Ethernet switch

dCS Products
47

Now we are talking, Marc @poseidon77.

H2 - as it is - has an unspecified number of ad hoc hypotheses as it only says ‘other factors’. But which and how many? Occam’s razor would let us choose H1 because of that.

I propose different hypotheses:

H1: Perceived sound quality improvement is caused by changing the ethernet switch. Ad hoc hypotheses: (1) Only part of the listeners perceive the change as large enough. (2) Ancillary systems have not been standardized among survey sample.
H2: Perceived sound quality improvement through changing the ethernet switch is caused by self-delusion.

Occam’s razor would have us prefer H2, as it does not rely on ad hoc hypotheses. But would H2 fit a survey sample and prove its predictive power? You will notice that I introduced perception more prominently into the mix.

With these hypotheses we could go ahead and survey a sample of audiophiles and find out empirically. For that purpose we could debate how to structure the survey, or what percentages need to be reached to make H1 or H2 tenable. Also, we could argue about methods to employ in the survey to account for the possibility of delusion being the cause. As I see it, for H1, there would need to be a percentage above 5% of respondents including a check for sanity. These checks for sanity have been provided in the forum before - like letting a significant other listen or going back to an earlier configuration or simply the expenditure of upgrades. If we assume basic sanity and a certain level of education, people part with their money for good reasons. H2 is a strong statement. We would need to demonstrate that the vast majority of people who reported a positive change in sound quality are deluded.

In my experience there is a certain magnitude of effect a purely mental event can have. When I sit down to listen, after a while the music sounds better, because I have settled my mind into listening and distractedness has subsided. For others a glass or two of wine will do the same. That’s the threshold I would accept for self-delusion, not more.

You’re not reacting to three important parts of my above post. The tactics to deal with complex systems as a superior way to approach the hifi hobby and at the end the reference to empiricism, introspection and neuroscience to achieve valid insight. And last but not least the differences between individuals’ minds.

48

This discussion is very interesting. My job is doing research and teaching, and I am afraid that with hi-end improvements we are in a swamp where usual scientific paradigms do not apply.

Trying to design a reproducible experiment is hopeless, the quantity of control variables is overwhelming (and for some of them we have no control), and confounders (some of them due to legitimate subjectivity) are all over the place. I would never attribute a scientific value to any experiment conducted in this way.

So I gave up on classical Hypothesis Testing and developed a Bayesian approach.

I start with a baseline neutral level of confidence in a possible improvement due to an arbitrary component X: p(X is good). Then I update this value with all the information available (what I can find). Example: if there is a sound scientific rationale, p(X is good) increases, otherwise it decreases. If my ears are happy of what they hear, p(X is good) increases, if many people on the web (no reviewers, just passionate people) agree on how good X is, with no commercial mumbo-jumbo or pseudoscience, p(X is good increases), and so on. If there is a perfectly good scientific reason why an improvement doesn’t exist, p() decreases wildly, and so on. To give an example of this last one: when the seller says that a given Ethernet switch improves the accuracy of the bits that are presented to the DAC, or their timing over an Ethernet cable, this sends p(X is good) close to 0 immediately.
At a certain point, if the evaluation (which is always a distribution) is significantly far from the confusion area (in which anything could be true), toward the “good” area, where “significantly” must be weighed with price, then I reach for my wallet.

It works. It’s how I decided for some expensive cables, for the SPEC amplifier, for everything I have now, and I am pretty happy about that.

Of course, it works if one is well balanced. When you fall in love with some speakers, during the honeymoon you only see (hear) the princess, only to wake up one month later with a witch, and nothing can help you. But it’s part of the game, isn’t it?

A.

3 Likes
49

Ahhh, excellent, thanks for clarifying the real world pitfalls of hypothesis testing for hifi, @Zapp. So I probably have to drop that proposal. What I want to get away from is the rationale “lack of apparent explanation plus perceived improvement equals delusion”. This is not the case as our minds and how they interact with our sense consciousnesses differ greatly.

I believe it to be good to permit complexity in research. Complex fields like the human mind don’t lend themselves well to reductionistic research. Same as nutrition. The human body is too complex. The reductionist paradigm is mechanical in nature and resorts to simple causality. To state that A causes B it has to be true that (1) A always precedes B (2) B always follows A (3) There is no C that could also cause B. Simple models of causation can lead to unexpected and unexplained findings. Reductionistic research uses prospective experiments and case control studies. In contrast a wholistic evidence source is observational research. Contemplative science or neuroscience are others.

