klettermann
Senior Member
In some circles the rage seems to be replacing normal 1Gb wired ethernet with a 10Gb optical system. Allegedly this prevents dreaded internet noise, RF, etc etc from getting to the streamer, polluting the signal and degrading the SQ. Seems like a logical next step if you think Etherregen is a good idea. So, I recently ran a small before/after test to see whether changing part of my playback network from copper Ethernet to optical Ethernet produced any measurable change at the DAC analog output.
This is not intended as a universal proof that Ethernet, switches, SFPs, optical links, power supplies, or network topology can never matter. It is a practical consequence test in one system. The question was simple: If changing copper Ethernet to optical Ethernet matters in this playback chain, does any repeatable difference appear at the DAC analog output?
System path:
Server / local FLAC files → network switches → Lumin U1 Mini → miniDSP SHD Studio → Schiit Yggdrasil LiM → RCA analog output → Focusrite Scarlett 16i16 → REW
The system uses the miniDSP SHD Studio for crossover, sub integration, and EQ. The DAC measured was the Ygg’s RCA output. The comparison was between my normal wired Ethernet topology and an optical/fiber topology inserted into the relevant network legs feeding the audio chain. The analog measurement setup was unchanged between tests: same DAC output, same RCA cable, same Scarlett input, same gain, same REW settings.
Test signals:
REW setup:
Focusrite ASIO driver
44.1 kHz / 24-bit capture
64k FFT
Hann window
75% overlap
No smoothing
Forever averaging
Results:
For the 1 kHz -12 dBFS tone, the wired and optical measurements were identical to the displayed precision in REW. Across repeated runs, the fundamental, noise, N+D, SNR, THD, THD+N, and ENOB values remained the same within the displayed resolution. The visible spur pattern was also effectively unchanged.
For digital silence, both wired and optical again produced the same practical result. The broadband floor and small mains-related residual components were stable and did not track the Ethernet topology. The 60/120 Hz residuals appear to be part of the analog measurement/system grounding environment, not a copper-vs-optical network effect.
Wired Ethernet, 1 kHz -12 dBFS
Optical Ethernet, 1 kHz -12 dBFS
Wired Ethernet, digital silence
Optical Ethernet, digital silence
My conclusion is narrow:
In this system, under these test conditions, changing from copper Ethernet to optical Ethernet did not produce a repeatable measurable change at the Yggy's analog output. I also did not hear a difference.
There are several possible interpretations. One is that the relevant network-borne noise was not meaningfully present in my system, so there was nothing for optical isolation to fix. Another is that some upstream noise or leakage may have existed, but the Lumin / miniDSP / Yggy chain rejected or filtered it before it became an analog-output consequence. A third possibility is that my measurement setup or test signals did not capture the relevant mechanism. That is possible. This was not an Audio Precision measurement, and 1 kHz plus digital silence do not exhaust all possible test conditions. A wider-band 96 kHz capture, multitone testing, higher-frequency tones, or different network-stress conditions could be useful follow-ups. Still, the practical result here is straightforward. In this playback chain, I found no repeatable DAC-output difference between copper and optical Ethernet using these tests.
My reason for posting this is not to claim finality, but to encourage consequence testing. If a switch, cable, SFP module, power supply, optical link, or direct-attach cable is claimed to produce a clear sonic difference, then it should be useful to measure the DAC analog output before and after the change. It is NOT because such a measurement fully predicts sound quality. It does not. But, because if a change is audible, it's reasonable to ask whether there is any corresponding downstream physical consequence: noise floor, spurs, sidebands, ultrasonic content, modulation products, distortion, level change, or something else.
In my case, for copper vs optical Ethernet, I did not find one. Frankly I didn't expect to even if I was hoping to see something. It was a fun little project anyway!
Make of it what you will and cheers,
This is not intended as a universal proof that Ethernet, switches, SFPs, optical links, power supplies, or network topology can never matter. It is a practical consequence test in one system. The question was simple: If changing copper Ethernet to optical Ethernet matters in this playback chain, does any repeatable difference appear at the DAC analog output?
System path:
Server / local FLAC files → network switches → Lumin U1 Mini → miniDSP SHD Studio → Schiit Yggdrasil LiM → RCA analog output → Focusrite Scarlett 16i16 → REW
The system uses the miniDSP SHD Studio for crossover, sub integration, and EQ. The DAC measured was the Ygg’s RCA output. The comparison was between my normal wired Ethernet topology and an optical/fiber topology inserted into the relevant network legs feeding the audio chain. The analog measurement setup was unchanged between tests: same DAC output, same RCA cable, same Scarlett input, same gain, same REW settings.
Test signals:
- 1 kHz sine, -12 dBFS
- Digital silence
REW setup:
Focusrite ASIO driver
44.1 kHz / 24-bit capture
64k FFT
Hann window
75% overlap
No smoothing
Forever averaging
Results:
For the 1 kHz -12 dBFS tone, the wired and optical measurements were identical to the displayed precision in REW. Across repeated runs, the fundamental, noise, N+D, SNR, THD, THD+N, and ENOB values remained the same within the displayed resolution. The visible spur pattern was also effectively unchanged.
For digital silence, both wired and optical again produced the same practical result. The broadband floor and small mains-related residual components were stable and did not track the Ethernet topology. The 60/120 Hz residuals appear to be part of the analog measurement/system grounding environment, not a copper-vs-optical network effect.
Wired Ethernet, 1 kHz -12 dBFS
Optical Ethernet, 1 kHz -12 dBFS
Wired Ethernet, digital silence
Optical Ethernet, digital silence
My conclusion is narrow:
In this system, under these test conditions, changing from copper Ethernet to optical Ethernet did not produce a repeatable measurable change at the Yggy's analog output. I also did not hear a difference.
There are several possible interpretations. One is that the relevant network-borne noise was not meaningfully present in my system, so there was nothing for optical isolation to fix. Another is that some upstream noise or leakage may have existed, but the Lumin / miniDSP / Yggy chain rejected or filtered it before it became an analog-output consequence. A third possibility is that my measurement setup or test signals did not capture the relevant mechanism. That is possible. This was not an Audio Precision measurement, and 1 kHz plus digital silence do not exhaust all possible test conditions. A wider-band 96 kHz capture, multitone testing, higher-frequency tones, or different network-stress conditions could be useful follow-ups. Still, the practical result here is straightforward. In this playback chain, I found no repeatable DAC-output difference between copper and optical Ethernet using these tests.
My reason for posting this is not to claim finality, but to encourage consequence testing. If a switch, cable, SFP module, power supply, optical link, or direct-attach cable is claimed to produce a clear sonic difference, then it should be useful to measure the DAC analog output before and after the change. It is NOT because such a measurement fully predicts sound quality. It does not. But, because if a change is audible, it's reasonable to ask whether there is any corresponding downstream physical consequence: noise floor, spurs, sidebands, ultrasonic content, modulation products, distortion, level change, or something else.
In my case, for copper vs optical Ethernet, I did not find one. Frankly I didn't expect to even if I was hoping to see something. It was a fun little project anyway!
Make of it what you will and cheers,
. Receiving a 1 kHz file from a NAS over a 1 Gbps wired network to RPI5/Moode. NAS->Router->Router->Switch->Switch->RPI5/Moode->Topping DX1-II. Fiber optics are a good thing, but not in terms of audio noise. Regular wired Ethernet is also galvanically isolated, and its noise is neither visible nor audible.
