|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
My comments are interlaced...
From: Rich Taborek [SMTP:rtaborek@xxxxxxxxxxx]
Sent: Saturday, March 25, 2000 5:21 PM
Subject: SONET/Ethernet clock tolerance
Dave Martin, Norival Figueira,
I've been looking at the various requirements for transporting Ethernet over
SONET and one of them in particular is bothering me. That requirement is the one
to bridge the clock tolerance of Ethernet (+/-100 ppm) with that of SONET (+/-
The root question I have is whether or not current SONET framing or that
proposed for the Ethernet WAN can accommodate a clock of +/- 100 ppm? This would
be required to transport the proposed WAN PHY (or UniPHY) across clock domains
tolerance. I'm asking this question because I'm trying to come up with a
proposal for SONET framing for the UniPHY which is 100% compatible and compliant
with SONET OC/192c and I am using all the WAN PHY information presented by
yourself and others as a model.
My understanding from a previous presentation by Paul Bottorff,
slide 9, is that the payload clock tolerance is 320 ppm. Since I'm unfamiliar
with the many nuances of SONET framing, can you please acertain that this is
You understanding is correct.
If the proposed SONET framing for Ethernet is adequate to support +/-100 ppm
clock tolerance compensation, the second question I have is as to the mechanism
for performing clock tolerance compensation. It seems to me that the mechanism
involves at least the rewriting of SPE pointers and the modification of Line
Overhead Bytes (H1 and H2).
Correct again. The mechanism of pointer adjustment involves the H1-3 bytes.
My further understanding is that the clock tolerance compensation process is
referred to as one of the 3 "R's" (Re-Amplify, Re-Shape, Re-Time). The specific
process is Re-Timing and is usually reserved to SONET Regenerators and LTE's
(Line Terminating Equipment).
Here there is a misunderstanding. The "R" for "Retime" in a "3R" (i.e. Regenerator) is
not for clock tolerance compensation. The function of each "R" is:
(Amplified Spontaneous Emission), like shot noise, that accumulates thru optical amps.
jitter accumulated from propagation (e.g. Duty Cycle Distortion).
It has been proposed that Transponders for
coupling an Ethernet WAN PHY to SONET OC-192c can be either "Passive" or
"Active" according to our agreed upon "WAN PHY Definitions":
observation is that a transponder which contains both 100 ppm and 4.6 ppm optics
MUST perform re-timing. Therefore, it must be an Active Transponder.
An active transponder may include some LTE functionality (e.g. pointer processing).
also the case for all WAN PHY elements which cross clock domain boundaries.
Please help me validate or invalidate my observations.
If my observations are correct, my suggestion is to not bridge Ethernet to SONET
until the SONET boundary is encountered. A WAN PHY, SONET Lite or UniPHY which
transports SONET framed Ethernet in any manner may require significant
re-framing at any point that retiming is required.
David W. Martin
+1 613 765-2901
+1 613 763-2388 (fax)
> David Martin wrote:
> How about "SONET-compatible PHY". While I could agree to drop the "WAN"
> portion of the name, we still need to be clear that our proposal is not a
> SONET-compliant PHY. There is a significant cost/feature difference. To
> the "SONET-compatible PHY" has the following key differences:
> 1. minimal OH processing (i.e. only 4 OAM bytes, not SONET's
> 2. wider clock tolerance (i.e. the usual +/-100ppm, not SONET's +/-4.6ppm)
> 3. higher jitter tolerance (i.e. >0.15UIpp of SONET, exact value still TBD)
> 4. low cost optics (i.e. for <40km, not the 80/120km of SONET OC-192)
> All of which will "bring the cost down out of the stratosphere" to paraphrase
> committee member and in line with the 3x1GE target.
> David W. Martin
> Nortel Networks
> +1 613 765-2901
> +1 613 763-2388 (fax)
> -----Original Message-----
> From: Bruce Tolley [SMTP:btolley@xxxxxxxxx]
> Sent: Friday, March 24, 2000 2:50 PM
> To: stds-802-3-hssg@xxxxxxxx
> Subject: Renaming the WAN PHY
> To follow up on the suggestion made by Jonathan during the New Mexico
> plenary, may I be so bold as to suggest that we change the name of the
> PHY" to something very simple like PHY with SONET framer.
> We need to get the word WAN out of the name of the PHY
> o Most common folk outside the esteemed IEEE process think long distance
> when they think WAN and on the basis of Paul Bottorf of Nortel,'s
> presentations the initial application of the WAN PHY would be for short
> links between collocated equipment often in the same room.
> 2) There are ways to build MANs/WANs that do not require SONET. For 10GbE
> some of these MANs/WANs will use the LAN PHY and a 1550 PMD over dark
> or dark wavelengths. The proof point for this is the 1000s of long
> 1310 nm 1550 nm 1000BASE-X GBICs that are being deployed today over dark
> fiber. The WAN PHY as stated in the goal does not address the total
> possible 10GBE WAN market and confuses people.. This makes it a bad name
> my opinion
> I do not think we necessarily need a motion to change the objective, but
> think we need to choose our words carefully when we name the PHY that the
> objective signifies.
> Cisco Systems
Richard Taborek Sr. Phone: 408-845-6102
Chief Technology Officer Cell: 408-832-3957
nSerial Corporation Fax: 408-845-6114
2500-5 Augustine Dr. mailto:rtaborek@xxxxxxxxxxx
Santa Clara, CA 95054 http://www.nSerial.com