Re: [802.3_NGMMF] what offsets occur for connectors which meet our speci

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

All, this thread is being moved to the Reflector, in case there is further discussion.

From: Kolesar, Paul <PKOLESAR@xxxxxxxxxxxxx>
Sent: Saturday, September 29, 2018 9:10 AM
To: Petar Pepeljugoski <petarp@xxxxxxxxxx>; Dudek, Mike <Mike.Dudek@xxxxxxxxxx>
Cc: abbottjs@xxxxxxxxxxx; Adrian Amezcua (Prysmian) <Adrian.Amezcua@xxxxxxxxxxxxxxxxx>; Brad Booth <brbooth@xxxxxxxxxxxxx>; Brett Lane (Panduit) <Brett.Lane@xxxxxxxxxxx>; Bulent Kose (Panduit) <Bulent.Kose@xxxxxxxxxxx>; Chow, Bruce (Corning) <chowbc@xxxxxxxxxxx>; Chris Cole (Finisar) <chris.cole@xxxxxxxxxxx>; Carlo Mariotti (cmariott) <cmariott@xxxxxxxxx>; Carlo Tosetti (ctosetti) <ctosetti@xxxxxxxxx>; Dale Murray - LightCounting <dale@xxxxxxxxxxxxxxxxx>; Harris Dave, US <Dave.Harris@xxxxxxxxxxxxxxxxx>; David Lewis (Lumentum) <David.Lewis@xxxxxxxxxxxx>; David.Piehler@xxxxxxxx; David_Piehler@xxxxxxxx; derek.cassidy@xxxxxx; fmarques@xxxxxxxxxxxxxx; fymchang@xxxxxxxxx; George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>; Gary Nicholl (Cisco) <gnicholl@xxxxxxxxx>; greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx; Darryl Heckle <HeckleDC@xxxxxxxxxxx>; hyaku@xxxxxxxxxxxxx; Indris <Iseman85@xxxxxxxxx>; james.withey@xxxxxxxxxxxxxxxxx; Kamino, John T (John) <jkamino@xxxxxxxxxxxxx>; Jeffery Maki <jmaki@xxxxxxxxxxx>; John Johnson (Broadcom) <john.johnson@xxxxxxxxxxxx>; John F Petrilla <john.petrilla@xxxxxxxxxxxxxxxx>; john.yurtin@xxxxxxxxx; Jonathan David Ingham <jonathan.ingham@xxxxxxxxxxxxxxxx>; Jonathan King <jonathan.king@xxxxxxxxxxx>; Joost.grillaert@xxxxxxxxxx; Jose Castro (Panduit) <Jose.Castro@xxxxxxxxxxx>; Kenneth Jackson (Sumitomo) <kjackson@xxxxxxxxxxxxxx>; Kobi Hasharoni <kobi.hasharoni@xxxxxxxxxxxxxxxxx>; kpjackson@xxxxxxxx; kshrikhande@xxxxxxxxxxxx; liyongyao@xxxxxxxxxx; Mark.ives@xxxxxxxxxxx; McCurdy, Alan H (Alan) <mccurdy@xxxxxxxxxxxxx>; Choudhury, Mabud (Mabud) <MChoudhury@xxxxxxxxxxxxx>; mgustli@xxxxxxxxxx; Mark Nowell (Cisco) <mnowell@xxxxxxxxx>; Paul E. Neveaux Jr. Ph. D. <paul.neveaux@xxxxxxxx>; Paul Vanderlaan (Nexans) <paul.vanderlaan@xxxxxxxxxx>; petejone@xxxxxxxxx; phongpham@xxxxxxxxxxx; Piers Dawe <piersd@xxxxxxxxxxxx>; Pondillo, Peter (Corning) <PondilloPL@xxxxxxxxxxx>; Samamra Raed, US <raed.samamra@xxxxxxxxxxxxxxxxx>; Rakesh Sambaraju Ph. D <rakesh.sambaraju@xxxxxxxxxx>; Ramana Murty <ramana.murty@xxxxxxxxxxxx>; rhorner@xxxxxxxxxxxx; Rick Pimpinella <Rick.Pimpinella@xxxxxxxxxxx>; Rick Pimpinella (Panduit) <rjpi@xxxxxxxxxxx>; Lingle, Robert L (Robert) <rlingle@xxxxxxxxxxxxx>; Tellas, Ronald <Ronald.Tellas@xxxxxxxxxx>; salvatore.rotolo@xxxxxx; sam_sambasivan@xxxxxxxxxxxx; scott.sommers@xxxxxxxxx; shawn.nicholl@xxxxxxxxxx; Steffen Koehler (Finisar) <steffen.koehler@xxxxxxxxxxx>; Xu, Yang (Sunny) <Sunny.Xu@xxxxxxxxxxxxx>; swansonse@xxxxxxxxxxx; takeshin@xxxxxxx; tommitcheltree@xxxxxxxxxxx; Ted Sprague <tsprague@xxxxxxxxxxxx>; ureddy@xxxxxxxxx; vbala@xxxxxxxxxxxx; vineets@xxxxxxxxx; Vipul Bhatt <vipul.bhatt@xxxxxxxxxxx>; wu.chengbin@xxxxxxxxxx; yang.zhiwei2@xxxxxxxxxx; yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx; zhang.yuanbin@xxxxxxxxxx; zhongqiwen@xxxxxxxxxx; zshen@xxxxxxxxxx
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

