True. However, he seem to have recognized that he was wrong and that scientific rigour must be applied.

I doubt this. I think if he were still the editor of Stereophile, he'd still be 'on message', even if he thought the magazine was heading in the wrong direction - or already there.

That's not just because the advertisers would pull out, it would be considered a betrayal by the readership, too. A magazine could potentially argue that its responsibility to its readers includes supporting the belief systems of those readers, even if those belief systems are highly questionable to non-readers. A 'pr0n' magazine, for example, is never going to include features about how its industry objectifies women.

What gets me is just how on message everyone is on magazines like Stereophile. What happens to those voices of dissent in print? I understand why they don't last long on audiophile forums - for the same reason audiophile opinions tend not to have long shelf-lives here - but what about the magazine itself? Is the subject forbidden, or does it simply engender so much animosity in the faithful readership that to publish anything apart from the received audiophile wisdom is fruitless? I really can't imagine that there's not one scientifically-savvy audiophile in the whole pack.

I've been reading this thread with great amusement seeing how those three people, [Fremer], Atkinson and Krueger (I don't know any of them) act and behave like children. I expected discussion, but instead, I read pages of childish fighting, the kind my daughter went through in 4th grade.
People, please, show some dignity. You are not children anymore, and are able to have civilized conversation.

Mr. [Fremer], Mr. Atkinson, do you even believe that ABX test can be used to show if there are really a difference between two amplifiers, or two codecs?

What gets me is just how on message everyone is on magazines like Stereophile. What happens to those voices of dissent in print?

Of course. But to organize such a test is not a trivial matter. The problem with the ABX protocol is that unless carefully implemented, it tends to produce false negatives, particularly if the number of trials is small - ie, the results are null even when a real but small difference exists.

Of course. But to organize such a test is not a trivial matter. The problem with the ABX protocol is that unless carefully implemented, it tends to produce false negatives, particularly if the number of trials is small - ie, the results are null even when a real but small difference exists. (See Les Leventhal's mid-1980s AES paper on this problem.)

Quote from: Gag Halfrunt on 2009-04-26 23:29:13

I doubt this. I think if he were still the editor of Stereophile, he'd still be 'on message', even if he thought the magazine was heading in the wrong direction - or already there.

Agreed. I wonder if this is the true reason why he resigned. .

This is because it is based on blind faith and in reality, a cult (no offense, Mr. Atkinson...)

...but this is my POV)- bra... bra.... It is a house of cards and cannot stand the slightest scrutiny with the light of science.

Editor: Emacs

I am not avoiding anything. I have been addressed points made by other posters who have referred to comments I have made elsewhere. I really don't see that TOS#8 applies to my writings that have published outside this group. If the moderating team wishes me not to comment, then perhaps they should request those posters not to quote my published work.

Quote from: Arnold B. Krueger on 2009-04-26 20:03:42

Quote from: Stereoeditor on 2009-04-26 17:50:41

Why do I have to? I have not said anything that is in dispute. It has been established that 16-bit LPCM at 44.1Hz sampling is not audibly transparent, ie, undetectable by all listeners under all practical circumstances with all kinds of program material (see Stuart, Fielder et al). As the performance of a lossy codec can only asymptotically approach that of the original LPCM file, thus it, too, is not audibly transparent.

John, to the best of my knowlege, the various writings of Stuart and Fielder that you have cited do not in fact themselves meet the standards of HA TOS #8. Therefore, they cannot be cited in your obvious efforts to circumvent TOS #8.

Sigh. I had assumed that academic  papers published in the Journal of the AES (in this case by two AES Fellows), could be cited on this forum.

Quote from: krabapple on 2009-04-26 02:32:33

Quote from: Stereoeditor on 2009-04-25 22:19:13

It is interesting with these files, not only to audition the character of what is removed by various codecs but also the level of the difference signal.  I fail to see why my doing so in these demonstrations should be "prohibited."  :-)

But *of course* mp3s aren't bit-identical to their source and *of course* there can be audible content in the difference file.

Correct. And I believe people who don't have the facilities to prepare such files for themselves should be able to experience them. Surely an informed listener is a better listener?

