The Absolute Sound (magazine) take on many aspects of computer assisted music reprodu

[sasully]

Thanks again. Again I opened the link you provide. There is a stating thread about the TAS article.
Opening sentence of the opening post slashing the article:

I haven't read the TAS article, but I followed up online and found one of its authors responding to queries about it, on another forum.

I will pass.

How very thoughtful of you.

If you'd read further -- even just to the end of that first post (which was mine, btw) -- you'd see that what is quoted, and commented on, are the words of one of the TAS article's authors, defending his article. They do not bode very well for the article contents themselves.

 

[sasully]

In your first post to this thread, I could hear you licking your chops all the way from here, Amir. Anyway:

The article we are discussing is based on blind tests.

From what I've seen so far, a co-author claims it used a mix of single and double blind tests. That's encouraging, but I have yet to see any of the details necessary to see if they mattered. Given how extraordinary their claims are, that the author admits "Our results lack proper statistical rigor" is not encouraging. (But that wasn't a strong point in your own private blind tests either, was it?) Will this turn out to be a report that adopts the forms and language of science without the substance?

So that remark doesn't apply. Indeed, that is what makes this report interesting in that it starts off with very high level of objectivity yet arrives at conclusions that are normally associated with not-only sighted testing, but wishful thinking! I will post more on this but first, I have to scan a chart they have or else, it is hard to show that. It is a shame the article is not online anywhere to quote and comment .

I do suggest people read the articles before commenting. That, I think is the objective method of evaluating their claims .

'Tis a shame TAS doesn't make the article public. Doubtless the 'meat' of it will become available eventually. But in the meantime is it not fair to comment on what one of the article's co-authors says online about the work?

 

[FrantzM]

Hi

Call me cynical.
I have compared, blind, wav to flac and ape. I can't hear any differences. What I have come to understand is that the audiophile psyche is bent on finding differences. If there is a measurement difference, even if is a millionth of a picovolt it will be used to validate the audiophile idea of differences. This articles is an illustration. I also think that manufacturers have a vested interest in pushing these "differences'.. If there is none and if it becomes clear that there aren't any, a good market ripe for the picking would disappear. When any person with the ability to read and follow simple instructions can put together a music server for less than $500 , that challenges and likely surpasses any transports out there ( I am hereby talking about the Music Server concoted by the people of the PNWAS, presented oin this very forum by Gary L Koh) .. There is apotential stampede toward using any pC to make music thus the the drying up of a perfectly good money stream.
I will not use one more electron on this article it is flawed to almost be suspect... Out for now

 

[sasully]

I was replying to the "hothouse flowers" comment in the context of the level of performance seeked, and expected of the system; not some seemingly bizarre connections between ability and conditions.

If flac sounding different from wav isn't a bizarre connection between 'ability and condition' in audio , then nothing is.

No-one expects an ordinary family car to operate to high standards, but everyone would be disappointed if the Ferrari was ill-mannered in significant ways. So for an audio system: if used to supply background music, no expectations; but if the volume is wound up to realistic, natural levels, and you're listening to how well the qualities of a solo violin or solid drum kicks are reproduced, then the slightest perturbations in the tonal qualities will be very audible, and disturbing if negative in quality.

You are putting the cart before the horse. The central question is whether a particular posited 'perturbation' is actually audible, or whether you are simply imagining that you hear it. Not whether there are ANY slight perturbations that can be audible.

So the argument falls back, as you point out, to which variations are relevant. And the two sides of the fence are those who claim seemingly nonsensical attributes matter, and those who want measurement "proof" for everything and discount all that is not readily amenable to such an approach.

You fail to note the quite well-grounded reasons why such an approach is requested, and the well-known flaws inherent in the 'alternate' approach favored by audiophiles.

The answer, as always, is in the middle, and I favour closer to the first lot than the second, for the reason that I highlighted the key word: the explanations for why something is relevant are often "wacko", but that doesn't invalidate the "truth" that there is an effect, the "problem" is that there still doesn't exist a full understanding of cause and effect in the matter...

Frank

The only 'effect' here that is 'true' is that the author believed they heard something after changing something. But the 'cause' for that 'effect' is what remained to be conclusively determined. Is it what the author believes it to be: something heretofore essentially unknown/unacknowledged? Or is the 'cause' something far more mundane and typical, to be selected from the list of usual suspects?

There is no reason to adopt a 'middle position' in such cases, unless the evidence is truly close to equal on both sides. You rarely find 'middle positions' in science, for example.

 

[rbbert]

...There is no reason to adopt a 'middle position' in such cases, unless the evidence is truly close to equal on both sides. You rarely find 'middle positions' in science, for example.

Surely you don't mean that?? In biological sciences, there are absolutely uncountable "middle positions". Even in physical sciences, in which I am not as expert, there are many "middle positions", even without delving into quantum physics. In behavioral sciences, such as psychoacoustics, I'm not sure there are any positions that are NOT "middle positions".

 

[rbbert]

... I have compared, blind, wav to flac and ape. I can't hear any differences...

