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

Re: [8023-10GEPON] Ad hoc on 10GEPON channel model - definitions



Dear Piers and all,
 

After some consideration and in-depth reading of the 802.3 standard, here is what I think about the issues related with the definitions:

  • since the existing standards use the current definitions and are successful, it would be unwise of us to try to modify the basic definitions and cause havoc connected with a number of revisions of individual clauses that would follow. The current versions of the clauses are relatively stable and we should not touch them if we can avoid that.
  • if during the process of defining the power budgets for the 10G EPON systems we come to the conclusion that we need a more detailed definition of the allocation for penalties, we could potentially introduce a more detailed (extended) definition of this term along with the sub terms which would originate from this concept. I think that we could do that safely since other clauses would remain intact and we would have the more detailed definitions that we may eventually need.
  • 1G EPON systems are less affected by the channel impairments (1 G data rate etc.) so the allocation for penalties was more relaxed. In the case of 10G systems with the highest provisioned power budget (28, 29 dB or above) we really need to look into penalties in detail - otherwise we risk creating vague specifications. I think that the carrier input would be very valuable in this point.
  • indeed, the EFM provides much more details than any other standards, though additional level of details in specifications brings further clarity rather than disturb the overall picture.
  • regarding clause 68 - please correct me if I am wrong but there is no clause 68 in 802.3-2005. Do You mean clause 64 ?
  • I would also welcome any feedback on methods used to estimate link penalties for 10G P2P Ethernet and 1G EPON systems - there seem to be some inconsistencies between the definitions and the values which are presented in the respective clauses. If anybody had any idea how the penalty allocation was carried out and its value estimated, please share it with us.
Thank You for Your attention
Best wishes
Marek
 

From: DAWE,PIERS [mailto:piers.dawe@AVAGOTECH.COM]
Sent: sexta-feira, 24 de Novembro de 2006 19:46
To: STDS-802-3-10GEPON@listserv.ieee.org
Subject: Re: [8023-10GEPON] Ad hoc on 10GEPON channel model

Marek,
 
You say:
 
The other way out is to provide >clause< specific definitions, and I would like to hear Your opinion on this option. I do not dislike it personally though it may be perceived as a violation of the general 802.3 definitions.
 
I think this would be difficult.  We should be much more advanced and precise in our definition of what if any problem we are trying to solve before attempting such a thing.
 
But what about the margins for aging, repair, rerouting etc. Should they be considered in the channel insertion loss (max value) or indicated together with the other penalties we will have to account for in the system.
 
You should ask the operators of 1 G EPON (802.3 clause 68) if they are confused.  I believe it's clear enough that margins of optical loss for aging, repair, rerouting are
are part of the loss allocated to the fiber plant (NOT the "allocation for penalties").  You may wish to have an explicit item for this margin, but it would be a new thing, not found in other 802.3 clauses as far as I know.
 
On the other hand, the existing confusion within the specifications of 802.3 is personally speaking, disheartenting. As Tanaka-san pointed out in his presentation last time, we have no idea in the end what are the allocations for individual transmission penalty parameters since the values are taken in an arbitrary way and we have no clue how to back track the reasons for their selection.
 
Standards are not text books.  You have the EFM link model, which is a lot more than most standards give you.
 
 are we however free to provide out extended definition applicable to PON environment? After all, we use P2M fibre plant rather than P2P links which are common in all other Ethernet specs. </Marek
 
Clause 68 addresses an Ethernet PON.
 
if the 802.3 provides some global definition of the term, why do some of the specs fail to follow it? The model that we are using should correspond to 802.3 definitions as closely as possible, or am I mistaken ??</Marek> 
 
It looks like a mistake.  Often the definitions are written after the clause.  We can look into the history of this.
 
Piers
-----Original Message-----
From: Hajduczenia, Marek [mailto:marek.hajduczenia@SIEMENS.COM]
Sent: 22 November 2006 08:20
To: STDS-802-3-10GEPON@listserv.ieee.org
Subject: Re: [8023-10GEPON] Ad hoc on 10GEPON channel model

Dear all,
please find my comments to the last 3 emails inline in <Marek></Marek> tags.
 
