[Resolve]

I don’t know why we keep seeing stuff like this in the audiophile world. I can only assume that audiophile marketing has a vested interest in keeping the audiophile community in the dark, through “lies of omission” and misrepresenting the facts, etc.?

To answer your point, “yes” that is indeed where “it makes sense to look - at least initially”, which is exactly why the early psycho-acousticians looked there initially in the 1890’s, then did so again in far greater detail in the 1930’s when advances in audio electronics allowed it and at numerous times since! So, it makes no sense to do even more research when there’s already been so much going back over 90 years. Nevertheless, we keep seeing points like this in the audiophile community.

G

I mean we have the work from Temme and GedLee on some of this stuff but if you have a paper on audibility thresholds for given distortion products across individual listeners by all means I'd love to see it. Like if there's something else to reference I'd love to be able to.

Edit: there's also this paper on the subject, which is a good read - although I don't know what's to really be gleaned from it fully.

 

[Doltonius]

All of these sounds can be broken down into a sum of singular frequencies called harmonics as the graph you provided notes it. The sum of these frequencies is called the "spectrum" of the sound. The frequency response shows how the IEM manipulates the spectrum of sound. Pressure changes over time can be calculated from the spectrum and the other way around also holds true, it could be even done on pen and paper for simple cases. Sound can be described by either as a function of pressure over time or as a function of pressure over frequencies. They are equivalent in a mathematical sense but certain things can be seen and be understood easier in one form than in the other. Just by looking at the squiggly line, you can't really tell how a trumpet would sound, can you? Even though the pressure changing over time perfectly describes sounds since that is the definition of sound. The data is there by definition, it just isn't easy to read it.

That @Frankie D talks about overtones and timbre but somehow thinks that FR graphs don't show this information (and that a lot of people agree with the claim by upvoting that post) is the most puzzling thing in this entire discussion.

 

[Resolve]

That @Frankie D talks about overtones and timbre but somehow thinks that FR graphs don't show this information (and that a lot of people agree with the claim by upvoting that post) is the most puzzling thing in this entire discussion.

Yeah, I mentioned this in the other thread but our audio engineer Blaine gave a presention at two canjams now that effectively mythbusted this notion by showing how the harmonic content of music comprises the same measured response you get from a sweep or an FFT. I'll post it here once that's public.

 

[gregorio]

I mean we have the work from Temme and GedLee on some of this stuff but if you have a paper on audibility thresholds for given distortion products across individual listeners by all means I'd love to see it. Like if there's something else to reference I'd love to be able to.

We have audibility thresholds “when isolating FR” going back famously to 1933 (Fletcher/Munson) and its first practical application in WWII was precisely for “the subjective goodness/badness” you mentioned. There has been considerable additional research since then of course and extensively in areas such as auditory masking. If we know these factors (and we do) then there’s limited value in specific research into say “given distortion products” beyond just confirming existing research because distortion products are of course just frequencies, they don’t have any magical properties that other frequencies don’t have.

G

 

[Resolve]

We have audibility thresholds “when isolating FR” going back famously to 1933 (Fletcher/Munson) and its first practical application in WWII was precisely for “the subjective goodness/badness” you mentioned. There has been considerable additional research since then of course and extensively in areas such as auditory masking. If we know these factors (and we do) then there’s limited value in specific research into say “given distortion products” beyond just confirming existing research because distortion products are of course just frequencies, they don’t have any magical properties that other frequencies don’t have.

G

Sure but as the paper I linked above from 2007 indicates, the existing norms around audibility for nonlinear distortion products were based on metrics they found to be insufficient. I don't know yet what that means, but to say this is a closed topic because of work done in the 1930s... I'm not so sure I can agree.

Edit: and just to be clear, while I have done some ABX testing with source gear, by no means is this my area of expertise. I just rely on what other people who understand this stuff better tell me, or the various papers I'm able to read. I won't claim to actually know anything on this topic. But it strikes me as... not really being an issue if additional audibility threshold testing gets done.

 

[gregorio]

Sure but as the paper I linked above from 2007 indicates, the existing norms around audibility for nonlinear distortion products were based on metrics they found to be insufficient.

