No. Please go ahead. You may have triggered some perseverant rubbish reactions here. But much more important, you do input and you do evoke useful further information. Indeed, all in all and unfortunately, this thread evolved to a rather mickey S/N ratio until now. But still there is enough useful S contained in it, being worth to individually filter the N out.
Last edited:
I have plotted distortion profiles of LME49860 for various gains/loopgains. Input signal is distorted 1kHz sine that contains only H3 (3rd) harmonic component except for the base frequency 1kHz. LME48960 is used with gains 0dB, 20dB, 40dB and 60dB. Corresponding loopgains (at 3kHz) are 88dB, 68dB, 48dB and 28dB. We can see how with decreasing loopgain distortion raises, and also that even order harmonics are higher (relative to odd orders) as loopgain is decreasing. This demonstrates that we need as much feedback as possible, of course must stay in stable safe zone. And it also demonstrates that in all cases feedback creates the new distortion harmonics that were not present in the original signal, though they are low in amplitude. Output signal was always set to 2Vp.
I happen to know someone who's got a degree in electronics, who is highly outspoken about his dislike for feedback. He is literally one of those mythical designers that is so gifted, he can design a circuit on the back of an envelope using a slide rule, faster than I can find an IC for the same job in the draws. However, he does no critical listening and is highly susceptible to 'religion', i.e. conspiracy theories and the like. When I visit him, his PC is always open on Facebook, and you can't help but notice the amount of 'fringe science' that's recommended to him on his homepage. The anti-feedback religion is just that. And as with any religion, objective evidence is invisible to the believer.
And this is where I think the standard symbol notation is just so valuable.
Because in a perfect world the substraction block would perfectly compensate everything.
In practice however, it's the performance of this element how well this goes.
Since it's the element being in control of the steering wheel (we will keep coming with the puns lol)
So yes, I totally agree with what you're writing.
If that control element isn't up to par and performance is worse than the amplifier itself, it can ruin the entire amplifier.
Which is kinda obvious if you understand the fundamentals.
It's quite literally that a car only performs and drives as good as the skills of the driver.
So if people don't understand this context, that's where all the voodoo stories start.
All that being said, some people enjoy the non-linearities in tube amplifiers.
Which is fine, but the main and important difference is that it's subjective taste, not objective data.
Totally agree.
I saw the immediate in depth discussion growing as well and was also thinking if we could please move that to another topic.
These details are just not important for understanding the concept as a whole and VERY specific on the performance of the element that controls the feedback.
But unfortunately it's what people have a tendency to do, immediately jump on some silly details like hungry wolves.
Maybe a good follow up on your articles, is actually to describe what elements and parameters are of importance and why they limit an amplifier with feedback eventually?
(or a feedback without as well actually)
...And solid-state amplifiers also. There are entire families of mondern-day tone boxes.
I think there is some objective data supporting that some people have a preference for certain non-linearities. I keep looking for it, and I have not found it, but "some study this... or some study that..." without links and/or real references keeps coming up in casual conversation.
Heck, to the novice (me), it seems as if the entire "audio world" is supported by continually perpetuated myths. It seems to me that amplifiers of almost any reasonable level of power delivery for general consumer / home use are available. They perform exceedingly well (objectively) at a price to performance ratio that was likely not possible in the "heyday" of audiophilia. If there ever was a 'problem', it seems to have been been solved. Amplifier tech of today is largely good enough to where there should arguably be no audible difference between "well-designed" amplifiers. So, how do marketing people differentiate? Myths... Nostalgia... etc. etc.
My intent is not to disparage those, like me, that like to play with tone boxes. It's fun! However, there are likely hundreds of exceptionally-well-performing amplifiers on the market that can be had for comparatively little money. If everyone in "the industry" publicly agreed on the goals of an "audio" amplifier, the industry would shrink to almost nothing. It would be the fastest race to the bottom on price. They NEED consumers to think that their box has something special in it. <end wild speculation>
Sorry, for continuing a little of the OT, Jan. Selfishly, yes, continue with more myth-busting, please!
In terms of basic theory and in terms of trying to make things intuitive without a lot of math, there is an old fashioned approach. Its more or less in line with what Earl Geddes has said, to the effect that we should be looking at transfer functions.
Regarding the changes in curvature that transfer functions go through as feed back is increased, the harmonic content changes can be analyzed graphically.
The two files below explain a couple approaches. Think the correlation method is more intuitive, myself. A few examples make it pretty clear why harmonics can increase before decreasing as feedback is increased, since the effect of feedback is to straighten out the transfer function.
Several other references on graphical methods are available if there is any interest.
Regarding the changes in curvature that transfer functions go through as feed back is increased, the harmonic content changes can be analyzed graphically.
The two files below explain a couple approaches. Think the correlation method is more intuitive, myself. A few examples make it pretty clear why harmonics can increase before decreasing as feedback is increased, since the effect of feedback is to straighten out the transfer function.
Several other references on graphical methods are available if there is any interest.
Last edited:
The times have changed since we were discussing here with Scott Wurcer, Walt Jung, John Curl, syn08, AndrewC. Understood. Will not pollute this thread anymore. Good luck, Jan.
Sorry for responding a little late, for some weird reason my phone just flat out refused to download the pdf.
Anyway, it always goes to show how EXTREMELY important it is to read and understand certain papers within context, but also keep an eye on some specific details, especially related to certain conclusions.
Even more so when those conclusions are kinda out of the ordinary.
side note: (OT)
I've mentioned this in another topic as well.
I have read many papers where they just missed a very important detail, in particular with BEM/FEM methods.
