Superclocks

[Elberoth]

On 22.04.2017 at 10:48 PM, Superdad said:

 

Reasonable points and questions: Yes, the oscillator manufacturers are measuring their clocks--and the performance of those is of course not at all the same as the jitter performance of a complete DAC.

 

I will tell one interesting story:

Crystek has gotten a LOT better making their XOs!  18 months ago John and I requested and received samples of their CCHD-575 (the one Jonathan linked to above).  Got 25.0MHz version (for testing in the ISO REGEN).  They sent us 3 samples, and they each came with individual phase-noise plots.

Whereas the data sheet on Crytek's web page shows -100dB for 10Hz offset (in the mid-20MHz f-range we are using), the plots for the 3 samples we got were -108dB, -110dB, and -112dB at 10Hz!   

Of course our first thought was that they hand-selected these and that production runs could not be anywhere near as good as these.  So I immediately picked up the phone and called Crystek (they are actually not that large a company; one sales engineer seems to handle the whole country).  They said no, the samples were not at all specially selected--just measured right off the line. And he said that their processes have improved so much that the $9.60 CCHD-575 (that's what I paid for the first 500 pieces; goes to $9.30 ea. @ 1,000) now outperforms their big $27 CCHD-957 as used in a lot of top-tier DACs.   I told him they ought to update the 575's datasheet on the web to reflect that.  Guess he did not relay that suggestion to their marketing department. 

 

Of course if you want to see eye-popping phase-noise performance, then check out the Pulsar Clock. -118dB to -123dB at 10Hz!  But it is 420 Euros...

 

What makes you think that the processes improvments he is talking about did not result in better real world numbers for the CCHD-957 as well ?

[Elberoth]

On 27.04.2017 at 3:20 AM, JohnSwenson said:

Rubidium clocks are usually very bad to use for audio. They have very good long term stability, but high phase noise. The long term stability has nothing to do with audio but the close in phase noise is what is important. So a rubidium is exactly the wrong oscillator to use.

 

Another problem is that the rubidium is probably NOT going to be outputting a frequency that can be used directly by audio circuitry, so some for of frequency synthesizer is going to have to be used, and these ALWAYS increase the phase noise.

 

A rubidium is great for an actual clock (you can read the time) that you want to be accurate to the microsecond over years of run time, but not so good for audio.

 

John S.

+1

 

I recommend everyone here to read this 2 pages long essey on clocking:

 

http://www.grimmaudio.com/site/assets/files/1088/picoseconds_or_ppm.pdf

 

It is very easy to read for non techies like us and explains why crystals are better than Rubidium Clocks for audio and why long term stability (in ppm) does not metter one bit.

[Elberoth]

9 hours ago, Beolab said:

 

This is very interesting to discuss  Alex, and i got two topics for you: 

 

If you elaborate a little with your knowlage and experience on how much better is the  acclaimed 26Nm 1ppm MSB Galaxy clock vs let say the Pulsar / Crystek CCHD595 clocks, and what is the fabrication cost of the Galaxy clock ? Is it the best of the best or could a company like Texas Instruments or Intel make much better for 1/100 of the price? 

 

 

Secondly, what is your thaughts about using a poorer streamer like a BlueSound Node II MQA using two W4s Remedy ( CCHD957 ) SPDIF clocks in a chain with LPS-1 Ultracap vs the top tier streamers like Auralic Aries , Aurrender W20, Antipodes , Sonorus, do you think it is the best result in the end? 

 

My experience is that the Bluesound + 2x clocks is on pair with my other top tier streamers, but i want to know your input on it. 

 

Have an great evening!  

 

/ Fredrik

 

 

What do you mean by 26Nm ?

 

Nm is Newton Metre - unit of torque

 

MSB used to call their clocks by jitter numbers in fs (Femtosecond) - chence models 140, 77 and 33. 

 

It is impossible to compare the MSB jitter numbers to any other oscilators, as for jitter to have any meaning you need to know the bandwith of the measurement.  

 

MSB used to give the phase noise numbers of their clocks on their webpage, which would make the comparo possible, but I cannot find it right now.  

 

[Elberoth]

40 minutes ago, Beolab said:

 

=Nano meter , but forget it 

 

Here you got the Specs in a PDF 

 

http://www.msbtech.com/support/GalaxyClock_Lit_Press3.pdf

 

Thanks, that was the page I was looking for. The specs for their clock are as follows:

 

FemtoSecond Galaxy Clock Specifications:

Phase Noise at 0.1 Hz -67 db

Phase Noise at 1 Hz -99 db

Phase Noise at 10 Hz -134 db

Phase Noise at 1 kHz -157 db

Phase Noise at 10 kHz* -157 db

Phase Noise at 100 kHz -157 db 

 

If you get me the phase noise figures at the same freq for other oscillator (Crystek or other) I will calculate the jitter for you to make the comparo possible.  

[Elberoth]

 

The plot from MSB is indeed measured on the analog output of the DAC. However, the specific numbers published (the ones I quoted), must have been obtained using a phase noise meter.  

[Elberoth]

40 minutes ago, bibo01 said:

It doesn't the carrier though.

The numbers for Pulsar clock you can find them here:

http://www.pulsarclock.com/ds/Pulsar_Clock.pdf

 

My guess is it is either 44.1 or 48 x 512

[Elberoth]

25 minutes ago, JohnSwenson said:

Be very careful with these numbers, they are not directly comparable to any other clock phase noise numbers. Note the numbers are in dB, NOT dBc/Hz which is the standard for phase noise measurements. Looking at that page from MSB it looks like those numbers are from an FFT. Which means this is a sampling of some waveform with some clock. Exactly what is undefined. Did they use the Galaxy clock to sample the analog wave from a DAC? If so is the Galaxy clock also used in the DAC? These unknowns make it very difficult to convert these numbers into meaningful dBc/Hz numbers that can be compared to other clocks.

 

John S.

 

Good point. I have assumed by default that those numbers are in dBc/Hz, which as you said is the standard for phase noise measurements.