I would also like to make a general comment on this issue.
 
I personally believe that some form of consensus between IEEE and ITU specs would be desirable since it would facilitate the analysis of the specifications from both standardization bodies. We must however exercise extreme caution since some of the definitions that we may want to change, are already buried in the 802.3 specs and their alternation may require a very strong support from 802.3 voters - in other words - either we have a good reason to do that and convince a bunch of people (which is strictly speaking - tough) or we leave it where it is and provide definitions for elements which are still undefined in the 802.3. The other way out is to provide >clause< specific definitions, and I would like to hear Your opinion on this option. I do not dislike it personally though it may be perceived as a violation of the general 802.3 definitions.
What do You think about it?
 
Additionally, I would like to know what is Your opinion on the channel insertion loss: there are some obvious components like fibre loss, connectors, splices and PSC module(s). But what about the margins for aging, repair, rerouting etc. Should they be considered in the channel insertion loss (max value) or indicated together withthe other penatlies we will have to account for in the system. Please note that in the previous ad hoc we did indicate that as for now, we are unable to reach any conclusion as to whetere such margins should be allocated.
 
best wishes
 

Marek Hajduczenia (141238)
SIEMENS Networks S.A. - IC COM D1 R
Rua Irmãos Siemens, 1
Ed. 1, Piso 1
Alfragide
2720-093 Amadora
Portugal
* Marek.Hajduczenia@siemens.com
http://marekhaj.easyisp.pl/index.php
(+351.21.416.7472  4+351.21.424.2082

 

 


From: Frank Effenberger [mailto:feffenberger@HUAWEI.COM]
Sent: terça-feira, 21 de Novembro de 2006 22:46
To: STDS-802-3-10GEPON@listserv.ieee.org
Subject: Re: [8023-10GEPON] Ad hoc on 10GEPON channel model

Dear Piers,

I think that there is plenty of common ground to work on here, and I’m sure that this time, we will work out a common framework of terms that will satisfy both the IEEE and ITU formalisms.  At least, that is my hope. 

<Marek>That is definetely what we would like to see - bridging the gap between ITU and IEEE terminology could be advantegous and limit the confusion which is creating when reading and comparing IEEE and ITU specs. I am however concerned that at some point, we may end up trying to change too much as Dawe was pointing out - certain definitions are already hard-coded into 802.3 and we will have to have either a very good reason for their modification or avoid touching them at all. Update of all existing standards which refer to certain parameter values may be an overkill and a task that will not find too many supporters in the 802.3 voting group. On the other hand, the existing confusion within the specifications of 802.3 is personally speaking, disheartenting. As Tanaka-san pointed out in his presentation last time, we have no idea in the end what are the allocations for individual transmission penalty parameters since the values are taken in an arbitrary way and we have no clue how to back track the reasons for their selection. It would be very helpful for the development of any future specifications, if we finally settled for some stable set of definitions and provided detailed information on how the given power budget was estimated.</Marek>  

As for the last sentence you mentioned, the idea is not that the optical penalty would be measured… Indeed, this would be impossible.  The idea is that the operator should design his fiber route and figure out what his penalties should be.   I think this approach is especially important for the PON application, where we may have various other systems sharing the same fiber.  In such a system, the path penalty is really beyond the control of the equipment manufacturer.  

