Are you Euphonophile?

[Blumlein 88]

Where does the idea come from that the application of crude, bent transfer functions to a composite recorded signal should, in any way, create new meaning in it, or add musical content to it, or change the performance, or change the recording venue? It is like claiming that an algorithm consisting of two lines of code has been developed that can automatically process poetry to make it better.

People who are seduced by this fantasy are clearly starting from systems that are deficient, and casting around for a magic bullet to do something, anything, to jazz up a lacklustre sound. The recent reviews of the Kii Three (just as an example), express the opposite view: that when the system is straight and lacking distortion, all such ideas evaporate.

I told you above, by listening to one of those euphonic systems. That is where the idea bent transfer functions can be euphonic.

I've told of this here before, but will do so again. I had a nice triode amplifier fed by a Meridian 563 (I think, I may have had the Wadia 25 by then) directly into the amp which fed Quad ESL63 speakers. You can decide if that is a system lacking and having lacklustre sound that needs jazzing up. The triode amps sounded very nice at least a couple of my friends thought it the best amp they had ever heard and a few more at least thought highly of it. It put up a very 3D sound, with lots of aliveness and dynamics, and seemed to be see thru transparent as if you were hearing the recording much more directly vs most other amps. Very pleasing and transparent which is an excellent combination.

So I decided to load the output with resistors similar to the load of the speakers, tap it with a resistor divider network to get unity gain vs input so I could feed it to another amplifier. My belief was the triode amp let more of the music through than most amps. But how much? Was it letting 50% thru, or 90% or 10%? And how much of the good quality of the recording did other amps lose? I hoped to get an idea of how much musicality was lost by this other amp as I knew the sound of the triode amp. My plans were then to reverse positions and try to gauge a little bit if the triode amp had much room for improvement or was nearly best possible already.

Big surprise however. I found that this triode amp feeding a Spectral DMA-50 was reproduced with all the qualities intact. The Spectral seemed to diminish it none whatsoever. That also told me the Spectral was fully capable of throwing up the 3D, spacious, dynamic, detailed, transparent sound that I had assumed was beyond its abilities.

Reversing positions, you couldn't tell if the Spectral was in the loop or not. It didn't diminish or change the sound of the triode amp following it. If you pulled it out and had only interconnects between source and triode there was no change in sound. It was actually the transparent high fidelity amp. While the Spectral sounded quite good on its own, it didn't have some of those other nice characteristics the triode did. But rather than being better in transmitting musical signals the triode was in fact adding its own delicious colorations. Those characteristics weren't in the source to start with. It sure was euphonic however.

So in that one experiment I found the Spectral had a transfer function that was straight, and the bent one sounded better. So the question is what sort of transfer function gives this sound? Some of it can be discerned because similar designs with similarly bent transfer functions sound similarly.

 

[andreasmaaan]

It apparently adds more richness, dynamics and space.

But starting with what sort of system? The lacklustre, deficient one I mentioned above, I should think...

In this study in which the distortion profiles of various opamps were modelled, the authors concluded (p. 176) that:

An important observation from both the op amp and capacitor studies is that listener preference was not necessarily reduced as a result of detectable levels of distortion. In listening tests, where participants could hear a shift in the character of the sound, they typically preferred the more distorted signal. The research in this dissertation began for the purposes of evaluating the sonic character of the distortion created by op amps and capacitors. The result that some amount of distortion from these components might improve listener perception of sound quality on certain musical sources was unexpected.

There is an in-depth discussion of this on pp. 176-179, including a review of past studies that have made similar findings, as well as hypotheses as to the reasons for the findings.

The author speculates that a preference for some level and type of distortion may result from the fact that distortion is a cue for loudness, and loudness is generally preferred:

If the harmonics produced by electronic components are similar to those typically produced in nature or within the human hearing system, a slightly nonlinear system may be perceived to be louder than a linear system with the same objective dynamic range. An increase in perceived loudness could be the primary reason for listeners' preference towards distorted signals in the listening tests...

But whatever the reason, there is decent evidence for the proposition that many listeners prefer a certain kind of "bent" transfer function to one that is not.

 

[andreasmaaan]

It depends on how you define distortion. Jean Hiraga or some old Stereophile guy are not the arbiters.

