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

[802.3_NGAUTO] Explanation on comment 143 - upper frequency for PSANEXT/PSAFEXT

On draft 2.1, I submitted comment 143, mostly as a hedge against bad experiences.  I'd like to get you all thinking about this, because it is a value judgement on our parts.
Comment 143 is on the frequency range for the alien crosstalk specifications.  Sometime in the past couple of drafts when we limited the frequency range for the transmission parameters (IL, RL) we also we cut back on the frequency range for these noise coupling parameters and made them PHY-speed dependent.
As you will see, the noise coupling parameters are different, because they impact compatibility between PHY types in installations.

Comment 143 says:
It is important to limit the noise ingress even outside the bandwidth of the PHY, especially if multiple rates of PHYs are to be used together in the same system.  As such, the PSANEXT and PSAFEXT characteristic needs to be specified to the same frequency for all PHY types

The suggested remedy was: Replace Fmax on Page 169 line 9 and Page 170 line 6 with 4000 MHz.
(this makes the upper frequency for these two specifications 4000 MHz, independent of the PHY type.)

The issue here is simple.  If you do not limit PSANEXT and PSAFEXT coupling at higher frequencies, you must consider the noise that can couple from higher-rate BASE-T1 systems to be unbounded.  If you don't, you are assuming some alien crosstalk specification, without stating it, OR, you are specifying that 2.5GBASE-T1 cannot reliably coexist with 5GBASE-T1, and, likewise, neither 2.5G nor 5GBASE-T1 can coexist reliably with 10GBASE-T1.  Otherwise, you are assuming that there is an unspecified out-of-band noise rejection filter built in to the PHYs, AND, that this filter bandwidth is sensitive to the PHY rate. (FYI , the alien crosstalk noise rejection test is also band limited, the noise there stops at 1.25x Nyquist, but extending the frequency of that test would not fix the problem, because the cabling could couple at ANY level out of band).

In addition to putting compatibility requirements on fixed, engineered configurations, this would also virtually preclude reliable autonegotiating multi-speed PHYs (as there is no out-of-band noise rejection requirement).  In reality, we don't usually consider this a problem, because we assume the cabling to be well-behaved in alien noise coupling as the frequency increases.  However, assuming a specification is always a bad idea.  Much better to write it down.  Even if it requires some slope change or relaxation beyond the currently specified level (contributions or suggestions welcome from cabling guys), having some spec, any spec, at higher frequencies for the lower rate PHYs fixes the problem and enables a reliable, compatible, multi-speed environment.

George Zimmerman, Ph.D.
President & Principal
CME Consulting, Inc.
Experts in Advanced PHYsical Communications

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