Arny, your cavalier attitude on the simple topic of communication is breathtaking. I have been nearly ostracized - literally - for making far less inflammatory accusations between groups of friends, than what you have accused Michael of.

And then you have the nerve to respond not by actually backing your sh*t up,

but by attacking Michael's semantics!?!

I think I understand why I am (and Canar is) so flummoxed about this entire topic, going all the way back to when B0RK was the center of attention. You, and also Steven to a certain degree, have turned this thread into what seems like a word-for-word rematch of the exact same argument, made against the exact same people, that has been replayed on Usenet for what seems like a decade - an argument I have absolutely no intention of supporting.

I said this to jj, I alluded to this to Michael, and I'll say it to you: HA IS NOT RAO.

It isn't even RAHE.

Some of us are quite willing to take John and Michael as being rational, intelligent, educated, experienced people

which is an assertion to some degree backed up by their rather good and coherent replies (particularly John's which I have always admired). I cannot say the same about your replies.

Let me put it this way, when your debating style is criticized as being too coarse on Stereo Central, something is deeply wrong.

Mr. Atkinson,
I have twice put versions of this question to you, but it was apparently lost in the shuffle.  Your published position on the use of lossy compression is as follows:  "MP3s and their lossy-compressed ilk do not offer sufficient audio quality for serious music listening. This is not true of lossless-compressed formats such as FLAC, ALC, and WMA lossless"

If an individual listener is unable to differentiate a lossy file, either under the conditions of an ABX test or in a more relaxed and conventional listening setting, from its lossless counterpart then in what way, precisely, does the lossy file "not offer sufficient audio quality for serious music listening"?

I do just about everything including personal listening with pure .wav files and redbook CDs.

Quote from: 2Bdecided on 2009-04-21 12:06:50

But that doesn't mean that no difference exists. 320kbps mp3s can be ABXed. The Emperor has clothes in this case - but very few people can see them!

Don't count me into that. My digital music players are loaded with .wav files whereever possible.

For that person at that time, then yes, the lossy compressed file is presumably good enough.

But everyone's threshold is going to be different and perhaps more significantly, their threshold of defects will change with time.

Hence my blanket recommendation to which you refer: lossless or uncompressed for "serious" listening, to which I would add archiving. Why not when hard drive capacity is now so cheap.  And for portable listening, I personally use AAC at 320kbps and recommend that, even if it might be thought overkill.

PS: I always recommend to my readers that they keep their CDs, as inconvenient as that might be. They are the backup of last resort.

For that person at that time, then yes, the lossy compressed file is presumably good enough. But everyone's threshold is going to be different and perhaps more significantly, their threshold of defects will change with time. For example, when I was younger, I wasn't bothered by scrape flutter in affordable analog tape machines. Just as well, as that's all I could afford at that time. But over the years, I have become much less tolerant of it, presumably because I have learned to identify it, and that is something that can't be unlearned.

I think the reason why so many of us are focusing on the lossy vs. lossless debate and are giving Mr. Atkinson no quarter is due to his article, MP3 vs AAC vs FLAC vs CD.  It demonstrates little more than utter ignorance of lossy formats, how they're developed and how they're tested.

• It's pretty obvious from the MP3 plots that dither or noise shaping was not employed on the output of the MP3 decoder, as the distortion on Figure 2 is clearly quantization noise. While it is likely that some MP3 decoders do not have a ditherered output, some decoders do dither or noise shape, or use 24-bit output (particularly foober2000's, and likely many other mainstream players - although I do not believe a survey has been made on the matter). This doesn't exactly make for a level playing field for comparison with lossless - for subjective or objective comparisons. For those who are educated enough to choose an MP3 decoder with dithered output, the fact that JA chose an undithered decoder output makes this comparison largely meaningless.
• The entire notion of using frequency response plots to estimate the performance of lossy encoders is fundamentally flawed, because they only show one source of distortion while being ignorant of many far more important ones, like pre-echo and stereo imaging. They are almost useless for such evaluations. Relying on plots to tune coders - especially at lower bitrates - is a direct cause of the crappy MP3 sound that JA comments on, and yet he makes the exact same mistake that so many others have made in the past! In a magazine that prides itself on the primacy of subjective evaluations! HA as a forum is profoundly against the use of objective measurements (like frequency response) to evaluate coders, as is jj IIRC, who is about the best expert opinion one is likely to find on the matter...
• Too little information is provided on the Fraunhofer codec that JA uses. Is it fastenc or l3enc? Which version? CBR or VBR? Which other settings? All these things matter tremendously. As Sebastian's latest test results pointed out, 128kbps can vary quite tremendously in quality - and at least one Fraunhoder codec is considered among the worst codecs out there.