Not directly on topic, but try burning audio CD's at your burners maximum speed from FLAC, APE and WAV files. I'd be surprised if you didn't find the CD burned from WAV files not only to sound better, but to be the only one which extracts accurately (i.e., giving you bit-identical rips to the WAV files you started with.)

 

[fas42]

If flac sounding different from wav isn't a bizarre connection between 'ability and condition' in audio , then nothing is.

There is a very reasonable explanation for such, a chain of causality: it's been well established, though you may wish to dispute this, that audio equipment is very sensitive to the quality of the the mains power. For all sorts of reasons: as an electrical engineer I have no problem modelling how conventional circuitry and power supplies are very susceptible to distortion and noise incoming on the AC, on a theoretical setup which has components with real life characteristics as part of the model. A computer processes different files in quite different ways, there is a different pattern of electrical noise and interference occurring within the confines of the processing environment, some of which escapes into the air and some back through the power cord into the general mains supply. And ultimately into the analogue, audio area of things. Hence the two file formats sound "different" ...

The levels may be low, but if you have "sensitive" equipment, and your hearing is attuned to these subtleties, then it all makes perfect sense ...

You are putting the cart before the horse. The central question is whether a particular posited 'perturbation' is actually audible, or whether you are simply imagining that you hear it. Not whether there are ANY slight perturbations that can be audible

.
Once you bring the human organism into the picture, all bets are off as regards whether conventional, scientific, testing will establish whether something is possible or not. As an example, in Australia, the Aboriginal has historically had by Western standards, phenomenal eyesight. Trained by the cultural environment of course, the fact remains that such an individual can be sensitive to seemingly impossible subtle differences in the visible environment, and handles it with aplomb.

You fail to note the quite well-grounded reasons why such an approach is requested, and the well-known flaws inherent in the 'alternate' approach favored by audiophiles.

No-one can say that this whole matter is easy: there are many difficulties to be resolved, not least of which is that testing is still done in "crude" ways, for example, using audio setups which are not up to scratch by audiophile standards. It's a journey to be taken, not a set of of near instant, totally conclusive answers to be extracted.

The only 'effect' here that is 'true' is that the author believed they heard something after changing something. But the 'cause' for that 'effect' is what remained to be conclusively determined. Is it what the author believes it to be: something heretofore essentially unknown/unacknowledged? Or is the 'cause' something far more mundane and typical, to be selected from the list of usual suspects?

Again, this is the hard bit. Enough people "know" that a difference exists. Hence it is the job of others to not just pooh pooh their experiences because it's going to be "hard" to establish the true causes, but be prepared to investigate thoroughly. Everyone is jumping to quick, premature, conclusions, but the whole audio industry is not going to move forward convincingly until someone finally gets to the bottom of these interactions and behaviours ...

Frank

 

[amirm]

From what I've seen so far, a co-author claims it used a mix of single and double blind tests. That's encouraging, but I have yet to see any of the details necessary to see if they mattered.

Well, the detail they provide in the article is that they only use the results where both of the authors in single blind tests arrived at the same conclusion. Beyond this, they are not that specific.

Given how extraordinary their claims are, that the author admits "Our results lack proper statistical rigor" is not encouraging. (But that wasn't a strong point in your own private blind tests either, was it?) Will this turn out to be a report that adopts the forms and language of science without the substance?

Actually, I would be ecstatic with two people in many trials, providing conclusions that agreed with each other. Because that is one heck of a lot more data than we have otherwise where one camp is providing only subjective sighted results, and the other, no data at all other than, "this can't be true!" It is not like these guys ran one or two tests. They ran a ton. All blind. If all of those outcomes point to such consistent views regarding changes in sample rates and such, is pretty compelling relative to what else gets thrown around as substitutes. So no, that is not my main issues with the tests.

It would have been a complete impossibility for them to hire an army of people to conduct so many tests as to satisfy people who are looking for statistical rigor. Me? I am just looking to be more informed than the day before. If dozens of tests point to something that can be substantiated with science of how they work, I am thinking I am way ahead. Not trying to cure a disease, or bring world peace . Just making the hobby a more informed one. Realize that is not your aspirations relative to interpreting such data.

'Tis a shame TAS doesn't make the article public. Doubtless the 'meat' of it will become available eventually. But in the meantime is it not fair to comment on what one of the article's co-authors says online about the work?

I think the issues with the test require more data than they have shared in the remarks. They are assuming people have read the article. But sure, you can comment on anything specific they are saying as long as it doesn't run foul of what else is said in the article.

 

[rbbert]

...If dozens of tests point to something that can be substantiated with science of how they work, I am thinking I am way ahead. Not trying to cure a disease, or bring world peace . Just making the hobby a more informed one...

If you are going to buy into the idea that randomized DBT's are a "gold standard" here, then you also need to have reproducibility of results. History has clearly demonstrated that even large apparently well-designed and controlled tests can have results that are not reproducible due to unappreciated errors in testing methodology (including equipment). We've all already noted that if there are problems getting identical sound from identical audio files, there are almost certainly errors in testing methodology, which then invalidates ALL their findings. I for one think that computer science is well enough established that "identical files" means just that and doesn't really require further testing (except of the hardware used or test design used).

