Conclusive "Proof" that higher resolution audio sounds different

[Al M.]

Poor hi-frequency reproduction was once considered part & parcel of early redbook reproduction, but those days are long behind us.

Apart from problems with early DACs, some of the early analog/digital converters were reported to have had a resolution in the high frequencies no better than 12 bit.

 

[Phelonious Ponk]

Not at all ... in fact (and with all due respect to Amir hearing) ... I think many experienced audiophiles, especially those who do a lot of hi-quality digital recording, could repeat & verify his claims.

However, whats this lesson taught us ... that we didn't already know?

I'd like to see many experienced audiophiles tested. And I think we've learned something important about the significance of the difference between "hi-res" and Redbook.

Tim

 

[amirm]

Thanks for the discussion guys .

Just another tidbit of data. I repeated the 32 Khz vs 96 Khz test using my Stax differential tube amp (SRM-007t) and the "pro" headphone. Here are the results:

-------------------
foo_abx 1.3.4 report
foobar2000 v1.3.2
2014/07/16 08:31:45

File A: C:\Users\Amir\Music\Arnys Filter Test\keys jangling full band 2496.wav
File B: C:\Users\Amir\Music\Arnys Filter Test\keys jangling band resolution limited 3216 2496.wav

08:31:45 : Test started.
08:32:28 : 01/01 50.0%
08:32:41 : 02/02 25.0%
08:32:55 : 03/03 12.5%
08:33:08 : 04/04 6.3%
08:33:21 : 05/05 3.1%
08:33:34 : 06/06 1.6%
08:33:47 : 07/07 0.8%
08:34:03 : 08/08 0.4%
08:34:14 : 09/09 0.2%
08:34:26 : 10/10 0.1%
08:34:45 : 11/11 0.0%
08:34:50 : Test finished.

----------
Total: 11/11 (0.0%)

-----------------------

So there is nothing funny going on with the Etymotic headphone I was using. At least for this test case (I am travelling so no time to run more tests).

 

[amirm]

A few thoughts based on this test data:

1. People do hear differently. And trained listeners as a class hear better. The Harman tests were mentioned earlier. I don't think this is the same thing as I have had the same discussion with Dr. Toole. That is, when it comes to these small non-linear distortions, they can be picked up by trained listeners but not at all for large population of people and audiophiles. This aspect was not tested at Harman as they were evaluating speakers and the differences are always large there so the nature of the tests were different.

2. Given #1, then we cannot rely negative outcome from double blind tests and extrapolate to all people and say any difference they are hearing is imaginary. Tests must have trained listeners and critical content used per ITU BS1116 recommendation for detecting small differences. Otherwise, it matters not that it got published in Journal of AES like Meyer and Moran.

3. We tested a large group of audiophiles at Microsoft and many of them did not have listening abilities to find such differences. On the other hand, there was an occasional audiophile who beat our trained listeners with no training.

4. The test content in this case could be flawed. But flawed or not, it shows that when small differences exist, not everyone can hear them. Specifically, some of the vocal advocates like Arny Krueger could not. His follow on remarks on AVS now says he can't hear above 8 Khz due to his hearing being damaged in the Army years and years back!

Net, net, as was mentioned earlier in the thread, this should cause reflections on both sides. If you can't tell the differences, and you hold the extreme position of inaudibility or subjectivity, this data should compel you to move from that position -- at least slightly. Objectivists should no longer point to random tests run and claim these differences don't exist. Extreme subjectivists should think whether they really are able to hear such differences if they can't do so in this test. Whatever problems there are with ABX testing, did not interfere with my ability to hear and find the differences.

 

[RogerD]

In my system, the HDCD CD proved superior ... but the differences were tainted by the quality of the equipment they were played on, my CDP much superior to my Blu-Ray player (which can reproduce 24/192).



