On Fri, 14 Jan 2011 09:57:44 -0500, "Arny Krueger"
wrote:

"Don Pearce" wrote in message


And of course the term jitter is used with cables -
wrongly in my view. Jitter is a random perturbation of
the data edges caused by noise events. The inaccuracies
caused by cables are systematic and identical on each
data edge. This means that they can be corrected. Either
matching the cable better or using channel estimation
(all mobile phones have this and it is dirt cheap) to
measure and cancel the inaccuracies that shift the edges
out of place.

The digital signal's edges tend to wander around because the cable is
ultimately a low pass filter and the spectral content of the digital data
passing through the cable varies as the data varies. So matching the cable
better can't be of much help.

No, this simply isn't so. Matching a cable properly results in a flat
frequency response and a flat group delay. This reshapes the signal
perfectly.

That leaves the more complex methods like channel estimation or buffering
and reclocking.


More complex, sure, but not exactly taxing these days.

d

In article , Arny
Krueger
wrote:
"Don Pearce" wrote in message


And of course the term jitter is used with cables - wrongly in my
view. Jitter is a random perturbation of the data edges caused by
noise events. The inaccuracies caused by cables are systematic and
identical on each data edge. This means that they can be corrected.
Either matching the cable better or using channel estimation (all
mobile phones have this and it is dirt cheap) to measure and cancel
the inaccuracies that shift the edges out of place.

The digital signal's edges tend to wander around because the cable is
ultimately a low pass filter and the spectral content of the digital
data passing through the cable varies as the data varies. So matching
the cable better can't be of much help.

Well a better match might improve things in principle. The snag is that
even trying to do this is questionably given the frequency range involved
as the cables don't necessarily have a frequency-independent impedance. :-)

And as you say, the real problem is the finite bandwith.

Yes. I think PM and others refer to it as 'data induced jitter' because it
depends on the details of the waveform. Generally tested using the 'J test'
waveform that toggles the LSB of an otherwise steady fs/4 waveform.

Slainte,

Jim

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

In article , Don Pearce
wrote:
On Fri, 14 Jan 2011 09:57:44 -0500, "Arny Krueger"
wrote:

"Don Pearce" wrote in message


And of course the term jitter is used with cables - wrongly in my
view. Jitter is a random perturbation of the data edges caused by
noise events. The inaccuracies caused by cables are systematic and
identical on each data edge. This means that they can be corrected.
Either matching the cable better or using channel estimation (all
mobile phones have this and it is dirt cheap) to measure and cancel
the inaccuracies that shift the edges out of place.

The digital signal's edges tend to wander around because the cable is
ultimately a low pass filter and the spectral content of the digital
data passing through the cable varies as the data varies. So matching
the cable better can't be of much help.

No, this simply isn't so. Matching a cable properly results in a flat
frequency response and a flat group delay. This reshapes the signal
perfectly.

Not if the cable loss changes with frequency. You can optimise by playing
with the matching, but not necessarily get a perfect output.

That leaves the more complex methods like channel estimation or
buffering and reclocking.


More complex, sure, but not exactly taxing these days.

It is true that the bandwidths involved aren't exactly out into the THz
region. But as Arny has said, it makes sense in practice to get the dac
(receiver) to deal with this by some mix of reclocking, bufferring, etc.
That way the results become less dependent on the choice of source and
cable... and if someone has bent or stood on the coax. :-)

Slainte,

Jim

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

Do you think Russ might be able to supply me with some decent relays for my
Denon amp. They tend to get dodgy contacts and the sound distorts untill you
thump them. I feel sure he has a solution for this problem.

Brian

--
Brian Gaff....Note, this account does not accept Bcc: email.
graphics are great, but the blind can't hear them
Email:
__________________________________________________ __________________________________________________ __________


"tony sayer" wrote in message
...

http://www.asa.org.uk/ASA-action/Adj.../Russ-Andrews-
Accessories-Ltd/TF_ADJ_49597.aspx
--
Tony Sayer

"David Looser" wrote in message
...
"Fed Up Lurker" wrote

But if it is subjectively
discernable then in this day and age it is measurable.

