I started a new thread cause this is unrelated. But you really got me
to thinkin' ....
What you mention in item 1 below is not exactly true, right. The
spectral content of the signal will have harmonics and possibly several
different velocity groups based on the channel phase effects. So where
am I going with this ... well, this being true for any speed and code
scheme, the trace geometry and via geometry could certainly add a
considerable amount of phase, depending on the geometry used. If you
look at what Dan Dove suggested in another thread regarding some kind of
xgbic device and a connector, a 2.5gbit speed would allow for ( and I
don't know that we would need it ) or be more tolerable of attenuation
and phase delay in the geometry and the connector scheme.
Have to think about this one some more. I keep having dreams of the
'killer via' stocking me.
A couple of minor points on XAUI:
1) Speed: XAUI serial lanes are self-timed (i.e. no clock accompanies
in either direction) and operate at 3.125 Gbaud. Because of 8B/10B
highest frequency observed on PCB traces is 1.5625 GHz, not 3 GHz.
2) Pin Count: It's exactly 16. 4 in each of 2 data directions,
Member of Technical Staff
High Speed Communications VLSI Research
Bell Laboratories, Lucent Technologies
10250 Valley View, Suite 113
Eden Prairie, MN, 55344
Direct: (612) 996-6932
Cell: (612) 670-5930
Fax: (612) 996-6995