would have been great to hear what Scott Wurcer had to say.

 

[pma]

There is a hint, just hint, how to do it even with zero bias and zero idle current.

1) push-pull OP stage and its distortion, no correction



2) corrected by NFB circuit

amplitude and phase

As one can see, THD is down from 3% of the OP stage to 0.0007% with NFB correcting circuit.

 

[LKA]

That is correct. So, 1-2mA you are saying. Then NFB can correct it easily. I was under impression that you use much lower idle current, far below 100uA. Or is it just Danhard who used zero bias?

I use few mA for better high freq thd (above 5kHz)
Danhard usually uses one Vbe drop between bases, so not zero.

 

[rongon]

There is another issue with class B definition. Douglas Self in his books on power amplifier distortion defines class B as optimally biased class AB. Not underbiased and not overbiased, just biased to get lowest distortion in the class AB output stage, depending on Re resistor value an voltage drop across this resistor. This makes even more confusion in semantic.
And it is nothing more than semantic. Semantic to be used by marketing departments and other product promoters.

Which book and on what page, please? Thanks.
That's a unique definition, and not understood as such by most people who work with this stuff.
Semantic confusion indeed.

 

[pma]

Which book and on what page, please? Thanks.

I hope you do not want for me to go through a set of his books. Several examples below.



Please feel free to do your own literature research .

 

[rongon]

Thanks for the snippets.

What's in those quotes reinforces what I've already posted, but they do not support your claim that "Douglas Self in his books on power amplifier distortion defines class B as optimally biased class AB. Not underbiased and not overbiased, just biased to get lowest distortion in the class AB output stage, depending on Re resistor value an voltage drop across this resistor."

It looks like those of us who have worked with vacuum tubes have a slightly different definition of the term "class B" than those who have only ever worked with transistors. In this article (https://www.electronics-tutorials.ws/amplifier/amp_6.html) the author states, "...what is commonly termed as a Class B Amplifier, also known as a push-pull amplifier configuration.

"Push-pull amplifiers use two “complementary” or matching transistors, one being an NPN-type and the other being a PNP-type with both power transistors receiving the same input signal together that is equal in magnitude, but in opposite phase to each other. This results in one transistor only amplifying one half or 180o of the input waveform cycle while the other transistor amplifies the other half or remaining 180o of the input waveform cycle with the resulting “two-halves” being put back together again at the output terminal.

"Then the conduction angle for this type of amplifier circuit is only 180o or 50% of the input signal. This pushing and pulling effect of the alternating half cycles by the transistors gives this type of circuit its amusing “push-pull” name, but are more generally known as the Class B Amplifier"

No wonder everyone is confused. This author implies that any push-pull amplifier is by definition a 'Class B Amplifier'.

In the meantime, you'll find in the classic texts from the vacuum tube era:

(from the Radiotron Designers Handbook, 4th Edition)

So I'm going to stop talking about this amplifier class stuff in this thread. I learned this stuff playing with tube output stages, in which it's possible to bias the output device so that its grid (analogous to the gate of a transistor) is drawing significant current (which I think is not possible with transistors). Meanwhile, in the solid state world, the definition of class B has been stretched so wide that it includes just about any push-pull transistor output stage.

So I give up. I agree in general with DonH56's definitions and I don't care to belabor these points any more. (I hear people out there exclaiming "FINALLY! Hallelujah. Give it up already, would ya?")

 

[DonH56]

Push-pull long ago morphed into meaning two devices operating 180 degrees out of phase, whether complementary as common in modern SS amplifiers, two tubes (or transistors) at opposite ends of the transformer windings driven out of phase, or quasi-complementary circuits common in the early transistor days when a high-power "complementary" PNP was created by a combination PNP/NPN pair since high-power PNPs were uncommon back then.

Plate or collector current, my definition of the amplifier classes was from a college class 40+ years ago, and I'm too hold and obstinate to change.

Probably part of the confusion lies in that most class B circuits are actually B2 with a little trickle current, but not all.

 

[AnalogSteph]

Don, this amplifier (about 1971), is class B or not?

(BTW, R313 was a design nonsense. Omitting it distortion goes 1 - 2 orders down).

Technically, this kind of zero bias circuit that's traditionally been called "Class B" is actually class C. If run open-loop, each half of the output stage conducts less than 50% of the time. With feedback, the rest of the circuit has to slew like mad at each zero crossing as there is a dead zone of over 1 V to be bridged every time before the output stage starts conducting again. You can imagine what this does to HF distortion. You can make the problem progressively smaller by replacing the direct tie between both output transistor bases by one silicon diode or one silicon diode plus a Schottky, respectively - all while maintaining zero output stage bias.

The concept of "gm dead-zone filler resistor" R313 is fundamentally sound in that it does work in eliminating the excessive crossover distortion at low levels (output impedance only spiking to 22 ohms beats near infinity any day of the week), but it obviously puts a burden on the driver stage by limiting output stage current gain, so performance at higher levels will be degraded.

