Pio2001

I think that the process of error correction, with the red lights, consists in rereading 16 times until 8 at least (or 9 at least ?) match.

Is this also correct when C2 is enabled ? It would be better if it would reread until no C2 is reported anymore.

A test ( http://www.hydrogenaudio.org/index.p...ST&f=20&t=9049 ) showed that reading twice might not be as secure as previously thought. Note that when you get wrong results in non C2 mode, the CRCs are likely the same (both wrong), thus there is no way to detect the problem.
In this case, it would be much better to reread until no C2 is reported, otherwise, enabling C2 would be unuseful, since the extra (consistent) errors detected by C2 would not be corrected.

In this test,a reference file was burned on CD, then the CD was damaged and read twice in burst mode.

There were 10641 errors.
1930 (18.1 %) of them were consistent, thus would have not been detected without C2.
22 of them were consistant and isolated (far away from any inconsistent error). Thus would not have been corrected for sure (other consistent ones may be corrected when a nearby inconsistent error is detected).

Thus the accuracy of EAC without C2 in this test would have been between 81.9 and 99.8 %.
Now if we take a drive with 99.9 or 100% C2 accuracy, if the error correction is done by rereading until a majority match, the accuracy with C2 will be 81.9 to 99.8 % for this test. If rereading is done until no C2 occurs, it would be 99.9 or 100 %

__________________
Pio2001

Pio2001

Well, actually, if 10641 is the total numbers of errors and if 1930 were at the same place, and if the drive properly interpolates, then a file should have 1930+(10641-1930)/2=6085 errors in average.
Thus the proportion of consistent errors is 1930/6085*100 = 32% not 18 !

__________________
Pio2001

JeanLuc

Hmmm ... I switched over to read audio data in secure mode the following way some time ago:

C2: On
Caching: On (except for the drives that have been reported as non-caching by both Feurio! and EAC)
Acc. Stream: On

My Read Mode is now "Test & Copy" for single tracks and manual double extraction for CD images (copy image & cue first, then copy range afterwards and compare CRC's) ... that way, I can still detect random errors by double-reading everything and benefit from C2 error reporting ...

Up to the present day, I haven't encountered a single CRC mismatch on "normal wear" CD's yet ...

__________________
The Name was "Plex The Ripper", not "Jack The Ripper"

JeanLuc

BTW ... Isn't that what Plextools do (with the exception that the maximum number of re-reads is limited) ?

__________________
The Name was "Plex The Ripper", not "Jack The Ripper"

Pio2001

Yes, Plextols does it.

But what you call "random errors" are "inconsistent errors". When they are corrected, if some consistent errors are left, you get "no errors occured", because the rereading believes it has corrected them, and same CRC, since they are consistent.

Thus you can't be sure that your rips are error free, since you can't detect the 32 % consistent errors with C2 off, nor with C2 on (they will flash the red light, but be left uncorrected), nor by comparing CRC (same and both wrong).

__________________
Pio2001

JeanLuc

If I use two different drives (which I do) ... I think this 32%-value is somewhat lower ... I don't think that any given drive will read the same data when inconsistent errors occur ... and, after all ... CIRC has not been designed for perfection in terms of data integrity ...

I will do some testing on my own because there have been several occasions where I have encountered that plextools simply failed to correct errors, even when reporting everything was all right ... (while EAC could repeatedly rip the same results on different drives) ...

__________________
The Name was "Plex The Ripper", not "Jack The Ripper"

Pio2001

It depends if the consistent error is isolated. If so, any drive should interpolate it the same way. And CIRC is designed so as it is very hard to have contiguous errors.

A consistent error just means that the data is completely destroyed, thus the error always occurs.

Edit : Anyway, only tests will give useful informations.

__________________
Pio2001

JeanLuc

Right ... CDS200 works that way ... isolated and repeated CU errors which will be treated differently by most drives.

CIRC itself IMHO has been invented with a single paramount object ... to make occuring errors due to media degradation/scratches/fingerprints inaudible ... nothing more, nothing less. C1 errors can be fully corrected, as well as C2 if there are not too many of them ... otherwise, Interpolation/Muting/Holding/Skipping will kick in with different quality levels of error correction/concealment capabilities according to the chipset manufacturer.

If the music industry would have chosen a different format 25 years ago, they would probably have had less available space on a CD (see advanced yellow book error correction) but more redundant data that could be used for error correction and not error concealment ... but no one could (or wanted to) foresee the problems we are facing now with problematic media or dying CDR's (optical or magnetic pcm studio tech was brand new at that time as well so no experience with older media was available) ...

For the time being, I ripped the same erraneous track in 24 different combinations (drive/software/speeds) now ... and i will carry on ... there is some SP/DIF recording to do ... ;-)

__________________
The Name was "Plex The Ripper", not "Jack The Ripper"

Pio2001

Ah yes, the different drives handling error differently. I forgot about this. An error can be uncorrectable for one drive, and not for another.

But to study this, you need to analyze what kind of error correction your drive can perform. There are not many different ways to correct errors, while there are many drives on the market.
Thus many drives will necessarily correct errors the same way.

__________________
Pio2001

DrDoogie

Um, C2 can be corrected?

Can anyone point me to the section of ECMA 130 that states this?

If not, considering that the "CI" in CIRC stands for "Cross-Interleaved", don't you need a LOT of erroneous adjacent blocks in order to get a C2 error?

In short, does the scratched CD really have a chance of being read correctly simply by re-reading, given that there are lotsa errors to begin with?

edit: spelling and content

JeanLuc

IIRC, most chipsets can correct 2 (some 4) wrong symbols per block on the C2 stage ... but I may as well be wrong

By the way ... Plextools are unable to retrieve consecutive matching CRC's while reporting that every error has been corrected by re-reading on that special test track that I use for my personal extraction test ...

__________________
The Name was "Plex The Ripper", not "Jack The Ripper"

Pio2001

ECMA-130 is not Red Book. It doesn't states this. However, what is written in section C4 page 33 implies it : "The error correction encoder C2 genrates..."

Since yes, let's consider the number of erroneous adjacent blocks needed in order to get an uncorrectable C2 error.
It depends on the chipset of the drive used. It varies from 8 to 17 adjacent blocks.

It depends on the amount of damage that the data has suffered. We saw that more than 2/3 of the errors on a lighly damaged CDs are not consistent.

__________________
Pio2001