No, it only gets better with more feedback, in the sense that it gets more and more transparent. Music in -> music out, with less and less change.
Jan
Heavy feedback cannot go bad.
More feedback is less distortion.
One can argue that it makes a bad type of distortion brandishing odd harmonics.
The answer is: Those supposed bad harmonics do get lower with higher NFB. There will be a point where the distortion will be swamped under the noise floor. Then, harmonics exist as mathematical calculated artifacts, but they do not exist in the real world of physics.
At very low THD, thanks to heavy NFB all amplifiers sound the same and that same is THE perfect amplifier.
More feedback is less distortion.
One can argue that it makes a bad type of distortion brandishing odd harmonics.
The answer is: Those supposed bad harmonics do get lower with higher NFB. There will be a point where the distortion will be swamped under the noise floor. Then, harmonics exist as mathematical calculated artifacts, but they do not exist in the real world of physics.
At very low THD, thanks to heavy NFB all amplifiers sound the same and that same is THE perfect amplifier.
What's "heavy feedback"? 40 dB? 60 dB?
It depends on the individual circuit or chip. Careless application of feedback can result in instability and poor transient response. There is no one size fits all.
Tube amplifiers employing global feedback often only had 12 to 20 dB of feedback in the upper frequencies. The phase margin of these amplifiers was dictated by the output transformer.
Some chips are "unconditionally" stable (they're not really) and can be configured for unity gain. Power chips like the 3886 are stable for a higher gain only.
Discrete circuits can optimize the dominant pole, circuit stage gain and degeneration, and local phase response for whatever gain the engineer requires.
It's not unusual for modern circuits to employ 40-60 dB of feedback. Heresy to a purist I know, but if done correctly it isn't an issue.
Read up about Bode plots for a description of how these concepts are used. Read "How to Tame the 3886 Amplifier" by one of our members for a look at dealing with real world nasties. This will give you an idea of how feedback works, and its limits.
will try not to get too involved but since we are already sharing opinions (hopefully based on experiences), to put it all in a perspective: at the point when the Japanese OEMs wanted to pursue better audio amps and had actually invested in developing power jfets made for 6-8 Ohm impedance loads to make those, amps also featured equalizers to adjust for a better listening experience. That is how much they thought of the recording signal which was presented to their amp and of the speakers which their amps drove. And now 30 years later when amps are made around less then optimal run-of-the-mill components and call out some 0.00029% for some non standard distortion metric you all come in glorifying them as the end of the road? And they all sound the same as all printed copies of an oil painting could be made to look the same but never as nice as the original in different lights etc.
I deal professionally with people involved with writing software for voice recognition and who are struggling to come up with metrics for pitch, timbre etc. and I can tell you upfront: there is no single number metric that can capture anything that relates to how the sound will be perceived by a human brain. Too bad not more of those people are into audio or if they are it is a private affair for them.
Like I mentioned however, one can readily start experimenting with FW amps for example; a lot of them are designed for the same power supply so boards can be swapped in and out if you are looking to save. And their distortion patterns have already been characterized by others if you want to understand a little of what makes them sound different. Of course one needs really good speakers too, and as many as possible as there are magic combinations in gear matching.
I get most gratification from non-audio friends visiting or my family members when they throw an unsolicited comment or just go silent listening because the music grabs them with presence and detail never experienced before.
I deal professionally with people involved with writing software for voice recognition and who are struggling to come up with metrics for pitch, timbre etc. and I can tell you upfront: there is no single number metric that can capture anything that relates to how the sound will be perceived by a human brain. Too bad not more of those people are into audio or if they are it is a private affair for them.
Like I mentioned however, one can readily start experimenting with FW amps for example; a lot of them are designed for the same power supply so boards can be swapped in and out if you are looking to save. And their distortion patterns have already been characterized by others if you want to understand a little of what makes them sound different. Of course one needs really good speakers too, and as many as possible as there are magic combinations in gear matching.
