Re: [8023-10GEPON] Upstream Wavelength
Glen,
Sorry for the late reply; I consulted with some experts here at Corning on this - here on some of our thoughts:
Corning would recommend that IEEE 802.3av reconsider using the 1340-1360 band for the following reasons:
1. For 1260 nm, one must add 0.07 dB/km to the value of attenuation coefficient specified at 1310 nm. This is stated in both IEC 60793-2-50 and in G.652.D. Using low water peak fiber, the attenuation coefficient is less than the value specified at 1310 nm for all wavelengths up to 1625 nm. For 1340 - 1360, certainly less since these are lower than the water peak wavelength of 1383 nm. Virtually all "modern" fibers are low water peak ITU-T G.652.D or IEC B1.3. So, why go to 1260 when it imposes a penalty of 1.4 dB over 20 km?
2. Cable cutoff wavelength is defined with a 22 m length and two single turn 80 mm diameter bends, one at each end. Matched clad type fiber has relatively little length dependence. Depressed clad fibers have a distinct length dependence. So as Frank noted below, using shorter patchcords with offset connections could induce some multimode behavior. Significant additional testing would be necessary to ensure that there are no deleterious effects associated with the use of the 1260 nm to 1280 nm operational window.
3. Cutoff measurements have variability. Even if the cutoff is measured at <=1260 nm, the actual cutoff may be higher. We would be reluctant to push the window to the specification limit.
In summary, we would recommend the 1340-1360 window given some uncertainty around the cutoff, the lower attenuation, and better connectorization losses.
Best regards,
Steve
-----Original Message-----
From: Glen Kramer [mailto:glen.kramer@TEKNOVUS.COM]
Sent: Friday, May 09, 2008 4:02 PM
To: STDS-802-3-10GEPON@LISTSERV.IEEE.ORG
Subject: [8023-10GEPON] Upstream Wavelength
All,
E-mail below expresses my opinion as a TF member, not that of a chair.
1) In July 2007, we discussed the choice between 1260-1280 and 134-1360 bands. We decided to use 1260-1280 because the dispersion was negative in this region (see http://www.ieee802.org/3/av/public/2007_07/3av_0707_uematsu_2.pdf)
But this discussion never mentioned the issue of proximity to the cut-off band.
2) I understand that cut-off wavelength is a function of core diameter and refractive indices of the core and the cladding. It is difficult to lower the cutoff wavelength without reducing the fiber core diameter or reducing the difference between refractive indices of core and cladding. Both of these have undesirable effects.
3) Consider ITU-T recommendation G.957 "Optical interfaces for equipments and systems relating to the synchronous digital hierarchy". Of interest are 3 table (attached) listing operating parameters of various interfaces. In most case, the O-band starts with wavelengths slightly higher than 1260, i.e., 1261-1360 or 1263-1360. The tables also show a footnote that says that some users may require the shortwave limit of 1270. Further, this document says:
"The wavelength regions permitting system operation are partially determined by either the cut-off wavelength values of the fibre or of the fibre cable. For G.652 and G.653 fibres, these values have been chosen to allow single-mode operation of the fibre cable at 1270 nm and above, with values as low as 1260 nm permitted by some Administrations. For G.654 fibre cables, the cut-off wavelength values have been accepted for single-mode operation at 1530 nm and above."
So, my questions are:
1) Are we taking unnecessary risk specifying 1260-1280 nm band for our use for 10G EPON?
2) Is all currently-deployed fiber based on IEC 60793-2 B1.1 and B1.3?
Are there areas where older fiber is deployed? What about the bend-insensitive fiber that goes into the apartments? Does it have the same cutoff?
3) What are the disadvantages of using 1340-1360nm? Dispersion is positive instead of negative. But would dispersion penalty be noticeably higher? Would we need to recalculate power budgets if we use this region?
4) Is it correct that using 1340-1360 is still friendly to G984.5?
I do not have a very strong opinion about the wavelength we should choose, but I don't want to make a mistake of overlooking something, and least of all I want 10GEPON to use a suboptimal wave band, leaving better real estate to some other use.
Please, comment and let's hear all the pros and cons.
Thank you,
Glen
________________________________________
From: Hajduczenia, Marek (NSN - PT/Portugal - MiniMD) [mailto:marek.hajduczenia@NSN.COM]
Sent: Thursday, May 08, 2008 1:02 AM
To: STDS-802-3-10GEPON@LISTSERV.IEEE.ORG
Subject: Re: [8023-10GEPON] Upstream Wavelength
Dear Jim,
some thoughts on the comment You are referring to ... I am the one responsible to its rejection so I believe some more explanation is justified, especially if we do not want to have this discussion during the meeting.