Thank goodness we finally figured out how to do proper thermally-compensated biasing. I strongly suspect most designers implementing zero bias output stages merely wanted to keep their amps from blowing up from thermal runaway.

The LM358 opamp is a famous example of a circuit with a dead zone output stage, a choice which its designer has publicly regretted later in life. Its crossover glitches (plainly seen on a scope) are legendary... as you might have guessed, it's not exactly slewing very fast either.

@wwenze is on the right track... if you want to properly linearize a dead zone output stage, you need a rather Class-D-like approach with enough (PWM) HF bias to dither around the dead zone and higher-order loop compensation. This may enable a hybrid class of amplifier more efficient than a Class AB but with less problematic output filters than a pure Class D (and wider bandwidth, too). Whether we even need still that given how good Class D has become with some tricks is another matter, but it's food for thought.

 

[DonH56]

Technically, this kind of zero bias circuit that's traditionally been called "Class B" is actually class C. If run open-loop, each half of the output stage conducts less than 50% of the time. With feedback, the rest of the circuit has to slew like mad at each zero crossing as there is a dead zone of over 1 V to be bridged every time before the output stage starts conducting again. You can imagine what this does to HF distortion. You can make the problem progressively smaller by replacing the direct tie between both output transistor bases by one silicon diode or one silicon diode plus a Schottky, respectively - all while maintaining zero output stage bias.

The concept of "gm dead-zone filler resistor" R313 is fundamentally sound in that it does work in eliminating the excessive crossover distortion at low levels (output impedance only spiking to 22 ohms beats near infinity any day of the week), but it obviously puts a burden on the driver stage by limiting output stage current gain, so performance at higher levels will be degraded.

Curious... Did you find it actually limited the output impedance to the value of the extra resistor? Ages ago I tried using R313 in a similar design but found it didn't help all that much. During the interval both output followers were off, impedance still spiked, despite the boot-strapping effect. I went to a feedforward circuit but found I needed a fair amount of driver power and it was better to implement a simple sliding bias scheme to keep the output followers on.

Thank goodness we finally figured out how to do proper thermally-compensated biasing. I strongly suspect most designers implementing zero bias output stages merely wanted to keep their amps from blowing up from thermal runaway.

The LM358 opamp is a famous example of a circuit with a dead zone output stage, a choice which its designer has publicly regretted later in life. Its crossover glitches (plainly seen on a scope) are legendary... as you might have guessed, it's not exactly slewing very fast either.

Widlar's comments on it were pretty funny. Not sure I have seen them in print, especially since, well, Widlar.

@wwenze is on the right track... if you want to properly linearize a dead zone output stage, you need a rather Class-D-like approach with enough (PWM) HF bias to dither around the dead zone and higher-order loop compensation. This may enable a hybrid class of amplifier more efficient than a Class AB but with less problematic output filters than a pure Class D (and wider bandwidth, too). Whether we even need still that given how good Class D has become with some tricks is another matter, but it's food for thought.

Modern self-oscillating designs are able to include the output filter in the feedback loop so that's pretty much a nonissue, at least in the audio band.

 

[wwenze]

To deal with a transistor with 2 regions of operation, you need a design with 2 regions of operation. (Or you parallel the second region with another transistor in the "on" region and pretend the second region doesn't exist, until people come around telling you class AB has crossover distortion)

Is that what Topping is talking about? By finding a way around the crossover distortion they allowed a bigger overlapping range of biasing, so they can really bias things in class B (or C) and let feedforward take care of the rest.

 

[DanielT]

Regarding the technical details, I just follow, as best I can, what y'all are discussing. Interesting and I'm trying to learn more.

BUT what I'm wondering, which as far as I can see has not been answered or discussed is why Topping B100 chose the design they did? IF it is Topping's intention to be class leading in the SINAD war why choose a class B design? Given that it's class B, which, to be honest, I don't really understand if it is. Or maybe it is, like a little, in a way (that's how I interpret the answers with my limited knowledge).

Like this. If you were to design an amplifier with the power that Topping B100 has with the same low distortion, would you choose a class B design? Is it the easiest, most cost-effective way forward?
Note you cannot choose class A because a criterion, which I now add, would be a physical size and weight similar to that of Topping B100. Maybe that also excludes class AB? So the choice then is between class B and class D, right?

Excuse my word salad, I don't really know what I'm talking about, but I can always ask.

 

[JSmith]

why choose a class B design?

I think it's just because they can... they've done D, AB, so now B. They seem to like covering all bases in the market.


JSmith

 

[wwenze]

We don't know what Topping did, so we don't know why they did what they did.

 

[ryanmh1]

I'm sure it's been mentioned, but just look up what Douglas Self calls "class B" in his handbook. It's just an optimized bias point which actually involves very little bias current. On further reflection, that's probably all they mean and they call it Class B because Douglas Self says they can since he wrote the book on amplifiers.