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

RE: [10GBASE-CX4] Comments on latest revision




Hi Clark,

Good comments. We will address them directly in the Dallas
meeting. I have a few specific comments which I will note below.

Please keep the comments coming in via the reflector. These will
help us to nail down any final issues.

BTW: Funny thing, while looking into your points, I found a typo
in clause 45 that we carried over. Ooops! We will fix that and 
inform 802.3 for maintenance.

Thanks,

Dan

-----Original Message-----
From: Clark Foley [mailto:clarkf@mxim.com]
Sent: Wednesday, February 26, 2003 2:45 PM
To: 'stds-802-3-10GBCX4@ieee.org'
Cc: Clark Foley (E-mail)
Subject: [10GBASE-CX4] Comments on latest revision



To the CX4 SG,

Congratulations on a great job of pulling all of this together.  Thank you 
for the opportunity to comment.

sections
54.6.7 and 54.6.8 Global and Lane-by-lane transmit disable functions.
---------------------------------------------------------------------
I think that the "Global" disable function should be mandatory and the 
"Lane-by-lane" disable should be optional.  Because this is CX4 and not 
four asynchronous channels, the link should disabled with one bit.  It 
doesn't really matter if you have MDIO, one bit is as easy as four bits to 
set.  In designs without the optional MDIO, one bit becomes much easier to 
implement than four.  The priority should be placed on disabling the link 
in as simple a manner as possible.  One bit (a.k.a. one pin) does this. 
 "Lane-by-lane" disable is helpful for debugging and evaluating designs, 
but it does not make sense at runtime for a XAUI link.  Please make the 
"Lane-by-lane" disable the option for those who need to tinker off line.

DD: I think we need to look this over one more time. I think I agree with
you but would like to hear it discussed in the larger group.

section
54.7.3.4 Amplitude and Swing (and Table 54-6)
----------------------------------------------
Line 37 - Dc-reference logic level is meaningless because of the ac 
coupling.  This makes sense.

DD: Yep.

Line 38 - If dc-reference logic level is meaningless because of the ac 
coupling, then the common mode voltage should also be meaningless.  A dc 
common-mode voltage requirement does not make sense.  I recommend deleting 
the common-mode voltage requirement.

DD: The point of distinction is that line 37 is addressing *logic levels*
while line 38 is addressing common-mode voltage. The reason for spec'ing 
the common-mode DC voltage is to provide direction for the AC coupling 
requirements of the receiver. We wanted to limit it to a number that allows
either AC coupling at the transmitter (VDC_CM=0) or DC coupling with a 
source terminated up to 1.8V rails and anything in between.