I'm exploring this topic on a forum concerned with such matters as well as audiophile switches, special ethernet cables, etc. My thesis is that such things probably do introduce some kind of electrical noise, but even if they do it doesn't work in such a way as to mangle all those little tiny bits. I used ChatGBT 5.2 to propose a simple test method to see if any of the stuff upstream of the DAC output can carry through the system in a way that moves the noise floor. It's kind of brute force method but seems to make sense. Please do have a look and see if you agree. Suggestions and comments welcome! Thanks and cheers,
--------------------------------------------------------------------
ChatGBT:
Equipment required:
What is measured:
"
--------------------------------------------------------------------
ChatGBT:
Equipment required:
- DAC or streamer under test with analog outputs
- Low-noise audio interface / ADC (RME, MOTU, Focusrite-class or better)
- Computer running REW
- Ethernet switch and Ethernet cables (UTP and optional STP)
- Optional: fiber media converters, battery or alternative supply for the switch
- Connect the DAC’s analog output directly to the audio interface line input
- Disable any DSP, volume normalization, or software gain changes
- Set a fixed gain on the audio interface and do not change it
- Stream digital silence (zero signal) so the DAC is active
- Use RTA or Spectrum mode (no sweep)
- FFT length: 65k or 128k
- Window: Hann
- Averaging: Infinite (or ≥30 seconds)
- Frequency range: 0–96 kHz (or interface limit)
- Vertical scale: fixed, same scale for all measurements
- Disable autoscaling and smoothing
What is measured:
- Broadband noise floor at the DAC’s analog output
- Presence or absence of discrete tones (hum, switching noise, HF hash)
- Changes in spectral shape rather than single-bin fluctuations
- Ethernet connected vs disconnected
- Switch powered on vs powered off
- UTP vs STP cable
- Copper Ethernet vs fiber Ethernet
- Switch on same AC circuit vs different circuit
- Optional: switch powered from battery or alternative supply
- Repeat the FFT measurement under each condition
- Overlay or directly compare spectra using identical scaling
- If noise level or spectral content changes repeatably when an Ethernet condition changes, the noise is said to track that condition
- If spectra are identical within measurement repeatability, Ethernet is not contributing to analog noise
- Autoscaled plots
- Single-shot measurements
- Differences below the measurement system’s noise floor
- Subjective listening impressions without correlated output changes
- No change at analog output → Ethernet irrelevant to audio
- Changes correlated with shielding or grounding → grounding issue
- Changes visible on analog power rails → internal isolation or design issue
"
