Forced Class A opamp

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
By all means do your experiments and measurements using op amps that were designed more recently than 40+ years ago. Be warned that it is hard to resolve noise and distortion as low as the modern uber- op amps deliver using a typical test bench.
==
Thanks for the clarification Winston, I was scratching my head about how to bias a class B output stage to be lower distortion than class AB. That said modern IC designers have access to tricks we discrete designers can only dream about.

JR
 
By all means do your experiments and measurements using op amps that were designed more recently than 40+ years ago. Be warned that it is hard to resolve noise and distortion as low as the modern uber- op amps deliver using a typical test bench.
==
Thanks for the clarification Winston, I was scratching my head about how to bias a class B output stage to be lower distortion than class AB. That said modern IC designers have access to tricks we discrete designers can only dream about.

JR
Fair point; let me rephrase my conclusions, then: with this test setup, I was able to detect a difference in performance in the older TL071, not in the newer parts.
Tommaso
 
In theory class B is fine, in reality semiconductors do not turn on and off instantaneously, so the transition from negative to positive gets a little messy while one doesn't want to turn off right away, and the other is slow to turn on. Class AB keeps both devices turned on while passing through 0V so we trade a little efficiency loss for much lower distortion.

Inside ICs they can do tricks I can only imagine, and modern processes are much faster which helps at the relatively slow audio rate of change.

JR
 
Yep that makes sense John.
I cited B. Oliver's paper earlier but, in reality, it isn't as simple as his 26mV/I theory would predict.
In reality, thermal effects/drift are only one area that can play havoc whereby all bets are off as far as zeroing in on optimum class B bias. And life is too short for anyone to be trimming things after a few hours of use.

As you said earlier, it's incredibly hard to resolve distortion below what's achieved with modern devices. About the only area of discrete vs IC I've done OK is wrt noise in that we're free to use a beefier (lower Rbb) input pair for low source Z's and and/or a higher current output drive for lower feedback impedance.

Other than that, I dunno.

Edit: One area that I've found interesting in terms of distortion measurements is regarding the loading on IC op amps. While some are benign down to even 600 ohms, some are downright ugly below a couple of K. All sorts of higher order products start peeking up out of the floor. .
 
Last edited:
What I'm saying thus far is: a perfectly biased, ideal world, class B output stage would result in lower distortion than a class AB biased output stage. No Gm doubling and all that other crud and jazz. D.
Self has a rundown of this in his amplifier book and my own, much cruder experiments support this.
Class B is not a working reality, it's a concept. As John says, nothing goes from zero to something with discontinuity.
Now, I don't know the actual tricks that IC designers have utilised in recent times to eliminate output stage distortions, but a pure class B would certainly maintain a constant as far as distortion vs load, level etc.
IC mfgrs do not publish internal schemo anymore, but I assume they use a derived form of the class A+AB+B output stages that were the de facto standard when power amps were analog.
 
Class B is not a working reality, it's a concept. As John says, nothing goes from zero to something with discontinuity.

Well, yes of course. But I think we're (I'm) maybe not defining what is meant by AB or B in the same way as each other.


I'm defining class B as a properly biased output stage (such as adheres to requirement that external RE be equal the internal re' by Oliver's definition), simply the mode of operation that 99% of analogue amplifiers in the world are biased to.

And AB as the back up plan that allows a class A amplifier to work reasonably well if presented with a lower impedance external load.
One can reasonably argue that AB is not even a proper class in itself, but simply a combination of A and B.
 
Well, yes of course. But I think we're (I'm) maybe not defining what is meant by AB or B in the same way as each other.
My definition is class A both devices conduct complementarily all the time, AB one of the devices does not pass current for a fraction of the time, B is each device conducts strictly alternatively (which is not a real possibility).
I'm defining class B as a properly biased output stage (such as adheres to requirement that external RE be equal the internal re' by Oliver's definition), simply the mode of operation that 99% of analogue amplifiers in the world are biased to.
I'm not familiar with this. Turn on the light, please. :)

And AB as the back up plan that allows a class A amplifier to work reasonably well if presented with a lower impedance external load.
Class A is a concept with a validity domain (load current cannot exceed idle). Class B is a simplification of a more complex process.
AB is the practical answer.
When I wrote "A+AB+B", I should have writ "A+ A->B + B->C". In the Crown amps I remember, the output stages had the predrivers in class A, the drivers in AB and the outputs very close to class C.

