This is a real interesting point of view, to which I am not disagreeing, especially given you have clearly spent a lot of time developing a switch. I have personally found that adding a clock had a profound effect on my pair of SOtM switches. I should also note that the power supply also had a big impact, I am not really able to countify which had the biggest impact, I would lean towards the clock having the biggest in my set-up, although could that be a result of the post switch streamer etc; I am using! Who knows.
Best audiophile switch
- Thread starter Frankie67
- Start date
I can confirm that.
Years ago, I read an interview with Ed Meitner where he made fun of reclocking. He said that the important clock must be located next to the DAC chip. This seemed so obvious to me for many years that I looked for DACs with the highest quality clocks possible. But I never really heard a difference.
In scientific research, new solutions are not found by repeatedly treading the same well-worn paths. So I tried a network switch with an integrated OCXO clock from Afterdark. The gain in instrument separation was immediately audible.
Later, I purchased the 10 MHz reference clock generator from Mutec REF10 SE120. This has extremely low phase noise in an important range, which I will discuss below.
Of course, I wanted to understand why this sound improvement was possible. Because actually, the sender clock plays no role in Ethernet. This is because data transmission is asynchronous. It doesn't matter whether the data reaches the buffer slower or faster. This is because the buffer is taken from the receiver clock at exactly the right time. See image below.
Let's turn to John Swenson, who was a senior project manager at a large integrated circuit company for 30 years. He brought Wander into play, i.e., phase noise in the range between 1-10 Hz.
The phase noise spectrum of the incoming data overlaps the phase noise spectrum of the local clock (receiver clock). This is how phase noise travels from one device to the next, even over optical connections. This phase noise spreads – both with interfaces such as Ethernet, USB, S/PDIF, I2S, and in the chips on the boards. Even a DAC with a perfect clock sitting right next to the DAC chip is affected by all the upstream jitter/phase modulation that has taken place in the chips and on the circuit board just before it.
I don't know whether this theory is correct or not. But every guest I play music to over the network, once with and once without the Mutec REF 10, hears the difference. And wants to continue listening with reclocking.
@StreamFidelity great post.
Mutec REF-10 SE120
You start by mentioning DACs and this is of course a completely different domain where different rules apply. It's important we are clear about which domain is under discussion and we avoid blurring boundaries and/or extrapolating inaccurately from one domain to another.
Q1: Is your Mutec attached to your DAC or your switch?
If your DAC, I can absolutely see how it can make a difference. Nothing to argue with here! The streamer rips open the data packets to construct its bitstream to which it applies a timing element and jitter becomes a thing. Clock accuracy inside the streamer and at the DAC makes a huge difference because here we are in the land of the synchronous signal. There are clearly understood mechanisms by which this can occur.
Back to the network
Q2: When you tried the network switch with integrated OCXO clock from After Dark and heard an improvement, were you comparing it with the same switch without the OCXO clock? Or a different switch entirely?
In other words, were you able to isolate the clock as the only possible source of the improvement you heard?
You clearly understand how ethernet is asynchronous, with error checking and data retransmit if required. Excellent.
If a quiet external clock is used on a switch which disables the internal clock, it's conceivable that a quiet external clock might generate less RFI noise than an internal one. Ditto if a quiet internal clock were to replace a noisier internal one. I'd argue this would be an RFI effect, not a timing effect.
Given how the multi-layer ethernet protocols work, I have yet to see any explanation as to how this timing is captured nor how it (and what Swenson calls phase noise which is distortion in the timing domain) somehow survives the data transformation which is made by the streamer. Because it apparently manifests itself even over an optical link, it cannot travel as an electrical signal like RFI does. It must be baked into the data itself. How? At what level of the 7-layer OSI model? These are clearly rhetorical questions and, equally clearly, any manufacturer is free to prioritise his/her design criteria as he/she sees fit.
All the best,
Nigel
Mutec REF-10 SE120
You start by mentioning DACs and this is of course a completely different domain where different rules apply. It's important we are clear about which domain is under discussion and we avoid blurring boundaries and/or extrapolating inaccurately from one domain to another.
Q1: Is your Mutec attached to your DAC or your switch?
If your DAC, I can absolutely see how it can make a difference. Nothing to argue with here! The streamer rips open the data packets to construct its bitstream to which it applies a timing element and jitter becomes a thing. Clock accuracy inside the streamer and at the DAC makes a huge difference because here we are in the land of the synchronous signal. There are clearly understood mechanisms by which this can occur.
Back to the network
Q2: When you tried the network switch with integrated OCXO clock from After Dark and heard an improvement, were you comparing it with the same switch without the OCXO clock? Or a different switch entirely?
In other words, were you able to isolate the clock as the only possible source of the improvement you heard?
You clearly understand how ethernet is asynchronous, with error checking and data retransmit if required. Excellent.
If a quiet external clock is used on a switch which disables the internal clock, it's conceivable that a quiet external clock might generate less RFI noise than an internal one. Ditto if a quiet internal clock were to replace a noisier internal one. I'd argue this would be an RFI effect, not a timing effect.
