Hi,
This thread is not about optimal class B biasing, whatever you want to call it.
Emitter resistors do not greatly improve linearity, and for high bias class AB
it has been shown the lower the value of emitter resistors the better. Also
switching performance is not an issue for high bias AB, as basically the
output stage hardly ever turns off, and with no Re's the be junctions
are not significantly reverse biased in the case of 1/4 A class AB.
(Unless severe low impedance loads are encountered.)
Why can't you understand the intended scope of this thread ?
rgds, sreten.
Whether the output stage will turn off mainly depends on the bias current and the load current. What bias current level are you thinking of?
Hi,
Your being lazy. 1/4 A class AB is very explicit, half current / voltage.
It will turn off at high power as its 1/4 A, but it won't turn off hard.
rgds, sreten.
Hi,
Jesus, read this thread !
At typical volume levels (< -6dB) it won't turn off at all.
That is one of the fundamental points of this thread.
rgds, sreten.
Last edited:
Guys please keep things civil..
I think a reiteration of the important point of the thread might be a good idea 🙂
From what I've gathered the intention is to run a single pair of output transistors with a high bias such that you will get around 20 watts into 8 ohms class A. However it will be run off of voltage rails high enough so that you can get 80 watts out of it, where the amplifier will then transition out of class A and into Self's class AB.
The point and question of the thread? Is that sreten wants to know peoples thoughts on certain ideas that could possibly give the best linearity when the amplifier transitions into the class AB region. Is that what's supposed to be going on here?
From what I've gathered the intention is to run a single pair of output transistors with a high bias such that you will get around 20 watts into 8 ohms class A. However it will be run off of voltage rails high enough so that you can get 80 watts out of it, where the amplifier will then transition out of class A and into Self's class AB.
The point and question of the thread? Is that sreten wants to know peoples thoughts on certain ideas that could possibly give the best linearity when the amplifier transitions into the class AB region. Is that what's supposed to be going on here?
Hi,
Yes. but the power output does not matter, its class A up to a 1/4 of output.
And yes, given that, what is the best arrangement to the class AB transition.
rgds, sreten.
Last edited:
Hi Magnoman,
The performance and behavior of the output stage are very important, regardless of how much negative feedback is used. Self's Blameless amplifier is relatively blameless up to the output stage, but not beyond, in my opinion. No matter how good the input stage and VAS, a so-so output stage will prevent realizing a really good amplifier. Self's explanation of optimal bias of the output stage is worthy and important, of course. The CFP output stage is almost impossible to properly bias for optimal crossover characteristics, however. In my book I delve into the shortcomings of the CFP output stage, which go far beyond its difficulty in stabilizing.
It is also important to keep in mind that there are two types of crossover distortion - static and dynamic. Static results from the total transconductance of the output stage varying with current as the signal current goes through crossover. Dynamic crossover distortion largely results from the difficulty of turning off the output transistors at high speed - sometimes called switching distortion.
Cheers,
Bob
20W class A, 80W max into 8R implies Iq=1.1A and V+/V-rail=38V
so output device quiescent = 42W
Why is it that some of the best commercial bipolar designs (Sugden, DNM etc) are simple, with good enough measurements to avoid euphonic distortion, but relatively few components (but ordinary, not boutique) allowing a good layout. Complex amps with VERY low distortion never seem to be preferred.
so output device quiescent = 42W
Why is it that some of the best commercial bipolar designs (Sugden, DNM etc) are simple, with good enough measurements to avoid euphonic distortion, but relatively few components (but ordinary, not boutique) allowing a good layout. Complex amps with VERY low distortion never seem to be preferred.
Rectification: crossover distortion includes any cause of a nonsmooth transition between devices thus switching effects. However, switching implies completely tuned off /on devices so switching distortion does not occur in class AB. I cleave to this definition and apologize for being wrong previously.
Last edited:
What's wrong with building an amplifier that is biased to remain in ClassA upto 25% of maximum rated power output?
I believe that is the question. But am I interpreting you correctly?
Let's look at what you are proposing while sticking with Self's definitions of ClassB, ClassAB and ClassA.
D.Self tells us that ClassA is the pinnacle of these three options in a Blameless Push-Pull amplifier.
Starting with a 40Vpk output for the proposed design. We can get 200W into 4r0 or 100W into 8r0 or 50W into 16r, if each is designed to deliver exactly that voltage into that load.
