RE: [802.3ae_Serial] XAUI update from Plenary
I was asked by some to explain why Dj is almost always out of band. It
might be easier to post my response here:
I assert that the main source of Dj is ISI. To see why Dj spectral content
concentrates out-of-band, we can consider the minimum ISI frequency, at
3.125Gbps=1.56GHz:
PRBS31 min. ISI frequency is 1.56GHz/31~50MHz
8B10B min. ISI frequency is 1.56GHz/5=312MHz
where 31 and 5 are the longest run length for PRBS31 and 8B10B,
respectively.
Conclusion:
PRBS31 min. ISI frequency is already out-of-band (at least 50MHz). 8B/10B
bit stream would be further out (at least 312MHz)
In 8B10B stream, run length of 5 is rare. "Typical" run length would be
shorter than 5, making most ISI frequency component even higher.
Hence our conclusion that Dj is usually out of band.
Leo
http://www.bitblitzcom.com
-----Original Message-----
From: owner-stds-802-3-hssg-serialpmd@majordomo.ieee.org
[mailto:owner-stds-802-3-hssg-serialpmd@majordomo.ieee.org]On Behalf Of
Leo Wong
Sent: Wednesday, July 25, 2001 3:44 PM
To: Kesling, Dawson W; Serial PMD reflector (E-mail)
Subject: RE: [802.3ae_Serial] XAUI update from Plenary
Dawson,
Thanks for the updates. I support the our direction in issue 1 & 2.
We've been doing some jitter measurements and our data disagree with the
necessity to add random jitter limitation (issue 3).
Dj is almost always out-of-band, while Rj may come from in-band or
out-of-band jitter.
Jitter tolerance of any CDR is always higher for in band jitter than
out-of-band jitter, because CDR PLL can track in-band jitter, but reject
out-of-band jitter. In other words, if a CDR device has a certain jitter
tolerance number for Dj, it would have even higher jitter tolerance for Rj.
(Pre-emphasis and equalization techniques not considered here)
Draft 3.1 specifies far end jitter Dj of 0.37 UI. Rj should be higher. The
0.55UI total jitter tolerance limit should suffice. An additional Rj limit
is unnecessary.
BitBlitz can demonstrate technical feasibility tolerating at least 0.55UI
out of band jitter.
Leo
http://www.bitblitzcom.com
-----Original Message-----
From: owner-stds-802-3-hssg-serialpmd@majordomo.ieee.org
[mailto:owner-stds-802-3-hssg-serialpmd@majordomo.ieee.org]On Behalf Of
Kesling, Dawson W
Sent: Friday, July 13, 2001 3:29 PM
To: Serial PMD reflector (E-mail)
Subject: [802.3ae_Serial] XAUI update from Plenary
There are three remaining technical issues from the IEEE Plenary meeting
this past week that must be resolved by vote at the September Interim.
1. Differential return loss at the driver
Credible simulation results confirm what many have suspected, that the 10 dB
return loss limit at the driver is impractical. We need measured data
showing how much relaxation is needed for practical implementations, and
solid information concerning how much relaxation is possible without
breaking the link integrity in the presence of connectors.
2. Common mode return loss at the receiver
One participant proposed that the common mode return loss be unspecified at
the receiver to eliminate the need for common mode termination. Information
on the impact to XAUI system performance is needed to determine if this is
possible.
3. Random jitter limit.
If deterministic jitter is very low, the present draft allows random jitter
to take up almost the entire jitter budget. It was argued and agreed that
this is not practical or desirable, and that random jitter should have a
maximum limit (like deterministic jitter does) to ease compliance
requirements on the receiver. Information on the maximum random jitter that
can be seen in practice is needed to establish a practical limit.
The Plenary summary for XAUI is posted at
http://grouper.ieee.org/groups/802/3/ae/public/jul01/kesling_2_0701.pdf.
Thank you to all participants!
-Dawson