And then when you've got them gasping and shaking their heads in dismay, you *do* plan to explain what perceptual encoding *means*, right?  Which is to say, why the differences you hear in isolation, and see on graphs, can be perceptually *irrelevant* to what the mp3 sounds like? And why DBT would be *necessary* to determine if a given listener can actually hear the difference?
 
If you don't, you're misleading and misinforming your public...and that would be *shameful*.

You seem very ready to argue by projection, "krabapple." Why wouldn't I explain what perceptual coding is and why it can be useful under some circumstances? Why wouldn't I allow people to hear MP3s at various bitrates and the equivalent Red Book and hi-rez versions?  Even if they can't perceive any difference between any of the versions I play them, again, an informed listener is a better listener.

Of course these comparisons will be level-matched. Why wouldn't they be? But given that these demonstrations are open to the public there may by up to 20 people in the listening room, a formal DBT is out of the question. And please note that, as I keep saying, these are demonstrations, not tests. There will be no scoring of listeners' preferences. As I have said, I am only interested in exposing listeners to the the various formats. This is so that they can decide for themselves whether a) hi-rez formats are necessary, b)  whether CD is good enough for serious listening, and c) whether the lossy versions are sonically compromised or not. Who could argue that that would be a bad or, in your emotionally loaded term, a "shameful" thing.

As I have said, Hydrogen Audio members who live in Colorado are welcome to attend one of these sessions and debate this subect in person.

John Atkinson
Editor, Stereophile

Why do I have to? I have not said anything that is in dispute. It has been established that 16-bit LPCM at 44.1Hz sampling is not audibly transparent, ie, undetectable by all listeners under all practical circumstances with all kinds of program material (see Stuart, Fielder et al)

As the performance of a lossy codec can only asymptotically approach that of the original LPCM file, thus it, too, is not audibly transparent.

From my reading of the literature and my discussions with some of the engineers involved in designing lossy codecs, audible transparency does not appear to be the goal. Instead, it is that a lossy codec be undetectable enough of the time with enough listeners with enough kinds of program that it will be appropriate for use in circumstances where storage space or transmission bandwidth is at a premium.

Ah, yes. Not one of the JAES's better days when they published the Meyer-Moran paper, not the least because of its lack of experimental detail. YMMV, of course.

Quote from: greynol on 2009-04-26 21:12:56

ABX is the appropriate test to demonstrate that you can distinguish a lossy encode from the source from which it was created.

I disagree. You need to establish the degree of departure from transparency using an impairment scale. As I said, the literature appears mainly to feature ABC/HR testing for codec testing and from what I have observed of tests that use that protocol, it does appear a more fruitful tool for blind testing of small but real differences.

It is certainly not true and I have no problem answering specific points made on this forum.

Quote from: cpchan on 2009-04-26 21:31:59

Quote from: greynol on 2009-04-26 21:12:56

Why you're evading the issue is somewhat a mystery to me.

I agree. I would also like to see him address Gordon Holt's comment. He seem to be evading that too.

I have commented on Gordon's essay elsewhere, but not on Hydrogen Audio. And as the person who interviewed Gordon and published his essay in my magazine, perhaps you might want to give me some credit for that.

While on the face of it, Gordon's statement at http://stereophile.com/asweseeit/1107awsi is powerful, it must be pointed out that Gordon never performed any blind testing to support his review conclusions when he edited Stereophile nor after I took over from him as  editor in 1986. (Perhaps paradoxically, I have been involved in a considerably greater amount of blind testing than Gordon.)  He did review the ABX Comparator - see http://www.stereophile.com/asweseeit/121 - but ultimately decided it would not be a useful reviewing tool - see http://www.stereophile.com/thinkpieces/141 - writing "We never purchased an ABX comparator for several reasons. First, we have never felt the need for it. Second, we are finding that, regardless of "controls," an A/B test doesn't reveal small differences between components as well as does prolonged listening."

It is fair, therefore, to point out that Gordon hadn't practiced what he now preaches.

John Atkinson
Editor, Stereophile

Of course. But to organize such a test is not a trivial matter. The problem with the ABX protocol is that unless carefully implemented, it tends to produce false negatives, particularly if the number of trials is small - ie, the results are null even when a real but small difference exists. (See Les Leventhal's mid-1980s AES paper on this problem.)

