Is the entire audio industry a fraud?

[Dan]

If you are not listening blind, then you are using your eyes, not your ears, to judge.

No need for fancy ABX tests and statistics. Just listen blind. Do whatever you usually do to decide that Amp A sounds better than Amp B, but without knowing which is which.

The number of audiophiles who have actually tried such an experiment is vanishlngly small.

It's not so easy to do. I am probably like a lot of audiophiles, I know virtually no one interested in Hifi. They couldn't care less. Even my friend who owns a chain of movie theaters thinks it's all silliness. If I asked my wife or daughter to switch whatever in a blind test they'd laugh their asses off.

 

[Mr. Widget]

No need for fancy ABX tests and statistics. Just listen blind. Do whatever you usually do to decide that Amp A sounds better than Amp B, but without knowing which is which.

I have done this numerous times over the years and my conclusions do not completely align with the outspoken majority here.

I hope to have time to set up much more rigorously controlled unsighted comparisons at some point. Until then I will withhold my opinions, but I will not rule them out as "extraordinary" until I have had the opportunity to explore this in more depth.

 

[ahofer]

I have done this numerous times over the years and my conclusions do not completely align with the outspoken majority here.

I hope to have time to set up much more rigorously controlled unsighted comparisons at some point. Until then I will withhold my opinions, but I will not rule them out as "extraordinary" until I have had the opportunity to explore this in more depth.

Fair. Level matching is always a question (see my signature quotes), so do address that at the electronics level to control adequately.

 

[Mr. Widget]

Fair. Level matching is always a question (see my signature quotes), so do address that at the electronics level to control adequately.

Oh yeah... I've been paying attention.

Unfortunately I don't see having time to do this in the coming weeks, but I promise a full report including the controls when I get to this.

 

[atmasphere]

Wall of words notwithstanding, you still haven’t provided any data or evidence to back up your claims. Yet again.

Well I knew it was likely you'd dig in your heels.

Why supply 'evidence' when its self-evident?? I can see I have five fingers on each hand, do you doubt that?

What I want to know, which you've not shown, is since harmonics define how instruments sound, how can adding more not affect how they sound? I've asked you that in one form or another three times now and all I've gotten is obfuscation and insults. It doesn't become you- normally you have good info. Is the reference not available?

 

[MarkS]

Level matching is always a question

I claim that level matching is not needed if
(1) the volume starts at zero after each switch and
(2) the listener adjusts the volume to preference without any visual or tactile indication of the precise volume.

 

[MarkS]

since harmonics define how instruments sound, how can adding more not affect how they sound?

If the added harmonics are low enough in level, they cannot be heard.

 

[atmasphere]

If the added harmonics are low enough in level, they cannot be heard.

Right- we are talking about higher ordered harmonics; so how low?

Keep in mind that the ear uses higher orders to sense sound pressure. And has a 120-130dB range so is keenly sensitive to them.

 

[Blumlein 88]

I claim that level matching is not needed if
(1) the volume starts at zero after each switch and
(2) the listener adjusts the volume to preference without any visual or tactile indication of the precise volume.

Absolutely WRONG.

Matching by ear is likely to be within about .5 db to 1 db. A difference of .2 db is enough to mess up results. As in two identical tracks with identical gear with one .2 db louder will be reliably chosen in blind testing. Not to mention the extra issue of resetting volume with each trial. Quick switching makes a big difference in how sensitive the test is. By forcing the listener to readjust volume it takes time and just will screw up the result.

 

[Blumlein 88]

No, it means you got a specific number of correct/incorrect trials in the test and then a specific set of statistical analysis was applied to draw a conclusion.

The conclusion has two confidence numbers attached, namely the risk of false positives and the risk of false negatives.

This, like so many things has a triangle with three sides were two exclude third:

small number (< 100's) of trials
low risk of false positives
low risk of false of negatives

Choose any two. Accept that the third item will be out of the picture.