Mike,

Good point about the Rayleigh distribution. The upper tail theoretically extends to infinity, much like a normal distribution. In practice, the upper end gets truncated by manufacturing screens. But in the model for the channels it likely has not been truncated. That would drive connection losses above 0.75 dB 5% of the time, and the maximum loss to be unbounded. It would be insightful to see the impact of applying that truncation on the MN prediction.

Petar,

If a connection falls out of spec due to dirt build up sufficient to cause bit errors ratios to exceed limits because of diminished power and/or added MN and/or reflections, then diagnosis of the channel troubles would lead one to clean the connection to remedy. I do not think we need to concern ourselves with ensuring the channel functions flawlessly under those out of spec conditions.

Regards,

Paul

From: Petar Pepeljugoski <petarp@xxxxxxxxxx>
Sent: Friday, September 28, 2018 9:28 PM
To: Dudek, Mike <Mike.Dudek@xxxxxxxxxx>
Cc: abbottjs@xxxxxxxxxxx; Adrian Amezcua (Prysmian) <Adrian.Amezcua@xxxxxxxxxxxxxxxxx>; Brad Booth <brbooth@xxxxxxxxxxxxx>; Brett Lane (Panduit) <Brett.Lane@xxxxxxxxxxx>; Bulent Kose (Panduit) <Bulent.Kose@xxxxxxxxxxx>; Chow, Bruce (Corning) <chowbc@xxxxxxxxxxx>; Chris Cole (Finisar) <chris.cole@xxxxxxxxxxx>; Carlo Mariotti (cmariott) <cmariott@xxxxxxxxx>; Carlo Tosetti (ctosetti) <ctosetti@xxxxxxxxx>; Dale Murray - LightCounting <dale@xxxxxxxxxxxxxxxxx>; Harris Dave, US <Dave.Harris@xxxxxxxxxxxxxxxxx>; David Lewis (Lumentum) <David.Lewis@xxxxxxxxxxxx>; David.Piehler@xxxxxxxx; David_Piehler@xxxxxxxx; derek.cassidy@xxxxxx; fmarques@xxxxxxxxxxxxxx; fymchang@xxxxxxxxx; George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>; Gary Nicholl (Cisco) <gnicholl@xxxxxxxxx>; greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx; Darryl Heckle <HeckleDC@xxxxxxxxxxx>; hyaku@xxxxxxxxxxxxx; Indris <Iseman85@xxxxxxxxx>; james.withey@xxxxxxxxxxxxxxxxx; Kamino, John T (John) <jkamino@xxxxxxxxxxxxx>; Jeffery Maki <jmaki@xxxxxxxxxxx>; John Johnson (Broadcom) <john.johnson@xxxxxxxxxxxx>; John F Petrilla <john.petrilla@xxxxxxxxxxxxxxxx>; john.yurtin@xxxxxxxxx; Jonathan David Ingham <jonathan.ingham@xxxxxxxxxxxxxxxx>; Jonathan King <jonathan.king@xxxxxxxxxxx>; Joost.grillaert@xxxxxxxxxx; Jose Castro (Panduit) <Jose.Castro@xxxxxxxxxxx>; Kenneth Jackson (Sumitomo) <kjackson@xxxxxxxxxxxxxx>; Kobi Hasharoni <kobi.hasharoni@xxxxxxxxxxxxxxxxx>; kpjackson@xxxxxxxx; kshrikhande@xxxxxxxxxxxx; liyongyao@xxxxxxxxxx; Mark.ives@xxxxxxxxxxx; McCurdy, Alan H (Alan) <mccurdy@xxxxxxxxxxxxx>; Choudhury, Mabud (Mabud) <MChoudhury@xxxxxxxxxxxxx>; mgustli@xxxxxxxxxx; Mark Nowell (Cisco) <mnowell@xxxxxxxxx>; Paul E. Neveaux Jr. Ph. D. <paul.neveaux@xxxxxxxx>; Paul Vanderlaan (Nexans) <paul.vanderlaan@xxxxxxxxxx>; petejone@xxxxxxxxx; phongpham@xxxxxxxxxxx; Piers Dawe <piersd@xxxxxxxxxxxx>; Kolesar, Paul <PKOLESAR@xxxxxxxxxxxxx>; Pondillo, Peter (Corning) <PondilloPL@xxxxxxxxxxx>; Samamra Raed, US <raed.samamra@xxxxxxxxxxxxxxxxx>; Rakesh Sambaraju Ph. D <rakesh.sambaraju@xxxxxxxxxx>; Ramana Murty <ramana.murty@xxxxxxxxxxxx>; rhorner@xxxxxxxxxxxx; Rick Pimpinella <Rick.Pimpinella@xxxxxxxxxxx>; Rick Pimpinella (Panduit) <rjpi@xxxxxxxxxxx>; Lingle, Robert L (Robert) <rlingle@xxxxxxxxxxxxx>; Tellas, Ronald <Ronald.Tellas@xxxxxxxxxx>; salvatore.rotolo@xxxxxx; sam_sambasivan@xxxxxxxxxxxx; scott.sommers@xxxxxxxxx; shawn.nicholl@xxxxxxxxxx; Steffen Koehler (Finisar) <steffen.koehler@xxxxxxxxxxx>; Xu, Yang (Sunny) <Sunny.Xu@xxxxxxxxxxxxx>; swansonse@xxxxxxxxxxx; takeshin@xxxxxxx; tommitcheltree@xxxxxxxxxxx; Ted Sprague <tsprague@xxxxxxxxxxxx>; ureddy@xxxxxxxxx; vbala@xxxxxxxxxxxx; vineets@xxxxxxxxx; Vipul Bhatt <vipul.bhatt@xxxxxxxxxxx>; wu.chengbin@xxxxxxxxxx; yang.zhiwei2@xxxxxxxxxx; yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx; zhang.yuanbin@xxxxxxxxxx; zhongqiwen@xxxxxxxxxx; zshen@xxxxxxxxxx
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

