Can You Trust Your Ears? By Tom Nousaine

[FrantzM]

Hi

These discussions are philosophical in nature. They also point toward our fear of the unknown. It is an uneasy feeling to ponder the depth of our ignorance or worse our fallibility. Our ears are the devices through which we perceive sound. As for many of our limbs and perceptual organs they need to be trained to perform well. Not "used', trained . Untrained our ears are poor. Trained they become better but remain inaccurate.
As an audiophile, I have had to face my relatively plebeian hearing. I have had for many years believed that my ears were superior. It dawned to me around 2006 that they weren't. I have post this elsewhere but bear with me and let me repost it . I was in a room with a CRT TV (remember those? ) and there was no picture on it. It looked off and I was relaxing .. A young cousin perhaps 16 years old then came to the room and told me the TV was on .. I looked at the CRT, no picture and there wasn't any indication of TV being on... the cable box ws of course off... She insisted the Tv was on and proceeded to turn it off ... Well, she heard the TV from about 20 feet .. the flyback transformers tend to emit some sounds in the over 16 KHz range and for her it was pretty loud , for me it was entirely absent. I did not hear it at all. It was the beginning. a few years later there was a test of an audiophile comparing his super audio speaker cables with run of the mill cables. He couldn't tell them apart. I on my side was sure he would or that I would if I performed similar test ... I failed miserably.. Nothing scientific. Had a few friends, we level matched grossly and compared 6 AWG electrical wire with my seriously audiophile cables and I couldn't tell them apart either .. Later I tested myself on various software amongst them the foobar ABX ... Failed often to distinguish 256 KBps mp3 from CD .. Finally downloaded Harman "How to Listen" and am able to distinguish some of the artifacts.
Audiophile are often engaged in other activities involving the notion of "best". Be it car or wine or food, etc .. There is a search. I have some audiophile friends that would not use anything but a full -frame camera ( that includes me BTW) but can hardly tell the difference between a picture taken with a full frame and and a half-frame. The same people will talk about the special picture a Leica camera produces, of course once they know it was taken with a Leica... I have a photographer friend who played a cruel joke on some of us with pictures taken with an iPhone 7 Plus ( talk about a seriously good camera in this smartphone!!! ???). He was on the countryside and had only , of course, his cellphone .. and took some superb pictures and proceed to enlarge some of them ... We all thought it had to be his his Hasselblad or Nikon !! Well it wasn't ...

The point of all this writing is that this is a much more unsettling for most audiophiles to accept that our vaunted hearing is ...not. In the absence of serious measurements our hearing is what we have to use. Consider training our ears/hearing to make them/it more accurate. That is all we can do. It is thus IMO wiser not to trust them.

 

[Blumlein 88]

Sounds like you trusted and trusted and trusted. Then events pricked your certainty. So you decided to verify. Trust, but verify. In time trust, but verify only verified you have to be careful trusting your hearing.

Training and education lead to the truth.

Congratulations as so many aren't ready for the truth.

Excellent post.

 

[Cosmik]

Without any scientific proof whatsoever, my experience tells me that like most of our senses, hearing is as much a change detector as it is a sensor of absolute levels etc. Stare at an image for long enough and it will fade away; listen for long enough to a system with a poor frequency response and after a while it will seem normal - but it will take quite a long time to get there, and will be reset back towards normality if you listen to other things at other times. Time constants from minutes to months, I would guess.

If DSP can do it, I should imagine that our hearing probably can, too. So it could maybe even learn to adapt to nonlinear distortion (to some extent), or correct phase errors in speakers, but I would guess it would be a painful transition to get there and maybe never completely effortless unless learned from birth. Exposure to 'normality' would constantly tend to reset the training.

 

[Jakob1863]

I think the issue here is a matter of scale.
Sighted listening can detect differences of the order 5-10dB.
Blind testing is needed to reliably detect smaller differences.

It depends......
Even if not taken literally, "we" know that people can reach incredible sensitivity under the conditions of a "blind" controlled listening test (in this context i mean tests that are well planned, well done and well documented).
Otoh "we" know too, that a lot of people do not detect even quite large differences before getting used to the certain conditions of a "blind" controlled listening test.