(where garbage-in=garbage-out changes are super high)
Therefor their end-conclusions are wrong, yet the follow-up papers continues with this same error.
I have seen a company relying their whole existence on this error a decade later.
It's almost scary how these things go, seriously.
Not loosing history out of sight is so extremely important!
Go watch the new SmarterEveryDay video about NASA's new rocket to the moon, very interesting watch on the subject of the importance of history.
Which loops us back on-topic, with a bit of overshoot (lol), why articles like Jan's are important reminders! 🙂
It's great for people who are new to these topics, plus it's great for veterans to dust of the old books again and start discussing things.
Which keeps us sharp! 🙂
Pretty good excerpt from Self on Audio in EE Times, about input stage feedback-related problems.
https://www.eetimes.com/distortion-in-power-amplifiers-part-ii-the-input-stage/
https://www.eetimes.com/distortion-in-power-amplifiers-part-ii-the-input-stage/
Maybe some addition to the first feedback piece that seems underexposed, and may give better understanding.
With an open loop gain of just 1000, you see that - obviously - the amp input signal is 1/1000 of Vout.
In the attached picture, it's 21.5V Vout and 21.5mV Vin.
Since that amp is still affecting the output, there must be still some distortion appearing at Vout; feedback hasn't totally eliminated the distortion.
And it cannot, unless, in the ideal case (which we cannot achieve), if the amp had an open loop gain of infinity, the input voltage would be zero.
In that case, the amps' amplification would no longer be relevant and distortion would really be zero.
But in our non-ideal world, all we can do is work towards the highest open loop gain we can get away with, to benefit maximally from available feedback.
And in turn, that is important if we strive for HiFi in the original sense - an amp that faithfully reproduces what we throw at it, adding nothing, taking nothing away.
Jan
With an open loop gain of just 1000, you see that - obviously - the amp input signal is 1/1000 of Vout.
In the attached picture, it's 21.5V Vout and 21.5mV Vin.
Since that amp is still affecting the output, there must be still some distortion appearing at Vout; feedback hasn't totally eliminated the distortion.
And it cannot, unless, in the ideal case (which we cannot achieve), if the amp had an open loop gain of infinity, the input voltage would be zero.
In that case, the amps' amplification would no longer be relevant and distortion would really be zero.
But in our non-ideal world, all we can do is work towards the highest open loop gain we can get away with, to benefit maximally from available feedback.
And in turn, that is important if we strive for HiFi in the original sense - an amp that faithfully reproduces what we throw at it, adding nothing, taking nothing away.
Jan
Attachments
'Sounds better' is just an opinion, not a repeatable measurement. It might be that someone finds an amplifier that distorts to be euphonic, if not accurate.
Without any context, it doesn't say much at all.
Someone could have compared two amplifiers in two different rooms while he was just recovering from a cold with one of them.
Or more realistic scenario, "compared two amps" in two entirely different rooms with entirely different speakers. 🤦♂️
Anyway, I don't have a problem that people have a subjective preference.
We all do.
But like I said, there is a difference between subjective taste and objective data.
They are not even the same thing nor comparable.
It's understood that when comparing two amplifiers' S.Q., they are auditioned at matched volume levels that the lower-powered amp can easily handle, and the same speakers, location and listening position are used. Both amps must be capable of working into the speakers' impedance. The only thing different is the two amplifiers themselves, nothing else. As for having a cold, I'd wait until I recovered from that before I would decide on a speaker purchase from my short list based on audition.
This "philosophical" angled discussion would be a more interesting read for me, than nuts and bolts theory illustrated in LTSpice. I've heard of such over the past 30 years, but never had a rigorous clarification. Particularly why "That definitely will sound bad.". Or why "The worst you can do is". These are art of design elements, coming from experience. Not having to personally grind through failing experiences to learn what and what not to do is valuable.
Same, for the topic of speakers; why "That definitely will sound bad.", why "The worst you can do is".
For speakers some aspects are a lot more straightforward.
For example, a resonance always sounds "bad", well I guess unless people are not receptible to it. (aka it's not audible for them)
Which is also fine.
But it will never sound "good".
Similar to an enormous mismatch in directivity.
So to some extend, you can also design with that in mind.
Not because it will be better, but the changes that it will bad are becoming smaller or minimal.
Therefor the changes to succes are higher.
Which is nothing more than a tool.
Like every tool, for some people that starts to become the goal instead.
Nothing wrong with that either I guess, but it's good to understand at what wavelength people are talking.
Otherwise you end up in a never ending feedback loop, in danger of oscillating out of control 😉 (hihi)
Jan,
good work on your articles! I really like the approach of using the "hobbyist standard" LTspice to describe some easily grasped and real-world examples, as I find that to be much closer to a breadboarded prototype than any book on Control Theory could ever be. It is a pretty hands-on approach without the need for a full-fledged workshop and no risk of blowing things up right from the start, and it's not a huge leap into the real world once you have access to a workshop.
Maybe you could devote a thread to each of the articles here on the forum and link to it there, to be more specific or expand on those points that people may have further questions on. And of course "we" all may chime in and help out, too.
good work on your articles! I really like the approach of using the "hobbyist standard" LTspice to describe some easily grasped and real-world examples, as I find that to be much closer to a breadboarded prototype than any book on Control Theory could ever be. It is a pretty hands-on approach without the need for a full-fledged workshop and no risk of blowing things up right from the start, and it's not a huge leap into the real world once you have access to a workshop.
Maybe you could devote a thread to each of the articles here on the forum and link to it there, to be more specific or expand on those points that people may have further questions on. And of course "we" all may chime in and help out, too.