@Cloclo … I think mains service power/ground/earthing quality is most important in concert with optical isolation… Also mitigating potentials for RFI and EMI being introduced into the device topologies… this is why I employ ferrite beads on power cables…
We have to take this investigation into account and apply this to our subjective experiences, when cognitive bias has been tempered by the more reasonable familiarity with our playback system sound-quality in our environment.
I presume that he already has an audio system with a decent power, and ground quality. He is not a beginner. He just didn’t know how to set an optical isolation for his endpoint.
@Cloclo @bitracer …This should be on the same power distribution of your computer… Any ground-loop is eliminated by the optical interface and DC conversion… You don’t want to introduce noise back into the Ethernet port and into the topology of the computer via a ground-loop from the transmitting media converter… so the computer and optical media converter should be sharing the same ground/earth point…
If you use a single computer as a server, and you have more than one endpoint in your home, you need an optical isolation for each of them. And you can not power all the media converters on the same power distribution as your computer, because they are not in the same room.
If noise enters inside the computer, it’s not an issue, like with a direct USB connection to a DAC. Because, all the noise that will be streamed by the computer to the endpoint will be filtered by the media converter during the conversion of the Ethernet packages from electrical to optical.
In my home, I have a special setting with two servers, and a Host. They are all connected to the same switch, and this switch has a SFP module. This module is connected by fiber to the EtherRegen of my endpoint.
If I, or another family member, stream to another endpoint, in another room, the sound is channeled from the server through the LAN by electrical Ethernet up to the media converter of the required endpoint. Then, the media converter of the endpoint filters the noise, and streams the data through fiber to the “clean zone” of the endpoint with the same kind of cheap converters that I advised @bitracer to buy.
Think of the first media converter optical interface connected to the computer as the two being integrated together… functioning as one unit. therefore sharing the same ground/earth point…My point is regarding noise migrating back into the computer topology of the data bus and power topology of the computer architecture that is computing and handling the data before it gets transmitted on the output buses…
What you suggested is what I’m actually doing for my own endpoint.
It is the only endpoint that really matters, because the other family members listen to music, and don’t care as much about SQ as I do.
The LAN is organized like that also because I’m in charge of the computers and the LAN for everybody at home.
Nevertheless, if an endpoint benefits from an optical isolation, the electrical noise from the server, the LAN or whatever… does not matter at all. Because the conversion of the DATA from the analog electrical current that channels it to the digital rays of light removes all the electrical noise that was present in the electrical current.
The media converter converts only the packages of data to rays of light. It does not convert the electrical noise, because the noise is not data.
it’s not about transmitted noise along the transmission network… it’s about noise that affects the computational performance of the computer architecture and output buses (intrinsic jitter) and the integrity of the data presented to the output buses… The DAC will only respond to the data it receives…
There’s no risk at all for the integrity of the data. Whatever is the level of the electrical noise, there can not be loss of data, neither inside the computer nor in the Ethernet network.
But jitter does matter. That’s why I set the servers near my own endpoint, and they stream through the minimal number of switches, and not through the router. (Actually, they stream through a single switch, the one to which they are connected and who converts the signal to optical.)
The flat is already wired with copper. If I had to set its network a new, I would have done it with fiber.
That’s not really true, jitter can not be transmitted through the switches. On every leg the signal is regenerated. So the total jitter is at most what’s generated on the last leg before reaching the target device. Even then there is buffering at play, so the jitter is nullified and at the end depends on the internal design of the streamer, the clock used, etc.
I know that in theory, you are completely right. But in practice, the jitter may be reduced by the switches but not nullified.
What is the justification for these expensive “audiophile” supposed switches?
To clock better the incoming Ethernet stream, and to reduce the jitter.
If the generic switches were able to nullify the jitter, what’s the benefit of an “audiophile” grade switch?
And if the “audiophile” switch was able to nullify the jitter, why am I hearing an improved sound when I use it with an external 10 MHz reference clock?
Even then, it’s only on the last leg. It doesn’t matter how many switches there are before it. I’m not sold on this, but if you invest couple thousand bucks on a streamer, why not. It doesn’t hurt.
I don’t think that it’s only the last leg that matters.
As a last leg, after the streamer, and just before the DAC, I have a very good digital interface with the best available word clocks. And it clocks the I2S sound.
And yet, the sound is still improved when I connect the EtherRegen to the external 10 MHz reference clock, and the EtherRegen is located before the streamer.
