"Bedouin" wrote in message

"Jim Lesurf" wrote in message
...
In article ,
"Bedouin" In principle, losing some of the 'squareness' of the edges
should not matter if the receiver can still read the data OK.
Although there may be a possible effect of 'data induced jitter' of
the kind that Julian Dunn and others have discussed in the past. In
practice I have my doubts about this being a serious problem,
though, in most casts.

It was the introduction of Jitter due to the rounding of the signal
that I was thinking about.

I'm not sure what part of the "square wave" a DAC typically detects.

I've probed real-world SP/DIF lines and found that there typically is no
square wave on a SP/DIF line. More often than not, it most resembles a
modulated sine wave. This is because the output of most SP/DIF coax outputs
is routed through a transformer with controlled losses at high frequencies.
The low-pass characteristic is there for so the device passes DFCC part 15.
Signal detection is based on something like a Schmidt trigger doing level
high-hyseterisis level detection.

This scheme is of course sensitive to noise pickup. Any jitter that is added
at this point is supposed to be removed further on down the line.

If it is a zero crossing point which in many ways seems the most
sensible then loss of the high frequencies will have little effect,
if it works on a point away from the zero crossing then loss of the
high frequencies could be very significant.

Here's a spec sheet for a relevant part:

http://www.sc-elec.demon.co.uk/cs8411.pdf

"Arny Krueger" wrote in message
...
"Bedouin" wrote in message
I'm not sure what part of the "square wave" a DAC typically detects.

I've probed real-world SP/DIF lines and found that there typically is no
square wave on a SP/DIF line. More often than not, it most resembles a
modulated sine wave. This is because the output of most SP/DIF coax
outputs
is routed through a transformer with controlled losses at high
frequencies.
The low-pass characteristic is there for so the device passes DFCC part
15.
Signal detection is based on something like a Schmidt trigger doing level
high-hyseterisis level detection.

This scheme is of course sensitive to noise pickup. Any jitter that is
added
at this point is supposed to be removed further on down the line.


But I believe that such waveforms are a significant cause of jitter, and
most of the CD Clock upgrades work in part by squaring off the waveform. In
practice DACs could reduce the jitter by processing the input, but the
reported benefits of the clock upgrades suggest that that is not done as
effectively as it might be.

"Arny Krueger" wrote in message
...
"Bedouin" wrote in message
I'm not sure what part of the "square wave" a DAC typically detects.

I've probed real-world SP/DIF lines and found that there typically is no
square wave on a SP/DIF line. More often than not, it most resembles a
modulated sine wave. This is because the output of most SP/DIF coax
outputs
is routed through a transformer with controlled losses at high
frequencies.
The low-pass characteristic is there for so the device passes DFCC part
15.
Signal detection is based on something like a Schmidt trigger doing level
high-hyseterisis level detection.

This scheme is of course sensitive to noise pickup. Any jitter that is
added
at this point is supposed to be removed further on down the line.


But I believe that such waveforms are a significant cause of jitter, and
most of the CD Clock upgrades work in part by squaring off the waveform. In
practice DACs could reduce the jitter by processing the input, but the
reported benefits of the clock upgrades suggest that that is not done as
effectively as it might be.

On Mon, 01 Dec 2003 23:20:58 GMT, "Bedouin"
wrote:

"Jim Lesurf" wrote in message
...
In article , "Bedouin"
In principle, losing some of the 'squareness' of the edges should not
matter if the receiver can still read the data OK. Although there may be a
possible effect of 'data induced jitter' of the kind that Julian Dunn and
others have discussed in the past. In practice I have my doubts about this
being a serious problem, though, in most casts.

It was the introduction of Jitter due to the rounding of the signal that I
was thinking about.

I'm not sure what part of the "square wave" a DAC typically detects. If it
is a zero crossing point which in many ways seems the most sensible then
loss of the high frequencies will have little effect, if it works on a point
away from the zero crossing then loss of the high frequencies could be very
significant.

The point is not the nominal value of the trigger voltage, but the
*slope* of the edges of the 'square' wave. Many poorer S/PDIF input
receivers have trigger points which are a simple fraction of the rail
voltage, so that noise on the rail voltage directly affects the
trigger point. It follows that jitter is directly proportional to the
slope of edge, since an instantaneous transition (i.e. a truly
vertical edge) would cause no jitter. How much of this jitter is
*audible*, is another question, but it's a problem which was solved
many decades ago in the comms industry. As ever, so-called 'high end'
audio is decades out of date with real engineering practice.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On Mon, 01 Dec 2003 23:20:58 GMT, "Bedouin"
wrote:

"Jim Lesurf" wrote in message
...
In article , "Bedouin"
In principle, losing some of the 'squareness' of the edges should not
matter if the receiver can still read the data OK. Although there may be a
possible effect of 'data induced jitter' of the kind that Julian Dunn and
others have discussed in the past. In practice I have my doubts about this
being a serious problem, though, in most casts.

It was the introduction of Jitter due to the rounding of the signal that I
was thinking about.

I'm not sure what part of the "square wave" a DAC typically detects. If it
is a zero crossing point which in many ways seems the most sensible then
loss of the high frequencies will have little effect, if it works on a point
away from the zero crossing then loss of the high frequencies could be very
significant.

The point is not the nominal value of the trigger voltage, but the
*slope* of the edges of the 'square' wave. Many poorer S/PDIF input
receivers have trigger points which are a simple fraction of the rail
voltage, so that noise on the rail voltage directly affects the
trigger point. It follows that jitter is directly proportional to the
slope of edge, since an instantaneous transition (i.e. a truly
vertical edge) would cause no jitter. How much of this jitter is
*audible*, is another question, but it's a problem which was solved
many decades ago in the comms industry. As ever, so-called 'high end'
audio is decades out of date with real engineering practice.
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On Tue, 02 Dec 2003 06:54:10 GMT, "Bedouin"
wrote:

But I believe that such waveforms are a significant cause of jitter, and
most of the CD Clock upgrades work in part by squaring off the waveform. In
practice DACs could reduce the jitter by processing the input, but the
reported benefits of the clock upgrades suggest that that is not done as
effectively as it might be.

It varies. Established engineering-led companies like Meridian and the
old Audiolab made precision DACs with excellent jitter suppression,
but of course the idiot 'high end' audiophiles thought that
incompetent crap like the Mark Levinson 'reference' DAC was better
because you could hear the effect of different transports. Priceless!
--

Stewart Pinkerton | Music is Art - Audio is Engineering

On Tue, 02 Dec 2003 06:54:10 GMT, "Bedouin"
wrote:

But I believe that such waveforms are a significant cause of jitter, and
most of the CD Clock upgrades work in part by squaring off the waveform. In
practice DACs could reduce the jitter by processing the input, but the
reported benefits of the clock upgrades suggest that that is not done as
effectively as it might be.

It varies. Established engineering-led companies like Meridian and the
old Audiolab made precision DACs with excellent jitter suppression,
but of course the idiot 'high end' audiophiles thought that
incompetent crap like the Mark Levinson 'reference' DAC was better
because you could hear the effect of different transports. Priceless!
--

Stewart Pinkerton | Music is Art - Audio is Engineering

In article , "Bedouin"
bedouin@yonderblue wrote:
"Jim Lesurf" wrote in message
...
In article ,
"Bedouin" In principle, losing some of the 'squareness' of the edges
should not matter if the receiver can still read the data OK. Although
there may be a possible effect of 'data induced jitter' of the kind
that Julian Dunn and others have discussed in the past. In practice I
have my doubts about this being a serious problem, though, in most
casts.

It was the introduction of Jitter due to the rounding of the signal that
I was thinking about.

OK. This has been considered as a potential source of problems by some
people. In theory, you can see mechanisms for this, but in practice my own
experience is that it isn't really much of a problem in the domestic
systems I've encountered. Hence I have my doubts that we need loose sleep
worrying about it.

I'm not sure what part of the "square wave" a DAC typically detects.

This depends upon the DAC I assume. However there are two (related) isses
here. One is reading in the correct series of '1's and '0's. The other is
then being able to convert these onto analog levels with the required
regularity. There may well be some 'jitter' in terms of the instants at
which each successive bit is read in, but this does not necessarily mean
that the jitter then appears at the output. It may, or may not. Depends
upon the DAC/receiver design, and the details of the jitter. Also, it is
debatable if low levels of jitter matter as much as some magazines, etc,
assume.

If it is a zero crossing point which in many ways seems the most
sensible then loss of the high frequencies will have little effect,

I don't know what systems are most popular. Arny is probably correct that
Shmidt trigger systems may be common. These don't specifically work on the
crossing point as they use a pair of levels (this suppresses noise to some
extent and avoids spurious multiple-detections, etc).