One can reasonably argue that AB is not even a proper class in itself, but simply a combination of A and B.
+1
 
Back several decades ago I wrote an article explaining the sundry amplifier classes. I just did a search for it and the Peavey server is down for maintenance. Here is a link to it if the server comes back up. amplifier article

Note my article is targeted toward Peavey customers and includes me pimping some Peavey amplifiers of the period.

JR
 
My definition is class A both devices conduct complementarily all the time, AB one of the devices does not pass current for a fraction of the time, B is each device conducts strictly alternatively (which is not a real possibility).

Yep, and I will admit that you're probably in the majority as far as what we call them




I'm not familiar with this. Turn on the light, please. :)

I suspect you’re more than familiar with the theory involved 😉 But you want others to be on the same page too, which is fair enough and what this forum is about. So:


When we talk about classes of amplifier, we’re really talking about the conductance angle of the output transistors. Or tubes for that matter.

In a class A amplifier, both output transistors are continuously conducting throughout the full cycle of audio signal and we say that the conductance angle is 360 degrees (all of the cycle)

In the other type of amplifier we’re discussing here, each output transistor conducts for about 50% of the audio signal cycle and we say that the conductance angle is 180 degrees (half the cycle).

Our amplifier is like a relay race with the output transistors being two runners in that race, and the audio signal being the batton that’s usually passed between the runners at an appropriate point.

If both runners were running around the track together, each runner simultaneously holding one end of the batton, then this is analogous to the class A amplifier. It gets the job done, but isn’t efficient in terms of energy usage.

If we have a well trained pair of runners, each runner carrying the batton for half the time, and there is a well executed hand off of the batton between the two, then this is analogous to what I am calling a class B amplifier.
I do believe it’s an issue of semantics as to whether this should be called AB rather than B.
A quick poll of a few books on amplifier design and I see I’m in in a minority, but am in agreement wth Doug Self in calling this class B. On the other hand, Bob Cordell calls it AB but admits that semantics are involved.
Most other folks call it AB with the exeption of Barney Oliver whose paper from the 1970’s explicitly talks about this class of amplifiers as B.


On the optimum bias for our non class A amplifier:
It isn’t so much the absolute current that we are placing a value on, but the voltage drop across our emitter resistors. When this is optimum, our small signal and large signal output impedance is the same, or as near as we can get it to being the same.

We’ll only ever get close, never exactly, to this ideal - but this condition is met when Re (the resistor attached to our transistor’s emitter) and our transistor’s intrinsic resistance (re’) is the same.
This condition is met when there is an approx 26mV drop across our emitter resistor.

Edit: I just noticed that PRR touched on this last part in an old post above. However, he says 30mV (then adds "or 26mV whatever") across the emitter resistor.
Without wishing to disagree with PRR, I will add that my own experience of Iq = .026/(Re+re+(rb/hFE)) generally errs on the ideal value being between 14mV and 26mV drop across the output transistor's emitters rather than being higher than the 26mV figure. Not always hence, measure it and A.O.T.
 

Attachments

  • Cross-Over-Distortion-in-Class-B-Amplifiers.pdf
    663.8 KB · Views: 22
Last edited:
You may be conflating a couple different issues. The widely accepted definition of amplifier classes is all about conduction fraction.

For some oversimplified definitions;

"Class A" doesn't even need two devices if just means the output device is constantly conducting so there is no turn-on/off transition. Lots of simple old discrete low level circuit paths are pure class A. Not because its better, but because its cheap. The low cost availability of modern ICs has displaced most discrete design class A or otherwise.

"Class B" involves pairs of output devices that alternately conduct one sourcing current during the positive half of the waveform, the second sinking current during the negative half of the waveform. Pure class B is rarely used because of perturbations related to power devices not turning on and off instantaneously, colloquially known as "crossover distortion", most apparent in low level high frequency signals .

