On Mon, 10 Jan 2022 08:26:26 +0000 (UTC)
Norman Reitz via time-nuts time-nuts@lists.febo.com wrote:

Try contacting Bernd Neubig at Axtal. He is a fellow time-nut and likely to be
open to supply you with what you need.

That said, are yo sure you need such stringend phase noise requirements?
It's audio. Nobody is going to hear whether the noise is -60dBc or -80dBc @ 1Hz,
much less -120dBc. As you have noticed, what you want is not readily available.
There are only a handful OCXO available that reach that level. One of them is
the famous Oscilloquartz 8607 and its successor the Rakon HSO13/HSO14. But be prepared
to pay the price of a small car for each of them. And there is the NEL ULPN OCXO 1714a,
but I don't know how much that one costs.

			Attila Kinali

--
In science if you know what you are doing you should not be doing it.
In engineering if you do not know what you are doing you should not be doing it.
-- Richard W. Hamming, The Art of Doing Science and Engineering

Hi

Do you need both outputs from the same part? If so it likely will
cost less to synthesize the 25 MHz off of the 10 MHz on your side
of things rather than have it done in the OCXO.

Is a price above $5,000 each at 100 pieces reasonable to you?
If not, you might need to look at your spec’s a bit. As mentioned
earlier, your phase noise specs are more than a bit tight.

Bob

You can have a look at morion.com.ru
They sell small quantities, no trade license. I think the only product that
can meet the -120 dBc at 1 Hz (10 MHz carrier) is the MV336, which is
expensive

Il giorno lun 10 gen 2022 alle ore 10:03 Norman Reitz via time-nuts <
time-nuts@lists.febo.com> ha scritto:

On Mon, 10 Jan 2022 12:35:17 +0100
Attila Kinali attila@kinali.ch wrote:

To give here a bit more background: psychoacoustic masking, which is the relevant
metric here, mans that we cannot discern sounds that are close to eachother with
one of them being louder than the other. Depending on who you listen to, it's
usually a sound masking another sound at a distance of 100Hz up to 20dB to 40dB
lower. Even if we account for someone with golden ears and use 60dB, that would
translate to a noise spec of -60dBc @ 100Hz offset. That's a spec that almost all
XO do fulfill. A good VCXO (40-100MHz) is somewhere around 90-100dBc @ 100Hz.
Any OCXO will fulfill that spec too, even the tiny DIL-14 ones (most are at
-110-140dBc @100Hz @10MHz).

And this doesn't take into account that we are arguing about audio frequency specs
at HF frequencies. I.e. if we use the 10MHz clock and use it to derive a sampling
clock for an ADC to sample a 20kHz signal, the noise performance improves by another
~25dB... at least (if the design is done right, it can be up to 50dB)

What is more important than close in noise, though, is broadband noise performance
and spurs. For someone with good ears, it's not unheard of to be able to discern
far away noise and spurs down a -100dB-120dB. Especially the spurs can be quite
hard to control, depending on what clock synthesis system is used.

Another important spec, especially for recording, is accuracy of frequency. An
offset of just 1ppm becomes 3.6ms if you record for an hour. That's something
most people can hear already. But whether this actually elevant or not depends
on how the recording is done. The usual way is to have a central master clock
that feeds all clocked devices, such that all of them have the same notion of
time/frequency. In that case, quite high frequency deviations can be tolerated,
way beyond what a simple XO would deliver.

			Attila Kinali

--
In science if you know what you are doing you should not be doing it.
In engineering if you do not know what you are doing you should not be doing it.
-- Richard W. Hamming, The Art of Doing Science and Engineering

I'm curious, too, what kind of audio application requires this level of
phase noise performance, and why.

Cheers
Stefan

Am 10.01.2022 um 12:35 schrieb Attila Kinali:

We are receiving such inquiries from "Audio nuts" rather frequently, but
also from professional high-end audios-studio equipment makers. There
argument is often, that the spatial transparency of the sound, i.e. how
exactly you can locate the sound source (instrument in an orchestra) would
be noticeably improved by such ultra-low noise OCXO sources. So it should be
more about time or phase (jitter?) than about frequency....

As the customer and his belief is "king" at AXTAL - as long as doable and
payable - we have developed our AXIOM45ULN series, where the best phase
noise option guarantees a PN level of -115 dBc/Hz @ 1 Hz. But this kind of
performance can only be achieved by a crystal selection with rather low
yield. Therefore, as a manufacturer you need enough customers who accept
that "less is sufficient" and will buy the OCXO made from the other
crystals. We also are getting a few parts with -120 dBc/Hz @ 1 Hz out of a
larger lot, but we rather keep them than selling them to everybody.

