PDA

View Full Version : Modifying squeezebox clock



Arthur Cheng
2005-02-28, 06:34
Dear hackers!

I have noticed that the slim box uses Micronas MAS3539F chip with and
clock of 18.432M. I would if an replacement of this clock will reduce
jitter and improve performance to give a truly audiophile results?

Arthur

Robin Bowes
2005-02-28, 07:40
Christian Pernegger wrote:
>> I have noticed that the slim box uses Micronas MAS3539F chip with and
>> clock of 18.432M. I would if an replacement of this clock will reduce
>> jitter and improve performance to give a truly audiophile results?
>
>
> Disclaimer: I'm not an expert ...

So it would seem :)

> but I don't quite see what replacing the clock chip would help. From
> some of the previous messages I gather that the jitter is a lot higher
> with .flac than with .mp3. I haven't decided on a DAC yet, so I can't
> test it myself but how does raw PCM / .wav fare?

jitter is independent of source format.

> The amount of jitter at the SqueezeBox's output should depend almost
> exclusively on the capacity of the buffer used. The higher the used
> bitrate, the greater the chance of a buffer underrun if your network or
> server can't deliver the data at exactly the right time. You need about
> 4 times the buffer of even a very high quality mp3 vs formats that are
> transmitted uncompressed like .flac.
> Upgrading the buffer in the Squeezebox is likely tricky, but a buffer in
> your DAC has exactly the same effect. (Beats me why not all DACs have
> one - RAM costs next to nothing) In that case the DAC will show jitter
> at the input but it won't be there at the output.

jitter is nothing to do with the buffer.

I suggest you do some googling to see what jitter actually is andwhat
causes it.

R.
--
http://robinbowes.com

Phil Karn
2005-02-28, 09:49
Arthur Cheng wrote:
> Dear hackers!
>
> I have noticed that the slim box uses Micronas MAS3539F chip with and
> clock of 18.432M. I would if an replacement of this clock will reduce
> jitter and improve performance to give a truly audiophile results?

Crystal oscillators, by their nature, have very low phase noise, but
they can have frequency errors and long term drift that, depending on
the application, sometimes has to be corrected. This is well known to
any electronic engineer.

So it seems to me that the D/A jitter caused by crystal oscillator phase
noise is going to be completely undetectable by the human ear unless,
perhaps, the oscillator is quite defective. The small errors in
frequency could accumulate over time to cause buffer overrun or underrun
on live streams, but short term jitter, as in audible frequency
modulation distortion of a signal? I *seriously* doubt it.

Have you conducted any properly controlled (e.g., ABX) tests to see if
you can really detect a given amount of jitter? Maybe you just *think*
you can...

Robin Bowes
2005-02-28, 10:34
Phil Karn wrote:
> Arthur Cheng wrote:
>
>> Dear hackers!
>>
>> I have noticed that the slim box uses Micronas MAS3539F chip with and
>> clock of 18.432M. I would if an replacement of this clock will reduce
>> jitter and improve performance to give a truly audiophile results?
>
>
> Crystal oscillators, by their nature, have very low phase noise, but
> they can have frequency errors and long term drift that, depending on
> the application, sometimes has to be corrected. This is well known to
> any electronic engineer.
>
> So it seems to me that the D/A jitter caused by crystal oscillator phase
> noise is going to be completely undetectable by the human ear unless,
> perhaps, the oscillator is quite defective. The small errors in
> frequency could accumulate over time to cause buffer overrun or underrun
> on live streams, but short term jitter, as in audible frequency
> modulation distortion of a signal? I *seriously* doubt it.

Phil,

Jitter is the single biggest factor affecting digital audio.

This link [1] explains it rather well.

[1] http://www.jitter.de/english/engc_navfr.html

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 11:50
Christian Pernegger wrote:
> Quotes from:
> http://www.digido.com/portal/pmodule_id=11/pmdmode=fullscreen/pageadder_page_id=28/

First paragraph on that page:

"October 2002
I hesitate to remove this older article from our website, as it is still
informative, but I highly recommend that those interested in the latest
word on this subject please read the chapter on jitter in my new book.
Some questions that this previous article has raised have been clarified
in our letters section, and of course are covered much better in the
book. -BK"

I have that book ("Mastering Audio - the art and the science" by Bob
Katz) and in it he does indeed make a better job of explaining jitter,
although it's not brilliant.

A better explanation is given here [1].

[1] http://www.jitter.de/english/engc_navfr.html

> "Jitter is time-base error. It is caused by varying time delays in the
> circuit paths from component to component in the signal path." and "The
> only effect of timebase distortion is in the listening; as far as it can
> be proved, it has no effect on the dubbing of tapes or any digital to
> digital transfer (as long as the jitter is low enough to permit the data
> to be read."

This definition is erm, incomplete, to say the least.

> Jitter is a timing problem, the digital data is not altered. In theory
> if we can manage to output the correct data at the exactly right time,
> there is no jitter.

Correct. However, that is precisely the problem - it is not easy to
output the correct data at exactly the right time.

>> jitter is nothing to do with the buffer.
>
>
> "Playback from a DAT recorder usually sounds better than the recording,
> because there is less jitter. Remember, a DAT machine on playback puts
> out numbers from an internal RAM buffer memory, locked to its internal
> crystal clock. A DAT machine that is recording (from its digital input)
> is locked to the source via its (relatively jittery) Phase Locked Loop.
> As the figure above illustrates, the numbers still get recorded
> correctly on tape, although their timebase was jittery while going in.
> Nevertheless, on playback, that time base error becomes irrelevant, for
> the numbers are reclocked by the DAT machine!" and
> "I repeat: jitter does not affect D-D dubs, it only affects the D to A
> converter in the listening chain"
>
> Buffers can eliminate any jitter you might have picked up up to this
> point in the signal path. But you still have to read from the buffer at
> the exact right intervals. If your power supply is not clean or the
> clock is inaccurate for some other reason you cannot do that and get
> jitter in the output. If you use the Squeezebox's DAC, the Squeezebox's
> clock is relevant, if you have a good seperate DAC only that matters as
> far as jitter is concerned.

No, buffers have no effect on jitter. Also, jitter introduced at the
source stage affects subsequent digital stages, e.g. and external DAC.
One of the reasons expensive CD transports can sound so much better than
cheaper units is that they reduce jitter by using high-quality power
supplies.

>> jitter is independent of source format.
>
>
> There have been multiple messages on this list station that output
> jitter from the 'box was a lot higher with .flac files (transferred from
> the server as PCM) than with .mp3 (decoded inside the 'box to PCM. Since
> I don't think there are two clock chips the clock chip can not account
> for the difference. Leaves the processing inside the squeezebox (mp3
> decoding vs pass-through) and the bitrates of the data passed to it.
>
> Yes, to reduce the jitter of .mp3s further you'd need to provide cleaner
> power or a better clock chip. But I was looking for ways to bring the
> jitter level of uncompressed down to .mp3 levels.
>
> According to my tests the 'box has 256kb of buffer (A 128kbit/s mp3
> plays for a tick over 15 seconds if I pull the network connector in
> mid-play.) The same buffer can not even hold one and a half seconds of
> uncompressed audio. A CPU spike on the server, a network usage spike or
> a bit of load can easily cause interruptions of this magnitude. Network
> speed is never constant even under the best conditions.

You are confusing "jitter" with playback interruptions. Perhaps this
diagram will help:

slimserver---buffer---decoding---DAC---analogue out
|
+-------Digital out

Jitter is introduced at the decoding stage and is caused by the timing
errors in the bitstream.

> Even after having googled around a bit I stand by my assumption that the
> difference in output jitter is due to the different bitrates and that a
> bigger buffer would help towards closing the gap.
>
> Constructive and specific criticism welcome :)

You don't understand what jitter is.

Read the link given above [1].

R.
--
http://robinbowes.com

Phil Karn
2005-02-28, 12:52
Robin Bowes wrote:

> Jitter is the single biggest factor affecting digital audio.
>
> This link [1] explains it rather well.
>
> [1] http://www.jitter.de/english/engc_navfr.html

I'm very skeptical. And that's putting it mildly.

First of all, that page is maintained by a company that sells a rather
expensive ($500) solution to the "problem" of jitter. That alone doesn't
make it bogus, but it sure does call for close scrutiny.

Sure, jitter exists. If it's extreme enough, you'll hear it. Jitter
frequency modulates the audio components with a noise process, and this
creates sidebands around each spectral component that effectively smear
that component in frequency.

And yes, with a scope you can see jitter on real signals if you expand
the horizontal sweep to an extreme degree (5ns/div), as shown on that
page. And yes, you can even clean up that jitter with an expensive box.

But can you really hear the difference?

I see lots of anecdotal testimonials from satisfied customers, but where
are the double-blind tests? I see no proper scientific evidence
whatsoever that the jitter seen on real signals is even audible *at
all*. Audiophiles have repeatedly proven themselves capable of "hearing"
all sorts of incredible improvements from expensive snake-oil
accessories that disappear completely when properly controlled listening
tests are done.

James Randi, the well-known magician and skeptic, has had a lot to say
about dubious audiophile accessories in his recent commentaries. Check
him out at www.randi.org.

--Phil

T
2005-02-28, 13:37
>> This link [1] explains it rather well.
>>
>> [1] http://www.jitter.de/english/engc_navfr.html
>
> I'm very skeptical. And that's putting it mildly.

Especially since they use two different vertical settings for the 'good' and
'bad' signals!

Tom

Robin Bowes
2005-02-28, 13:49
Phil Karn wrote:
> Robin Bowes wrote:
>
>> Jitter is the single biggest factor affecting digital audio.
>>
>> This link [1] explains it rather well.
>>
>> [1] http://www.jitter.de/english/engc_navfr.html
>
>
> I'm very skeptical. And that's putting it mildly.
>
> First of all, that page is maintained by a company that sells a rather
> expensive ($500) solution to the "problem" of jitter. That alone doesn't
> make it bogus, but it sure does call for close scrutiny.
>
> Sure, jitter exists. If it's extreme enough, you'll hear it. Jitter
> frequency modulates the audio components with a noise process, and this
> creates sidebands around each spectral component that effectively smear
> that component in frequency.
>
> And yes, with a scope you can see jitter on real signals if you expand
> the horizontal sweep to an extreme degree (5ns/div), as shown on that
> page. And yes, you can even clean up that jitter with an expensive box.
>
> But can you really hear the difference?
>
> I see lots of anecdotal testimonials from satisfied customers, but where
> are the double-blind tests? I see no proper scientific evidence
> whatsoever that the jitter seen on real signals is even audible *at
> all*. Audiophiles have repeatedly proven themselves capable of "hearing"
> all sorts of incredible improvements from expensive snake-oil
> accessories that disappear completely when properly controlled listening
> tests are done.
>
> James Randi, the well-known magician and skeptic, has had a lot to say
> about dubious audiophile accessories in his recent commentaries. Check
> him out at www.randi.org.

Phil,

Jitter does not fall into the same category as other "snake-oil"
audiophile accessories - it is a well-defined and relatively
well-understood engineering phenomenon. Suggesting otherwise is rather
disingenous. Believe it, it exists, and can be easily demonstrated.

One thing I will say is that the effects of jitter can be very subtle,
only noticeable in relatively high-end equipment, i.e. the inadequacies
of run-of-the-mill hifi equipment will mask any jitter-induced problems.

Most folk on this list will not notice jitter at all (and are probably
sick of us banging on about it! Perhaps it's time for an audiophile list
:) )

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 13:53
T wrote:
>>> This link [1] explains it rather well.
>>>
>>> [1] http://www.jitter.de/english/engc_navfr.html
>>
>>
>> I'm very skeptical. And that's putting it mildly.
>
>
> Especially since they use two different vertical settings for the 'good'
> and 'bad' signals!

To which figures are you referring? I can't see any such discrepancy,
but then I'm maybe not looking at the right images.

R.
--
http://robinbowes.com

T
2005-02-28, 14:09
> To which figures are you referring?

http://www.jitter.de/english/engc_navfr.html and then click on 'how does
jitter sound'

Tom

Jack Coates
2005-02-28, 14:15
Robin Bowes wrote:
....>
> Jitter does not fall into the same category as other "snake-oil"
> audiophile accessories - it is a well-defined and relatively
> well-understood engineering phenomenon. Suggesting otherwise is rather
> disingenous. Believe it, it exists, and can be easily demonstrated.
>
> One thing I will say is that the effects of jitter can be very subtle,
> only noticeable in relatively high-end equipment, i.e. the inadequacies
> of run-of-the-mill hifi equipment will mask any jitter-induced problems.
>
> Most folk on this list will not notice jitter at all (and are probably
> sick of us banging on about it! Perhaps it's time for an audiophile list
> :) )
>
> R.

and how :) I currently auto-delete anything with "audiophile" or "forum"
in the subject.

--
Jack at Monkeynoodle dot Org: It's a Scientific Venture...
Riding the Emergency Third Rail Power Trip since 1996!

Robin Bowes
2005-02-28, 14:18
T wrote:
>>> This link [1] explains it rather well.
>>>
>>> [1] http://www.jitter.de/english/engc_navfr.html
>>
>>
>> I'm very skeptical. And that's putting it mildly.
>
>
> Especially since they use two different vertical settings for the 'good'
> and 'bad' signals!

No, the vertical settings appear to be the same - this suggests that
signal levels are different, i.e. the output from their gadget is higher.

R.
--
http://robinbowes.com

T
2005-02-28, 14:37
> No, the vertical settings appear to be the same - this suggests that
> signal levels are different, i.e. the output from their gadget is higher.

Probably, but it no longer a fair test to show rise time and jitter if the
amplitude/time ratio is not constant. It is irrelevant if it is due to
scope settings or signal amplification.

Tom

Robin Bowes
2005-02-28, 14:45
Christian Pernegger wrote:

[healthy skepticism snipped]

> Either the bit stream is tranmitted intact or it is not. There is no
> in-between. Sure, the waveform is distorted but there would have to be
> in insane lot of this "line induced jitter" to actually flip bits.

You're missing the point completely. It's not as simple as "the
bitstream is transmitted intect or it is not". At the implementation
level, this "bitstream" is actually a sequence of precisely-timed
voltage transitions. This stream is decoded by the DAC to produce the
analogue waveform. If the timing of those transitions is slightly out
then the analogue waveform is distorted.

> Everything else can be fixed by reclocking and hinges on the quality of
> the reclocking.

Ah, but who said anything about re-clocking? Not all DACs do
re-clocking. Those that do it well are not cheap. For example:
http://www.av123.com/products_product.php?section=processors&product=1.1

> The rest of the article is about recording, which is, espacially with
> multiple sources, a whole different thing.
>
> Last not least, from that very site:
>
> QUOTE
> Dual stage Clock recovery can also be combined with a data buffer memory.
>
> The first PLL writes to the buffer, and the second PLL reads from the
> buffer and tries to keep it always half filled.
>
> If we use a FIFO (first in first out) memory as a data buffer, we have
> more time to react to clock variations of the input signal and thus are
> able to attenuate lower jitter frequencies. The larger the buffer, the
> lower the jitter frequency that can be attenuated.
>
> The disadvantage of the FIFO memory solution is (beside its higher
> price) that the output data will be delayed (the buffer has to be half
> filled before data is output).
> UNQUOTE
>
> Higher price: What higher price? $150 would buffer you a CD. Even SRAM
> is not prohibitively expensive. It's true that buffers delay the output
> but they shift it CONSTANTLY meaning you don't want to use this
> technique for recording but for playback it's just fine.
>
> Show me scientific proof that using the above quoted jitter reduction
> method directly before the DAC (on the same circuit board) the jitter at
> the DACs output is influenced by the jitter at the input. Till then I
> suggest reading up on "placebo" :)

I'm not sure what point you're making here. You seem to be saying that
you agree that jitter exists as a problem and are presenting several
mechanisms to reduce it.

I never said that jitter is *always* a problem, merely that in
low-budget equipment it often is. Consumer audio equipment is designed
to a budget - there's no other reason that all CD players don't produce
audiophile quality output.

You still seem to be confused.

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 14:51
Jack Coates wrote:
> Robin Bowes wrote:
>> Most folk on this list will not notice jitter at all (and are probably
>> sick of us banging on about it! Perhaps it's time for an audiophile
>> list :) )
>>
>> R.
>
>
> and how :) I currently auto-delete anything with "audiophile" or "forum"
> in the subject.

I was thinking of you as I typed, Jack!

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 14:55
T wrote:
>> No, the vertical settings appear to be the same - this suggests that
>> signal levels are different, i.e. the output from their gadget is higher.
>
>
> Probably, but it no longer a fair test to show rise time and jitter if
> the amplitude/time ratio is not constant. It is irrelevant if it is due
> to scope settings or signal amplification.

No, the sweep time appears to be the same (5ns/division - the figure in
the top left). You would see the same effect on the trace if jitter was
present in the second signal.

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 16:06
Christian Pernegger wrote:
>> You're missing the point completely. It's not as simple as "the
>> bitstream is transmitted intect or it is not". At the implementation
>> level, this "bitstream" is actually a sequence of precisely-timed
>> voltage transitions. This stream is decoded by the DAC to produce the
>> analogue waveform. If the timing of those transitions is slightly out
>> then the analogue waveform is distorted.
>
>
> Only if you don't buffer. :) For any sane transmission protocol it is
> just as simple as bit-identical or not.

Ah, but the AESEBU/SPDIF Digital Audio Interface is *not* a sane
transmission protocol.

> How well this data is
> transformed into in anologue waveform and what distortions the analog
> signal might suffter on its way to the speakers is another chapter.

Almost. The general concensus is that a DAC supplied with a jittery
input stream will not sound good.

>>> Everything else can be fixed by reclocking and hinges on the quality
>>> of the reclocking.
>
>
>> Ah, but who said anything about re-clocking?
>
>
> I did. IIRC in every post from the very start. You probably just read
> my, admittedly stupid, bitrate --> jitter conclusion and discarded the
> rest of my posts.

Check again. You may have been *thinking* reclocking, but you didn't
mention it.

>> Not all DACs do re-clocking. Those that do it well are not cheap.
>
>
> I remember having written something along the lines of "why don't all
> DACs do this, when RAM is so cheap?"

That relates to buffering, which is *not* what we're talking about; at
least, it's not what *I'm* talking about.

>> http://www.av123.com/products_product.php?section=processors&product=1.1
>
>
> $1000 for a DAC with a bit of RAM? I guess they can get away with it...
> makes me wonder how something like the M-Audio 66 stacks up, jitter-wise :)

Actually, The M-Audio SuperDAc is particularly susceptible to jitter.
Check the archive of M-Audio_SuperDAC_2496 (AT) yahoogroups (DOT) com for more details.

>> You seem to be saying that you agree that jitter exists as a problem
>
>
> Sure
>
>> and are presenting several mechanisms to reduce it.
>
>
> Spot on again. I think this whole thread was about reduction of jitter,
> seeing as it started with the suggestion of replacing the clock chip.

It was. But some folk seemed skeptical about whether or not jitter was a
real problem.

> Personally I find buying a DAC that properly buffers and reclocks
> preferable to a clock chip upgrade.

That's one solution, but not the only one. A clock chip upgrade (29 +
12v PSU) might be more cost-effective.