I get most gratification from non-audio friends visiting or my family members when they throw an unsolicited comment or just go silent listening because the music grabs them with presence and detail never experienced before.
and a separate post on the subject of binaural recording and audio playback (since a few showed interest here): the best recordings have been made by years by Aachen head microphones where the resonance of the ear canal, high frequency shadow thrown by the head itself, etc. were all accounted for. Nowdays the mics to mount on your own head are also available. A few years back when developing a speaker, I took some mics built into a headset home and kept putting them on and taking them off to alternate between listening to binaural audio playback in headphones and the speakers in the room. After a few hours at one point I still had a feeling of the muffs around my ears when I took them off and I thought I was listening to headphones when listening to the speakers in the room; that is how close it gets with VERY expensive software and hardware (typically not deployed in the commercial audio field).
If you are ever in Brighton, MI pay a visit to HeadAcoustics to demonstrate it for you. For example one sales pitch usually entails walking across behind you while jingling keys then placing the keys in front of you and playing a recording back for you where you will be made aware of the spatial effect where you can still hear the keyset moving around you and point to its location.
Commercial audio will never see this as an average customer does not demand it. moreover the improvements on the recording end are not encouraged by the market either. Check out “The Art of Listening” (2017) on Youtube. (and note tube amplifiers in one of the recording studios in the movie
)
If you are ever in Brighton, MI pay a visit to HeadAcoustics to demonstrate it for you. For example one sales pitch usually entails walking across behind you while jingling keys then placing the keys in front of you and playing a recording back for you where you will be made aware of the spatial effect where you can still hear the keyset moving around you and point to its location.
Commercial audio will never see this as an average customer does not demand it. moreover the improvements on the recording end are not encouraged by the market either. Check out “The Art of Listening” (2017) on Youtube. (and note tube amplifiers in one of the recording studios in the movie
)
Last edited:
Op amps are not as interchangeable as some information on this forum might make you think. Op amp rolling is often touted as a way to improve a circuit, and often the op amp of choice is the OPA2134. But the fact is that the 2134 is kind of fussy, while the favorites of industry (5532 being the pinnacle in this world) not so much. Circuits have to be optimized for a given chip and a lot of commercial designs are just plain lazy and subpar; the 2134 will be underdamped if you just solder it in.
There's a reason that there's dozens of op amps that are popular for audio circuits. They're not all the same, at all. FET and BJT devices will give different performance in a given circuit. Some have better output stages than others. Some have lower input current than others. You have to look at all this stuff if you want to design an optimum circuit.
How go you know this is the case? Is it the case?
Did you read SY's article about audibility of opamps in series? That was a controlled test, and those are the only ones that have any value.
Everything else is personal, subjective and unrepeatable.
Jan
Last edited:
Somewhat off-topic, but: Does anybody know of properly conducted recent controlled listening tests where people were able to tell amplifiers apart, given a similar frequency response? The only ones I konw of are the very demanding pass-through tests that the Swedish Audio Society has done. But I'm not aware of other tests from recent years. It seems somewhat peculiar to me that psychoacoustic researchers conduct listening tests on loudspeakers all the time, but noone seems to bother to do such listening tests on amps.
I appreciate your view and respect your experience.
I’m not knowledgeable enough to confirm one way or the other, but it seems the most educated here all lean towards a similar outlook.
But given this, I’d like to pose a question:
(Jan please don’t take this as an underhanded dig, I’d simply like insight)
Just to choose an example, that is NOT an amplifier (and correct me if I’m wrong) Marcel VdG was asked to write an article for a LA and design a DAC using tubes. He believed that using solid state devices would provide the same performance. The tube DAC would cost a DIYer well north of 1K to construct. The solid state, much less.
(Marcel has contributed to this thread and will likely chime in, which is why I chose this example)
He states pretty conclusively that he can’t tell the difference and thinks the performance equal in the thread which originally launched the GB in the forum.
Is the reasoning that the tube article would “sell” or reach a different demographic LA might want attract?
He states in the article preface here:
“Although Marcel believes that excellent audio equipment can be made using solid-state technology, he considers it a nice intellectual challenge to see how far he could get with building a DAC that has all critical analogue and mixed-signal functions realized in valve technology. “
So, if it’s all the same, and snake oil and folk engineering pervades, why then request someone to publish material for an expensive, seemingly unnecessary and ecologically insensitive option when low cost, equal performance options are plentiful? Why sell access to something you don’t believe in? You are retired now I believe, so I expect it not to be a commercial consideration?
One has to ask themselves: why?
This can raise the same sort of trust issues that proponents of amplifier topologies frequently get into here, at least from the perspective of a person without vast comprehension of amplifier topologies. Entertainment and self knowledge purposes don’t seem to justify it. It seems to point to commercial considerations, as has some of the discourse in the Blowtorch thread and elsewhere.
Of course, it’s responsible as an editor to try and cull content for your readership. I’m not finger pointing. I also genuinely applaud LA for the wealth of incredibly interesting material it’s churned out. The very little I’ve managed to understand has really enriched my process.
Also, I say this as someone who has built some vacuum tube devices and really love them. Perhaps I like the distortion. If that’s the case, it’s fine by me. I don’t want to suffer in my leisure time in the pursuit of correctness. By the same token, part of me feels rather stupid building something so wasteful and vain when others tell me I can accomplish the same result as a more sensible, actualized human being. So, I’m not against tubes either, or against being totally wrong and taking up another angle- I’m wide open to suggestions.
I’m just connecting the dots here trying to understand, based upon your beliefs why suggest someone build such a thing?
If I have this right, all EE’s here should be:
A. Using the same amplifier since ~1985
B. Or maybe built a class D or T amp along the way to save on their electricity bill.
C. Focusing 100% of their energy on loudspeakers and DSP, where there seems to be universally agreed upon room to improve.
If this isn’t the case, there’s something for me that I either don’t understand or that doesn’t add up.
Why are people still publishing on preamps and dacs and amps?
I’m only humbly offering the perspective of someone here on the “outside” of things, as a dabbler and an introductory hobbyist without an EE background, and to perhaps illuminate for you how difficult it is to separate the wheat from the chaff with so many seemingly disjointed statements and views from both sides of the fence.
To me in these kinds of disputes it’s the novice who suffers and is turned away, because they lack the knowledge to make their own educated conclusions given the raw data.
I think this kind of uncertainty kills the DIY spirit for many, because they can’t get the first word settled: what should I DO. Then they wind up listening to the snake oil peddlers because they are much friendlier and more enticing- it’s part and parcel of the snake oil trade ;-)
DIY in my opinion is less an act of trust than an act of faith for beginners, and faith doesn’t just pop right out into the lap of the winner of a logical debate. There’s much more nuance at play which a PhD won’t do much for. For example: while I have come to enjoy DF96’s insights and his background would suggest to many he’s someone to take seriously, it’s not very surprising his conversion rate is low.
Why not beat the folk engineers at their own game rather than play into the trap?
Last edited:
Well, I can't answer for Jan, but consider me one of the ultra-skeptics on the forum. I actually think a lot of the EE's or objectivists or whatever the label-of-the-week-is are chasing psychoacoustic delusion. Why am I here? The hobby is no less interesting. I'd like to own both a '66 Chevelle and a C7 Corvette. I just don't try to tell myself the Chevelle is superior in some way. By analogy, 90% of the bandwidth of this forum is dedicated to just that.
Well, yes but I don’t necessarily agree. I see it regularly with a different mechanism of action.
Yes it’s especially unlikely when they can’t interpret those facts and figures or one comes across as hostile to their beliefs and they choose not to examine the information deeply.
Influencing and transforming belief systems is 100% possible, but facts and figures is the wrong playbook. Like trying to jam a floppy disk in a cd rom drive.... right intention, flawed execution.
Hello Koja, thanks for sharing your experience with binaural audio. I will check the art of listening after this post.
Perhaps the 'more feedback' all amplifiers sound the same' group thought about the first requirements, or laws of feedback.
1. You need more gain and a higher input signal (superposed with feedback)
2. The gain comes at a price, the higher you push the gain of a design, the higher the distortion before feedback is applied right?
3. Stability, more gain stages reduce bandwidth, requires also more filters for oscillation and parasitic, right?
4. Ok transistors don't require inverters, they have a 'free' gain stage, good, they still makes more distortion than tubes, most with a higher noise floor.
5. So more bad output will equal more quality of sound? answer is: no, see 6.
6. It cant because: any closed-looped amp injecting a feedback signal equal or higher than the source at the gain node will not sound better than the open-loop amplifier.