Quote from: cliveb on 2009-04-20 11:32:15

What exactly do you mean by "making stuff up"? Are you saying that they are lying, or that there's something going on in the brain that generates the 3D experience? I believe it's the latter.

Well, potentially both.

I don't think 'lying' is the right term, though. I suspect audiophiles (make that 'most' audiophiles) are sincere in their belief that they can hear in 3D, but whether that is merely self-delusion or the result of some psychoacoustic processes is unclear. What would be interesting, however, is if you could 'deprogram' an audiophile, would they stop with the 'holographic imagery' too? Some of the less febrile are starting to question the audiophile articles of faith... will 'soundstaging' disappear along with the need to spend crazy cable cash?

Back up: to the best of my knowledge, none of us has actually offered a public criticism of that article to JA

Quote from: Axon on 2009-04-28 07:41:49

Back up: to the best of my knowledge, none of us has actually offered a public criticism of that article to JA

Perhaps this would be a good time to resurrect the discussion:
http://www.hydrogenaudio.org/forums/index....showtopic=61839

Quote from: Canar on 2009-04-20 22:25:40

Hm. My own experiences with the iPod are that it's a touch on the noisy side...

[...]

Get an in-line attenuator and adjust the effective sensitivity of your earphones to suit your needs for noise-free listening. Or, obtain earphones that are simply less acoustically efficient.

The basic problem that you are worrying about has been around for at least 50 years that I know of. It is very easy to set up a monitoring system that will have an audible noise floor with virtually any piece of equipment.

I think I first experienced with my own equipment  using a pair of Telex headphones and an Eico ST-70 back in 1962 or so. But I worked for an audio store and I had encountered it with the store's equipment before that.

The effective solution that I  found way back then was to build an in-line attenuator and adjust the effective sensitivity of my earphones to suit my needs for relatively noise-free listening.

I think there are two tendencies relevant here. One is the deterioration of the hearing with age; the other is our capacity to learn to discriminate better. Clearly, people can learn to hear lossy-coding artifacts...

It also seems to be the case that, once you get to a reasonable quality level, problems with lossy are occasional glitches, rather than general subtle degradation. One exception might be low-pass filtering, for the young of ear. This at least is the impression I have gained from reading the reports of people with better hearing than me.

So I have a question for Mr Atkinson, and a genuine and straight one: would he agree with this characterisation of (potential) problems with lossy, that there are episodes of glitch, or is it a more pervasive effect?

If it's the odd bit of pre-echo, flanging or whatever, then it would seem to me a question of what you get used to; I'm very very happy not ever to have to hear a sudden pop from vinyl, or anxiety that someone walking heavily in an old house will cause a skip. I'm very happy not to hear tape hiss, though I also admit I must have got more sensitive to it now that it only turns up on old recordings transferred to digital. Maybe it's the case that some people don't mind lossy artifacts, when they happen, because they just don't mind them, rather than because they've got cloth ears or no money; vinyl enthusiasts don't mind the odd vinyl hiccup.

Is it justified to literally carpet bomb others with posts to convince them to refrain from any statements about audio that can't be backed up blind testing, when the same person is putting "pure" .wav files onto his iPod?

Let alone AAC, where probably over 99% of all releases are transparent to 99% of the population*. But preferring "pure .wav" even over losslessly compressed files (about half the storage space, twice as fast transfer speed, less battery consumption) seems quite odd to me when you demand such high standards of rationality from others.

*I have yet to find one person who could ABX non killer samples for a modern AAC encoder with ~192kbit/s. The share of killer samples in my collection is about 1 in 4000. Even the flanging-sensitive 8 kHz cutoff tester  in the c't setup could not tell apart most 256 kbit/s mp3s from WAV. And the available mp3 encoders in 2000 were much worse than what we have today.

