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

Re: [802.3BA] XLAUI / CAUI Ad Hoc

My concern is that specifying a chip to chip implementation sets one set
of requirements focused on low signal power, low emissions, and fairly
tightly specified channel definitions.  A chip to module implementation
set of requirements focuses on slightly greater signal power, controlled
rise times, managed pre-emphasis and equalization, and a broader range
of channel definitions that include connectors.
If we succeed in a chip to chip implementation that does NOT have
the chip to module implementation requirements considered, then all
architectures will require intermediate transceivers on the motherboard
between the protocol ASIC and the module, measurably impacting
cost, complexity, reliability, and power.  XFI was focused specifically
on chip to ASIC implementations, using a module CDR to improve the jitter
budget in the tightly constrained SR environment.  As such, it may be a
good starting model.
It would be better to have no chip to chip implementation in this revision
of the standard if that implementation precludes chip to module implementations
using a single definition of ASIC I/O circuitry.
Bob Snively

From: Chris Cole [mailto:chris.cole@xxxxxxxxxxx]
Sent: Monday, December 29, 2008 6:02 PM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3BA] XLAUI / CAUI Ad Hoc

Hi Ryan,


You are right that the nAUI chip to chip specification methodology in draft 1.1 can be leveraged to build retimed modules. It just requires writing a chip to module specification somewhere which specifies test points at the modules pins. The presentation you sent is an excellent starting point for this and a lot of the XFI interface specifications can be leveraged in writing the nAUI chip to module specifications.


Best Wishes for the New Year to Everyone




From: Ryan Latchman [mailto:Ryan.Latchman@xxxxxxxxxx]
Sent: Tuesday, December 23, 2008 10:07 AM
To: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3BA] XLAUI / CAUI Ad Hoc


Hi Ali, Chris,


This is an important discussion which needs to get resolved quickly.


I would like to ensure that XLAUI / CAUI maintains its broad market applicability as a simple retimed interface.  I don’t think the current specification methodology prevents it from being leveraged to build retimed modules.  I’ve put together the attached material to show how retimed interfaces were specified in the past (namely XFI).  In XFI, you’ll notice that the Before Connector and After Connector specs are similar.  40/100GbE modules may have an analogous situation, depending on their size and electrical characteristics. 


If we need to change the XLAUI / CAUI specification, we need solid contributions on what needs to change. 


Happy Holidays,




From: Ali Ghiasi [mailto:aghiasi@xxxxxxxxxxxx]
Sent: December 22, 2008 5:59 PM
To: Chris Cole
Cc: STDS-802-3-HSSG@xxxxxxxxxxxxxxxxx; John DAmbrosia; Ryan Latchman
Subject: Re: [802.3BA] XLAUI / CAUI Ad Hoc



I just wanted to illustrate the difficulty members of 802.3ba would have in defining host-module interface for 100GBase-LR4/ER4 based on
publicly available information rather than in any way pointing to you for argument you did not make. 

You also say that specific implementation detail are inappropriate for the IEEE, but CR4/CR10, SR4/SR10 are based on very specific set of
assumptions.  The presentation I gave in Dallas, I made some very specific assumption on the module-host implementation which may be correct
or completely wrong, but we have to make some specific assumption please see page 5
Currently xAUI adhoc is defining transmitter mask instead of testing transmitter with compliance channel, in case the group decides to define
module-host compliance points then a 2nd transmitter mask must be defined at the output of module compliance board (see 2nd diagram on page 5).
In summary we have to replicate xAUI transmitter and receiver table for host and module.