Hi folks.
Is CD-DA to .wav conversion really a losless procedure ? I would be interested in the technical background.

If it's really losless, if I rip a audio CD to wav files and burn another CD from these wav files and rip those wav files again ... would CD1 be 100% identical to CD2? And would the first set of wav files be 100% identical to the second ?

Thanks.

CDDA is essentially raw, headerless wave data, so a wave file is just the same data but with headers to describe the content. So, yes, in simple terms, both will be the same. There's a bit more to it in practice, but the theory is correct.

It is very well possible to create a bit identical copy of a CDDA disc, as John33 said. If you take good care in all the individual steps involved in the process.
A good starter could be the EAC article in the HA wiki.

Perhaps this thread would fit better in CD Hardware/Software

Not unless the combined read/write offset of the process was zero and samples where overreading/overwriting was required were faithfully reproduced, otherwise certainly not.

There is more to it than just wave files. In some instances a perfect copy may not even be possible.

As far as the concept of sound quality, there will be nothing lost if you go

cdda -> wave (or comprable lossless format) -> cdda

and the rip that created the lossless audio data was free from ripping errors, regardless of the finer points I have raised.

Thanks for all your replies, the move greynol and the link Hanky.

(1) Set aside hardware (CD drive) inaccuracies, resulting in offset, ripping errors etc. it is possible then, right? I was more interested in the theoretic possibility then the practical feasibility.

(2) And greynol, what are you referring to with "it's more than just wave files" and in which cases it might not "even be possible" ?

(3) So, and my last question - I know - will be *very* vague, but I don't know how to make it clearer .... maybe you will manage to read something into it. Some years ago I read an article here on hydrogenaudio (which unfortunatley I can't find anymore) which somehow suggested that an audio CD describes the audio in a different way than the .wav format does and therefore the respective resulting audio signals are not the same. It said that the audio format on the CD is not just the bytes inside a .wav files in a rearranged form (sub-bytes, error-correction data,etc.), but that the audio format itself is different. Both .wav and CDDA are uncompressed, but the transformation is not losless.
To explain it, the article used the analogy of color spaces in digital image processing. Uncompressed image data (like in tiff images) compare to uncompressed audio data (for example .wav) and lossy compressed images (for example .jpeg) compare to lossy compressed audio data (for example .mp3) [... although the lossy part of the analogy is not used for our example here].
The article went on by saying that much like the CMYK and sRGB color space cannot reproduce the same colors, CDDA and .wav don't produce exactly the same audio signal.

I'm convinced it's not true, but wanted to see what the experts say here ...

It's not true.
Both CDDA and WAV uses same technique - PCM (Pulse Code Modulation), so, in terms of colour space, they are both in CMYK (or RGB or smth else).

Yes, but depending on the disc reader, the music can be more or less colorful ;-)

Nicely played the analogy


Relieved to get this confirmation. But what about "it's more than just wave files" then ? Is it about the sub-channels and those other things you have on the CD, but not in the wave file ?

Noooooooooh... the transport must not influence the bits.



The only significant difference is in hidden tracks and track changes.

Play one audio CD on two different CD players and there might well be a difference in what you hear. The easiest to quantify difference, and usually the most obvious, is likely to be loudness (i.e. volume level, not vis a vis the loudness war). This has nothing to do with the CD or the audio on it, it is because of differences in the DACs and output circuits of the CD players; the audio data is identical. If this kind of difference fits your definition of “lossy” ... well that’s ridiculous. Go chase your own tail, I have nothing relevant to say.

First consider a wav file on a computer. It is easy to make one or more copies of it, on different drives, on different partitions, in different folders, in the same folder with different names. There are various tools that can tell you the copies are identical to the original, bit for bit. Hopefully you can agree this copying is lossless.

Write the file to an audio CD-R. Extract it back to disk. Use the tools to compare the extraction to the file used to write the CD. The extracted file will be identical to the original. Do it a thousand times and the result will still be identical, bit for bit.

It is possible for errors to occur reading or writing to the optical media, but if the equipment and blanks are good, these are rare. If they do occur, they are not “lossy” they are just errors. Except for the few bytes with errors, the rest will be bit for bit identical to the original.

The audio data is encoded on the audio CD in a very different manner than it is on a hard disk or on a data CD. It is not simply a matter of a different header or other control information. Also, the data is arranged in a very different way along the audio track. These are mainly tricks for ensuring the data gets off the disk during playback with as few errors as possible. If you were to make an exact image of the audio CD on your hard drive, you could not compare it to the original hard disk wave file with those tools I mentioned and find any kind of match -- unless you used some program that first converted the image to a wav file(s).

