$customHeader
Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

Version 1 Next »

G.8262 (clause 9.2): Where did the Jitter Tolerance mask for 1G Electrical interfaces come from?

G.8262 does not specify a jitter tolerance mask for 1G electrical interfaces (Table 6, note 2 and Table 11, note 1 both state that these interfaces are  “for further study”). However, the 4-lane, multi-level physical layer scheme used for 1G electrical Ethernet has a clock rate of 125MHz (versus 1.25GHz for 1Gbit/s optical Ethernet), which implies a 10:1 ratio for jitter parameters. We have therefore implemented 1G electrical Ethernet jitter testing using a scaled jitter tolerance mask to an upper frequency of 5kHz and a jitter generation measurement filter of 250Hz to 1MHz using this implied ratio.

Mapping between SyncE jitter period and UI give us:

  • 1G optical: clock rate = 1.25GHz and so one UI = 0.8ns
  • 1G electrical: clock rate = 125MHz and so one UI = 8ns

The 1G optical mask is:

From this mask, we derived an equivalent for 1G electrical as follows:

The 10Hz point must aligns with the G.8262 10Hz wander tolerance limit, which is 250ns. 250ns equates to 31.25UI. We then simply scaled the higher-frequency breakpoints and max frequency down by a factor of ten (because the clock rate is 1/10 compared to 1G optical). Since UI is already a normalised unit, we retained the 1.5UI requirement at the higher frequencies

This gives us:

  • No labels