Ideally the main reference frequency should ideally be located as close to the dac as possible, this was stated by Wadia in the early 90's alos they ran a synthesied clock for the DSP as well. They used dedicated 16.934Mhz & 18.432Mhz for the two mulriples of 44.1 & 48Khz As many manufacturers did.

However as with all things in the real world its not plain sailing. These days with great fabication techniques and small components, circuit board layout, board stacking and veer placement, plus many clock now are synthectically derived.

The quality & placement of the oscilliator, supply rails, length of track to the dac chips itself (or the precision ladder) is all taken into account.

I can only think of three (dacs) that really do run with this philosophy right to ever aspect of integration and have what i would call a outstandingly low phase noise & jitter figures all of these dac's are in the six figures. Does that make they so much better than the others, in most cases not so much but with the right set up its quite breath-taking.

The most accurate internally fitted clock in a dac is the now the Wadax which claims a phase noise of 12fS which is around 2 more than the clocks I use in the lab, MSB have a 33fS device in there Select II and you can upgrade it on the reference for a mere £12K

"Zepto Reference clock: doubled the master clock frequency with only 12fs total jitter." which in itself is not correct, it should be halved lol I would suggest that is beyond 10-13th power in terms of accuracy



So does have infinitesimally small phase noise & jitter really deliver positive up-turns in SQ given that the hearing threshold of humans is 120s/n can we really hear such differences well plenty of folks will tell that the maths don’t stack up its impossible to hear any differences what so ever and that you are all barking mad.
Conversely the LB clock has quite a low phase noise reference point @ 1Hz its -70dBc getting on towards -160dBc @ the 10Mhz point better than a lot of Rb clocks in audio, however it does exhibit one trait that a great many so called reference standard clock don't it has lot smaller noise spurri on the plotted trace. I could name; more than a few that are shall we say a touch more 'deviant' than the LB


Now quite a few dac use frequency dividing networks to achieve the various individual sample rate frequencies and every time you go through a splitter or introduce another outlet or deviation into the clock pathway you introduce phase noise, how much? purely depends on how the circuit has been implemented.

IMHO the best dac's just have two clocks plus one master and they are sited within 30mm of the dac chips or ladder.

However, to really create a truly decent clock it’s no small task and its going to take a linear LPS, many stages of regulation and smoothing, rock-solid stability, individual supplies for the various temperature-controlled sections, all circuit pathways impedance matched to within a one trillion of a nat’s chuff and that takes a lot of effort, time and R&D.
Also when these clock are touted and measurement given, this is a laboratory environment, all test equipment would have been on for at least six ours, temperature controlled, isolation mains supplies, test ground planes ultra slick cabling all used in the lab. Remember, you are measuring these devices from their output through a specialist cable to a very precision measurement instrument which has been designed to remove as much of internal device noise /temperature and power supply (ac & dc) crap in an RF free area possibly a faraday cadge or clean room as in the CISPER testing.

This does not resemble your audio system at home in any shape or form, and as I mentioned at the start the 'real world' clock performance in your DAC or Steamer is NOT the same imho.

So why does attaching an aftermarket clock make that difference?

Could be a multiple of factors, some manufacturers design in the word clock input to sell you and off board unit, so it could be a marketing exercise.

The clock being attached is superior even through a cable and pair fo BNC interfaces

The original design of the dac/streamer/transport is average and the money was spent on the casework and remote.

It just merely changes the presentation not for the better or worse (a change is as good as rest).

I have recently come across dac's that are wired for 10Mhz inputs and even with the top clocks ZERO difference can be detected despite the company actually making a master refence clock, that one I'm looking into lol

A very good & industry respected fellow once said to me many years ago when challenged about 'could he hear any difference' replied I can clearly here a difference, therefore there is one, although whether you like that difference is a matter of personal taste. But don't tell I can't hear something where I clearly can!


Some idea of what is involved in clocks & Phase noise  grading

For Tony
So spurs have a effect on sound quality. Is this similar to jitter or something a bit different

Yes l would like to know about Spurs, so no Harry Kane references 🤣

All joking aside, it would great to see some good and bad examples of Spurs, and indeed Tony's take on it.

Martin
I am not surprised the AD clock makes a big difference on the U18 I remember hearing something similar in Vic setup when Stellabagpus put his AD first on the U18 and then on the DI20HE. It certainly helped both DDC's With the Audio gd DI20HE no longer being just a great DDC for Jazz as it suddenly also kicked arse 🤘. Just a shame the software is a bit buggy.

How much of this correlates with Gustards and their K2 Synthesiser? I can’t help feeling there’s something in the way the K2 works that means external clocks, and possibly the LB in particular benefit. I can’t speak for how external clocks work out with other DACs that have an external clock facility.

I can only relate additionally to the EtherREGEN reclocker.  This has an internal Crystek 575 which limits its performance quite a lot.  With an external clock, the ER really delivers as a pre-stage to the DDC/DAC combo and its benefits are very audible, even to changing its PSU and clock cable.

This shows the Crystek 575 phase performance.  Hard to believe as it holds back both the LKS DAC and EtherREGEN in sound quality.

I agree, the 575 sounds bad (very Naim like) in both the ER and my old LKS. My point is, though, that it 'measures' well with a low published phase noise and spike-free plot.

John the blue triangles are markers. The plot itself is remarkably spike-free. Possibly massaged to look good.

Just having a break from drilling lol, I've noticed that the PN plot on the 575 is way up on amplitute its +14dBm that high, most measurments are around -30 to 0dBm for RF based devices (apart from amplifiers lol)

That is 25mW of power 3.16V pk<>pk and 1.12Vrms assuming a nomial impedence of 50 Ohms.

I elduded top the rise edge of the SQ wave is one indicator of possible SQ along with pre & overshoot, I am more interested in + & - (ve) width's