Poor hi-frequency reproduction was once considered part & parcel of early redbook reproduction, but those days are long behind us. If your CDP is "severely attenuating", the highs, then consider that an issue with your CDP and not necessarily the responsibility of the format. As for higher-rez formats having superior highs by simple default, that's proved to be as much of a moving target as CD; I've heard my share of hi-rez digital products with very much compromised extended frequencies.

tb1

Now most Redbook can sound really good. If you take the exact same recording and compare them between redbook and a higher resoluton version the difference is similiar to a generation difference between a master and a 2nd generation tape. Maybe not to the same degree but just to explain that is what I am talking about.
It is all system dependent. If a system can resolve better than another then other then a listeners limitations,that would or should explain the different outcomes,other then the format itself.

 

[MylesBAstor]

Now most Redbook can sound really good. If you take the exact same recording and compare them between redbook and a higher resoluton version the difference is similiar to a generation difference between a master and a 2nd generation tape. Maybe not to the same degree but just to explain that is what I am talking about.
It is all system dependent. If a system can resolve better than another then other then a listeners limitations,that would or should explain the different outcomes,other then the format itself.

Out of curiosity, where did you do your comparisons and with what gear? That's definitely not my experience using gear like Playback Design, Aurender and a few others. RBCD sounds like a skeleton - no stick men - next to 24/192 with the Aurender and Tidal gear. (In one case using a sampler from Channel Classics.) Something is wrong somewhere.

 

[RogerD]

Out of curiosity, where did you do your comparisons and with what gear? That's definitely not my experience using gear like Playback Design, Aurender and a few others. RBCD sounds like a skeleton - no stick men - next to 24/192 with the Aurender and Tidal gear. (In one case using a sampler from Channel Classics.) Something is wrong somewhere.

Myles,

You disagree that RBCD is lower resolution than HDCD?

 

[MylesBAstor]

Myles,

You disagree that RBCD is lower resolution than HDCD?

No I think the difference between the two formats is much larger.

And I'm tired of people constantly obfuscating the discussion between the formats by constantly bringing up the quality of the original mastering. You have to be an idiot not to realize that. We are talking all things being equal, the highest, not the lowest common denominator recording.

 

[RogerD]

No I think the difference between the two formats is much larger.

And I'm tired of people constantly obfuscating the discussion between the formats by constantly bringing up the quality of the original mastering. You have to be an idiot not to realize that. We are talking all things being equal, the highest, not the lowest common denominator recording.

I agree that the differences are much larger and I am not in any way talking about the original master. All things being equal,yes there is a big difference in informational quality.

 

[TBone]

Now most Redbook can sound really good. If you take the exact same recording and compare them between redbook and a higher resoluton version the difference is similiar to a generation difference between a master and a 2nd generation tape. Maybe not to the same degree but just to explain that is what I am talking about.

Depends on the recording, it's provenance, and/or subsequent conversions ... I can name many a recording in which the hi-rez version failed to deliver the so-called default promises of high-resolution. Also, I can name many a different CD pressings which compares the "master-tape" difference. Look, I'm not denying the potential advantages of hi-rez digital, I've heard the results first hand within my system, but certainly, I wouldn't consider all hi-rez as universal advantaged based on higher# alone. That's a default that's far too dismissive of the many other issues surrounding digital reproduction.

tb1

And BTW, the NY 1971 concert CD was initially an HDCD release, like many of NY albums. So if you heard that CD (and compared it to the DVD) in a player not capable of HDCD decoding, that comparison becomes moot.

 

[TBone]

You disagree that RBCD is lower resolution than HDCD?

it's not "lower" resolution by default, HDCD with all encrypted HDCD extensions/filters off (chosen during the recording stage) would result in a HD-CD that would in fact remain identical to any regular CD, even though the HDCD tag would still indicate HDCD "processing". Therefore, the quality of HDCD was as much a moving target - for both software & hardware - as regular CD/SACD/HI-REZ/LP.

tb1

 

[microstrip]

Isn't that a bit of a red herring in the context of this discussion?
If the mastering is different then all you can conclude is that different masters sound different.

Exactltly - this is nowhere related to the intrinsic sound quality of the digital format. It just introduces noise in the debate.

 

[microstrip]

I'd like to see many experienced audiophiles tested. And I think we've learned something important about the significance of the difference between "hi-res" and Redbook.