Schmidt is probably better than a simple fixed-level system. However it
would perhaps be better to recover the clock, then detect at the 'mid time'
level of each bit slot by running a sampling clock phase-shifted, based
upon the recovered clock. This keeps you as far as possible from any edge
or noise effects. No idea if any commercial makers use this approach,
though as it is quite complex.

if it works on a point away from the zero crossing then loss of the high
frequencies could be very significant.

Yes. In terms of information theory, the precise shape of the 'edges' does
not matter, The reciever only has to decide if the the signal during each
time slot is a '1' or a '0'. In practice, though, the shape may matter to
some extent, depending upon how the data is read in. I'd expect most decent
DAC/recievers to deal with most inputs without problems, 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

In article , "Bedouin"
bedouin@yonderblue wrote:
"Jim Lesurf" wrote in message
...
In article ,
"Bedouin" In principle, losing some of the 'squareness' of the edges
should not matter if the receiver can still read the data OK. Although
there may be a possible effect of 'data induced jitter' of the kind
that Julian Dunn and others have discussed in the past. In practice I
have my doubts about this being a serious problem, though, in most
casts.

It was the introduction of Jitter due to the rounding of the signal that
I was thinking about.

OK. This has been considered as a potential source of problems by some
people. In theory, you can see mechanisms for this, but in practice my own
experience is that it isn't really much of a problem in the domestic
systems I've encountered. Hence I have my doubts that we need loose sleep
worrying about it.

I'm not sure what part of the "square wave" a DAC typically detects.

This depends upon the DAC I assume. However there are two (related) isses
here. One is reading in the correct series of '1's and '0's. The other is
then being able to convert these onto analog levels with the required
regularity. There may well be some 'jitter' in terms of the instants at
which each successive bit is read in, but this does not necessarily mean
that the jitter then appears at the output. It may, or may not. Depends
upon the DAC/receiver design, and the details of the jitter. Also, it is
debatable if low levels of jitter matter as much as some magazines, etc,
assume.

If it is a zero crossing point which in many ways seems the most
sensible then loss of the high frequencies will have little effect,

I don't know what systems are most popular. Arny is probably correct that
Shmidt trigger systems may be common. These don't specifically work on the
crossing point as they use a pair of levels (this suppresses noise to some
extent and avoids spurious multiple-detections, etc).

Schmidt is probably better than a simple fixed-level system. However it
would perhaps be better to recover the clock, then detect at the 'mid time'
level of each bit slot by running a sampling clock phase-shifted, based
upon the recovered clock. This keeps you as far as possible from any edge
or noise effects. No idea if any commercial makers use this approach,
though as it is quite complex.

if it works on a point away from the zero crossing then loss of the high
frequencies could be very significant.

Yes. In terms of information theory, the precise shape of the 'edges' does
not matter, The reciever only has to decide if the the signal during each
time slot is a '1' or a '0'. In practice, though, the shape may matter to
some extent, depending upon how the data is read in. I'd expect most decent
DAC/recievers to deal with most inputs without problems, 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

In article , Ian Molton
wrote:
On Mon, 01 Dec 2003 23:20:58 GMT "Bedouin"
wrote:


I'm not sure what part of the "square wave" a DAC typically detects.
If it is a zero crossing point which in many ways seems the most
sensible then loss of the high frequencies will have little effect, if
it works on a point away from the zero crossing then loss of the high
frequencies could be very significant.

If the jitter is considered noise, and thus phaseless, then it may have
no effect on the reconstruction at all, as 16 bits are needed to form a
sample and the jitter may very well totally cancel itself out in that
time (plus there plenty of time for a PLL to smooth things over there
AIUI.

This should be so *if* the jitter is unrelated to the signal, and *if* a
PLL is used that has a long enough smoothing time to suppress any clock
problems due to medium term and short term jitter induced by shape changes.

The problem (in principle) is that detected jitter may be signal dependent
as it stems from the changed shape of the waveform and the effect this
might have upon the instants at which the receiver decides it has seen a
transition. Also, I am not sure that all DAC/receivers have effective PLLs.

TBH though I recon USB audio would be a decent alternative to all this
unidirectional crap ;-)

I have the feeling that we have had this discussion before, elsewhere...
:-)

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