CLOCKS, what should we look for in next generation

[zephyr24069]

20 hours ago, mozes said:

I just wanted to clarify that master clocks ca  be beneficial to SQ. I don't know why or how. This is based on my subjective listening only.

I have the same experience over the last 10+ years; my playback experience has been enhanced by good quality external clocks with the right spec-compliant and well-built cables and attaching to equipment designed for use with external clocks (also properly built, good power supplies, etc...).

[zephyr24069]

1 hour ago, jabbr said:

 

I can't easily tell from your profile which stand alone DAC you are using (because this matters). Esoteric seems to be designed for external clocking and it is very reasonable that equipment explicitly designed for external clocking would perform better.

 

For example RME ADI-2 FS does not provide for an external clock.

MSB provides internal upgrade: http://www.msbtech.com/support/GalaxyClock_Lit_Press3.pdf ... I like the provision of specs including 0.1 Hz offset -- note that these specs are now achievable at a lower price point given the increased affordability of great clocks -- they also show the base of peak widening effect of close in phase noise. MSB has enough confidence in their onboard clock that they provide a clock output for sync but not input.  

 

Again, it all depends on your system

Agreed...it definitely matters.  From 2006-2007 through 2015, I had an Esoteric P-03U with D-03 DAC at first with the G-0s Rubidium master clock for a couple years, then sold the G-0s and went without a clock for 1-2 years slaving the P-03U to the D-03's internal clock (176Mhz output), then the G-03x OCXO clock for a few years through 2015 when I upgraded to the Esoteric P-02 Transport and D-02 DAC.  I sold off the G-03x and used the D-02's internal clock as master with the P-02 slaved to it finding that the D-02's 176Mhz output was good but it's "MCLK OUT" frequency at 22.xxx Mhz was my preferred setting.  Then a couple years ago after running through a number of clocking options (external clocks) including Esoteric's own G-0Rb and G-01 as options and thinking about SRS and others, I ran across Cybershaft as an option from a search for master clocks on HiShark.  Been using various Cybershaft clocks every since.

 

My DAC and Transport today are the still the Esoteric D-02 and P-02 respectively. Later this year I will sell the D-02 and step up to dual Esoteric D-01s (not the new X generation) and eventually upgrade the P-02 with a P-01 transport.  All boxes since the mid-2000s that I have utilized are designed to be used with external clocks as a reported upgrade option.

 

As an aside, I've hung out with friends that have the Esoteric Grandioso K-01/K-01x and the later Grandioso K1 (a fantastic 1-box as is the K-01X). All these seem to benefit in musical playback quality in one way or another from using a good quality external clock, properly executed. The key presumption in all setups I've used since the mid-2000s or that friends have gone with is that a key element is the overall quality of the external clock itself, not just the specs on the clock module, the power supply, etc....as well as the components it is to clock. Another critical element is the use of proper 50-ohm or 75-ohm clock cables.  

 

Thanks to all who have put forth some great replies on this thread historically and in the last few days,...I appreciate the chance to learn more from many different points of view and of course all the technical facts that have been contributed!

[jabbr]

41 minutes ago, zephyr24069 said:

From 2006-2007 through 2015, I had an Esoteric P-03U with D-03 DAC at first with the G-0s Rubidium master clock for a couple years, then sold the G-0s and went without a clock for 1-2 years slaving the P-03U to the D-03's internal clock (176Mhz output), then the G-03x OCXO clock for a few years through 2015 when I upgraded to the Esoteric P-02 Transport and D-02 DAC.

 

In the last few years, while the physics of crystal oscillators and rules of electronics have remained remarkably stable  , the cost of high quality, so-called femtosecond clocks, has, well, plummeted, and an $700-800 external clock needs to compete with a $10-20 internal clock. No doubt you have an affinity for Esoteric products, and Esoteric admits up front that an external clock can beat their internally provided clocks. As long as you are happy with your system that's all that matters.

 

I'm not suggesting that anyone replace a current system that works well for them, rather that for new purchases a DAC which is well designed and with a great on-board clock should be considered. For me the analog stage and overall design remain the most important factor. Folks like @PeterSt have upgraded the analog electronics in the "G3" which is clearly audible (I don't even know if the clock has changed), and iFi iMicro BL, or the newer iFi Pro, or the RME ADI-2 Pro FS ... I haven't heard the latter two but those would be at the top of my list if I were looking for a new DAC in the $1500 range.

[zephyr24069]

2 hours ago, jabbr said:

 

In the last few years, while the physics of crystal oscillators and rules of electronics have remained remarkably stable  , the cost of high quality, so-called femtosecond clocks, has, well, plummeted, and an $700-800 external clock needs to compete with a $10-20 internal clock. No doubt you have an affinity for Esoteric products, and Esoteric admits up front that an external clock can beat their internally provided clocks. As long as you are happy with your system that's all that matters.

