In article , Paul

wrote:


If you are referring to the format in which data is represented on the
audio disc:

It is not correct to say that the information recorded onto audio CDs
is "simply streamed". Audio CD employes quite a complex and powerful
set of systems for redundancy and error detection and correction. If
you don't understand CIRC then you may not know the details of this,
but it is present. Indeed, this is why you can drill a 'small hole' in
the information layer, yet then be able to read the audio information
with no actual errors or loss to the audio data.

No actual errors or data loss????? Clearly there are errors and data
loss. I assume you are suggesting that the loss can be 'rebuilt' to be
100% as the original?

You may be confusing two distinct issues.

1) Random and unrepeatable errors - e.g. due to noise in the reading system
- always lead to a non-zero chance of there being some errors at the
channel bit level which are either undetected by the correction processes,
or can't be corrected.

2) Systematic errors - e.g. a hole in the information layer - do *not*
always have a non-zero chance of producing undetected or uncorrectable
errors. They either will, or will not, depending on the specific details of
the systematic error.

Thus if a a hole has lost data which means that the original information
cannot be recovered correctly, then no amount of re-tries will help. The
data is lost. But if the hole has *not* lost data in this way, then the
hole does not prevent a single read from recovering the orginal
information.

Type (1) can lead to data loss regardless of if any type (2) problem is
present or not.


I feel if this were indeed possible the computer industry would be using
it. What may be possible is interpolation. This may be acceptable for
audio but disaster for computer data.

Accordng to the Red Book standards you can expect interpolation for burst
errors (e.g. holes) which extend over more than 2.5 mm of the information
layer. But for shorter burst errors, no interpolation is needed to correct
for the hole.

Interpolation may be needed to deal with random errors. However if you look
at the Red Book standards this should be very rare. Here 'rare' can be
quantified as follows:

For the sake of example, assume a channel bit error rate (BER) of 0.0001.

This means that interpolations will tend to arise only at the rate of 1
sample value per 10 CDs. In practice it is likely that these will pass
unnoticd when they occur.

It means that undetected errors (a 'click') will occur at a rate Philips
described as 'negligable'. This may seem vague, but the reason was that it
is a rate of 1 sample per many hundreds of thousands of CDs. Even a channel
bit BER of 0.001 only produces a 'click' for one sample per 750 CDs.

The snag, of course, is that not all CDs are well produced. :-)


In the case of audio, we are still left with corrupt, inaccurate
information. Clearly then, this is of no help in the search for High
Fidelity. It is interesting to read the CRC from an audio CD. Play a
track and note the CRC value. Play the same track again and see if the
CRC is consistent. I use Audiograbber to do this little test.

Well, I have in the past done the test of playing a CD, and recording the
output via SPDIF using a CD audio recorder. I did this a few times for a
few different tracks. Then loaded the copies onto my computer. They were
all bit-for-bit identical to each other, and the original.

I have also read the same track repeatedly onto the computer. Same results.

There will be occasions when this does not occur. No real system can
totally exclude the chance that random non-systematic errors will corrupt
the result. But in reality this seems rare. In practice, you may find that
faint clicks and pops caused by next door's fridge are a worse problem.



Hence even though most domestic audio players read the disc at x1 rate
with no 're-tries', they still do not have to actually read every
single channel bit from the disc. They can accidentally 'miss' bursts
of discs *and the result following error detection and correction will
generally be the intended audio data with no errors.* This is the
purpose of the systems Philips/Sony built into the Red Book specs.

'generally be the intended audio'? Not good.

Depends on your definition of "not good". Is not a rate of audible errors
of well below one sample per 1000 CDs not good enough? If so, then I am
afraid that no system anyone is using would suffice for your purposes.

Engineering isn't about making systems 'perfect'. That is left to
theologians. I don't know if they have made any better audio systems than
engineers, though... :-)

Slainte,

Jim

--
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Audio Misc http://www.st-and.demon.co.uk/AudioMisc/index.html
Armstrong Audio http://www.st-and.demon.co.uk/Audio/armstrong.html
Barbirolli Soc. http://www.st-and.demon.co.uk/JBSoc/JBSoc.html