Email Security Warning:

The following message was sent from an external e-mail address. Exercise caution when opening attachments, clicking links, or exchanging information.

Thanks for all this information. I appreciate it. Very enlightening. The summary seems to be that the answer to my original question is 7um or close to it. I’d like to check whether the simulated channels used to develop OM3 and VCSEL launch specs 17 or so years ago are the same channels that have been used for the modal noise analysis. Also I may have misunderstood something but your quote says “We used a Rayleigh distribution for the connector offset with parameters adjusted such that 95% of the values were less than 7 µm” To me that means that 5% of the values exceeded 7um and are likely to be exceeding the 0.75dB loss limit for the connectors. If the same channels were used for the modal noise analysis then there is a reasonable argument that these 5% of the channels should have been removed and I wonder whether that would have removed all the higher modal noise results.

From:Kolesar, Paul <PKOLESAR@xxxxxxxxxxxxx>
Sent: Friday, September 28, 2018 11:57 AM
To: Petar Pepeljugoski <petarp@xxxxxxxxxx>; Dudek, Mike <Mike.Dudek@xxxxxxxxxx>
Cc: abbottjs@xxxxxxxxxxx; Adrian Amezcua (Prysmian) <Adrian.Amezcua@xxxxxxxxxxxxxxxxx>; Brad Booth <brbooth@xxxxxxxxxxxxx>; Brett Lane (Panduit) <Brett.Lane@xxxxxxxxxxx>; Bulent Kose (Panduit) <Bulent.Kose@xxxxxxxxxxx>; Chow, Bruce (Corning) <chowbc@xxxxxxxxxxx>; Chris Cole (Finisar) <chris.cole@xxxxxxxxxxx>; Carlo Mariotti (cmariott) <cmariott@xxxxxxxxx>; Carlo Tosetti (ctosetti) <ctosetti@xxxxxxxxx>; Dale Murray - LightCounting <dale@xxxxxxxxxxxxxxxxx>; Harris Dave, US <Dave.Harris@xxxxxxxxxxxxxxxxx>; David Lewis (Lumentum) <David.Lewis@xxxxxxxxxxxx>; David.Piehler@xxxxxxxx; David_Piehler@xxxxxxxx; derek.cassidy@xxxxxx; fmarques@xxxxxxxxxxxxxx; fymchang@xxxxxxxxx; George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>; Gary Nicholl (Cisco) <gnicholl@xxxxxxxxx>; greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx; Darryl Heckle <HeckleDC@xxxxxxxxxxx>; hyaku@xxxxxxxxxxxxx; Indris <Iseman85@xxxxxxxxx>; james.withey@xxxxxxxxxxxxxxxxx; Kamino, John T (John) <jkamino@xxxxxxxxxxxxx>; Jeffery Maki <jmaki@xxxxxxxxxxx>; John Johnson (Broadcom) <john.johnson@xxxxxxxxxxxx>; John F Petrilla <john.petrilla@xxxxxxxxxxxxxxxx>; john.yurtin@xxxxxxxxx; Jonathan David Ingham <jonathan.ingham@xxxxxxxxxxxxxxxx>; Jonathan King <jonathan.king@xxxxxxxxxxx>; Joost.grillaert@xxxxxxxxxx; Jose Castro (Panduit) <Jose.Castro@xxxxxxxxxxx>; Kenneth Jackson (Sumitomo) <kjackson@xxxxxxxxxxxxxx>; Kobi Hasharoni <kobi.hasharoni@xxxxxxxxxxxxxxxxx>; kpjackson@xxxxxxxx; kshrikhande@xxxxxxxxxxxx; liyongyao@xxxxxxxxxx; Mark.ives@xxxxxxxxxxx; McCurdy, Alan H (Alan) <mccurdy@xxxxxxxxxxxxx>; Choudhury, Mabud (Mabud) <MChoudhury@xxxxxxxxxxxxx>; mgustli@xxxxxxxxxx; Mark Nowell (Cisco) <mnowell@xxxxxxxxx>; Paul E. Neveaux Jr. Ph. D. <paul.neveaux@xxxxxxxx>; Paul Vanderlaan (Nexans) <paul.vanderlaan@xxxxxxxxxx>; petejone@xxxxxxxxx; phongpham@xxxxxxxxxxx; Piers Dawe <piersd@xxxxxxxxxxxx>; Pondillo, Peter (Corning) <PondilloPL@xxxxxxxxxxx>; Samamra Raed, US <raed.samamra@xxxxxxxxxxxxxxxxx>; Rakesh Sambaraju Ph. D <rakesh.sambaraju@xxxxxxxxxx>; Ramana Murty <ramana.murty@xxxxxxxxxxxx>; rhorner@xxxxxxxxxxxx; Rick Pimpinella <Rick.Pimpinella@xxxxxxxxxxx>; Rick Pimpinella (Panduit) <rjpi@xxxxxxxxxxx>; Lingle, Robert L (Robert) <rlingle@xxxxxxxxxxxxx>; Tellas, Ronald <Ronald.Tellas@xxxxxxxxxx>; salvatore.rotolo@xxxxxx; sam_sambasivan@xxxxxxxxxxxx; scott.sommers@xxxxxxxxx; shawn.nicholl@xxxxxxxxxx; Steffen Koehler (Finisar) <steffen.koehler@xxxxxxxxxxx>; Xu, Yang (Sunny) <Sunny.Xu@xxxxxxxxxxxxx>; swansonse@xxxxxxxxxxx; takeshin@xxxxxxx; tommitcheltree@xxxxxxxxxxx; Ted Sprague <tsprague@xxxxxxxxxxxx>; ureddy@xxxxxxxxx; vbala@xxxxxxxxxxxx; vineets@xxxxxxxxx; Vipul Bhatt <vipul.bhatt@xxxxxxxxxxx>; wu.chengbin@xxxxxxxxxx; yang.zhiwei2@xxxxxxxxxx; yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx; zhang.yuanbin@xxxxxxxxxx; zhongqiwen@xxxxxxxxxx; zshen@xxxxxxxxxx
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

External Email

Mike,

When we built the simulated channels used to develop OM3 and VCSEL launch specs 17 or so years ago, we did consider the 0.75 dB max loss limit for individual connections. I believe we used nominal CD and NA 50um fiber (with 5000 different mode delay structures) and limited the connection offset the maximum tolerable to meet the 0.75 dB loss spec. Here is a quote from the JLT article that documented the development of the system specification.

We used a Rayleigh distribution for the connector offset with parameters adjusted such

that 95% of the values were less than 7 µm. This translates into a mean of approximately 3.58 µm.

Since that time there have been contributions to IEC showing the loss between nominal NA and CD 50um fibers carrying mode power distributions meeting the lost test set EF launch condition at 850 nm. This curve hits 0.75 dB at 7 um offset corroborating our choice.

The mismatch in NA and CD between fibers causes mode mixing and mode selective loss. There may be some small amount of uniform loss also, but I think it is dominated by MSL. Fibers of different NA and CD carry different numbers of modes, hence there is not a perfect correspondence between such fibers. However, the loss going from a small CD and NA fiber into a fiber with larger CD and NA can be zero when they are well aligned. This shows that the modes don’t have to match, but the receiving fiber has to be able to carry the ones it sees from the transmitting fiber. Going in the opposite direction, losses will exist, caused predominantly by the highest order modes that can’t be carried by the smaller CD and NA fiber, thus it is mode selective.

Regarding the dependence of MN on the number of modes, I think it’s the number of modes the fiber can carry at the wavelength of interest, not the number of modes launched into the fiber, that matters, at least for sources that are centered on the core. I believe that fibers with smaller number of modes, either due to smaller NA and CD (e.g. 50 um vs 62.5 um) or due to wavelength (1300 nm vs 850 nm) can have larger speckles and therefore more MN generation potential.

It is not difficult to create fiber assemblies with fiber from the same spool. You probably can’t easily order them from a manufacturer, but they can be made in the lab.

Years ago the cabling industry tackled the problem of assessment of loss on multimode channels that are used for VCSEL applications. We moved away from overfilled launch to a well specified launch based on the encircled flux metric. The launch spec tightly constrains the EF in the outer radii of the fiber to be near the maximum fill limit that VCSELs may produce according to 802.3 transmitter launch specifications, thus providing a conservative assessment of the channel loss for VCSEL applications, because it is the high order modes that are selectively lost at connections. As you said, that removes the margin that was discussed yesterday for worst-case VCSEL launches.

But most VCSEL launches are not at this limit. Instead they are less filling, with mode power distributions that evolve thru connections towards more filling conditions. I believe that the mode mixing events at connections not only expand the mode power distribution, but also reduce speckle contrast due to mixing of fields that are not in phase due to differential mode delay and/or different wavelength. So by the time that a typical VCSEL launch power distribution is scattered at connections sufficiently to excite high order modes with the potential to create significant MN at subsequent connections, it may also have been reduced in speckle contrast that mitigates the generation of MN. I think simulating all of that is quite difficult. Yes, we have connector coupling matrices that predict mode power redistributions. But I do not recall such simulations predicting speckle contrast. I may be wrong, as Petar said that his simulations carried phase information. If true, I’d like to understand this better.

Regards,

Paul

From:Petar Pepeljugoski <petarp@xxxxxxxxxx>
Sent: Friday, September 28, 2018 1:03 PM
To: Dudek, Mike <Mike.Dudek@xxxxxxxxxx>
Cc: abbottjs@xxxxxxxxxxx; Adrian Amezcua (Prysmian) <Adrian.Amezcua@xxxxxxxxxxxxxxxxx>; Brad Booth <brbooth@xxxxxxxxxxxxx>; Brett Lane (Panduit) <Brett.Lane@xxxxxxxxxxx>; Bulent Kose (Panduit) <Bulent.Kose@xxxxxxxxxxx>; Chow, Bruce (Corning) <chowbc@xxxxxxxxxxx>; Chris Cole (Finisar) <chris.cole@xxxxxxxxxxx>; Carlo Mariotti (cmariott) <cmariott@xxxxxxxxx>; Carlo Tosetti (ctosetti) <ctosetti@xxxxxxxxx>; Dale Murray - LightCounting <dale@xxxxxxxxxxxxxxxxx>; Harris Dave, US <Dave.Harris@xxxxxxxxxxxxxxxxx>; David Lewis (Lumentum) <David.Lewis@xxxxxxxxxxxx>; David.Piehler@xxxxxxxx; David_Piehler@xxxxxxxx; derek.cassidy@xxxxxx; fmarques@xxxxxxxxxxxxxx; fymchang@xxxxxxxxx; George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>; Gary Nicholl (Cisco) <gnicholl@xxxxxxxxx>; greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx; Darryl Heckle <HeckleDC@xxxxxxxxxxx>; hyaku@xxxxxxxxxxxxx; Indris <Iseman85@xxxxxxxxx>; james.withey@xxxxxxxxxxxxxxxxx; Kamino, John T (John) <jkamino@xxxxxxxxxxxxx>; Jeffery Maki <jmaki@xxxxxxxxxxx>; John Johnson (Broadcom) <john.johnson@xxxxxxxxxxxx>; John F Petrilla <john.petrilla@xxxxxxxxxxxxxxxx>; john.yurtin@xxxxxxxxx; Jonathan David Ingham <jonathan.ingham@xxxxxxxxxxxxxxxx>; Jonathan King <jonathan.king@xxxxxxxxxxx>; Joost.grillaert@xxxxxxxxxx; Jose Castro (Panduit) <Jose.Castro@xxxxxxxxxxx>; Kenneth Jackson (Sumitomo) <kjackson@xxxxxxxxxxxxxx>; Kobi Hasharoni <kobi.hasharoni@xxxxxxxxxxxxxxxxx>; kpjackson@xxxxxxxx; kshrikhande@xxxxxxxxxxxx; liyongyao@xxxxxxxxxx; Mark.ives@xxxxxxxxxxx; McCurdy, Alan H (Alan) <mccurdy@xxxxxxxxxxxxx>; Choudhury, Mabud (Mabud) <MChoudhury@xxxxxxxxxxxxx>; mgustli@xxxxxxxxxx; Mark Nowell (Cisco) <mnowell@xxxxxxxxx>; Paul E. Neveaux Jr. Ph. D. <paul.neveaux@xxxxxxxx>; Paul Vanderlaan (Nexans) <paul.vanderlaan@xxxxxxxxxx>; petejone@xxxxxxxxx; phongpham@xxxxxxxxxxx; Piers Dawe <piersd@xxxxxxxxxxxx>; Kolesar, Paul <PKOLESAR@xxxxxxxxxxxxx>; Pondillo, Peter (Corning) <PondilloPL@xxxxxxxxxxx>; Samamra Raed, US <raed.samamra@xxxxxxxxxxxxxxxxx>; Rakesh Sambaraju Ph. D <rakesh.sambaraju@xxxxxxxxxx>; Ramana Murty <ramana.murty@xxxxxxxxxxxx>; rhorner@xxxxxxxxxxxx; Rick Pimpinella <Rick.Pimpinella@xxxxxxxxxxx>; Rick Pimpinella (Panduit) <rjpi@xxxxxxxxxxx>; Lingle, Robert L (Robert) <rlingle@xxxxxxxxxxxxx>; Tellas, Ronald <Ronald.Tellas@xxxxxxxxxx>; salvatore.rotolo@xxxxxx; sam_sambasivan@xxxxxxxxxxxx; scott.sommers@xxxxxxxxx; shawn.nicholl@xxxxxxxxxx; Steffen Koehler (Finisar) <steffen.koehler@xxxxxxxxxxx>; Xu, Yang (Sunny) <Sunny.Xu@xxxxxxxxxxxxx>; swansonse@xxxxxxxxxxx; takeshin@xxxxxxx; tommitcheltree@xxxxxxxxxxx; Ted Sprague <tsprague@xxxxxxxxxxxx>; ureddy@xxxxxxxxx; vbala@xxxxxxxxxxxx; vineets@xxxxxxxxx; Vipul Bhatt <vipul.bhatt@xxxxxxxxxxx>; wu.chengbin@xxxxxxxxxx; yang.zhiwei2@xxxxxxxxxx; yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx; zhang.yuanbin@xxxxxxxxxx; zhongqiwen@xxxxxxxxxx; zshen@xxxxxxxxxx
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

Hi all,

The fiber standards (TIA-492-xxxx) specify:

4.2.1 core diameter: 50+/- 2.5um
4.2.5 core cladding concentricity error: <3um
4.2.3 cladding diameter 125 +/- 2.0um
4.2.2 core non-circularity <6%
4.2.4. cladding non-circularity <2%

example: on one side fiber with concentricity error towards the left, worst case received fiber with same concentricity error aligned on the opposite side. Total offset 6um

On top of these, you need to consider your connector effects,i.e ferule size, ferule hole diameter (fiber needs to fit in so there will be some misalignment there), as well as sleeve size. Ceramic might be better but are more expensive.

Anyway, my point, as I said during the teleconference is that these can add up. I am sure the statistics will take care so we do not just add them, but then the 3 or whatever sigma we need will push us towards the larger offsets. I think that offsets up to 7.5 um are reasonable and can be frequently encountered, and the question is how often those in the range 7.5-10 occur.

As for the alignment tolerances on the TX side to fiber, we were considering up to 10um to be typical, however I would defer to the module manufacturers to supply the details if they can.

Regards,

Peter

Petar Pepeljugoski
IBM Research
P.O.Box 218 (mail)
1101 Kitchawan Road, Rte. 134 (shipping)
Yorktown Heights, NY 10598

e-mail: petarp@xxxxxxxxxx
phone: (914)-945-3761
fax: (914)-945-4134

From: "Dudek, Mike" <Mike.Dudek@xxxxxxxxxx>
To: "Kolesar, Paul" <PKOLESAR@xxxxxxxxxxxxx>, John F Petrilla <john.petrilla@xxxxxxxxxxxxxxxx>, "Lingle, Robert L (Robert)" <rlingle@xxxxxxxxxxxxx>, Jeffery Maki <jmaki@xxxxxxxxxxx>, "zhongqiwen@xxxxxxxxxx" <zhongqiwen@xxxxxxxxxx>, "petejone@xxxxxxxxx" <petejone@xxxxxxxxx>, "Mark.ives@xxxxxxxxxxx" <Mark.ives@xxxxxxxxxxx>, "McCurdy, Alan H (Alan)" <mccurdy@xxxxxxxxxxxxx>, "Kamino, John T (John)" <jkamino@xxxxxxxxxxxxx>, "Choudhury, Mabud (Mabud)" <MChoudhury@xxxxxxxxxxxxx>, Vipul Bhatt <vipul.bhatt@xxxxxxxxxxx>, "Mark Nowell (Cisco)" <mnowell@xxxxxxxxx>, "John Johnson (Broadcom)" <john.johnson@xxxxxxxxxxxx>, "Rakesh Sambaraju Ph. D" <rakesh.sambaraju@xxxxxxxxxx>, "David_Piehler@xxxxxxxx" <David_Piehler@xxxxxxxx>, Petar Pepeljugoski <petarp@xxxxxxxxxx>, Dale Murray - LightCounting <dale@xxxxxxxxxxxxxxxxx>, "Paul E. Neveaux Jr. Ph. D." <paul.neveaux@xxxxxxxx>, Jonathan David Ingham <jonathan.ingham@xxxxxxxxxxxxxxxx>, Darryl Heckle <HeckleDC@xxxxxxxxxxx>, Jonathan King <jonathan.king@xxxxxxxxxxx>, Piers Dawe <piersd@xxxxxxxxxxxx>, "Pondillo, Peter (Corning)" <PondilloPL@xxxxxxxxxxx>, Ted Sprague <tsprague@xxxxxxxxxxxx>, "David Lewis (Lumentum)" <David.Lewis@xxxxxxxxxxxx>, "abbottjs@xxxxxxxxxxx" <abbottjs@xxxxxxxxxxx>, "Tellas, Ronald" <Ronald.Tellas@xxxxxxxxxx>, "Gary Nicholl (Cisco)" <gnicholl@xxxxxxxxx>, "Steffen Koehler (Finisar)" <steffen.koehler@xxxxxxxxxxx>, "Bulent Kose (Panduit)" <Bulent.Kose@xxxxxxxxxxx>, "Brett Lane (Panduit)" <Brett.Lane@xxxxxxxxxxx>, "swansonse@xxxxxxxxxxx" <swansonse@xxxxxxxxxxx>, "Xu, Yang (Sunny)" <Sunny.Xu@xxxxxxxxxxxxx>, "Jose Castro (Panduit)" <Jose.Castro@xxxxxxxxxxx>, "Adrian Amezcua (Prysmian)" <Adrian.Amezcua@xxxxxxxxxxxxxxxxx>, Brad Booth <brbooth@xxxxxxxxxxxxx>, "mgustli@xxxxxxxxxx" <mgustli@xxxxxxxxxx>, George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>, "David.Piehler@xxxxxxxx" <David.Piehler@xxxxxxxx>, "Paul Vanderlaan (Nexans)" <paul.vanderlaan@xxxxxxxxxx>, "phongpham@xxxxxxxxxxx" <phongpham@xxxxxxxxxxx>, "Kenneth Jackson (Sumitomo)" <kjackson@xxxxxxxxxxxxxx>, "Joost.grillaert@xxxxxxxxxx" <Joost.grillaert@xxxxxxxxxx>, "Chow, Bruce (Corning)" <chowbc@xxxxxxxxxxx>, "Chris Cole (Finisar)" <chris.cole@xxxxxxxxxxx>, "Rick Pimpinella (Panduit)" <rjpi@xxxxxxxxxxx>, "james.withey@xxxxxxxxxxxxxxxxx" <james.withey@xxxxxxxxxxxxxxxxx>, "liyongyao@xxxxxxxxxx" <liyongyao@xxxxxxxxxx>, "zhang.yuanbin@xxxxxxxxxx" <zhang.yuanbin@xxxxxxxxxx>, "yang.zhiwei2@xxxxxxxxxx" <yang.zhiwei2@xxxxxxxxxx>, "fmarques@xxxxxxxxxxxxxx" <fmarques@xxxxxxxxxxxxxx>, "wu.chengbin@xxxxxxxxxx" <wu.chengbin@xxxxxxxxxx>, "zshen@xxxxxxxxxx" <zshen@xxxxxxxxxx>, "rhorner@xxxxxxxxxxxx" <rhorner@xxxxxxxxxxxx>, "vbala@xxxxxxxxxxxx" <vbala@xxxxxxxxxxxx>, "yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx" <yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx>, "shawn.nicholl@xxxxxxxxxx" <shawn.nicholl@xxxxxxxxxx>, "hyaku@xxxxxxxxxxxxx" <hyaku@xxxxxxxxxxxxx>, "john.yurtin@xxxxxxxxx" <john.yurtin@xxxxxxxxx>, "takeshin@xxxxxxx" <takeshin@xxxxxxx>, "derek.cassidy@xxxxxx" <derek.cassidy@xxxxxx>, "kpjackson@xxxxxxxx" <kpjackson@xxxxxxxx>, "scott.sommers@xxxxxxxxx" <scott.sommers@xxxxxxxxx>, "salvatore.rotolo@xxxxxx" <salvatore.rotolo@xxxxxx>, "vineets@xxxxxxxxx" <vineets@xxxxxxxxx>, "kshrikhande@xxxxxxxxxxxx" <kshrikhande@xxxxxxxxxxxx>, "ureddy@xxxxxxxxx" <ureddy@xxxxxxxxx>, "sam_sambasivan@xxxxxxxxxxxx" <sam_sambasivan@xxxxxxxxxxxx>, "greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx" <greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx>, "tommitcheltree@xxxxxxxxxxx" <tommitcheltree@xxxxxxxxxxx>, Indris <Iseman85@xxxxxxxxx>, Rick Pimpinella <Rick.Pimpinella@xxxxxxxxxxx>, "Samamra Raed, US" <raed.samamra@xxxxxxxxxxxxxxxxx>, "Harris Dave, US" <Dave.Harris@xxxxxxxxxxxxxxxxx>, Kobi Hasharoni <kobi.hasharoni@xxxxxxxxxxxxxxxxx>, Ramana Murty <ramana.murty@xxxxxxxxxxxx>, "fymchang@xxxxxxxxx" <fymchang@xxxxxxxxx>, "Carlo Tosetti (ctosetti)" <ctosetti@xxxxxxxxx>, "Carlo Mariotti (cmariott)" <cmariott@xxxxxxxxx>
Date: 09/28/2018 12:38 PM
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

To John’s point. The reason I was asking this question was that my understanding is that there is a large data base that contains lots of different VCSEL launches, fibers, connector offsets etc. with Modal noise calculated but that that analysis had not constrained the connector offsets. This data base was used as at least one input to determining what the modal noise allowance should be in the development of the 10GBASE-SR specification. For that specification the modal noise without the connector offset constraint was sufficiently small (0.15dB) that using this conservative approach wasn’t an issue. Pier’s presentation leveraged the 10GBASE-SR specification and found inconveniently high modal noise predictions. I’m suggesting that we should re-visit the analysis as it is possible/probable that the data base includes situations where the connector offsets are so large that the connector would have failed the insertion loss spec due to connector offset. These situations should be removed from the analysis.