If there is a real difference between "sighted" and "blind" , we only know if it was already researched, was it?
Based on my personal experience, which is exclusively related to difference that matter in everyday life, they do both quite well, but it is much more difficult doing "blind" controlled listening tests....

 

[fas42]

If DSP can do it, I should imagine that our hearing probably can, too. So it could maybe even learn to adapt to nonlinear distortion (to some extent), or correct phase errors in speakers, but I would guess it would be a painful transition to get there and maybe never completely effortless unless learned from birth. Exposure to 'normality' would constantly tend to reset the training.

Perhaps for some people. For me, FR and phase are not an issue, but underlying, disturbing distortion artifacts are - I find it impossible to listen to PA sound unless the energy of the music is sufficiently high to offset that, and at a more subtle level conventional audiophile systems very quickly reveal their gremlins - with the latter the longer one listens the more aware one is of the nonlinear deficiencies; the conventional phrase for this is "listener fatigue". These are the artifacts which, when one is about to enter a room with music happening, make one 100% certain that that it's a hifi working, without sighting anything.

 

[fas42]

It depends......
Even if not taken literally, "we" know that people can reach incredible sensitivity under the conditions of a "blind" controlled listening test (in this context i mean tests that are well planned, well done and well documented).

The sensitivity that counts is often the somewhat unconscious one - when the quality of what one is listening to makes one uncomfortable, you want to turn the volume down, change to a 'better' track, change to a 'better' recording - live acoustic music draws one in, you want to experience the impact even more intensely; typical audio reproduction is more often "how much of this can you take?".

 

[Jakob1863]

The sensitivity that counts is often the somewhat unconscious one - when the quality of what one is listening to makes one uncomfortable, you want to turn the volume down, change to a 'better' track, change to a 'better' recording - live acoustic music draws one in, you want to experience the impact even more intensely; typical audio reproduction is more often "how much of this can you take?".

It´s something like awareness to grasp a feeling for the reproduction overall. Concentrating to soon on narrow aspects leads quite often to reduced sensitivity.
But, as other already pointed out, training for evaluation purposes helps, getting used to the certain conditions of a "blind" controlled listening test is extremely important, otherwise chances are quite high that you don´t get correct results.

 

[Jakob1863]

Hmmm... I am an engineer, and some say a pretty good one, but I do not think I have seen as much talk about philosophical stuff and engineers since college days of post-bar-closing early morning drink-a-thons after a big test or finals. Pretty sure most of us do not think about "neo-positivist tradition" and such when doing our designs. A cross section of audio theory both technical and practical should include some background on hearing and perception, but at the end of the day I'm likely to be working from a data sheet to realize a design, then don't get close to the whole "how do we hear" debate until listening tests and tweaks driven as much or more by listening panels and marketing than engineering.

Anyway, this one's over my head and beyond the ken of my little pea brain, have at it! - Don

Obviously in most cases it doesn´t harm if engineers don´t think about the philosophical foundation of science, although our way to work is based on that.
But in these audio discussions about "debateable effects" it seems imo that a lot of engineers don´t realize that they are discussing things that don´t naturally belong to their knowledge. Test theory, or statistics for behavioral sciences or cognitive psychology.
Audio reproduction is from a technical point of view a sort of a broken system - when comparing the reproduced sound fields to the original ones.
It works, but everthing we know about the direction of development ist based on perceptual evaluation and therefore we have to know about philosophy of science because it matters. Or as Bech and Zacharov wrote:

"Almost everyone listens to soundmost of the time, so there is often
an opinion that the evaluation of audio qualitymust be a trivial matter.
This frequently leads to a serious underestimation of the magnitude of
the task associated with formal evaluations of audio quality, which can
lead to compromised evaluations and consequently the poor quality
of results. Such a lack of good scientific practise is further emphasised
when results are reported in journals or at international conferences
and leads to a spread of scientific darkness instead of light."
(Source: Bech, Zacharov; Perceptual Audio Evaluation–Theory, Method and Application, 2006, John Wiley & Sons, page xii)

 

[amirm]