Gosh, something from "Fed Up Lurker" (an entirely inappropriate user name
if ever there was one) that I can agree with!

Admit it, you do find me addictive reading......

It's hard to decipher your long, rambling post, but you appear to be
claiming that there is a measurable and audible difference between
different SPDIF interconnects that isn't simply a matter of bit errors.
SPDIF merely transmits a series of numbers, and as long as those numbers
arrive at the far end unaltered then the transmission is perfect and there
*cannot* be any differences between different interconnects. There's
nothing subjective about numbers!

You're wrong yet again! The subject was toslink and coax, the SPDIF
and CD's 16 binary digits.
Jitter is a timing based error, can be a problem with any feed but with
16bit cd data, as the design flaw with the SPDIF is that it is an unclocked
interface then for that data to transfer bit-for-bit everything else has to
be right. Frequently the in and out interfaces of each device are not an
ideal match, noted by PM was that even if both output and input were
true 75ohm but the interconnect was not, that would result in both
measureable and subjective differences, and vice versa, 50-75-50ohm etc.
There are a whole heap of other factors but you wouldn't understand.


I've recently been experimenting with long distance transmission of SPDIF
over Cat5 cable using RS422 driver and receiver chips. I've been able to
send error-free SPDIF over 200m of this stuff

200m of Cat5 - what were you doing, tapping into a nieghbours ethernet
socket? What has that got to do with CD's 16bit format via the *SPDIF*?

and, as expected, there is absolutely no subjective difference whether the
source player and DAC are directly connected via TOSLINK, or the signal is
diverted via 200m of Cat5.

You haven't quite grasped that toslink is not same as Cat5.
Upload some pictures of your experiment then I'll believe you and will
be able to then tell you what you're doing wrong - I will anyway, the
digital interface of CD, DVD, DAT, DAB etc, dosen't involve Cat5.
The interface of such devices is unclocked for starters.....


David.


You're probably feeling silly now, try and have a stress free weekend.

On Fri, 14 Jan 2011 17:13:42 +0000 (GMT), Jim Lesurf
wrote:

In article , Don Pearce
wrote:
On Fri, 14 Jan 2011 09:57:44 -0500, "Arny Krueger"
wrote:

"Don Pearce" wrote in message


And of course the term jitter is used with cables - wrongly in my
view. Jitter is a random perturbation of the data edges caused by
noise events. The inaccuracies caused by cables are systematic and
identical on each data edge. This means that they can be corrected.
Either matching the cable better or using channel estimation (all
mobile phones have this and it is dirt cheap) to measure and cancel
the inaccuracies that shift the edges out of place.

The digital signal's edges tend to wander around because the cable is
ultimately a low pass filter and the spectral content of the digital
data passing through the cable varies as the data varies. So matching
the cable better can't be of much help.

No, this simply isn't so. Matching a cable properly results in a flat
frequency response and a flat group delay. This reshapes the signal
perfectly.

Not if the cable loss changes with frequency. You can optimise by playing
with the matching, but not necessarily get a perfect output.

Over the kinds of frequency we are dealing with in audio, cables are
sensibly flat regards loss. Sure there will still be errors, but of
minuscule magnitude. I don't think I've ever seen a cable that wasn't
good for hundreds of MHz if used properly.

That leaves the more complex methods like channel estimation or
buffering and reclocking.


More complex, sure, but not exactly taxing these days.

It is true that the bandwidths involved aren't exactly out into the THz
region. But as Arny has said, it makes sense in practice to get the dac
(receiver) to deal with this by some mix of reclocking, bufferring, etc.
That way the results become less dependent on the choice of source and
cable... and if someone has bent or stood on the coax. :-)

As I said also - my first point about buffering and generating a clean
clock rather than using the data edges.

d

"Fed Up Lurker" wrote in message
...

"David Looser" wrote in message
...
"Fed Up Lurker" wrote

But if it is subjectively
discernable then in this day and age it is measurable.

Gosh, something from "Fed Up Lurker" (an entirely inappropriate user name
if ever there was one) that I can agree with!

Admit it, you do find me addictive reading......