While I’ve only quickly scanned that paper, it does NOT appear to indicate “the existing norms around audibility” are insufficient. What it actually indicates is what was already well known, that the THD, THD+N and IMD metrics do not necessarily correlate well with audibility. This is hardly surprising as they are just measurements of distortion and noise, not measurements of the audibility of distortion or noise. Hence the point of the referenced thesis paper, a proposal for a new metric (in 2007) that does incorporate audibility.

I don't know yet what that means, but to say this is a closed topic because of work done in the 1930s... I'm not so sure I can agree.

Then your potential disagreement would also disagree with the paper you cited! Ironically, your quoted paper cites Fletcher’s/Munson’s work (paper from 1933) as well as Fletcher’s work from the late 1930’s and to a certain extent relies on it. One of the reasons the typical measurements of distortion do not necessarily correlate with audibility is because they do not consider/incorporate the Fletcher/Munson “loudness contours” or other well known/established audibility factors, such as auditory masking (which I also mentioned previously)!

G

 

[Resolve]

While I’ve only quickly scanned that paper, it does NOT appear to indicate “the existing norms around audibility” are insufficient. What it actually indicates is what was already well known, that the THD, THD+N and IMD metrics do not necessarily correlate well with audibility. This is hardly surprising as they are just measurements of distortion and noise, not measurements of the audibility of distortion or noise. Hence the point of the referenced thesis paper, a proposal for a new metric (in 2007) that does incorporate audibility.

Then your potential disagreement would also disagree with the paper you cited! Ironically, your quoted paper cites Fletcher’s/Munson’s work (paper from 1933) as well as Fletcher’s work from the late 1930’s and to a certain extent relies on it. One of the reasons the typical measurements of distortion do not necessarily correlate with audibility is because they do not consider/incorporate the Fletcher/Munson “loudness contours” or other well known/established audibility factors, such as auditory masking (which I also mentioned previously)!

G

Well... my takeaway from it was more that conventions around THD and intermodulation distortion did not line up with subjective reports, and it made more sense to use newer, better correlated metrics. But if you say that's to reinstate older work, by all means I'm fine with that too. My point was more that there's still work being done on this topic today.

I have no issues with saying FR is all there is when all else is within the masking window or 'below the audible threshold' if that is in fact the case. That's... why this topic got started in the first place. I just keep an open mind to new information should it arise, and I do know of instances where the ABX results aren't as straightforward as we might think. One of the common ills in this hobby is that people get very dogmatic about a given perspective, and even though my personal opinions end up more on the side of "these other metrics don't matter much unless they're awful", I'm perfectly happy to be wrong about that if someone can show an example to the contrary.

 

[bigshot]

Well... my takeaway from it was more that conventions around THD and intermodulation distortion did not line up with subjective reports, and it made more sense to use newer, better correlated metrics. But if you say that's to reinstate older work, by all means I'm fine with that too. My point was more that there's still work being done on this topic today.

What has changed about distortion or human hearing to indicate that the established thresholds aren't accurate? The fact that the facts don't line up with subjective reports should just indicate how inaccurate subjective reports are.

 

[theveterans]

What has changed about distortion or human hearing to indicate that the established thresholds aren't accurate? The fact that the facts don't line up with subjective reports should just indicate how inaccurate subjective reports are.

Preference alone makes the whole subjective reports inaccurate. You can NEVER correlate or line up subjective reports to established thresholds EVER no matter how you slice and dice it up

 

[castleofargh]

While I’ve only quickly scanned that paper, it does NOT appear to indicate “the existing norms around audibility” are insufficient. What it actually indicates is what was already well known, that the THD, THD+N and IMD metrics do not necessarily correlate well with audibility. This is hardly surprising as they are just measurements of distortion and noise, not measurements of the audibility of distortion or noise. Hence the point of the referenced thesis paper, a proposal for a new metric (in 2007) that does incorporate audibility.

