Paper: Sonic Differences Between Numerically Identical CDs

[jefemeister]

Edit: See rauer's post below for a more up to date, freely available paper by the same Authors on the same subject.

I just got a very interesting paper from our company librarian. It's an AES paper presented back in 1996 titled "An Investigation of the Sonic Differences Between Numerically Identical Compact Discs." The abstract:

Quote:

Originally Posted by AES Paper It has been observed that CD media apparently containing identical data can have varying subjective sound qualities on replay. It is difficult to suggest a theoretical basis for this. This paper describes an analytical study of a variety of CD media and players in an attempt to identify the differences and to isolate their causes. The research is still in progress.

I'm not allowed to post the paper but I will list the section headings under "Some Possible Mechanisms for Audible Differences"

Quote:

Originally Posted by AES Paper Mechanisms for disc-dependent variations through numerical differences: - Accidental processing of the digital audio data - Errors introduced in the transfer of data to the master disc - Variations in the readability of the disc introducing data errors on playback Possible mechanisms for disc-dependent variations without numeric differences: - Variations in Pit geometry introducing EFM jitter - Variations in pit geometry affecting the player's servo system - Variations in the encoded data signal affecting the player's servo system

In general, I don't think there's anything terribly new here, but the explanations under the main headings are worth reading should that be a possibility for you.

 

[rickcr42]

but....but....but...jeff !

Is not the CD perfect ?

at least that is what we were told in the late seventies/early eighties when they stopped making vinyl

BTW-all Sony's fault !

If they did not buy CBS/columbia records taking control over an extensive music library and the main force behind pushing the CD technology this would never have happened

 

[jefemeister]

Quote:

Originally Posted by rickcr42 but....but....but...jeff ! Is not the CD perfect ?

"Perfect Sound Forever"
Wouldn't be the first thing Sony's screwed up over the years!

 

[TWIFOSP]

What kind of measuring device are they using to determine the sound is "different". Pray tell it's not the human ear. So what exactly are they using to detect sonic differences?

I of course haven't done any studies, but intuitively I would think there'd be more variation between the DAC, drivers on the sound reproduction devices, than the 1s and 0s on the CD.

It's quite concievable that there is a very minute difference everytime a driver is used. It's a moving part, so therefore technically wearing out even if it's on a timescale and level that is too small for us to detect.

I dunno, I don't have a hard time buying sonic differences from test to test, but I do have a hard time buying that the 1s and 0s are the cause of variation. There is entirely more to go wrong with the measuring and reproduction devices than binary data.

I would be very much interested in reading the test procedures and the conclusions. Hell, the entire paper

 

[rickcr42]

Quote:

I of course haven't done any studies, but intuitively I would think there'd be more variation between the DAC, drivers on the sound reproduction devices, than the 1s and 0s on the CD.

Obviously for any test to be valid it would have to have a "control" system in place of a certain lvel of quality where the results could not only be compared but MUST be repeatable or the test would not be valid.

Considering the paper is one submitted to the AES I highly doubt it was just some audio hobbyist basement fly by night test on a radio shack mini system but a valid experiment and food for thought.

You or I do not have to beleive the results but the more info you aquire on a subject and compare to the already present knowledge base the better informed your choices can be.

I always like a good read on a new thing,especailly when it agrees with something i already knew instinctivly

 

[rauer]

I found a pretty similar text

 

[jefemeister]

Quote:

Originally Posted by rauer I found a pretty similar text

Nice find! That's by the same author, with a later date, and even mentions the 1996 article as a starting point.

edit: upon looking through the newer paper, a lot of the same info is there but the older paper goes into more discussion about the possible causes of the problem as well as why each cause may or may not have an overall effect. The newer paper seems to leave these "possibilities" out.

 

[eyeteeth]

[size=large]Now that paper should be a *sticky*!!![/size]

 

[TWIFOSP]

The test sounds flawed to me. The paper says that they used DIFFERENT systems and relied on the human ear.

And the author never offers a technical explanation on why it's the CD and not the reproduction system. The article raises the same concerns I did in an earlier post, but never goes as far as to draw conclusions.

They offer statitiscal proof that the tests are valid, which is fine. But they never used a control setup and they relied on a flawed measuring device.

The overall conclusions seem to agree, in that they are more suspicious of the DAC than the CD itself.

 

[markl]

All this does is confirm (to the extent the testing is valid) what many have known all along. If you are a regular at the stevehoffman.tv site, the mecca for knowledge about how different masterings sound, you'll see that for many years, many uber-geeks have sought out specific pressings *of identical masters* for the additional quality achieved under different manufacturing conditions. Many German and Japanese pressings of the *exact same CD data* are often reported to sound subjectively "better" and therefore are very sought-after and expensive. They theorize that the pressing plants in those countries put out a superior product, leading to the audible sonic differences.

Also, geeks on the CD-R burning side of the coin also report that different CD-Rs sound different, even when they are burned with the exact same data. Even in the relatively limited tests that I've done, I agree that different media really can sound different, I've commented about this elsewhere.

So, for me, I am receptive to the idea that different pressings of the same data produced in different plants can sound different, given that different CD-R media can also sound different.

EDIT: I'm going to post the link to the paper over at the stevehoffman site and see what they have to say...