>> I never said that jitter is *always* a problem, merely that in
>> low-budget equipment it often is.
>> Consumer audio equipment is designed to a budget
>
>
> K, maybe I'm really confused... I think we were talking about combining
> a Squeezebox (high priced consumer item) an external DAC (hardly a
> consumer item) and perhaps other upgrades (clock) to get something
> audiophile out of the Squeezebox.

Not really, we're discussing how to get the best out of the Squeezebox.
My agenda also covers "bang-for-buck" ie. "champagne" quality at
"vinegar" prices, to butcher a metaphor.

>
> A consumer just lives with the analogue out. It's not _that_ bad really.

It isn't bad at all. Before I modded my preamp, neither myself or
another contributor to this list could hear any difference between the
SB analogue outs and either my Art DI/O or his Arcam Delta DAC or his
Perpetual Technologies P3-A with P1-A correction engine.

On his higher-end system (Naim) there were clear differences. I can also
hear a difference now I've modded my amp.

> For the audiophile enthusiast, what's $1000 for a DAC. If it's done
> right it's the only thing you'll need to get great sound quality out of
> almost any kind of digital input.

Erm, audiophiles aren't always idiots with bottomless pits of money to
throw at snake-oil solutions. Some of us want to get "the sound" with
minimum possible outlay - I have four kids who come first!

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 16:39
Julian Alden-Salter wrote:
> Having done a number of comparisons between various cd transports (dpa,
> arcam) and the squeezebox into my dac using the same digital interconnect
> I'd be interested to know peoples thoughts on what is causing the clear
> differences heard if not jitter. Power supplies have been mentioned and to
> an extent I would agree that they make a difference. Another theory I've
> heard is that somehow interference / noise is being transferred from some
> transports to the dac and thus causing degradation to the sound. Anyone got
> any alternatives?

What difference do you hear?

> The thing that doesn't really make sense is that flacs (pcm / wavs) have a
> worse digital output than mp3's.

Erm, mp3s are lossy so they will inevitably sound "worse". How are you
comparing?

> I'm wondering if this is an internal rf
> noise issue as something will have to be working harder to process more data
> than with mp3's - or perhaps not. The thing I like about the tent xo3 is
> that it doesn't just feed a high quality clock to the decoder but it also
> corrects the spdif output after decoding. There is another board by a
> company called trichord which does a similar thing but it's more expensive.
> The biggest problem that I've got at the moment is how to fit the extra
> gubbins inside the squeezebox. I guess I'd have to re-box it but finding
> something that's a suitable size is difficult.

I've got to say, the Tent clock chip upgrade looks very appealing. I'm
planning an upgrade to my Art DI/O (4 x ALWSR PSUs: +/-15V, +5V digital
and +5V analogue) and I might slip a clock chip upgrade in too.

Let me know how you get on.

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 16:43
Christian Pernegger wrote:
>> Check again. You may have been *thinking* reclocking, but you didn't
>> mention it.
>> [...]
>> That relates to buffering, which is *not* what we're talking about; at
>> least, it's not what *I'm*
>> talking about.
>
>
> Once you use a buffer / FIFO you'll want to read what was written to it
> and you'll want to do that in sync with a clock signal - that's the
> whole point of buffering in the first place. Did you think I wanted to
> poll the poor buffer at arbitrary intervals and then pass on whatever
> was in there as fast as possible. (I almost want to know how that sounds
> now :))

My point is that buffering is another solution to the jitter problem. If
the digital source/DAC you're working with doesn't have a buffer with an
accurate clock then you're stuck with the accuracy of the clock
information encoded in the digital stream.

>
>> That's one solution, but not the only one. A clock chip upgrade (29 +
>> 12v PSU) might be more
>> cost-effective.
>
>
> Possibly. Not for me once you factor in I'll probably butcher three
> squeezeboxes in the process.
>
>> Erm, audiophiles aren't always idiots with bottomless pits of money to
>> throw at snake-oil solutions.
>
>
> Ah, they are not? I stand corrected and apologize for having you pegged
> as one after the "de-magnetize your CDs" site.

D'you know, I think we're possibly in raging agreement here, once we
start talking about the same thing!
--
http://robinbowes.com

Phil Karn
2005-02-28, 16:48
T wrote:

> Especially since they use two different vertical settings for the 'good'
> and 'bad' signals!

That's not the worst part. I can read a scope display for myself and see
that the settings were changed.

The big flaw with this particular measurement (other than the horizontal
sweep has been blown up to greatly exaggerate the "problem") is that it
shows the jitter at the end of an S/P-DIF link before it has been
processed by the receiver chip. The receiver tracks and reconstructs the
incoming clock with a phase locked loop that substantially reduces the
jitter (noise).

Reducing noise on narrowband signals like clocks and carriers is exactly
what phase lock loops *do*; they behave like tunable, very narrowband,
high-Q filters that automatically track the signal's frequency.

Sure, it's possible to be sloppy in the design of a PLL. But if you are,
chances are it wouldn't work at all. And if it does work, it probably
works well.

Again, before any of this is even relevant, one must conduct the
properly controlled studies to determine whether clock jitter is even a
real, audible problem. I see no evidence that it is.

The ironic thing about this discussion is that the Squeezebox is not s
S/P-DIF receiver that has to track an external clock. The Squeezebox
requests TCP/IP packets containing audio data, and it plays them at its
own pace using its own internal crystal oscillator. While I do not have
a schematic of a Squeezebox, I think it quite likely that this same (or
another) crystal oscillator also clocks the DACs providing analog audio
to the jacks. So there's no S/P-DIF link and no PLL reconstructing clock
anywhere in the Squeezebox's audio path. But if you connect an external
DAC, perhaps with one of those $500 jitter removers, now you have
introduced one or even two "flawed" S/P-DIF links and clock recovery
circuits, each adding their own jitter!

So it seems to me that if you're really worried about jitter, just use
the analog outputs on the Squeezebox and don't even connect anything to
the S/P-DIF outputs!

--Phil

Phil Karn
2005-02-28, 17:01
Robin Bowes wrote:

> Jitter does not fall into the same category as other "snake-oil"
> audiophile accessories - it is a well-defined and relatively
> well-understood engineering phenomenon. Suggesting otherwise is rather
> disingenous. Believe it, it exists, and can be easily demonstrated.

I didn't say that jitter doesn't exist. I could demonstrate what it
sounds like in large amounts by writing some software to intentionally
jitter PCM data. And I can certainly see some jitter on real signals
with a scope that has its horizontal sweep rate turned all the way up.

But that's not the question here. The question is whether the level of
jitter found in ordinary digital audio equipment is so large that it can
*actually be detected by the human ear*. I've seen no proper scientific
evidence that this is so; just a lot of laudatory anecdotal testimonials
about a high priced gizmo that will solve this "problem". Given that
similar glowing testimonials have been written about everything from
magic speaker wire to to small passive devices you put near your
speakers to rings that you place on your CDs, please forgive me if I
don't find such testimonials very convincing.

> One thing I will say is that the effects of jitter can be very subtle,
> only noticeable in relatively high-end equipment, i.e. the inadequacies
> of run-of-the-mill hifi equipment will mask any jitter-induced problems.

Ah, another one of Langmuir's classic signs of pathological science: the
"effect" in question is extremely small and always at the limits of
detection no matter how many experiments are run. See
http://www.cs.princeton.edu/~ken/Langmuir/langmuir.htm

--Phil

Robin Bowes
2005-02-28, 17:21
Phil Karn wrote:
> T wrote:
>
>> Especially since they use two different vertical settings for the
>> 'good' and 'bad' signals!
>
>
> That's not the worst part. I can read a scope display for myself and see
> that the settings were changed.
>
> The big flaw with this particular measurement (other than the horizontal
> sweep has been blown up to greatly exaggerate the "problem")

What makes you think that?

> is that it
> shows the jitter at the end of an S/P-DIF link before it has been
> processed by the receiver chip.The receiver tracks and reconstructs the
> incoming clock with a phase locked loop that substantially reduces the
> jitter (noise).
>
> Reducing noise on narrowband signals like clocks and carriers is exactly
> what phase lock loops *do*; they behave like tunable, very narrowband,
> high-Q filters that automatically track the signal's frequency.

Yes, but the AESSBU/SPDIF standard is such that even the best PLL can't
do this perfectly.

>
> Sure, it's possible to be sloppy in the design of a PLL. But if you are,
> chances are it wouldn't work at all. And if it does work, it probably
> works well.

....possible...chances are...probably...

>
> Again, before any of this is even relevant, one must conduct the
> properly controlled studies to determine whether clock jitter is even a
> real, audible problem. I see no evidence that it is.

Triode posted a link to a paper published by the AES that discussed the
problem of jitter concludes:

"It can be shown that, compared to low oversampling multi-bit designs,
pulse density modulation converters are much more sensitive to jitter
when producing low frequency audio signals. This phenomena may explain
certain subjective characteristics of PDM DACs which cannot otherwise be
rationalised. A simple model of jitter error audibility has shown that
white jitter noise of up to 180 ps can be tolerated in a DAC, but that
even lower levels of sinusoidal jitter may be audible."

> The ironic thing about this discussion is that the Squeezebox is not s
> S/P-DIF receiver that has to track an external clock. The Squeezebox
> requests TCP/IP packets containing audio data, and it plays them at its
> own pace using its own internal crystal oscillator. While I do not have
> a schematic of a Squeezebox, I think it quite likely that this same (or
> another) crystal oscillator also clocks the DACs providing analog audio
> to the jacks. So there's no S/P-DIF link and no PLL reconstructing clock
> anywhere in the Squeezebox's audio path. But if you connect an external
> DAC, perhaps with one of those $500 jitter removers, now you have
> introduced one or even two "flawed" S/P-DIF links and clock recovery
> circuits, each adding their own jitter!