Conclusion, there are no miracles in nature because the closer you get to nature's sound the harder it gets. If you think multiplying an increasingly deteriorating signal will result in 'perfect' sound, maybe it works for satellite transmission, hard drives, and FM, but not for driving a complex speaker with xo's.
If you want to truth about feedback: feedback (well applied of course!haha) removes mostly efficiently 2n harmonics (pleasurable, harmless harmonics). The cancelation of the second harmonics by the injection feedback signal cause IMD with the input signal generating a 3rd harmonics content which destroys low level signals (yes, not only noise).
If a 'dead' sound is the goal, please do use as much feedback as possible. It will be a perfectly dead sounding amp.
Open you mind beyond THD numbers, there is much more there: low signal details, ability to track multiple signals, power supply which all sound different even in feedback amps.
Perhaps the 'more feedback' all amplifiers sound the same' group thought about the first requirements, or laws of feedback.
1. You need more gain and a higher input signal (superposed with feedback)
2. The gain comes at a price, the higher you push the gain of a design, the higher the distortion before feedback is applied right?
3. Stability, more gain stages reduce bandwidth, requires also more filters for oscillation and parasitic, right?
4. Ok transistors don't require inverters, they have a 'free' gain stage, good, they still makes more distortion than tubes, most with a higher noise floor.
5. So more bad output will equal more quality of sound? answer is: no, see 6.
6. It cant because: any closed-looped amp injecting a feedback signal equal or higher than the source at the gain node will not sound better than the open-loop amplifier.
Conclusion, there are no miracles in nature because the closer you get to nature's sound the harder it gets. If you think multiplying an increasingly deteriorating signal will result in 'perfect' sound, maybe it works for satellite transmission, hard drives, and FM, but not for driving a complex speaker with xo's.
If you want to truth about feedback: feedback (well applied of course!haha) removes mostly efficiently 2n harmonics (pleasurable, harmless harmonics). The cancelation of the second harmonics by the injection feedback signal cause IMD with the input signal generating a 3rd harmonics content which destroys low level signals (yes, not only noise).
If a 'dead' sound is the goal, please do use as much feedback as possible. It will be a perfectly dead sounding amp.
Open you mind beyond THD numbers, there is much more there: low signal details, ability to track multiple signals, power supply which all sound different even in feedback amps.
Last edited:
If you want the truth about feedback. It does not remove. It prevents distortion. It is really important to look at this as an engineer. Valves, transistors, transformers and even resistors are flawed. Feedback is used to fix these flaws.
There is no hostility towards people who prefer listening to or building any topology. I just can't understand where all the incorrect information comes from. The above posts are a good example. Which circuits get the predicate closer to nature or have a dead sound? This is ridiculous of course.
There is no hostility towards people who prefer listening to or building any topology. I just can't understand where all the incorrect information comes from. The above posts are a good example. Which circuits get the predicate closer to nature or have a dead sound? This is ridiculous of course.
Jan only asked me to design a DAC, it was my choice to make one with valves. The reason for that is that I considered it fun to gain some knowledge of and experience with valve oscillators and switching circuits (it's a hobby for me, so it should be fun - optimizing out every last cent while just meeting a long list of specs is work rather than hobby). To my opinion, the digital signal processing of the valve DAC is better than for most commercial DACs: most commercial DACs have no headroom whatsoever for intersample overshoots, mine has a selectable amount of headroom and a clipping neon lamp that warns the user when the headroom setting is too low.
There are a lot of misconceptions going on. People don't understand how feedback works, and its limitations. There is a lot of pontificating about the evils of feedback, which likely originates with someone who doesn't know how to use it.
There is also a notion of op amps being interchangeable, for the sake of comparison. Often rolling op amps results in disappointment, and the start of an internet rumor ("such and such op amp sounds veiled " It's oscillating fool!). Some op amps work fantastic in one type of circuit but won't work very well in other circuits.
My high feedback, op amp packed system sounds very "live" when I play live music. Reverberation, echoes, even boomy live bass sounds live. No second harmonics were deleted.
Ridiculous indeed.
