Hi JonoB, Simon,
Lets take a typical network installation found in a lot a UK houses. Many properties will receive their broadband internet feed from their ISP via cable modem. These modems utilise DOCSIS which is the interface standard technology that determines how a modem receives Internet signals from the cable ISP and translates them into the Internet service used in the home. There are 2 DOCSIS standards, the older 3.0 and the more modern 3.1. One of the more important specs for DOCSYS is the channel bonding number ie how many channels can operate simultaneously to achieve maximum speed. DOCSIS 3.0 may have 8X4 (8 upload, 4 download) which limits speed to 100Mbps (I believe), 16x8, 24x8, or 32x8 (for 1Gbps max). DOCSIS 3.1 has 32x8 to achieve its max speed of 10Gbps.
The DOCSIS interface converts the incoming streams to the Ethernet standard used in the home. This conversion process requires power (usually 12V) and creates noise, as does any process that performs work. If you are interested to understand the structure of ethernet data packets, hereās a link
For a music file you can imagine how much processing is required for a standard 16Bit 44.1 KHz CD. Now add all the other household internet traffic and maybe you begin to understand why limiting the amount of traffic on your Audio network can remove large sources of processing noise of all forms and reduce demands on power supplies etc.
Ethernet packages are output from the Modem via an RJ45 plug/socket. This interface also requires power and also generates a degree of noise. The ethernet packages may now go to a router. Routers based on the Intel Puma chipset have had serious bugs in the firmware that would cause dramatic rises in latency when traffic volume increases. For audio, routers based on a Broadcom chipset avoid these issues and sound a lot better. Within the router, ethernet packages, currently in voltage form, are re-formatted as Wi-Fi frames, converted to a radio signal and transmitted via built-in antennae.
Again a lot of processing noise is generated, along with transmission noise, as with any radio signal. Somewhere within the range of the transmission another antennae, maybe from a network bridge picks up the signal and the bridge translates the radio signal back to an ethernet-package-formatted voltage (using whatever it has for a power supply)ā¦.again with accompanying noise. The newly generated ethernet package stream is probably now sent via RJ45 ethernet cable to a switch. The switch usually requires 12V DC so is powered by its own supply. A degree of filtering is applied to the incoming signal. Central to the switch architecture is the switching chip, which itself produces noise, some more that others. Optimized-for-audio switches like for example the Innuos PhoenixNET has highly optimised power supplies, switching chip selected for its low noise characteristics, extremely accurate oscillators and audio grade magnetics for the RJ45 isolation. Potential sources of RFI ingress are screened, internally generated EMI is minimised by design and component selection and all sensitive electronics are isolated from vibration. From the switch, the data stream may then travel via an RJ45 ethernet cable to the server, where again noise can be removed and minimised. The less traffic that has travelled though the network, the less work has been done by all network components, the less noise of all types is included with and within the bitstream. The server has a major impact on sound quality, depending on how much noise it can remove and how clean and well timed a data stream it can output. The server may output an ethernet packet stream or may convert the incoming ethernet to the USB standard for transmission directly to the DAC. What sounds like a trivial conversion can have a major impact on sound quality. Things like clock accuracy and power supply both play exceptionally important roles.
With the above in mind, the less network traffic processed, the lower the noise generated. The better the power supplies, the lower the noise level and the higher quality all the conversion steps and voltage structures. The more accurate and precise the conversion steps, the better the downstream processes run. The better the network operates, the higher the quality of the data stream that reaches the DAC. The higher quality the DACās input, the better its output. Based on what Iāve learned about networks related to sound quality, power supplies, cables, oscillator accuracy, vibration control, EMI and RFI control all matter. The better you can get every module of your network to operate and the more noise OF EVERY TYPE that you can take out of the system or avoid, the better the final music quality will be.
Audiophile network devices seek to minimize noise of all types, which leads to a higher standard of musical presentation.
My guess is that audiophiles the World over would be utterly stunned at just how amazing a completely noise-free digital source would sound. Unfortunately such a thing is likely impossible to achieve but the potential for improving sound quality by reducing noise is still far greater than most audiophiles would imagine. This is why so many companies are designing, building and successfully selling devices to achieve major improvements in this regard.