Innuos ZENith "STATEMENT"

[barrows]

On 4/15/2018 at 5:50 AM, rafa said:

Sotm never stated all parameters so hard to say which is better.

 

3PPB clock stability is very very good. But still, it doesn't mean it will be the best audio server.

Hahahaha!  3 pp/B is long term clock stability and means nothing for performance of this device.  Long term clock stability is important if you want your clock to be accurate over a period of months or years (as for an atomic clock to tell time).  Since audio listening is a short term thing, it means nothing for audio performance and is pure marketing.

[barrows]

On 5/23/2018 at 12:23 PM, Blackmorec said:

Whether a clock has good long term stability in a DAC is irrelevant, unless you’re using your DAC to wake you up every morning. The fact that you switch it off every day renders long term accuracy moot. What’s important it that when its switched on it defines the sampling rate accurately, which means how accurately does it produce its designated frequency. 

Not really.  All clocks used for audio have good enough accuracy to to be running at the correct speed for audio reproduction with proper pitch (this would be analogous to turntable speed accuracy).  For audio playback accuracy there is no difference between a 2 ppb XO or a 50 ppb XO, they are all good enough to produce perfect pitch accuracy well beyond what is audible. What is important for audio clocks is the short term variations which give rise to jitter, specifically the phase noise at low frequencies.  Now, it may be true that some 2 ppb XOs also have excellent phase noise characteristics, but it may not be as well.  It is true that some OCXO Also have excellent phase noise characteristics (like the Pulsar clock and the NDK DuCoLon clock) but it not necessarily true just because the XO is oven stabilized.

Interestingly, it is also true that the use of low phase noise clocks improve the performance of things like Ethernet, hence the use of low phase noise XOs in Sonore's best Ethernet Renderers.

It maybe that Innuous' "Statement" uses an XO which does have low phase noise, but telling us the ppb gives no indication of this.  If Innuous wants to give a meaningful specification they should give the phase noise plot of their OCXO, or at least the phase noise at 10 HZ.

 

And for heavens sake, if you are concerned about clock performance, you are going to defeat the purpose of a high performance clock by turning it off every day.  The better the clock, the longer it takes to stabilize at its lowest phase noise.  Keep anything with a critical clock powered up 24/7 for best performance.

[barrows]

8 hours ago, Blackmorec said:

My original post addressed your assertion that the parts per billion spec for an OCXO clock is a long term stability spec, which it isn’t. Its a short term accuracy spec. 

Regarding oscillator clocks, jitter is a measure of variations in the time domain

Phase noise is the resulting product of jitter in the frequency domain. The more time variation, the more frequency variation, hence the spectrum of phase noise. 

If you look at all the major oscillator manufacturers and resellers, they all spec their clock’s accuracy in ppm or ppb.  The lower the number, the more accurate the clock (the closer it comes to always producing the desired frequency) the less jitter and resulting phase noise.

 

I have several devices with high end clocks. They all benefit from a warm-up but there’s no real SQ benefit that I can hear from leaving them on 24/7 

Generally speaking, in terms of sound quality, a 30 minute warm up with signal flow is far more effective than leaving a device idling 24/7, as the idling device still requires the 30 minutes with signal to sound its best.  The only exception I’ve found was with my Naim Active System, which seemed to need several days to reach optimum performance and did indeed benefit from remaining on 24/7. And that system had no oscillators; until I added a CDS. 

I suspect you might learn something from this:

 

http://www.ndk.com/en/ad/2013/001/index.html

 

ppm/ppb indeed is a measure of long term clock accuracy (and clock frequency accuracy in general as you suggest), but is has no correlation to what matters for our purposes: clock phase noise, especially at low frequencies.  The NDK Ducolon is one of the best OCXO for audio purposes (but not necessarily for a server, as the frequencies available are the audio frequencies) although the ppm/ppb is not all that impressive, the low frequency phase noise is.

I am not trying to suggest the clock in the Innuous is bad at all, just that we do not know much of anything unless Innuopus publishes a phase noise curve for it.  It does make me wonder though, that they only specify the ppm/ppb spec...

[barrows]

@Blackmorec, read your link.  Phase noise is the specification which matters for our purposes here.  Phase noise is the frequencies present outside the nominal operating frequency of the XO.  All XOs used for audio purposes, and for clocking of computers, interfaces, etc, have enough frequency accuracy for our purposes: ppm/ppb.  

Especially what we would want to know is the Phase Noise at low frequencies: 10 Hz and below, so the best way to evaluate XO performance for our purposes is to look at the Phase Noise plot.

There is no need to be concerned about the ultimate frequency accuracy, as all the XOs are good enough.  A little bit slow XO, or a little bit fast XO is not going to make a difference.  Indeed I would rather spec this XO: Phase Noise at 10 Hz= -115 dBc/Hz and 50 ppm, than this one: Phase Noise at 10 HZ= -90 dBc/Hz and 25 ppm.

 

Another example: if we have an XO that is nominally 45.1584 MHz, and it is actually running at 45.1583 MHz it makes no difference to us, but if we can get an XO with few dB better Phase Noise at low frequencies, that is something worth doing.

 

This is also why "Atomic" clocks are generally overkill and not an advantage for audio purposes.  Great for super precise measurements of time over the longer term, but we are not perceiving musical notes over a period of years.