 

I'm not suggesting that anyone replace a current system that works well for them, rather that for new purchases a DAC which is well designed and with a great on-board clock should be considered. For me the analog stage and overall design remain the most important factor. Folks like @PeterSt have upgraded the analog electronics in the "G3" which is clearly audible (I don't even know if the clock has changed), and iFi iMicro BL, or the newer iFi Pro, or the RME ADI-2 Pro FS ... I haven't heard the latter two but those would be at the top of my list if I were looking for a new DAC in the $1500 range.

Yes I do have an affinity for Esoteric products and they have always been improved in terms of expanse and accuracy of the full 3D soundstage, pacing, etc...by use of an external clock (of the right design, properly executed, great power supplies, etc...more than just a great oscillator), their sound quality with their own onboard clock circuits is very good indeed. This is particularly true IMHO from P-02/D-02 forward.  There was a big jump in the sound quality overall in the out of the box experience with the improved clock circuit in the D-02 versus the 03 prior generation IMHO and the inclusion of the 'MCLK' frequency from an inherently better onboard clock circuit.

 

If I was a DCS fan I'd probably be similarly focused.  In the case of Esoteric, they apparently go to great pains to engineer a total component such that a good quality clock input results in improvement in the overall function of the device, far better than just using a better clock chip onboard. Whether that is happenstance or limits of physical design or their decided belief (and marketing/ability to profit by taking the path) in also designing bespoke external clocks will probably never be known exactly.

 

From what I've seen on the market and from other manufacturers' use of them (e.g. MSB, TotalDAC, others), the availability of high-quality femtosecond-grade assemblies is much better than it ever has been and the cost of including them has dropped proportionately. The results in world-class playback with these stacks and standalone components speak for themselves.

 

I am having to live without my clock since this past Monday as it's on its way to Japan for an upgrade to multiple 50-ohm outputs and an even better power supply (i.e. a step-up to "Limited v2" level with Cybershaft).  Only reason I mention this is that the change in the system after removing the clock and going back to the clock circuit set to MCLK OUT frequency on the D-02, slaving the transport to the Esoteric D-02 was extremely noticeable.  Tonally, musical involvement and 2D, the system still sounds great.  3D imaging is also still very good however the depth of the soundstage and it's extent from the stage out to and wrapping around the listening position on many recordings that are immersive like that (Roger Waters Amused To Death for example and others) is not as expansive as it was with a high-quality external 10Mhz clock generator in the system.  In short, it sounds very good and if I'd never heard otherwise I'd be very happy but having lived with 'better' I cannot wait for a good clock to get back in this system.

 

As an aside, using measurement data sent to me based upon testing at the output of the clock I currently utilize, I am quite happy with the femto-second level results of my clocking circuit. Most importantly, if I'd never seen any of this, what it does for the musical playback experience is what definitely gives me satisfaction that for what I've chosen as playback gear, this is one of the ways (along with good power cords, interconnects, power conditioning, etc...) to get the most out my playback experience;

 

 

*Reference:  https://abracon.com/phaseNoiseCalculator.php. and https://www.maximintegrated.com/en/app-notes/index.mvp/id/3359 "Relation between RMS Period Jitter and Phase Noise"

 

FWIW....thanks for a great discussion!  If you ever want to see the entire system as it exists today and many discussions about the journey thus far, look up my userID here (Zephyr24069) on Audiogon and you'll find a virtual system posted there.

 

 

[barrows]

iFi PRO looks to have some engineering problems, likely caused by trying to offer so many features at a low price:

 

https://www.stereophile.com/content/ifi-audio-pro-idsd-da-processorheadphone-amplifier-measurements

 

Evaluating jitter and/or clock performance by listening tests alone is fraught with difficulties.  Every system is different, and virtually every system has some existing "problems" which may obscure the results of listening tests.  I participated in some listening tests where we correlated various jitter levels and distributions (measured) with actual listening observations.  I came away from these tests realizing that in some systems, some listeners are definitely going to prefer the sound of the higher jitter option.  I suspect this effect is very likely responsible for "improved subjective results" some people report when applying certain external clock solutions.  As JS has mentioned, an external clock solution with extremely low phase noise, combined with a DACs which implements a "perfectly chosen" DDS device, and does perfect implementation, may be able to improve on the jitter performance of a good (say Crystek 575/957) internal clock.  But it is highly unlikely that the internal clock receiving/distribution is done perfectly.

 

I am getting great results in my DACs with both Crystek 575/957 and NDK SDA series XOs, with great deal of attention paid to power supply for the XO and good distribution techniques.  Could it be better, sure the same approach with a really, really good (local) OCXO could be better technically, would that difference be audible... that is the question.

 

On other trick one can do, is they can measure and hand select clocks for lower phase noise at low frequencies-the clocks referred to above all vary by as much as +/- 5 dBc/Hz.  If one has the ability to measure for phase noise, and can afford to discard (or use for less demanding circuits) a handful of XOs, you can achieve better performance.  Unfortunately, the tools for measuring phase noise are prohibitively expensive for most.  For a project awhile back, Sonore got a small batch of hand selected NDK oscillators-the few customers who received these got a very special upgrade indeed!  Unfortunately I do not believe the source who made those selections is offering the service anymore.

[Superdad]

48 minutes ago, barrows said:

For a project awhile back, Sonore got a small batch of hand selected NDK oscillators-the few customers who received these got a very special upgrade indeed!  Unfortunately I do not believe the source who made those selections is offering the service anymore.

 

I assume you mean Patrick in Texas.  Have not spoken to him in at least a year.  He can be a bit ornery.  

Last we spoke I suggested he go ahead an measure some Crystek CCHD-575 to verify for himself my claim that they provided us with 3 random production samples of 25.0MHz with paper plots showing 10Hz performance of -108, -110, -112 dBc/Hz--which is vastly better than what Crystek's own published plots show.  Luiz says that their engineer is very conservative and does not want to boast about 575 performance.  I always tell him that his competitors are not shy about doing so.  I wish that John could find those original printouts that came with the samples, but am afraid they got lost in his house move.

[Superdad]

BTW, here's the plot for an eye-poppingly great OCXO we have been looking at.  Have no idea the cost yet--likely $600 a piece.  But check it out: -110dBc/Hz at 1Hz, -143 at 10Hz!  (this is 10MHz part, but even in the mid-twenties audio clock range these are spectacular).

 

[barrows]

6 minutes ago, Superdad said:

BTW, here's the plot for an eye-poppingly great OCXO we have been looking at.  Have no idea the cost yet--likely $600 a piece.  But check it out: -110dBc/Hz at 1Hz, -143 at 10Hz!  (this is 10MHz part, but even in the mid-twenties audio clock range these are spectacular).

 

Yup, pretty good even for 10 MHz!  Do they have plots for audio rate OCXOs?

[austinpop]

7 minutes ago, Superdad said:

BTW, here's the plot for an eye-poppingly great OCXO we have been looking at.  Have no idea the cost yet--likely $600 a piece.  But check it out: -110dBc/Hz at 1Hz, -143 at 10Hz!  (this is 10MHz part, but even in the mid-twenties audio clock range these are spectacular).

 

 

Clock porn!

[asdf1000]

OCXXXO

[rickca]

33 minutes ago, austinpop said:

 

Clock porn!

I've heard enough about clock pics lately.

[jabbr]

2 hours ago, Superdad said:

BTW, here's the plot for an eye-poppingly great OCXO we have been looking at.  Have no idea the cost yet--likely $600 a piece.  But check it out: -110dBc/Hz at 1Hz, -143 at 10Hz!  (this is 10MHz part, but even in the mid-twenties audio clock range these are spectacular).

 

 

Those specs look reasonable at that price point. There are a few well known co that cut crystals and then others that package them eg:

https://www.vectron.com/products/ocxo/ox-204.pdf

(not sure how new this is)

 

and Wenzel of course: http://www.wenzel.com/wp-content/parts/501-26719.pdf

[PeterSt]

3 minutes ago, jabbr said:

Those specs look reasonable at that price point. There are a few well known co that cut crystals and then others that package them eg:

https://www.vectron.com/products/ocxo/ox-204.pdf

 

Maybe it is not a good idea to compare with specs of a lower base frequency with what audio requires ? (specs will worsten the higher the frequency)

Also it may not be a good idea to take synthesis for granted now (it really is not for the better as far as I can tell).

And thus it is apples and oranges and (IMO) the Vectron is not the very best. Add to this that you'd thus need audio frequencies and as long as it hasn't been made ... you can be the first and pay the $ for that. Haha.

 

Additionally, never look at oscillators which can be pulled (like this one). That's doing unnecessary no-good to specs (unless you want to make a PLL of course, but who would want that ? ).

[rickca]

2 hours ago, Superdad said:

BTW, here's the plot for an eye-poppingly great OCXO we have been looking at.  Have no idea the cost yet--likely $600 a piece.  But check it out: -110dBc/Hz at 1Hz, -143 at 10Hz!

This is almost as good as the Mutec REF 10 specs.  Do you know what clock they use?  I can't imagine Mutec is spending $600 for their clock.

[jabbr]

Right just pointing out that those specs have been available at some time for s price. 

 

The AT cut was developed ?1934 and the higher Q SC cut in 1974. Growing a crystal with high Q takes time & care. The OCXO is not important rather the SC cut/high Q crystal — and size of crystal — but generally the better crystals are used in OCXO, then there’s also the design and materials used in the oscillator circuit. 

 

In the same same way you can parallel transistors together to reduce noise, you can also tie multiple crystals together to further reduce phase noise —- but what’s the point? For a motherboard that does spread spectrum anyways? 😂 For Ethernet switch? None of this extraordinary effort is necessarily (just use a Crystek or NHK and be done w it) ... for a DAC? Perhaps if the rest of the circuitry has also been optimized. 

[jabbr]

10 minutes ago, PeterSt said:

Additionally, never look at oscillators which can be pulled (like this one). That's doing unnecessary no-good to specs (unless you want to make a PLL of course, but who would want that ? ).

 

Um to parallel several clocks together  

[Superdad]

13 minutes ago, PeterSt said:

Also it may not be a good idea to take synthesis for granted now (it really is not for the better as far as I can tell).

 

Not taking them for granted, but the better synths do not add much phase noise at all.  While for a DAC it’s best to go with native clocks, there are times when one needs several frequencies.  An example is our upcoming EtherREGEN, where we need both 25MHz (in several locations) and 250MHz, and we have two completely separate domains.  So two synths and two Crystek 575s fill the bill.

[PeterSt]

12 minutes ago, jabbr said:

Um to parallel several clocks together  

 

There is "networked clocks" for that (heard of that ? I am serious - and the company who makes them is IMO the best candidate for new ultra low jitter oscillators; I forgot the name by now, but have a quotation somewhere for only a 100 and "doable").

[PeterSt]

2 minutes ago, PeterSt said:

There is "networked clocks" for that

 

PS: With master and slave clocks. I considered that instead of a clock distribution system (see earlier post about the 2500 component DAC). I forgot why I didn't choose that. Probably something with skew. Anyway, this would be a new kind of approach (for a purpose I again forgot but instead of distribution).

[Miska]

11 hours ago, rickca said:

This is almost as good as the Mutec REF 10 specs.  Do you know what clock they use?  I can't imagine Mutec is spending $600 for their clock.

 

My question would be how good the clock performance is after feeding 10 MHz clock through DPLL to generate actual clocks needed for audio...

 

I'm hardly getting excited about any clocks anymore until I see measured proof of the final performance from DAC output...

 

[barrows]

10 minutes ago, Miska said:

 

My question would be how good the clock performance is after feeding 10 MHz clock through DPLL to generate actual clocks needed for audio...

 

I'm hardly getting excited about any clocks anymore until I see measured proof of the final performance from DAC output...

 

While I tend to agree, I would get excited if i could purchase a DIP packaged XO @45.1584 with <115 dBc/Hz @ 10 Hz, or a 27.5792 with <120 dBc/Hz @ 10 Hz for $500.00 USD or less.  Just for bragging rights and subjective testing.  I would want to receive a phase noise plot for each individual clock as well.

[Miska]

7 minutes ago, barrows said:

While I tend to agree, I would get excited if i could purchase a DIP packaged XO @45.1584 with <115 dBc/Hz @ 10 Hz, or a 27.5792 with <120 dBc/Hz @ 10 Hz for $500.00 USD or less.  Just for bragging rights and subjective testing.  I would want to receive a phase noise plot for each individual clock as well.

 

I'm quite happy with my CCHD-957's. Anyway the clock module is rarely the performance limiting factor. Especially on wider scale there are so many bigger things to work on in the D/A converter itself...

[barrows]

3 minutes ago, Miska said:

 

I'm quite happy with my CCHD-957's. Anyway the clock module is rarely the performance limiting factor. Especially on wider scale there are so many bigger things to work on in the D/A converter itself...

Like I said, I agree, mostly...  But I would like to try something better here anyway, just to satisfy my curiosity.  If I did then hear a difference I do have access (not really super easy access) to an AP for testing the DAC output...

[jabbr]

38 minutes ago, barrows said:

While I tend to agree, I would get excited if i could purchase a DIP packaged XO @45.1584 with <115 dBc/Hz @ 10 Hz, or a 27.5792 with <120 dBc/Hz @ 10 Hz for $500.00 USD or less.  Just for bragging rights and subjective testing.  I would want to receive a phase noise plot for each individual clock as well.

and a puppy 😂

 

But suppose you want to use the clock with eg DSC-2 ... that board isn’t clock optimized because you need to be sure all the clock traces are buffered, properly terminated and are the same path length, and there are many other things that can be done to optimize circuitry, crosstalk, impedance, switching noise etc ... and then you don’t even know that it matters ... but if you don’t carefully optimize the rest of the DAC with a truly low phase noise implementation you’ll never know. 

[Miska]

24 minutes ago, barrows said:

Like I said, I agree, mostly...  But I would like to try something better here anyway, just to satisfy my curiosity.  If I did then hear a difference I do have access (not really super easy access) to an AP for testing the DAC output...

 

I'd be curious to see J-test24 results for various different clock modules if you can easily swap them!