"Class AB" is an elegant solution to crossover distortion. Instead of a hard transition between sinking/sourcing current at exact zero crossings, a small class A current keeps both power devices turned on during the zero crossing. This amount of current can be pretty small perhaps little as 25mA in a relatively high power stages. I suspect Winston's discussion may be related to setting class A current for an optimal class AB output stage. There are other considerations and I promised to keep this simple so we will ignore thermal stability etc.

"Class C (E/F?)" these are RF amplifiers where the source/sink output stages conduct less than 50% of the duty cycle, relying upon resonant circuits to deliver full sinusoidal output. Don't worry this class will not be on the quiz.

"Class D" This technology uses saturated switch output devices to alternately pull the output up hard to positive rail, then hard down to negative rail. This PWM encoded audio signal gets extracted and turned back into audio by a LP filter. There are variant Class D designs but they are beyond this overview.

"Class G/H" To improve efficiency of class AB amplifiers multiple PS voltage rails allow parallel output stages to draw current for small signals from a lower voltage rail saving much thermal power dissipation. Higher power output signals draw from higher voltage rails.

Class D is the most efficient audio amplifier class which means all else equal it requires less power supply and heat sink for equivalent power output. Only recently has class D technology matured to the point where it delivers on that cost saving promise.

That's the simple version :cool:

JR
 
Yep thanks John. I conveniently omitted the single device version of class A as it didn't seem relevant to this discussion of how two opposite polarity devices are operating, but your point is an important one to make 👍

I'm happy to acquiesce for the common consensus of understanding on here regarding what I call class B :)
Again, I rmean this only in regard to the practical audio amplifier, one in which there is indeed a transition, one that is optimum in terms of bias.

The issue as I see it, and the reason I went with B. Oliver's and D. Self's definition, is because class AB is a very loose definition indeed.
How much overlap do we include in this definition? Does it include the optimum bias as well as the over biased? And also under biased?

Its safe to say we're all talking about the same thing anyway, that being an amplifier in which there is as smooth a transition from the NPN to the PNP as possible.

👍
 
Last edited:
Here's how Self initially outlines amplifier classes as we're discussing here, he even stole some of my words 😉
Class AB :
 

Attachments

  • Self on  Class AB.png
    Self on Class AB.png
    85.7 KB · Views: 45
I think B. Cordell's explanation is the best. If we stick to the definition that in class B each active element conducts exactly 50% of time or 180 deg, D. Self is wrong. When I explain the differences in classes to my students, I actually show it through a change in THD depending on the bias of the output transistors.
Excerpt from Cordell's book:
"The formal definition of classes A and B is in terms of the so-called conduction
angle. The conduction angle for class A is 360 degrees (meaning all of the cycle), while
that for class B is 180 degrees. More accurately, the definition should really be the angle
over which the transistor contributes transconductance to the output stage and signal
current to the output. This precludes many so-called non switching amplifiers from
being called class A. Such amplifiers include bias arrangements that prevent the power
transistor from completely turning off when it otherwise would.
Most power amplifiers are designed to have some overlap of conduction between
the top and bottom output transistors. This smoothes out the crossover region as the
output current goes through zero. For small output signal currents, the output transistors are in the overlap zone and the output stage effectively operates in class A. These
amplifiers are called class AB amplifiers because they possess some of the characteristics
and advantages of both class A amplifiers and class B amplifiers. Most push-pull vacuum
tube amplifiers operate in class AB mode. Class AB output stages have a conduction
angle that is greater than 180 degrees, although sometimes only slightly so.
There is some semantic controversy in the definition of class B and class AB output
stages. This arises partly because transistors do not turn on and off abruptly, so the
definition of 180-degree conduction is fuzzy. There is a grey region between class B and
class AB. I view class B as an amplifier that is underbiased. I also say that a class AB
output stage has transitioned into its class B region when it exits its class A region.
I will use the term class AB to describe the optimally biased output stage, as it has
important historical origins in push-pull vacuum tube amplifiers. Optimally biased
class AB output stages can have a very substantial quiescent current when multiple
output pairs and low-value emitter resistors are used. They have a class A region that
extends to double the value of the output stage quiescent bias current."
 
