[802.3AZ] Frequency stability for 10GBASE-T EEE
Colleagues,
The 10GBASE-T EEE ad hoc has spent some time discussing the length of quiet and refresh periods during the 10GBASE-T LPI state.
For 10GBASE-T the values of the quiet and refresh periods will be constrained by the frequency stability of the master transmitter. The allowable quiet time (and hence the overall power savings) is dependent on the frequency drift during the quiet period, since the SLAVE PHY must stay timing locked to the MASTER PHY over this quiet period.
More details on this subject are contained in
http://www.ieee802.org/3/az/public/mar08/grimwood_01_0308.pdf
and
http://www.ieee802.org/3/az/public/adhoc/aug08/grimwood_01_0808.pdf
In the second presentation Mike Grimwood proposed that the rate of change of the symbol transmission rate on each pair of the MASTER PHY shall not exceed 0.17 ppm/second. This rate corresponds to a linear drift over the entire 100 ppm transmit symbol clock operating range in 10 minutes.
The 10GBASE-T ad hoc would like to solicit systems vendor feedback on this requirement.
Does this drift rate seem to be a reasonable limit or is it too constraining for the system? Feedback would aid us in selecting timing parameters that maximize power savings while guaranteeing link stability, and would be greatly appreciated.
Best regards,
Gavin.