Then your potential disagreement would also disagree with the paper you cited! Ironically, your quoted paper cites Fletcher’s/Munson’s work (paper from 1933) as well as Fletcher’s work from the late 1930’s and to a certain extent relies on it. One of the reasons the typical measurements of distortion do not necessarily correlate with audibility is because they do not consider/incorporate the Fletcher/Munson “loudness contours” or other well known/established audibility factors, such as auditory masking (which I also mentioned previously)!

G

I'm guessing we've all read the same papers at some point(although I might be talking about us 3 and 2 other guys in the thread when I say all), and indeed what should concern us is limited to what is audible, and if it bothers us or not. Which is where we all agree, even above audible levels, THD is rarely of any help for subjective predictions. IMD hardly does any better, and we pretty much never look at that in transducers anyway. Temme with Olive and sans Olive, did seem to come up with measurements that do better to relate to preferences, but I have no clue how to set up those measurements.
That's where I agree with @Resolve, here they were talking about an IEM with bass so low that it seems to reach into dark energy and create gravity waves(at least that's what some reviews made me believe^_^). We have a lot of work done recently on sound in cars, noise-cancelling, more lossy codecs, generic approaches to sound preferences like MDAQS(promoted a few times on Headfi)... It is being done because people thought we didn't know enough.

I don't expect any revolution in that domain, for sure. People often already had some ideas about what's what long before research came and said it's not just you, "it's 48% of the population!" or something. Plus, for this thread in particular, IEMs tend to have distortions a good deal lower than speakers(the better known/studied transducers).
IDK about that IEM in particular, I couldn't even find an impedance graph after reading @theveterans' question.

 

[gregorio]

Well... my takeaway from it was more that conventions around THD and intermodulation distortion did not line up with subjective reports …

And that, IMHO, is the problem. Not that THD and IMD did not line up with subjective reports but that the audiophile community developed false “conventions around” them, largely (or entirely) driven by marketing and, this isn’t just the case with THD and IMD but with numerous things!

You can NEVER correlate or line up subjective reports to established thresholds EVER no matter how you slice and dice it up

I’m not sure I agree with this. It’s certainly true of some thresholds but not necessarily of all thresholds, given certain conditions. Of course though, the point of thresholds is that they represent the limits of human hearing, they don’t represent what we should expect to achieve in our sitting room, when listening to music (rather than test signals) and with say 40+ year old ears. The problem with many subjective reports in the audiophile world is that they often exceed the thresholds, not uncommonly by one or more orders of magnitude!

I'm guessing we've all read the same papers at some point(although I might be talking about us 3 and 2 other guys in the thread when I say all), and indeed what should concern us is limited to what is audible, and if it bothers us or not. Which is where we all agree, even above audible levels, THD is rarely of any help for subjective predictions.

The bolded part is the problem IMHO, depending on who exactly is this “us”, then “what is audible” can have a significantly different meaning. It’s not that “audible” itself has a different meaning but that “what can be audible” does, because the “us” for me isn’t necessarily other consumers but can be other music/sound engineers (and other musicians). And, this isn’t due to some golden eared, superhuman hearing ability of engineers but because we’re routinely editing and processing recordings (with EQ, compression and other tools) which raises the levels in some places by 40dB or more, thereby sometimes converting the inaudible into the audible. So, when “us” engineers state that something “can be audible” we don’t mean that if you have great hearing/listening skills and TOTL equipment you’ll hear it, but that conditions may arise where the required processing/mixing/mastering can, in some portion/s of a particular recording, result in a large enough boost of artefacts, distortions or noise to render them audible.

Of course, none of the above is applicable to the “us” as consumers, we’re never (for example) boosting portions of our recordings by 40dB or more and/or repeatedly running them through a conversion process and thereby audibly revealing artefacts/distortions caused by say jitter or converters. Unfortunately though, for nearly half a century, audiophile marketers have been trawling the pro audio world for discussions, descriptions and terms that they can misrepresent as applicable to the audiophile community and it’s now packed with them!

G

 

[bigshot]

the point of thresholds is that they represent the limits of human hearing, they don’t represent what we should expect to achieve in our sitting room, when listening to music (rather than test signals) and with say 40+ year old ears. The problem with many subjective reports in the audiophile world is that they often exceed the thresholds, not uncommonly by one or more orders of magnitude!

