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

Re: [802.3_NGAUTO] potential objectives change



Amir,

The single pair auto-negotiation does not add substantial time to the startup process. Less than 1ms compared to a target of 100ms to link up.  Without autonegotiation a handshaking scheme is still needed between master and slave to establish the start of link training.

 

Regards,

Brett

 

From: Amir Bar-Niv [mailto:Amir.Barniv@xxxxxxxxxxxx]
Sent: Tuesday, February 07, 2017 1:31 PM
To: STDS-802-3-NGAUTO@xxxxxxxxxxxxxxxxx
Subject: [EXT] Re: [802.3_NGAUTO] potential objectives change

 

External Email


Question (this is not a suggestion, just a question):  is there a need at all for auto-negotiation in this type of environment (automotive)?

 

The biggest disadvantage of auto-negotiation is the additional setup time of the link that is needed to finish the auto-negotiation process, every time we turn on the power.

 

When you use Ethernet in non-automotive environment/network, then for example when you have a switch that supports multiple speeds (e.g. 10G, 1G and 100M), you must have auto-negotiation since you never know what speeds are supported by the link partner that is connected to this switch. The auto-negotiation detects the highest common denominator of the switch and the link partner, and set the speed accordingly.

In automotive, once we define the final topology/structure/cables of the in-car network then we know the speed that is required on each link. Even if a PHY supports multiple speeds we can set it manually to work in the desire speed, and save the extra link setup time of the auto-negotiation.

 

Regards,

Amir

 

From: NATALIE WIENCKOWSKI [mailto:NWIENCKOWSKI@xxxxxxx]
Sent: Tuesday, February 07, 2017 9:58 PM
To: STDS-802-3-NGAUTO@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGAUTO] potential objectives change

 

All,

 

As we discuss auto-negotiation and multiple speeds, is it necessary for all of the speeds to use twisted pair to do this?  If coax or twinax is chosen for 10 Gbps is it possible for this phy to operate at and communicate with PHYs at lower speeds that use a form of twisted pair?  If auto-negotiation is truly needed by some parties between these speeds we need to ensure that we don't select a cable type that makes this impossible.

 

It would be  helpful if PHY vendors could provide contributions on whether this is possible and if end users could provide information on their needs for auto-negotiation.

 

Thanks,

 

Natalie Wienckowski

 


From: Geoff Thompson <thompson@xxxxxxxx>
Sent: Wednesday, February 1, 2017 6:10 PM
To: STDS-802-3-NGAUTO@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGAUTO] potential objectives change

 

Colleagues

What I would like  to see is some technical presentations on the impact
of HIGHLY ASSYMETRICAL data rates on:
        - Relative implementation complexity (e.g. gate count, die size)
        - Operational power dissipation
and (therefore presumably)lower relative cost.

It seems from the discussions that in this particular case,
we have high enough prospective volume and
enough cost sensitivity
and enough sensitivity to power dissipation
that this case is unusual enough in our history to be worth extra attention.

In particular, the back channel requirements seem low enough that they
could be satisfied by what exists in other standards as an auxiliary
or management channel.

Geoff Thompson