Why 24 bit audio and anything over 48k is not only worthless, but bad for music.

[eq1849]

Given the caveats of statistics (sample size, etc.), then yes. But you’ve still not answered either of my questions!

G

Your questioning seems to assume a definitive conclusion has been drawn on the matter, which it doesn’t appear to have been, so I think you may be Begging the Question here a bit.

 

[gregorio]

Your questioning seems to assume a definitive conclusion has been drawn on the matter, which it doesn’t appear to have been, so I think you may be Begging the Question here a bit.

And still you have not answered the questions!

Your point here is nonsensical. We know what 16/44 and 24/192 are, we know properties of them, what they can contain and what they cannot contain and therefore the differences between them, because we (science) invented them. We then simply have to look at those differences and compare them to the exceptionally well known/established thresholds of hearing. That alone allows for a valid conclusion but in addition we also have direct evidence, that controlled listening tests fail to falsify that already valid conclusion. So, it doesn’t get much more of a “definitive conclusion on the matter” than that, no assumption involved and certainly not “begging the question”!

Now you’re answers please.

G

 

[danadam]

If you repeated the same test across multiple sessions and overall indicated say 80 out of 100 that would be pretty solid evidence

Yes, about 6σ if I calculated it correctly. That "pretty" in "pretty solid" seems like underselling it

 

[bigshot]

Why does no one ask you to prove you FAILED a blind listening test?

No one asks you to prove anything. Proof would require another person designing the test, documenting and administering it, with peer review follow up. If you do a test yourself, it isn’t to prove anything to us. We know the answer already. It’s established and documented. It’s for you to find out for yourself. If you can’t hear a difference that you believed you could hear sighted, then you can chalk up the difference to bias or perceptual error. If you can reliably hear a difference with a proper test, you’re going to want to contact someone from the AES because you can do something no one else can. That should be documented.

 

[eq1849]

No it’s not. That “real question” might have been real many decades or centuries ago but it ceased to be “real” when it was answered. Today it’s not a “real question”, it’s not a question anywhere at all, except in the audiophile world where it is still a question but it’s a fake, marketing driven question, rather than a real one! The only “real question” today is exactly how the brain manufactures those perceptions and how can that information be used to potentially manipulate perception.

So, what is the “real question” here then? As it’s not a “real question” of whether you actually passed a valid 16/44.1 vs 24/192 test, then the “real question” is how you achieved that result. For which there are various potential answers: A test methodology fault (such as not accurately level matching, etc.). A fault with one of the files. A fluke from not performing enough trials. A faulty piece of test equipment. Not meeting required conditions (such as a reasonable listening level). Etc.

G

When I first joined, I setup the blind listening tests per instructions from you and big shot(level matching, the while works).

The biggest factor for me was relaxing, in my experience.

 

[theveterans]

When I first joined, I setup the blind listening tests per instructions from you and big shot(level matching, the while works).

The biggest factor for me was relaxing, in my experience.

Similar to mine. As with all of my experiences with DBT ABX, the trick is to not to concentrate too hard since it's gonna give me fatigue sooner hence everything will sound homogenous to my perception. Usually, the biggest difference I hear is when I'm actually caught off guard during these test. That also applies to my sighted listening to some degree

 

[castleofargh]

When I first joined, I setup the blind listening tests per instructions from you and big shot(level matching, the while works).

The biggest factor for me was relaxing, in my experience.

If you manage to get statistical significance for a reasonably set up test, aren't you curious as to why you're hearing a difference?
Even in the research papers that support some more or less clear impact on listeners from hires, they keep also showing that listeners failed to pass the blind test. Nearly all research works reject conscious notice of a sound difference.

If I was really getting reliable clues while testing CD against hires, I would wonder about the actual cause. And there could be quite a few depending on how little effort was put into creating the listening conditions.
I could have one of those DACs that roll off so early, the treble takes a hit of several dB when playing 44.1kHz. That could easily be checked by converting the hires file to 16/44 and then back to its original resolution. That way while the track is missing all the ultrasound content, the DAC will apply the same filter on both tracks, so no chance of treble roll off on one. Other possibilities can be excluded that way too, like some idea about a track using a bigger bandwidth/sample rate having better timing than 44.1kHz, or some idea about the DAC(if delta sigma), resampling at a fixed sample rate so it would be doing less resampling on the hires file than on 44.1. We could have any sort of idea, potentially correct or ludicrous, so long as we can correctly isolate the hypothesis and test for it, that's part of the process. If maybe you have special ears that do perceive higher frequencies(youngsters or genetic rarity, or robot in a flesh suit), that's something you surely would want to test on the side.
Maybe if the files come from a PC, it's poorly configured, and it applies a crappy resampling to only one of the files, maybe it's audible, maybe it just adds enough delay to unconsciously identify the tracks that way.

