Myths, tricks and hey, that's neat!

[PMA]

I don't see how that's useful for an article that's talking about generic details?

I think it is, these are real, serious issues that give the feedback back reputation. From my point, another high school textbook analysis does not help. We have dozens of that.

 

[69383]

It happens at any loopgain level, more or less. The more loopgain, tighter feedback, the better and effect is lesser. However, it is always here. You may have pure H2, after application of feedback it will be reduced but also and always higher harmonics are created, though very low in level. Easy to show in simulation.

Well if bad amplifiers get worse and good amplifiers get better, I expect a sweetspot somewhere in between.

 

[MarcelvdG]

It happens at any loopgain level, more or less. The more loopgain, tighter feedback, the better and effect is lesser. However, it is always here. You may have pure H2, after application of feedback it will be reduced but also and always higher harmonics are created, though very low in level. Easy to show in simulation.

That's also what you see in the graph (and Feldtkeller didn't even need a simulator for it). The absolute value of the quantity on the vertical axis is proportional to the third harmonic distortion (I think it is the third-order Taylor coefficient, but I'm not sure), the loop gain is on the horizontal axis. The input signal plus bias is raised to the power gamma.

gamma = 2 is the theoretical case of a forward path that generates only second-order distortion.

The gamma = 1.5 case is for a valve that strictly follows Child's law - which no valve ever does, but it's closer to reality than a pure quadratic function. In this case, you get a minor bump in the third-harmonic distortion around a loop gain of -0.7.

 

[jan.didden]

It is a good article as far as it goes. One thing I learned about negative feedback, and correct me if I am wrong, if you start with a single stage amp and from the input signal to output you measure its total harmonic distortion, then apply negative feedback and the total harmonic distortion amount is reduced. But the original output distortion spectrum before feedback is 2nd and/or 3rd order harmonics, with the negative feedback the total distortion product maybe reduced but now you have added 4th and 5th order harmonics, at a lower level but there, and around you again and now you have an even smaller amount of 6th and 7th order harmonics added to the output signal.

I and better people than me (Bruno Putzeys, Bob Cordell) have debunked that again and again, but I will do it again in a future column.

Jan

 

[jan.didden]

In this particular case about feedback, I would have used the more standard symbols from control theory.

They are just much easier to understand and follow for the average reader compared to these specific LTSpice blocks. (they actually always confuse me)

I am aware that they do the same, but in my opinion anything that's control loop related, is just easier with adding/summing/subtract blocks, in combination with proportional blocks (= gain), integrator, differentiators and other transfer functions.

This just gives a much better overall understanding, also about the fact that these things are not just only related to electronics.
Which gives the opportunity to use more practical things as an example.
Making it a lot less abstract.

By default, those can't be found in LTSpice, but the famous "zzz library" has them.
Which can be found here, if I am not mistaken http://www.bordodynov.ltwiki.org/

Also, what I was missing, is that if you talk about feedback one also has to (briefly) mention feed forward as well.
Because feedback is not the only to prevent things from crashing into other things.

Good points. I deliberatedly used LTspice symbols hoping they would look familiar to the non-EE people on this forum who all seem to use LTspice even if they are not sure what it tells them.
And the format was trying to give a 'closed' explanation in less than 1000 words. I could have gone on about feed forward, positive feedback and how to design a star ship, but that would overshoot the target. 😎

Jan

 

[jan.didden]

Well if bad amplifiers get worse and good amplifiers get better, I expect a sweetspot somewhere in between.

ALL amplifiers eventually get better, it just takes more feedback if you start out with a bad one. The worst you can do is take a mediocre amp to begin with and add a low level of feedback. A tube amp with 3% distortion and add 12dB feedback. That definitely will sound bad.

Jan

 

[PMA]

That's also what you see in the graph (and Feldtkeller didn't even need a simulator for it).

I am speaking about a bit different phenomenon. Look at this, please, when tight FB is closed around pure H2 nonlinearity. The rise of high order harmonics is inevitable, you will find it in many publications, Douglas Self on amps to name one.


Pure H2, no FB


Same nonlinearity in opamp loop


We all know that distortion was suppressed to negligible level. However, existence of new, high order spectrum components is a fact.

 

[Daihedz]

... The article is being praised by those who already understand. ...

You might have written "... The article is being praised by those who already understand a bit ...". Or also "... The article is being praised by those who are willing to understand ...".

... don't kid yourself that you have reached any musicians , trades people, doctors, history teachers and other diyers without a career investment in math who will continue to ignore your forum conversations in favour of those threads rehashing the supremacy of zero feedback, non-oversampling and fatter speaker cables ...

I am a doctor as well, with only relatively sparse mathematical career investment, and I would like to thank very much forists like Jan, and thank you to all other forum authors willing to make the effort to teach and share all kind of in-depth theoretical and also practical background knowledge about audio. This is why I am on this forum: First place, to steadily learn.

 

[MarcelvdG]

I am speaking about a bit different phenomenon. Look at this, please, when tight FB is closed around pure H2 nonlinearity. The rise of high order harmonics is inevitable, you will find it in many publications, Douglas Self on amps to name one.


Pure H2, no FB
View attachment 1244624

Same nonlinearity in opamp loop
View attachment 1244625

We all know that distortion was suppressed to negligible level. However, existence of new, high order spectrum components is a fact.

That's exactly what Feldtkeller analysed in 1936, although his main interest was the third harmonic rather than the higher ones, and besides a purely quadratic function, he also looked at a bit more realistic non-linear function.

 

[PMA]

Purely quadratic function was used for clarity purpose, easy to understand. We can see realistic nonlinear function when we analyze real amplifiers, however direct FB effect on new high order harmonics is then masked and messed up with the signal without FB. High order harmonics are mentioned because they become more audible than low order ones, ear masking. Simulators give us new view not available in 1936.