The data is still the same, however, there is nothing lost or “lossy” because of the encoding method. Likewise, offset and gaps between tracks are irrelevant to lossy/not lossy. These later may be important to someone who cares about minuscule differences between optical disk copies, but they are not part of the audio data.

Right.

Non-null samples that exist well into the lead-in and lead-out. I can't think of any burning software that can recreate this phenomena. This doesn't include hardware limitations such as the ability to overread and overwrite.

No. The transformation definitely is lossless. (please, no Clinton-isms)

Thanks for your patience greynol.
whooo ... never heard of that ... is that dangerous ?
I mean, how does this manifest itself when I listen to the CD? And if burning software cannot reproduce it, then my song will be truncated ?
Sorry, I'm not English. What are Clinton-isms? Twisting of words ?
Anyway, I wrote that I was convinced it wouldn't be true, but I wanted to free myself (and others) of any possible doubts ... thx.

No,but their interpretation.If I'm not mistaken the same CD three different CD players (let's say a very old one, a contemporary Hi-Fi device and a Disc-man) can result in different listening experience even with the same headphones attached, no ?

Even if the audio material is gapless (for example a live recording) but still split into different tracks ? If there are minuscule differences, you might hear clicks at those gaps ...

<searches for the crassest of jokes, gives up>

It doesn't; it's essentially silence.

You'll be able to reproduce enough samples as defined by the TOC minus the half-dozen or so offset samples unless your burner can write these. It's less than trivial.

"It depends upon what the meaning of the word 'is' is."

Nope.

EDIT: Ok, I'm going to bed now.

Hm, I'm confused now ... are you talking to me, or to kjoonlee ? Remember we have expressed opposite points of view ...

ok, is that good or bad for me? LOL
OK,I guess the bottom line is that in the worst case I loose a bit of near-silence, right ?
Ah yes, I laughed a lot when I heard that. There were a lots of others but of course that's the most obvious one ... Well, I have already had gap issues due to inappropriate choice for gap handling in EAC, but I'm probably replying out of context.

Well, that's what I said, too, wasn't it?

Nooooooooo .... AndyH-ha shows up ... you know what Andy, you've been haunting me since the first day I've poked my nose into HA. One of the first threads I read here was this monster-philosophical-flame-thread and - funnily enough - I just happened to spend my last two hours with re-reading it and albeit very interesting I swore to myself that the one thing I'll never ever do is start a philosophical discussion with you about the meaning of lossless or lossy ....

Not only because, in this case here, I have never given any definition whatsoever of lossy and certainly did not imply that resulting audio signal of different hardware devices was any prove of it not being lossless, which is absurd of course ... please be aware that this point was brought up much later and not even by me ...

Yes, that's basically the same example that I've used.

Hmmm ??? Encoded differently ? I thought both are encoded via PCM ?

You mean EFM? Well very obviously there are differences in the physical storage of the data, but this thread was not about the physical but the logical level.

And here I was getting the distinct feeling that you never sleep.

AndyH-ha means the way data is written on the media.


And there's your mistake. Audio CD is a very simple format, it doesn't have physical and logical levels. And because of it such programs as EAC exists - CDDA has no sector markup on the media, no CRC data... Just lead-in and lead-out markers, where's the beginning and where's the end - it's all that you know for sure. In the middle of the track you have to control your reading all by yourself.

Ok, then I was using "logical" inappropriately. I actually wanted to distinguish between the physical level (channel bits, EFM) and the audio-encoding (PCM).

contradicts:and(which translates to:)... no ?

Both, WAV and CDDA are PCM encoded audio data.
They contain exactly the same data representing the binary code of the quantized values of the audio signal - just the container format is different.

Best regards,
Chris

Yes, audio CD has error correction bits, but they are not meant to always correct errors 100%, but rather to minimize uncorrected errors. Uncorrected errors are then handled by interpolation etc.

A data CD, on the other hand, has additional CRC correction information to make the data (hopefully) 100% corrected. The additional CRC data explains why the capacity of a data CD is less than that of an audio CD.

>data (hopefully) 100% corrected

not just corrected but detected if there is an error, audio cds have less correction and less detection of errors when compared to data cds.

My intention was not to ruffle any feathers, merely to point out that it is easy to empirically determine for oneself that the audio can be transferred from media to media indefinitely without any changes. However, if one actually want to observe this, there is a detail that must be accommodated. Possibly some means of comparing the original with the DAE copy automatically takes account of this, but I’m not sure that is possible.

The extracted copy will have some small number of additional samples at the beginning of each file, and possibly also at the end, samples that were not in the original. These may all be digital zeros but it has been too long since I did the experiments; I don’t remember if there may be some small amount of very low level noise.

