Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [802.3af] clocking over ethernet


NTP can be used to sync up a local clock, of course, but jitter/wander
(phase discontinuity?) remain troublesome for many applications.

I exchanged e-mail with Joerg.  It seems like Joerg needs a
microsecond-epoch transition for his application.  The .af group's
fastest transitions are several orders of magnitude slower than that,
and besides, timing pulses are not in the .af group charter.  (Remember,
we couldn't even agree on an AC detection scheme ;->  )

That being said, I agree with you about this value of this idea.  A
pleisiochronous embedded clock (say, in the Ethernet preamble) would do
wonders for synchronous application development.  Combine that with .af
power, and you have a whole new world of products you could build.

What's called for here is a whole new study group:  how to add
pleisiochronous clock references to existing PHYs.  I did that sort of
work myself several years ago when working on ATM UNIs.

I am overcommitted these days, so I can't volunteer to run this.  But
it's a great idea.


Steve Jackson

* * *

Raymond Gass wrote:
> Steve,
> the problem I see with NTP is jitter and wander that makes
> it not really appropriate to generate a sync pulse
> I like the proposal from Joerg; should be investigated
> regards
> Raymond