DVX
Instruction
Manual
37
However, BLER only tells you how many errors were generated per second, it doesn’t tell you anything about the
severity of those errors. Therefore, it is important to look at all the different
types
of errors generated. Just because a
disc has a low BLER, doesn’t mean the disc is good. For instance, it is quite possible for a disc to have a low BLER,
but have many uncorrectable errors due to local defects. The smaller errors that are correctable in the C1 decoder are
considered random errors. Larger errors like E22 and E32 are considered burst errors and are generally caused by local
defects. As you might imagine, the sequence E11, E21, E31, E12, E22, E32 represents errors of increasing severity.
Why E32 is Considered Uncorrectable
Although it is possible under some circumstances to correct up to four bad symbols at the second stage, not all players
can do this. Until recently, most players could only correct two bad symbols at the C2 stage. For these players, E32
would be uncorrectable. In order to have a high probability of a disc working in any drive, we consider E32 an
uncorrectable error, even though some drives may be able to correct it.
This is also the rationale for not allowing E22 or E32 errors on a data disc. The earliest generation of CD players could
only correct one bad symbol at the C2 stage. As a result, an E22 error (two bad symbols at the second stage) would be
uncorrectable on these drives. In order to have the highest confidence in a data disc, it should have no E22 or E32
errors. Also, keep in mind that this requirement is for new discs, as made. Obviously, the quality will degrade with use
and age. Making discs with E22 or E32 errors does not leave adequate margin for future degradation. This is not an
onerous requirement, because with modern equipment, there is no reason to make discs with E22 or E32 errors.
How DVD Error Correction Works
DVD error correction also uses a Reed-Solomon product code for error correction. The primary difference from CD is
the size of the correction block, and the lack of interleaving.
CD error correction uses an error correction block of only 24 user bytes. CDs were designed for audio, and you don’t
want an uncorrectable block to be too noticeable. DVDs, on the other hand, use an error correction block of 32 kB. It
turns out that the error correction capability of Reed-Solomon product codes increases with the size of the block, so this
gives a much greater error correction capability, which is required due to the small size of the pits.
DVD error correction also works in two stages, but instead of C1 and C2 as on a CD, DVD error correction arranges
the data into 208 rows and 182 columns. Each row and column has it’s own parity bytes. Correction of the rows is
known as “Inner Parity” or PI. Correction of the columns is known as “Outer Parity” or PO. The rows are corrected
first, so uncorrected bytes in a row may still be corrected by the outer parity correction.
Therefore, PI Fails (PIF) are not fatal. But PO fails cannot be corrected, and should not be allowed.
Summary of Error Names
For CDs
, each count of:
E11means there was one bad byte in a data frame at the C1 stage, which was corrected.
E21 means there were two bad bytes in a data frame at the C1 stage, which were corrected.
E31 means there was three or more bad bytes in a data frame at the C1 stage, which were not corrected.
E12 means there was one bad byte in a data frame at the C2 stage, which was corrected.
E22 means there were two bad bytes in a data frame at the C2 stage, which were corrected.
E32 means there were three or more bad bytes in a data frame at the C2 stage, which were not corrected.
BLER is the total number of errors encountered at the C1 stage (=E11 + E21 + E31)
AVE values displayed are the averages over the whole of the disc that has been tested.
PK values displayed are the largest one-second samples encountered over the whole of the disc that has been tested.
10PK values are the largest 10-second samples encountered over the whole of the disc that has been tested.
CD Errors are reported for each one second of program time. Since there are 7350 data frames per second, the
maximum number of errors per second is 7350.