Since the comparison between the copies is on a sample by sample basis, mis-aligning the two files by even one sample gives very different results. One must therefore determine what has been added and start the comparison after it, or delete it. Therefore, as in so many other situations, anyone attempting to play with computer audio without a good audio editor that functions at the sample level seems destined for frustration.

These extra samples do not change an of the original audio data, they are simply extra. They may or may not be on the CD; they could simply be an artifact of extracting. Their presence, or number, may depend on the hardware and or the software used. I only know they were on my test tracks.

Because of these extra samples, finding, in the DAE, the place where the original sample string starts can sometimes be quite tedious. It is thus helpful to mark the beginning, or near the beginning, of the original file(s) by pulling up (or down) a single sample (before writing to CD-R) and using that as the alignment point.

And you have your wild fantasies backed up by some facts?
Reed-Solomon != CRC.

You are correct, Reed-Solomon is not the same as CRC, but to call his statement "wild fantasies" seems a bit harsh.

A data CD has additional layers of (Reed Solomon) error correction which an audio CD does not. Many people misspeak and call all digital error correction "CRC". Replace all instances of "CRC" with "Reed-Solomon" and his statement is true. Regardless, his point is correct, even if his terms are not.

Maybe I was using "logical" inappropriately. As said before (see above), I actually wanted to distinguish between the physical level (channel bits, EFM) and the logical one (tracks, PCM audio-encoding, etc.). Wikipedia (which of course is not authoritative) also distinguishes between "physical frames" and "logical format"/"logical structure" :

Ouch, the colours, they burn.

I wanted to make sure that the two concepts are unambiguously shown in different colors ... (apparently it worked )

Yes, I realized that just after I switched the PC off and went to sleep. Before, I just reacted to the claim as a whole which was kind of funny since CRC is not designed/employed to correct errors at all just to detect them.And spreading such misinformation in a technical forum where other can read it doesn't help it either.

Are you sure data CDs feature data correction at all ?? The checksums are normally only used to detect errors. Checksums don't provide information for replacing missing data as far as I know.

So he left out the term "interleaved". I can see how someone could mistake CRC for CIRC, even if they are completely different things.

>Are you sure data CDs feature data correction at all ??
Yes, 3 levels worth, your previous post said as much. Did you actually read what you quoted?

sorry, I can't follow you greynol. You mean my wiki quotes (in red and green)? They weren't about data CDs.

Start reading them at

oh yes ... I can see that now ... must have had a blond moment ...

I did an experiment where I :
1. Ripped an audio CD (Katie Mehlua)
2. Burned those tracks as a audio CD-R
3. Played back that CD on my Sony stand-alone DVD-player, and recorded it using spdif into my PC
4. Did a sample-by-sample comparision of the wav file from 1) and the wav file from 3)

The comparision turned out to be identical as long as I shifted one or the other by an integer amout of samples (aligning the tracks).

Does this answer the questions?

-k

Just my two cents. I haven't trhoughtfully what others have said, so sorry if this has been said already.

CD-DA, in terms of the audio data is equal to a 44100hz 16bit stereo WAV file.
So, in theory, given you have the ability to make a perfect data transfer, YES, you can make a perfectly lossless transfer from CD-DA to WAV and back to CD-DA.

The problem is the physical format of CDDA is not as reliable as your hard drive's filesystem or that of a CDROM/DVDROM etc and retrieving the digital audio data from it without any loss can prove difficult, specially with consumer level equipment. This is one of the reasons, I guess, that when it comes to CDDA you don't copy the tracks but EXTRACT them.
Proof of this is the fact that, after decades of the format's release and after a long long time of computer CD drives and recorders being commercially available there's still development going into trying to extract CD audio with the least amount of errors/loss possible.

Now, if you have some 44.1khz 16bit stereo wave data in your computer and you author that into a CD in CDDA format, that process is lossless (unless your recorder screws up or you use a bad disc), but once you try to retrieve that data from the CD you COULD run into one of the issues that plage CDDA extraction.

The final answer to your question is that, yes, it's lossless, but there are shortcomings that may thwart your results, and these are quite common.

If you want to embark yourself into the eternal quest for the perfect CDDA extraction you should get into the likes of EAC and dBpoweramp, and get an adecuate CD drives.

knutinh and radorn, thanks for your valuable posts and the experiment. All good things to know. Together with the previous replies the I think the initial question can be considered solved by now. But in the course of the discussion there have been a couple of interesting side-issues, so feel free to comment on them further (for example I find the distinction between physical and logical level as quoted above not quite clear yet) or anything else which fits here. Thanks to everybody who helped me understanding the initial issue.