Tim

IMHO we learned that Amir could distinguish the formats in the conditions outlined. We do not know enough about Amir subjective scales to valuate the subjective differences accurately, each of us is just interpreting his words. I was particularly interested because he found differences in aspects I valuate a lot.

 

[Phelonious Ponk]

The voice of reason...

A few thoughts based on this test data:

1. People do hear differently. And trained listeners as a class hear better. The Harman tests were mentioned earlier. I don't think this is the same thing as I have had the same discussion with Dr. Toole. That is, when it comes to these small non-linear distortions, they can be picked up by trained listeners but not at all for large population of people and audiophiles. This aspect was not tested at Harman as they were evaluating speakers and the differences are always large there so the nature of the tests were different.

2. Given #1, then we cannot rely negative outcome from double blind tests and extrapolate to all people and say any difference they are hearing is imaginary. Tests must have trained listeners and critical content used per ITU BS1116 recommendation for detecting small differences. Otherwise, it matters not that it got published in Journal of AES like Meyer and Moran.

3. We tested a large group of audiophiles at Microsoft and many of them did not have listening abilities to find such differences. On the other hand, there was an occasional audiophile who beat our trained listeners with no training.

I imagine we will have many of these occasions here at WBF. Or at least a few that will never take the test, deny the results if they do, and strenuously insist that they hear what you know only the very rare untrained listener hears.

4. The test content in this case could be flawed. But flawed or not, it shows that when small differences exist, not everyone can hear them. Specifically, some of the vocal advocates like Arny Krueger could not. His follow on remarks on AVS now says he can't hear above 8 Khz due to his hearing being damaged in the Army years and years back!

Poor Arny. This is bound to leave a mark.

Net, net, as was mentioned earlier in the thread, this should cause reflections on both sides. If you can't tell the differences, and you hold the extreme position of inaudibility or subjectivity, this data should compel you to move from that position. Objectivists should no longer point to random tests run and claim these differences don't exist.

I moved from that position a couple of years ago. Now my question is "what is this difference? What's the cause?"

Extreme subjectivists should think whether they really are able to hear such differences if they can't do so in this test. Whatever problems there are with ABX testing, did not interfere with my ability to hear and find the differences.

Some will. But many will continue to question the test, insist that they can hear in longer listening sessions what cannot be heard in short samples (but they won't test this themselves...), or assume they are among the few (without ever testing that assumption). And the debate will continue unabated. I have faith.

Tim

 

[Phelonious Ponk]

Hey Mike - remember a few pages back when you questioned my assertion that there were a lot of hyperbolic audiophile descriptions of the weakness of Redbook compared to hi-res? I knew it wouldn't take long...

RBCD sounds like a skeleton - no stick men - next to 24/192...

Tim

 

[TBone]

Ah, that typical skeleton vs stick-men audio comparison has somehow eluded my ears ... perhaps I need to put my system on a diet?

 

[Orb]

I keep reading that this "is" from those who wish to question the results of AB/X testing. What I haven't read is any evidence in support of it. AB/X testing is pretty broadly used, outside of audio, for some things much more critical than our hobby. You'd think if it was as unreliable and problematic as many audiophiles claim, there would be some studies,confirmation of that. None have ever come forward in any of audiophile conversations that I recall. On the contrary, it continues, as it has for decades, to be a valued research tool.



I don't think anyone here is questioning what Amir heard.



Agreed. If I followed that, I think where we may depart is on the possible results of long-term listening. You seem to be saying that which is extremely subtle to a trained expert listener, concentrating on differences, focusing on passages that reveal those differences, may somehow become much more obvious to the un-trained listener in long-term, casual listening. Occam's razor; It seems much more likely that the "night and day" differences are a result of sighted listening. You know you are listening to the superior file, therefore you hear superior quality.

Tim

Tim,
you probably forgot but last year I spent a lot of time providing a lot of diverse research papers explaining various mechanisms for testing and methodology for JND/biases/cogntive decision process/etc.
But coming back to less scientific and more practical example; look how Amir and a few manage to pass this ABX and yet quite a few did not.......
Anyway I give up now, was a waste of my time back then and a waste of time now
But AB and even ABX comparisons do have their own issues-challenges-limitations and this been reported in scientific research I linked in the past.

