Re: [802.3_MAINT] Follow up on proposed Comment to add distance specific

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

Hi Brad,

Yes, we had some discussion on these points during the call. I’ll post notes soon (over the weekend?).

Overall, we had a healthy discussion on the merits & demerits similar to the dialogue that has been present on the reflector. However, after that debate, my plan is to submit a comment based upon “traverso_1_0711” as it is posted.

One possible compromise (only discussed at a hazy outline level) was to have interested parties pursue a non-IEEE standard route with the detailed specifications required to achieve >400 m reach over OM4.

I’ll try to capture the details in my notes.

cheers

--matt

From: Brad Booth [mailto:bjbooth@xxxxxxxxx]
Sent: Wednesday, August 17, 2011 4:07 PM
To: STDS-802-3-MAINT@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_MAINT] Follow up on proposed Comment to add distance specification for 10GBASE-S over OM4 media

I was surprised to see no response to this. Not sure if it got addressed on today's conference call.

I do want to address the aspect of willingness. In my humble opinion, there would be greater willingness to support the proposal if the full set of 10GBASE-SR compliant parts could be used rather than a "subset". Without a distinct nomenclature to distinguish an optics device that complies with the subset versus one that complies with the 10GBASE-SR specification, there is no way to know how to achieve the greater reach than 400 m on OM4.

If you wish to create a new port type, then a new class of PMD is being created. To put it in terms of cabling, OM4 is a subset of OM3. But why is it called OM4 and why does it have its own specification if it is just a better performing OM3? (rhetorical question) If the cabling standards bodies have to create a new specification for this improved subset of OM3 (aka OM4), then IEEE 802.3 should be required to create a new specification for this new PMD. The other option is to create a proprietary version as you mentioned. MSA's, consortia, etc. do this all the time and that is their choice.

What impedes my willingness to support this modification to the standard is that in my opinion it is creating a new PMD. It creates a new SKU in manufacturing, in creating parts, in building ports, etc. And to create a new PMD, even if it is a subset, should be done as its own project and not through the revision of the standard. Just as Marek did a CFI to modify the EPON capabilities, so should this effort be handled in a similar manner. Only then can the working group decide if this should be studied and if so, later decide if it warrants a project.

Thanks,
Brad

On Tue, Aug 16, 2011 at 2:13 PM, Kolesar, Paul <PKOLESAR@xxxxxxxxxxxxx> wrote:

Rick Pimpinella of Panduit is working to subscribe to this reflector. In the mean time he asked me to share his spectral width data from their assortment of transceiver samples. With about twice the number of samples, and representation of eight manufacturers, it validates the Corning distribution.

All this data is simply addressing the potential availability of devices that could meet the proposed engineered link spectral width. With ~40% of the samples passing the 0.30nm proposal, it seems clear that the subset of 10GBASE-SR compliant parts is rather abundant. So the industry is quite capable of supporting this proposal.

Putting that aspect of the question aside as resolved, the context of the discussion shifts from capability to willingness.

While I must agree that it would be helpful (as well as the right thing to do) to give a reduced spectral width transceiver a name to distinguish it from its minimally-compliant 10G-SR brethren, it is not absolutely necessary to do so within the standard for this is not a new PMD. Rather, is a completely compliant subset of the existing PMD. There is no design change, nor a different parts list, technology, transmission technique, or wavelength. The manufacturer simply sorts the parts from existing production yield. It interoperates with its counterparts to the full capability of the less-capable part. This interoperability aspect would be clarified by granting a new name, but let’s look at the converse first.

There are plenty of examples of PMDs in the market that do not appear in the standard. -SR-lite, -USR, and –ZR are among the suffixes that come to mind. None of these need be compliant to anything because they are proprietary. Every proprietary offer has no assurance of interoperability, which is a cornerstone of why we develop standards. Yet the market copes with all that, and the problems it brings will not stop future proprietary offers.

Now contrast those cases with that of the engineered link proposal where the part remains completely compliant and non-proprietary. The compliance aspect is powerful reason to give it a nod in the standard.

So what it is called can be defined in the standard or left to the whims of proprietary offers. I propose that rather than let this blossom into a whole range of new proprietary suffixes, to take charge of it within the standard. If the remaining rub is really just what to call it, let me propose some names:

10GBASE-SRX

10GBASE-SRE

Here X stands for eXtended and E stands for Enhanced. Both of these should ring well for those who like to think of the definition of these suffixes in ways that relate to the customer rather than technology. Placement in the last position is well suited, as that location is often used for embellishments such as in –SR4, -SR10, -LX4.

Now some of you will likely come back at me with the argument that we don’t have a project to develop a new PMD or we can’t do a new PMD under a maintenance activity. To both arguments I say again that this is not a new PMD, it is a subset of an existing PMD. We are operating under a project called 802.3bh with a scope encompassing the entirety of the standard. No one has disputed the proposal based on technical merit. The numbers are good. The changes are minimal. And the benefit is harmony rather than possible confusion.

Regards,

Paul Kolesar

From: Brad Booth [mailto:bjbooth@xxxxxxxxx]
Sent: Tuesday, August 16, 2011 9:28 AM
To: STDS-802-3-MAINT@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_MAINT] Follow up on proposed Comment to add distance specification for 10GBASE-S over OM4 media

Steve,

The scenario you listed is not an apples-to-apples comparison, but rather an apples-to-oranges comparison.

I'm sure I don't need to remind you that the compliance point in an Ethernet network is at the medium dependent interface (MDI).

A customer that purchases a switch or server that has a port type 10GBASE-SR or 10GBASE-ER assumes that the device is compliant with the specifications at the MDI. In the case of the ER channel, if the channel is compliant with the specification - whether it is 30 km or an engineered 40 km link - then the system will be interoperable.

As for your concern, if the 500 m OM4 link was compliant to the channel specifications at the MDI, then there would be no issue. From what has been presented, the specification at the MDI looking into the PHY and into the channel needs to be changed to permit a 500 m OM4 link. This is radically different than what 10GBASE-ER requires, hence the inability to do an apples-to-apples comparison.

To try to require a tighter specification at the MDI while still calling the port type 10GBASE-SR is misleading. As an industry, we would be unable to guarantee interoperability to an end user that has purchase 10GBASE-SR optics for their devices and expect them to operate over 500 m of OM4, because the only requirement for 10GBASE-SR is to be "minimally" compliant with the specification at the MDI. Therefore, a new port type is required for your proposal.

To create an apple-to-apple comparison, don't touch the spectral width of the 10GBASE-S PMD. If you can "engineer" the 500 m OM4 channel to have the same performance specifications as a 400 m OM4 channel, then you would be creating a system similar to what 10GBASE-ER does.

This is just my take on it.

Cheers,

Brad

On Tue, Aug 16, 2011 at 8:49 AM, Swanson, Steven E <SwansonSE@xxxxxxxxxxx> wrote:

Dan,

Thanks for the comment; I think this subject will be discussed further in a future call but I would like to understand the concern here.

If someone releases a product tomorrow that is 100% compliant, but minimally meets the laser specs for SR, then he can support a minimum of 400m per the standard.

If someone releases a product tomorrow that is 100% compliant, but exceeds the minimum laser specs, fiber specs and connector specs for SR by a defined amount, then he can support a minimum of 500m per the standard (by definition, an engineered link).

Contrast this with:

If someone releases a product tomorrow that is 100% compliant, but minimally meets the single-mode fiber specs for ER, then he can support a minimum of 30km per the standard.

If someone releases a product tomorrow that is 100% compliant, but exceeds the minimum single-mode fiber specs for ER by a defined amount, then he can support a minimum of 40km per the standard (by definition, an engineered link).

But it is still 10GBASE-ER, right?

Steve

Steven E. Swanson
Corning Incorporated
800 17^th Street NW, HI/ES
Hickory, NC 28603-0489

t       828-901-5328
f       828-901-5533
c       607-725-1129

swansonse@xxxxxxxxxxx

From: Dove, Daniel [mailto:dan.dove@xxxxxx]
Sent: Monday, August 15, 2011 5:42 PM
To: STDS-802-3-MAINT@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_MAINT] Follow up on proposed Comment to add distance specification for 10GBASE-S over OM4 media

Hi Steve,

I tend to agree with Brad. We cannot specify the reach based upon empirical laser data. We must consider the specifications and assume that somebody is going to release a product tomorrow that is 100% compliant, but minimally meets the laser specs for SR.

The goal of this maintenance effort should stick to confirmation that existing SR optics (per spec) will support the reaches defined and I think we have done that. We can debate about how much margin should be necessary as this might allow a slight increase in reach, but 400m is a nice round number and retaining some margin will increase confidence in passing the spec readily.

Dan

From: Brad Booth [mailto:bjbooth@xxxxxxxxx]
Sent: Monday, August 15, 2011 12:35 PM
To: STDS-802-3-MAINT@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_MAINT] Follow up on proposed Comment to add distance specification for 10GBASE-S over OM4 media

Thanks Steve.

BTW, you make reference to last week's presentation. I remember reviewing it on the conference call but was it ever uploaded? I'd like to be able to reference back to it.

As mentioned on the call, the concern is lack of 100% coverage with available 10GBASE-S PMDs. If only a subset of PMDs would satisfy the spectral width requirements for achieving 500m reach, then that subset would require a new PMD name to distinguish them as exceeding the 10GBASE-S requirements. I believe some felt that we don't add a new PMD specification to 802.3 without a PAR that permits use to do so. A new PMD is probably considered by many to be outside the scope of the revision PAR.

Cheers,

Brad

On Mon, Aug 15, 2011 at 12:56 PM, Swanson, Steven E <SwansonSE@xxxxxxxxxxx> wrote:

Brad,

I didn't do it for spectral width because I attached a plot of all of the VCSEL spectral widths but here is the raw data:

0.2	1	1.43%
0.25	3	5.71%
0.3	25	41.43%
0.35	26	78.57%
0.4	11	94.29%
0.45	3	98.57%
More	1	100.00%

So in the proposal we made last week, ~40% of the lasers would support 500m with 1.1 dB margin; with a little less margin, I think .35nm will work suggesting that ~78% of the lasers would support the proposal.

Steven E. Swanson
Corning Incorporated
800 17^th Street NW, HI/ES
Hickory, NC 28603-0489

t       828-901-5328
f       828-901-5533
c       607-725-1129

swansonse@xxxxxxxxxxx

From: Brad Booth [mailto:bjbooth@xxxxxxxxx]
Sent: Monday, August 15, 2011 12:23 PM
To: STDS-802-3-MAINT@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_MAINT] Follow up on proposed Comment to add distance specification for 10GBASE-S over OM4 media

Steve,

In bullet points #2 and 3, you list the worst case wavelength and OMA power. Can you also do the same for bullet point #1 for spectral width?

Thanks,
Brad

On Mon, Aug 15, 2011 at 10:23 AM, Swanson, Steven E <SwansonSE@xxxxxxxxxxx> wrote:

Matt,

I had an error in my calculation:

1. The average spectral width is 0.317

2. The average wavelength is 850.6; worst case 845.7

3. The average OMA power is -1.41; worst case -2.42

Using these numbers in the IEEE model yields 4.9 dB of margin at 400m.

Using these numbers in the IEEE model yields 2.6 dB of margin at 550m.

Steven E. Swanson
Corning Incorporated
800 17^th Street NW, HI/ES
Hickory, NC 28603-0489

t       828-901-5328
f       828-901-5533
c       607-725-1129

swansonse@xxxxxxxxxxx

From: Swanson, Steven E [mailto:SwansonSE@xxxxxxxxxxx]
Sent: Monday, August 15, 2011 10:50 AM
To: STDS-802-3-MAINT@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_MAINT] Follow up on proposed Comment to add distance specification for 10GBASE-S over OM4 media

Matt etal,

I had our guys look at the VCSELs we have purchased on the open market used in our test facilities; there are 70 transceivers from 6 different manufacturers. 78% of the spectral widths are less than 0.35. The triple trade-off numbers are as follows:

1. The average spectral width is 0.286

2. The average wavelength is 849.4; worst case 845.7

3. The average OMA power is -1.24; worst case -2.42

Using these numbers in the IEEE model yields 5.2 dB of margin at 400m.

Using these numbers in the IEEE model yields 3.3 dB of margin at 550m.

The more I look at this, I think we are being REALLY, REALLY conservative in IEEE, maybe too much so; here is the distribution of spectral width:

Steven E. Swanson
Corning Incorporated
800 17^th Street NW, HI/ES
Hickory, NC 28603-0489

t       828-901-5328
f       828-901-5533
c       607-725-1129

swansonse@xxxxxxxxxxx

From: Matt Traverso (mattrave) [mailto:mattrave@xxxxxxxxx]
Sent: Thursday, July 28, 2011 9:05 PM
To: STDS-802-3-MAINT@xxxxxxxxxxxxxxxxx
Subject: [802.3_MAINT] Follow up on proposed Comment to add distance specification for 10GBASE-S over OM4 media

Hello to the now very active IEEE Maintenance reflector!

As I mentioned during my presentation at the IEEE plenary in July I would like to schedule a few calls to follow up on the questions/comments raised during the meeting. (see http://www.ieee802.org/3/maint/public/traverso_1_0711.pdf)

My goal for these calls per the discussion on the floor is to try to build more confidence in the proposed reach value. I believe that there was broad consensus at the meeting that (a) goal of specifying value is worthwhile,(b) theoretical analysis was sound. I hope to confirm my perception at the first call & discuss further justification for 400m reach.

My intention is to submit a comment by the comment deadline of Aug 26^th.

Call timing:

Wednesday Aug 10th, 10 – 11:30 am PT

Wednesday Aug 17^th , 10 – 11:30 am PT

Meeting login details below

best regards

--matt traverso

-+-----+-----+-----+-----+-----+-----+-----+-----+-

Meeting Number: 203 920 382

Meeting Password: OM4

-------------------------------------------------------

To start this meeting