What is the point of expensive CD players?
 

On Mon, 13 Nov 2017 09:59:52 +0000, Jim Lesurf wrote:

In article , Woody
wrote:


Interesting observation.

For some reason I always thought my first 14-bit Philips (CD104?)
sounded better than anything I had later, and that the one that I
bought to replace it some years later (16-bit parallel) also sounded
better. That machine now sits with a very elderly lady we know and I
will reclaim it when she passes. Comparison with my present Marantz
CD5400SE will be interesting.

The first player I had was the first gen Marantz using the 14-bit x4
Philips chipset. Happy with it for about a decade. Although I did add
some 'Toko' analogue low pass filters that rolled off at about 19 kHz as
that seemed to make the results sound nicer to my ears. Possibly because
it cut down the signal levels slightly going into the amp.


You seem to have forgotten that one of the benefits of 4x oversampling
eliminated the need for a brick wall anti-aliasing filter to allow a
filter with a much gentler roll off slope to be used which produced much
less in-band ripples in its response curve.

It's just possible that your Toko analogue filter may have been
filtering off low level supersonic products in the 20 to 60KHz range that
were upsetting the amplifier's stability, perhaps creating
intermodulation products of its own, leading to a slightly dirtier sound
as a result. The other thing about Philips's rather neat use of 4 times
oversampling with 14 bit DACs to achieve the same accuracy and dynamic
range of a perfect 16 bit DAC was the improved accuracy of monotonicity
over that of the typical consumer grade 16 bit DACs of the day. It really
was a very clever move on the part of Philips at the time.

Nowadays, this oversampling principal has been taken to its ultimate
conclusion with very high speed single bit DACs that oversample with a
factor of 65536 (or is it 32768? - 1440MHz is the sampling frequency ISTR)
times the 44.1KHz sampling rate which corresponds to a sampling frequency
of some 2.88GHz. Whatever it is (32768 or 65536) it's an extremely high
sampling rate whichever way you look at it - makes a 44.1KHz sampling
rate look positively pedestrian indeed.

The oversampling frequencies might seem rather extreme but the big
payback is that a single bit DAC doesn't need the extreme accuracies
required by the last two or three MSBs used by 16 and 14 bit parallel
converters of old. Indeed, not even the accuracy of the next to LSB of
such converters, just a reasonable accuracy to avoid clipping in the
following analogue stages of the DAC which error can be compensated for
with a simple 'volume control' trim pot if required. Monotonicity
guaranteed, absolutely! :-)


--
Johnny B Good

What is the point of expensive CD players?
 

Johnny B Good wrote:

---------------------

Jim Lesurf wrote:

Interesting observation.

For some reason I always thought my first 14-bit Philips (CD104?)
sounded better than anything I had later, and that the one that I
bought to replace it some years later (16-bit parallel) also sounded
better. That machine now sits with a very elderly lady we know and I
will reclaim it when she passes. Comparison with my present Marantz
CD5400SE will be interesting.

The first player I had was the first gen Marantz using the 14-bit x4
Philips chipset. Happy with it for about a decade. Although I did add
some 'Toko' analogue low pass filters that rolled off at about 19 kHz as
that seemed to make the results sound nicer to my ears. Possibly because
it cut down the signal levels slightly going into the amp.


You seem to have forgotten that one of the benefits of 4x oversampling
eliminated the need for a brick wall anti-aliasing filter to allow a
filter with a much gentler roll off slope to be used which produced much
less in-band ripples in its response curve.


** Jim has not "forgotten" - impossible since it is not true.

The filter coming after D to A conversion is called a "reconstruction" filter and has nothing to do with aliasing.

The primary filter used by Philips in its dual DAC 14 x 4 players was a digital filter IC ( SAA7030 ) that created a high order LPF that still needed analogue filtering afterwards to reduce supersonic artefacts to tolerable levels. However, enough remained to prevent THD analysers reading the residual properly - hence very few reviewers could achieve the claimed 0.005%.

OTOH, Sony CD101s achieved 0.0006% THD using a spectrum analyser.
Later testing with a "dithered" disk showed no measureable harmonics.