 

[microstrip]

(...) It would have been a complete impossibility for them to hire an army of people to conduct so many tests as to satisfy people who are looking for statistical rigor. Me? I am just looking to be more informed than the day before. If dozens of tests point to something that can be substantiated with science of how they work, I am thinking I am way ahead. Not trying to cure a disease, or bring world peace . Just making the hobby a more informed one. (...)

Great . As people often say (write) ,
+1

 

[fas42]

I for one think that computer science is well enough established that "identical files" means just that and doesn't really require further testing (except of the hardware used or test design used).

Identical only means that while everything remains within the digital domain: once you introduce an analogue element to the situation all bets are off. As a simple example: the reading of a CD at different times on a certain player -- there is a purely analogue mechanism doing the actual reading, the level of cleaniness of the CD, how warmed up the servos and lubricants of the drive are, and myriads of other subtleties all affect whether the digital info is read easily or not: the digital content never changes, but the electrical stress within the mechanism is varying by large degrees. Just saying that something is identical without covering all the bases will only get the testing methodology into hot water ...

Frank

 

[rbbert]

...the digital content never changes...
Frank

That is my point

 

[fas42]

That is my point

But you can't listen to digital files, identical or otherwise: at some point you have to cross over into the analogue world for sound waves in air to be generated, for your ears to register as to what those files mean. And that's when things can, and will go pear-shaped ...

Frank

 

[amirm]

If you are going to buy into the idea that randomized DBT's are a "gold standard" here, then you also need to have reproducibility of results.

I hope to be able to both explain and or duplicate their results before believing them .

 

[rbbert]

I hope to be able to both explain and or duplicate their results before believing them .

This is of course the key. I think I've mentioned that in my limited attempts so far (perhaps affected by expectation bias, assuming that is real), without any statistical rigor whatsoever, I have been unable to confirm many of their findings. Specifically, the different ripping speeds hypothesis and the different brands of CDR's sounding different. "Upsampling" with Izotope RX Pro does seem to produce some audible differences in my different playback attempts, but I would hesitate to call the upsampled versions "better", and that seems to me to be very hardware dependent (also).

 

[DonH50]

Upsampling changs the information content. Exact files should sound the same; something else must have changed if there is a difference in playback sound or measurements.

 

[microstrip]

After people found CD players could sound different, industry and consumers created the idea that the main source of sound differences was the readout mechanism and that a good index of performance was the capability of error correction - every player test included the length of the gap that the reader was able to "read" without "mistracking". A reason was found for differences in sound quality.

However later error analysis carried in CD readers equipped with ICs allowing error diagnosis has shown that the few errors existing in the digital content of the CDs could not explain the difference between the sound of different players or even different copies of the CD - the readout is almost digitally perfect. Also there was no correlation between the readable gap and the sound quality Until now, no one has fully explained exactly what is the correlation between the CD transport and the sound quality.

Two digital files in computers can be bit by bit identical, but are located in different areas of the hard disk. Why should we believe that the complex magnetic reading system of a computer can behave differently from the optical system reading of a CD reader in terms of affecting the sound quality? Everyone knows that after running defragmentation or file access optimization programs files are read much faster - this does not prove it sounds better but shows that although "digital identical" files can behave differently in time.

Surely, if you do not believe that CD transport quality affects the sound quality of a DAC, my questions and arguments have no meaning at all. But IMHO the Mark Levisnon ML31 sounded better than the Madrigal CDT.

 

[rbbert]

...

Two digital files in computers can be bit by bit identical, but are located in different areas of the hard disk. Why should we believe that the complex magnetic reading system of a computer can behave differently from the optical system reading of a CD reader in terms of affecting the sound quality? Everyone knows that after running defragmentation or file access optimization programs files are read much faster - this does not prove it sounds better but shows that although "digital identical" files can behave differently in time...

Not impossible, perhaps, but unlikely. Perhaps some true computer experts can address this? Hard drive information is read into a buffer before being accessed by a program. In the case of Jplay (or several other good music players), this data is then read from the buffer into memory, and only then converted to PCM (from the computer's memory, not the hard drive). That's not even taking into account the fact that the hard drive can read a second's worth of digital audio data in a very small fraction of a second, making sure it is correctly read.

 

[rbbert]

Upsampling changs the information content. Exact files should sound the same; something else must have changed if there is a difference in playback sound or measurements.

Yes, that's why I'm not surprised that the changed files sometimes sound different; "better" or more accurate is another thing entirely.

 

[fas42]

Micro's comment about "digitally identical" files behaving differently in time was most apposite. People seem to forget that all the actual, electronic processes taking place in the computer in extracting and processing the data are all inherently analogue in nature; it's just at a higher, more abstract level that what has taken place is now considered to be a digital process. Again, electrons, and electromagnetic radiation know nothing about the human desire to apply the label "digital" to some activity ...

Frank