FCFIFRAGA%CIUC2.UC.RCCN.PT@CUNYVM.CUNY.EDU (FRANCISCO AMARAL F FRAGA) (07/02/90)
Technical aspects of high end CD players
The following lines were writen with the idea of strating some
exchange of views about technical aspects of CD players and may be,
much more interesting, their relation to sound quality.
High end CD players seem to be separated in two groups : the
"upgraded" and the "innovative". What I call the "upgraded" are
good machines ofmedium performance which are tweaked by small
manufacturers in order to improve the sound of the basic machine.
Main areas of improvement are : laser servo-systems and reading
optics, system clock, power supply quality, DAC selection and
analogue filtering and buffering. Some manufacturers build separate
decoders with similar improvements. One important aspect is that
the digital processing chips are the same as in the cheap players
and the sonic signature of digital oversampling filter of the basic
player will always be present.
"Innovative" machines are almost always separate processors
(it is claimed to improve sound to keep the transport and decoder
in separate boxes, and besides, some of the manufacturers didn't
have the capacity to built transport machines, but the scene is
changing quickly). The innovative aspect of the machine is in the
digital oversampling filter : they don't use low cost chips
available in the market and because they don't have neither access
nor the skills to design custom made chips they use digital signal
processors (DSP's) to perform filtering and oversampling. These
chips are very fast microprocessors optimized to perform high speed
processing of signals in real time. With these chips the designer
can built algorithms to improve the sound of the player and even
tailor the sound of the player to his preferences, much like a
record cartridge.
Just a few examples from recent magazines press releases and
reviews
Krell CD 64x player : Motorola 56000 family DSP's 64x
oversamplingand BurrBrown 16 bits dac's.
Krell CD 16x player : Motorola 56000 family DSP 16x
oversampling software and BurrBrown 18 bits high grade PCM58K DAC
(PCM58 is available in several grades, but only K selection
approaches true 16 bit performance even untrimmed)
Wadia converters use ATT 32000 series of DSP's. The Wadia 1000
is 16x oversampling and uses 16 bits BurrBrown DAC's, the Wadia
2000 is 64x oversampling and uses a custom made dac.
Cambridge will launch briefly a 12x oversampling converter
based on Texas Instruments TMS 32020 dsp, unknown dac. May be
Philips TDA1541 as they have been using in their players?
Madrigal (the people from Mark Levinson) converter uses a
Motorola 56000 machine with BurrBrown DAC's.
Just to end Roland Research is developing a CD converter, but
its technical aspects seem to be one of the best kept secrets in
audio industry.
Some very personal foot notes:
1. DSP's seem a proper way to built high end CD players. They have
a lot of computing power which when properly used allow the
designer to tailor the sound of the equipment he is developing and
even deal with problems which happen in later stages such as DAC's
and buffers. Let's hope that DSP based converters become cheaper in
the future.
2. The manufacturers say DSP based players can easily be upgraded:
all you need to upgrade your machine is a new set of ROM or EPROM
each time the designer improves its software. It's true, but this
year DSP's are much more powerful than last year state of the art
and manufacturers will be using them very soon. DAC performance is
improving too, so DAC's should be easily exchangeable (plug in
boards as used by some converters seems a good idea).
3. Dynamic behavior of DACs is quite important and has been under
specified /studied. Manufacturers usually only specify settling
time for worst case all 0's to all 1's or something similar. This
kind of transition never happens in a CD player and something as
"small signal settling time" would be meaningful. One must remember
that 64x oversampling means 2.822 Mhz DAC frequency and that an
audio signal can't change a lot in less than 400ns!
4. Texas Instruments new TMS 32030 seems to be a very promising DSP
for CD converters. If someone knows something about such a player
please inform me!
5. One awkward aspect of high end CD players is the standard serial
link between transport and converter. If there is a standard 3 wire
bus (the so called I2S bus) why don't high end manufacturers use
it for linking their systems? They should need to add the error
signal available in every player and it seems unwise to decode a
signal already existing somewhere to economize some cable and plugs
or some optical fibers. Sony had an interesting idea in their
Reference 2 box machine : the master clock was placed in the
converter box and sent back to the transport.
6. This last point was raised to me when recently hearing some CD
transport and separate converters at a high end dealer. The
differences between transports were much more noticeable than
expected and even some incompatibilities were evident. My guess is
that they should be traced to the connection between the separate
units. May be the input signal separator is quite sensitive to
level and shape of the digital input?
7. I would like to exchange opinions about error correction in CD
players. This seems to be a high risk subject but seems to be
completely forgotten in high end magazines, who seem to be more
interested in the capabilities of playing holes almost never
existing in real CDs unless you have tested your new Black and
Decker drill on your CD collection! Is there anyone else interested
in this matter?
Francisco Fraga