I really like your Bayesian approach. It simplifies in an apparently valid way. We had a loosely related and less scientific thread here:

2 Likes
50

This discussion has quickly taken an epistemological turn. Which is interesting, but also makes it impossible to settle the issue of ethernet switches in a way that would satisfy everybody. After thousands of books and tens of thousands of articles, there is still no agreement on how we can be certain to know what we are talking about.

Still, just for the sake of it, I do want to clarify my reasoning behind some of my assertions.

It’s best to drop the whole Occam Razor thing. The concept (or heuristic tool) doesn’t apply here and/ or needlessly obfuscates the matter at hand.

So let’s start from scratch and anchor our discussion in a real life situation. I change my ethernet switch and I detect an improved SQ. What’s going on? In other words, I make an observation and would like to have/ formulate an explanation. Most of the time, I won’t really bother; I’ve changed something in my system and that has produced the desired effect. Case closed. Why spend time and energy on something that seems obvious? It’s like changing a defective bulb and wondering why the new bulb works.

In this case, though, I know that changing the ethernet switch shouldn’t make an appreciable difference. Ethernet switches and cables work according to particular data transfer protocols and those don’t impact a DAC negatively, unless they are badly implemented, but then the transfer wouldn’t happen in the first place or in a clearly deficient way, such as producing plops and stutter — that’s what science is able to tell me. And yet there’s a difference. I can hear it after all. I could explain this scientifically unexpected result in a variety of ways:

  1. I’m on to something — ethernet switches do make a difference
  2. (related) There are more things in heaven and earth, Horatio, …
  3. I’m imagining things
  4. A miracle (amplifies the Horatio hypothesis)
  5. Besides changing the switch, I’ve also moved the rug half an inch to the right
  6. It’s rained a lot recently and my electricity has gotten cleaner
  7. I’ve developed the magic touch
  8. A butterfly flapped its wings in the Amazonian forest
  9. The parrot has finally expired
  10. and so on (one could generate an infinite sequence of hypotheses/ explanations, only to be cut short by following the parrot to one’s final quietus.

None of these explanations can be cursorily discounted. I grant you that some of them will strike you as pretty wild, far out there (perhaps on the astral plane). Perhaps we can agree that 6) for instance is plain bonkers. But we will only agree because we inhabit the same world, in this case a world that we take to be governed by certain universal sequences of causes and effects. That’s not yet a scientific world view, as we would also have to eliminate 4) — no miracles allowed in scientific explanation. And so on — but perhaps at one point we will disagree and our worlds will begin to diverge from each other. That is one of the hottest issues in epistemology, sadly undervalued in much of the epistemological literature — the inescapable horizon of our explanatory models, that which grounds what we consider to have explanatory force.

I will posit, though, that we are both interested in evidence based explanations. And that means that we will only consider explanations than can be empirically verified (another issue here: what are the criteria for that verification?). 9) for instance can’t be verified as a dead parrot can’t be resuscitated.

To simplify and work towards a tentative conclusion, I will consider 1) and 3). That’s my world, my horizon, as somebody who discards purely subjective impressions as carrying sufficient weight when deciding whether a statement is true or false.

I already know that ethernet switches shouldn’t make a difference. But to allow for the possibility that they might, I will have to eliminate the possibility of 3). That by the way also accords well with Hume’s dictum that extraordinary claims require extraordinary evidence.

  1. is a scientifically reasonable assumption. We imagine lots of things pretty much all the time. It’s called confirmation/ expectation bias, priming, … How do we test this (as we’re interested in evidence based results)? As far as I can see, the only available test, one that would take into account the possibly very special hearing of mine, would be to have someone switch the switch (horrible pun) without my knowledge. Will I detect a difference in SQ after the switch?

(I also know that when we test new gear, we listen more carefully than usual. So we often hear details that had escaped us before, and we ascribe hearing those details to the new gear, although we should also consider the fact that we are just listening more attentively. This is a case not of imagining things but of false causative ascription).

In my case — as I know that 1) has been scientifically debunked and 3) has been clearly established through numerous studies — I will privilege 3). Unless clear evidence is produced that 1) might after all be possible. Such evidence will never come from testing every person who claims that ethernet switches make a difference. How could we possibly test everybody, in such a way as to eliminate biases? And that’s why we will happily continue to assert stuff that has no basis in evidence. We can’t ever be proven wrong, at least not conclusively.

As to not addressing certain points you’ve made.

  1. how to deal with complex systems — your recommendations are great, especially in the way of implementing protocols of courtesy and respect. But I don’t think they will be of much assistance when it comes to deciding whether a truth claim can be validated.
  2. the neurosciences — not my domain (I’ve only read up on that with regards to issues such as the self as an illusion or free will). Moreover, different neuroscientists say different things.
  3. individual minds — sure, our minds are all different to some extent. We have for instance different responses to different frequency ranges, but I’m not sure that this matters in our present case, one I take to be based on (self)induced illusion.

We will never agree, but that’s OK.

51

I think you wrote a lot of interesting things, especially “extraordinary claims require extraordinary evidence”.

Which is not too far from my Bayesian method after all, is it?

To remain in context: at baseline, I am pretty much sure that a switch won’t make any difference. In order to convince me that it does, I will need A LOT of SUBSTANTIAL evidence from my own ears, with different kinds of music, at different times, in my own system.

When some website says that their switch will improve the bitstream quality, sorry but it doesn’t work that way. I can’t buy it: there is solid evidence that this is not how things work.

When they say that their switch isolates much better from noise coming from different sources, I think “hm … that’s odd, there is supposed to be a galvanic isolation in the Ethernet card, but I might give it a try”. And then I will need some very adequate SQ improvement as before.

I don’t think we may expect anything more scientific than this …

And of course, there will always be people who buy alarm clocks, or magical pebbles (and will actually hear an incredible SQ improvement) and people who, for various reasons, refuse to accept that a dCS Rossini, or a top of the line MSB, sound definitely better than some 500$ DAC.

That’s just the way we are.

A.

52

Good, Marc @poseidon77, let’s agree to disagree.

These proposed tactics are not meant to be a validation tool for truth claims at all. They are the dance that is called for in order to successfully engage with complex systems (aka the listener’s mind in relation to hifi). They are a blueprint for going about our hobby - much like the Bayesian approach @Zapp put forward.

You might enjoy reading ‘This is your Brain on Music’ by neuroscientist Daniel Levitin.

It appears that it requires a certain level of contemplative experience to appreciate how very different our individual minds are. The usual view is ‘Mind? Yeah, it thinks, so what. It’s just a function of the brain anyway…’ Both not true. The reality is more complex, less reified. Western philosophy is limited in its scope here. That’s because the interesting findings come from insight beyond cognition.

Frequency range of hearing as an example for variability between minds is orders of magnitude smaller than the actual differences. The differences matter, because the way my mind interacts with my ear consciousness can very well be so different from yours that we both can truthfully describe the opposite experience (hearing improvement, not hearing improvement). Our moment to moment perception is grounded both in mental dispositions existing since birth and listening experience acquired during our lifetime. Moreover, what I listen for in music reproduction will almost certainly be different from what you listen for. Those are the real reasons why we have these debates. No illusion involved. My firm belief is, that dead parrots, miracles, rainfall, butterfly effects, wearing tin foil hats and so on are not considered an ingredient by any participant here. In other words I believe in the basic sanity of all involved. Plus experienced audiophiles know biases, self delusion and so forth and account for them. BTW, we are not debating coffee makers or pebbles in relation to hifi, we debate a component that is a direct part of the audio chain.

Unfortunately it is only hypothetical and I am back to personal experience. If you @poseidon77 and @Zapp came over we could compare. The cumulative differences in setup would be switch + LPS + power chord and maybe ethernet cables. I would change more than the switch, because then it becomes really obvious. All of the above changes should make no difference, because it’s IP traffic. So I believe all would be permissible in a test setup. I am confident both of you hear the improvement. If not, you would each get a box with six bottles of Louis Roederer.
:champagne:

Actually, in terms of testing there is little to lose but time and effort. Procure an audiophile switch (Melco, English Electric, Innuos), get a good power supply (Keces, Ferrum, Paul Hynes), a decent power cord and power distributor and try. The achievable result is a more natural, analog sound. Less of what is attributed to ‘sounding digital’. It won’t make a Bartók a Rossini. But if one enjoys the sound of vinyl, it’s worth considering.

If the only alternative to debunked is illusion, testing a switch upgrade is a waste of time. I give to consider that the comfort of the armchair might be the wrong place to resolve the issue…

53

Galvanic isolation is not as simple as that. The two sides may very well be isolated for DC, but at higher frequencies the capacitive elements of the physical isolation device come into play and the isolation is no longer perfect, hence HF noise may get through to the analogue circuits in the DAC (not the digital ones).

54

Exactly! That’s why I say “I can give it a try”. Because there are compromises in every device, so why not test if we can have an improvement here? As you explained, there is a technical rationale behind it.

A.

55

True in theory…

However, in the case of the Rossini for example, it uses a discrete part for its Ethernet port Magnetics; Pulse Electronics H5062FN <— Click on that link to have a look at its specifications; signifiant noise attenuation well into 200MHz!!

I’ll leave aside subjective discussions of the audibility of those kinds of frequencies ever impacting the DAC. I will just say though, IMHO, the dCS Engineering folks know what they’re doing :wink:

1 Like
56

I find your extension of the Bayesian method intriguing. Maybe in the end it just boils down to good old common sense.

How would you distribute values?

IMG_1456
IMG_14561857×596 77.5 KB

Because that’s really the issue, isn’t it? Would we all agree on the same values in our calculation of probability? I posit that our respective values will largely depend on priors that ground our sense of reality/ what to expect; those priors fall outside the rational parameters of the system to be employed in calculating probabilities. In other words, the values would need verification/ rationalization that Bayes’ theorem cannot provide by itself. (An extension of Gödel’s incompleteness theorem).
You may know that Biblical scholars have had some fun with Bayes’ theorem, especially with regard to the historicity of Jesus. As you can imagine, that proved highly contentious and didn’t settle anything.

57

Well, of course you cannot factor subjectivity out of the equation. I put the financial factor in it, saying that I also weigh my decision on the cost of the device. That’s quite subjective of course, otherwise we all would be listening to full stack Vivaldi …

That of choosing priors correctly has always been the main issue with the Bayesian approach. But in this case we can safely assume a non informative uniform prior: p(device is good) dist= Unif([0,1]).
Then, by applying Bayes Theorem repeatedly, you see the distribution change, hopefully become something narrow with a well defined modal value …

All this, of course, happens in on’e mind. I wouldn’t make the computations, although I have the libraries to do MonteCarlo sampling and as a pure exercise it would be interesting to see.

If you put in the model evaluations from many different people, in some cases you will obtain wide distributions with no easily identifiable mode, in others you’ll see bimodal distributions. It reflects the fact that it can be hard to reach a consensus, due to a variety of subjective evaluations.

A.

58

I’m truly happy how this discussion unfolds. I find it enjoyable and rewarding to read, think through and weigh our different perspectives and approaches. All have something going for them. Thank you!

2 Likes
59

Just to contribute my slightly different angle on this subject.

I’ve become addicted to Qobuz streaming over the last few years - I just can’t live without that infinite universe of music to explore! But streaming never sounded as good as locally sourced music. It was comparatively glary, sometimes a bit harsh, soundstage a bit fuzzy… Consequently, sorting out the network connection has been a high priority.

I tried some different RJ/45 type copper cables, cheap ones and expensive ones, including both shielded and unshielded types (all CAT certified) - didn’t make that much difference to be honest. Disappointing.

I then tried a couple of the new ‘audio class’ switches between the router and the dCS Bartok. That did make a worthwhile improvement - I was getting somewhere at last. One of the switches I had on loan, the Melco S100 has optical inputs. I had read about the advantages of full galvanic isolation that optical ensures. So as optical cables are cheap and easily obtainable I tried it.

Now that genuinely made a big difference. All aspects of sound quality were instantly better - and so easily achieved. Powered by my new found enthusiasm for all things optical, I experimented with different optical cable types and transceivers. As the SPF is an international standard interface, this is easy and safe.

I’ve ended up with Single Mode, OS2 1310nm optical cable from Commscope and transceivers from Finisar, I have the FTLF1324. The S100 is on my hifi rack in the listening room. My Ubiquiti router is 15m away in another room. Next to the router I have the Sonore Optical Module Deluxe, then 15m of optical cable. All components have their own LPS. I’ve also isolated the WIFI access point optically using very cheap components sourced from Amazon.

It’s the best ‘bang for buck’ upgrade I’ve ever had. I’m truly shocked and delighted at how good Qobuz streaming now sounds - easily as good (if not better when hi-rez) than any local source, CD, SACD or vinyl.

My takeaways from this are: an optical link in the home network is a critical necessity to achieve the best streaming SQ. You still need RJ/45 CAT cables to make connections, but it really doesn’t seem to matter much what type you use.

2 Likes
60

Using fiber optic technology improves the sound quality. I have been using it for a week. I chose a slightly different way. Behind the PPA Studio Duo router I have placed a GigaFOILv4-INLINE - EMI / RFI Filter. From GigaFOILv4-INLINE - EMI / RFI Filter, the signal flows directly to Bartók via the Polish cable of GFmod Advanced PRO LAN - Ethernet cable. The very use of the Gfmod cable between the router and Bartok was a significant improvement in sound and after using the GigaFOILv4 filter there was another improvement. If I hadn’t heard it with my own ears, I would have been ashamed to write, thinking that I had succumbed to the suggestions and rationalized the spending of $ 1,500 on items that would only make me feel better. I encourage you to test this solution in your own system - because only your own ears can appreciate or not appreciate the change that the solution described by me brings. Although from the point of view of hard science there can be no change, and probably because the signal path is more complicated, there may be deteriorations in signal transmission. Meanwhile, the change is beneficial.

61

That’s the solution I was toying with. I was thinking of an optical connection from the router to the Rossini, except I need two transducers: one from copper to optical near the router and another one from optical to copper at the Rossini. I think the transducer is powered, isn’t it?

In my approach, the optical solution is much more plausible than some hi end cable.

A.

62

There’s an extensive thread from early this [year] that goes into details on how to go about this. You might find it very useful to go through the full thread;

1 Like
63

My feelings exactly! It’s refreshing to have a discussion of a good intellectual standard.

I’ll get my hands on a copy. Thanks for the tip.

That’s tempting. But in the middle of the pandemic …

“little” is relative. Here’s what I mean.

  1. Financial aspect: I’m very reluctant to put money into something about which there is only anecdotal evidence. Not everyone can shrug off a few thousand Euros if the product doesn’t work as desired. There’s a thread on this forum re ethernet cables, and one participant says that after spending tens of thousands of dollars on cables, he has finally come to understand that it may not have been altogether worth it. Tens of thousands of dollars splashed on tweaking your system?! Great for that person, but that’s hardly relevant for the vast majority of hifi enthusiasts. It’s just financially unfeasible for most of us.
  2. Time: indeed, how much time are you prepared to spend on optimising your system? I personally find that the more I read up on hifi components, post on forums, arrange for demos, etc (all of that is fun, no doubt and can be rewarding), the less time I actuallly spend on listening to music. And I may begin to worry too much how my system sounds, as opposed to just enjoying the music. Finally, I have other interests as well. As for instance reading Levitin’s ‘This is your Brain on Music’. :wink:

I agree. Testing something as a component of your system, in your listening space, stress free, etc. is preferable to making armchair judgments. But see above. And then there’s of course the question of how much you should trust your own ears. I’m not going to get into that again, though.

64

Yes the transceiver is powered by the media converter which has the SPF port (or ‘cage’ as it’s sometimes called) into which you slot the transceiver. The optical cable then plugs into the transceiver. Optical cables do come with different connectors, but it’s simplest to use the most common LC type. So with the same type of transceiver with LC sockets at each end, the cable required is LC to LC.

You are right, you place one media converter with its transceiver next to the router and the another next to the streamer - optical cable in-between. Any CAT ethernet cable is fine for connecting-up the media converters. They don’t make an appreciable difference in my experience because the optical link has ensured full isolation.

1 Like
65

A less complicated solution that brings exactly the same, that is thanks to the use of optical filtration, can be obtained using Gigafoil V4. The advantage of this solution is one device and only one LPS power supply is needed. I was considering an optical solution in my system, but I was scared off by the amount of additional devices in the case of the Gigafoil filter we have it in one box. Of course, I am not a distributor of devices from the company I am writing about. No distribution in Poland. It is worth trying both solutions and checking their effectiveness in your own system. I am convinced that you only have to trust your own ears. So what if someone else hears the improvement of the sound in auer system, if we cannot hear it, it is negligible for us.

1 Like
66

The Gigafoil and Keces was never stable for me. I returned them and went to a ~30ft run of fiber optic cable from my switch to a SGC opticalModule just in front of my Upsampler. Since then rock solid Roon.