It's been a while since I decided to write a post about offsets.
The offset is the amount of time the drive that reads the audio CD will mistake the beginning and end of the CD. A read offset of +600 means that the extracted track will begin 600 samples into the resulting wav. An offset of -600, and the extraction will miss the 600 first samples of music.
A sample is 1/44100th of a second, thus a -600 offset misses about 1/75 of a second.
Why don't the drives read properly ? At the beginning, the technology was unable to read exactly at a given place on a CD. It was not the same each time.
Engineers being aware of the limitations of CD readers, the CDs were pressed with about 600 samples of null audio data at the beginning (588 according to Eric Charton "La gravure des CD 2e édition, MacMillan ed.), and several thousand at the end. Then no drive should miss anything.
In addition, the red book standart says that before the first track should stand 2 seconds of silent audio : the "pregap", improperly called lead-in in EAC ( the lead-in is in fact some non-audio data standing before the pregap). And after the last track, the "lead-out" should also begin with silent audio data. So if the drive overread the CD because of a big offset, there will be no error : just silence added to the result.
Now, with the improvments in accuracy, especially in computers, a CD Reader is able to read exactly at the same place each time. I've not got the Red Book, but I think the place where the audio begins is not officially defined. As far as I know, it is defined in millimeters from the center of the CD ! Then, the groove having a coil shape, it doesn't define the angle at which the audio begins. So a little spread is allowed, not all CDs beginning exactly at the same place.
Copying CDs, the error about the position will increase each time a copy of a copy is made, until it gets bigger than the silence buffers at the ends, then some music will be cut. That's why EAC features the offset correction ability : comparing a copy to an original, it tells how much offsetted they are, and then allows to pad the extracted wavs so that the copy made with them is exactly at the same place as the original.
But moving data inside a fixed-lenght window, there is one end that will disappear, while some silence will be added at the other end.
Then EAC allows for separate read and write offset correction. Andre Wiethoff seek and found some CDs made from analog sources where nobody payed attention to the original background noise. His drive allowing overreading into the pregap and the lead out, he could read some more data at the end, in order to see the whole tracks, and to see where the noise was cut, marking the real end of the CD. He found several different positions, but 6 of them were the same. He then included some of the data of the CD in a database featured in EAC, so that any other people having the same CDs could run an automatic offset detection that will tell how much their drives are offsetted from the 6 matching CDs reference.
Lead by curiosity, I performed the same test, and found +26 two times, and +2350 one time for an Artec 34x CDROM drive (see my report at
http://pageperso.aol.fr/Lyonpio2001/offset.htm ) with three CDs where I could see some analog noise being cut at the end of last track. With the +26 value, I deduced the read offset of my Yamaha 6416S drive, is was 12 samples different from what is reported in SatCP's database for similar drives from Andre's reference. (
http://www.ping.be/satcp/ , links at the bottom of the page)
But those are not the only different references known !
Since countless days and week and monthes (and years), people keep asking the mailing list (
http://groups.yahoo.com/group/eac ) why they never get the same offsets using CDs of EAC's database. I can't find back the post where someone gave about 10 results, with only two or three matching. Andre also warns that only the same pressings as the ones recorded in the database will give correct results.
So why do we blind ourselves thinking there should be only one valid offset, while the evidence that 80 % or the official CDs use a different offset is before our eyes ??
We should adapt the read offset differently for each CD we read, since it has been burned with a different offset !!
There are some especially sensitive cases : Cds made on computers from MP3s or tapes or vinyls etc... The files used to make the tracks have no silence at the ends, and the user usually use an automatic 2 seconds gap insertion between the tracks. So at the beginning and at the end of the CD, the music will begin and end exactly at the places defined by the burner offset. This CD must be read with trial and error until the correct correction is found, looking with a wav editor where the music stops or begins.
There is also the case of CDs where the original analog noise is present in all the pregap and lead-out. Then a fade-in and fade-out should be performed, together with inserting silence, to achieve a perfect result.
There are also some CDs that begins too early, eg : Mike Oldfield Discovery EMI 786426 CDM01 begins 7643 samples before track 1, into the pregap. A special offset correction of -8000 is needed to properly rip it. And the cut I very obvious, listening to the CD!
In conclusion, the offset correction can help preventing the shifting of data if several generations of copies of copies of copies are made, but it doesn't help to achieve a perfect CDR. If you seek for perfection, use an accurate wav editor, and check by yourself what you burn !
Links :
Offsets are arbitrary (pio2001) :
http://groups.yahoo.com/group/eac/message/6593
Mathematical war between Pio2001 and Matthias trying to figure out ofsets without external reference :
http://www.digital-inn.de/showthread.php3?threadid=3079
How to "read" your read offset with noisy CDs (pio2001) :
http://pageperso.aol.fr/Lyonpio2001/offset.htm