I also have some queries on 10BASE-T1L Low Power Idle (LPI) functionality of
The approach taken in Draft 2.0 is for asymmetric LPI at the PHY level, and is largely described in Figure 146-15 PHY Control state diagram (part b). My concerns here are as follows:
- There does not appear to be any defined fail-safe/timeout behavior in the event that refresh-quiet signalling from the link partner is not as expected. The maximum time between refreshes should be governed by the lpi_quiet_timer duration (nominally 2050
us). If a refresh is not observed from the link partner after some longer time (longer than
lpi_quiet_timer), then the PHY should consider that the link partner might no longer be active, and the link should then be brought down. I understand that this is the purpose of of Clause 220.127.116.11 Refresh Monitor function of 1000BASE-T1 EEE.
Does 10BASE-T1L EEE need a similar function, or is there something there already?
- With asymmetric LPI a scenario where the SLAVE is transmitting data frames while the MASTER is in QUIET arises. Here the SLAVE has to transmit in the absence of actively recovering timing from link partner signalling. There might be a need to place additional
requirements on the MASTER transmit clock in LPI mode; frequency drift when the MASTER is in QUIET cannot be tolerated to any degree.
- There is no predictability to refresh timing that a PHY implementation can work with. Refreshes will be asynchronous between the PHYs; they might coincide, partially overlap, or be completely separate. The PHY will have to be able to handle all cases.
Other PHY EEE standards appear to address this, e.g. 1000BASE-T1 EEE, where refreshes between MASTER and SLAVE are staggered and synchronized, and 1000BASE-T EEE ensures that refreshes always coincide.
- PHY power saving in LPI mode might be limited. The refresh/quiet ratio is 1/10 (lpi_refresh_timer/lpi_quiet_timer). The PHY receiver will have to be active for a local refresh (for echo canceller update), as well as for link partner refresh (for channel
equalizer update, and timing recovery update). Considering receiver activity then, the ratio becomes closer to 2/10. Other PHY standards were better in this regard, e.g. the corresponding ratio for 1000BASE-T EEE was closer to 1/100.
- Was any consideration given to adopting a symmetric LPI scheme for the 10BASE-T1L PHY? This would address some of the concerns listed, but maybe there were aspects of it seen as undesirable.
Thanks & Regards,
To unsubscribe from the STDS-802-3-10SPE list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-10SPE&A=1