Given how the multi-layer ethernet protocols work, I have yet to see any explanation as to how this timing is captured nor how it (and what Swenson calls phase noise which is distortion in the timing domain) somehow survives the data transformation which is made by the streamer. Because it apparently manifests itself even over an optical link, it cannot travel as an electrical signal like RFI does. It must be baked into the data itself. How? At what level of the 7-layer OSI model? These are clearly rhetorical questions and, equally clearly, any manufacturer is free to prioritise his/her design criteria as he/she sees fit.
All the best,
Nigel
Thank you very much for your openness.
The Mutec Ref10 SE120 is connected to the switch. I use a GUSTARD N18 PRO.
I compared it to a SOtM switch that didn't have a clock input. But that's not the point, because it's like comparing apples to oranges.
More important is the comparison with the same switch with and without clock. This works very well with the Mutec REF10 SE120 and the GUSTARD N18 PRO switch (see above). At the touch of a button, the clock can be removed from the signal path without causing any delays. And here, all my guests (including myself) have always preferred reclocking.
Have you ever performed a reclocking?
While jitter indicates deviations in the time domain, phase noise uses the frequency domain. The same drift can be derived from both values.
When phase noise converted into jitter, the comparison (the lower, the better) shows that the clock generators from Mutec (REF10) and Afterdark (Giesemann) are of excellent quality, depending on the characteristics:
Source: https://www.aktives-hoeren.de/viewtopic.php?p=240701#p240701
Forget the usual Crystek quartz crystals ore others; they can't compete with good OCXO clocks. You won't hear the difference if you use cheap clocks.
We had heard to identically switches, but one with a better clock. It is quiet easy to hear the difference. Also if you have Etheregen an you use a external clock like Mutec. The clock is better next to the switch asic and better not external. If it is external you need PLLs to clock your 25MHz signal for example with 10MHz reference.
If you have external lps, the are offen as better as the internat ones. But the best is to put the regulator next to the consumer, less cable, no plugs,...
In the end just one thing is important. Whatever you did (or a manufacturer), do it as good as possible. All the things matter.
I think all the difference is on layer 2, not above. Maybe this courses RFI as an side effect and let us hear the difference.
I did not hear any good explanation of what we hear.
If you have external lps, the are offen as better as the internat ones. But the best is to put the regulator next to the consumer, less cable, no plugs,...
In the end just one thing is important. Whatever you did (or a manufacturer), do it as good as possible. All the things matter.
I think all the difference is on layer 2, not above. Maybe this courses RFI as an side effect and let us hear the difference.
I did not hear any good explanation of what we hear.
Excellent, thanks for confirming.
I have reclocked in the (post-streamer) digital domain when I had a "pimped" Bluesound Node 2i streamer with a high end LPSU and added a Mutec MC3+ (USB designation though I didn't make use of this) and liked what it did. I prefer my Innuos devices without.
My point is not to suggest anyone who hears a difference with an external clock on a switch is imagining things! It's simply that we don't understand the mechanisms via which the addition of a clock improves sound quality; it could be noise-related or timing-related, and my money is on the former because I can see a mechanism for this whereas I can't for timing. IMHO it would therefore be unfortunate if people jumped the conclusion "the addition of a clock improves sound quality therefore this improvement arises from an improvement in clock accuracy".
Last edited:
I understand what phase noise is, and post-streamer I'm with you.
On jitter, we have the pre- vs post-streamer domains to consider. Jitter in ethernet is a term used by network engineers, typically running heavily loaded corporate networks, to describe any variability in the timing of the arrival of data packets/frames at the receiving network node. This is the same word but with an entirely different meaning to in the post-streamer (bitstream) domain.
Let's no forget that (a) under ethernet protocols, errors are checked at the receiving end and if a packet needs to be re-sent then it will be, (b) the streamer will buffer the packets before (c) extracting the digital data from them and constructing a continuous (and timed!) bitstream which the DAC can recognise. This is why clock accuracy is not relevant, though clock quietness may well be (or some other mechanism we are not yet aware of).
This is why I think we can dispense with timing as a concern. It's just the way this stuff works which makes it a non-concern.
Phase noise on the other hand? This is why I'd like to see an explanation of how this gets coded into the data. And how it impacts or survives through the massive data transformation inside the streamer.
Last edited:
My (non-technical) working assumption is that reduction of RFI is the main mechanism to getting better sound via digital.
Do you think there is a RFI-related difference between 10 Mbps and Gigabit transmission? The internet tells me they use different encoding schemes for voltages.
I agree with this.
There may be a difference - and I know some switch designers chose to limit bandwidth to 10Mbps for this reason - but I guess the real questions are (a) is it consistent across all implementations and (b) whether it matters in the great scheme of factors which one can choose to focus on in seeking to design stuff that reduces RFI ultimately ending up at the DAC.
On (a) I suspect it's a bit like arguments about amplifier classes or SMPS vs linear PSUs and ultimately more about the individual implementation than any universal statement of fact. On (b) it would seem crazy to choose one max bandwidth over another while still festooning your device with LEDs which have notoriously noisy drivers adding RFI of their own back into the mix.
Thank you for your thoughts. I found a recently published series of tests by alpha-audio interesting.
