Re: [802.3_NGEPON] Call for Contributions on PMD issues

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

John,

Please refer to slide 5 in http://www.ieee802.org/3/ca/public/meeting_archive/2016/03/harstead_3ca_1a_0316.pdf. This is from LightCounting, who gave me permission to show the chart publicly, but only if I removed the numbers from the vertical axis. Perhaps BRCM receives these reports and you can check the numbers for yourself (remember to multiply units by 4 to get the number of 25G lasers). Also, this does not include 25G Ethernet.

Anyway, you can see proportionally that about 2/3 of the forecasted units are SMF.

There were similar ramps of 100G Ethernet forecasted at datacenter-focused sessions at OFC. 40G Ethernet (4x10G) is going to be swapped out for 100G Ethernet.

From: John Johnson [mailto:john.johnson@xxxxxxxxxxxx]
Sent: Friday, July 15, 2016 5:01 PM
To: Harstead, Ed (Nokia - US)
Cc: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGEPON] Call for Contributions on PMD issues

Hi Ed,

Welcome back.

Are you including multi-mode lasers (VCSELS) in your datacom total? I think if you only consider single-mode lasers then the datacom volume is much smaller. It's a good discussion to have at our upcoming meeting.

Regards,

John

John E. Johnson, Ph.D.

Manager, III-V Component R&D | Fiber Optics Products Division
Broadcom

office: 484.397.2368 | mobile: 908.329.5710
9999 Hamilton Blvd. | Breinigsville, PA 18031 USA
john.johnson@xxxxxxxxxxxx | broadcom.com

On Fri, Jul 15, 2016 at 3:48 PM, Harstead, Ed (Nokia - US) <ed.harstead@xxxxxxxxx> wrote:

Hi all, sorry to respond so late to this email trail (was on vacation).

I do want to state a difference of opinion with John on the relative volumes of PON and telecom/datacenter lasers.

Without researching it, I beleive the cumulative volume to date of 10G telecom/datacenter lasers is orders of magnitude larger than 10G PON, and this is 6-7 years after 10G PON standardization. I expect similar for 25G. By the time 25G EPON volume starts to ramp, there will already have been 10s of millions of 25G lasers deployed in datacenters, maybe 100M (according to analysts). It will take a very long time for 25G EPON to catch up with that.

From: John Johnson [mailto:000007ff7d378f43-dmarc-request@xxxxxxxx]
Sent: Tuesday, June 28, 2016 5:14 PM

To: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGEPON] Call for Contributions on PMD issues

Hi Frank,

Certainly the 100G Ethernet applications have driven a lot of the technology development for 25Gb/s per channel optics - EML, DML and mux/demux - and that is the starting point for development of NG-EPON optics. How much could be directly shared remains to be seen given the very different specs and cost targets of the two applications. As you point out we will reuse whatever pieces we can consistent with the objectives.

My position is that we should be careful to not align NG-EPON with existing optics for the sole purpose to have some leverage during the early deployments if it would compromise the economics in the long run. NG-EPON volume will eventually dwarf the telecom/datacom volume. In fact I think it more likely that future 100G+ point-to-point optical components will be driven by the high volume, low cost infrastructure that we develop for NG-EPON rather than the other way around. There are many examples of infrastructure and technology put in place for PON being adopted for telecom/datacom applications, including cooled TO-can 10G EML, 40GBASE-LR4 and more recently 100G-CWDM4.

I look forward to further discussion on this topic.

Regards,

John E. Johnson, Ph.D.

On Tue, Jun 28, 2016 at 10:44 AM, frank effenberger <frank.effenberger@xxxxxxxxxx> wrote:

All,

It is true that these components are not going to be the same exact chip. But I think we crossed that bridge once we said the work “PON” – our loss budget is so much bigger, it’s not even funny. However, even if there is a different die design, using the same wavelengths could give us the possibility of reusing some sub-elements. The same wafer design should be usable, for instance, as well as the wavelength multiplexing filters. I think we all have a natural tendency to reuse what already exists, if it works for us (and that is a question).

That’s why on page 8, I state that reuse of LR4 is one option, but that the spacing (and by inference the actual channels) is a question-mark.

Hopefully we can get more input on that and the availability of high power DMLs, and their impact on the wavelength choice.

Frank E

From: John Johnson [mailto:000007ff7d378f43-dmarc-request@xxxxxxxx]
Sent: Tuesday, June 28, 2016 9:57 AM

To: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGEPON] Call for Contributions on PMD issues

All,

One comment on the component technology:

Commonality with 100G Ethernet optics would be a great benefit, but the huge difference in required output powers for point-to-point and PON links makes it unlikely. It's not possible to make 100G-LR4 optics with 7dB more power without fundamentally changing the underlying technology. From a component vendor's point of view these are completely different products even if they share the same wavelength plan. More likely the development of high power optics for NG-EPON could enable future lower cost 100G Ethernet optics with longer reach.

Regards,

John

John E. Johnson, Ph.D.

On Wed, Jun 22, 2016 at 5:15 PM, Marek Hajduczenia <marek.hajduczenia@xxxxxxxxx> wrote:

Shawn,

Some thoughts inline in red

Regards

Marek Hajduczenia, PhD | Network Architect, Principal Engineer

4145 S. Falkenburg Rd | Riverview Fl 33578

================================

Cisco certified: CCNA CSCO12874393

ccna_routerswitching_sm

From: Shawn Esser [mailto:shawn.esser@xxxxxxxxxxx]
Sent: Wednesday, June 22, 2016 4:49 PM
To: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_NGEPON] Call for Contributions on PMD issues

Frank,

Great job in putting this together. Some comments:

1. On Slide 3: There also has been discussion that NGEPON has upstream options of 25Gb/s and 10Gb/s on one or more 25G layers:

a. 10Gb/s option has not been decided. How do we come to a decision? [mh0622] A simple motion would suffice in this case. It is an option that has been discussed for a while, covering the reuse of by then legacy 10G-EPON optics for upstream direction.

b. Is the 10Gb/s option on all four 25G layers? For example, would there be an option for 100G/40G? [mh0622] My take on it would be simple: we have 10G defined today under 802.3av and only full reuse of this optics makes sense. Adding 4 x 10G upstream requires new optics altogether. At that point, it is better to focus on 25G optics and avoid market fragmentation.

c. How would NGEPON upstream at 25Gb/s and 10Gb/s co-exist with each other on the same 25G layer on the same PON segment? TDM or WDM? Or they don’t co-exist on the same PON segment? [mh0622] That depends on where new 25G wavelengths are placed. If WDM coexistence with 10G-EPON upstream could be achieved, it would be beneficial, allowing first gen NG-EPON ONUs running 25x10 and 25x25 to coexist on the network.

d. How would NGEPON 10Gb/s upstream coexist with 10GEPON upstream? TDM or WDM? Can TDM work for this? Could the 10GEPON OLT and NGEPON OLT distinguish between 10Gb/s data from 10GEPON ONU or NGEPON ONU if they are on same wavelength? [mh0622] If we reuse 10G-EPON upstream optics, any coexistence would have to fall under TDM. Otherwise, we’re talking about new type of 10G optics, which defeats the purpose of reuse altogether.

2. On Slide 3: This is confusing to me and may need further explanation for others also:

How do 25G, 50G, and 100G coexist?
- It seems “TDM” type coexistence is preferred

The four 25G layers will each be on their own wavelength pair so on the surface it seems like “WDM”. [mh0622] it seems like a loaded question: if 25G, 50G and 100G have to coexist via WDM, they have to have different wavelengths (7 in total). Otherwise, we need TDM and share some wavelengths between 25G, 50G, and 100G devices. We had pretty involved discussion on this at the last meeting.

3. On Slide 8: In my mind, It has not been settled that uncooled lasers cannot meet the power budget requirement or will not meet it in the future. One vendor stated that uncooled laser could achieve +6dBm output power today under some conditions. The technology could improve further over a few years so uncooled lasers could meet the power budget. Since uncooled lasers can offer significantly lower ONU costs, the team may not want to preclude uncooled lasers for upstream wavelength 0 by specifying a narrow wavelength range. Since there is only limited spectrum available, most likely cannot have 8 wavelengths, or even 4 wavelengths, with a wide wavelength range (~20nm) – maybe only for upstream wavelength 0 which is the most cost-sensitive.

4. Add call for further study on relative cost comparison for options on 25G ONU wavelength pair 0 since this could influence decisions. Specific areas:

a. Uncooled DML vs. Cooled DML vs. Cooled EML upstream laser

b. ONU diplexer: Focus Beam Filter versus Collimated Beam along with the optimal size of the minimum edge gap between the US wavelength and downstream wavelength. The edge gap will impact the wavelength plan and ONU cost. Do we rule out other filter technologies (i.e. thin film) that can provide much smaller edge gaps?

5. BTW, what does the acronym “PMD” mean? [mh0622] PMD = Physical Medium Dependent

Thanks,

Shawn

From: frank effenberger [mailto:frank.effenberger@xxxxxxxxxx]
Sent: Wednesday, June 22, 2016 11:34 AM
To: STDS-802-3-NGEPON@xxxxxxxxxxxxxxxxx
Subject: [802.3_NGEPON] Call for Contributions on PMD issues

All,

Please see attached the “call for contributions” on PMD issues.

Unfortunately, I will be unable to attend the meeting tomorrow, as there is another standard meeting happening at that time.

But I would love to receive Email correspondence on this CFC – to either enhance it, or to express interest in working on a topic.

It would be highly efficient if we can identify the groups of people interested on each topic, so that they can create joint contributions that are more definitive. This is particularly true of the topics that are mainly analytic (e.g., analysis of fiber nonlinearity impacts).

Thank you,

Frank Effenberger