| This is quite long:
I'm no expert, but this is how I see it:
Unfortunately, 1010 <> 1010 in any language - there are many things that can
affect the bit stream ariving at the DAC, even if we ignore RF noise, Power
supply modulation by the digital IC's (stick a 'scope on the +Vcc line of a
VAXstation 3100 and look!), PSU modulation and noise from the servos etc.
Let us assume, for now, that the CD transport reads the data with some errors.
Since there is redundancy in the data - CRCs, multiple copies on the disk etc,
some of these errors can be completely corrected - so you might think that they
would be inaudable. In a perfectly designed and executed player that might be
so, but most (all) the current players are flawed in some way (see below).
If there are too many errors (ie the error burst is too long for the error
CORRECTION algorithm (which is the same for all CD players - it's math) to be
used, the player goes into error CONCEALMENT mode - that is, it attempts to
calculate what the missing data might/should/would have been by interpolating
between the last correct data and the next correct data - which could take more
time (see below again). The error CONCEALMENT strategy used by different player
may differ, but I suspect that most of the budget players use one of the
commonly available chip-sets and thus do it much the same way. Some very
expensive DACs (WADIA et al) I believe make use of custom programmed (ROM)
microprocessors and may have several different concealment strategies, which
may/will have different effects on the sound if brought into play. If the error
is too big for concealment, the player simply mutes the output - hence the
clicks you get with a bad disk.
Question - why are some players better at error correction than others.
Answer - they aren't - they may be better at error CONCEALMENT, and they may
have better quality or better aligned transports, resulting in fewer
errors off disk, and thus less need to use the concealment functions.
The first problem that appears in digital systems, even when all data follows
the same path, is that tiny (ie. fractions of a bit time)
timing errors can be introduced - thus even though the data rate may average
44.1Khz over several frames, the position of each 16 bit word will waver about
the ideal position; this is called JITTER and is thought by many to be one of
the main reasons for CD player differences. Even reclocking the data through
a set of shift registers will not remove all the jitter - it's a very difficult
job to design jitter free digital systems. This jitter does not affect the
content of the data, but it does affect the time that the data arrives at the
DAC, and thus the time that the DAC is strobed to start the conversion. Thus
there are small time domain distortions in the analogue output which can be
heard.
The Transport-DAC link is another problem. Most CD players have a lousy drive
circuit for the digital co-ax output. It usually has a series resistor to
protect against shorts, and a fairly high output impedance. Similarly, many
DACs may not have the ideal input impedance to match the transport driver.
In this sort of situation, the transmission characteristics of the cable used
can provide a better or worse match. Mis-matching on this link will change
the shape of the signal, which may/will cause the logic thresholds to shift in
relation to the original pulse generation time. This can lead to futher timing
errors. The US based AUDIO conference has a detailed description of the effect
of fitting a properly designed output stage to a CD transport. (I can't yet
think why most hi-fi reviewers find co-ax better that optical though - it
ought to be better as it avoids most of the problems above, and also cannot
pass out-of-band electrical noise etc - or can it?)
To summarise - even given fully corrected digital datastreams (ie error
concealment was not used), if you feed a given DAC from a number of different
CD transports, there will be differences in the analogue output due to
different amounts of jitter (timing error) produced by each transport, and by
the effects of the data link between the transport and the DAC. This is before
any out-of-band (noise, RF, PSU modulation etc) are considered.
By the way - I am about the biggest cynic I know when it comes to hi-fi. I
still have problems with DIRECTIONAL cables etc! and other funny tweaks.
Dave
|