In my opinion, it implicitly confirms John Swenson's theory that Wander also travels via fibre optics.
But I think we are moving into an area that few people understand. I am one of those who do not understand it. For me, it is like quantum physics. Everyone knows it exists, but only a few understand it.
Therefore, my recommendation is to simply try reclocking. Scientific progress is not based on an explanatory model, but rather on researchers simply trying something new. The explanatory model is developed afterwards.
Will have a poke, thanks.
For me it's not. Quantum physics arose from an iteration between theory and practice. As far as I can see (above Alpha Audio link notwithstanding as I've yet to view), all we have here is practice with no coherent theory of how "Wander" is encoded in data packets
As above, it needs to iterate. I'm already struggling with how a high accuracy clock in an asynchronous domain might affect signal timing/distortion (jitter) in a synchronous domain. The suggestion that increasing clock accuracy might not only decrease timing distortion but might also decrease frequency distortion is a bridge too far for me.
Still, thanks for engaging, and as we both know, it's quite possible to disagree and to do so politely!
Best wishes.
Nigel
Some explanation from UpTone Audio:
"high frequency stuff (RFI) on PCBs is the easiest to get rid of. It is the very low frequency stuff--on the ground-plane, modulating the DAC's clock--which is hard to get rid of."
"phase-noise and amplitude noise actually become each other and cause ground-plane bounce which propagates all the way through the chain and onto the DAC's ground-plane--where it then does affect the jitter of the DAC's master clock."
"high frequency stuff (RFI) on PCBs is the easiest to get rid of. It is the very low frequency stuff--on the ground-plane, modulating the DAC's clock--which is hard to get rid of."
"phase-noise and amplitude noise actually become each other and cause ground-plane bounce which propagates all the way through the chain and onto the DAC's ground-plane--where it then does affect the jitter of the DAC's master clock."
So neither phase-noise nor amplitude noise travel as part of the signal (including over an optical connection as mentioned by an earlier correspondent), they both travel via the ground plane? This is the first time I've seen this postulated.
RFI is not easy to get rid of!
Hello all you switch loving audiophiles.
I know some of you have disdain for the audiophile press; however, it is a data point. The September issue of TAS reviewed the Reiki switch and optical bridge as well as the JundoStream cable. Very nice comments. Of particular note is the comparison Robert Harley made with another audiophile grade switch.
In my experience, adding Reiki products before the steamer has a profound effect on the system's noise floor. I have stated this elsewhere, but I will say it again. I am constantly amazed at how much noise enters the system through an Ethernet cable. Anyone going from a router directly to their streamer with a cheap CAT 8 cable is flooding their system with noise.
in full disclosure, I am an Reiki Audio dealer (as is stated in my signature).
I know some of you have disdain for the audiophile press; however, it is a data point. The September issue of TAS reviewed the Reiki switch and optical bridge as well as the JundoStream cable. Very nice comments. Of particular note is the comparison Robert Harley made with another audiophile grade switch.
In my experience, adding Reiki products before the steamer has a profound effect on the system's noise floor. I have stated this elsewhere, but I will say it again. I am constantly amazed at how much noise enters the system through an Ethernet cable. Anyone going from a router directly to their streamer with a cheap CAT 8 cable is flooding their system with noise.
in full disclosure, I am an Reiki Audio dealer (as is stated in my signature).
And if you evaluate clocks in audiophile switches in the context of total system noise, not in isolation?
Clock quality matters only to the extent that it reduces noise — not because Ethernet data requires better timing.
It’s absolutely fair to be skeptical — after all, Ethernet is an asynchronous protocol, and in theory, shouldn’t benefit from a higher-grade clock.
But some argue the improvements don’t come from more accurate data timing, but from reduced phase noise and electrical interference within the switch itself. That noise can propagate downstream and subtly influence the environment in which the DAC reconstructs the signal. In those cases, the clock isn’t “timing the audio” — it’s helping to quiet the electrical backdrop.
Well Said.
In a certain sense, it doesn't matter how it works. If I connect something to my system and it lowers the noise-floor I am happy. I no longer feel the need to understand all of the science behind how it works. I simply view all of these things as tools to improve the music. Of course, some tools work better than others.
Regarding clocks used in audio, it is interesting that two very different designs -- clock as close as possible (to DA converter, for example) and clock in a separate enclosure -- can work very well.
Electronic interference is certainly an enemy in the Ethernet path. But I think John Swenson means something else. He has published a white paper: https://cdn.shopify.com/s/files/1/0660/6121/files/UpTone-J.Swenson_EtherREGEN_white_paper.pdf
I quote some from it:
I don't think you'll find a setup that basic in an audiophile system, not if they have any interest at all in high-fidelity network audio. I know some just use Spotify or internet radio for background music, because their interest is in CD or vinyl.
CAT8 is always shield-tied to ground at both ends, therefore it requires specific grounding conditions to avoid a ground loop. Even though it has the best shielding capability, it should not be used universally.
Similar threads
- Locked
 | Steve Williams Site Founder | Site Owner | Administrator |  | Ron Resnick Site Owner | Administrator |  | Julian (The Fixer) Website Build | Marketing Managersing |