Yep I can see that Moamps, I have the book and certainly agree with what's written there in regard to the operation of things. I admit that I made an omission above when writing about a conduction angle of 180 degrees. I should have written "slightly more than" 180 degrees, or even to be more precise as Bob writes:
"the angle over which the transistor contributes transconductance to the output stage and signal
current to the output."


He does say this:
"There is some semantic controversy in the definition of class B and class AB output
stages. This arises partly because transistors do not turn on and off abruptly, so the
definition of 180-degree conduction is fuzzy. There is a grey region between class B and
class AB. I view class B as an amplifier that is underbiased."


The implication being that this is not necessarily set in stone as far as others are concerned.


I'm actually convinced that AB as a definition was a carry over from vacuum tube circuitry when there was AB1 and AB2 etc., for with and without grid current.
Vacuum tubes have a much smoother transition between on and off states of course, therefore not really directly applicable to solid state.

However, as I said earlier, I'm happy to acquiesce on the name on here so that folks are on the same page. :)
But then, how exactly are you defining this optimum class AB? By Oliver's definition of class B or?

When you demonstrate to the students in your class the changing of THD against the change in bias, I assume you are showing this for both small and large signals yes?
 
Last edited:
There are three areas when bias the output transistor.
The first region of large THD and very small current Ic which is actually class C, the second region where THD changes greatly (falls) with the change of current Ic (knee in Ic = f (Vbe)), the beginning of this region is B bias, the end of this region is the so-called optimal AB bias, the third region is A where THD is the smallest and does not change significantly. Parts of this area are sometimes called deep class AB.

I do not use a large output signal during the demonstration, a few volts is usually enough to clearly see the significant change in THD.

The name AB2 class is an unfortunate choice because it does not say nothing about the bias of the output tubes but only about the way the output tube is excited by the voltage and grid current.
 
OK thanks Mo. So it seems, if I understand your description correctly, we're talking about the same in terms of optimum bias. The only difference being the naming convention in that you name the end of your second "B bias" region the "so-called optimal AB bias".

That's all I care about, that we're all on the same page in agreeing that there is an optimum as laid out by Oliver for the minimising of crossover distortion. I'm happy to use "optimal AB" on here from now on to avoid confusion :)

The issue then becomes the murky area of what you say is "sometimes called deep class AB" in your third A region. This is the area where, if the overbias is well beyond that stated by Oliver, we see gm doubling distortion on large signal currents.
 
Last edited:
Back to the topic:
Just as tomgam1 stated in his post, back a couple of decades ago when I played with this stuff, I also noticed that force biasing by turning off one output device didn't always result in much change regarding distortion. The outcome being IC device dependent.

The esteemed Walt Jung proposed a circuit in the late 90's whereby there was an active current source used for the "forcing", followed by a diamond buffer which handled the heavier lifting as far as output load. Circuit attached.

I no longer have the measurements I took of any tinkering around as my means of storage back then consisted of taking Polaroids of FFT displayed on the ancient HP analogue spectrum analyser.

With a diamond type buffer attached to the IC, it was sometimes not even necessary to current source "force" the IC's own output since, for all practical purposes, the buffer presented a high enough input impedance that the IC was possibly already operating class A for any reasonable voltage level.
Who knows? Without access to the IC's schematic it's anyone's guess regarding tricks that are utilised within to minimise distortion.
 

Attachments

  • Walt Jong Buffer.png
    Walt Jong Buffer.png
    293.2 KB · Views: 44
My slightly upgraded version of the Diamond buffer which keeps the voltage across the driver transistors constant and thus, gives a bit less distortion.
Looks like emitter degeneration values shown here for the output pair are a typo by me, I doubt the 12 R values shown will have been the same as the driver degeneration R's since output current is a ratio of the driver transistor quiescent so, scale accordingly.
Faster output devices would be good too.
 

Attachments

  • Modified Diamond.png
    Modified Diamond.png
    60.1 KB · Views: 33

Latest posts

Back
Top