This is the core of the problem with the audiophile community. You can define a threshold in such a way that you would have to bend over backwards twice to run into it in the real world, and you can be satisfied that you've solved the problem. But then an audiophile will come along and up the threshold a chunk "just in case". Then another audiophile will come along and concoct some solution to a problem that doesn't exist and knock they numbers a bit further. Then yet another audiophile will theorize that *maybe* we can hear the unhearable, and another will cheat the test to prove how golden their ears are and shove it even further into overkill. We see all of this in Head-Fi every day. No one talks about listening to a Beatles album playing on their home stereo in their living room, instead they talk about test tones, anechoic chambers, super audible frequencies, noise floors that extend into the far distance and abstract measurements on a page that are never correlated with actual human hearing. Sound Science can be as guilty of this doubling down on abstract theory as much as subjective woo-woo audio forums.

It's refreshing when someone actually mentions practical reality when it comes to all this. There are things that matter and things that don't, and the distance between what kind of real world sound is important to one person and what is important to another isn't that far apart. After all, the purpose is to listen to commercially recorded music in a way that sounds good to us. That really isn't all that hard to figure out how to achieve... and it is perfectly achievable.

 

[Doltonius]

This is the core of the problem with the audiophile community. You can define a threshold in such a way that you would have to bend over backwards twice to run into it in the real world, and you can be satisfied that you've solved the problem. But then an audiophile will come along and up the threshold a chunk "just in case". Then another audiophile will come along and concoct some solution to a problem that doesn't exist and knock they numbers a bit further. Then yet another audiophile will theorize that *maybe* we can hear the unhearable, and another will cheat the test to prove how golden their ears are and shove it even further into overkill. We see all of this in Head-Fi every day. No one talks about listening to a Beatles album playing on their home stereo in their living room, instead they talk about test tones, anechoic chambers, super audible frequencies, noise floors that extend into the far distance and abstract measurements on a page that are never correlated with actual human hearing. Sound Science can be as guilty of this doubling down on abstract theory as much as subjective woo-woo audio forums.

It's refreshing when someone actually mentions practical reality when it comes to all this. There are things that matter and things that don't, and the distance between what kind of real world sound is important to one person and what is important to another isn't that far apart. After all, the purpose is to listen to commercially recorded music in a way that sounds good to us. That really isn't all that hard to figure out how to achieve... and it is perfectly achievable.

In short, the "sound science" that is relevant for the audiophile community must be based on psychoacoustics, and not pure acoustics. Data whose correlation with subjective perception in realistic scenarios has not been experimentally established is indeterminate for our purposes at best, irrelevant practically, and misleading at worst.

 

[bigshot]

One of the common ills in this hobby is that people get very dogmatic about a given perspective, and even though my personal opinions end up more on the side of "these other metrics don't matter much unless they're awful", I'm perfectly happy to be wrong about that if someone can show an example to the contrary.

I don't know why people argue about it at all.

With speaker systems 5% distortion is perfectly acceptable and sounds good when playing back commercially recorded music. In certain frequencies, decent speakers can have as much as 10% and still sound OK. Have you ever heard of IEMs that come anywhere close to that?

Here are test results for audibility of speaker distortion that someone posted online. That big blue bar in the middle represents 9%. -33 is 2.24%. Why are people worrying about distortion levels in DACs and amps when the transducers totally swamp them in that department and it still sounds great?

 

[bigshot]

Data whose correlation with subjective perception in realistic scenarios has not been experimentally established is indeterminate for our purposes at best, irrelevant practically, and misleading at worst.

I have good news for you. Just about every decent DAC and amp are audibly transparent in normal use. Headphones a mostly a matter of finding a response curve that pleases you. Speakers are a balancing act between the sound produced by the speakers and the envelope the room places on it. In short, if you focus on getting your transducers in line, you are almost all of the way there.

There is a factor that is rarely discussed by audiophiles that is more important than most of the things they do discuss... usability. A system that is simple and has an intuitive interface is very important, yet audiophiles love to create complex Rube Goldberg contraptions that break the process of sound reproduction up into bits and have a separate black box for each bit. They love fiddling with equipment more than listening to music.