Best regards
Bernd

-----Ursprüngliche Nachricht-----
Von: Attila Kinali [mailto:attila@kinali.ch]
Gesendet: Montag, 10. Januar 2022 15:42
An: Discussion of precise time and frequency measurement
time-nuts@lists.febo.com
Betreff: [time-nuts] Clock specs for audio (was: High precision OCXO
supplier for end costomers)

On Mon, 10 Jan 2022 12:35:17 +0100
Attila Kinali attila@kinali.ch wrote:

To give here a bit more background: psychoacoustic masking, which is the
relevant metric here, mans that we cannot discern sounds that are close to
eachother with one of them being louder than the other. Depending on who you
listen to, it's usually a sound masking another sound at a distance of 100Hz
up to 20dB to 40dB lower. Even if we account for someone with golden ears
and use 60dB, that would translate to a noise spec of -60dBc @ 100Hz offset.
That's a spec that almost all XO do fulfill. A good VCXO (40-100MHz) is
somewhere around 90-100dBc @ 100Hz.
Any OCXO will fulfill that spec too, even the tiny DIL-14 ones (most are at
-110-140dBc @100Hz @10MHz).

And this doesn't take into account that we are arguing about audio frequency
specs at HF frequencies. I.e. if we use the 10MHz clock and use it to derive
a sampling clock for an ADC to sample a 20kHz signal, the noise performance
improves by another ~25dB... at least (if the design is done right, it can
be up to 50dB)

What is more important than close in noise, though, is broadband noise
performance and spurs. For someone with good ears, it's not unheard of to be
able to discern far away noise and spurs down a -100dB-120dB. Especially the
spurs can be quite hard to control, depending on what clock synthesis system
is used.

Another important spec, especially for recording, is accuracy of frequency.
An offset of just 1ppm becomes 3.6ms if you record for an hour. That's
something most people can hear already. But whether this actually elevant or
not depends on how the recording is done. The usual way is to have a central
master clock that feeds all clocked devices, such that all of them have the
same notion of time/frequency. In that case, quite high frequency deviations
can be tolerated, way beyond what a simple XO would deliver.

			Attila Kinali

--
In science if you know what you are doing you should not be doing it.
In engineering if you do not know what you are doing you should not be doing
it.
-- Richard W. Hamming, The Art of Doing Science and Engineering

time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an
email to time-nuts-leave@lists.febo.com To unsubscribe, go to and follow the
instructions there.

On 1/10/22 12:26 AM, Norman Reitz via time-nuts wrote:

I assume you've tried the usual suspects like Vectron (part of
Microchip, now), Bliley, Wenzel, Q-Tech, Abracon, MTI-Milliren.

You might check to make sure you're not edging close to the limits in
the US Munitions List (ITAR) - that will tend to make things harder -
even if it's not space qualified or for radar: (15) Space-qualified
oscillator for radar in paragraph (a) of this category with phase noise
less than −120 dBc/Hz + (20 log10(RF) (in GHz)) measured at 2 KHz* RF
(in GHz) from carrier;

Most of these places only sell through distribution, not through
Mouser/Digikey/Farnell kinds of paths. But small quantities should be no
problem (I've bought single or 3-5 oscillators at one time from most of
these mfrs, and gotten quotes from all of them). Delivery time will be
long (months, if not a year).

Just curious: Do you, from your customer interactions, have the
impression that your customers are equipped to verify those performance
figures? Or do they essentially have to trust you?

Cheers
Stefan

Am 10.01.2022 um 16:12 schrieb Bernd Neubig:

On 1/10/22 7:12 AM, Bernd Neubig wrote:

And I suppose this is why it's worth talking to the mfr than looking
through the catalogs. There might well be some key requirement that if
relaxed slightly would work out quite well in terms of availability.

We run into this all the time in the space business - someone does a
structured requirements flowdown, allocating design margin to subunits,
and winds up with a performance spec that is difficult to meet, and
nobody wants to go back up the flowdown chain and ask if the requirement
can be changed.  Indeed, the cost of doing the waiver might be more than
just buying the unnecessarily expensive part.

Hi,

On 2022-01-10 15:41, Attila Kinali wrote:

A classical paper on jitter for audio is written by late Julian Dunn:

http://www.nanophon.com/audio/jitter92.pdf

He wrote a range of good papers which is worth digging up.

AES-3 (also known as AES/EBU) is the professional version of the digital
audio interface that in the consumer version is called S/P-DIF. Much of
the reasoning for AES-3 applies directly onto S/P-DIF.

In professional audio we tolerate +/- 10 ppm for larger productions and
for all my experience, the exact frequency have not been as much trouble
as that of different frequencies and the degradation that comes from
slips, resync or Sample Rate Conversion (SRC). It has been interesting
to teach the TV and radio techs about basic synchronization and how
doing it well once save money, time and quality. There is a few war
stories to be told.