The other thing about Philips's rather neat use of 4 times
oversampling with 14 bit DACs to achieve the same accuracy and dynamic
range of a perfect 16 bit DAC was the improved accuracy of monotonicity
over that of the typical consumer grade 16 bit DACs of the day. It really
was a very clever move on the part of Philips at the time.


**Philips used a 10 bit active plus 4 bit passive DAC ( TDA1540 ) which was not particularly linear. It was not near as good in those respects as the single 16 bit DAC used in early Sony players. In order to measure the linearity and THD of a Philips /Marantz player one needed to install a filter like to Toyo in the signal path.

I studied the topic carefully at the time and did my own testing too.


..... Phil

What is the point of expensive CD players?
 

In article , Johnny B Good
wrote:
On Mon, 13 Nov 2017 09:59:52 +0000, Jim Lesurf wrote:

In article , Woody
wrote:


Interesting observation.

For some reason I always thought my first 14-bit Philips (CD104?)
sounded better than anything I had later, and that the one that I
bought to replace it some years later (16-bit parallel) also sounded
better. That machine now sits with a very elderly lady we know and I
will reclaim it when she passes. Comparison with my present Marantz
CD5400SE will be interesting.

The first player I had was the first gen Marantz using the 14-bit x4
Philips chipset. Happy with it for about a decade. Although I did add
some 'Toko' analogue low pass filters that rolled off at about 19 kHz
as that seemed to make the results sound nicer to my ears. Possibly
because it cut down the signal levels slightly going into the amp.


You seem to have forgotten that one of the benefits of 4x oversampling
eliminated the need for a brick wall anti-aliasing filter to allow a
filter with a much gentler roll off slope to be used which produced much
less in-band ripples in its response curve.

Actually, the rather depends on the implimentation. In practice did this
because at the time they couldn't mass-manufacture good 16 bits DACs and
digital reconstruction filters. Turned out to be an ingenious trick. :-)

But it didn't change the point that adding the filters gave a sound which
I preferred. Nor that the real DACs still did produce some HF aliasing.

It's just possible that your Toko analogue filter may have been
filtering off low level supersonic products in the 20 to 60KHz range
that were upsetting the amplifier's stability, perhaps creating
intermodulation products of its own, leading to a slightly dirtier sound
as a result.

Yes. It is also possible that the way the filter dropped the signal level
by about 6dB made it easier for the following amplifier stages to cope. I
was using an Armstrong 626 at the time and this feeds inputs though gain
buffers before the volume control. The amount of feedback is modest by
modern standards, so reducing the level may have reduced the distortion.

FWIW I used to use the orginal chipset as the basis of my lectures to
undergrads on this topic. You can find the notes on the "Scots Guide". I
got copy of the Phiips Tech Rev that described them when Audio CD was
lauched.


Nowadays, this oversampling principal has been taken to its ultimate
conclusion with very high speed single bit DACs that oversample with a
factor of 65536 (or is it 32768? - 1440MHz is the sampling frequency
ISTR) times the 44.1KHz sampling rate which corresponds to a sampling
frequency of some 2.88GHz. Whatever it is (32768 or 65536) it's an
extremely high sampling rate whichever way you look at it - makes a
44.1KHz sampling rate look positively pedestrian indeed.

1 bit dacs have some basic problems with limited noise shaping and idler /
latchup risks. Many systems are likely to be 'low bit' to avoid this. DSD
has to play tricks to dodge this bullet.


The oversampling frequencies might seem rather extreme but the big
payback is that a single bit DAC doesn't need the extreme accuracies
required by the last two or three MSBs used by 16 and 14 bit parallel
converters of old. Indeed, not even the accuracy of the next to LSB of
such converters, just a reasonable accuracy to avoid clipping in the
following analogue stages of the DAC which error can be compensated for
with a simple 'volume control' trim pot if required. Monotonicity
guaranteed, absolutely! :-)

Not absolutely. A claim wise engineers would avoid. :-)

The change in design simply displaces what causes problems to another area.
e.g. if the pulse density distribution affects the reference level or
generates idler patterns in the noise shaping. You can model these effects
but can be quite difficult to reliable predict them in real DACs. Hence one
of the reasons Lip****z warned people against using DSD for archiving, etc.

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics https://www.st-andrews.ac.uk/~www_pa...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html