The discussion was originally about the possibility for improving the
performance of the SB by adding a more accurate clock/crystal
oscillator. As you rightly say, with no external DAC there is no SPDIF
link but there is still the possiblity that the internal clock can be
improved upon.

> So it seems to me that if you're really worried about jitter, just use
> the analog outputs on the Squeezebox and don't even connect anything to
> the S/P-DIF outputs!

That is of course not feasible if you want to use an external DAC, in
which case you need to minimise jitter, which brings back to where we
started.

R.
--
http://robinbowes.com

Robin Bowes
2005-02-28, 17:31
Phil Karn wrote:
> I didn't say that jitter doesn't exist. I could demonstrate what it
> sounds like in large amounts by writing some software to intentionally
> jitter PCM data. And I can certainly see some jitter on real signals
> with a scope that has its horizontal sweep rate turned all the way up.
>
> But that's not the question here. The question is whether the level of
> jitter found in ordinary digital audio equipment is so large that it can
> *actually be detected by the human ear*. I've seen no proper scientific
> evidence that this is so; just a lot of laudatory anecdotal testimonials
> about a high priced gizmo that will solve this "problem". Given that
> similar glowing testimonials have been written about everything from
> magic speaker wire to to small passive devices you put near your
> speakers to rings that you place on your CDs, please forgive me if I
> don't find such testimonials very convincing.

I too am skeptical of most such claims. jitter is not one of them.

>> One thing I will say is that the effects of jitter can be very subtle,
>> only noticeable in relatively high-end equipment, i.e. the
>> inadequacies of run-of-the-mill hifi equipment will mask any
>> jitter-induced problems.
>
>
> Ah, another one of Langmuir's classic signs of pathological science: the
> "effect" in question is extremely small and always at the limits of
> detection no matter how many experiments are run. See
> http://www.cs.princeton.edu/~ken/Langmuir/langmuir.htm

That's not what I said.

I merely pointed out (explicitly, in fact, in a part of my post you
chose not to include) that most folk won't ever be bothered by the
effects of jitter because it is masked by other flaws in their systems.

The effects of jitter are clearly audible, and reproducible, on any
sufficiently accurate, high-end audio system. This is not pathological
science.

R.
--
http://robinbowes.com

Phil Karn
2005-02-28, 20:52
Robin Bowes wrote:

>
> The effects of jitter are clearly audible, and reproducible, on any
> sufficiently accurate, high-end audio system. This is not pathological
> science.

Okay, so where are the properly controlled experiments that prove it?

Robin Bowes
2005-03-01, 02:53
Julian Alden-Salter wrote:
> Robin,
>
>
>>What difference do you hear?
>
>
> Without launching off into audiophile rambling about 'inner detail' and
> 'timbre'. The dpa just sounded more realistic, detail was easier to hear and
> transients (i.e. drum strikes) had more impact whilst the squeezebox sounded
> muffled in comparison.

That sounds like classic symptoms of jitter - smearing of the transients.

> I'd have no qualms about trying to discern the
> differences in a double blind test and I'd be pretty damn confident that I'd
> be able to get a statistically significant result telling the two apart. To
> give myself a bit of leeway with hyperbolae - the difference was like night
> and day. The fact that 4 of us were in total agreement (even though the
> tests were sighted and we all have wildly different tastes in music and
> systems at home) should give some clue as to the magnitude of the
> difference.

Have you got anyway to measure jitter? If so, you may be in a position
to perform an experiment to keep Phil happy :)

Something like this:

1. Measure jitter from Squeezebox
2. Measure jitter from the dpa
3. Perform double-blind testing between the two sources

I'd be very interested to hear about the results of such an experiment.

>>Erm, mp3s are lossy so they will inevitably sound "worse". How are you
>>comparing?
>
>
> I'm not actually talking about sound quality here - flac/pcm still has the
> legs on mp3 (just at 320 but it's a tough call at that bit rate). As I said
> before my dac has low and high quality locks - one for high jitter signals
> which isn't very picky and can lock onto pretty much any standard digital
> signal you throw at it. There is also a higher quality lock which will only
> 'XLOCK' if the signal received is low enough jitter / high enough quality.
> Mp3's and the transports tested all 'XLOCK'ed. Flac / pcm's do not. This is
> 100% verifiable and consistent and not subject to any subjectivity - I can
> post pics somewhere if you like.

So, you're saying that the SB digital out is different for flac vs. mp3,
and that your dac will only XLOCK onto mp3s (and the output from your
other transports)? That is certainly interesting information. Can you
measure the jitter in the output when playing mp3s vs playing flac/pcm?

>>I've got to say, the Tent clock chip upgrade looks very appealing. I'm
>>planning an upgrade to my Art DI/O (4 x ALWSR PSUs: +/-15V, +5V digital
>>and +5V analogue) and I might slip a clock chip upgrade in too.
>
>
>>Let me know how you get on.
>
>
> I'm going to be borrowing a prototype alwsr psu specifically designed for
> the squeezebox in the near future. I'm not sure how big a difference it will
> make as I feel a psu will mostly benefit the analogue output stage. I tried
> a monarchy dip reclocking device but this made little or no difference so it
> looks like the tent clock will be my next thing to try but the casing issue
> makes things difficult.

Yeah, you need to pull the guts out of the SB and mount it in another
case if you're doing this level of modification.

I haven't decided exactly what I'll be doing, but one option is to build
the SB and the digital stage of my Art DI/O into a new chassis, along
with several ALWSR (I'll need five, or six if I add a Tent clock!)

That's on the back-burner right now until after I've moved house/changed
jobs. I'm going to work for Farnell so I'll great access to components :)

R.
--
http://robinbowes.com

Phil Karn
2005-03-02, 15:53
>> The big flaw with this particular measurement (other than the
>> horizontal sweep has been blown up to greatly exaggerate the "problem")
>
>
> What makes you think that?

Because I can read a scope legend. The scope has been blown up in the
horizontal direction to exaggerate the "jitter" You're seeing only a
tiny fraction of a bit time.

Not only that, but it's showing us the raw signal jitter, before the bit
clock has been reconstructed by the receiver's PLL and divided down to
the sample rate, which decreases the jitter accordingly. And
furthermore, there's no indication of the exposure time, so we don't
know anything about the frequency spectrum of the jitter. That's
important too.

> Yes, but the AESSBU/SPDIF standard is such that even the best PLL can't
> do this perfectly.

Not true. The jitter is clearly the result of constraining the bandwidth
of the cable so that intersymbol interference is introduced. With a good
cable, there wouldn't be any intersymbol interference.

And even if what you said is true, so what? It doesn't *have* to do a
perfect job. Nothing can. It only has to do it well enough so that you
can't hear a degradation.

> Triode posted a link to a paper published by the AES that discussed the
> problem of jitter concludes:
>
> "It can be shown that, compared to low oversampling multi-bit designs,
> pulse density modulation converters are much more sensitive to jitter
> when producing low frequency audio signals. This phenomena may explain
> certain subjective characteristics of PDM DACs which cannot otherwise be
> rationalised. A simple model of jitter error audibility has shown that
> white jitter noise of up to 180 ps can be tolerated in a DAC, but that
> even lower levels of sinusoidal jitter may be audible."

I'm not up on *all* the various DAC topologies, but I still can't
understand why this should be such a problem. If anything, I'd expect a
given amount of jitter to have *less* effect on a low frequency signal
than on a high frequency one. That's because jitter is really just
undesired phase modulation by a noise process, and a variation by a
given amount of time represents a much smaller number of degrees in a
cycle of a low frequency waveform than a high frequency waveform. This
is analogous to the fact that phase modulation and frequency modulation
are closely related, but with a 6 dB/octave pre-emphasis or de-emphasis
because of the fact that frequency is the derivative of phase, and phase
is the integral of frequency.

Note that the paper analyzed the effect of jitter on full amplitude 20
KHz sine waves. That's about as worst-case as you can possibly get. Can
you still even hear 20 KHz? I know I can't.

> The discussion was originally about the possibility for improving the
> performance of the SB by adding a more accurate clock/crystal
> oscillator. As you rightly say, with no external DAC there is no SPDIF
> link but there is still the possiblity that the internal clock can be
> improved upon.

Yet according to that same AES paper, internal clock phase noise is
utterly negligible compared to the jitter source he was investigating,
namely data-dependent jitter introduced by intersymbol interference.
This is exactly what I suspected given what I know about crystal
oscillators -- they have the lowest phase noise of any kind of RF
oscillator, and suffer only from frequency inaccuracy and slow drift.

>
>> So it seems to me that if you're really worried about jitter, just use
>> the analog outputs on the Squeezebox and don't even connect anything
>> to the S/P-DIF outputs!
>
>
> That is of course not feasible if you want to use an external DAC, in
> which case you need to minimise jitter, which brings back to where we
> started.

One could ask why you even need an external DAC in the first place, as
the Squeezebox already provides exactly that function. At least it gets
the DAC out of my computer where 100A ground currents circulate!

I know some people think that something costing only $200 can't
*possibly* be any good, and that at least $1K of fancy, gold-plated
high-end accessories is needed to even be worthy of their consideration.
God bless em', they do keep the audiophile industry in business.

--Phil

Robin Bowes
2005-03-02, 17:12
Phil Karn wrote:
>
>>> The big flaw with this particular measurement (other than the
>>> horizontal sweep has been blown up to greatly exaggerate the "problem")
>>
>> What makes you think that?
>
>
> Because I can read a scope legend. The scope has been blown up in the
> horizontal direction to exaggerate the "jitter" You're seeing only a
> tiny fraction of a bit time.

What information on that image tells you that? It seems to me you're not
understanding what you're seeing.

> Not only that, but it's showing us the raw signal jitter, before the bit
> clock has been reconstructed by the receiver's PLL and divided down to
> the sample rate, which decreases the jitter accordingly. And
> furthermore, there's no indication of the exposure time, so we don't
> know anything about the frequency spectrum of the jitter. That's
> important too.