Let me use this to answer other questions aimed at me above. The reason I asked Marcel for a DAC design was that there wasn't one published in the earlier issues and I like to cover all relevant audio fields. I left it to Marcel which design/how he wanted to do it. And I fully understand Marcel's reasoning: even if it does not lead to a better sounding DAC (which I believe it does not) it would be a very good opportunity to home in on the various functionalities, compromises and issues in a DAC design, and such investigating is something that a lot of tech people enjoy.
Which brings me to the next point. Contrary to what is suggested above, I doubt it that the average EE-type guy here tries to design an amp that sounds better than the previous amp. Most would be fully aware that competently designed and not overloaded amps pretty much sound the same. Rather, the challenge is to design a 'good' amplifier with different devices ('I want to try my hand at MOSFETs') or with less devices or with a novel circuit concept or smaller and/or cooler, etcetera. In other words, an intellectual challenge.
When an EE-type designs a new type of amp, the ONLY way to know for certain that you have reached your goal, or not, or what is still missing or wrong, is by measuring, measuring and measuring. With objective measurement equipment, that is consistent, reasonably calibrated, and that measures the same today as what it measured yesterday. In other words, those designers operate essential in 'scientific method' mode. Listening to the result can be part of it, but more of a confirmation type of check. If you amp measurements and tests show that it can almost perfectly reproduce whatever signal you throw at it at the speaker terminals under realistic loads, how can it sound anything else than 'good'?
And that is why they do not accept, on the face of it, reports that something sounds much better than something else. Because your hearing is about the most unreliable set of 'measurement equipment' you can think of. Influenced by many factors that have nothing to do with the sound, changing in sensitivity and focus all the time and having pretty much zero reliable acoustic memory.
Part of the eternal confusion comes from the fact that if you listen to music, you, ehh, listen to music. You enjoy the music, and people can enjoy music on anything from a 100k/pair speaker to the kitchen clock radio.
But when you want to listen to two amps trying to decide whether they sound differently, you only can do that reliably with some sort of controlled test. If you don't do that, your 'result' only has some worth to you personally, in that particular case and situation only.
So don't come to me saying that you replaced cap A for another one 'specifically designed for audio' and a veil was lifted, 'we' don't take that seriously. It is an extraordinary claim, and unless you provide an extraordinary iron clad proof, we just ignore it, at best.
More later.
Jan
Let me clarify. I am aware that for anyone listening, it does feel like you have a sort of direct line to the sound. Just as for you, for me it sounds like I really hear what's going on, I hear all details in my ear, so how can I be mistaken? I am not deaf!
But that goes past the hard fact that you do NOT hear what is in your ear. You become aware of a certain sound, and that awareness, that perception is constructed by your brain and uses a large and rich repertoire of data, of which the actual signal from your ears is just one, not even the largest. But we are not aware of that process, we just 'hear' .
You need to make a quantum jump in understanding how this works, and unless you 'get' that, you keep reading complete nonsense like 'I trust my ears'.
A well-known philosopher once commented that the history of science is the history of man learning to disregard the six senses and trust what his measurements and controlled observations are telling him.
A very good example is our growing understanding of the solar system. We started by 'they all turn around the earth, because you can clearly see it', to comprehensive understanding of the planets around the sun, the moons around the planets, the sun through the milky way, wobbles in trajectories from micro-variations in gravity, etcetera. In other words, and that is important, our understanding increased in step with our ability to disregard those 'obvious' things our senses told us.
Jan
But that goes past the hard fact that you do NOT hear what is in your ear. You become aware of a certain sound, and that awareness, that perception is constructed by your brain and uses a large and rich repertoire of data, of which the actual signal from your ears is just one, not even the largest. But we are not aware of that process, we just 'hear' .
You need to make a quantum jump in understanding how this works, and unless you 'get' that, you keep reading complete nonsense like 'I trust my ears'.
A well-known philosopher once commented that the history of science is the history of man learning to disregard the six senses and trust what his measurements and controlled observations are telling him.
A very good example is our growing understanding of the solar system. We started by 'they all turn around the earth, because you can clearly see it', to comprehensive understanding of the planets around the sun, the moons around the planets, the sun through the milky way, wobbles in trajectories from micro-variations in gravity, etcetera. In other words, and that is important, our understanding increased in step with our ability to disregard those 'obvious' things our senses told us.
Jan
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.