To me, distortion of the fundamental frequency response is the most basic form of distortion, though it is not typically called distortion. It is distortion, but it is usually just called frequency response. It might have been called frequency response distortion, and perhaps it once was. But, any deviation of output vs. input signal levels, aside from signal level amplification in volts and/or amperes, to me fits any classic definition of distortion.

I agree completely in terms of what "distortion" should, and technically does (although in practice almost never) mean. But the term "euphonic distortion" has never IME referred to anything other than nonlinear distortion.

I'm not saying Stereophile is the arbiter, it's just one example. I've never come across an example of linear distortion being described as "euphonic distortion".

But it's all semantics anyway. We agree that some distortion, linear or nonlinear, can be preferred by some people

 

[Cosmik]

I told you above, by listening to one of those euphonic systems. That is where the idea bent transfer functions can be euphonic.

I've told of this here before, but will do so again. I had a nice triode amplifier fed by a Meridian 563 (I think, I may have had the Wadia 25 by then) directly into the amp which fed Quad ESL63 speakers. You can decide if that is a system lacking and having lacklustre sound that needs jazzing up. The triode amps sounded very nice at least a couple of my friends thought it the best amp they had ever heard and a few more at least thought highly of it. It put up a very 3D sound, with lots of aliveness and dynamics, and seemed to be see thru transparent as if you were hearing the recording much more directly vs most other amps. Very pleasing and transparent which is an excellent combination.

So I decided to load the output with resistors similar to the load of the speakers, tap it with a resistor divider network to get unity gain vs input so I could feed it to another amplifier. My belief was the triode amp let more of the music through than most amps. But how much? Was it letting 50% thru, or 90% or 10%? And how much of the good quality of the recording did other amps lose? I hoped to get an idea of how much musicality was lost by this other amp as I knew the sound of the triode amp. My plans were then to reverse positions and try to gauge a little bit if the triode amp had much room for improvement or was nearly best possible already.

Big surprise however. I found that this triode amp feeding a Spectral DMA-50 was reproduced with all the qualities intact. The Spectral seemed to diminish it none whatsoever. That also told me the Spectral was fully capable of throwing up the 3D, spacious, dynamic, detailed, transparent sound that I had assumed was beyond its abilities.

Reversing positions, you couldn't tell if the Spectral was in the loop or not. It didn't diminish or change the sound of the triode amp following it. If you pulled it out and had only interconnects between source and triode there was no change in sound. It was actually the transparent high fidelity amp. While the Spectral sounded quite good on its own, it didn't have some of those other nice characteristics the triode did. But rather than being better in transmitting musical signals the triode was in fact adding its own delicious colorations. Those characteristics weren't in the source to start with. It sure was euphonic however.

So in that one experiment I found the Spectral had a transfer function that was straight, and the bent one sounded better. So the question is what sort of transfer function gives this sound? Some of it can be discerned because similar designs with similarly bent transfer functions sound similarly.

But I've heard similar tales about trying out different USB cables..!

 

[Blumlein 88]

But I've heard similar tales about trying out different USB cables..!

Well then like those USB cable aficionados maybe you should try it for yourself.

 

[Blumlein 88]

I think Vladimir Lamm has the best idea of how this works with tube amps anyway.

He wants 2nd or 3rd order HD and nothing else. He wants that distortion to rise nearly linearly with signal level. It should be around the inaudible level at your basic average sound levels and rise over the last 20 db or so of output. I forget whether it should max out at 2% or 3% in his opinion. It ideally should have the same FR at all power levels according to him though many xfmr coupled amps don't manage that. He believes they sound better if they come close to that ideal. He also believes that at each power level the distortion should be the same across the 20 khz bandwidth. Presumably this gets you the benefits of euphony without becoming noticeably colored in a bad sense. From hearing different bits of gear I think he is onto something.

I'm not sure how to separate out xfmr distortions from this, but they seem important too. If they do this at the recording end, we don't need any of it at the playback end. Emulator software is getting pretty good. I simply don't know if anyone's emulation is a serious attempt at creating the sound of a good push-pull triode or SET. My preference is push-pull btw. SET's always seemed a bridge too far.

Unfortunately, too much of all this is empirical at best.

 

[andreasmaaan]

I think Vladimir Lamm has the best idea of how this works with tube amps anyway.

He wants 2nd or 3rd order HD and nothing else. He wants that distortion to rise nearly linearly with signal level. It should be around the inaudible level at your basic average sound levels and rise over the last 20 db or so of output. I forget whether it should max out at 2% or 3% in his opinion. It ideally should have the same FR at all power levels according to him though many xfmr coupled amps don't manage that. He believes they sound better if they come close to that ideal. He also believes that at each power level the distortion should be the same across the 20 khz bandwidth. Presumably this gets you the benefits of euphony without becoming noticeably colored in a bad sense. From hearing different bits of gear I think he is onto something.

I'm not sure how to separate out xfmr distortions from this, but they seem important too. If they do this at the recording end, we don't need any of it at the playback end. Emulator software is getting pretty good. I simply don't know if anyone's emulation is a serious attempt at creating the sound of a good push-pull triode or SET. My preference is push-pull btw. SET's always seemed a bridge too far.

Unfortunately, too much of all this is empirical at best.

I've played around with a few VST emulators when mixing music. I've only found ones that emulate classic studio gear (not classic "audiophile" amps). Hard to say how good a jub they do without having used the original gear. Do you have any interesting recommendations?

In terms of giving very specific control over the type and extent of harmonics added, this one looks very good.

 

[Blumlein 88]

I've played around with a few VST emulators when mixing music. I've only found ones that emulate classic studio gear (not classic "audiophile" amps). Hard to say how good a jub they do without having used the original gear. Do you have any interesting recommendations?

In terms of giving very specific control over the type and extent of harmonics added, this one looks very good.

I don't have a good one to recommend. I wish Universal Audio would do one of the Manley triode amps. Then again it would cost too much if they did.

Supposedly Universal models circuit behavior even down to looking at physical arrangement of the gear to include coupling between transformers and other bits. Which would include effects which vary with signal level. I don't know how the others go about such things. Antelope's emulations seem to be pretty high quality, but none of tube amps. They do have some of tube compressors that seem about right. Just not the same effect however. Well actually Antelope does have some of guitar amps. Maybe I need to try those. I've mostly ignored them so far.

 

[Cosmik]

All of this disguises something: what is being recommended is not scientific, in that it is the 'measurement without theory' that was mentioned in another thread. Without a hypothesis or explanation, there is no science. Retro-fitting an explanation about human hearing to some listening test results is not science. If there is a genuine hypothesis, it must be tested, not with a selection of fixed boutique 'amplifiers', but in isolation as some lines of DSP code, for example.

But the irony is that once a system is set up to allow the hypothesis to be tested in isolation, the system is so 'straight' that anyone who hears it in its pure form is blown away by the sound and wouldn't dream of scuzzing it up with arbitrary bits of distortion. All these other anecdotes about the miraculous properties of transformers, or ex top secret soviet military technicians-turned-audiophiles and their 'theories', are starting from the position of a non-straight system.

Sure, one of my early teenage creations comprising a single full-range driver behind a grille of curtain material with crude transistor amplifier would sound so bad that I might cast around for some 'secret formula' to give it 'musicality'. Similarly if I want something to cut through the sound of the kids and washing machine. But a pure hi-fi system doesn't need any of that:

...it is dominated by such a sense of realistic clarity, imaging, dynamics and detail that you begin almost to forget that there’s a speaker between you and the music.
... every time I played some old familiar material I heard something significant that I’d never noticed or appreciated before. The review period became a near continuous frenzy of, “I wonder what that will sound like.”

To say to the author of that piece: "I could improve that experience with two lines of DSP code or a resistor and diode." would be ridiculous. "Of course I would destroy some of the imaging, clarity and detail, but who needs those? Instead I would be applying top secret military research (so secret it has never been published!), tapping directly into your psychoacoustic pleasure centres by scientifically synthesising an intoxicating spray of background sizzle and discordance, filling in those awkward spaces surrounding the instruments, imbuing the sopranos' voices with the rich phlegm-like gargle and wheeze that they are so badly missing."

 

[Cosmik]

If I have a composite recording of two sources, then just as in real life I can perceive, track, monitor those two sources separately in my brain. And if this is aided by the time-of-arrival cues of a stereo system (as discussed in another thread), this is even more compelling.

What is being proposed with 'euphonic distortion' is to add a third 'source' that, if listened to in isolation would be a sizzling, shrieking, discordant noise, crudely mimicking the two genuine sources and not emanating from a stable position in the stereo field. Forgive me for questioning the wisdom of this!

 

[Blumlein 88]

snippage

But the irony is that once a system is set up to allow the hypothesis to be tested in isolation, the system is so 'straight' that anyone who hears it in its pure form is blown away by the sound and wouldn't dream of scuzzing it up with arbitrary bits of distortion. All these other anecdotes about the miraculous properties of transformers, or ex top secret soviet military technicians-turned-audiophiles and their 'theories', are starting from the position of a non-straight system.

snippage

Again, that all reads fine, except, well that isn't what happened.

I owned the Spectral amp and used it for a couple of years. Even after hearing it and taking some time to grok what my experiment meant, I liked the scuzzed up version (which btw despite your imaginative casting the sound in a negative light is not how it sounds). I'd let other people use my amp, they liked it. I let other people hear both versions which was surprising to them, but they liked the "scuzzed up" version. So your thought experiment fails the reality test.

 

[Blumlein 88]

If I have a composite recording of two sources, then just as in real life I can perceive, track, monitor those two sources separately in my brain. And if this is aided by the time-of-arrival cues of a stereo system (as discussed in another thread), this is even more compelling.

What is being proposed with 'euphonic distortion' is to add a third 'source' that, if listened to in isolation would be a sizzling, shrieking, discordant noise, crudely mimicking the two genuine sources and not emanating from a stable position in the stereo field. Forgive me for questioning the wisdom of this!

As long as you create straw man descriptions of the euphonic sound you'll continue failing to understand what people are experiencing.

Firstly because it isn't listened to in isolation. Secondly because it isn't shrieking, discordant and sizzling, nor is it a third source, it is always related to primary sources coming and going as they do with a direct relationship to them.

I also don't think at this point euphonophiles are claiming science of the idea exactly. So damning the idea for something it hasn't yet claimed doesn't make any sense. Yes some additional investigation might prove fruitful or might prove this is another psycho-acoustic dead end.

 

[Cosmik]

Firstly because it isn't listened to in isolation. Secondly because it isn't shrieking, discordant and sizzling, nor is it a third source, it is always related to primary sources coming and going as they do with a direct relationship to them.

Just assertions divorced from the reality of what is going on. A signal comprising two pure sine waves can be perfectly dissected into the original sine waves. But 'something' is being added to those two sources. The 'something' is, by any definition, a third source. It is discordant (it probably has no harmonic relation to either of the individual sine waves). On a more complex signal, IMD sizzles and shrieks - just think of what characterises an old 78 recording of an orchestra.

The trouble starts when people think of 'the signal' as the entity that is being listened to. It isn't: the listener is tracking and monitoring the continuity and/or separation of the sources within that composite wiggly line. No one knows what can be contained within that signal: a sine wave, two square waves, a single voice, a choir, a symphony orchestra, a drum. There is no single prescription of 'a bent transfer function' that can have universal relevance to all those possibilities. It is just meaningless rubbish being plastered over the sound.

Only by starting from the idea that the signal is a singular thing - a stream of flavoured paste - can it be argued that a bent transfer function is anything more than a source of irrelevant rubbish.

 

[Blumlein 88]

snippage..........

Only by starting from the idea that the signal is a singular thing - a stream of flavoured paste - can it be argued that a bent transfer function is anything more than a source of irrelevant rubbish.

More than 99.9% of all recordings you can possibly get to listen to have very bent transfer functions. More than 99% of them have horribly bent, disfigured, multi-level numerous sources of bending and breaking the transfer function. Does that mean that more than 99.9% of all recordings are irrelevant rubbish? Or does the un-scientific sourcing invalidate the entire approach of fidelity? Following your lead the answer is yes. I don't think so.

 

[Cosmik]

More than 99.9% of all recordings you can possibly get to listen to have very bent transfer functions. More than 99% of them have horribly bent, disfigured, multi-level numerous sources of bending and breaking the transfer function. Does that mean that more than 99.9% of all recordings are irrelevant rubbish? Or does the un-scientific sourcing invalidate the entire approach of fidelity? Following your lead the answer is yes. I don't think so.

I can't help what has gone into the creation of the signal, but in the case of a purist recording the distortion of the transfer function and therefore the amount of irrelevant rubbish is at least as small as possible. In the case of rock/pop/studio-created recordings, the producers and musicians may have used bent transfer functions and other techniques for creative reasons - but they have access to the individual sources/mics/feeds/tracks and so their distortion may genuinely be harmonically related to the sources if that's what they choose, or not.

What I can help, is how much extra rubbish I add. It used to be called high fidelity, but that seems to be a quaint notion now!

It seems fundamental to me (so to speak!) that the singular signal is a 'multiplex' of multiple sources, and my brain (just as it does with composite pressure waves reaching my ears in real life) is 'demultiplexing' the sources. Synthesising extra 'stuff' (even worse that it is derived from some arbitrary, changing, arithmetic relationship between the sources rather than being independent) cannot do anything but hinder that process.

 

[svart-hvitt]

I can't help what has gone into the creation of the signal, but in the case of a purist recording the distortion of the transfer function and therefore the amount of irrelevant rubbish is at least as small as possible. In the case of rock/pop/studio-created recordings, the producers and musicians may have used bent transfer functions and other techniques for creative reasons - but they have access to the individual sources/mics/feeds/tracks and so their distortion may genuinely be harmonically related to the sources if that's what they choose, or not.

What I can help, is how much extra rubbish I add. It used to be called high fidelity, but that seems to be a quaint notion now!

It seems fundamental to me (so to speak!) that the singular signal is a 'multiplex' of multiple sources, and my brain (just as it does with composite pressure waves reaching my ears in real life) is 'demultiplexing' the sources. Synthesising extra 'stuff' (even worse that it is derived from some arbitrary, changing, arithmetic relationship between the sources rather than being independent) cannot do anything but hinder that process.

Is dither «euphonics»?

 

[Theo]

As a non scientific hypothesis (I can't demonstrate nor falsify....), would an added distortion give a false sense of depth or space by adding elements related to the signal in a stereo random looking pattern, like reverb would do? Tube afficionados usually pretend their amp produces a 3D image...
BTW, let's not forget that a significant number of people do prefer listening to "euphonic" equipment, so even if not technically demonstrated, can't this be discussed as a sociologicaly relevant scientific observation?

 

[Wombat]

Anything other than an accurate transcription is a subjective example.

 

[andreasmaaan]

If there is a genuine hypothesis, it must be tested, not with a selection of fixed boutique 'amplifiers', but in isolation as some lines of DSP code, for example.

The study I posted did use DSP-processed signals. There are detailed graphs showing the effects of the processing on each signal, including the baseline unprocessed signal (pp. 190-209).

I agree though that research should be undertaken to repeat the results and to further investigate the reasons for preferences in favour of distortion. The current hypotheses proposed are no more than speculation (but that does not change the findings).

But the irony is that once a system is set up to allow the hypothesis to be tested in isolation, the system is so 'straight' that anyone who hears it in its pure form is blown away by the sound and wouldn't dream of scuzzing it up with arbitrary bits of distortion.

That's an argument based on your own experience which is contradicted by the study I posted, which uses exactly the kind of processing you insist must be used.

All these other anecdotes about the miraculous properties of transformers, or ex top secret soviet military technicians-turned-audiophiles and their 'theories', are starting from the position of a non-straight system.

Nothing I'm saying relies on any "evidence" of this kind.

 

[Blumlein 88]

I can't help what has gone into the creation of the signal,

No you can't help that.

but in the case of a purist recording the distortion of the transfer function and therefore the amount of irrelevant rubbish is at least as small as possible.

But those purist recordings are not even .1% of all that is available. Probably not even .01%. So how often does this idea of yours even come into play?

In the case of rock/pop/studio-created recordings, the producers and musicians may have used bent transfer functions and other techniques for creative reasons - but they have access to the individual sources/mics/feeds/tracks and so their distortion may genuinely be harmonically related to the sources if that's what they choose, or not.

Sorry, rose colored glasses. The processes used, even in a relatively unprocessed recording are way beyond the bit of THD and IMD an amp adds to it.

What I can help, is how much extra rubbish I add. It used to be called high fidelity, but that seems to be a quaint notion now!

It seems fundamental to me (so to speak!) that the singular signal is a 'multiplex' of multiple sources, and my brain (just as it does with composite pressure waves reaching my ears in real life) is 'demultiplexing' the sources. Synthesising extra 'stuff' (even worse that it is derived from some arbitrary, changing, arithmetic relationship between the sources rather than being independent) cannot do anything but hinder that process.

The process is way beyond simple hindering in the vast majority of recordings. It is indeed a quaint idea. One I fully support. But it rarely comes into the picture. Very, very, rarely. We have the potential now to provide end to end transmission of the original recording, and yet damnably it is so close to never happening that the difference isn't worth bothering with.