It's the same for both cases. Level playing field, anybody?

>
>> can't do this perfectly.
>
>
> Not true. The jitter is clearly the result of constraining the bandwidth
> of the cable so that intersymbol interference is introduced. With a good
> cable, there wouldn't be any intersymbol interference.

But you don't believe in all this audiophile "mumbo jumbo" about good
cables and bad cables, do you?

> And even if what you said is true, so what? It doesn't *have* to do a
> perfect job. Nothing can. It only has to do it well enough so that you
> can't hear a degradation.

You're general POV seems to be that if *you* can't hear a difference
then no-one else should either.

>> Triode posted a link to a paper published by the AES that discussed
>> the problem of jitter concludes:
>>
>> "It can be shown that, compared to low oversampling multi-bit designs,
>> pulse density modulation converters are much more sensitive to jitter
>> when producing low frequency audio signals. This phenomena may explain
>> certain subjective characteristics of PDM DACs which cannot otherwise
>> be rationalised. A simple model of jitter error audibility has shown that
>> white jitter noise of up to 180 ps can be tolerated in a DAC, but that
>> even lower levels of sinusoidal jitter may be audible."
>
>
> I'm not up on *all* the various DAC topologies, but I still can't
> understand why this should be such a problem. If anything, I'd expect a
> given amount of jitter to have *less* effect on a low frequency signal
> than on a high frequency one. That's because jitter is really just
> undesired phase modulation by a noise process, and a variation by a
> given amount of time represents a much smaller number of degrees in a
> cycle of a low frequency waveform than a high frequency waveform. This
> is analogous to the fact that phase modulation and frequency modulation
> are closely related, but with a 6 dB/octave pre-emphasis or de-emphasis
> because of the fact that frequency is the derivative of phase, and phase
> is the integral of frequency.

Have you ever heard the effect of phase modulation on an audio signal?
The sort of improvements audio enthusiasts wax lyrical about can often
be attributed to phase. Soundstage, depth, clarity, all that sort of
stuff. Even a small phase error can radically smear the sound.

> Note that the paper analyzed the effect of jitter on full amplitude 20
> KHz sine waves. That's about as worst-case as you can possibly get. Can
> you still even hear 20 KHz? I know I can't.

I don't hear discrete sounds or tones, no, but I do hear relationships
between sounds and positional information that is encoded in the HF
band. If you lose that you lose clarity and detail in the sound.

>> The discussion was originally about the possibility for improving the
>> performance of the SB by adding a more accurate clock/crystal
>> oscillator. As you rightly say, with no external DAC there is no SPDIF
>> link but there is still the possiblity that the internal clock can be
>> improved upon.
>
>
> Yet according to that same AES paper, internal clock phase noise is
> utterly negligible compared to the jitter source he was investigating,
> namely data-dependent jitter introduced by intersymbol interference.
> This is exactly what I suspected given what I know about crystal
> oscillators -- they have the lowest phase noise of any kind of RF
> oscillator, and suffer only from frequency inaccuracy and slow drift.

And what happens if the crystal is inaccurate?

>>> So it seems to me that if you're really worried about jitter, just
>>> use the analog outputs on the Squeezebox and don't even connect
>>> anything to the S/P-DIF outputs!
>>
>> That is of course not feasible if you want to use an external DAC, in
>> which case you need to minimise jitter, which brings back to where we
>> started.
>
>
> One could ask why you even need an external DAC in the first place, as
> the Squeezebox already provides exactly that function. At least it gets
> the DAC out of my computer where 100A ground currents circulate!

Because I can hear that my DAC sounds better than the DAC in the Squeezebox.

> I know some people think that something costing only $200 can't
> *possibly* be any good, and that at least $1K of fancy, gold-plated
> high-end accessories is needed to even be worthy of their consideration.
> God bless em', they do keep the audiophile industry in business.

It's not about spending money for the sake of it. I want to get the best
possible sound with the least possible outlay. I don't go chasing "snake
oil" gadgets. I have a Squeezebox, which is stock (190). I have an
external DAC, which cost me 70 plus 10 for a UK power supply (I bought
it from the US); I've tweaked my (inexpensive) amplifer (cost new 120,
15 years ago. 20 on tweaks) and added a power amplifier for biamping
(90). The sound I get from my setup now is at a level comparable to
systems costing well into four figures for an outlay of only 190

You seem to be very closed to the possibility that others can hear
things that you obviously can't. You're not wrong if you can't hear it,
but you are wrong to insist that others can't either.

Please don't dismiss as impossible everything you can't explain or don't
understand in engineering terms.

R.
--
http://robinbowes.com

Phil Karn
2005-03-02, 18:23
Robin Bowes wrote:

> What information on that image tells you that? It seems to me you're not
> understanding what you're seeing.

The legend says the time trace is 5 ns/div. A pair of measuring lines
implies that the jitter on the uncorrected is somewhat less than that,
about 3 ns.

As I recall (I can't find good, recent references), each subframe in
S/P-DIF carries a (usually) 16-bit PCM sample in a 32-bit subframe, so
the clock rate of the composite stereo signal would be 64 times the 44.1
KHz CD sampling rate, or 2.8224 MHz. At this rate, a bit time is 354.31
ns, so the <5 ns jitter is a tiny fraction of the bit time. That's why
you only see one bit transition on the scope -- because its horizontal
sweep rate has been blown way up to make the small jitter visible.

Even if that jitter were directly imposed on the local VCO, which it is
not because of loop filtering, it would still be reduced by a factor of
64 as the VCO clock is divided by 64 to produce the DAC sample clock.
5ns of jitter would become 78 ps. Even tinier when you consider that a
cycle of 20 KHz (the highest frequency the CD can reproduce) takes 50
*microseconds* to complete. What fraction of an audio cycle is that?
What's the FM modulation index? What is the resulting spectrum of
sidebands around the 20 KHz signal? What about lower frequency audio?
I'll leave the precise numbers to the reader, but it should already be
obvious that they're already tiny for the 20 KHz signal, and even
smaller at the lower frequencies.

>> Not only that, but it's showing us the raw signal jitter, before the
>> bit clock has been reconstructed by the receiver's PLL and divided
>> down to the sample rate, which decreases the jitter accordingly. And
>> furthermore, there's no indication of the exposure time, so we don't
>> know anything about the frequency spectrum of the jitter. That's
>> important too.
>
>
> It's the same for both cases. Level playing field, anybody?

Except that, in both cases, it's just too small to matter!

> But you don't believe in all this audiophile "mumbo jumbo" about good
> cables and bad cables, do you?

Don't put words in my mouth. You don't need cables that can pass 2 MHz
if you're carrying baseband analog. If you're carrying S/P-DIF, which
has its spectral peak at 2.8 MHz, then you do.

It would be hard to find a coaxial cable that couldn't do an adequate
job of passing the spectral components of a 2.8 MHz S/P-DIF signal over
a few feet in a home stereo system. We regularly use even smaller coaxes
to carry far higher frequencies from cell phones to external antennas.

I *can* see how there might be a problem with ordinary shielded,
twisted-pair microphone cables such as the kind long used in
professional work to run 600 ohm analog signals over significant
distances. Here you'd probably want a redesigned cable better suited for
megabit digital signals. Something like Cat-5, for example, which is
really cheap and goes up to 100 MHz. It doesn't have to be expensive or
gold plated to be good.

> You're general POV seems to be that if *you* can't hear a difference
> then no-one else should either.

Not at all! For one thing, I'm 48 and my hearing is not what it was at
18. But if I can't hear a difference, and some calculations cast strong
doubt on *anyone* hearing the difference, then I think it reasonable to
ask those who claim to hear a difference if they have conducted any
proper blinded listening tests. If not, then I question their assertion.
Audiophiles have a very long history of "hearing" all sorts of amazing
differences that seem to disappear as soon as proper controls are
introduced. That's a fact, and it would be foolish to ignore it here.

> Have you ever heard the effect of phase modulation on an audio signal?

Sure I have. Remember I said I help design modems for a living. Phase
modulation (e.g., PSK) is one of the modem designer's standard methods.
I'm well aware of what large amounts of phase modulation sound like;
I've spent many hours listening to these things while developing and
using modems on satellite links. But that doesn't mean very tiny amounts
of phase modulation sound at all alike.

> The sort of improvements audio enthusiasts wax lyrical about can often
> be attributed to phase. Soundstage, depth, clarity, all that sort of
> stuff. Even a small phase error can radically smear the sound.

Yeah, but can you do it in a properly controlled test? How do you know
you're not just fooling yourself?

> I don't hear discrete sounds or tones, no, but I do hear relationships
> between sounds and positional information that is encoded in the HF
> band. If you lose that you lose clarity and detail in the sound.

Again, so you say. People can claim to hear anything. Prove it with a
properly controlled test. That's all I ask.

> And what happens if the crystal is inaccurate?

If it's in a wall clock, the clock runs a little fast or slow. If it's a
local oscillator in a radio receiver, then its dial frequency
calibration is a little off. If it's the oscillator in a PLL recovering
the clock from a digital signal, then what happens depends on how far
off frequency it is, and the bandwidth of the loop filter. If it's
sufficiently far off and/or the loop bandwidth is too narrow, then the
loop will never lock, or it might take a very long time to lock. (BTW,
this was the big fear about the recent Huygens probe to Titan. They were
so concerned about a design flaw in the receiver PLL causing loss of
lock and all the data to be lost that they completely redesigned the
mission.) But if it's not so far off, then the oscillator will still
lock to the exact frequency of the incoming signal. There will be a DC
"stress" in the error control voltage to account for this offset, but
that won't affect the loops' ability to track the signal.