I already posted that, in the other thread.

Thor

I get in a sense what you are talking about. For instance you think you hear a difference you do a blind test and get 12 of 20. No where close to only 5% chance your result was random. Do this five times and with 60 of 100 you are within the range to say there is only a 5% chance the result was random. There are sure to be some edge cases where you miss something.

What is often claimed is some highly dubious possibility and the description is "night and day", " you must be deaf if you don't hear this" and the ever popular "my wife heard it from another room". Or even, "the difference is so huge I don't have to match levels". Those guys should not need 100 trials. Yes, maybe in some cases there is something that would show up in 100 trials, but if it goes against rational reasonable knowledge about how things work, it is very unlikely. At a minimum something you need 100 to reach 60 correct and show something is not a "night and day" difference.

Having done a few amateur tests, it has always astounded me how something can seem so certain, so clear, to the point I'd feel like it was blindingly obvious (pun intended) only to have it disappear completely with blinding. I have over time done some of those with recorded digital files to 100 trials. The more I did the closer my results get to exactly 50/50. As real as it seemed there was nothing to it. Of course I've also detected real differences so the test certainly can work.

With the claims made, and experience about such things, and how things work you could hold out hope some of these improbable claims are real. I think the odds are very low. No one can say until such testing is done I suppose.

On the other hand, if it takes 60 of 100 to finally tease out a result, then just how important is the difference in general music playback for enjoyment? 50/50 vs 60/40 must be getting to trivial levels.

I like the way 2AFC testing works, but in audio the listeners always complain about not having a choice that says it sounds the same. No matter how I explain why this will work, and that people have scored beyond chance even when they think they hear no difference, they are never satisfied. I also like triangle testing. Here are three samples two are the same one is different. Listen however you like and pick the odd one out. Up/down testing seems useful for determining thresholds. You have to have a characteristic you know is audible and can be varied however. There are some online blind tests where you can find your personal threshold for THD and IMD. Up down testing to me seems cognitively easier than other types I suppose because you start with something audible and get feedback on the results.

I don't agree that ABX is some evil that is designed to return a null result glossing over real differences. Like anything it can be misused. Like much testing the temptation to use too few trials or samples is there. When done with few trials it is better to confirm differences than to rule them out. You just need more trials. If you can do 10 of 10 and especially 20 of 20 you likely don't need to continue.

 

[Blumlein 88]

Right- we are talking about higher ordered harmonics; so how low?

Keep in mind that the ear uses higher orders to sense sound pressure. And has a 120-130dB range so is keenly sensitive to them.

We have masking curves. They will give at least some general guidance. I think 5th harmonic at 1 khz is masked until it rises above -50 db of the 1 khz tone. It will be less at higher harmonics, but this gets limited by our insensitivity at the highest frequencies. Plus the 5th or higher orders won't be heard above 4 khz. You then have the other end to look at which is IMD. The fortunate result of IMD is when it comes from higher frequencies there is usually more and higher levels at those frequencies where the IMD shows up. So it gets masked pretty well too. One can always find some strange edge case like the Beatles song that ends with a long period of a 15 khz tone at fairly high levels.

 

[SIY]

Why supply 'evidence' when its self-evident??

It very much is not. And the argument, "Instrument spectra show 20-30-50+ percent harmonics, so an amp adding 0.05% harmonics should be audible" makes absolutely no sense.

Since we can add harmonics in software (@pkane has a particularly useful program), it would be trivial for you to take a music file, add the same harmonic content as the Adcom amp you claimed had "a perception of 'harsh and bright' at some volume setting," then post the files and an ABX log showing identification. That would be actual evidence.

Do that successfully and I (and others who look at 40+ years worth of testing showing no significant audible differences between amps with the same frequency response and not clipping) will humbly salute you, cite you regularly, and not question this claim again.

Evidence. That what separates solid claims from fairy tales.

