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

[802.3_10SPE] 10BASE-T1L PHY Control synchronization


I have a query on operation of the 10BASE-T1L PHY Control function of IEEE P802.3cg/D2.0.

I understand that 10BASE-T1L provides for two modes of PHY configuration (MASTER/SLAVE):

  - Auto-negotiation

  - FORCE mode, auto-negotiation disabled

With auto-negotiation enabled there are a number of interactions between arbitration and PHY Control, but two in particular are:

  1. Auto-negotiation will complete at about the same time in the two PHYs, and so the PHY Control functions will start at about the same time.

  2. Auto-negotiation provides for fail-safe/timeout behavior in the form of the link_fail_inhibit_timer.  If link_status is not observed to be OK before this timer expires, i.e. link startup is taking too long, then auto-negotiation will restart and PHY Control will be disabled.

In the FORCE mode auto-negotiation is disabled.  The training_timer of PHY Control appears to provide fail-safe/timeout behavior in the event that link startup takes too long to complete.  However, the FORCE mode does not seem to provide any mechanism for PHY Control synchronization between the two PHYs.  PHY Control start will depend on when the PHY exits from reset/powerdown, or management agent enable.  The PHY Control functions of the two PHYs will be asynchronous, and the training_timer of the PHYs will start at different times.  They might start off out of synch., with the training_timer of one PHY starting (at worst) 1500 ms after the other PHY.

The PHYs then would not have the full training_timer duration (3000 ms) in which to complete training.  Training must complete on both sides (loc_rcvr_status and rem_rcvr_status needed), and so training_timer_done in one PHY will frustrate link startup for both PHYs.

A badly out-of-synch starting condition between the PHYs might persist, resulting in a much longer time to achieve link up.

Has the synchronization of PHY Control functions in the FORCE mode been considered?  Should a change be made to the draft standard to address this?

Thanks & Regards,

Niall Fitzgerald

To unsubscribe from the STDS-802-3-10SPE list, click the following link: