THE TRUTH ABOUT OFFSETS

Offsets seem to confuse a lot of people. In this article I'll try to explain what offsets are and what you can do to correct the offsets. Almost every drive cannot position its reading head exact at the wanted sector. For data CDs that isn't much of a problem since every sector has location information so the drive can easily find the correct sector. Contrary to data CDs, audio CDs contain no sector location information. In other words it's very difficult too locate a certain sector on an audio disc. That means that drives will have an offset when reading audio CDs. For most modern drives this offset is constant so you can correct it once you know it.

We'll use a Plextor PlexWriter 8/20 (PX-W820T) CD recorder as example. This writer has a read offset of -355 samples and a write offset of -30 samples. A negative offset means that the drive starts reading or writing too early and a positive offset that it starts too late.

Let's have a closer look at this. Assume a perfect writer with a write offset of 0 samples.

As you can see and probably expected, the perfect recorder writes all 3 tracks to the CD without any offset or missing samples. However, no writer I know of has a write offset of 0 samples which matches the above example. A more realistic situation is the PlexWriter with -30 samples write offset. If we want to burn the 3 tracks from the above example to CD we get following situation:

What happened? The burning software commands the writer to start writing at position 0 (Addressed position) but due to the write offset the Plextor starts burning the audio data 30 samples too early (Write position). In other words, the whole CD is shifted 30 samples to the left! The effect of this is when you play the burnt CD in a CD player you'll miss the first 30 samples. These samples are written to the CD, but for the CD player in the pregap which is not played. Because of the shift of the audio data the first 30 samples of Track 2 are added to the end of Track 1. The same for Track 3 and Track 2. Since the writer started burning to early it will also stop burning 30 samples too early. That means that the CD player will read the 30 last samples of the CD in the Lead-Out which is silence. The conclusion is thus: no audio data is lost on write, but the first 30 samples are not playable. For the rest there's a 30 samples negative shift over the whole CD what means that the track indices no longer are at the correct positions. It's obviously that a positive write offset would give similar results but with the data shifted in the opposite direction.

Now you know what the write offset is, it's time to take a closer look at the read offset. The read offset is identical to the write offset except that it only occurs during reading (sounds logical, not?). Below is the extraction of a perfect drive with 0 samples read offset.

As expected the drive extracts all 3 tracks from the CD perfectly. No missing samples or shifted data. The extracted tracks are identical to the tracks on the CD. Of course, a drive with read offset 0 is an utopian dream thus below you can find a more real life situation of a Plextor PlexWriter with -355 samples read offset. Before you start e-mailing me that the Plextor has read offset offset of +355 samples in Exact Audio Copy you should know that EAC is somewhat inconsistent with the signs of the offsets. While the write offset in EAC is the real write offset, the read offset is actually the read offset correction, thus the opposite of the read offset. Since EAC 0.9 prebeta 8 the read offset in EAC has the more correct read samples offset correction value name. Anyway, back to our situation with a -355 samples read offset.

The negative read offset means that the Plextor starts reading 355 samples too early. Thus while the software addresses the Plextor (Addressed position) to start reading at Absolute position 0, it will actually start reading 355 prior to that point (Read position). This has as effect that the extracted Track 1 has 355 samples silence from the 2 seconds pregap (PG) at its beginning. This because the Plextor starts reading 355 samples in the pregap where no audio data is (though the Plextor "thinks" it started reading at 0). The last 355 samples of Track 1 are missing. Track 2 has the missing 355 samples of track 1 at its beginning. Similar for Track 3 which has 355 samples of Track 2 at its beginning and is missing the last 355 samples because the Plextor stops reading 355 samples too early. Unlike with the write offset, the read offset introduces irretrievable data loss! The last 355 samples of the CD are lost. on extraction. For a drive with a positive read offset a similar story applies.

In the example of the PlexWriter with write offset -30 we burnt 3 perfect extracted tracks. But we just saw that there's also an offset introduced during extraction, thus what would happen if we want to burn the extracted tracks from the above example to CD. This is the situation that applies to most of us who want to copy a CD but don't know about the offset issue. The extracted tracks all have the 355 samples shift and the last track is 355 samples too short. When burning this with write offset -30 the tracks will be written 30 samples too early. 

Because the tracks already have a 355 samples offset this results in a total offset of 325 samples, the so-called combined read/write offset:

combined read/write offset = read offset - write offset = -325

In English that means that the copy of the CD has a 325 samples offset shift to the original. On writing no data was lost, but since the last 355 samples were already missing the copy is not perfect. Due to the offset the track indices are also positioned wrong and every track starts 325 samples too late. Thus Track 1 starts with 325 samples of the original CD's pregap (silence) and misses its last 325 samples. Track 2 has Track 1's 325 missing samples at it's beginning. Track 3 is missing its last 355 samples due to the read offset but during play 30 samples of the Lead-Out (silence) are added.

Now before you start to panic you should know that 30, 325 and 355 samples are very short times. To get an idea about the real length in time of a sample use following formula:

time = samples / 44,100

Thus 355 samples is 0.00805 seconds, or 8.05 milliseconds, 325 samples is 0.00737 seconds or 7.37 milliseconds and 30 samples is 0.00068 seconds or 0.68 milliseconds. The offset is a time shift of the data and thus it does not affect the audio quality in any way! Since all tracks on a CD end with silence or fade in each other there's no way you can hear the offset. Therefore the whole offset issue may be ignored if you don't care about that...

