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

Re: [802.3_10SPE] Autonegotiation objective clarification



Hi Geoff,

 

So you are saying that the fact that, over a given cable, a particular autoneg scheme will resolve to a technology which is unable to establish a link on that cable is not a reason to reject that autoneg scheme. Fair enough.

 

Oisín.

 

 

From: Geoff Thompson [mailto:thompson@xxxxxxxx]
Sent: 26 January 2017 22:56
To: Cuanachain, Oisin <Oisin.Ocuanachain@xxxxxxxxxx>
Cc: Geoff Thompson <thompson@xxxxxxxx>; STDS-802-3-10SPE@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_10SPE] Autonegotiation objective clarification

 

Oisin-

 

While the questions you raise have some validity,

they are essentially no different than the issues that we have with clause 28 auto-negotiation.

That legacy AutoNeg has no cable testing ability, either for frequency response or cable length.

If you try to run 1000BASE-T on Cat3 cabling, it will not help you at all.

OTOH, I feel that cl. 28 AutoNeg as it exists has been essential to the market success of 10/100/1000BASE-T.

 

Therefore, I reject your argument.

 

Geoff Thompson

 

 

On Jan 26, 2017, at 10:51 AMPST, Cuanachain, Oisin <Oisin.Ocuanachain@xxxxxxxxxx> wrote:

 

Hi, 

 

I have been thinking about yesterday’s presentation (http://www.ieee802.org/3/cg/public/adhoc/brandt_012517_3cg_01_adhoc.pdf) and it does raise some interesting questions. Primarily what exactly are we hoping to negotiate ? 

The presentation highlighted the idea of  forward vs. backward compatibility. I’m not sure I understood the distinction between these. At the moment my assumption is

Backward compatibility = compatibility with 1000BaseT1 autoneg (Clause 73)

Forward compatibility  = compatibility with any future single-pair 802.3 PHYs.

 

For me, Backward compatibility is problematic for at least two reasons:

(i)                  The signaling frequency is very ill-suited to the 1000m channel response. Slide 7 of the above presentation acknowledges the Clause 73 signalling ‘Will be severely attenuated by 1000 m link segment (>25dB)’ but I think even this understates the problem, Slide 5 ofhttp://www.ieee802.org/3/cg/public/Jan2017/McClellan_3cg_01a_0117%20xBASE-T1%20Autoneg.pdf shows that roughly 90% of the signal power is above 6.7MHz (0.4fBaud) but our current Link Segment proposal shows an Insertion Loss of 33.8dB at 6.7MHz.

(ii)                Assuming we can achieve reliable reception of Clause 73 signaling over 1000m what happens if two chips which implement both 802.3bp and 802.3cg are connected via 1000m? The autoneg process will resolve to a HCD technology of 1000BaseT1 and then promptly fail to bring up a link as 1000BaseT1 only reaches to 40m.

 

Forward compatibility, as I have interpreted it above, seems far easier. We could still re-use a lot of the work in Clause 73 but simply shift the clock frequency down to a range more appropriate to a 1000m cable. As to what we are negotiating at the moment that would simply be Master/Slave resolution.

 

The above assumes that the same autonegotiation will be used for the 15m PHY and the 1000m PHY and this may be a mistake. I can certainly see how Clause 73 autoneg would make a lot of sense for a .cg PHY intended to meet the 15m objective where two other PHYs (100/1000BaseT1) already exist. I guess this leads us back to the question of whether the 15m and 1000m reach objectives would be met with a single PHY or two different PHYs.

 

Oisín