Or, once well-sorted (and not clipping) and not judiciously being used as an effects box, many amplifiers cannot actually be told apart while blinded? Peeking has magical effects. (I use magic in the most tongue in cheek fashion)
But it is likely that you will hear differences in frequency response due to the loudspeaker's loading with different amplifiers, if some of those amps do not have low output impedance. I guess in that way the high Z ones are to some extent effects boxes.
Among the tests that JA does in his sidebars in Stereophile, I think the amp frequency responses with various loads are among the most useful.
IM or harmonic, they are based on the same non-linearity and are essentially harmless until they get relatively high in value, like 0.1% or more at listening levels. Ampex proved this over 50 years ago in an AES paper.
Does make for a fine line between amp design and calling it a loudness effects box. 🙂
Then again, I'm not botique about my audio in general. Far too boring (and poor) for that.
Then again, I'm not botique about my audio in general. Far too boring (and poor) for that.
There can be differences between several amplifiers of good quality but this will due to peaking as you point it, or more likely to tests conditions that are changed without the testing guys noticing the thing...
Let s assume that we listen a piece of classical music that require 50W for one minute and then 0.1W for an intrument that play a solist part..
From the moment the solist part start and till it ends the speakers will slowly cool down and the system caracteristics is evoluting within much higher variations than the ones that could be provided by changing the amplifier..
A double blind test that do not take account of thermal variation is null and void as the first amplifier to start the test will be deemed more dynamic.
Of course speakers are heated before doing comparisons (i hope that it s the case) but then the room must be under thermic and hygrometric control to get the speakers in iso conditions and the sound velocity constant through the tests..
Putzeys duplicated Baxandall's results, and used similar modeling, global feedback around non-linear amplifier.
I haven't found local feedback to exhibit harmonic multiplication. Here's another example: a pentode, say a 6BQ5, is connected as a triode, giving what I think we could all agree is local feedback. Its gain is reduced and its linearity is increased, but its harmonic distortion structure is now the simpler one of a triode.
If you were to argue that this is *not* feedback, then I would agree to the same extent that degeneration is not feedback, and is instead a new device.
At some level, feedback is not feedback; it's a different (composite) device.
The analogy that convinced me of this is Putzeys' description of SID, where the SID mechanism is shown (ignore the error in the drawing Fig. 11) as operating outside the feedback loop, in a perfectly conventional amplifier.
There are different kinds of "feedback" and some are more equal than others.
Thanks very much for any comments or corrections,
Chris
But that paper is a simulation based on the global feedback model, just like Baxandall and Putzeys. I contend that the model doesn't apply everywhere, just as Putzeys does for SID.
Thanks very much for any comments or criticisms,
Chris
Attached is an example of degeneration I took myself.
I think that when we talk about feedback - local, global, degenerative etc,
it's useful to succinctly describe what we mean, as the terms have fallen
into vague meaning in many of these conversations.
One aspect of the "harmonic multiplication" thing is that it's very dependent
on the actual amount of the distortion. Baxandall's graphic starts at 10%,
which illustrates the phenomenon but assigns an unrealistic 10% figure to
the 0 dB feedback value. It shows that with about 10 dB feedback the 3rd
harmonic rises to about 10 dB below the 2nd. By contrast, my curve which
starts at 1% shows that with about 10 dB feedback the 3rd is about 30 dB
below the 2nd. Take the initial figure down to .1%, and I expect it to
become insignificant.
This indicates to me that the problem is one associated with circuits which
are inherently very nonlinear before feedback. Like a Sunday in TJ, feedback
is cheap but it's not free.
😎
Attachments
Nice demo and the prove that non linear phenomenons have by the definition a non linear evolution, hence the THD starting point is not neutral in respect of the demonstration validity...
That's pretty convincing. Don't know why I don't see it in my own poor observations. Maybe I don't look at "pathological" enough cases to get results up to where I can see them. Your example makes it clear that the effect is there.
Yes! And we're talking about a model, which just makes things more fraught.
So, do you think that the intrinsic resistances inside active devices should be considered to cause degenerative feedback?
Thanks, as always,
Chris
try Linear local feedback and get back to us on that
its hard to write anything at all with out every possible qualification, assumption spelled out in every sentence
the internal plate V feeds back according to the same Langmuir-Childs Law that the gm follows for similar geometry
so the triode strapped pentode example uses a nonlinear "replica" feedback that does give considerable "extra" distortion cancellation
True, of course, but exemplifies the softness of the word "feedback". What is included, and what excluded?