<Marek>I also interpreted that statement this way - we do not want to tell operators to test the line insertion loss every time they switch on a new Ethernet port. This should be a plug&play rather than plug&pray feature - on the other hand, if we do not provide a clear specification on how to define the channel iserntion loss + its margins, we may end up again with people sharing completely different views on whether the margins should be included in the optical fibre plant insertion loss or not. I believe that we could provide a set of extended definitions to be used on the scope of the future 802.3av, which do nto necessarily cover other specs. The PON environment will require us to provide more detailed specifications of the channel insertion loss when compared to P2P links, where the margins are less stringent.</Marek> 

Regards,

Frank E

 

 


From: DAWE,PIERS [mailto:piers.dawe@AVAGOTECH.COM]
Sent: Tuesday, November 21, 2006 3:22 PM
To: STDS-802-3-10GEPON@listserv.ieee.org
Subject: Re: [8023-10GEPON] Ad hoc on 10GEPON channel model

 

Frank and all,

 

With regard to "who has the responsibility", 802.3 is aiming for an interoperability specification.  There are not just two parties: manufacturers of transmitters and receivers, and operators.  There are three: implementers of transmitters, implementers of cable plant, and implementers of receivers.  802.3 specifications aim to make clear who is responsible for what.  In particular, they make the transmitter take responsibility for its own performance and the dispersion penalty that it causes (because the transmitter implementer knows what dispersion it is to comply over).  I would imagine that some nonlinear effects which are strong functions of spectral width or dither would similarly be the transmitter's responsibility, but would also be a concern for the fiber specification.  Also, the explicit stressed sensitivity specifications makes the receiver requirements much clearer to all parties.

 

I agree with you that a standard should specify "worst-case, end-of-life parameter values".  Large unallocated margins are not usual (in Table 38-5, for example, the margins in three out of four cases arise because the same transmitters and receivers are used over different fiber types with different characteristics).

<Marek>That is indeed a good point - I found so far no indiciation whether IEEE specifies BOF type parameters and then accounts for some aging penalties or whether EOL parameters are provided plus some yet undefined penatlies. This should be clarified, otherwise we may end up having serious problems with the definitions for high power budgets (28, 29 dB) </Marek> 

 

I notice that "power budget" is defined for all of 802.3 as follows:

1.4.273 power budget: The minimum optical power available to overcome the sum of attenuation plus power penalties of the optical path between the transmitter and receiver calculated as the difference between the transmitter launch power (min) and the receive power (min).

<Marek>As I nocited before, we may not want to change this one for sure - are we however free to provide out extended definition applicable to PON environment? After all, we use P2M fibre plant rather than P2P links which are common in all other Ethernet specs. </Marek> 

 

This is not the same as the link model's power budget, Tx min - unstressed Rx sensitivity max, which is consistent with e.g. Tables 38-7, 38-8 and 38-9 (and similar tables in other clauses).  As folks will not be willing to change numbers in established standards without great debate, the definition above might need to be brought into line.  IEEE Std 802.3aq does not use the term "budget" at all.  Power budgets are informative.  If you find it more helpful to focus on the maximum/minimum channel insertion losses at this stage in the project and avoid the word" budget", you are free to do so.

<Marek>Then the question is as follows: if the 802.3 provides some global definition of the term, why do some of the specs fail to follow it? The model that we are using should correspond to 802.3 definitions as closely as possible, or am I mistaken ??</Marek> 

 

This sentence from G.957:

"For each installation, it should be verified that the total optical path penalty, which includes combined dispersion and reflection degradations, does not exceed the value..."

is concerning.  802.3 in general aims at "plug and play": without a need for a measurement of optical path penalty for each installation.

 

Piers

-----Original Message-----
From: Frank Effenberger [mailto:feffenberger@HUAWEI.COM]
Sent: 21 November 2006 18:34
To: STDS-802-3-10GEPON@listserv.ieee.org
Subject: Re: [8023-10GEPON] Ad hoc on 10GEPON channel model

Dear all,

I’ll start the ball rolling with a discussion of definitions.

I would like to speak in favor of the system of definitions that are established in G.957.  

The text below is cut from the document (I hope this snippet falls within the bounds of ‘fair use’.  In any case, the document will become open on Jan 1st, 2007, so we needn’t be too concerned.)

I have highlighted the key text that, in my humble opinion, have been the biggest source of difference. 

The biggest “style statement” that this approach makes is one of ‘who has the responsibility’.  The manufacturers have the responsibility for the transmitters and receivers.  The operators have the responsibility for the fiber plant.  No crossing of the lines is allowed. 