If the discussion was can you feel getting hit by a baseball bat, we'd have no reason to suspect another cause when you reply yes. But here, those who set serious experiments suggest we won't hear a difference. At the very least, not a conscious one where we can identify the track consistently. If then someone comes and says he can, isn't it completely normal to suspect some confounding variables creating a false positive in his test?

 

[theveterans]

@castleofargh

If you have a DAC that relay clicks on sample rate switching then you cannot perform the hi-res against CD DBT ABX because the clicks are a giveaway.

What's genuine about goldensound's abx tracks is they're both the same sample rate of 176.4 KHz and the only difference between them is the oversampling filter used for those files and both are linear all the way to 20KHz with the high performance filter just being much, much steeper attenuation at nyquist half band of 44.1 KHz sample rate compared to the "standard" (i.e. the typical oversampling used by AKM or ESS DAC chips)

 

[bigshot]

It may seem obvious to us, but you can’t compare an HD track or SACD to a CD. The mastering might be entirely different. I suspect that this is at the root of most of the claims of “night and day” differences. Most of the rest can be chalked up to the fact that the HiRes track is probably a hair louder than the StdDef one.

 

[castleofargh]

@castleofargh

If you have a DAC that relay clicks on sample rate switching then you cannot perform the hi-res against CD DBT ABX because the clicks are a giveaway.

What's genuine about goldensound's abx tracks is they're both the same sample rate of 176.4 KHz and the only difference between them is the oversampling filter used for those files and both are linear all the way to 20KHz with the high performance filter just being much, much steeper attenuation at nyquist half band of 44.1 KHz sample rate compared to the "standard" (i.e. the typical oversampling used by AKM or ESS DAC chips)

Yes, I've had DACs like that in the past, the noise and/or delay bothered me to the point of resampling everything from the PC just so I could keep shuffling my playlists without whining about it.

Goldy checked 2 particular filters, and seems to have isolated them from nearly everything else, which is the correct way to try and show causal impact. I haven't kept track of this so IDK if there are some findings about test issue or not, I think at the time the most likely explanation was that he was young and could still hear a little above 20kHz(at what SPL? IDK).
I sure can't hear 20kHz at my usual listening level somewhere in the 60 to 65dB SPL, or a little louder if I'm listening to someone talk in English(funny how I need to turn up the volume for English but not for French). Now my limit is closer to 14kHz.

He got a lot a crap for that test, and while he sure click baited everybody by bringing up DAC when it was not a listening test of 2 DACs, I think he approached the question in a very reasonable way that drastically limited the number of extra variables affecting the outcome.
The extra variables remaining are:
- the track used.
-the listening level(were any EQ and gain used to help hear what happened near the filter frequencies?).
-the headphone(how loud it can go at and above 20kHz and how much it can distort).
-the listener.

Acting on each of those is susceptible to alter the outcome of the test. So while, to my knowledge, I can't see a reason to doubt his test and outcome, I still believe the conclusion requires those 4 conditions to be specified. Can a filter be audible, yes, depending on those 4 extra criteria beside the filter choice itself. Those conditions are important because TBH, we all knew that the filter choice could be audible, you just have to pick a terrible one or no filter for one of the test samples and most people would hear the change.

 

[gregorio]

What's genuine about goldensound's abx tracks is they're both the same sample rate of 176.4 KHz and the only difference between them is the oversampling filter used for those files and both are linear all the way to 20KHz with the high performance filter just being much, much steeper attenuation at nyquist half band of 44.1 KHz sample rate compared to the "standard" …

But that is NOT “genuine”, because it is not true! I’ve seen the same thing stated by others but GoldenSound himself, in his video, stated the difference was not only the steepness of the “high performance filter” but also the dither. In this respect it appears to be a rehash of the infamous Stuart paper on the “Audibility of digital filters”.

There are various ways to reliably tell 16/44 from hi-res in a controlled DBT, I mentioned them previously and that I had done so. The easiest is simply to find a very quiet section (a fade out is good), crank the volume and listen for the dither. That’s what the famous Boston Audio paper did nearly 20 years ago. Another way is to use a NOS DAC, as mentioned.

Unfortunately, GoldenSound has a business to run and therefore regularly overstates (or over implies) the audibility of differences. I’ve had this out with him here in this forum in the past. It’s a real shame, on the one hand it’s extremely useful and does a great service having someone publishing objective measurements of consumer audio equipment but on the other it’s also disingenuous to indicate differences as being audible when they’re not. Had a similar issue with Amir many years ago and for the same reason but he was much more sophisticated when trying to defend his stance than GoldenSound. Fortunately, he’s softened his stance significantly since starting ASR but I personally still take his subjective conclusions with a pinch of salt.

G

 

[JamoBroGuy]

Fortunately, he’s softened his stance significantly since starting ASR but I personally still take his subjective conclusions with a pinch of salt.

Me too, partially. I’m not sure how much I agree with the whole “IEMs and headphones must follow Harman” standpoint, but that aside his measurements are obviously excellent.