PS: I always recommend to my readers that they keep their CDs, as inconvenient as that might be. They are the backup of last resort.

What audible difference is so small as to be impossible to ABX?

For example I think it is not fair to call mr Atkinson's 5/5 result a "lucky coin".

Where is that based on?

How can you blame "subjectivists" to interpret results towards a result they like if we do the same?

I'm no statistic expert, but as far as I know a 5/5 result in an ABX test is only a lucky coin if this happens in multiple test runs for the same subject.

So, if one subject does multiple runs of 5 trials; a 5/5 result may indeed be interpreted as random luck. However, if one person does one run of 5 trials and gets a 100% score, the pval is something like 3%, period. That is all the statistics is able to say.

So although it is very unfortunate that no more test runs were performed, we can say (and should, if we are consistent) that based on these results it was likely that Mr. Atkinson heard a difference.

I think Fremer is XXXX. I think he's a XXXX XXXX. The Beach Boys review(s) quoted earlier are clear evidence, not to mention pretty much all we know about the guy.

That's exactly what I do with the iPod. Thanks for your comment though, it does give me some peace of mind that I was doing the right thing. I have tried the Shure earphones up to the SE530, but liked better the Westone UM2, which is too efficient for its own good. The attenuators that come with those though are very crappy, at some attenuation level they start to mess with high frequencies.

Do you think it is fair to call it a definitive test?

IMO it is too small of a sample to have high confidence in. I always go for 1%.

I think that if I would throw away a 5 trial test on the grounds that it is too small, then applying that standard elsewhere is at least not hypocritical.

I wouldn't call that a lucky coin, I would call that a biased coin. But, I don't see how that applies to the 199x AES demos.

Quote

So, if one subject does multiple runs of 5 trials; a 5/5 result may indeed be interpreted as random luck. However, if one person does one run of 5 trials and gets a 100% score, the pval is something like 3%, period. That is all the statistics is able to say.

Agreed.

The solution to questionable results is to simply run more trials. Run enough independent trials and the results will converge to something that is very reliable.

• A "properly" built amp has the following properties: Low output impedance, flat FR, any noise and distortion well below the threshold* of human hearing.
  [blockquote]
• An absolute threshold of hearing can be agreed on, at least in terms of a rather "safe" lower limit. Relating research is already quite old and nothing that would matter in this context has been discovered for a long time. Current and recent research has discovered many interesting properties of human hearing, but it mostly relates to masking and spatial processing. Both aren't of interest when looking at supposedly neutral elements of a playback chain as amps, DACs, cables, CD players, etc., which are already accurate at much smaller time scales than this research is concerned (exclude lossy encoding)
  
• Any review that states audible differences between neutral components must either be misinformation or the component isn't "properly built" and adds distortion (may sound pleasing to the reviewer for some content).
  [/blockquote]

• Distortion may sound pleasing. An audible hiss can soften your experience, mask artifacts and reflections. A non flat frequency response may sound warm for some content (dull for other). When people prefer hissy, technically severely limited LPs over digital they probably prefer the cloud of imperfections enclosing the signal (as much as they would like to call it being closer to the original).

Don't both sides just come down to faith? Objectivists believe they know what to measure, that they can measure it and that they know the effect it has (if any). Subjectivists believe that's not true and there must be other things involved that haven't been identified or measured yet.

One side trying to convince the other appears fruitless to me

IIRC (and I may not) in his 'real life' Fremer was/is a psychiatrist, and in in NYC that can pay pretty well.

So I have a question for Mr Atkinson, and a genuine and straight one: would he agree with this characterisation of (potential) problems with lossy, that there are episodes of glitch, or is it a more pervasive effect?

If it's the odd bit of pre-echo, flanging or whatever, then it would seem to me a question of what you get used to; I'm very very happy not ever to have to hear a sudden pop from vinyl, or anxiety that someone walking heavily in an old house will cause a skip. I'm very happy not to hear tape hiss, though I also admit I must have got more sensitive to it now that it only turns up on old recordings transferred to digital. Maybe it's the case that some people don't mind lossy artifacts, when they happen, because they just don't mind them, rather than because they've got cloth ears or no money; vinyl enthusiasts don't mind the odd vinyl hiccup.