[newzooreview]

I swapped out an ethernet cable in my system and heard an immediate and noticeable improvement. I have a Roon ROCK server, NAS, and streamer (Eversolo DMP-A8) all connected to the same Ubiquiti switch. I did some digging to satisfy my curiosity about why I might be hearing a difference and I thought I would write up what I found in case it helps anyone.

The idea that digital audio streaming is impervious to network infrastructure components—owing to the error-corrected nature of digital data transmission—overlooks analog-domain interactions that affect sound quality.

Network components introduce audible distortions from analog-domain interactions between electrical noise, clock integrity, and power supply contamination. While TCP/IP ensures bit-perfect data transmission, the physical realities of signal transmission and network device behavior introduce artifacts that produce audible distortions. Mitigating these effects requires a systemic approach: noise isolation, power supply optimization, and clock domain separation. The network is best treated as an analog component—rather than a neutral data conduit—to provide measurable improvements in temporal precision and analog signal integrity.

Reconstructing the Analog Waveform into 0s and 1s
Ethernet communication relies on the transmission of analog square waves, which the receiving device must reconstruct into digital bits (0s and 1s). This process occurs at the physical layer (PHY) of the network interface in the streamer, where clock recovery circuits extract timing information from the waveform’s edges. Distortions to these edges, caused by electromagnetic interference (EMI) or radio-frequency interference (RFI), force the receiver’s clock recovery mechanism to operate suboptimally, increasing timing uncertainty (jitter). Shielded Ethernet cables can mitigate (but not completely eliminate) external noise entering the cable (the antenna effect), but improper grounding—such as connecting shields at both ends—can create ground loops, allowing low-frequency noise currents to enter analog circuitry.

TCP/IP Interactions and Indirect Artifacts
The TCP/IP protocol ensures data integrity through retransmission of lost or corrupted packets, a process managed independently of the physical layer’s signal reconstruction. While corrupted Ethernet frames are discarded at the data link layer before reaching higher protocol levels, retransmission events increase network activity, elevating high-frequency noise generated by switching power supplies (SMPS) in routers and switches. This noise couples into Ethernet cables as common-mode interference, modulating the streamer’s master clock oscillator and degrading phase noise performance. Congestion control algorithms, such as TCP’s BBR, further compound timing irregularities by inducing variable latency (bufferbloat), which forces the streamer’s buffer to expand and contract, subtly altering temporal precision.

Phase Noise and Clock Destabilization Mechanisms
The stability of the streamer’s internal clock oscillator—critical for digital-to-analog conversion—is vulnerable to noise contamination from network components. Switching power supplies in consumer-grade network equipment generate broadband noise across the 10–100 kHz range, which propagates through shared ground planes and Ethernet cable shields. This noise modulates the oscillator’s power supply rails, elevating phase noise levels and increasing jitter at the digital-to-analog converter (DAC).

Mitigation Strategies
Optimizing network infrastructure to minimize analog-domain artifacts requires addressing noise generation, propagation, and coupling pathways. Optical fiber links eliminate conductive noise by decoupling the streamer from copper-based Ethernet networks. Dedicated network segments, isolated via VLAN configurations, can reduce cross-traffic-induced retransmissions and associated power supply noise. Directionally shielded Ethernet cables, grounded exclusively at the streamer end, can minimize noise ingress while avoiding ground loop formation.

Critical Factors Influencing Sound Quality
The audible impact of network-induced artifacts depends on system resolution and listener acuity. Entry-level DACs may mask subtle timing errors, but more revealing systems reproduce distortions as degraded spatial coherence, dynamic compression, and high-frequency harshness.

[newzooreview]

Given the nature of the technical issues, I am going to make some changes to how the Eversolo is connected to the network.

As a first cut, I'm going to connect an optical isolator (Dawn Technology Effe-01) to the main switch (the 20 port Ubiquiti with a bunch of stuff connected to it).

The output of the optical isolator will go to a NuPrime Omnia SW-8 switch (which has a high quality clock crystal and electrical shielding).

I'll connect the Roon ROCK server and the Eversolo DMP-A8 to the Omnia. As a start, I won't connect the NAS to the Omnia since it's using a stock switching power supply that might transfer noise into the Omnia, lowering the benefit of the optical decoupling. I might move the NAS to the Omnia switch later on to see if it helps.

The Roon ROCK server will be connected to the Omnia through a Stack Audio SmoothLAN ethernet filter. The NUC is running from a linear power supply, but it still generates internal noise. The Eversolo will be connected via a Network Acoustics Muon Pro Streaming System (cable plus Muon Pro filter).

So, I have gone completely overboard with this. Swapping the Ethernet cable and looking into the technical issues suggests that a lot of what I'm chasing in improved sound is related to getting rid of the high-frequency glare and errors in spatial cues generated in the Eversolo by poor attention to the signal path feeding it.

[Rusty Jefferson]

Charles,

You are 100% on point with your observations.  The optical break you're implementing is mandatory, but if I can make a recommendation, put your ROCK and NAS on your first switch. Put only the Bridge (Eversolo) on the second switch. Keep the Ethernet cable from the second switch as short as possible to your Bridge.

If you're still in the DMV area I have some parts/pieces I'd be happy to let you experiment with. Shoot me a PM.

[brj]

One option to consider since you already have a Unifi setup - consider the $269 USW-Aggregation.  Cheaper than many isolators, it may allow a better switch layout and connectivity options.  I have my main and office switches LAGed into the aggregation switch, which leaves 4 ports left over.  My NAS, containing my music, connects to one @ 10GbE, and the other 3 have 1GbE transceivers to serve my upstream Unifi cable modem and the two downstream audio devices - my Fitlet3 HQPlayer NAA and a media converter that fronts my subwoofer management device (though Ethernet is only used for configuration, not data/music, so it's mostly idle).

There are no switching power supplies in the audio system (behind the optical connections) because all of the low voltage devices are battery powered, including the Fitlet3 and media converter.  I can't say that I've ever done an A/B comparison of this setup to a CAT6A/SMPS setup, as I had to implement the functionality regardless, and it was just as simple to do it this way.  And SOOO much nicer to run fiber through the walls than copper!  (Frankly, the 100 Ah SLA battery, Pi Audio BatteryBUSS and DC cables were likely cheaper than getting a bunch of quality linear power supplies.  Though I did have to pay attention to grounding, on the Fitlet3 in particular.)

Just be aware that the USW-Aggregation does have some power limitations in aggregate across its optical ports.  The SFP-to-RJ45 adapters in particular are power hogs (and thus heat engines), and can quickly get you into trouble if you use more than a few of them.  I only have one of those to feed my NAS, while I have (3) DAC cables for in-the-rack connections (the most power efficient option) and Unifi SFP or SFP+ optical transceivers for the remaining (4) ports.  It's been working without issues for at least a year now, and the system sounds great. 

[newzooreview]

Thank you both. I have a bunch of gear en route, so for now I think I will focus on playing with the new items before introducing anything else.

Regarding the placement of the optical decoupler and the main switch (Ubiquiti). What do you think about connecting the second switch (Primare SW8) to the main switch using an iFi LANSilencer coming into the Primare.

So,

Ubiquiti Switch <===>iFi LANSilencer---Primare SW8

SW8
     iFi LAN Silencer <=====> NAS
     iFi LAN Silencer <=====> Roon ROCK NUC
     iFi LAN Silencer <=====> Optical Isolator (Effe-01)

Optical Isolator
     <===Muon Cable===>Muon Pro Filter<===>Streamer (Eversolo)

This seems to allow the NAS and Roon ROCK NUC to communicate with each other over the SW8 which is isolated to some degree from the main switch, then places the optical isolator in front of the streamer so the final leg is just looking at the noise from the Optical Isolator power supply and induced EMI/RFI from the cabling out of the Optical Isolator and into the streamer.

Any further thoughts appreciated since the sources of noise are understandable but the ways to best address those sources are more trial and error as far as I can tell.

[Rusty Jefferson]

Initially, I would start with the NAS and ROCK on the first switch with the fiber break to the second switch and just the Bridge on the second switch.  Very basic. Then, after listening, you could try the iFi devices (I have no experience with them) to see if there's an improvement. Then, I'd probably put a linear power supply on the second switch if it doesn't have one. From there, it's listening to a variety of fiber converters and switches if possible. They all do sound different.

Myself and most of my friends doing this are using the Sonore Optical Module Deluxe converter(s) and either Innuos or Synergistic Research switches. Zero digititus.

[newzooreview]

Thank you, again.

While waiting for equipment to arrive, I pulled out a small switch powered by an iFi Power Elite.

I connected the small switch to the Ubiquiti and then the Roon NUC, NAS, and Streamer to the small switch.

The Roon NUC did not like this at all: the Roon client on my Mac kept trying to reconnect to the Roon NUC. The only solution was to move the Roon NUC back to the main switch. When doing that, it made sense to move the NAS back to the main switch as well since running the data from the NAS through the satellite switch and back to the main switch seems unproductive.

There is no VLAN, so that might be something to investigate for the second switch, but I think when the equipment arrives I will do what you suggest:

Ubiquiti Switch <iFi LAN Silencer<=====> NAS
                      <iFi LAN Silencer<=====> Roon ROCK NUC
                      <=====>iFi LANSilencer>Primare SW8
                     
Primare SW8.  <=====>Optical Isolator

Optical Isolator<===Muon Cable===>Muon Pro Filter<===>Streamer (Eversolo)
                     
Swapping the iFi LAN Silencers in and out is something I will likely play with. I am also unsure about the benefit of the SW8 in this setup. It does have a femto clock, so it's possible it could help improve the signal going into the Optical Isolator.