Question on audiophile Ethernet cables and switches


I understand that cable construction and EFI/EMI conditions can affect the quality of an analog signal carried on speaker, interconnects, and Ethernet cables. In the Ethernet case, poor quality can result in bit errors which result in frame re-transmissions.
I am having a hard time understanding how audiophile Ethernet cables and switches affect sound quality.
The typical Ethernet controller has two physical interfaces. One to the system bus for reading and writing frame contents to system memory. The other physical interface is to the transceiver for transforming an analog signal to/from bit(s).
A generic flow for reading a frame: Ethernet controller receives a read request via the system bus. The read request will contain the address in the main memory for the frame contents. The Ethernet controller engages the transceiver to receive bits. These bits are assembled into an Ethernet frame. Once a frame is identified the CRC is validated to ensure there are no bit errors. The controller will also perform MAC filtering to ensure that the frame is destined for this machine. The frame contents are then written to system memory. Once the contents are completely written to memory the operating system is notified that a new frame has been read and the next process in the chain takes over. In this scenario, there is no jitter between bits in a received frame.
I have a separate music LAN which uses cheap CAT-5E and the cheapest gigabit Ethernet switch I could find. In monitoring my music server I find there are no Ethernet frame errors or re-transmissions on the transmit or receive sides.
So can someone explain with technical details how an audiophile Ethernet cable or switch would affect sound quality?
128x128welcher
welcher,  time jitter of D/A conversion basically adds noise in frequency domain.  Ethernet data cannot directly introduce jitter, since data delivered is asynchronous to D/A conversion clock and is placed in the buffer.  Ethernet cable, as any other cable injects electrical noise - computer/router electrical noise and induced ambient electrical noise.  This electrical noise is most likely RF, well above audio band, but product of modulation of such frequencies on any non-linear element might become audible.  
Injected noise might also add jitter to internal D/A converter clock.  This clock is crystal based accurate and stable, but electrical noise will affect its edges, either internally (before squaring/buffering) or it will add to output edges (that have limited slew rate).  Noise added to transition will produce time uncertainty of threshold point - a jitter.  Time jitter of digital conversion clock will add sidebands to root signal.   Pure 1kHz tone with added 120Hz jitter will produce 880Hz, 1000Hz and 1120Hz frequencies.  There will be more sidebands, but much smaller and practically only first two count.  They have very small level, well below -60dB (since amplitude of time jitter is very small), but are still very audible not being harmonically related to root frequency.  Such jitter will be more audible when correlated to particular noise frequency and less audible when this frequency is low (sidebands are close to root) or with multiple frequencies (uncorrelated jitter).

Ethernet has transformer isolation (or optocouplers) as part of the IEEE 802.3 standard, but there is always small capacitance across, that will conduct high frequency noise (like switching noise).
If the measurements on the analog side of the DAC on IMD and jitter are below human threshold of what's audible I wouldn't worry about it. There are a lot of what's and maybe but the problem of DACs dealing with these issues isn't a concern anymore unless your DAC is poorly made.