Again, crystals -- even cheap ones -- have the cleanest phase noise
spectra of just about any oscillator out there. Even atomic clocks. Such
devices actually generate their output with crystals, and they use the
exotic rubidium or caesium stuff to slowly "steer" the crystal's
frequency to compensate for frequency offset and slow drift.

> Because I can hear that my DAC sounds better than the DAC in the
> Squeezebox.

Once again: how do you *know* this?

> It's not about spending money for the sake of it. I want to get the best
> possible sound with the least possible outlay.

Quite frankly, you could have fooled me.

> You seem to be very closed to the possibility that others can hear
> things that you obviously can't. You're not wrong if you can't hear it,
> but you are wrong to insist that others can't either.

Not at all. I am still open to the possibility that others can hear it,
but if that goes strongly against what is known in the scientific
literature, then I insist on proper scientific proof, not a mere
anecdote. I think I'm quite right to insist on that.

> Please don't dismiss as impossible everything you can't explain or don't
> understand in engineering terms.

Not at all. It's just that before we can try to understand something in
engineering terms, we have to first find out if it's even real. Science
gives us the tools we need to determine that, even when it's something
as subjective as audio quality.

--Phil

Phil Karn
2005-03-02, 21:14
Phil Karn wrote:

> Even if that jitter were directly imposed on the local VCO, which it is
> not because of loop filtering, it would still be reduced by a factor of
> 64 as the VCO clock is divided by 64 to produce the DAC sample clock.

I'm going to have to revise and correct this. (In my defense, I've been
home sick with the flu for the past few days, and I'm not firing on all
cylinders.)

When the VCO clock is divided by 64 to obtain the 44.1 KHz sample clock,
the jitter time is *not* divided by 64.

With this correction, I believe the rest of my analysis remains valid.
That is, the clock division by 64 reduces the modulation index of the
jitter on audio components by that same factor. So 5ns of jitter (a
little more than that scope showed) represents only 1/10,000 of a cycle
of a 20 KHz sine wave (period 50 microsec). And the modulation index
would be proportionately lower at lower audio frequencies; down at 1
KHz, where there's far more energy in a typical audio signal, 5 ns would
be only 5/1000000 (5 millionths) of an audio cycle! These are truly
*tiny* phase modulation indices that would produce very little in the
way of PM/FM sidebands. I just can't imagine that anyone could hear them.

And I've left out the effects of PLL smoothing, which reduces jitter
even further.

But if you feel otherwise, conduct some properly controlled listening
tests. I'm all ears.

--Phil

Robin Bowes
2005-03-03, 01:47
Phil Karn wrote:
> Robin Bowes wrote:
>
>> What information on that image tells you that? It seems to me you're
>> not understanding what you're seeing.
>
>
> The legend says the time trace is 5 ns/div. A pair of measuring lines
> implies that the jitter on the uncorrected is somewhat less than that,
> about 3 ns.

Yes, that's perfectly reasonable. It still doesn't suggest that the
scope scale has changed between traces. You're looking at multiple frame
transitions superimposed on top of each other, not whole frames.

>
> As I recall (I can't find good, recent references), each subframe in
> S/P-DIF carries a (usually) 16-bit PCM sample in a 32-bit subframe, so
> the clock rate of the composite stereo signal would be 64 times the 44.1
> KHz CD sampling rate, or 2.8224 MHz. At this rate, a bit time is 354.31
> ns, so the <5 ns jitter is a tiny fraction of the bit time. That's why
> you only see one bit transition on the scope -- because its horizontal
> sweep rate has been blown way up to make the small jitter visible.
>
> Even if that jitter were directly imposed on the local VCO, which it is
> not because of loop filtering, it would still be reduced by a factor of
> 64 as the VCO clock is divided by 64 to produce the DAC sample clock.
> 5ns of jitter would become 78 ps. Even tinier when you consider that a
> cycle of 20 KHz (the highest frequency the CD can reproduce) takes 50
> *microseconds* to complete. What fraction of an audio cycle is that?
> What's the FM modulation index? What is the resulting spectrum of
> sidebands around the 20 KHz signal? What about lower frequency audio?
> I'll leave the precise numbers to the reader, but it should already be
> obvious that they're already tiny for the 20 KHz signal, and even
> smaller at the lower frequencies.

Your entire hypothesis is based on your prejudiced belief that jitter is
not audible, or rather the level of jitter likely to be present in
digital audio systems is not audible. I could turn this whole thing
round and say "prove that jitter is not audible at that level."

>> It's the same for both cases. Level playing field, anybody?
>
>
> Except that, in both cases, it's just too small to matter!

See above.

>> But you don't believe in all this audiophile "mumbo jumbo" about good
>> cables and bad cables, do you?
>
>
> Don't put words in my mouth. You don't need cables that can pass 2 MHz
> if you're carrying baseband analog. If you're carrying S/P-DIF, which
> has its spectral peak at 2.8 MHz, then you do.
>
> It would be hard to find a coaxial cable that couldn't do an adequate
> job of passing the spectral components of a 2.8 MHz S/P-DIF signal over
> a few feet in a home stereo system. We regularly use even smaller coaxes
> to carry far higher frequencies from cell phones to external antennas.

Again, you're defining your own standards. You use the word "adequate";
that suggests that you accept that some cables will do a better job of
carrying SPDIF than others?

> I *can* see how there might be a problem with ordinary shielded,
> twisted-pair microphone cables such as the kind long used in
> professional work to run 600 ohm analog signals over significant
> distances. Here you'd probably want a redesigned cable better suited for
> megabit digital signals. Something like Cat-5, for example, which is
> really cheap and goes up to 100 MHz. It doesn't have to be expensive or
> gold plated to be good.

You're the one with the "expensive, gold-plated" fixation.

>> You're general POV seems to be that if *you* can't hear a difference
>> then no-one else should either.
>
>
> Not at all! For one thing, I'm 48 and my hearing is not what it was at
> 18. But if I can't hear a difference, and some calculations cast strong
> doubt on *anyone* hearing the difference, then I think it reasonable to
> ask those who claim to hear a difference if they have conducted any
> proper blinded listening tests. If not, then I question their assertion.
> Audiophiles have a very long history of "hearing" all sorts of amazing
> differences that seem to disappear as soon as proper controls are
> introduced. That's a fact, and it would be foolish to ignore it here.

It doesn't follow that every reported subjective improvement in sound
need to be subjected to blind listening tests to prove it. I actually
share your skepticism, but I also believe what I hear.

>> Have you ever heard the effect of phase modulation on an audio signal?
>
>
> Sure I have. Remember I said I help design modems for a living. Phase
> modulation (e.g., PSK) is one of the modem designer's standard methods.
> I'm well aware of what large amounts of phase modulation sound like;
> I've spent many hours listening to these things while developing and
> using modems on satellite links. But that doesn't mean very tiny amounts
> of phase modulation sound at all alike.

And it doesn't mean that small amounts are inaudible.

>> The sort of improvements audio enthusiasts wax lyrical about can often
>> be attributed to phase. Soundstage, depth, clarity, all that sort of
>> stuff. Even a small phase error can radically smear the sound.
>
>
> Yeah, but can you do it in a properly controlled test? How do you know
> you're not just fooling yourself?

Because I'm very objective when doing this sort of thing. If I can't
hear any difference then I don't fool myself into thinking that I can.
For example, I did some tests comparing the SB analogue output with a
couple of outboard DACs: an Arcam Delta and a Perpetual Technologies
P3-A with P1-A correction engine ($2500 worth of kit).

In my test setup, I couldn't hear any difference. On a friends high-end
system there was a clear difference. This was one of the reasons I
modified my amp setup.

>> I don't hear discrete sounds or tones, no, but I do hear relationships
>> between sounds and positional information that is encoded in the HF
>> band. If you lose that you lose clarity and detail in the sound.
>
>
> Again, so you say. People can claim to hear anything. Prove it with a
> properly controlled test. That's all I ask.

It is not necessary to subject all perceived improvements/differences to
controlled testing. At the end of the day, it comes down to a subjective
decision whether, firstly, you can tell the difference betweeen A and B
and, secondly, whether you prefer A or B. The differences I heard in my
system before/after modifications are significant, not small placebo
effects. It's night and day.


>> And what happens if the crystal is inaccurate?
>
> If it's in a wall clock, the clock runs a little fast or slow. If it's a
> local oscillator in a radio receiver, then its dial frequency
> calibration is a little off. If it's the oscillator in a PLL recovering
> the clock from a digital signal, then what happens depends on how far
> off frequency it is, and the bandwidth of the loop filter. If it's
> sufficiently far off and/or the loop bandwidth is too narrow, then the
> loop will never lock, or it might take a very long time to lock. (BTW,
> this was the big fear about the recent Huygens probe to Titan. They were
> so concerned about a design flaw in the receiver PLL causing loss of
> lock and all the data to be lost that they completely redesigned the
> mission.) But if it's not so far off, then the oscillator will still
> lock to the exact frequency of the incoming signal. There will be a DC
> "stress" in the error control voltage to account for this offset, but
> that won't affect the loops' ability to track the signal.

I meant, what happens if the crystal that controls the sending signal is
inaccurate? Answer: the timing information in the digital stream is
wrong and the signal suffers from jitter.

>> Because I can hear that my DAC sounds better than the DAC in the
>> Squeezebox.
>
> Once again: how do you *know* this?

Erm, listen to A, then listen to B; ah! A sounds better than B. Easy really.

>> It's not about spending money for the sake of it. I want to get the
>> best possible sound with the least possible outlay.
>
>
> Quite frankly, you could have fooled me.

What have I said that suggests I want to "spend money for the sake of
it" on audio equipment? I have been involved in audio engineering for
around 20 years. I have a degree in ElectroAcoustics. I have worked in
hifi retail. I have had extensive exposure to both high-end and
professional audio systems. I *know* that my home system doesn't sound
as good as I would like it to but I don't have the $$$ to buy anything
better at the moment. I also have a hobbyist interest in electronics, so
enjoy combining the two by experimenting with modifying my equipment,
thus giving me better performance without paying silly prices for
so-called "audiophile" equipment.