Cheers,
Magnus

Hoi Bernd,

On Mon, 10 Jan 2022 16:12:44 +0100
"Bernd Neubig" BNeubig@t-online.de wrote:

This argument is still bogus.

Sound localization works on the difference in arrival time at the ears.
We are talking here about dozens to hundres of µs of time differences.
The average human can discern time differences in the order of 10µs
(in terms of a localization setting, other settings have a different threshold).
Let's assume we have someone who is especially trained at this and
can discern time differences down at 1µs. Now let's further assume
for jitter to have little influence it needs to be at least a factor
of 1000 lower... Then we'd be at a jitter spec of 1ns at a bandwidth
of, let's be generous, 1Hz to 20kHz. Even a lowly XO will fulfill
this requirement. And it would be still broadband white noise that
would dominate this jitter, not the close in noise.

Besides, we know from experiments [1,2] that the noise levels can
be pretty substantial (like close to 0dB SNR) before localization
performance degrades noticably.

But yes, the customer is king and if he pays, why not deliver what he wants.

		Attila Kinali

[1] "Sound localization in noise: The effect of signal-to-noise ratio",
by Good and Gilkey, 1996
https://doi.org/10.1121/1.415233

[2] "Sound localization in noise in normal-hearing listeners",
Lorenzi, Gatehouse and Lever, 1999
https://doi.org/10.1121/1.426719

--
In science if you know what you are doing you should not be doing it.
In engineering if you do not know what you are doing you should not be doing it.
-- Richard W. Hamming, The Art of Doing Science and Engineering

Hi Stefan,
Bernd said it allready. If you once notice the result of a "better" clock with lower phase noise, you dont want to miss it anymore. Even if there is a lot of vodoo and pseudo-sciences in the audio sector - and audio-nuts can talk  hours about things no scientist ever heard about. But the improvement of phase noise is not an imagination. We are not talking about making something just "work". 
In worst case, phase noise can lead to bit errors. better/lower phase noise clocks will produce a better musical flow, better dynamics, better representation of details and details that were notbeeing noticed before. And don't forget - the quality of the power supply is not entirely unimportant ;-)
kind regards
Norman
Am Montag, 10. Januar 2022, 15:52:59 MEZ hat Stefan Heinzmann stefan_heinzmann@gmx.de Folgendes geschrieben:

I'm curious, too, what kind of audio application requires this level of
phase noise performance, and why.

Cheers
Stefan

Am 10.01.2022 um 12:35 schrieb Attila Kinali:

time-nuts mailing list -- time-nuts@lists.febo.com -- To unsubscribe send an email to time-nuts-leave@lists.febo.com
To unsubscribe, go to and follow the instructions there.

Hi

An equally important part of this:

What are you driving with this OCXO and what is it’s measured noise floor
at 1 Hz (or 10 Hz or what ever ….). In some cases a “crazy” OCXO is actually
quieter than the device it is driving. That means that the last 5 or 10 db in phase
noise improvement really has zero impact on the system performance. I’ve run
into this a lot of times over the years.

Bob

Hi Norman,

I was hoping for an answer that was more quantitative. We're engineers,
right? For us, the question is not so much "whether?" but "how much?",
and particularly "how much do I need for how much of an effect?".

I am especially skeptical of your assertions because you don't seem to
address at all how the frequency spectrum of the phase noise impacts the
effect. If phase noise does have an effect, I would expect to see a
strong dependency here, and Attila has hinted at the reason why this is.
Yet you mention phase noise specs very close to the carrier (so close
that any artifacts are bound to mix into the subsonic range!), which by
all educated reasoning should be the least relevant.

So, again, what "hard facts and numbers" do you have available that
would allow us to estimate how much degradation is to be expected from a
certain performance level of the oscillator, and in what application?

And, please, don't assume we're rookies here! The hint that the power
supply can be important is quite unnecessary and can very easily be
interpreted as offensive. This is not an audiophile mailing list!

Cheers
Stefan

P.S.: By the way, having brought up the topic, you could bolster your
reputation by showing how phase noise relates to bit error rate
(quantitatively!).

Am 10.01.2022 um 18:07 schrieb Norman Reitz via time-nuts:

On 1/10/22 9:56 AM, Bob kb8tq wrote:

This comes up a lot with ADCs.. Wideband driver amplifiers on the clock
inputs may put more noise on the digitized signal.