I'll start with the 50W into 16r version. Bias the output stage to full ClassA for best crossover distortion (paraphrasing Self's words - none). Ib = 1.25A
Max Class A current is 2.5Apk.
Into a 16r load that equates to 40Vpk and we have a Blameless Push-Pull 50W into 16r ClassA amplifier.
Now change the load to 8r0.
Into the lower impedance the maximum peak voltage will be reduced. Let's assume 38Vpk for a -0.45dBV loss when loaded at half impedance. Not an unusual result for a ClassA specified amplifier. This just happens to meet Cordell's rule for 180% of maximum power into half impedance load for a well designed ClassAB amplifier.
The maximum power is 90W into 8r0. This output is ClassAB.
If we listen to normal Audio Signals and restrict the peak output voltage to 20Vpk feeding that 8r0 resistor then the current sent to the load is never more than the ClassA limit of 2.5Apk. We have a ClassA output that performs as well as any ClassA amplifier where we ensure that Current peak is kept below two times output bias current. We have lost nothing in performance relative to any well designed ClassA amplifier driving our 8r0 load. It has Blameless performance without any crossover distortion.
What is the maximum ClassA output power? It is ClassA current limit squared times resistive load divided by two = 2.5^2 * Rload / 2 = 25W.
We have the 1/4 Power into the 8r0 load that was suggested for the hypothetical 100W amplifier in post1.
So our 50W into 16r ClassA amplifier doubles up as a 25W into 8r0 ClassA amplifier and there are no downsides.
If we listen to music into a real speaker load and limit the output current to <2*Ib then we are in ClassA all the time and we have lost nothing in performance. D.Self confirms that throughout his book.
If we listen to normal music and allow for transient peaks that can be 10 (+20dB) times higher voltage than the average level then we are listening at 0W9 into 8r0 with the capability of never clipping a 90W transient. For most of the time the amplifier is ticking over between 90mW of ClassA output and 9W of ClassA output. We don't stress the amplifier at normal listening levels.
We will get frequent peaks that exceed the 9W level (+10dB ref. average level of 0W9) and much less frequent peaks that exceed the 18W level (+13dBV ref. average level of 0W9) These less frequent peaks are still in ClassA if the speaker does not draw excess current due to transient frequency coinciding with a reduced effective impedance.
When the output current exceeds 2.5Apk then the amplifier transitions from ClassA to ClassAB, it does not clip. It simply increases distortion due to the added crossover that was absent at currents below 2.5Apk.
But crossover distortion is more evident at low output levels. At high output levels it is less audible, if at all. So we have our 1/4ClassA amplifier that transitions to very infrequent transitory periods where crossover distortion is evident in measurements but may only just be audible.
I see nothing wrong with that mode of operation. I contend that D.Self told us all that in his book.
I have seen others suggesting that a high biased amplifier can never perform as well as a fully ClassA biased amplifier, but in our normal listening mode we will probably never be aware of that slight increase in distortion on very loud transient peaks. I don't think the High Bias Class AB amplifier can be maligned if protagonists actually think about what is being proposed.
Any faults in this commentary?
I believe that is the question. But am I interpreting you correctly?
Let's look at what you are proposing while sticking with Self's definitions of ClassB, ClassAB and ClassA.
D.Self tells us that ClassA is the pinnacle of these three options in a Blameless Push-Pull amplifier.
Starting with a 40Vpk output for the proposed design. We can get 200W into 4r0 or 100W into 8r0 or 50W into 16r, if each is designed to deliver exactly that voltage into that load.
I'll start with the 50W into 16r version. Bias the output stage to full ClassA for best crossover distortion (paraphrasing Self's words - none). Ib = 1.25A
Max Class A current is 2.5Apk.
Into a 16r load that equates to 40Vpk and we have a Blameless Push-Pull 50W into 16r ClassA amplifier.
Now change the load to 8r0.
Into the lower impedance the maximum peak voltage will be reduced. Let's assume 38Vpk for a -0.45dBV loss when loaded at half impedance. Not an unusual result for a ClassA specified amplifier. This just happens to meet Cordell's rule for 180% of maximum power into half impedance load for a well designed ClassAB amplifier.
The maximum power is 90W into 8r0. This output is ClassAB.
If we listen to normal Audio Signals and restrict the peak output voltage to 20Vpk feeding that 8r0 resistor then the current sent to the load is never more than the ClassA limit of 2.5Apk. We have a ClassA output that performs as well as any ClassA amplifier where we ensure that Current peak is kept below two times output bias current. We have lost nothing in performance relative to any well designed ClassA amplifier driving our 8r0 load. It has Blameless performance without any crossover distortion.
