lotus

previous page: 4-18-1] Why doesn't the audio data on the copy match the original?
  
page up: CD-Recordable FAQ
  
next page: 4-19] Digital audio extraction of a track is shifted slightly

4-18-2] The audio data matches exactly, why do they sound different?




Description

This article is from the CD-Recordable FAQ, by Andy McFadden (fadden@fadden.com) with numerous contributions by others.

4-18-2] The audio data matches exactly, why do they sound different?

(2004/12/10)

Suppose you extract the audio track from the copy, and it's an exact binary
match of the track you wrote from your hard drive, but the CDs don't sound
quite the same. What then?

Most people don't notice any difference between originals and duplicates.
Some people notice subtle differences, some people notice huge differences;
on better CD players, the differences are harder to hear. Some say CD-R is
better, some say worse. While it's true that "bits are bits", there *are*
reasons why CD-Rs may sound different even when the data matches exactly.

An excellent paper on the subject is "The Numerically-Identical CD Mystery:
A Study in Perception versus Measurement" by Ian Dennis, Julian Dunn,
and Doug Carson, submitted to the Audio Engineering Society (Preprint
4339, 101st AES convention). It's available for download in PDF form at
http://www.prismsound.com/m_r_downloads/cdinvest.pdf. The paper is primarily
concerned with why pressed CDs created at different plants or with different
methods sound different, but the observations are relevant to CD-R as well.

The conclusions in the paper suggest that low-frequency modulations in the
disc affect the servo and motor electronics, causing distortion noticeable
to a critical listener.


One prominent theory is jitter. This isn't the DAE "jitter" described
in section (2-15), but rather a timebase error. A good overview can be
found in the jitter article on http://www.digido.com/. A brief explanation
follows.

The digital-to-analog ("D/A") conversion at the output of the CD player
is driven by a clock in the CD player. The clock is tied into feedback
mechanisms that keep the disc spinning at the proper speed. If the digital
signal being read from the disc has irregular timing, small errors can
be induced in the output clock. Even if the CD player gets all of the
digital bits accurately, it will produce inferior results if the timing
of the bits on the disc isn't precise. Put another way, something has to
send a sample to the speakers 44100 times per second, and if it's speeding
up and slowing down many times each second your ears are going to notice.

There is some question as to whether the clock driving the output will
actually be affected by the input. If the output clock in the CD player
is isolated and stable, jitter from the CD will not affect it.

If you play a CD digitally (e.g. by ripping it and then playing it through
a sound card), the quality of the CD doesn't matter, because it's the
timing of the clock in the sound card that drives the D/A conversion.

It has been asserted that the clocking of bits on a CD-R isn't as precise
as on a pressed CD. Writing at different speeds on different types of
media requires adjustments to the "write strategy" (section (3-31)) that
can result in individual "marks" being sloppier than at other speeds.
This could account for inferior -- or at least different -- sound.

Yamaha believes they have found a partial solution for jitter problems
with their Audio Master Quality feature. See section (2-41).

There do not appear to be any carefully constructed (double-blind)
tests published on the web that confirm that jitter is the cause of this
phenomenon. The "Numerically-Identical CD Mystery" paper rejects jitter
as a possible cause.


Some people have asserted that *any* two CDs, pressed or otherwise, will
sound slightly different. Some claim to hear differences in identical
CDs from different pressing plants. The former is silly, but the latter
has a lot of anecdotal evidence to support it.

The manual for the CDD2000 reportedly states that the drive uses 4x
oversampling when playing pressed CDs, but switches to 1x for CD-R.
This affects the quality of the D/A conversion, and can make an audible
difference.

http://www.mrichter.com/cdr/primer/losses.htm has some further thoughts,
including a table showing signal level differences.

An extremely technical introduction to CD reading is available at
http://www.tc.umn.edu/~erick205/Papers/paper.html. This may shed some
light on why reading audio CDs is difficult, as well as explain concepts
like aliasing and dither.

If you are finding your CD-Rs to be noticeably inferior, try different media,
different write speeds, a different player, or perhaps a different recorder.
There is some evidence that different brands of media and recorders may
work better for audio, but in the end it's a highly subjective matter.
Some people say CD-Rs sound worse, some people say they sound better (and
some people think vinyl records are still the best).


 

Continue to:













TOP
previous page: 4-18-1] Why doesn't the audio data on the copy match the original?
  
page up: CD-Recordable FAQ
  
next page: 4-19] Digital audio extraction of a track is shifted slightly