This is correct. No need to read Crowhurst, this is Control Systems Theory 101. Where you lose the plot is in forgetting that if you use enough feedback, those harmonics (which incidentally don't stop at the 81st but in theory extend to infinity) will be so low that no physical ear on this planet can hear them.
On the other hand, a so-called zero feedback amplifier*, while possibly not having harmonics to infinity, has relatively high level harmonics, easily measured, easily heard. That gives your precious amps their specific character - the slew of harmonics coming from a non-linear amplifier at audible levels.
So yes, a global feedback amp may sound cool or analytic compared to your amps - that's because the feedback amp is more high-fidelity in the literal sense of the word; no coloration, no audible harmonics.
Now of course anybody is entitled to his preference, but it's a bit funny to sell a performance shortcoming as an advantage.
Jan
* you probably mean 'no GLOBAL loop feedback'; as you probably know, EVERY amp ever build has feedback, even if it is hidden in a cathode resistor or a Cag or whatever.
In a well-designed amplifier these are very likely to be well below the existing distortion components due to open-loop non-linearity. The exception would be an amplifier with unusually high open-loop low-order distortion.atmasphere said:
Output impedance is a term in electrical engineering. It does not change its meaning for different 'paradigms'.
Silly me. I just thought they were examples of 'euphonic' distortion or 'tone control via equipment rolling'.
When I have time I will read your article, and respond with comments. I hope it contains more sense than your posts.
To the latter: I am very familier with ESL impedance curves- that is why I mentioned them! It appears you did not read my post completely. Go back and look for a list of speakers- its in that 'wall of words'... as you can see the words 'hardly any are' is not really correct at all. An entire industry of SET designers needs a set of speakers they can use...
I don't think any ESL designers designed with the Power Paradigm in mind, but what they did do is they built a loudspeaker whose impedance curve has nothing to do with the efficiency (resonance) of the speaker and has everything to do with a capacitor. As a result its efficiency at any given frequency is about the same, so how do you drive that right unless the amp has some ability to ignore what is typically a 10:1 difference in impedance over the range of the speaker? I can tell you that most of our MA-2 production is running on Sound Labs... and we have a lot of Quad owners as well.
Now to the former: look at how the distortion of the Halcro climbs as power is decreased. Look at how the Pass Labs amp has distortion decreasing as power is decreased. The latter is a very desirable trait in an amplifier, and in listening, this is proven out between the Halcro and the Pass.
This is in fact the basis for the 'First Watt' line of amps Nelson is known for.
Actually the ear hears this stuff easily, it uses the presence of higher ordered harmonics to calculate sound pressure.
That depends on if the amp is non-linear or not. Some amps are not so linear and others are. A typical SET has distortion that is unmeasurable at low power levels. To take advantage of this the amplifier usually has to be on a high efficiency loudspeaker.
Yes, I am using a short-hand there. However the 'cool or analytic' is actually a coloration- it is the presence of higher ordered harmonic distortion in trace amounts.
This is incorrect and easily proven.
In the voltage paradigm, the presence of negative feedback will affect the output impedance. This is well-known.
In the Power Paradigm, it is shown that negative feedback does not affect the output impedance at all. This is because if it did, it would violate Kirchoff's Law.
To illustrate Atmasphere's point: the attached is a single JFET stage.
Just for illustration - this is the famous Peter Baxandall graph everybody gets so excited about.
On the vertical scale is the distortion of that stage for several harmonics. On the horizontal scale is the feedback factor.
For instance, on the left side, where the feedback is zero, the 2nd is at -25dB (a bit over 5%), the 3rd is at -55dB (about 0.05%), the 4th at -98dB etc.
The interesting bit is that when you increase feedback (go from the left bottom corner to the right) you see that the 2nd and the 3rd decease as expected, but, surprise! the 4th INCREASES! Similarly for the higher harmonics.
So, aha! feedback increases higher harmonics!
Yes, IF you have low feedback. The graph also shows that if you make the feedback high enough, ALL harmonics eventually decrease.
Now remember, this is a simple single FET stage, very crooked transfer curve.
Got that?
To be continued.
Jan
Just for illustration - this is the famous Peter Baxandall graph everybody gets so excited about.
On the vertical scale is the distortion of that stage for several harmonics. On the horizontal scale is the feedback factor.
For instance, on the left side, where the feedback is zero, the 2nd is at -25dB (a bit over 5%), the 3rd is at -55dB (about 0.05%), the 4th at -98dB etc.
The interesting bit is that when you increase feedback (go from the left bottom corner to the right) you see that the 2nd and the 3rd decease as expected, but, surprise! the 4th INCREASES! Similarly for the higher harmonics.
So, aha! feedback increases higher harmonics!
Yes, IF you have low feedback. The graph also shows that if you make the feedback high enough, ALL harmonics eventually decrease.
Now remember, this is a simple single FET stage, very crooked transfer curve.
Got that?
To be continued.
Jan
Attachments
So now let us look at the same situation with a reasonably linear amplifier, not the single crummy JFET. See attached, courtesy Bob Cordell (read his book).
Aha! When we NOW increase feedback, ALL harmonics decrease, no new ones emerge from the noise!
This my friends is the situation with a reasonable linear audio power amp. The point with the simple, no feedback (global) SET (tube) amps is that they have NOT ENOUGH feedback - they are like that JFET. That gives you coloration and 'emotion' that some like, and that's entirely their prerogative. And this also explains why some SET amps may start to sound worse when you aply feedback, unless you apply A LOT.
Reminds me of Bruno Putzeys' 'flight of fancy' where he modified a tube amp for 60dB feedback at 20kHz - and that's HUGE.
Some say they never heard a tube amp so clean, so clear, so, yes, analytical.
Jan
Aha! When we NOW increase feedback, ALL harmonics decrease, no new ones emerge from the noise!
This my friends is the situation with a reasonable linear audio power amp. The point with the simple, no feedback (global) SET (tube) amps is that they have NOT ENOUGH feedback - they are like that JFET. That gives you coloration and 'emotion' that some like, and that's entirely their prerogative. And this also explains why some SET amps may start to sound worse when you aply feedback, unless you apply A LOT.
Reminds me of Bruno Putzeys' 'flight of fancy' where he modified a tube amp for 60dB feedback at 20kHz - and that's HUGE.
Some say they never heard a tube amp so clean, so clear, so, yes, analytical.
Jan
Attachments
Last edited:
And only if you have a severely poor quadratic transfer function to begin with. But you knew that...
Part 6, Figure 2 will be your next post, I predict.
edit: Cross-posted! GMTA
Negative feedback always lowers the output impedance. But if you go into constant power, your feedback is no longer straight negative thus output impedance is less or not at all affected. You know that.
Jan
Last edited:
Others just start out poor. Here's what you get for $140,000:
Lamm Industries ML3 Signature monoblock power amplifier Measurements | Stereophile.com
No shortage of high order distortion to begin with, despite the absence of global feedback. Perhaps the square law paradigm is not appropriate...
Nelson even put the figure from Baxandall in the article. Perhaps a close reading of the Baxandall series might be useful for you?
It really sounds to my like you have no experience with this. Specifically, at this point you don't know how the Power Paradigm works and are making an incorrect association to some abstraction.
If I apply enough feedback to one of our amps, it will behave like a voltage source.
But no matter how much feedback is added, its output impedance will not be lowered. As I said before, such would violate Kirchoff's Law, the law of energy conservation. I have noticed that people operating solely in the Voltage Paradigm have trouble understanding this. But Kirchoff's Law is a basic law of electricity and just like Ohm's law, cannot be violated.
Interestingly though, in the voltage paradigm, adding feedback will indeed lower the 'output impedance' which I have in quotes as it is a charged term.
I hope you can sort this out, so I don't have to put up another 'wall of words' as SY put it, but I will explain if asked.
I think you confuse the actual 'internal' output 'resistance' which is part of every amp, and the let's say, 'black box' output impedance. Just bear with me for a second.
You are correct that no amount of feedback can change the 'internal' output impedance of anything, amp or otherwise.
Let me explain. Let us assume for the discussion that you have an amp with no feedback. The open output voltage , no load, is, say, 10V. Now you hang an 8 ohms resistor on the output, and the output voltage drops to 5V. You would correctly assume the 'output impedance' of that amp is 8 ohms.
Now you apply negative feedback, say 20dB. Again, at 10V unloaded, you hang the 8 ohms on it and, surprise! it only drops to 9V! Eureka you shout, the 'output impedance' has changed to 0.8ohms!
Of course there's nothing changed internal in the amp, yet it ACTS as if it suddenly has lost 90% of its output impedance.
This is how feedback does it.
Hope this helps,
Jan
PS You mention 'voltage source'. If your amp is as good as a real voltage source, the output voltage does not change whatever the load you hang off of it - a voltage source has zero output impedance - seen from the outside of course. What's internal in the voltage source, who knows! Maybe a super duper SET with a million dB feedback!
Last edited:
I had never seen that second figure, as posted by Jan, before; only the first one, with the standard Baxandall curves. So is the bottom line, then, that all that talk of negative feedback introducing higher-order harmonic distortion was only relevant in a rather extreme situation, such as Baxendall chose in order to illustrate a point? (I think he had something of order 10% open-loop 2nd harmonic distortion?)
That second plot Jan posted in #90 (from Bob Cordell's book, right?) shows no sign of any deleterious effects from applying negative feedback. That is quite impressive.
Chris
We don't make SETs, we make OTLs
I offered SETs as an example as their ability to have distortion drop to unmeasurable is something that is well-known. This is less well-known in OTLs; as far as I know we make the only OTL with such a distortion characteristic.
Your explanation is correct. Under the Power Paradigm, the internal impedance of the output section is 'output impedance'. In this way the formula that predicts the output impedance of the circuit also conforms with the measurements (using black box technique).
I can't say our amp would be that perfect a voltage source; a lot depends on the size of the amp- the smallest we make does 30 watts into 8 ohms while the biggest does over 500 into 3 ohms.
We found a long time ago that we had to make a choice. We had switchable feedback built in (and have been considering it again in a new product); with a flip of a switch we could go between 20 db and none (essentially bypassing a stage of gain in the process). When feedback was engaged, the soundstage was foreshortened and smaller, plus the highs were brighter with less detail.
(FWIW our amps have one stage of gain, are fully differential from input to output, have a direct-coupled driver and direct-coupled output section, are triode and class A2.)
A lot of people have thought that I made all this stuff about Voltage and Power up, but I didn't. Finding out what feedback did with that switch caused me to examine a lot of information over the last twenty years on distortion and output impedance; I think I learned about the power paradigm about 15 years ago. Thing is, no-one talks about this stuff even though it is part of audio history and stares any audio professional in the face nearly every day. Its led to a lot of confusion and debate on the web too
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.