I meant last leg of Ethernet as far as networking technology is concerned.
I understand.
And the sound with the switch that is the last leg, just before the streamer, still benefits from a reference external clock. That means that eliminating the jitter of the Ethernet stream is not that trivial.
I can’t really confirm that, but I would expect that it would also depend on the streamer. Similar to the USB.
If the streamer has an input for a 10 MHz reference signal, the sound is improved further.
But it’s the same thing even for the DAC.
EDIT
The sound with my digital interface is also improved with the external 10 MHz reference. Its word clocks are more efficient.
@Cloclo… It is not about loss of data… it is about the distortion of the data “reads” and “writes” (the interpretation of the “ON” and “Off” state of the individual bit in the word-stream) along the data-path, before it is delivered to the output buses of the computer, and when the data reaches the bus controller(s)…
The value of an external clock is that by nature of it’s implementation… the external device power topology is self reliant and not influenced by ripples in the power topology that would otherwise be precipitated by computational and input and output interface power dynamics… therefore providing a stable clock constant… Using a common clock reference for symbiotically related digital-audio components like a DAC and a SACD player or a DDC and a DAC makes sense… If a DAC design does not account for dynamic power demands of the CPU and input and output interfaces, then maybe the money is better spent on a better DAC component…
@bitracer is correct, that jitter in the Ethernet transmission is a non-issue due to the error-correction implementation and caching/buffering nature of the Ethernet protocol… It is probably overkill to use external clocking on a switch, if the DAC has a port for an external clocking device.
I’m not sure to understand.
When you stream, the computer does not render the sound. It is rendered by the network player, and in the case of my own endpoint, the Host renders it.
The computer streams the bit-perfect data of the file or the data of the file to which you applied DSP. But it is not a rendered sound, and there could be no distortion.
Furthermore, unlike other computing devices, switches spread a limited quantity of noise through the RJ45 cable, because they have a built-in galvanic isolation. The media converters also have a built-in galvanic isolation. Their galvanic isolation is necessary to regenerate the streamed data, so it won’t be buried by the electrical noise of the LAN.
These devices can not overwhelm a computer with noise.
Anyway, I will give you more details about my setup.
The servers are a Mini M1 and a i5 NUC that runs Windows 10. They are plugged to the same power strip, and are connected to the switch with ordinary RJ45 Cat.6 cables.
The Host is an i5 NUC that runs GentooPlayer + Diretta Host, and it is powered by a LPS of 19V. Its LPS is plugged in one of my two strips for the audio gear. I use a LPS for the Host, because it renders the sound for my own endpoint, and because powering it with a LPS brings a small sound improvement.
I bought this LPS long ago for another configuration of the setup. Otherwise, it would have been a luxury to buy it for the Host, because the improvement that it produces is quite small.
The switch with the SFP module to which the servers and the Host are connected is also powered by a LPS that I already had. This LPS has two outputs of 12V. I use one of them to power the switch, and the other to power EtherRegen that is located on the other side of the fiber.
This LPS is also plugged to one of the strips of the audio gear.
Clocking is a matter that I don’t know very well, because I’m not very experienced in it. It is a quite complex matter, because there are different kinds of clocks (reference and word), there are different wave forms (sine and square)… and I’m studying it.
Anyway, the external reference clock, in tests that other people did and shared their feedback, and that were confirmed by my own tests, benefits to the performance of each device of the setup. Even if it is the only device of the setup to be clocked.
The quality of the power supply of the device is always important, but I don’t think that it is the main reason for which a reference clock benefits to the performance of the device. For instance, my digital interface is powered by a regenerative regulated power supply. And yet, a reference clock improves the performance of its word clocks.
And DACs who have an input for a 10 MHz reference signal are usually good products with good power supplies. I don’t have yet a DAC with such input, but I’m planning to replace mine with such a model in the future.
More experienced users in clocking than me say that in their system, the device that benefits the most to the SQ from the external 10 MHz reference clock is the DAC. The second one is the DDC (digital interface). The third one is the streamer. And the last one is the switch.
This contradicts the rationale for using optical fiber to eliminate potential ground-loop inducted noise between components… You are connecting the chassis of both the switch and the EtherRegen by using the common earthing of the LPS… If you are looking to create a power-island for the endpoint components, you would have those components on their own isolated power/ground plane…