So, the manufacturer’s must provide a guarantee that their transmitter will be no worse than the specification (output power, eye-mask, and spectral characteristics, mainly), and that their receivers will work as specified, even given the worst-case transmitter.  No fiber impairments are included here, and the ultimate link-test is done using an attenuator.  

Similarly, the operators must guarantee that their fiber falls within the spec.  Note that this goes beyond setting the optical loss.  The operators are responsible for any other impairments that are caused by their fiber, such as dispersion, reflections, polarization, nonlinear self-impairments, and non-linear crosstalk impairments.  If it’s caused by the fiber, it is the operator’s engineering responsibility. 

<Marek>Correct, that is why we need to finally decide what we call "channel insertion loss" - whether the margins related with the fibre plant should be included in this general term or not. Otherwise we will always have a disrepancy between partical definition which say - yes, it is icnluded since it is the same fibre plant after all, and a theoretical one which says that penatlies should be grouped together. I would like to see some discussion on this issue if possible. </Marek>  

Now: of course, as we prepare the standard, we will consider all these impairments, and we will set the benchmark for how many dB’s should be allocated to them.  Also, when the standard finally boils down to listing a number, it is the ‘attenuation maximum’ that gets the top billing.  You will notice that there isn’t even a name defined for the (minimum launched power) – (receiver sensitivity).  That is not a number that is expected to be used in applications.  

<Marek>This may be difficult unless we have a clear position and understanding on where to allocate the given penatlies. With 1 G standard, the problems were less accented since the data rate was lower. With 10G we will have a higher launch power, higher data rate and that alone means that we will have to take more impairments into consideration. The issue is only what Tanaka-san indicated before - how to divide them and attribute them to individual system components. </Marek> 

Anyway, that’s my two cents.  

Sincerely,

Frank Effenberger

8      Optical engineering approach

The selection of applications and set of optical parameters covered by this Recommendation are chosen to reflect a balance between economic and technical considerations to provide the possibility for transverse compatible systems using the synchronous digital hierarchy. This clause describes the use of the parameters in Tables 2 to 4 to obtain a common system design approach for engineering SDH optical links.

8.1   Design assumptions

To meet the greatest number of application possibilities with the smallest number of optical interface component specifications, three-interface categories are assumed for each level of the SDH hierarchy. These are distinguished by different attenuation/dispersion regimes rather than by explicit distance constraints to provide greater flexibility in network design while acknowledging technology and cost constraints for the various applications.

Worst-case, end-of-life parameter values are specified in this Recommendation to provide simple design guidelines for network planners and explicit component specifications for manufacturers. As a result, neither unallocated system margins nor equipment margins are specified and it is assumed that transmitters, receivers, and cable plant individually meet the specifications under the standard operating conditions. It is recognized that, in some cases, this may lead to more conservative system designs than could be obtained through joint engineering of the optical link, the use of statistical design approaches, or in applications and environments more constrained than those permitted under the standard operating conditions.

8.2   Worst-case design approach

For a worst-case design approach, the optical parameters of Tables 2 to 4 are related as shown in Figure 3. In loss-limited applications, a system integrator may determine the appropriate application code and corresponding set of optical parameters by first fixing the total optical path attenuation, which should include all sources of optical power loss and any cable design margin specified by the system integrator. For those situations in which the system attenuation falls within the attenuation overlap region of two applications, then either set of optical parameters would apply. The most economical designs will generally correspond to the application code having the narrower attenuation range. For each installation, it should be verified that the total optical path penalty, which includes combined dispersion and reflection degradations, does not exceed the value given in 6.4.4 and Tables 2 to 4 since a higher value may lead to rapidly deteriorating system performance.

Figure 3/G.957 – Relationship of the optical parameters

For dispersion-limited systems, the system integrator may select an appropriate application code and corresponding set of optical parameters by determining the total dispersion (ps/nm) expected for the elementary cable section to be designed. The most economical design generally corresponds to the selection of the application having the smallest maximum dispersion value exceeding the dispersion value determined for the system design. Again, the total optical path power penalty should be verified as described above.

 

 


From: Hajduczenia, Marek [mailto:marek.hajduczenia@SIEMENS.COM]
Sent: Tuesday, November 21, 2006 6:02 AM
To: STDS-802-3-10GEPON@listserv.ieee.org
Subject: [8023-10GEPON] Ad hoc on 10GEPON channel model
Importance: High

 

Dear all,

so far, 10 people registered for participation in the 10GEPON channel model ad-hoc. Here is the complete list:

 

1. Marek Hajduczenia (chair), Siemens Networks S.A. (marek.hajduczenia AT siemens.com)
2. Glen Kramer, Teknovus Inc. (glen.kramer AT teknovus.com)
3. Frank Effenberger, Hauwei (feffenberger AT huawei.com)
4. Sergey Y. Ten, Corning, (TenS AT Corning.com)
5. Keiji Tanaka, KDDI Labs, (kj-tanaka AT kddilabs.jp)
6. Eric Lynskey, Teknovus Inc., (eric.lynskey AT teknovus.com)
7. Paul Kolesar, CommScope Enterprise® Solutions, (PKOLESAR AT systimax.com)
8. Tsutomu Tatsuta, NTT, (tatsuta AT ansl.ntt.co.jp)
9. Robert Lingle, OFS Optics, (rlingle AT ofsoptics.com)
10. Dawe Piers, AvagoTech, (piers.dawe AT avagotech.com)

 

I would like to first thank everybody for their will to participate. Let's hope that this ad-hoc is as successful as the last one.

 

Regarding the tasks related with this particular ad-hoc, I would like to suggest some of them. Please feel free to provide any comments / feedback / suggestions.

 

Task 1: Channel model definitions:

    - conformance between 802.3 definitions and channel model spreadsheet

    - establish definitions for loss budget, channel insertion loss, power budget margin and its individual components

 

Task 2: Identification of necessary channel model extensions to be included in the new spreadsheet

    - non-linear transmission channel impairments (e.g. SBS?)

    - RF video overlay (?)

    - realistic PSC loss modelling

    - any other ?

 

Task 3: Overhaul of the Excel spreadsheet and update to meet requirements of the 10GEPON

    - separation of the parameter sheet and results sheet to increase readability ...

    - add missing channel impairment parameters (e.g. SBS threshold)

    - add RF channel modelling and include all related transmission impairments (+ penalties)

    - add PSC component modelling as well as include PSC loss in the channel model

    - any other which are identified by the ad-hoc

   

Task 4: Model compatibility issues

    - is Excel spreadsheet sufficient to model all the effects in 10GEPON system?

    - what are other options for publicly available software which could do the job better?

    - do we need to improve the modelling precision to remove 1-2 dB overshoot which was signalled in the case of 10GE links ?

    - any other compatibility issues ?

 

I would appreciate if the participants could identify the tasks which are of their primary interest. Feel free to email me with any suggestions regarding changes, wording, addition of tasks etc.

 

If possible, I would also like to start the discussion on the defintions of individual channel parameters, since it was suggested that we ought to clarify finally concepts like channel insertion loss and whether power budget margins should be internally divided into channel insertion loss margins and other components related with Rx/Tx equipment. I would like to suggest discussion based on the presentation from Tanaka-san, available at: http://grouper.ieee.org/groups/802/3/av/public/2006_11/3av_0611_tanaka_1.pdf

 

I am looking forward to hear Your opinions on this topic.

 

Best wishes

 

Marek Hajduczenia (141238)
SIEMENS Networks S.A. - IC COM D1 R
Rua Irmãos Siemens, 1
Ed. 1, Piso 1
Alfragide
2720-093 Amadora
Portugal
* Marek.Hajduczenia@siemens.com
http://marekhaj.easyisp.pl/index.php
(+351.21.416.7472  4+351.21.424.2082