Anyway, for us perfectionists there must be a way to correct this issue. How would that be done? The answer is very simply: you just have to fool the drive! In the above example we saw a total offset of -325 samples over the whole CD. Thus what if we would apply a combined read/write offset correction of +325 samples?

combined read/write offset correction = - combined read/write offset  = - (-325) = 325

The software tells the drive to start reading at +325 (Addressed position) but physically it is of course reading at position -30 (-355+325 = Read position). That means that the Plextor starts reading 30 samples too early. Thus the first 30 samples of the extracted Track 1 will be silence from the pregap. To get the last 355 samples of Track 1 the drive will read 325 samples into Track 2. However, that means that still 30 samples are missing at the end of Track 1. These 30 samples are extracted at the beginning of Track 2. It's obvious that you will always lose the 30 last samples of the CD. There's no way to avoid that when using the combined read/write offset!

This introduces a new problem, though. For the last 325 samples of Track 3 (the last 30 are missed anyway), the Plextor would have to read into the next track, but there is no next track. Thus it must read into the Lead-Out. The Plextor is able to overread into the Lead-Out and can retrieve the last 325 samples of Track 3. However, many drives can't overread into the Lead-Out and will lose the last 355 samples of a CD. Thus a drive which supports overreading into the Lead-Out will miss only 30 samples while a drive which cannot overread into the Lead-Out will miss 355 samples (325 samples silence due to the overreading + 30 samples missing due to the offset)! Note that this only happens to the last track!

As we already lost 30 samples on extraction no matter whether we have a drive capable of overreading, it should be obvious that a 100% perfect copy is no longer possible. On the other hand, 30 samples is such a short time that you may ignore it, especially since it's at the end of the CD and almost every CD ends with silence anyway.

A similar story applies to a positive read offset where the CD-ROM must be able to overread into the pregap (in EAC the term Lead-In is used, but it actually is the pregap).

Even if we ignore the missing samples there is still a 30 samples offset in the extracted audio tracks. But thanks to the combined read/write offset correction this is corrected on writing as you can see in the example below.

Since the write offset is already included in the combined read/write offset, the CD is written with write offset 0. The data of the extracted audio tracks is shifted 30 samples to the right and because the Plextor starts writing 30 samples to early the resulting total offset is 0. In other words, the offset is corrected. Except for the 30 samples silence in the Lead-Out at the end, the copy is identical to the original!

When a drive was used which is not able to overread into the Lead-Out a similar result is achieved with the difference that the last 355 samples are silent (325 samples silence due to the overreading + 30 samples silence in the Lead-Out):

So, is a perfect copy (without missing samples) impossible then? No, but it's more complicated and Exact Audio Copy is the only software at this moment capable of making such a perfect copy without missing samples. In the above example we used the combined read/write offset correction to correct the offset caused by the read offset and write offset. Because the real read offset is -355 and we used -325 to extract the last 30 samples were missing.

Now this can be avoided when we would use the separate read offset on extraction and write offset on writing. While this sounds logical (and is), it is very difficult to determine the separate offsets and many people will not be able to use this method. Anyway, the read offset and write offset of the Plextor 8/20 are well known thus we can proceed in an attempt to make a perfect duplicate of a CD without missing samples.

As you can see on the image below the CD is now extracted with a read offset correction of +355 samples making the total offset 0.

The software tells the drive to start reading at sample +355 (Addressed position) but physically it is reading at position 0 (Read position = Absolute position). To get the last 355 samples of Track1 the Plextor reads 355 samples into Track 2.

As seen with the combined read/write offset this introduces a problem, though. For the last 355 samples of Track 3, the Plextor would have to read into the next track which does not exist. Thus it must read into the Lead-Out. The Plextor is able to overread into the Lead-Out and can retrieve the last 355 samples of Track 3. However, like said earlier many drives cannot overread into the Lead-Out and will lose the last 355 samples of a CD. Thus a perfect copy is impossible with a drive which cannot overread into the Lead-Out. Remember that this only happens to the last track! And once again, this is such a short time that you will not hear it. Since 99% of all CDs end with silence no music signal will be lost!

When writing these extracted tracks to CD we have to use the -30 samples write offset of the PlexWriter. We'll burn the tracks extracted with a drive supporting overreading into the Lead-Out. In the case of a drive which does not overread into the Lead-Out you just have to keep in mind that the last 355 samples of Track 3 are silence.

Normally the writer would start writing 30 samples too early but the +30 samples write offset correction compensates that resulting in a total offset of 0. The offset is corrected! Thus the software tells the writer to start recording at sample +30 (Addressed position) but physically it is writing at position 0 (Write position = Absolute position). To write the last 30 samples of Track 1 the drive overwrites into Track 2. A new problem raises however. To write the last 30 samples of the last track the drive would have to overwrite into the next track which does not exist. Thus the drive must be able to overwrite into the Lead-Out. If that is the case you can see that the copied CD is a 100% copy of the original with absolutely no missing samples!

Sadly enough, only few burners can overwrite into the Lead-Out and the Plextor PlexWriter 8/20 isn't one of them. That means 30 samples will be missing at the end of of the last track. Thus a perfect copy is impossible with this setup...

Special thanks to Erik Deppe, Andre Wiethoff and Mark Powell for the additional information.