"Almost everyone listens to soundmost of the time, so there is often
an opinion that the evaluation of audio qualitymust be a trivial matter.
This frequently leads to a serious underestimation of the magnitude of
the task associated with formal evaluations of audio quality, which can
lead to compromised evaluations and consequently the poor quality
of results. Such a lack of good scientific practise is further emphasised
when results are reported in journals or at international conferences
and leads to a spread of scientific darkness instead of light."
(Source: Bech, Zacharov; Perceptual Audio Evaluation–Theory, Method and Application, 2006, John Wiley & Sons, page xii)

That's the preface to sell the book so it is not all innocent itself either.

Here is the best motivation to do such tests right: make your livelihood depend on it as it did for me for a decade. You stop fooling yourself in a hurry. I had to analyze competitor tests to find weaknesses in them that biased their results, and they did the same to us. And I had to deal with reviewer tests in addition to test fueling our own development. And of course knowing how the industry as a whole conducts such research.

Yes, doing good tests is hard. But the alternative of giving up and relying on uncontrolled, subjective testing is just a disaster.

 

[Cosmik]

Obviously in most cases it doesn´t harm if engineers don´t think about the philosophical foundation of science, although our way to work is based on that.
But in these audio discussions about "debateable effects" it seems imo that a lot of engineers don´t realize that they are discussing things that don´t naturally belong to their knowledge. Test theory, or statistics for behavioral sciences or cognitive psychology.
Audio reproduction is from a technical point of view a sort of a broken system - when comparing the reproduced sound fields to the original ones.
It works, but everthing we know about the direction of development ist based on perceptual evaluation and therefore we have to know about philosophy of science because it matters. Or as Bech and Zacharov wrote:

"Almost everyone listens to soundmost of the time, so there is often
an opinion that the evaluation of audio qualitymust be a trivial matter.
This frequently leads to a serious underestimation of the magnitude of
the task associated with formal evaluations of audio quality, which can
lead to compromised evaluations and consequently the poor quality
of results. Such a lack of good scientific practise is further emphasised
when results are reported in journals or at international conferences
and leads to a spread of scientific darkness instead of light."
(Source: Bech, Zacharov; Perceptual Audio Evaluation–Theory, Method and Application, 2006, John Wiley & Sons, page xii)

And yet a DAC is still a device that precisely reproduces the waveform fed into it; an amplifier likewise; and the basic objective of speakers (although usually obscured by the problems associated with antiquated technology) is the same - the recent Kii Three gets closest to 'straight wire with gain'.

All this talk of science, psychoacoustics and listening tests is just superfluous. The engineers of the 1930s/40s/50s would have gone for digital audio, solid state amps, DSP crossovers and all the rest like a shot if they could have. There is no secret psychoacoustic golden formula that somebody is going to stumble upon by cobbling together permuations of valves, vinyl and passive crossovers. It is like trying to improve a television picture by interposing various layers of coloured, rippled glass. There is only one type of glass that is optimal: the clear, colourless one. There is no need to perform viewing tests to determine it, nor any reason to make the window from diamond.

 

[Jakob1863]

@ amirm,

That's the preface to sell the book so it is not all innocent itself either.

Maybe, but the more important point isn´t about innocence but about being correct or not.
And, having read a lot of publications, i think they are right with their statement.

<snip>Yes, doing good tests is hard. But the alternative of giving up and relying on uncontrolled, subjective testing is just a disaster.

Yes, doing good tests is hard. But the alternative of taking the easy road and relying on bad tests is just a disaster.
The class of tests, where someone tries to proof the socalled "audiophile" wrong, belongs quite often to the second variant...

PS. I´ll try to address your last longer posts tomorrow.

@ cosmik,

And yet a DAC is still a device that precisely reproduces the waveform fed into it; an amplifier likewise; .......

Whatever "precisely" means in this context, as i´d argue that DACs and amplifiers do that sort of "precise" reproduction since a long time ago, but nevertheless were quite often not transparent wrt audio quality.
That´s what i meant earlier, as i remembered the different notions of engineers wrt reproduction quality over the decades.
The argumentation was always nearly identical and of course if one constantly says "no difference can´t be heard" for 50 - 100 years, he will someday right. It´s a bit like telling that the end of the world is near.

All this talk of science, psychoacoustics and listening tests is just superfluous. The engineers of the 1930s/40s/50s would have gone for digital audio, solid state amps, DSP crossovers and all the rest like a shot if they could have. There is no secret psychoacoustic golden formula that somebody is going to stumble upon by cobbling together permuations of valves, vinyl and passive crossovers. It is like trying to improve a television picture by interposing various layers of coloured, rippled glass. There is only one type of glass that is optimal: the clear, colourless one. There is no need to perform viewing tests to determine it, nor any reason to make the window from diamond.

These are hypothesis and i´m fine with that, but if we don´t want to confuse plausibility with evidence something is missing and imo at that point the "superfluous" talk about science/psychoacoustics and listening tests" comes into play, or more precisely ( scr ) _knowledge_ about ...../.....and.... is important.

 

[Cosmik]

These are hypothesis and i´m fine with that, but if we don´t want to confuse plausibility with evidence something is missing and imo at that point the "superfluous" talk about science/psychoacoustics and listening tests" comes into play, or more precisely ( scr ) knowledge about ...../...../.... is important.

Yes, but do you have some examples where psychoacoustics and listening tests have come up with something beyond the idea of 'straight wire with gain'?

I can suggest one: the 'loudness' button. Maybe this is a rare example where science based on listening tests could have been said to have helped create a 'psychoacoustic compensator'. But the listening tests didn't use music, and the characteristics of the button could be worked out straightforwardly from those results. And manufacturers dropped the idea after a few years, simply reverting to the SWWG system.

The idea of the button was to get closer to perceived SWWG given that people are often forced to listen at less than optimal volume levels. Scientific listening tests might (after much pain) show that the button worked moderately well with some recordings and playback levels, but not others. However, a person could probably work the same answer out for you in just a few moments of thinking about it.

 

[Fitzcaraldo215]

Yes, but do you have some examples where psychoacoustics and listening tests have come up with something beyond the idea of 'straight wire with gain'?

I can suggest one: the 'loudness' button. Maybe this is a rare example where science based on listening tests could have been said to have helped create a 'psychoacoustic compensator'. But the listening tests didn't use music, and the characteristics of the button could be worked out straightforwardly from those results. And manufacturers dropped the idea after a few years, simply reverting to the SWWG system.

The idea of the button was to get closer to perceived SWWG given that people are often forced to listen at less than optimal volume levels. Scientific listening tests might (after much pain) show that the button worked moderately well with some recordings and playback levels, but not others. However, a person could probably work the same answer out for you in just a few moments of thinking about it.

I have been thinking about the same question. Another more fundamental answer is why are many component specs limited to 20-20k Hz, or slightly more? Obviously, we can limit much testing to this empirically determined frequency range for most humans, rather than testing into the GHz range or going below 20Hz on the misguided assumption that ultra-or infra-sonic frequencies matter. We have scientifically determined the limits of human hearing empirically, so we can focus our energy on a specific, relevant frequency band.

Agreed about the Loudness button, and yes it has disappeared because it was poorly applied and quite useless. But, there is an interesting twist as of 5 or more years ago. Audyssey introduced a Dynamic EQ feature, based on their research which parallels the famous Fletcher-Munson curves. Basically, in an Audyssey calibrated system with this feature turned on, the amount of bass and treble boost is proportionately, seamlessly and innocuously varied based on the volume control setting done by the user. This is unlike the old Loudness button, which stupidly applied a fixed amount of bass and treble boost no matter what.

Unfortunately, the Audyssey scheme required reliance on the standardized average volume levels used in the movie industry, which do not exist for music recording. Ergo, it will not work well for music, and, I did not use it in my old Audyssey days. Some movie fans thought it was terrific, though.

 

[fas42]

And yet a DAC is still a device that precisely reproduces the waveform fed into it; an amplifier likewise; and the basic objective of speakers (although usually obscured by the problems associated with antiquated technology) is the same - the recent Kii Three gets closest to 'straight wire with gain'.

All this talk of science, psychoacoustics and listening tests is just superfluous. The engineers of the 1930s/40s/50s would have gone for digital audio, solid state amps, DSP crossovers and all the rest like a shot if they could have. There is no secret psychoacoustic golden formula that somebody is going to stumble upon by cobbling together permuations of valves, vinyl and passive crossovers. It is like trying to improve a television picture by interposing various layers of coloured, rippled glass. There is only one type of glass that is optimal: the clear, colourless one. There is no need to perform viewing tests to determine it, nor any reason to make the window from diamond.

This is not the case at all. If one decides to become sensitive to how much the sound quality does vary, depending on the innumerable variables in the chain, then it becomes obvious that 'straight wire with gain' is extremely difficult to achieve; most importantly, in the areas where the impact is audible. The DAC and amplifier are only precise in the highly limited areas of behaviours that the industry has chosen to measure, and these metrics fall far short of characterising all the behaviours that matter, to the ears.

And, psychoacoustics is everything. The secret is extremely simple: eliminate all the audible artifacts - trouble is, the majority of people have never experienced a system that is capable of such "transparency", or if they chance upon one it goes right over their heads - they just "don't get it". So, if one is not attempting to reach such a goal, it's virtually guaranteed that it will never happen ...

The television picture of conventional audio may be colourless in some instances , but it's certainly not clear - it's actually a type of frosted glass; somewhat subtle, good enough to get a general sense of what's going on; but all the fine detail is completely obscured.

 

[fas42]

It´s something like awareness to grasp a fealing for the reproduction overall. Concentrating to soon on narrow aspects leads quite often to reduced sensitivity.
But, as other already pointed out, training for evaluation purposes helps, getting used to the certain conditions of a "blind" controlled listening test is extremely important, otherwise chances are quite high that you don´t get correct results.

Simplest technique is to "listen without listening". Turn up the volume, and deliberately turn your attention to something else, say, talking to the person next to you, and fully concentrate on the conversation. Poor reproduction, not matter how well it measures conventionally, will have your insides screaming in a very short time frame - you will have to run to the system and hit the mute button, to get some relief ...

 

[fas42]

I was just struck by a thought then, about the television picture being presented through a glass which is "clear, colourless" - we have a car here where the windscreen had wipers of very poor quality on at one stage, or they were operated for extended times in dry conditions, when there was a coating of dust on the surface. Result, a fine pattern of scratch lines cover the complete area of vision ... now, it would be extremely easy to take numerous measurements to show that the glass was clear and colourless, perfect in the optics if I used the right lighting conditions, etc, when getting some numbers. Yet, at night, when there are lights coming strongly towards one, from various angles, and during the day when the sun is vaguely forward of me - the seeing is terrible! The 'right' conditions for assessing the quality of the windscreen show it to be in terrible shape, when it may seem perfectly fine on a cursory inspection.

This is a pretty reasonable analogy to much audio replay - seems OK in the 'right' circumstances, but doesn't take much variation from these "ideal" conditions to present quite poorly ...

 

[DonH56]

I am a little surprised specs have not crept down if not up. Few past the age of 25 can hear over 20 kHz, but especially with the HT revolution and numerous subwoofers available sub-20 Hz performance is not at all uncommon. On the high end there is also the CD roll-off so not much point in going beyond 20 kHz (and yes I know about higher sampling rates, and a fair amount about the arguments for retaining ultrasonic content).

I have toyed with the loudness on/off button on my Emotiva XMC-1. I tend to like it when listening to music at lower levels though it does tend to over-emphasize the bass. But the sound is really anemic without it, and it works as well or better than a lot of earlier loudness controls (some of which were frankly hideous IMO). At least the dynamic implementations of today work much better than the single button we had in the past.