It's hard to decipher your long, rambling post, but you appear to be
claiming that there is a measurable and audible difference between
different SPDIF interconnects that isn't simply a matter of bit errors.
SPDIF merely transmits a series of numbers, and as long as those numbers
arrive at the far end unaltered then the transmission is perfect and
there
*cannot* be any differences between different interconnects. There's
nothing subjective about numbers!

You're wrong yet again! The subject was toslink and coax, the SPDIF
and CD's 16 binary digits.
Jitter is a timing based error, can be a problem with any feed but with
16bit cd data, as the design flaw with the SPDIF is that it is an
unclocked
interface then for that data to transfer bit-for-bit everything else has
to
be right.

Design flaw?? There is no "design flaw" in SPDIF. The design flaws are in
the DACs that cannot cope with jitter on the digital input. It's not rocket
science to get the clock-recovery circuitry right.

Frequently the in and out interfaces of each device are not an
ideal match, noted by PM was that even if both output and input were
true 75ohm but the interconnect was not, that would result in both
measureable and subjective differences, and vice versa, 50-75-50ohm etc.
There are a whole heap of other factors but you wouldn't understand.

I understand it all rather better than you do!

I've recently been experimenting with long distance transmission of SPDIF
over Cat5 cable using RS422 driver and receiver chips. I've been able to
send error-free SPDIF over 200m of this stuff

200m of Cat5 - what were you doing, tapping into a nieghbours ethernet
socket? What has that got to do with CD's 16bit format via the *SPDIF*?

CDs are neither here nor there. SPDIF will carry digital audio from any
source, and with any bit depth up to 24 bit.

What I was doing was establishing how far I could send SPDIF over Cat5
before I got problems. 200m was just a convenient length as I happened to
have a 200m drum of the stuff to hand and, as I discovered, I got no errors
sending over that distance. Strictly speaking this is a hybrid of SPDIF and
AES/EBU since it uses balanced 110ohm transmission which is specified for
AES/EBU but not SPDIF, but since the signal was sourced from a domestic
player the metadata is SPDIF rather than AES/EBU.

and, as expected, there is absolutely no subjective difference whether
the
source player and DAC are directly connected via TOSLINK, or the signal
is
diverted via 200m of Cat5.

You haven't quite grasped that toslink is not same as Cat5.

What a stupid thing to say! You spend so much time trying to be "clever"
that you put your foot in it time after time.

Upload some pictures of your experiment then I'll believe you and will
be able to then tell you what you're doing wrong -

I'm not doing anything wrong - that's the whole point.

I will anyway, the
digital interface of CD, DVD, DAT, DAB etc, dosen't involve Cat5.
The interface of such devices is unclocked for starters.....

Presumably by "unclocked" you mean that there isn't a separate transmission
path for the clock. Well there rarely is because of the cost. That's why
SPDIF uses bi-phase mark encoding to ensure that the receiver can recover
clock regardless of the data (assuming a competent DAC design of course).
Perhaps you are labouring under the misapprehension that Cat5 can only be
used for Ethernet (which is also "unclocked" as it happens). Cat5 is a very
useful type of cable for medium distance transmission of digital data (works
well for analogue audio as well). It's cheap, easy to install and has
excellent balance. Of course it needs balans at each end to interface to
co-ax, or media converters to interface to toslink, but it still has cost
and performance advantages over either for longish runs.

You're probably feeling silly now,

Err no, you might be though.

try and have a stress free weekend.

And you!

David.

In article , Don Pearce
wrote:
On Fri, 14 Jan 2011 17:13:42 +0000 (GMT), Jim Lesurf
wrote:

In article , Don Pearce
wrote:
On Fri, 14 Jan 2011 09:57:44 -0500, "Arny Krueger"

No, this simply isn't so. Matching a cable properly results in a flat
frequency response and a flat group delay. This reshapes the signal
perfectly.

Not if the cable loss changes with frequency. You can optimise by
playing with the matching, but not necessarily get a perfect output.

Over the kinds of frequency we are dealing with in audio, cables are
sensibly flat regards loss.

