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

RE: [10GBASE-T] Technical feasibility: AutoNegotiation


If it is something to negotiate, then proprietary won't work.  If the SG
chooses to focus strickly on 10G, then the only negotiation will be for
10M, 100M, 1G or 10G.  If the multi-rate is selected, then negotiation
for 2.5G and 5G would be added to the list.  The variant is the length
of cable and type of cable that can be used, and there are means to help
determine the length and quality of cable.  Do we write those into the
specification?  Too early to tell.  For the 10G PHY, length
determination would make a nice annex.  For multi-rate, cable type would
make a nice annex.  Alas, we don't have an approved PAR yet, so there's
not much we can do but speculate.


-----Original Message-----
From: [] 
Sent: Monday, August 04, 2003 4:56 PM
To: Booth, Bradley;
Subject: RE: [10GBASE-T] Technical feasibility: AutoNegotiation


I'm not sure if you are saying that you believe that shifting speed down
due to cable quality issues should remain proprietary or if you are
agreeing with Dan that it would be of interest either in 10GBASE-T or as
a separate project.

Please clarify.

One possibility is to do an informative annex on autonegotion because
the two sides don't necessarily need to cooperate. One could try
reducing the highest speed adverti$ed if one finds that the link doesn't
work after negotiating and connecting. (Ping can tell something about
attenuation/length, but it doesn't tell much about whether the link has
too much alien NEXT.)


-----Original Message-----
From: Booth, Bradley []
Sent: Monday, August 04, 2003 2:26 PM
Subject: RE: [10GBASE-T] Technical feasibility: AutoNegotiation


Interesting that you'd ask this, because it was a question I had a
meeting or two ago.  In most cases, the PHYs can determine the link
length via a simple electronic ping.  As for the grade of the cable,
extrapolation could be performed during the training sequence as the
number of taps would have a correlation to the quality of the cable and
the data rate or distance supported.  A lot of these methods are in use
today for 1000BASE-T, but they are implementation strategies, not based
on what is in 802.3.


-----Original Message-----
From: DOVE,DANIEL J (HP-Roseville,ex1) [] 
Sent: Monday, August 04, 2003 3:49 PM
Subject: RE: [10GBASE-T] Technical feasibility: AutoNegotiation


Perhaps someone can give me a quick pointer on how we are going to
the issue of <100m cable plants that meet ISO 11801 specs not being able
support 10GBASE-T.

For example, I have a building full of CAT5e and CAT6 cables. The cables
were installed to specification. Now, all I see are patch panels and
jacks. I don't have a record of which cable is how long, or what its
electrical specs are.

The current Auto-Negotiation protocol supports HCD resolution which
that the link, if configured to 10GBASE-T, will resolve to that speed if
both ends of the link support it. However, if the cable (85m of CAT6)
not, the link will bounce up and down as it negotiates to speed, then
to link. Maybe even worse is the link that actually comes up, but
unreliably due to thermal variation causing changes in parametric values
suffers excessive BER or drops link occasionally.

What is the fix?

Do I test the wire in advance of using 10GBASE-T?
Do we fix the Auto-Negotiation protocol to automatically drop speed if
SNR gets too low to guarantee the BER?

If these have already been addressed, please feel free to just send the
to the presentation. 


Dan Dove
HP ProCurve Networking