As I understand it all switches receive data transmissions, store the data packets (while working out where to send it and waiting to send it), and then sent the data in a transmitted analog signal, so they have:
-
clocks, to govern the frequency of transmission of data packets, but they don’t reclock (like the K50 reclocker that outputs a ‘stream’ of data)
-
RAM/memory to store data before sending.
The intent of a ‘better’ clock is to reduce phase noise of the transmission of data packets but does not affect the actual data (eg. audio data) in those packets. There is rafts of chatter about clocks on the likes of Audiophile Style AfterDark Audio pages, mostly about external reference clocks. Uptone Audio webpage about EtherRegen states:
There are two types of sound-degrading influences the EtherREGEN is designed to radically decrease: Leakage—both high-impedance and low-impedance—and clock phase-noise. The clock phase-noise travels on the Ethernet signal itself and on power- and ground-planes. [Every signal edge coming out of any digital device carries the jitter/phase-noise of the clock used to “clock out” that edge; this shows up on the ground-plane and affects the threshold of chips’ clock inputs. This is an oversimplification of a complex subject; here is the link to our ‘white paper’ about the technical mechanisms (to be followed with some measurement proof of the effects).
(Source: EtherREGEN (original, "Gen1") – UpTone Audio)
For some back to basics, these seem good intros to switches and network architecture …