On Wed, October 26, 2022 9:11 pm, Thomas Tammann wrote:

...

Yeah, the Ethernet clock is not related at all to the audio clock used in
the D/A converter, even for synchronous audio-over-IP designs like Ravenna
and Dante.  Consumer music players (as opposed to professional audio
production equipment) do not even have a synchronized clock, they
free-run.  Zero link between the Ethernet clock in the switch and audio
quality. The quality of the clock to the digital-to-analog converter does
influence quality, but again there is absolutely zero link between the
Ethernet clock and the audio clock.

According to that link you provided it does have a Crystek crystal
oscillator, which is relatively low noise.
You can see the specs here:
https://www.crystek.com/crystal/spec-sheets/clock/CCHD-575.pdf

If it is working well an ovenized SC cut oscillator should be able to
reach that.

What you are attempting to do is understandable.  The entire premise
behind it doesn't make sense from a technical standpoint, but the starting
point is relatively straight forward, compare the phase noise spec of your
surplus OCXO to the Crystek spec to make sure your new oscillator is
better than the oscillator internal to the device.

--
Chris C

Thank you all!

First and foremost, most people who actually heard comparison with different Ethernet setups agree that it sounds different. Technically, it does not make sense at the surface, zeros and ones are just that. But its not that simple. Nobody argues that the digital signal alters depending on a e.g. switch or shielded vs unshielded CAT6 cable. But the zeros and ones are only logical, not physical aka voltages...

Also, there is a huge gap between the “measuring is everything” crowd in audio and “hearing trumps everything” crowd. Both are wrong. The truth is somewhere in the middle and we try to figure out which measurements have actually an influence on audio and why. Bottom line, changes on all levels of the OSI model CAN have an influence in the resulting quality of the audio system.

Crystek vs my Symmetricom OCXO:
The producer of my switch says that the Crystek actually measures much better in real than on paper. He talks to the manufacturer and has insight knowledge.
The Symmetricom may or may NOT improve over the Crystek depending on the individual specimen and implementation.

Hence, my goal is to measure: I hope I would see
114@1HZ and 140db@10Hz. Not only would it confirm what I hear ;-) but also justifies the money I put down for a “preselected” (binned) specimen. Also, if I buy another clock with different phase noise and ADEV, I can hear for myself how much this makes a difference in real world audio pleasure vs upgrading eg a pre-amp….

But yes, in the meantime I learned a lot from you guys, thank you!!

Tom

PS:
If you wanna read how nerdy audio guys are ;-)
https://audiophilestyle.com/forums/topic/59419-master-clock-for-your-etherregen/ https://audiophilestyle.com/forums/topic/59419-master-clock-for-your-etherregen/
(A few guys would make great members of this mailing list ;-)

This guy is seriously interested to measure in oder to understand thinks he can hear but not explain
https://www-open--end--music-com.translate.goog/forum/privatforen/thomas-michael-rudolph-tmr/651284-messungen-von-ethernet-infrastruktur-switches-nur-lesen?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=de&_x_tr_pto=wapp https://www-open--end--music-com.translate.goog/forum/privatforen/thomas-michael-rudolph-tmr/651284-messungen-von-ethernet-infrastruktur-switches-nur-lesen?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=de&_x_tr_pto=wapp

Hi

( see below )

Per their data, the Crystek Crystek CCHD-575 at 125 MHz

https://www.crystek.com/crystal/spec-sheets/clock/CCHD-575.pdf

Comes in at -89 dbc / Hz at 10 Hz offset. That’s not a guess, that’s their
actual measured data on their test gear. The shot shown on page 5 of the
spec is the output of a phase noise test set.

As mentioned previously, phase noise goes as 20 log N when you multiply
or divide the frequency. 20 log of 12.5 is 22 db. An ideal divider would put
the 575 at 89 + 22 = -111 dbc / Hz divided to 10 MHz. ( = the highest possible
PLL comparison frequency ).

If the PLL circuits contribute little or no noise, a reference at about -120 dbc / Hz
should be able to noticeably improve the close in noise. If the PLL has a noise
floor above -120 dbc / Hz ( at 10 Hz …. ) , then a “better” reference will not do much.

The Symmetricom part you have most certainly is much better than -120 dbc / Hz
at 10 Hz offset. That’s based on testing a number of the same parts.

While I’m sure that Crystek is proud of their parts, there are basic reasons
why a VHF VCXO isn’t going to do as well as a low frequency overtone SC
based part. The much higher Q on the SC is one, there are others.

Floks have mesured the actual performance of the Crystek parts. They
have not observed them to be massively better than spec close in. A couple
db better … sure. Tens of db better close in …. nope.

Welcome to crystal oscillator design. Some of us on this list spent > 40 years
as engineers working in that field.

Bob

Hi

Moving back a bit: What can you expect an external reference to
improve on a clock system?

Looking for phase noise improvement typically isn’t what you are after.
The phase noise is determined by what’s already inside the box. The impact
of the reference will not be very apparent as you look at a full up phase
noise plot of the output clock.

What the reference will do is to improve the long term stability of the clock
signal. The devices that are already inside the box are not great for drift / aging /
temperature / warmup / wander / ( or whatever term you might wish to use ….).
There are a number of things you could look for an improvement in. All are
based on the assumption that the device currently is based on the Crystek part
or possibly a TCXO. ( If a TCXO is present, I have not seen mention of it ).

1. Temperature stability should improve by several orders of magnitude. Even in
   a normal room setting this improvement should be apparent and very measurable.

2. Aging should be much better with the OCXO. Yes, the OCXO should be kept
   on power all the time. Compared to the Crystek part it should be orders of magnitude
   better.

3. Short and medium term ( tau of 1 second to 1,000 seconds ) stability should be
   much better with the OCXO than with the stand alone box. Any of the typical DEV
   measurements should show this. Each one may highlight things in a slightly different
   way.

4. Turn on stability with the OCXO always on power and the box cycled, should be
   much better. Effectively the “warm up” process is short circuited when running in
   this mode.

How much any or all of these matter in the actual system is up to the user. However,
these are the areas that drive folks to put a reference on a clock generation device. If
they do matter, then they are the things that probably need to be characterized on
this or that reference ( and on the system).

Bob