Re: 16-bit 625Mbaud XGMII
Curt Berg wrote:
> Hi Jaime,
> If I recall correctly, last time a 16 bit interface
> was proposed, it was a differential interface. This would
> take about the same amount of pins as a single ended 32 bit XGMII.
> (Anyone know about a single ended PMA to PMD interface?)
> For design simplicity a single ended 312.5 Mbaud is very much preferred.
> -Curt Berg-
> Extreme Networks
You are rigth: the 16-bit interface would by of the LVDS-type.
Sorry: I made a mistake.
I do mind defining in the standard a XAUI that is not a
simple extender, with a XGMII at its input and output.
A XAUI that has a XGMII in only one side (the MAC) creates a
compatibility and interoperability problem since the other
side outputs encoded symbols, and everyone wants a
different coder (8b/10b, 64b/66b,PAM-5, etc) for the PMD.
in the optical fiber.
On the other hand, a XAUI as a simple extender, with a
XGMII on both sides, is of little value to anyone: adding
one code on one side and then removing the coding in the
other in order to insert the final coding that will be used
in the PMD is useless. In the absense of a clear
winning code over Copper, (and the presentations
and discussions on the Reflector appear to indicate that this
is the case and the various coding schemes on Copper seem
to give reasonable eye openings - I ignore, for simplicity,
packaging , EMI considerations, cost and maximum length),
it makes more sense to use on the Copper medium the same
coder as the one to be used on the optical fiber medium
and avoid the multiplicity and complexity of coding/decoding
in the XAUI with one coder and again coding for the PMD
with another coding scheme. After all, it is much
more important to use the optimum coding in the fiber, where
we do not have spare SNR to play with.
XAUI was introduced to solve the problem of sending
10 Gbps over long distances on a Copper backplane.
It is really a PHY for 10 Gbps over Copper. Spending
a significant time to find out what is the best coding
over Copper might not be the right direction. It should not
be part of the 10 GbE standard that deals with finding
the optimum coding/PMDs for optical fiber.
The only specified interface should be the XGMII.
Jaime E. Kardontchik
San Jose, CA 95131