What is the maximum ClassA output power? It is ClassA current limit squared times resistive load divided by two = 2.5^2 * Rload / 2 = 25W.
We have the 1/4 Power into the 8r0 load that was suggested for the hypothetical 100W amplifier in post1.
So our 50W into 16r ClassA amplifier doubles up as a 25W into 8r0 ClassA amplifier and there are no downsides.
If we listen to music into a real speaker load and limit the output current to <2*Ib then we are in ClassA all the time and we have lost nothing in performance. D.Self confirms that throughout his book.
If we listen to normal music and allow for transient peaks that can be 10 (+20dB) times higher voltage than the average level then we are listening at 0W9 into 8r0 with the capability of never clipping a 90W transient. For most of the time the amplifier is ticking over between 90mW of ClassA output and 9W of ClassA output. We don't stress the amplifier at normal listening levels.
We will get frequent peaks that exceed the 9W level (+10dB ref. average level of 0W9) and much less frequent peaks that exceed the 18W level (+13dBV ref. average level of 0W9) These less frequent peaks are still in ClassA if the speaker does not draw excess current due to transient frequency coinciding with a reduced effective impedance.
When the output current exceeds 2.5Apk then the amplifier transitions from ClassA to ClassAB, it does not clip. It simply increases distortion due to the added crossover that was absent at currents below 2.5Apk.
But crossover distortion is more evident at low output levels. At high output levels it is less audible, if at all. So we have our 1/4ClassA amplifier that transitions to very infrequent transitory periods where crossover distortion is evident in measurements but may only just be audible.
I see nothing wrong with that mode of operation. I contend that D.Self told us all that in his book.
I have seen others suggesting that a high biased amplifier can never perform as well as a fully ClassA biased amplifier, but in our normal listening mode we will probably never be aware of that slight increase in distortion on very loud transient peaks. I don't think the High Bias Class AB amplifier can be maligned if protagonists actually think about what is being proposed.
Any faults in this commentary?
Last edited:
Hi,
My only bone of contention would be though all that information is in D.Self's
books, its not made explicit, and Self often states AB is worse than B (Self),
which it is in terms of measured distortion at near full power.
He also states in his opinion its only legitimate use is a fallback mode for Class A.
It is my contention that this is not true. I don't care about measured distortion
very near full power at all, anything linear is much better than clipping, and there
are cogent reasons to deliberately use AB because its better than using B (Self),
and more efficient than using full class A, and / or, more powerful for the budget.
rgds, sreten.
My only bone of contention would be though all that information is in D.Self's
books, its not made explicit, and Self often states AB is worse than B (Self),
which it is in terms of measured distortion at near full power.
He also states in his opinion its only legitimate use is a fallback mode for Class A.
It is my contention that this is not true. I don't care about measured distortion
very near full power at all, anything linear is much better than clipping, and there
are cogent reasons to deliberately use AB because its better than using B (Self),
and more efficient than using full class A, and / or, more powerful for the budget.
rgds, sreten.
Last edited:
Quite HOT ! (below) with 3 pair V mosfets running at over 800ma , you can get 40V peak and stay class A (2.5A). It took 6 pair NJW21193/4 running at 400ma and the drivers running 150ma to satisfy Class A with a triple.
What you get - (below 2), is that $30,000 luxman class A beasty (1ppm @ 20khz - H2 dominant) and a warm room in the cold winter (higher power bill , as well 😀 ).
OS
Attachments
Last edited:
A vs. AB
Below 1 - Class A .. H2 totally in charge except at clip (amp transitions to AB at clip ) distortion PPM all the way.
Below 2 - H2 at real low levels, H3/5 become nearly equal at higher levels , distortion higher rising to .002-5% at moderate levels.
It seems you do get a "little" with class A , but AB is not much worse. H3/5 is most likely totally Xover related.
OS
Below 1 - Class A .. H2 totally in charge except at clip (amp transitions to AB at clip ) distortion PPM all the way.
Below 2 - H2 at real low levels, H3/5 become nearly equal at higher levels , distortion higher rising to .002-5% at moderate levels.
It seems you do get a "little" with class A , but AB is not much worse. H3/5 is most likely totally Xover related.
OS
Attachments
Hi Sreten,
rather than looking at the big picture, it seems to me you have homed in on the region of operation of a slightly overbiased (ClassAB) of an already very low bias ClassB amplifier.
Yes, if we have Re = 0r33 and aim for Vbias of 24mV, then we have Ib = 72mA giving a ClassA output limit of ~144mA. Exceed that 144mApk and you enter ClassAB and D.Self rightly points out, as many others have done, that avoidable crossover distortion is added.
ClassB Ib=72mA, ClassAB Ib>73mA, underbiased ClassB Ib<71mA.
D.Self ranks these three conditions underbiased ClassB is worst followed by ClassAB with ClassB the best of these three, but reminds us that ClassA is best of all four.
Now, let's think about commercial amplifier design that has to be productionised by, among others, the cost accountants. D.Self is proposing a design method for this group.
A ClassA limiting current of 144mA is just 83mW. Ask for any more and the amp transitions to ClassAB with it's attendant increase in crossover distortion. I think most will agree that at this low output level that increased crossover distortion is not only audible, but can ruin quality of performance. Taken in that very restricted context then your interpretation of Self is correct, that ClassAB is not nice and should be avoided.
But the big picture tells us otherwise. ClassAB designed to maximise quality of performance actually breaks that restricted view "rule" of avoiding ClassAB.
I think if you read D.Self again you will see that he does not hide the big picture and that the little picture warning is valid for that little picture set of conditions. Maybe D.Self could have made it more explicit, but the messages are in there, if you read carefully.
I don't think D.Self in particular has maligned ClassAB, nor do any others that warn of the consequence of over-biasing a ClassB stage into ClassAB.
I see at least two solutions, low Re and and multiple devices to make best use of ClassB and very high Ib such that the majority of listening is in the ClassA region. I see no conflict.
rather than looking at the big picture, it seems to me you have homed in on the region of operation of a slightly overbiased (ClassAB) of an already very low bias ClassB amplifier.
Yes, if we have Re = 0r33 and aim for Vbias of 24mV, then we have Ib = 72mA giving a ClassA output limit of ~144mA. Exceed that 144mApk and you enter ClassAB and D.Self rightly points out, as many others have done, that avoidable crossover distortion is added.
ClassB Ib=72mA, ClassAB Ib>73mA, underbiased ClassB Ib<71mA.
D.Self ranks these three conditions underbiased ClassB is worst followed by ClassAB with ClassB the best of these three, but reminds us that ClassA is best of all four.
Now, let's think about commercial amplifier design that has to be productionised by, among others, the cost accountants. D.Self is proposing a design method for this group.
A ClassA limiting current of 144mA is just 83mW. Ask for any more and the amp transitions to ClassAB with it's attendant increase in crossover distortion. I think most will agree that at this low output level that increased crossover distortion is not only audible, but can ruin quality of performance. Taken in that very restricted context then your interpretation of Self is correct, that ClassAB is not nice and should be avoided.
But the big picture tells us otherwise. ClassAB designed to maximise quality of performance actually breaks that restricted view "rule" of avoiding ClassAB.
I think if you read D.Self again you will see that he does not hide the big picture and that the little picture warning is valid for that little picture set of conditions. Maybe D.Self could have made it more explicit, but the messages are in there, if you read carefully.
I don't think D.Self in particular has maligned ClassAB, nor do any others that warn of the consequence of over-biasing a ClassB stage into ClassAB.
I see at least two solutions, low Re and and multiple devices to make best use of ClassB and very high Ib such that the majority of listening is in the ClassA region. I see no conflict.
Last edited:
Hi AT,
My point, if any, is how would you go about building a class A amplifier.
IMO Self implies full class A is the only way to go and I disagree.
As explained by yourself designing for AB has rational arguments*.
IMO 1/4 A either halves your heatsinking or doubles your power.
As BJ stated these concerns are most apposite 10W to 30W.
And IMO EF will be better than CFP for 1/4A class AB,
(but TBH not that much ...).
rgds, sreten.
*not lost on manafacturers ....
My point, if any, is how would you go about building a class A amplifier.
IMO Self implies full class A is the only way to go and I disagree.
As explained by yourself designing for AB has rational arguments*.
IMO 1/4 A either halves your heatsinking or doubles your power.
As BJ stated these concerns are most apposite 10W to 30W.
And IMO EF will be better than CFP for 1/4A class AB,
(but TBH not that much ...).
rgds, sreten.
*not lost on manafacturers ....
Last edited:
- Status
- Not open for further replies.