SID is described (by B. Putzeys in the F-word article) as being generated "outside" the global feedback of a stone conventional amplifier. What other exceptions exist? That's what I'm trying to understand.
Thanks very much for any comments or criticisms,
Chris
There's a distinction between the active portions and the "dead" ones, e.g. contact resistances.
The pretty-good models for JFETs include the contact resistances, and if they aren't there the devices adhere better to the well-known quadratic formulae.
Thank you. And I guess I'm stretching too many analogies too thin about this matter of "feedback, its care and feeding". Occam's razor says to cleave to the simplest model, which mine ain't.
And yet, and yet... Something still bugs me. Need to work on it some.
Thanks very much, as always,
Chris
I would imagine any SID or TID is generated outside the feedback loop. At the onset, feedback is much reduced, and as it progresses, it drops off rapidly to the point where the amplifier is operating open loop.
You can make a reasonable argument that there's "no such thing as no
feedback." Drain or Collector impedance could be seen as feedback, the
internal transconductance, and so on.
Of course, if there's no such thing as no feedback, maybe there's little
entertainment in arguing about feedback.
Much earlier in another thread, several knowledgeable types agreed, at least
temporarily, that the use "no feedback" was allowable for marketing
purposes of amplifiers which did not have a global loop.
But we know better, don't we? 😉
feedback." Drain or Collector impedance could be seen as feedback, the
internal transconductance, and so on.
Of course, if there's no such thing as no feedback, maybe there's little
entertainment in arguing about feedback.
Much earlier in another thread, several knowledgeable types agreed, at least
temporarily, that the use "no feedback" was allowable for marketing
purposes of amplifiers which did not have a global loop.
But we know better, don't we? 😉
one thing I've not seen discussed as much is the input side of power amps.
in listening to various amps, i find that some of the time, most of what i am hearing as differences between two amps is just down to the interaction between the PA with the preamps. When i change the pre - some of the differences I hear just go away. And other times they dont.
I think some of this is due to how the power amps load the pre's. While the "DC" impedance presented by the PA may be large, there are probably some strange things that happen at higher frequencies/higher amplitudes.
in listening to various amps, i find that some of the time, most of what i am hearing as differences between two amps is just down to the interaction between the PA with the preamps. When i change the pre - some of the differences I hear just go away. And other times they dont.
I think some of this is due to how the power amps load the pre's. While the "DC" impedance presented by the PA may be large, there are probably some strange things that happen at higher frequencies/higher amplitudes.
What ,if the people who claimeid that back emf from the speakers are interracting with gnfb loop and cables ,were right?
Did anyone measure and compared harmonics spectrum taken from resistive load vs real speakers with cables at the same output power?
Did anyone measure and compared harmonics spectrum taken from resistive load vs real speakers with cables at the same output power?
Last edited:
One real problem with that is, to make a good comparison, one would have to use frequencies and levels that will fry your speakers, but you have a legitimate point there. While you can measure using a 'loudspeaker simulator' (RLC model of a speaker), this still does not account for (sometimes really large) nonlinearities at levels that cannot be considered 'small signal'.
Another way it could be done is to 'back-drive' the DUT amplifier through a speaker by a reference amplifier, and look at the voltage on the DUT output. This can severely load the DUT amplifier as it has to keep it's output at zero regardless of current, but it can be done for levels that are well within reasonable listening conditions. Still, this assumes that OLG is constant regardless of amplifier output voltage, which, as the song goes, ain't necessarily so...
Do we? If true, that would be wonderful news. It isn't true, of course, because local feedback is feedback and feedback is subject to the same laws of mathematics however you do it.Chris Hornbeck said:
You haven't looked hard enough. Unfortunately, measuring this stuff is difficult because the extra products are usually hidden beneath existing open-loop products and simulating this stuff is subject to the usual warnings about simulation of nonlinearities.
Both a pentode and a triode have distortion products at all orders, so the triode is not simpler but maybe lower amplitude at higher orders.
Disagree. Feedback is always feedback.
Yes. Just to be clear, the phenomenon always happens but it is only a problem in situations where the open-loop amp is significantly nonlinear. I'm clarifying that because I don't want someone to go away with the idea that the effect only occurs in certain places. The effect itself doesn't even require feedback - two components in series, with one of them nonlinear, will exhibit it.Nelson Pass said:
But they are not right! 'Back emf' appears as part of the speaker impedance. It is not a separate phenomenon which the 'enlightened ones' have discovered.padamiecki said: