I think these discussions are somewhat pointless. I suggest, if you want to try to improve SQ, you purchase a device with a good return policy and give it a try. Some people hear, or think they hear, differences. Others don’t.
For me, a Roon Nucleus works great, but I can tell no difference in SQ using ethernet or HDMI to Oppo 203. It just doesn’t matter.
Not yet.
One could probably achieve something equal by implementation of 10GB standards and fiber. And that may also introduce some challenges, depending on your system. (And cost as always).
You may still struggle to get measurements from manufacturers, like stressed eye pattern and properly data sheets. I must admit it’s hard to understand over 600 pages in a standard related to network transfer 
I found Finisar SFP’s at least seems to have some proper and valid information.
I also think 1310 nm singel mode fiber is a good choice. Mikrotik has a very reasonable 4 port 10GB switch with SFP+ cages.
Unless you building your own PC, no audio equipment as of today support SFP+
There exist SPF+ modules that are compatible with one GB network. I only know about Lumin and Sonore that supports optical SFP (1GB).
I do hope fiber will be more widely adopted in audio network.
Let’s talk in 5 years from now. Who knows, maybe IoT and 5G will take the lead
Hi Richard,
I had the same opinion.
MacMini -> (via USB) McIntosh D100 -> MC275
I tried to change.
Despite the fact I added a very expensive layer (Optical Rendu - Sonore), apparently only a “complication” in the connection flow, at the end of the day, the end results, after many comparisons between the two configurations, at least in my case, the second one, with OpticalRendu is much better. It’s worth the additional cost, I don’t know, but it’s better.
MacMini -> network via ethernet -> switcher with FiberOptic -> optical Rendu -> (via USB) D100 -> and finally MC275 (any either effective or potential electrical noise disappeared via FiberOptic).
In addition you can move MacMini (or any NAS) you can put it away from the listening room.
But my final suggestion is trust only your ears, nothing better exists in order to measure the final results.
Enjoy listening to your music.
Sante
Honestly I facepalm every time a post begins with an expertise declaration. There are experts who disagree all of the time. Having expertise in a field is not dispositive. If I ever got cancer I’d not go to one expert who said to me “trust me, I’m an expert”. I’d be sure to get second and third opinions from other experts and probably reject the one who began and ended his analysis with “trust me I’m an expert”. I’m not doubting either your expertise or the validity of what you have written. I just don’t think it’s effective to start with the declaration of expertise.
I agree. Being an “expert” or “audiophile” does not give you magic hearing. People need to decide for themselves.
There is no such thing as a “golden ear”.
The OP’s question was answered a long time ago!
Yes, yes, yes, for f#$ksake YES, you can connect directly from Nucleus to DAC via USB and avoid a streamer. Try it. If you like the sound, you are finished and you can just enjoy your music.
The rest of this topic is just…argh…painful.
No one should be so full of themselves that they believe they have the divine expertise to tell anyone else what the other individual can or can’t hear or what they should or shouldn’t hear.
Research your topic of interest, read reviews, ask questions…all good. But then go build your system to a budget YOU are comfortable with, with components YOU like, with a sounds that YOU like, and then tune out all the damn “experts” and sit down and enjoy YOUR music.
Enjoy the damn music. That’s what this is all supposed to be about. The music. The rest is just extraneous noise and hot air…and not via some USB cable or Ethernet cable.
OK, stepping off my soapbox. Nothing to see here folks. Move along.
High end audio is so full of bullsh:t and snake-oil products, it’s the worst hobby for this sort of crap of any I can think of. So lots of people have zero trust or confidence in vendors (there’s no correlation between cost of a product and it’s performance) and the products they peddle. Because of that, you see a lot of people buying things that some of us feel are a complete waste of money and we lash out partly out of frustration that those products exist and partly as a warning, and maybe as a poorly aimed way to try to help, etc.
Sheldon
Wake me when they happen. I’d like to see them.
I doubt it. The electrical isolation issues you seem so concerned about have already been dealt with in less complicated and less expensive ways.
Well, except for priests and other societally-blessed shamans. I suppose psychiatrists might also consider they have a sort of secularly divine expertise.
I didn’t intend to argue here, but there is an important point.
@danny is touching on it, but I’ll belabor it.
There is a subjective element to this, but there are parts that are objective and inarguable. It has to do with layers of abstraction — if you don’t take this concept into account you are led astray. And these layers of abstraction are objective and inarguable because they are invented, not measured and discovered. This is how we have built our systems.
If you have studied this at all, you know the OSI Model for networking:
What does it mean to define these layers? It isn’t just a pedagogical diagram.
Layers of abstraction means that we can innovate and modify in one layer, as long as it conforms to the abstraction, without harming the overall system.
How is it that the whole network stack continues to work when we replace the physical Ethernet layer with WiFi, or cellular modems, or USB? Because of the abstraction.
And the abstraction isolates the upper layers from the implementation details of the lower layers. This is why @R1200CL is wrong about jitter (which is what triggered this post). There is indeed a concept of jitter in Ethernet, and the White Rabbit thing he mentioned addresses that. But if you read the advertising copy for it (and @R1200CL’s notes), you see that all of this is about the bottom layers of the stack. This jitter may be perfectly valid for the performance and reliability of the Ethernet layers, but it has nothing to do with the jitter we are concerned about for DACs. Because of the layers of abstraction.
If an implementation of Gigabit Ethernet suffers from out-of-spec jitter, it will fail to deliver its functionality, it will not meet the abstraction, but that has nothing to do with jitter in the payload, the data stream jitter that may cause distortion in the DAC output.
The issue of Ethernet jitter is not wrong, but it is irrelevant.
Similarly, it has been popular in the recent year to explain that the “digital” signal Is actual an analog electrical signal that can be distorted by various flaws. Again, true but irrelevant. If we want to go crazy, we might say that the signal has fine grained noise because it is made up of electrons. Silly: it isn’t silly for physical reasons (even though the electron noise is 380 dB down), it is silly for engineering reasons because the abstraction hides that noise.
Problems in the bottom layers do not shine through to the top layers.
This doesn’t mean that problems in the bottom layers are not significant. (If a backhoe cuts the Layer 1 Physical cable, the top layers don’t work.) But layer 1 jitter has nothing to do with Layer 7 jitter (actually the digital to-analog conversion itself is above layer 7, it isn’t a network function at all, it’s a consumer of this stack.)
Note that I did not say that network implementations and cable have no audible effect. I take no position on that, I this post. What I am saying is that any audible improvement from a “USB cleanser” is not caused by jitter reduction.
If you want to see the effect of a femtoclock ISB cleanser on a modern DAC, don’t look at this diagram because these DACs are fed straight from a PC (Audio Science Review):
Those are thousand dollar DACs. Here is a $400 DAC:
As I said. Reinforced my point. Show a bunch of graphs, from ASR, and all is taken care of. No need to listen, or even own any audio. You are now an expert 
You sneer at expertise?
And @Edwin_Zargarian talked of street smarts vs. book smarts?
Can I recommend Frankie’s Tax Returns and Heart Surgery down at the gas station?
My Mac mini was previously connected to my DAC via USB, and I was absolutely certain that there would be no change to the sound quality. Much to my surprise, using a streamer via ethernet absolutely transformed my system. It was like I just spent a few thousand dollars to upgrade it.
Anders,
It has been pointed out repeatedly on this forum that physical layer jitter in network protocol audio transport is irrelevant. Your explanation of why this is true is truly comprehensive and accurate, but I fear it will not be accepted by the “audio jitter and noise is bad” camp as they seem to have no appreciation for the levels of abstraction involved. It seems they view network protocols as just another physically different SPDIF link.
Thanks for trying though 
What’s wrong with John’s Swenson description in his white paper ?
The OSI was never mentioned. In any case all data has to pass by the PHY, right ? So clock jitter and phase noise will be equal on any layer, right ?
@R1200CL thank you for the excellent pointers. What Superdad had quoted in the introduction is so true
The paper does point out what potentially can go wrong in the digital domain. It is particularly important like you had pointed that PHY receives and handles bits, all sort of things can influence digital signals, change the way how they are interpreted, before they enter the DAC chip, and after as well. The challenge perhaps is to put a system together within a reasonable budget.
Clock jitter will not pass up the abstraction of the OSI layers. If it did, the bits would come out wrong. This is a reality with spdif, but not ethernet. This is where bits are bits.
Noise will pass through the signal and possibly affect other connected analog systems. This is where the bits can be the same yet your DAC is going to produce different analog waves.
Jitter does not matter on ethernet. It can’t, because unlike systems where jitter can affect the audio (spdif), there is error detection and correction on TCP/IP/Ethernet. If the jitter was so bad to mess up the data, it’d just create a retransmit until it got it right.
This is in alignment with Swenson’s white paper.
Sorry dude, you are wrong and the paper you sited does not say what you think it does. Swenson is saying exactly what I am. Jitter only affects going to analog to digital or digital to analog. And it can’t creep up the OSI layer.
The other day I measured various voltages internally to the PC where Roon lives and what was disturbing was the chassis connections for say the HDD and the PSU case were several hundred millivolts and should be zero.
The shell of the USB connector is also connected to ‘PC chassis’ and should be at the same potential as the 0V but they were not.
Clearly this voltage of crud and noise can and does connect to the DAC at the other end of the PC. A circuit is made within the DAC components back to the PC, the current always flows back to the source.
Unless the noise is filtered in the DAC or along the way with various boxes/cables, this noise is added to the DAC’s processing and the output will be affected by coupling the noise to the power supply within the DAC.
This is fundamental how voltages and currents work, so unless the source shuts up the noise, the downstream components must deal with it, to date a feat that is difficult to resolve.
Data corruption manifests itself as ticks and pops, IIRC the wrong cable (far too long, >2m and capacitance too high) can cause these problems as well as setting buffers too low a value.
So the differences noted by various cables and USB fixers is is due to the differing electrical characteristics of devices acting as a rather (poor) filter between the source and the DAC. Not a data, OS issue unless deliberate EQ or DSP is involved.
