Mutec REF 10 Masterclock

[modmix]

Hi Thomas,

 

I am testing and measuring hardware and software in my system as some of us

I've done some tests using various 10 MHz OCXOs attached to my MC-3+.

 

REF 10 phase noise:   1 Hz:               ≤  -116 dBc/Hz
None of the OCXOs tested so far has such a low phase noise @ 1Hz; this value is important for good sound quality according to my findings.

My best OCXO (KS-24361) has according to spec a 2-3 times higher temporal jitter - keen to have REF 10 in my gear

 

Ulli

 

[modmix]

Let's do it step-by-step:

• Vivaldi Master Clock:
  External Reference Input on 1x 75Ω BNC connector. Accepts either Word Clock or AC coupled signals at 1MHz, 5MHz & 10MHz. Lock range is +/-300ppm
   
• Mutec REF 10:
  Provides simultaneous reference outputs with 50 and 75 Ω impedance
  galvanically isolated, individually switchable BNC clock outputs

=> REF 10 can be used @ Vivaldi Master Clock.

 

The rational for that would be to have a more stable reference to generate the digital audio clock frequencies  (44.1, 48, 88.2, 96, 176.4 or 192kHz) - depends on the quality of the Vivaldi internal clock, the capabilty of the Vivaldi clock to track the external reference etc.

 

Sound quality increases in case a dCS Scarlatti is give a quite good 10 MHz reference (KS-24361).

 

Might be the same with a better reference (REF 10) feed into a Vivaldi Master Clock...

Should be worth a trial, if you can

 

Ulli

 

[modmix]

2 minutes ago, k-man said:

The Vivaldi Upsampler and DAC will only accept Clock frequencies from the Master Clock at 4.1, 48, 88.2, 96, 176.4 or 192kHz. The Master Clock itself can accept a Reference signal at 10Mhz (as well as 1Mhz and 5Mhz).

 

Since the Mutec MC3+ USB accepts the 10Mhz signal, Mutec's main goal is to 'improve' the performance of that unit with the REF10.

Oh, I'm not so familiar with dCS...

The RME AES card also only accepts audio clock signals as an external reference, too.

A MC-3+ controlled by a KS-24361 10MHz used as an external reference at this card did clearly improve SQ.

Only draw back: as the external rate must match the sample rate of the track being played, some adjustment at the MC-3+ is required in case the next track uses a different rate (having some remote feature to adjust MC-3+'s sample rate would be fine in this case).

[modmix]

1 hour ago, svart-hvitt said:

 

cannot and will not

For being alive for several years now, it is proven that I "cannot and will not" die

Sorry - couldn't resist...

 

Hugh is right in one aspect: the circuit adjusting the internally used clock to an external reference may deteriorate the external signal.

 

A not so good external reference might easily prove lower quality sound quality compared to the internal clock.

 

The (discontinued) Antelope 10M rubidium atomic clock make use of a module rated in the respective manual as "designed for low cost mass production." It is a good exampel of what I mean with "so good external reference".

 

Attached to a MC-3+, Antelope 10M did change sound quality by far less than a better RB clock.

 

Rb clocks are made to get better long term stability. For audio short term stability is much more improtant. Many OCXO have much better short term stabilty compare with almost any Rb clock.

 

hth

Ulli

 

PS:
You may also read Paul McGowan's remark on Atomic clocks:

[modmix]

German aktives-hoeren has a picture showing REF 10 inside  and a report:

" Durch die immer wieder überarbeitete Stromversorgung --- habe man diesem bessere Rauschwerte entlocken können, als sie vom Hersteller des Oszillators spezifiziert seien."  source

My translation: phase noise is lower than manufacture spec due to several times improved power supply.

 

Modification of this section quite likely will ruin excellent phase noise behaviour, I guess.

 

Ulli

 

 

[modmix]

From some tests since 2014 using various 10 MHz source, I too came to the conclusion that Jrms close to the carrier is one of the best figures to determine sonic quality of such a device used as reference to a MC-3+.

 

From Cybershaft rubidium clock...low priced option?:

On 28.9.2016 at 8:18 PM, Elberoth said:

Here is the revised clock graph from the first page. I have also added the Sforzato PMC-01 BVA clock, which is the only clock in the world based on the Oscilloquartz BVA-8607 OCO module:

 

 

The numbers used in this table, are based on minimum manufacturer guaranteed values. The actual samples may exceed those numbers.

 

REF 10 gives just 7 fs (10-100Hz) - using phase noise figures given above in this thread and http://www.abracon.com/phaseNoiseCalculator.php.

 

One has also take into account the signal form (eg. sinusoidal, squared) for one simple reason: the receiver has to detect at which point in time the signal did pass the threshold region. A faster transition gives less jitter added by the input stage. From my experience, a factor of 3 is a good number to compare these types of signals.

 

The devices listed above give sinusoidal signals, too my best knowledge.

REF 10 has a square wave output signal, => < 2 fs (10-100Hz), if my experience applies here, too.