 

[TWIFOSP]

I don't disagree with the idea that having different media can produce different sound. My theory is that it's not the media itself responsible for the change, it's a flaw with the way the reproduction devices handle it.

If we were to take cdplayer with an optical out and measure the data streams between 2 cds that were stamped with the same data, but were suspected of sounding different, we could compare the raw bits coming out. We could also compare the rate and throughput of the data stream.

Assuming they weren't equal since they are supposed to sound different, you could conclude that the laser and reading mechanisms aren't buffering the true data that is coming off the cd. This would be the fault of the player, not the media.

Assuming they were equal, you would then need to test the raw signal coming out of the DAC. If these signals hit the exact same frequencies at the exact same time, then they should sound equal. If they still do not sound equal, the only logical explanation is that there is natural variation with the drivers of the reproduction equipment or the listeners ears. But if there is natural variation, you'd also see it using the same cd over and over, and not just different cds. So this is unlikely.

Assuming the data streams were equal and the frequencies coming out of the DAC were different, then that would prove that the DAC has some kind of variation in the way it's going to translate the numbers into signals.

I still call foul on the test methods outlined in that article. They never once looked at the raw data stored on the cd, or hypothosized failure points outside of the media. They relied on entirely too many variables to draw their conclusions.

Bottom line really is, if the media sounds different, then it's the sources fault for not reading the 1s and 0s correctly and precisely every time.

 

[rickcr42]

Quote:

I don't disagree with the idea that having different media can produce different sound. My theory is that it's not the media itself responsible for the change, it's a flaw with the way the reproduction devices handle it.

OK cool.

But theory based on what man ? Just a feeling ?
Or is this somehow backed up with some more info as in the above examples.Not trying to be a pain here man,just interested in the how and why of your conclusions .

Let the discussion begin !!!!!!

 

[eyeteeth]

"The Medium is the Message"
Marshall McLuhan
(bastardized)

 

[TWIFOSP]

Quote:

Originally Posted by rickcr42 OK cool. But theory based on what man ? Just a feeling ? Or is this somehow backed up with some more info as in the above examples.Not trying to be a pain here man,just interested in the how and why of your conclusions . Let the discussion begin !!!!!!

Based on how digital audio works.

There is more variability in how the electronics work than there is room for variance on the cd. The cd is just full of binary which is later can be read and translated into an audio signal. Which can then be picked up and amplified and reproduced.

So it stands to reason that the reading devices, the various lasers, the chips to buffer the data ect ect, can only read the format one way. 1s and 0s. Assuming there is no difference in the cd's 1s and 0s, then there is absolutely no room for interpertation on the reading aspect of it. It either reads a 1 or 0. There is no logic for anything else.

So if there is no room for alter the incoming bits on the data stream, the only thing left is the audio signal translation of it.

The only thing I can think of would be that some media is just harder to read then others. So the source has to provide more power to read it correctly, and that disrupts the other functions, namely the translation, buffer timing, or the DAC processing, which is what actually alters the sound.

CDs aren't played in real time. It's more or less fast enough for us to think it is, but they aren't. The data has to be read from the cd, buffered into a timing sequence, compiled into a code which can be translated, translated into an audio signal, and the signal is the output.

So technically, the media, while the underlying cause for altering the sound, isn't the reason why the sound is altered. The media is just a representation of the signal that is responsible for creating the sound.

So back to my other post, there are various points of failure, and the media is basically completely ruled out if you can verify that the 1s and 0s are constant.

That is the key point.

Assume we have 2 cds that are identically stamped but sound different.
Is the device reading the same amount of 1s and 0s? If not, the media is probably flawed. If so, and the sound is still different, than some process in the error checking, translation, DAC, or signal strength is being altered within the source.

The other theory is that this is probably just a case of tiny nano scracthes on the cd's surface where some of the 1s and 0s are unreadable. And the error correction algorithms on the source fill in the blanks so to speak, and the sound is altered.

Given that there are basically 783,216,000 imprints on a full cd, statistically given an astronomically high level of quality control there should still be at least 1 flawed imprint per cd (the fabled 6 sigma 2 failures per billion rule, as governed by the rules of natural variation). Given that the publishers are printing out CDs by the millions, I would wager that the quality control is not astronomically high.

Either way, the article never verified that the cds contained the same exact readable information. A fatal flaw if you ask me. Sure, the cds are supposed to contain the same bits and bytes since they are stamped the same. But they never verified if the readable content on the cd was the same. I.E. If you hooked the source up to an optical cable and did some kind of data dump, would the data dump be the same? Or I suppose you could compare them with an electron microscope, but that might take a while

So anyway, to reasons why the test is bunk:
1. It assumes exactly the same data: A CD is just basically a plastic disk stamped with 700 million pits to represent 1s and 0s. The room for failure is quite high.
2. It doesn't measure what happens internally to the data reading and signal assuming that the 783,216,000 bits on the 2 given cds are exactly the same.

 

[markl]

Quote:

Assume we have 2 cds that are identically stamped but sound different.

That's not what this is about though. It's about identical masterings, and therefore, theoretically identical data sounding different on different pressings. I've never read anyone say that two CDs from the same pressing sound different, though they may be out there!