jabbr

Oh believe me I started down this rabbit hole chasing ground loops! at least 99.9% of the benefit of fiber is the absence of ground loops ... but yeah fiberoptic is standard tool in the modern networking armamentarium. Its not that there isn't noise in opto-electronic conversion, its that noise from the rest of the system doesn't pass over a modern fiberoptic network. If you wish, the firm dividing line where upstream noise ends is where the glass fiber ends. That's essentially the end of the network and apropos this topic, we need if anything better DAC isolation, or signal filtering, or corrections in the digital domain, or multichannel synthesis etc. these are the areas where advanced in our hallucination of "being there" will occur.

Also let me reinforce that the DAC and anything directly connected to it either input or power supply is fair game as part of the audio specific system! My point however is that there are established ways to essentially 100% block noise transmission i.e. fiber cables, modern reclocking chips, clean power, good design. With proper design the idea that upstream noise from the disk drives of a NAS somehow makes its way into the DAC is complete fiction

Exactly and solved by enough bandwidth as well as protocol selection. Ravenna is almost the sole example of audio specific network hardware but come to think of it the very modern NICs e.g. NVidia/Mellanox and AMD/Xilinx are themselves programmable and Ravenna could be programmed onto these NICs if there were sufficient interest ... Ravenna is at the firmware level ... the trend is that what was hardware a few years ago becomes software/firmware today. I will say that using Ravenna in a concert hall with multiple distributed amps/speakers almost proves my statement that everything can be done with standard network equipment and specialized drivers/software. I do think that increasingly multichannel arrays of speakers with convolutions of spatial filtering as well as equipment corrections will be the future (mostly software). In the same way that post 1999 design ethernet receivers must handle imperfect inputs without increasing noise, DACs need to handle slightly imperfect inputs.

Sure, I mean first of all not all commercially available standard networking equipment is the same, nor all servers but that said the modern networking equipment has been designed to work with standard power supplies that plug into the wall. Fairly extraordinary noise elimination that happens on the boards and within the chips themselves. Extraordinary as in check out the eye pattern of a 200Gbe lane. Yes the DAC needs to be designed in the same way that all modern networking equipment is designed. It needs to lock onto an async clock domain and smooth out an imperfect input signal. This is done in every modern ethernet device. The DAC needs to be designed to either accept a specified clean external power source or clean up its own AC signal. But yes the DAC is an audio device and should be in a clean power environment. The DAC also needs to deal with filtering out high frequency digital signals/noise e.g. DSD ... that's how it works! The concept that noise flows down an Ethernet chain comes from the past centrury and was dealt with in the 1990s. Just use modern equipmemt which is explicitly designed to prevent that. I'm not talking about $5 ethernet switches with $3 power supplies.

Everything in the DAC and after the DAC matters. That's the only place we should really need bespoke hardware. Every thing else in audio can be done with software and commercial hardware.

Meh, not at all irrelevant. At 10GbE which itself is old, you can transmit an entire CD in 1 second, so I mean a DAC could buffer the entire CD and the network is off most of the time. At at more modern 100GbE network thats way less than the time between tracks. At 200GbE the network is as fast as an NVME drive. So optimize the DAC with a buffer.

Nah ... you been saying this for a decade without showing us a single measurement ... yeah I know its coming tomorrow. The fact is that there are tons of actual measurements showing this isn't the case with a properly designed/modern network. Modern Ethernet networks do not transmit noise from the NAS hard drives to the DAC. Its been proven over and over and over and over. Maybe idk with $5 switches but really not with modern well made commercial Ethernet equipment. Just not. Modern ethernet networks have extraordinarily low noise -- measured not just speculation. Its kinda like why I didn;t get that McLaren - racing cars need to be driven on smooth roads, not the pothole marked roads where I live. Modern networks have very smooth roads. Nvidia depends on this.

@Jud DAC has bluetooth output so yes my tongue in cheek comment addressed that. The actual point of course is that convolutions have the ability to apply corrections to the system whatever that might be. @Miska can correct me but I'm assuming that DAC corrections work as a measured convolution kernel.

Now if we can make the AirPods sound like the Sennhauser 800s then this will be a huge accomplishment! I'm saying this only slightly tongue in cheek because this area of processing has the possibility of vast improvements in our ability to get "there" in terms of the live or in studio hallucination. I can actually here the differences in modulators and filters as opposed to various widgets.

I'm crazy enough to have those vibration isolation cups under them

I our ongoing series of state of the art, yet never needed for home networking, devices, I present a ConnectX-7 200GbE NIC: https://www.servethehome.com/nvidia-connectx-7-ocp-nic-3-0-review-2-port-200gbe-and-ndr200-ib-mellanox/ What strikes me as incredible is that it matches the bandwidth of a local NVME SSD. What is also incredible is that they work with bog standard PSUs without causing the bits to get blown everywhere by stray voltage spikes.

Honestly if you power your audio area in an isolated fashion e.g. batteries but I use a heavy iron balanced transformer myself then there is zero need to power your network equipment via battery: glass doesn't transmit electricity.

I can't say what you are hearing, the possible causes approach infinite. Nor have I don't any listening myself to the effects of fiberoptic on what is ultimately a Wifi connection. Really the possibilities are such that I couldn't do a listening test and come to any conclusion for myself. If you are happy with the sound of your own system that's great. To the extent that you are making "conclusions" about fiberoptic x,y,z I will step in with data to the extent I have it. My recommendations are based upon my research and in rare cases my own listening experiences. If you are certain that the 10G switch does not eliminate upstream noise in your system, then you should compare different 10G+ switches yourself because these have different implementations. I have recommended Mikrotik only because it is cost effective and fanless. There are many other switches. I have tested a variety of 10-100G devices myself over the years and can't tell an SQ difference, but if your ears are vastly better than mine you should do your own comprehensive evaluation. I have ansolutely zero confidence in my own ears ability to detect femtosecond differences in jitter nor nanovolt differences in ground plane noise so yes I do defer to published measurements to make my assertions. Its not passion, nor theory.

To be clear it’s not a theory, the specification mandates compliance testing whereby noise is injected into the upstream signal and the noise needs to be eliminated by the receiver. true a model might not have been rigorously tested — none of the 1G Ethernet devices need have this testing done

There are actual measurements. Upstream noise isn’t transmitted. Whatever SQ differences you hear are not via the Ethernet stream transmitted along 10G+ fiber and most certainly not at 100G. what you hear is via a different mechanism and there are several possibilities. some of the possibilities are electrical.