Graham Maynard said:
This is utter nonsense. For practical purposes, in audio, the fb loop acts instantaneous, while speaker back emf (however you define it) is a mechanical phenomenon that operates with time constants orders of magnitude longer than that fb loop.
What do you think, the speaker surprises the fb loop because it was just not watching for a moment? Give me a break.
Jan Didden
But that all is clear visible in tests with reverse driven amplifier. It can show peaks around current zero transition ( small bias= narrow peaks-"disconected" NFB at transition, loss of control, big bias= wide dip -gm doubling) on elipse and it is in relation with crossover distortion in class B or AB. Angle of elipse axis show output impedance, width of elipse show amplifier's phase angle at test frequency. Ideally it should be horizontal line (zero output impedance,unlimited bandwidth), but there is no ideal amplifier in real world. There is no mystery with back-EMF, it creates voltage drop on momentary output impedance (included conectors and speaker cables!), independent on amplifier class, A or AB. Of course, class A is free from crossover high order's distortion, in this point it is better, but in other aspects it is the same for A or B(AB).. EMF can only be induced with real , bandlimited signal on amplifier's output, so it is also bandlimited, not as in simulation with steped signal from zero.
Hi Jan,
This is like making an audience laugh by mocking.
Did I say that back-EMF surprises the NFB loop?
Bandwidth limited back-EMF whether from capacitor tuned crossover-inductors or air-spring tuned drivers can be momentarily large, and occasionally series/parallel reinforced within composite loudspeaker systems. It is the amplifier topology that becomes momentarily 'surprised' when the music input simultaneously changes, such that the amplifier cannot continue working correctly (especially charge storage effects limiting open bandwidth class-AB), so how can any amount of NFB work as expected to make the problem go away before the waveform is damaged ?
Yes BV,
and that is why I have told everyone who is unhappy about suddenly starting sines on my simulated traces to ignore my first cycle energisation. Most audio responses become steady after one cycle, so why is everyone fighting me ?
*However* any amplifier capable of cleanly coping with a suddenly starting 10kHz sine, input or output, is going to be less easily 'surprised' than one which is of poorer topology and/or is open bandwidth limited at high output.
Edmond's modified class-AB has also been challenged in this thread. I do not like the look of the crossover spikes which are increased due to the hf reactive peak of the output stage loop (maybe a series resistor could flatten this), nor have I built, tested or auditioned that particular arrangement, but for class-AB it works very well, mainly I suggest because of lower impedances and adequate current around the differential stage.
We all have different ways of looking at things !
Cheers ........... Graham.
This is like making an audience laugh by mocking.
Did I say that back-EMF surprises the NFB loop?
Bandwidth limited back-EMF whether from capacitor tuned crossover-inductors or air-spring tuned drivers can be momentarily large, and occasionally series/parallel reinforced within composite loudspeaker systems. It is the amplifier topology that becomes momentarily 'surprised' when the music input simultaneously changes, such that the amplifier cannot continue working correctly (especially charge storage effects limiting open bandwidth class-AB), so how can any amount of NFB work as expected to make the problem go away before the waveform is damaged ?
Yes BV,
and that is why I have told everyone who is unhappy about suddenly starting sines on my simulated traces to ignore my first cycle energisation. Most audio responses become steady after one cycle, so why is everyone fighting me ?
*However* any amplifier capable of cleanly coping with a suddenly starting 10kHz sine, input or output, is going to be less easily 'surprised' than one which is of poorer topology and/or is open bandwidth limited at high output.
Edmond's modified class-AB has also been challenged in this thread. I do not like the look of the crossover spikes which are increased due to the hf reactive peak of the output stage loop (maybe a series resistor could flatten this), nor have I built, tested or auditioned that particular arrangement, but for class-AB it works very well, mainly I suggest because of lower impedances and adequate current around the differential stage.
We all have different ways of looking at things !
Cheers ........... Graham.
Graham Maynard said:
No. But you do now:
Graham Maynard said:
I was ahead of you.
Graham Maynard said:
Look, without a basic understanding of feedback circuits from your side, it is futile to try to communicate. I give up.
Jan Didden
Tell me Jan,
How do you expect NFB to counter momentary carrier/charge retentions within output stages during higher current class-AB crossovers, or the likes of differential tail current slew rate limiting/charging due to a Miller connected C.dom when the global loop momentarily overshoots because the output stage is incapable of following?
Well?
You appear to ASSUME that I lack understanding ?
Is it not the case - 'that which is understood during linear operation' - then literally goes - 'out of the (NFB) window'.
NFB error control is applied wrt input via the input filter, thus it cannot perfectly correct for load induced group delay errors due to non-linear internal activity.
I have just been heartened to recieve a private e-mail from someone who reads here, but does not post.
(He would probably get the same stressful barracking I do.)
His last sentence made me smile.
"We cannot learn what we think we already know."
Jan, you have challenged my understanding yet you have not commented on that multi-trace forward driven simulation I posted for a typical class-AB circuit, the output characteristics of which were modified to 'not have' common stage current/parallel capacitance generated output inductance.
How does your understanding explain the different error traces ?
Cheers .......... Graham.
PS. I really have wasted far too much time fending off non-technical challenge, so I am now unsubscribing this thread.
I have been here trying to encourage others to think about the aspects that are not being considered, not to have my different approach/thinking condemmed.
How do you expect NFB to counter momentary carrier/charge retentions within output stages during higher current class-AB crossovers, or the likes of differential tail current slew rate limiting/charging due to a Miller connected C.dom when the global loop momentarily overshoots because the output stage is incapable of following?
Well?
You appear to ASSUME that I lack understanding ?
Is it not the case - 'that which is understood during linear operation' - then literally goes - 'out of the (NFB) window'.
NFB error control is applied wrt input via the input filter, thus it cannot perfectly correct for load induced group delay errors due to non-linear internal activity.
I have just been heartened to recieve a private e-mail from someone who reads here, but does not post.
(He would probably get the same stressful barracking I do.)
His last sentence made me smile.
"We cannot learn what we think we already know."
Jan, you have challenged my understanding yet you have not commented on that multi-trace forward driven simulation I posted for a typical class-AB circuit, the output characteristics of which were modified to 'not have' common stage current/parallel capacitance generated output inductance.
How does your understanding explain the different error traces ?
Cheers .......... Graham.
PS. I really have wasted far too much time fending off non-technical challenge, so I am now unsubscribing this thread.
I have been here trying to encourage others to think about the aspects that are not being considered, not to have my different approach/thinking condemmed.
Graham Maynard said:
Why not?
In any case the crossover notch doesn't occur in Class-AB.
Graham Maynard said:
When is slew rate limiting a problem with an amp. of even moderate competence during normal use?
Graham, I am afraid you have fallen into the sticky pseudo-science trap inhabited by folks likes John Curl.
Graham Maynard said:
Graham,
Let me make one thing clear. None of us here dismiss the possibility of back EMF, or other related things, causing momentarily higher currents to flow than one might expect. Otala showed this and so did I. See my paper on IIM at www.cordellaudio.com. This is why we all agree that amplifiers should have adequate drive margin against higher currents flowing, and that amplifiers should not protect or current-clip under those conditions. Given an amplifier that satisfies that requirement, which is absolutely no mystery, there is not a further problem here. If all you are saying is that the amplifier should perform with very low distortion when sourcing unusually high currents, at whatever phase angle, there is no controversy - but there is nothing more to it than that.
Cheers,
Bob
Hi Mike,
My post #683 shows a terrible error, as I put a wrong quote.
Erroneously, I read it as: "...does not increase the amount of feedback applied to the INPUT stage...." So, in this sense, I agree with you.
Don’t ask me how this error could happen, perhaps I was in a hurry. Anyhow, forgive me this stupid mistake, which certainly did not contribute to any clarification on my part.
To all,
BTW, now that I come back to this topic, I take the opportunity to say that I’m not feeling comfortable when people talk about Edmond’s invention or scheme, as I merely re-re-invented this trick. I rather prefer to give it more neutral name. Given its transitional nature, that is, switching gradually between traditional Miller compensation (at HF) and Cherry’s compensation (at AF), I suggest to call it: “Transitional Miller Compensation”, in short TMC. Comments invited.
Cheers,
estuart said:
This is also true compared to single pole compensation.
estuart said:
Who invented this exercise in futility then and what did they call it?
estuart said:
Transitional Miller Compensation (TMC) seems to be a good name. Do we have any idea where or who it came from?
Bob
mikeks said:
Hi Bob and Mike,
As far as I know this "exercise in futility" (not my words) was first invented by the late P. Baxandall, then re-invented by D. Self (according to private communications with D.Self and me). Then we have Lars Clausen, see post #512, who essentially applied the same technique. Regrettably, I don't have any paper on this subject, so I don't know how the original inventor might have called it, hence my suggestion to call it TMC.
Sorry to say that Graham Maynard's litterature about distorsion of filters and amplifiers (six lengthy articles) published in Elctronics World had the consequence that I didn't renew my subscription. My prefered magazine since years didn't survive long, I think.
mikeks said:
Hi Mike,
First, The only difference is that Lars Clausen’s compensating capacitor (C6=2p2) is tied to the inverting side of the input stage, in stead of the input of the VAS. So what! If you are still regarding this scheme, opposed to all other members, as feed-forward or phase lead compensation, doesn’t matter. We all, except you, consider it just a variant on the traditional Miller compensation. We don’t start a semantic debate all over again, do we?
Second, You might consider to have a closer look at Lars Clausen’s diagram and maybe you will discover that R24, just like my R1 (see post #505), is tied to the output and a series cap, C6=1n (my C2), is tied to the VAS output. Apart from the 2p2 C6, I don’t see any essential difference. Lars Clausen's scheme is just an other example of TMC.
Maybe I've overlooked an other essential difference, that might justify your statement. Please, let us know.
Cheers,
estuart said:
You don't know what you're talking about.
Miller himself disagrees.
More info.
Moreover, semantics:
the study of meanings in a language.
Another example of misuse perhaps?