 

[ahofer]

We have masking curves. They will give at least some general guidance. I think 5th harmonic at 1 khz is masked until it rises above -50 db of the 1 khz tone. It will be less at higher harmonics, but this gets limited by our insensitivity at the highest frequencies. Plus the 5th or higher orders won't be heard above 4 khz. You then have the other end to look at which is IMD. The fortunate result of IMD is when it comes from higher frequencies there is usually more and higher levels at those frequencies where the IMD shows up. So it gets masked pretty well too. One can always find some strange edge case like the Beatles song that ends with a long period of a 15 khz tone at fairly high levels.

Do you have a citation for ’masking curves‘? I haven’t read about that (Not skepticism, intrigue).

 

[Killingbeans]

Masking Studies and Relationship to Nonlinear Distortion Audibility (?)

There was a discussion developing today on another thread about what is an acceptable (i.e. inaudible) level of nonlinear distortion from an electronic component. While we have a good sense of what kinds of claim are plainly ridiculous, the discussion reinforced for me just how unclear the...

www.audiosciencereview.com

 

[Blumlein 88]

Do you have a citation for ’masking curves‘? I haven’t read about that (Not skepticism, intrigue).

Here is a basic masking curve for a tone. The rest is just extending the idea. If you have fundamental at 4 khz, then the 5th harmonic is 20 khz. We don't need masking beyond that as we won't hear it.

I'm not sure what else is of interest. I can probably fine some treatment of it in more detail. EDIT to add: Killingbeans beat me to it and gave you a good thread about masking.

 

[MarkS]

Absolutely WRONG.

Matching by ear is likely to be within about .5 db to 1 db. A difference of .2 db is enough to mess up results. As in two identical tracks with identical gear with one .2 db louder will be reliably chosen in blind testing. Not to mention the extra issue of resetting volume with each trial. Quick switching makes a big difference in how sensitive the test is. By forcing the listener to readjust volume it takes time and just will screw up the result.

You are the one who is wrong here.

Matching is only necessary for RAPID switching. It is NOT necessary if the volume is reset to zero at the start of each trial, and then set by the listener by ear only.

It is correct that this sort of test is less sensitive. That is not the point. The point is to have a test that does not require special measuring equipment, and to mimic what audiophiles actually do to judge sound, which rarely involves rapid switching.

 

[Blumlein 88]

You are the one who is wrong here.

Matching is only necessary for RAPID switching. It is NOT necessary if the volume is reset to zero at the start of each trial, and then set by the listener by ear only.

It is correct that this sort of test is less sensitive. That is not the point. The point is to have a test that does not require special measuring equipment, and to mimic what audiophiles actually do to judge sound, which rarely involves rapid switching.

Then your results are convenient.........................and worthless.

I don't know where you got this idea or if you came up with it yourself. It is not the way to do a comparison and get any useful results. Level matching is pretty easy and only requires an inexpensive voltmeter. If you aren't willing to do that, your listening comparisons aren't worth telling anyone else about.

 

[Barrelhouse Solly]

Someone may have mentioned this, but one of the oldest principles of modern "scientific" marketing is, "Sell the sizzle, not the steak." The audio biz isn't a fraud, it's just that there's a lot of marketing involved.

 

[SIY]

Then your results are convenient.........................and worthless.

I don't know where you got this idea or if you came up with it yourself. It is not the way to do a comparison and get any useful results. Level matching is pretty easy and only requires an inexpensive voltmeter. If you aren't willing to do that, your listening comparisons aren't worth telling anyone else about.

To follow on, it’s easy to have a master volume control so that both sources can be raised and lowered simultaneously if the listener wants to continuously diddle with levels.

 

[MarkS]

The average audiophile does not own a voltmeter (and wouldn't be able to use it correctly if he did), and firmly believes that amps sound different, and that rapid switching is not needed to hear the difference.

My test procedure is intended to help such folks learn something.