>> You seem to be very closed to the possibility that others can hear
>> things that you obviously can't. You're not wrong if you can't hear
>> it, but you are wrong to insist that others can't either.
>
>
> Not at all. I am still open to the possibility that others can hear it,
> but if that goes strongly against what is known in the scientific
> literature, then I insist on proper scientific proof, not a mere
> anecdote. I think I'm quite right to insist on that.

Whilst I share your skepticism of the audiophile world, I think you are
rather too far in the opposite corner. Not everything can be explained
by science yet. Acoustics is still as much of an art as a science.

>> Please don't dismiss as impossible everything you can't explain or
>> don't understand in engineering terms.
>
>
> Not at all. It's just that before we can try to understand something in
> engineering terms, we have to first find out if it's even real. Science
> gives us the tools we need to determine that, even when it's something
> as subjective as audio quality.

It is not necessary to scientifically understand everying to establish
that it is real. Also, not all real phenomena are completely understood
in engineering terms.

R.
--
http://robinbowes.com

Marc Sherman
2005-03-03, 11:03
Robin Bowes wrote:
>
> Your entire hypothesis is based on your prejudiced belief that jitter is
> not audible, or rather the level of jitter likely to be present in
> digital audio systems is not audible. I could turn this whole thing
> round and say "prove that jitter is not audible at that level."

You could, but you'd be wrong. You can't prove a negative existential
claim. You can, however, easily prove a positive one (with a blind A/B
test in this case). The burden of proof therefore lies with the person
claiming the positive.

I /think/ that's Argumentum ad Ignorantiam, but I'm not 100% sure.
Anyone remember their first year Formal Logic better than I do? :)

- Marc

Phil Karn
2005-03-03, 21:37
Robin Bowes wrote:

> Yes, that's perfectly reasonable. It still doesn't suggest that the
> scope scale has changed between traces. You're looking at multiple frame
> transitions superimposed on top of each other, not whole frames.

Exactly my point! If the scope trace was slow enough to capture whole
frames, or better yet the multiple frames required to make up complete
cycles of most audio components, then the bit transitions would be razor
sharp because the jitter is far too small to be seen at that scale. So
they have to blow up the scope scale enormously just to make the
"problem" visible at all.

I never suggested that the scope scale changed between traces! Nor have
I suggested that the box under test doesn't indeed visibly reduce signal
jitter as measured on a scope set to 5ns/div. My whole point is that
this is all irrelevant if you can't hear the difference in the
reconstructed audio, and you've never proven scientifically that you can.

> Your entire hypothesis is based on your prejudiced belief that jitter is
> not audible, or rather the level of jitter likely to be present in
> digital audio systems is not audible. I could turn this whole thing
> round and say "prove that jitter is not audible at that level."

You could, but as Marc has already said, you'd be wrong. In science,
it's up to someone arguing for the existence of a phenomenon to prove
it. It's not the responsibility of the others who doubt its existence to
prove it doesn't exist.

It's often difficult or impossible to prove a negative. You merely have
to produce *one* person who can consistently hear the difference in a
controlled test, while I could test every person on the planet, find no
one who can hear the difference, and still I couldn't firmly conclude
that it is always inaudible. Someone might be born tomorrow who will
grow up able to hear the difference. Extremely unlikely, but possible.

If you feel that this degree of jitter is so plainly audible, why do you
so strongly resist the notion of conducting properly controlled
listening tests?

> Again, you're defining your own standards. You use the word "adequate";
> that suggests that you accept that some cables will do a better job of
> carrying SPDIF than others?

Of *course* some cables will carry S/P-DIF better than others! So what?
Some cables (e.g., "zip" cord from a table lamp) are probably wholly
unsuited to digital data signals in the 2 megabit/sec range. However,
they are rarely marketed for that purpose. Fortunately, an entirely
adequate coaxial cable can be had for very little money. It is not
necessary to spend a lot on this. And by "adequate" I mean a cable with
a bandwidth sufficient to reduce the data-dependent jitter to
practically zero, far below even levels that you'd agree are inaudible.

> You're the one with the "expensive, gold-plated" fixation.

What makes you say that? Many inexpensive cables, connectors and
components can do an excellent job of reproducing audio, especially with
digital technology. It is, rather, the "golden ear audiophiles" who seem
to think that no mass market audio device or component could *possibly*
ever be good enough for them, at least not without extensive and usually
expensive modifications. We saw quite a flare-up of this elitist
nonsense when the CD first came out in the early 1980s. Quite a few
audiophiles had their egos severely bruised when suddenly an inexpensive
mass-market device produced far better fidelity than all the expensive
turntables they'd been touting for years. While I haven't kept track of
this phenomenon since, it seems it's still alive and well.

> It doesn't follow that every reported subjective improvement in sound
> need to be subjected to blind listening tests to prove it. I actually
> share your skepticism, but I also believe what I hear.

Actually, it does. The history of audio is *full* of people honestly
convincing themselves that they could hear significant differences that
disappear completely when controlled tests are conducted. This is the
basic principle of experimental science -- to find ways to keep the
experimenter's own personal biases and mistakes from affecting the outcome.

> And it doesn't mean that small amounts are inaudible.

Actually, it does. Very small amounts of phase modulation *are* totally
inaudible. This can be confirmed both experimentally and theoretically;
as the modulation index is lowered, the power diverted from the
fundamental signal into its sidebands becomes monotonically less. At
zero modulation, all the power is in the fundamental and nothing is in
the sidebands.

> Because I'm very objective when doing this sort of thing. If I can't
> hear any difference then I don't fool myself into thinking that I can.

I don't doubt that you honestly believe this. But again, the history of
audio shows how easy it is to fool yourself.

The same is true in medicine. The greatest medical innovation of all
time wasn't anesthesia, or penicillin, or the polio vaccine or anything
like that. It was the double-blind scientific study. Otherwise we'd
never know for sure what really works and what doesn't. It's considered
so important to eliminate preconceived notions that not only does the
patient not know if he's getting the test drug or a placebo, but his
doctor doesn't know either. And it works.

I had this demonstrated very effectively recently when (at my
suggestion) I had my doctor order up a double-blind study on some
topical medications we were trying for chronic pain. I guessed wrong on
all three. And I was so *sure*! It was a real eye-opener.

> It is not necessary to subject all perceived improvements/differences to
> controlled testing. At the end of the day, it comes down to a subjective
> decision whether, firstly, you can tell the difference betweeen A and B
> and, secondly, whether you prefer A or B. The differences I heard in my
> system before/after modifications are significant, not small placebo
> effects. It's night and day.

It is necessary to do controlled testing when, as here, a serious
question arises as to the reality of the perceived difference. As for
testing, I suggest that if you cannot even tell the difference between A
& B in a controlled test, then it doesn't matter which one you prefer.

> I meant, what happens if the crystal that controls the sending signal is
> inaccurate? Answer: the timing information in the digital stream is
> wrong and the signal suffers from jitter.

Nope, you're wrong. If the *sending* clock is wrong, that means the data
will come out at a slightly different speed than nominal. The piece will
take slightly longer or shorter to play and the pitch will be slightly
affected. But one thing it *won't* do is add jitter to the reconstructed
signal! That is, assuming that the receiver can track the transmitter's
frequency offset. Otherwise it won't lock and you'll hear nothing.

> Erm, listen to A, then listen to B; ah! A sounds better than B. Easy
> really.

Except you *know* when you're hearing A and then B, and what A and B
correspond to. Wishful thinking thus has the opportunity to affect the
results. If the differences are truly real, then you'll still hear them
even if you don't know what A and B correspond to, or even if you don't
know when the switch occurs.

> What have I said that suggests I want to "spend money for the sake of
> it" on audio equipment? I have been involved in audio engineering for
> around 20 years. I have a degree in ElectroAcoustics. I have worked in
> hifi retail. I have had extensive exposure to both high-end and
> professional audio systems. I *know* that my home system doesn't sound
> as good as I would like it to but I don't have the $$$ to buy anything
> better at the moment. I also have a hobbyist interest in electronics, so
> enjoy combining the two by experimenting with modifying my equipment,
> thus giving me better performance without paying silly prices for
> so-called "audiophile" equipment.

If your audio budget is limited, like most people, then the scientific
method is even more important to you. You just can't afford to waste
money on high priced boxes like $500 "jitter removers" unless they
really work. And quite frankly, if you have worked in hifi retail, I'm
sure you know all the tricks of the trade that can be used to make a
customer prefer one device over another (the "better" device is, of
course, the one that will earn you a bigger commission). Not that *you*
ever used these tricks, I'm sure, but you must have at least heard of them.

The age-old advice still applies: in almost every case, if you have some
extra money and want to improve the sound of your audio system, spend it
on better speakers. Compared to loudspeakers, every other stereo system
component reached absolute perfection long ago.

> Whilst I share your skepticism of the audiophile world, I think you are
> rather too far in the opposite corner. Not everything can be explained
> by science yet. Acoustics is still as much of an art as a science.

Not everything can be explained or answered by science (e.g., is there
or isn't there a god?) but acoustics is *not* one of those fields.
Questions like "can people hear the difference that this box makes?"
falls squarely into the realm of science, and have been answered quite
effectively by scientific methods for many decades.

> It is not necessary to scientifically understand everying to establish
> that it is real. Also, not all real phenomena are completely understood
> in engineering terms.

