Nelson wrote:

> I have some old (well not that old) CD-R's which have developed a
> problem which looks like the dye is deteriorating. Because it is only
> in spots, some of the data is readable. Furthermore, Toast is able to
> create a disk image of the disc on some drives. Since read errors
> don't appear to be reported unless repeated attempts fail, I can't tell
> how bad the problem is. Furthermore, I'm not sure the "read" data is
> the "correct" data.
> The disc utilities I have only check the file structure and do not
> verify the individual files. I presume making an image forces reading
> of all the data on the disc, but just because it can be read doesn't
> mean it is not corrupt does it?
> - Does the CD contain error detection data, eg a checksum for every
> sector or block?

All CDs (audio and data) have an underlying two-stage error correction
mechanism at a very low level. This is good enough for audio CDs to
correct most minor errors.

Data CDs (including CD-R) have an additional error detection and
correction mechanism on top of the underlying one, to increase the
reliability compared to audio CDs, since it is possible to interpolate
missing data on an audio CD, but not on a data CD.

If you can create a disk image of a data CD-R with Toast on a particular
drive, and it didn't report any errors while creating the image, then
that should be a good copy of the CD-R. It will have copied every block
(sector) in the used portion of the CD-R.

Each individual sector on a data CD has a 32-bit CRC to detect errors,
and the chance of an uncorrected error slipping through the CRC is
pretty small.

An overview of the gruesome technical details can be found here:

> - Does anyone know of an application which will examine a damaged disc
> and associate damaged sectors/blocks with files?

Can't help with that one.

> This is a pretty depressing problem since the reason for writing the
> data to CD-R's in the first place was the unreliability of magnetic
> media.
> Thanks.

David Empson