I admit that the sighted listening practiced by my magazine can produce false positives. But in my view, that is preferable to false negatives. YMMV, of course. But in the end, if I publish a significantly high proportion of false positives, I will go out of business.

That being said, Arnold, I would personally hold works by the AES in high regard unless glaring flaws were found in them. AES is respectable if nothing else. Their focus seems to be scientific. I'd hold them to be true until proven otherwise.

Quote from: Gag Halfrunt on 2009-04-26 23:29:13

What gets me is just how on message everyone is on magazines like Stereophile. What happens to those voices of dissent in print?

As I thought would have been clear from the discussion of J. Gordon Holt's valedictorian comments, I do publish "voices of dissent" in Stereophile. Over the years I have published comments from such "objectivists" as Stanley Lipshitz (whom I count as a friend and mentor), Tom Nousaine, Arny Krueger, David Clark etc. I have even published reviews where the writer admits he can hear no significant difference between the item under test and his reference.

I think it important to remember that those of us whom you might consider to be on the other side of the fence are a heterogeneous bunch.

Quote from: cpchan on 2009-04-27 00:04:43

Quote from: Gag Halfrunt on 2009-04-26 23:29:13

I doubt this. I think if he were still the editor of Stereophile, he'd still be 'on message', even if he thought the magazine was heading in the wrong direction - or already there.

Agreed. I wonder if this is the true reason why he resigned. .

Not at all. While we did have some of the usual employer/employee disagreements, Gordon resigned from Stereophile in August 1999 primarily because he felt music reproduction in surround was the only valid way forward and he was frustrated by my refusal to abandon the magazine's coverage of 2-channel components and recordings.

• Original thread: Gizmodo article; Fremer; B0RK
• ABX discussion, primarily with JA

Quote from: Stereoeditor on 2009-04-26 21:37:34

Quote from: greynol on 2009-04-26 21:12:56

ABX is the appropriate test to demonstrate that you can distinguish a lossy encode from the source from which it was created.

I disagree. You need to establish the degree of departure from transparency using an impairment scale. As I said, the literature appears mainly to feature ABC/HR testing for codec testing and from what I have observed of tests that use that protocol, it does appear a more fruitful tool for blind testing of small but real differences.

ABC/hr is routinely used in Hydrogenaudio.org codec tests involving quality ratings of codecs at settings that stand a decent chance of NOT being transparent.  ABX , on the other hand, is entirely appropriate for establishing *difference* when *that* is in question.

AES is respectable if nothing else. Their focus seems to be scientific. I'd hold them to be true until proven otherwise.

Sigh. I had assumed that academic  papers published in the Journal of the AES (in this case by two AES Fellows), could be cited on this forum. If this is not the case, then I apologize to the moderators, of course. But it does seem inappropriate that the research of third parties is not available for reference by posters.

Btw, the AAC standard allows for coding of 24-bit/96-kHz PCM signals. So Mr. Atkinson, even by your stringent requirements: if we assume that 24-bit/96-kHz audio actually is transparent to everyone, a lossy codec can be transparent if you use a sufficient bitrate.

In order to assess the subjective sound quality of signals encoded and decoded at 24 bit resolution and 96 kHz sampling frequency, listening tests in the style of the ITU test specification BS.1116 have been performed.
.../...
The choice of bitrates has been performed to reach almost transparent quality at these bandwidths.
• Setting 1: 160 kbps for 21 kHz
• Setting 2: 192 kbps for 27 kHz
• Setting 3: 256 kbps for 42 kHz
• Setting 4: 160 kbps for 42 kHz.
Setting 4 has been included as a lower anchor. All these bitrates should be read as total bitrates for stereo.
.../...
As shown in figure 3 and the corresponding tables at the end of this paragraph, no statistically significant difference could be detected for settings 1, 2 and 3. Only at 160 kbps/ 42 kHz, the items ’cymbal’, ’applause’, ’guitar’ and ’triangle’ showed a degradation with a confidence interval not crossing the zero line. All of these results tend to be consistent for both, headphone and loudspeaker reproduction.

Another canny and prescient stance on Holt's part.  In terms of getting closer to the 'absolute sound', multichannel *is* the only valid way forward.  Unless you want to move entirely to binaural headphone listening.