You're attacking a straw man, as I *never* said or even implied that
every phenomenon must have a full scientific explanation to be
considered real. Countless drugs now on the market were approved by the
FDA as safe and effective, yet the mechanisms of their action are
unknown or only partly known. That's okay, because we don't have to know
*how* they work to know whether they actually work or not. If I run a
double-blind clinical study on some drug and the data says that a much
greater percentage of those who took the drug had remissions of their
cancers than those who took the placebo, then I don't have to understand
the precise mechanism by which the drug attacked the cancer just to know
that it did. Morphine has been used to relieve severe pain for thousands
of years, but only in the 1960s and 1970s did researchers begin to
unravel how it works. But that doesn't mean it didn't work until then!

And the same is true with the subjective audio phenomena we're
discussing. For if I can show with a blind study that you can't even
reliably tell the difference between A&B, then you cannot rationally
argue that you prefer A over B.

--Phil

momerath
2005-03-03, 22:24
I'm loving this conversation. Phil clearly knows what he's talking
about to a greater degree than any jitter-audibility-skeptic I've ever
encountered. Personally, I am quite sure that differences can be
heard between transports (even bit perfect ones, I believe), but am
less sure that the kind of jitter that might be present in a low end
bit-perfect transport vs a high end reclocker is. The thing is, I
plan to find out.

Phil is absolutely correct that the burden of proof is on the
audiophile, and that a double blind study is the only practical way to
prove the audibility of jitter. I intend to do this in the near
future with some skeptical friends.

Unfortunately, everyone that isnt present at the test will have to
take or leave my reported findings. It would be neat to have a large
funded study that more people might believe, though to my mind, the
funding of such a study makes it even more untrustworthy:) After all,
who would fund the study but those standing to make or lose money on
the results? Too bad we dont have some sort of science-oriented
society in which truth is the only commodity! ;)

~Michael

> And the same is true with the subjective audio phenomena we're
> discussing. For if I can show with a blind study that you can't even
> reliably tell the difference between A&B, then you cannot rationally
> argue that you prefer A over B.
>
> --Phil

Phil Karn
2005-03-04, 01:44
momerath wrote:

> Phil is absolutely correct that the burden of proof is on the
> audiophile, and that a double blind study is the only practical way to
> prove the audibility of jitter. I intend to do this in the near
> future with some skeptical friends.

By all means, if you can run a controlled test, then do it!

I'm not sure how to best go about that. I know of some audio ABX testing
software, but I think it's designed mainly to evaluate lossy codecs.
You're testing something at a much lower layer. It might be possible to
simulate the effect of jitter in software by using similar algorithms to
those that convert between sampling rates, although here the rates would
be very nearly identical. This might not satisfy everyone, though, as it
wouldn't actually test the real hardware, so an actual hardware switch
would probably be the best as well as the simplest.

The most crucial thing in any audio A/B test is level matching.
Otherwise, the louder signal will almost always sound "better". This
shouldn't be hard to do in an all-digital system; just make sure
identical bit streams come out both cables. Also, make sure that all
switching transients and propagation delays are the same; if you do a
test that involves A vs A and A vs B, you don't want the lack of a
switching transient or change in delay in the first case to give it away.

Keep in mind that a positive result will be less conclusive than a
negative result, at least with that particular subject. A positive
result could be due to some artifact other than the property you're
trying to test (e.g., different audio levels, different time delays,
etc) and you've got to carefully exclude every possible artifact other
than the one property (jitter) you're trying to test.

There's a classic example given in a MIT science textbook about just how
incredibly hard it can be to design a proper experiment on human
perception. It asked the question "Can humans detect magnetic fields?"
The answers kept coming up "Yes", but closer examinations kept finding
various bugs in the experiment that explained the positive results
(e.g., the test subjects could hear the buzzing of the electromagnets,
or saw the slight dimming of the lights as the switch was thrown, etc.)
Eventually, after going to very great lengths to eliminate all
artifacts, they got their answer: No, humans cannot detect magnetic
fields, at least not at the intensities tested.

Let me know what happens!

--Phil

kdf
2005-03-04, 02:04
Quoting Phil Karn <karn (AT) ka9q (DOT) net>:

> momerath wrote:
>
> > Phil is absolutely correct that the burden of proof is on the
> > audiophile, and that a double blind study is the only practical way to
> > prove the audibility of jitter. I intend to do this in the near
> > future with some skeptical friends.
>
> By all means, if you can run a controlled test, then do it!

just dont' fake it. I will not fall into this dispute, save to mention that I
recently attended a seminar to help with high speed data, and the speaker
actually presented a demo on jitter, trying to show how a certain IC managed it
well....

I did not return after lunch, as I noticed that the example trace was being
triggered on the edge displayed. The only time I have encoutered a case where I
might actually care specifically acout jitter in my designs...despite many
demands on the reps to produce real data...all I got was crap, including
admissions that their own spec was not met by any crystal that they could find
( in this case, it was Texas Instruments). Their (preliminary only)
application note resorted to specifying test cases, even though they did not
actualy meet the officially supported spec given for the IC.

since then, I've come to learn that even 1Gb is considered slow speed and
pointless for sweating details on anyting but HUGE backplanes.

having said that, I'll never put my career against what can be picked up by ear.
My field is visual, and I know that there are amazing things that you can pick
up by eye that NO analyser will find. I have no reason to believe that audio is
any different.

keep up the discussion/argument as you will. the end result will be interesting
:)

-kdf

Triode
2005-03-04, 09:01
Phil/Robin,

I've been holding off contributing to this thread, but what the heck - here's a few views to fuel the thread (but I don't predend to
be an expert in this area, just an enthusiastic amateur...)

1) Is jitter audable and where's the evidence:
From my personal experience, yes [i.e I can hear differences which I would put down to jitter - more of in minute]

However I would suggest you do a quick search on the AES pages for references to papers about jitter. The AES (Audio Engineering
Society) is definately a body which doesn't give much time to audiophile snake oil and hype - but does consider the serious
engineering behind audio. Unfortunately they charge for back copies, so the only access you can get to papers on line are ones
hosted by companies with a vested interest or accademics (e.g. Hawksford - worth looking at his list of papers
http://www.essex.ac.uk/ese/research/audio_lab/malcolms_publications.html#Journal)

Anyway, you may be interested in ref & graph 9 of the following: http://www.nanophon.com/audio/jitter92.pdf [Can't actually find
the paper referred to online]

2) Can't spdif be implemented in such a way to avoid jitter?
I would agree with this to a reasonably degree - essentially the early implimentations were naive as they didn't understand the
impact of jitter and in many cases this remains.
I agreed this comes down to how much engineering is put into the transmission line:
- consumer products bandwidth limit the output to ensure emissions tests are passed with cheap cables - in many cases they arguably
constrain it too much
- most links are poorly terminated so that reflections occur. In the case of most cheap consumer gear both ends are capacitively
coupled (no problem with that), but use cheap RCA connects etc, which ensure some reflections. It is argued by some that assuming
the cable is very short this will distort the signal seen at the receiver as it is the sum of the original plus reflections. [One
proposal I have seen is to always use 1.5m of cable to delay the reflection!]
- common older spdif digital receiver chips are good at capturing the signal, but are possbly limited in other respects. E.g. the
Crystal 8412 in my DAC (one of the industry standards from a few years ago which is easy to use and hence used by many!) - its PLL
doesn't attenuate any jitter in the audio band - corner frequency is at 25 KHz. Additionally the critical frame (L/R) transition
clock is derived straight from the input signal rather than from the clock produced by the PLL - dacs often reclock this!. (Some of
the audio diy community also suggest the 8412 injects significant noise back into the transmission line if coupled to it without
buffering due to its input stage using Smitt triggers)
Bottom line - the implementation of the input receiver is potentially a dominant factor in this conversation.

3) Why do different DAC chips impact things?
These days I think there is a dependance on the DAC conversion technology and many new DAC chips take steps to reduce sensitivity to
jitter.
Its relatively simple to comprehend a non oversampled multibit converter which changes its output on a transition of the L/R Clock -
in this case if there is jitter in the L/R clock transition then this has a direct relationship to the resulting audio signal - as
the correct voltage is produced, but not at the correct point in time. All the simple analysis consider this case.
With single bit and newer DAC technolgies (some of which are a small number of parallel multibits running at higher speed, or some
other hybrid of single and multibit), then there appears to be the opertunity to reduce this concern by clever processing/noise
shaping stuff. Hence the claims of many newer converter chips to have reduced jitter sensitivity.

4) What's my experience.
Well with my current DAC and high end system I can definately hear the difference between spdif sources. I put this down to the
fact that my current DAC has a very basic design for extracting the spdif clocks. Specifically between the Squeezebox and my
reference CD transport there is a small but noticable difference (I'll not bore you with the audiphile phrases). However I am happy
with this because the cost of the mods I have done on the CD transport exceed the cost of the Squeezebox! [The CD transport is
actually a much modified one from a few years ago which has a new low jitter clock and I reclock the spdif signal against this
imediately prior to the output buffer.]

In contrast the Squeezebox as a highly integrated device has quite a simple clock circuit and hence is always going to generate a
bit more jitter than a modern high end transport.
This is the reason I would recommend audiophiles looking for the best high end sound to use the Squeezebox with a modern dac with
reclocking. In this case there are many happy users who are getting a high end hifi experience + the all the usability and
convience benefits from the Squeezebox. [However do try any new dac against the Squeezebox as there are a very small number of
reports about certain reclocking dacs having problems locking to the Squeezebox in pcm mode]

5) Lets put this all into perspective.
Unless you care about hi end hifi, you won't care about this! (You need to be talking about >$500 dacs before you want to get
interested in this)
If you do, then you probably recognise that human hearing is very sensitive and to some degree personal taste comes into it... I
say this because I don't believe measurements are the final arbiter of a good audio experience. [I listen through valve (tube)
amplification which in simplistic terms measures poorly, but has that certain something to the sound which is highly engaging... In
contrast transister power amps of a few years ago measured very well in terms of distortion but sounded awful....]

Adrian