 

hth

Ulli

 

 

[modmix]

1 hour ago, Confused said:

My understanding is that AES protocol includes a clock signal.

Yes - implicitly.

Data are encoded prior to transmission in a so-called Biphase Mark Code format (more e.g. at http://scanlime.org/2011/04/spdif-digital-audio-on-a-microcontroller/).

 

A DAC has to extract the embeded clock => SPDIF decoder's job.

from: http://peufeu.free.fr/audio/extremist_dac/spdif.html

 

1 hour ago, Confused said:

Will the DAC's internal clock benefit from this feed and effectively use the super accurate clock in this feed?

In case the DAC just uses the recovered clock, of course having a more precise clock will help (given there is a good SPDIF decoder involved).

 

1 hour ago, Confused said:

Or does the DAC's internal clock (presumably less accurate) basically take precedence in the end, hence nullifying the benefit of the REF 10 & Mutec MC3?

There are concepts like this where clock delivered over the cable is used solely to corse adjust an internal clock (note that the source clock will never be exact 44.1kHz or so; the receiver must adajust to the real sample rate in order to avoid droped or missing samples).

from: http://www.latentlaboratories.com/blog/2015/2/2/dsp-01-part-7-totally-working-spdif-audio

In such a case adding a REF 10 might have less effect.

 

hth

Ulli

[modmix]

7 hours ago, Elberoth said:

The figure to look at, is the phase noise below 1Hz.

This is stated from time to time by various people, yes.

No rational seen so far, though...

 

From my unterstanding, phase noise below 1 Hz (e.g. 0.1 Hz) means time error takes place at a rate of e.g. 10 sec - correct?

 

When comparing various 10 MHz clocks the sonic impact could be easily heard at the attack of a single tone played on a piano - definitely not lasting 10 sec. ,-)

 

I'm of the impression that having a low noise figure below 1 Hz normally coincide with low figures at frequencies above 1 Hz - bad idea?

[modmix]

41 minutes ago, svart-hvitt said:

Seems like the MSB 33 Femto (i.e. 33 femtoseconds for the entire frequency range) is a bit ahead of the Ref10, does it?

Reminds me of playing Autoquartett when I was a child - more speed, more power and the hack.

But - let's play it ,-)

And, I guess, we do talk about electrical properties and leave money out of the game...

Luckily we can simply compare what the respective manufacture gives on their respective page:

  REF 10    femto 33       freq.range
  28.8fs           31.8fs             1-100
    7.3fs            7.8fs            10-100 

  51.4fs          42.3fs       1-100,000
  43.2fs          28.9fs     10-100,000
unknown         1ps      0.1-100,000

 

Looks like we do have two winners - depending on the frequency range

From some years of soundwise comparing OCXOs I do have an idea which frequency range is of more importance when it comes to sq.

BTW: is there a stand alone Femto 33 (including case, psu, clock distribution and the like)?
If so, what kind of output signal does it deliver?

[modmix]

The goal is to have as little jitter as possible when the digital-to-analog conversion is performed.

 

If neither the source nor the transmission adds a significant amount of jitter or, if the DAC has a good re-clocking implemented, adding a MC-3+ makes little or no sense at all.

 

In case the DAC uses it own clock, the clock must be adjusted to the actual sample rate (in order not to loose samples or to create drop-outs). Some DAC internal circuit (PLL or the like) will perform that based on the input data stream. Having an external wclock could be useful (but that has also some drawbacks in some cases).

 

In short: no simple answer for these "simple" questions - sorry.

 

Ulli

 

[modmix]

Don't know the dacs mentioned by you.

Wclock at the dac is not a "conditio sine qua non" - you might try...

 

[modmix]

41 minutes ago, barrows said:

I am pretty sure it is a 50 ohm impedance device, as are most clock inputs...

Mutec's MC-3+ clock input is not like most other inputs - they are 75 ohm

https://www.mutec-net.com/product_mc-3-plus.php?lng=en#data

[modmix]

Thanks for sharing these quite impressive figures.

 

ADEV (Allan Deviation) is a kind of standard deviation. It gives the average time errors when measured every τ seconds.

Allan Deviation Explained has a quite instructive diagram:

 

So, we now know how stable the REF 10 is when looking at it ever τ = 1 sec.

With respect to sample rate (or music) 1 sec is quite a huge interval.

 

Allan deviation is not just a single figure. Over τ you usaually find something like this:

source: municom Fachartikel zum Thema Quarze mit geringem Phasenrauschen

More information can be found eg. here https://www.nist.gov/pml/time-and-frequency-division/popular-links/time-frequency-z)

 

Question:

What is ADEV for smaller τ values? eg. 1 msec, 10 msec...

A plot would be great.

 

Ulli

 

BTW:

Good old HP10811 ADEV for comparison:

• http://z21.vfdb.org/Archiv/pdf-Dateien/H.Schmiedel_Frequenznormal.pdf p.20: 5e-12
  
• http://www.ke5fx.com/tpll.htm 8.6e-13