Synchronization in carrier class Ethernet

yes - but that is how it seems to work on the hardware we are using.

in practice i dont think it really matters, since we talking 100 ppm worst case, and it is only an issue if you run the link at wire speed under sustained load. we only found this when a test went over the scheduled time due to someone being called away.......

if you have Ethernet / IP applications that operate at 100% load continuously and it cannot tolerate a loss rate of 0.01% then the app is already broken - you just havent noticed yet.

Howdy Stephen,

Just curious - what you are seeing that makes it seem like that is the case?

I'd be surprised if the Ethernet frames coming out of a piece of SDH or SONET gear were timed to the SONET network. It's so much easier to just plop down a 125 MHz crystal oscillator and read the packets out of a FIFO than to try to time the output to the SONET network. I've designed commerical, carrier class Ethernet over GFP interfaces that are deployed , so I'm trying to imagine how I would do it if I were requested/required to time the output from SONET. The first thing that jumps to mind is using a DDS

Not exactly difficult, but is a more complex solution to a problem that barely exists (as you said, this only shows up at full line rate for extended periods of time).

Which, as you pointed out earlier in the thread, can happen with any gear that has an Ethernet interface (not just SONET or SDH boxes).

Marc

a conversation with an SE from the manufacturer - although thinking back he didnt state that the output was locked to the master clock, but the same meeting had some discussions about SDH clock. It may be i have added my own interpretation without sufficient reason...

so - we could be seeing the difference between 2 separate clock variations - same effect would come into play with an async output clock as well.

so worst case is + 100ppm into - 100 ppm clocks across an end to end link (or whatever the practical limits are - i suspect the manufacturers of expensive tend to stay well inside worst case)

which is the same type of problem - just twice the worst case.