Ethernet Clocking


i had previously reported that adding an Antelope 10m rubidium clock to the Etherregen results in major tightening of soundstage and location of individual instruments. To my great surprise adding filtering on the BNC 75 Ohms connection between clock and Etherregen results in substantial additional benefits. The filter used is a Mini-Circuits BLP-10.7-75+ DCto11MHZ model.

We are only beginning to understand how to maintain clean clocking on digital connections, it is of paramount importance to SQ.

antigrunge2

Except @lmcmalo cant help but chip in.    Let’s pass the point where some vendor has a RE-clocking product for a protocol without a clock :-0

Ethernet phy layer is an asynchronous.  Very roughly it works like this. Very roughly, it’s complicated in practice.  The line goes flat, there’s a gap between packets.  Then comes a fixed start sequence, the receiver locks on it and then runs its own clock starting from that point on the negotiated speed.  It works fine because it’s very brief.   The clock has to be in sync for that  1 packet.   Because there’s CRC, errors can be detected.   Phy errors are uncommon in normal situations   And no one cares, the missing packets get retransmitted by TCP..

 
This isn’t voodoo, but I must have wasted months of my life on « SPF modules ».   Those are Ethernet phy modules that can be swapped out in fancy switches instead of RJ-45.    Lots of compatibility problems.   The beauty is that mostly they work or they don’t…   Sometimes it’s in between and then some software detects the problem.   

 

 

Post removed