An-756 from Analog Devices

Sampled Systems and the Effects of Clock Phase Noise and Jitter

https://www.analog.com/media/en/technical-documentation/application-notes/AN-756.pdf

and
Clocking the RF ADC: Should you worry  about jitter or phase noise?

https://www.ti.com/lit/an/slyt705/slyt705.pdf

On Mon, 10 Jan 2022 19:19:09 +0100
Stefan Heinzmann stefan_heinzmann@gmx.de wrote:

A word of caution here: Human hearing is very very sensitive to some
things. One of those is two-tone discrimination. We can tell two
simultaneous tones appart quite easily, even if they are very close,
given they are at about the same loudness. 10Hz difference is nothing
and everyone can do that, even without training. Only once you get to
around 1Hz difference, will you need to get a musician or someone with
an equally trained ear. And it isn't the limit yet of what we found that
people, with training, can hear.

The reason why I am sceptical of the phase noise specs is because
artifacts would end up at much lower amplitudes. Which means the
signal that causes the artifact would be so much louder than the
artifact, that it would mask the artifact.

			Attila Kinali

--
In science if you know what you are doing you should not be doing it.
In engineering if you do not know what you are doing you should not be doing it.
-- Richard W. Hamming, The Art of Doing Science and Engineering

Hi

As noted in those app notes, there is a lot more to what goes on
with an ADC than a single spot noise spec would cover. Those
grubby details do indeed impact what sort of signal source makes
sense and how it limits system performance.

Sine to square conversion is always a bit messy. It is unlikely one
will find a < -120 dbc / Hz @ 1 Hz 10 MHz OCXO with a square
wave output. Like it or not, we live in a digital world. Simply converting
the sine to square and preserving that sort of noise level is going to
be a challenge.

My only concern here is that before one goes of and spends 5,10, or
20 thousand dollars on an OCXO, some of these details get talked
through. Also doing this sort of design without a bench set up to measure
these noise levels accurately is very risky. When you get to these
levels it becomes a “measure everything” sort of process.

How good is an OCXO spec’d at < -120 dbc likely to need to be? If
it’s a production part, they should be coming in at -123 to -126 to make
it a buildable spec. Any closer than that and yield goes to zero when
this or that very minor issue comes up. We love to look at plots showing
120.01 dbc and say “that does it”. In a production environment, not so
much. You need some margin.

If you can’t do a design that has some margin, then it becomes a selection
process. You build a thousand and test them all. Out come three pieces. The
cost of testing the other 997 all has to go into the price of those three. If
there is no market for the other 997, then the cost of throwing them away also
gets loaded into the 3 good ones. If it takes two years to find those three,
the delivery will likely be overdue ….

When you do find the magic 3, what happens if you wait a week / month
and retest them? Hmmm ….. maybe we shouldn’t have done that …. yikes ….
Getting accurate repeatable data at 1Hz offset is a very long and drawn out
process. Again a reason for needing that margin …..

Bob

Bob kb8tq writes:

Bob, you are kind of missing the point here...

The entire point of Audio-Homoepathy is to spend excessive amounts
of money, so that you can brag about how rich you are, and how
super-human your refined sense of hearing is, all without wasting
any time or effort on it.

How else can there be a market for $3000 CAT-6 ethernet cables ?

It is Audio-Homoepathy, because the crucial feature is "orders of magnitude".

They do not buy a slightly expensive ethernet cable, they buy one
which cost several orders of magnitude more than what it is worth,
so that nobody can doubt that money is no issue - and thereby
make the money the only real point.

I know several people who are quite comfortable parting these
newly-rich from their money, and I cannot really fault them...

One of those companies have a business model where they hand-build
quite competent amplifiers using outrageous raw materials and hawk
them at insane prices using industry-strength flim-flam.

When I say "indystry-strength flim-flam" I mean it.

At one time they had a role of teflon foil stored next to a resarch
nuclear reactor for some weeks, so that the "neutrons could equalize
the the tension in the micro-grid structure" before it got rolled
into capacitors with a gold foil with similar super-natural properties.

Often the invariably "very serious buyer" will persuade them, in
return for a stiff compensation, to never build any more, and refuse
to ever talk abut it, so he can (also) brag about having the only
one ever made, and make up some flim-flam about why that is so
("... but that reactor is closed now, and the guy who knew how to
reorient the rolls is dead.")

Then they rinse & repeat, this time with capacitors wound by virgins
from some tropical island or whatever.

The fact that the socalled "NFT" market has negatively impacted the
"high end audio" market recently tells you everything you need to
know about both of them.

So the question we are really being asked here, is what which
OCXO can be flim-flam'ed the most, when it becomes another dose
of audio-homoepathy.

Poul-Henning

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

Hi

Ok, well, the boys at Rakon / CEPE will be happy to sell you a ~ $30,000 OCXO
with a lead time of about 2 years. I suspect you will have to pay at time of order
unless you are a well known business …. Unlike a lot of this and that you see
tossed around for specs, I’m quite sure they do meet their published numbers.

The only thing I’m not 100% sure of is if the minimum order quantity is 1,3,5 or
10. If cost is no object, that should not really matter.

Bob