IME/IMO - Don

 

[Cosmik]

I was just struck by a thought then, about the television picture being presented through a glass which is "clear, colourless" - we have a car here where the windscreen had wipers of very poor quality on at one stage, or they were operated for extended times in dry conditions, when there was a coating of dust on the surface. Result, a fine pattern of scratch lines cover the complete area of vision ... now, it would be extremely easy to take numerous measurements to show that the glass was clear and colourless, perfect in the optics if I used the right lighting conditions, etc, when getting some numbers. Yet, at night, when there are lights coming strongly towards one, from various angles, and during the day when the sun is vaguely forward of me - the seeing is terrible! The 'right' conditions for assessing the quality of the windscreen show it to be in terrible shape, when it may seem perfectly fine on a cursory inspection.

This is a pretty reasonable analogy to much audio replay - seems OK in the 'right' circumstances, but doesn't take much variation from these "ideal" conditions to present quite poorly ...

I think you may have misunderstood what I was getting at. It may be the case that a particular system is defective under difficult circumstances, but if so, the problem is that it is deviating from 'straight wire with gain' - which is a simple engineering criterion.

I would like to know if, and where, scientific listening tests and knowledge of psychoacoustics have been used in order to get to something better than SWWG (for a hi-fi playback system). We see comments all the time that allude to psychoacoustics and the listening tests that serious audiophiles are allegedly carrying out. I am sceptical to say the least!

It is like a manufacturer of USB cables attending night school classes in English literature in the hope that this will result in more emotionally-moving printouts.

 

[Fitzcaraldo215]

I think you may have misunderstood what I was getting at. It may be the case that a particular system is defective under difficult circumstances, but if so, the problem is that it is deviating from 'straight wire with gain' - which is a simple engineering criterion.

I would like to know if, and where, scientific listening tests and knowledge of psychoacoustics have been used in order to get to something better than SWWG (for a hi-fi playback system). We see comments all the time that allude to psychoacoustics and the listening tests that serious audiophiles are allegedly carrying out. I am sceptical to say the least!

It is like a manufacturer of USB cables attending night school classes in English literature in the hope that this will result in more emotionally-moving printouts.

SWWG carries with it many possible interpretations. Real world audio systems do depart from SWWG frequency response at some point, but that is OK outside the normal range of human audibility, nominally 20-20k, which had initially to be determined empirically before that particular frequency range could be widely accepted. That seems a very fundamental, early part of psychoacoustics. We accept it today without thinking, but there was a time that was not so.

I think of psychoacoustics as being synonomous with the science of perceptual acoustics, which focuses on measuring human responses to audio stimuli, as opposed to measuring audio signals themselves electrically or acoustically.

But, you were apparently thinking more about alterations to flat frequency response or other alterations to the signal within the audible band indicated by psychoacoustics. Another example would be downward sloping acoustic target frequency response curve design for DSP EQ of rooms and auditoriums.

Another example is the current flap over MQA, based largely on findings about the correction of perceptual "time blurr" caused by a-d and d-a. Many traditional audiophiles seem unwilling to accept this concept. Perhaps MQA could do a better job of communicating their experiments and findings supporting their view. Or, perhaps, many audiophiles are just much more resistant to perceptual findings like this as opposed to traditional SWWG engineering concepts.

 

[Jakob1863]

Yes, but do you have some examples where psychoacoustics and listening tests have come up with something beyond the idea of 'straight wire with gain'?

Why should we narrow it to this point?

A lot of people seem to be interested in listening tests and you need some knowledge of psychoacoustic results and techniques to avoid repeating the same errors again and again. See for example the mentioning of an amplifier listening test in another thread where the experimenter noticed that a bias took place in favour of the second amplifier in the presentation.
Something like that is called a "time error" (if the first presentation is preferred it´s called a positive time error, if the second presentation is preferred it´s called a negative time error) and is known to exist since Fechner´s experiments around the 1870s . At first noticed wrt to single dimensional stimuli and later even for esthethical/preference decisions, but up to now not fully explored especially wrt multidimensional decisions.

Another example would be the incorporation of money prize or the experimenter´s influence. To do _good_ experiments one needs to know something about psychoacoustics and test theory and statistics as well.

The fiction of single wire with gain is nice, but obviously yet not achieved; so if one wants to know if the quality is nevertheless already sufficient he has to do some tests.
The next level would be to realize that we are not listening to amplifiers or DACs in isolation but to systems and have therefore to evaluate these.