BTW I am NOT per se questioning AB/X, but if you paid attention to all that I have said the challenge is the methodology, scope, control-monitoring-weighting of the test environment-test subjects, and critically understanding and managing the complexity of the associated JND.
Seems you are coming across with a strong position diametrically opposed to what you see "as those who wish to question the results of AB/X testing".

Cheers
Orb

 

[Orb]

A few thoughts based on this test data:

1. People do hear differently. And trained listeners as a class hear better. The Harman tests were mentioned earlier. I don't think this is the same thing as I have had the same discussion with Dr. Toole. That is, when it comes to these small non-linear distortions, they can be picked up by trained listeners but not at all for large population of people and audiophiles. This aspect was not tested at Harman as they were evaluating speakers and the differences are always large there so the nature of the tests were different.

2. Given #1, then we cannot rely negative outcome from double blind tests and extrapolate to all people and say any difference they are hearing is imaginary. Tests must have trained listeners and critical content used per ITU BS1116 recommendation for detecting small differences. Otherwise, it matters not that it got published in Journal of AES like Meyer and Moran.

3. We tested a large group of audiophiles at Microsoft and many of them did not have listening abilities to find such differences. On the other hand, there was an occasional audiophile who beat our trained listeners with no training.

4. The test content in this case could be flawed. But flawed or not, it shows that when small differences exist, not everyone can hear them. Specifically, some of the vocal advocates like Arny Krueger could not. His follow on remarks on AVS now says he can't hear above 8 Khz due to his hearing being damaged in the Army years and years back!

Net, net, as was mentioned earlier in the thread, this should cause reflections on both sides. If you can't tell the differences, and you hold the extreme position of inaudibility or subjectivity, this data should compel you to move from that position -- at least slightly. Objectivists should no longer point to random tests run and claim these differences don't exist. Extreme subjectivists should think whether they really are able to hear such differences if they can't do so in this test. Whatever problems there are with ABX testing, did not interfere with my ability to hear and find the differences.

Great summation Amir.
Cheers
Orb

 

[Phelonious Ponk]

Tim,
you probably forgot but last year I spent a lot of time providing a lot of diverse research papers explaining various mechanisms for testing and methodology for JND/biases/cogntive decision process/etc.
But coming back to less scientific and more practical example; look how Amir and a few manage to pass this ABX and yet quite a few did not.......
Anyway I give up now, was a waste of my time back then and a waste of time now
But AB and even ABX comparisons do have their own issues-challenges-limitations and this been reported in scientific research I linked in the past.

BTW I am NOT per se questioning AB/X, but if you paid attention to all that I have said the challenge is the methodology, scope, control-monitoring-weighting of the test environment-test subjects, and critically understanding and managing the complexity of the associated JND.
Seems you are coming across with a strong position diametrically opposed to what you see "as those who wish to question the results of AB/X testing".

Cheers
Orb

I remember, and I agree. And I'll still take imperfect, even casual blind listening over sighted all day, every day. And I know you know enough to understand why.

Tim

 

[Habanero Monk]

Hello everyone.

Arguments about value of sampling rate > 44.1 Khz and bit depth > 16 bits has filled thousands upon thousands of pages on various audio forums. There has been very little formal testing but what there is, such as Meyer and Moran point to no audible difference (yes, the test was flawed but this thread obsoletes what they did anyway).

I think the best way to end all this is for Studios to include a code on any Physical CD's to DL the 24/96 or 24/192 off their website for a nominal fee. Bandwidth is dirt cheap and they have ZERO amount of work in getting those formats up since that is what they were most likely natively recorded in.

Storage is cheap... I could easily store my entire album collection at 24/192 and still have room left over. Then it would be a non issue.

It's like people that debate cable burn in: Anyone with a system that they use routinely has 'burned in cables'. I think it's best to just make it all non-sequitur and let 'Hi-Rez' become the new normal.