Why "Identical" WAV Files Don't Compare the Same Way Twice in a Row

[Anderton]

This is a really interesting link:

 

http://web.ncf.ca/aa571/daefaq.htm

 

Some of the thoughts in here might explain the "I know these CDs were burned from the same program, but they sound different" phenomenon. Very thought-provoking, especially since they went to the trouble to do serious file comparisons.

 

Maybe this is why PC Audio Labs insisted that my computer have a Plextor drive...

[MikeRivers]

So this flies in the face of the proverbial "everybody" who knows that digital copies are perfectly accurate. But we already know that. If there's some form of computer system (including the error correction subsystem of a CD drive) it's going to have limitations if its primary goal is to keep playing audio as best it can.

 

Hard drivers and memory also use error correction. Does anyone question the accuracy of their reconstructed data? Does audio played off your hard drive sound the same every time? Are you really listening?

 

OK, so a hard drive has more error correction than a CD because they weren't trying to squeeze a specified amount of data on the drive (a CD has to accommodate a minimum playing capacity), they made the data checking and correction as robust as they thought was necessary and labeled the drive according to how much data would fit, allowing for the overhead for correction.

 

Could we make a CD capable of more accurate data correction by, for instance, using a blank with greater capacity and using that additional capacity for data correction? Would that help?

 

While this is interesting to discuss as perfectionists, doesn't the CD sound different in your living room than mine, or in your studio and in your car? Doesn't that difference swamp out the difference between two consecutive playings, or two pressings from the same master played on the same player?

[Lee Knight]

The article speaks about DAE. Extraction, right? Does this all hold true for writing?

[blue2blue]

Bait and switch.

 

I thought this was going to be about identical wav data file comparisons, not comparisons between the file products of digital audio extraction from audio CDs-- which I had assumed everyone realized was subject to serious accuracy issues. (One reason we've always told recordists to archive their music projects as data files [WAV, AIFF] on CD-ROMs, of course.)

 

Nothing new in this article but it's a worthwhile collection of information and it's nice they have at least a few links to more authoritative info sources.

 

We should hope that recording milieu people don't conflate DAE inaccuracy with data file comparisons or confuse the issues with CD-Audio format reads and error correction with the rigorous error-checking (not correction) in a CD-ROM read.

 

(We should hope but we do know that there are a lot of different folks in studio land and some of them would appear to lean more to the intuitive side of things. Some of the same intuitive qualities that may lead them to be good recordists/producers in practice may make them less than perfectly attuned to technical issues and the sort of logic one needs to use to sort them out.)

[Dean Roddey]

It's pretty easy to do accurate DAE. I've written an extractor as part of our product. The big issue with CD drives (in non-ROM mode anyway), is that they were intended for streaming, and the way they work (or used to work) can mean that they will sometimes return the same sector twice or skip one, because the read positioning isn't completely accurate. Though, to be fair, I don't think any modern drives have this problem really. Even when it happens, a single sector is only like 1/77th of a second, so it's a pretty short blip.

 

But, even then, it's easy to get around for DAE purposes. The DAE code just reads backwards a few sectors every time. Then it lines up the new block with the previous one by seeing where they overlap, by just a simple brute force comparison of the bytes. And for even more safety, you can do multiple reads of the data until you get the same data back a few times, in addition to the alignment stuff. Unless the CD is physically damaged, that will pretty much insure that the errors are quite minimal.

 

But certainly you should enver depend on the music style CD as a data storage format, since it just wasn't ever intended for that. When formatted for data then has all the usual redundancy and error correction.