However I was picking up your unqualified statement that mathcing would
give a 'perfect output'. I agree the departure from perfection should not
normally be an *audible* problem. Indeed, if peoplw want to worry about
they should worry about the LF cable impedance departing from the nominal
value at high freqencies. :-)

Sure there will still be errors, but of minuscule magnitude. I don't
think I've ever seen a cable that wasn't good for hundreds of MHz if
used properly.

What spdif cable bandwidth is required for, say 100 ps of jitter with the
J-test? I'm curious about this as I'm wondering about transferring
192k/24bit as well as ye olde 44.1k/16bit.

As I'm currently looking at some DACs I've noticed statements that optical
spdif is limited to 48k. That isn't coax of course, but it made me wonder
where this stated limit is coming from, or even if it is true. That in turn
makes me wonder about jitter and transfer bandwidth/dispersion when people
are using higher sample rates, etc, than the now-traditional CD standard.

Slainte,

Jim

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

On Sat, 15 Jan 2011 11:41:02 +0000 (GMT), Jim Lesurf
wrote:

In article , Don Pearce
wrote:
On Fri, 14 Jan 2011 17:13:42 +0000 (GMT), Jim Lesurf
wrote:

In article , Don Pearce
wrote:
On Fri, 14 Jan 2011 09:57:44 -0500, "Arny Krueger"

No, this simply isn't so. Matching a cable properly results in a flat
frequency response and a flat group delay. This reshapes the signal
perfectly.

Not if the cable loss changes with frequency. You can optimise by
playing with the matching, but not necessarily get a perfect output.

Over the kinds of frequency we are dealing with in audio, cables are
sensibly flat regards loss.

However I was picking up your unqualified statement that mathcing would
give a 'perfect output'. I agree the departure from perfection should not
normally be an *audible* problem. Indeed, if peoplw want to worry about
they should worry about the LF cable impedance departing from the nominal
value at high freqencies. :-)

I've always found it to be the other way round. Cable impedance is
pretty stable at high frequencies - it is only at low (kHz)
frequencies that the series and parallel resistance terms start
driving the impedance upwards. And of course at those low frequencies
it doesn't matter, because the electrical length is so short.


Sure there will still be errors, but of minuscule magnitude. I don't
think I've ever seen a cable that wasn't good for hundreds of MHz if
used properly.

What spdif cable bandwidth is required for, say 100 ps of jitter with the
J-test? I'm curious about this as I'm wondering about transferring
192k/24bit as well as ye olde 44.1k/16bit.


Not sure - length is just as important as impedance, of course. It
would take a pretty horrible piece of cable to have a bandwidth low
enough to induce jitter of those kinds of performance. Of course what
matters is not the cable, but the entire system of which it is a part.
You can point the finger almost anywhere when it comes to timing
errors.

As I'm currently looking at some DACs I've noticed statements that optical
spdif is limited to 48k. That isn't coax of course, but it made me wonder
where this stated limit is coming from, or even if it is true. That in turn
makes me wonder about jitter and transfer bandwidth/dispersion when people
are using higher sample rates, etc, than the now-traditional CD standard.


I have a newly designed piece of kit testing right now, a Ka band
(30GHz) transceiver linking up and down from a satellite. The system
contains about a kilometre of assorted cables, twenty miles of fibre,
many, many filters, mixers, vaguely linear amplifiers and microwave
dishes being blown around by the bad weather. I don't know what the
bit error rate is yet, because it has only been running since just
before Christmas and there hasn't yet been an error. What I'm saying
is that I don't understand how any designer could ever have a problem
with a few feet of cable. I'm astounded.

d

"Jim Lesurf" wrote

What spdif cable bandwidth is required for, say 100 ps of jitter with the
J-test? I'm curious about this as I'm wondering about transferring
192k/24bit as well as ye olde 44.1k/16bit.

The bit depth should make no difference, as SPDIF transmits 32 bits (24 of
which are available for audio data) per sample regardless of the bit depth
of the transmitted audio. Any unused bits are simply set to zero. OTOH the
bit rate of the SPDIF link will scale with the audio sample rate.

As I'm currently looking at some DACs I've noticed statements that optical
spdif is limited to 48k.

The toslink transmitters and receivers I've bought recently claim to be good
to 13Mb/s, which should allow a 96kHz sample rate without problems. 192kHz
would be pushing it.

David.