1. Your comment is submitted incorrectly, since D1.3 does not include Annex 91A - please check the latest version of the draft http://www.ieee802.org/3/av/private/draft_1_3/3av_d1_3.pdf which is subject to commenting per email http://www.ieee802.org/3/10GEPON_study/email/msg01027.html. Please bear in mind that editors have dozes of comments to process and we cannot look for the piece of text which You might refer to - that is interpreting the comment and we are not allowed to do it. If the commenter does not make sure I can trace the comment location, I will have to reject it. That is what happened with Your comments.
2. The second reason for the rejection is technical. We have been discussing the issue of the 1260 - 1280 nm transmission window for 10G signals in upstream (versus 1300 - 1320 nm or 1340 - 1360 nm) and it was decided that we would aim for this particular window, with the TF wide consensus. Various reasons can be mentioned and I am pretty sure that if You look through the past presentations, You will be able to trace it back quite easily. All in all, everyone (almost) seems to be in agreement that such a window allocation is feasible and will not warrant transmission problems.
As such, it is hard for me to decide on the validity of the claims of Your expert on this issue. The deployed fibre is supposed to have the cut off frequency below 1260 nm and the system should operate fine as long as the fibre type used in the PON is consistent with the spec (IEC 60793-2 B1.1 and B1.3). If someone wants to get hurt, we cannot prevent it. One can easily deploy fibre which will overmode in this range but that is already not standard compliant plant and we cannot provide for all the possible use cases.
IMHO we do not need to change anything. The restrictions which are placed on the fibre plant assure that the transmission in the selected window is feasible.
I will be looking forward to receiving any feedback from You
Best wishes
Marek Hajduczenia (141238)
NOKIA SIEMENS Networks S.A.
System Architect - COO BBA DSLAM R&D
Rua Irmãos Siemens, 1, Ed. 1, Piso 1
Alfragide, 2720-093 Amadora, Portugal
A marek.hajduczenia@nsn.com
m+351.21.416.7472 ++351.21.424.2337
________________________________________
From: ext Jim Farmer [mailto:Jim.Farmer@W7OPTICS.COM]
Sent: quarta-feira, 7 de Maio de 2008 22:00
To: STDS-802-3-10GEPON@LISTSERV.IEEE.ORG
Subject: [8023-10GEPON] Upstream Wavelength I'm campaigning for a comment submitted barely in time for the May meeting. The 10 Gb/s upstream wavelength is currently defined as 1260 to 1280 nm. See for example, Table 91-6 and many following references. This wavelength is also specified for the for 10/1GBASE-PRX-D3 (Table 91-7). According to my expert, this wavelength is dangerous because single mode fiber becomes overmoded at about 1260 nm, resulting in a second propagation mode at a different velocity. In addition, this wavelength will be overlapped by the normal 1310 nm 1 Gb/s upstream optical signal. This is acknowledged in the proposed document:
91A.3.2 Upstream wavelength allocation
The 1.25 GBd upstream transmission uses the 1270 - 1360 nm wavelength band, as specified in Clause 60. The 10.3125 GBd upstream transmission uses the 1260 - 1280 nm wavelength band, as specified in Clause 92. The two wavelength bands overlap, thus WDM channel multiplexign {sic} cannot be used to separate the two data channels.
An OLT supporting both upstream channels must use TDMA techniques to avoid collisions between transmissions originating from different ONUs.
We presume that the reason for specifying a separate wavelength band is to afford the opportunity to build a mixed-mode OLT receiver (capable of receiving both 1 Gb/s and 10 Gb/s) that is optimized for 10 Gb/s. However, this would imply a WDM at the OLT receiver, whose loss could rob a good bit of the improvement to be gained. This is acknowledged in Section 91A.5. Furthermore, this implies tight standards for the ONT 10 Gb/s laser, which could add cost. Since we have already acknowledged that the OLT will need to use TDMA to separate the 1 Gb/s from the 10 Gb/s upstream, we propose that the 1260-1280 nm upstream requirement be dropped, and that upstream transmitters for both speeds use the same wavelength band.
Figure 91A-5 shows some dual-rate receiver architectures that seem practical and which can provide optimization for each speed without having to use dual receivers. Implicit in these receivers is the ability to not only switch bias, but to also switch low pass filtering - since you have electrical amplification ahead of the filters, you can even leave both filters connected to the source and their respective loads all the time, so that the transient response of the filter is minimized.
This seems a better solution than using dual wavelengths, with the potential issues of the 1260 - 1280 nm band.
Thanks,
jim
Jim Farmer, K4BSE
Chief Technical Officer,
Wave7 Optics, Inc.
An Enablence Technologies Company
1075 Windward Ridge Parkway
Alpharetta, GA 30005 USA
678-339-1045
678-640-0860 (cell)
jim.farmer@w7optics.com
www.wave7optics.com