BTW, interesting to see your and Stuart’s great interest in nonlinear tube circuits.

 

[b_force]

I think it is, these are real, serious issues that give the feedback back reputation. From my point, another high school textbook analysis does not help. We have dozens of that.

I never said they aren't real?

But you first have to establish the basics and common ground for an article that's very obviously written for beginners.

For people with more experience you would have done something different entirely.

 

[b_force]

Good points. I deliberatedly used LTspice symbols hoping they would look familiar to the non-EE people on this forum who all seem to use LTspice even if they are not sure what it tells them.
And the format was trying to give a 'closed' explanation in less than 1000 words. I could have gone on about feed forward, positive feedback and how to design a star ship, but that would overshoot the target. 😎

Jan

Non EE-people are fine following the standard symbols of control theory 🙂

I totally get your point about feed forward.
But the way I read the article, was that it was tailored towards beginners or newcomers. In that case I think it's handy to establish some basics.
Which can be done in like one or two sentences.

but that would overshoot the target. 😎

I like that pun 👍😎

edit:
Here an example, there should be also 2nd order and 1st order block, but can't find them atm

and theoretically ideal non-inverting:

 

[krivium]

^ Or a small box explaining the different symbols and/or giving some short insight on principle too long to explain in the main text. I love when authors gives ways/guide to let me explore by myself ( reference to authors or books, name of other principle at play, etc,etc,...).

In other word let the chance to be active with learned content and not spoon feeded. It give me the feeling to be smart... satisfy my ego! Lol.

 

[b_force]

In other word let the chance to be active with learned content and not spoon feeded. It give me the feeling to be smart... satisfy my ego! Lol.

Well, what most people here seem to forget, or don't seem to understand, is that this article is written on Audioxpress.
A website for people with all kinds of backgrounds, the majority probably not with a strong physics/mathematics background.

It's NOT a website for EE veterans who have a master in theoretical electronics or similar by experience.

 

[rdf]

A website for people with all kinds of backgrounds

The banner of the page linked at the start of the tread suggests otherwise: Weekly Newsletter for the Audio Industry, Professional Designers and Audio Enthusiasts
That said even as an LTSpice user for a decade plus the voltage dependent voltage source symbol was unfamiliar and required searching. It's fair to say most users here work with models of physical devices. Also, how negative feedback arises in a network in which all negative terminals are grounded is at this moment beyond the reach of my morning coffee. Maybe an explanation of the advantage of using a VDVS over the familiar plus/minus/out three terminal block device and the signal flow phase relationships would ease entry.

 

[b_force]

The banner of the page linked at the start of the tread suggests otherwise: Weekly Newsletter for the Audio Industry, Professional Designers and Audio Enthusiasts

I guess you haven't been following Audioxpress?
They have been around for over two decades.

Yes, they provide some technical articles sometimes.

Btw, the suggestion that you just quoted is also perfectly in line what I just said.
It doesn't say: EE tech nerd who is deeply in circuit analyses.
So it doesn't even suggest otherwise, it literally just proved my point.

Or does every single person in the audio industry, who is a audio enthusiast or designs loudspeakers is all of a sudden a master in circuit design as well?

Also, how negative feedback arises in a network in which all negative terminals are grounded is at this moment beyond the reach of my morning coffee.

That only tells us again that we are talking about a brought and wide audience.

Any EE student, or student from other engineering studies (physics, acoustics, mechatronics etc), will get to know these symbols in the 2nd or 3rd year of their study at control theory lessons.
Or get at leas the very basics of it.

Again, it's 100% understandable that not everyone has that background (so therefor no offense), which only tells us again that using standard notation and explanation is mandatory.

So my apologies, but I am rather confused about your response, because you just literally explained what I just said;

A website for people with all kinds of backgrounds, the majority probably not with a strong physics/mathematics background.

 

[PMA]

The question then is for whom is the article. Those who know know and would not learn anything. Those who do not know are lost in orientation in symbols. I think that more usual and less general symbols like transistors and opamps, resistors and capacitors would help. It would not be that universal, however still working, clear and IMO more understandable. Explaining technical points to general public is not easy, it is a special discipline.

 

[rdf]

I guess you haven't been following Audioxpress?

Yes, for decades. No, those symbols weren't common when I got my BASc EE. No, they weren't common in four decades of practicing audio professionally. Yes, this specific implementation isn't ideally targeted to all interested technically adept audiences, which was the gist of my post.
And yes, presumptive social harping is tiresome.

 

[MarcelvdG]

In order to understand this

https://www.diyaudio.com/community/attachments/1702219797842-png.1244781/

you have to understand transfer functions and to get over the fact that it is neither a system diagram nor a schematic, but something in between. In particular, the subtraction point makes no sense: what's the point of subtracting zero from something? I understand what is meant because I know the op-amp circuit it is based on, if I didn't I would have a hard time understanding it.

 

[b_force]

Yes, for decades. No, those symbols weren't common when I got my BASc EE. No, they weren't common in four decades of practicing audio professionally. Yes, this specific implementation isn't ideally targeted to all interested technically adept audiences, which was the gist of my post.
And yes, presumptive social harping is tiresome.

Than there was simply miscommunication 🙂

I am confused why you didn't get control theory back than?
The book I used 20 years ago, was originally dated from the mid/late 70s and def being used with EE students.

Anyway, it doesn't matter, it proves the point even more that you make sure everyone is on the same page, before continuing a certain subject in an article.
Like others proposed, a very brief summary and explanation of the symbols etc.
If it's only for being nice as a little pick-me-up again, like oh yes, I used to know this stuff 😀 😀