Ethernet

Ethernet offers galvanic isolation by design.
All connections are transformed coupled.
However this does not guarantee that no noise at all might creep into a networked device.

 

The most obvious remedy would be total galvanic isolation.
WiFi is a nice example, no physical connection at all.
Some argue that because of the overhead required for WiFi to run, you get an even higher interface noise.

 

Ethernet can be done over optical as well.

Optical of course provides perfect galvanic isolation.
Today an optical converter won’t break the bank.

 

TP Link

TP-Link Gigabit Ethernet Media Converter, Up to 1000Mbps RJ45 to 1000M Multi-Mode SC Fiber $46.99

 

Note that the output is Ethernet hence electrical by design.
Any reason to assume this port produces a more pure Ethernet than any other normal port on a switch?
Don’t worry, there are audiophile switches as well.

 

AQ-SWITCH-8 Audiophile High-End Network Switch 398,00 EUR

Noise isolation

As streaming audio has become popular, it is easy to find raving reviews about all kind of products providing some kind of “isolation”.
The TP Link is reviewed here.

As usual, it is a sighted tests, objective measurements are rare.

 

Archimago is one of the few sources of objective measurements.

 

In his setup the 60 Hz hum disappeared when removing the network cable.
Indeed, despite the galvanic isolation as provided by Ethernet, some noise slips through

Notice this hum is at -130 dBFS. One might wonder if this is audible at all.
The green line is the same system running WiFi for networking.
No indication of added noise because of the WiFi.

Ethernet isolators

 

Although Ethernet is galvanic isolated by design, there are medical grade isolators.
They offer a higher protection against surges.

EMO systems medical grade (IEC 60601-1) isolator - $204

 

Products like these offer  6 kV AC dielectric strength and 8.5 kV DC dielectric strength.

The IEE 802.3 says:

a) 1500 Vrms at 50 Hz to 60 Hz for 60 s, applied as specified in Section
5.3.2 of IEC 60950: 1991.
b) 2250 Vdc for 60 s, applied as specified in Section 5.3.2 of IEC 60950:
1991.
c) A sequence of ten 2400 V impulses of alternating polarity, applied at
intervals of not less than 1 s. The shape of the impulses shall be 1.2/50 µs
(1.2 µs virtual front time, 50 µs virtual time of half value), as defined in IEC
60060. There shall be no insulation breakdown, as defined in Section 5.3.2
of IEC 60950: 1991, during the test. The resistance after the test shall be at
least 2 MW, measured at 500 Vdc.

 

Medical grade isolators offer a far higher protection compared with the standard.

 

Amir tested the EMO EN-60KDS.

 

No differences where found.
After wrapping the network  cable around a powercord, he noticed that a small 60 Hz spike disappeared when using the isolator.

Using a Shitt Modi2 also some small improvements in the jitter performance where found.

 

 

Depending on the DAC there can be a measurable improvement using a high grade isolator.
We talk about measurable effects at -128 dBfs.
As one forum  member pointed out:
128dB = 792.6 billionths of a volt, assuming a reference signal level of 2 volts.

One might wonder is such small fluctuations are audible at all.

Conclusion

It looks like Ethernet induced noise is a very small if not a insignificant problem.

Result will vary with the design quality of the DAC.

References

  1. On the value for ethernet "galvanic isolation" - Archimago's Musings
  2. Review and Measurements of EMO EN-60KDS Ethernet Isolator - Amir
  3. The Intercontinental Internet Audio Streaming Test - Archimago's Musings