To Paul’s comments. Thanks for the additional information. Would the Numerical Aperture and Core Diameter effects be an additional source of mode selective loss (causing modal noise) or is this likely to just create loss for all modes? I’m also wondering whether we could just use an additional insertion loss of 0.75dB above the loss with zero offset as an upper bound for the offset if it is difficult to obtain the fiber cables from the same spool. Do I understand from your text copied below that the connector loss measurement was changed from an overfilled launch to an LED launch that is close to meeting the max encircled flux spec in the IEEE specifications? If so the “hidden margin” for connectors having lower loss than in the optical budget that we were discussing on the call has disappeared in the worst case analysis. (It is still there statistically and there also might be something determistic if there is a correlation between VCSEL launches close to the max encircled flux (many launched modes?) and lower modal noise (Modal noise greater with fewer launched modes?)

Paul’s text. “Then, a source meeting the standard loss test Encircled Flux (which is defined to be near the maximum fill permitted of a transmitter (e.g. VCSEL) source)”

From:Kolesar, Paul <PKOLESAR@xxxxxxxxxxxxx>
Sent: Friday, September 28, 2018 5:19 AM
To: John F Petrilla <john.petrilla@xxxxxxxxxxxxxxxx>; Dudek, Mike <Mike.Dudek@xxxxxxxxxx>; Lingle, Robert L (Robert) <rlingle@xxxxxxxxxxxxx>; Jeffery Maki <jmaki@xxxxxxxxxxx>; zhongqiwen@xxxxxxxxxx; petejone@xxxxxxxxx; Mark.ives@xxxxxxxxxxx; McCurdy, Alan H (Alan) <mccurdy@xxxxxxxxxxxxx>; Kamino, John T (John) <jkamino@xxxxxxxxxxxxx>; Choudhury, Mabud (Mabud) <MChoudhury@xxxxxxxxxxxxx>; Vipul Bhatt <vipul.bhatt@xxxxxxxxxxx>; Mark Nowell (Cisco) <mnowell@xxxxxxxxx>; John Johnson (Broadcom) <john.johnson@xxxxxxxxxxxx>; Rakesh Sambaraju Ph. D <rakesh.sambaraju@xxxxxxxxxx>; David_Piehler@xxxxxxxx; Petar Pepeljugoski <petarp@xxxxxxxxxx>; Dale Murray - LightCounting <dale@xxxxxxxxxxxxxxxxx>; Paul E. Neveaux Jr. Ph. D. <paul.neveaux@xxxxxxxx>; Jonathan David Ingham <jonathan.ingham@xxxxxxxxxxxxxxxx>; Darryl Heckle <HeckleDC@xxxxxxxxxxx>; Jonathan King <jonathan.king@xxxxxxxxxxx>; Piers Dawe <piersd@xxxxxxxxxxxx>; Pondillo, Peter (Corning) <PondilloPL@xxxxxxxxxxx>; Ted Sprague <tsprague@xxxxxxxxxxxx>; David Lewis (Lumentum) <David.Lewis@xxxxxxxxxxxx>; abbottjs@xxxxxxxxxxx; Tellas, Ronald <Ronald.Tellas@xxxxxxxxxx>; Gary Nicholl (Cisco) <gnicholl@xxxxxxxxx>; Steffen Koehler (Finisar) <steffen.koehler@xxxxxxxxxxx>; Bulent Kose (Panduit) <Bulent.Kose@xxxxxxxxxxx>; Brett Lane (Panduit) <Brett.Lane@xxxxxxxxxxx>; swansonse@xxxxxxxxxxx; Xu, Yang (Sunny) <Sunny.Xu@xxxxxxxxxxxxx>; Jose Castro (Panduit) <Jose.Castro@xxxxxxxxxxx>; Adrian Amezcua (Prysmian) <Adrian.Amezcua@xxxxxxxxxxxxxxxxx>; Brad Booth <brbooth@xxxxxxxxxxxxx>; mgustli@xxxxxxxxxx; George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>; David.Piehler@xxxxxxxx; Paul Vanderlaan (Nexans) <paul.vanderlaan@xxxxxxxxxx>; phongpham@xxxxxxxxxxx; Kenneth Jackson (Sumitomo) <kjackson@xxxxxxxxxxxxxx>; Joost.grillaert@xxxxxxxxxx; Chow, Bruce (Corning) <chowbc@xxxxxxxxxxx>; Chris Cole (Finisar) <chris.cole@xxxxxxxxxxx>; Rick Pimpinella (Panduit) <rjpi@xxxxxxxxxxx>; james.withey@xxxxxxxxxxxxxxxxx; liyongyao@xxxxxxxxxx; zhang.yuanbin@xxxxxxxxxx; yang.zhiwei2@xxxxxxxxxx; fmarques@xxxxxxxxxxxxxx; wu.chengbin@xxxxxxxxxx; zshen@xxxxxxxxxx; rhorner@xxxxxxxxxxxx; vbala@xxxxxxxxxxxx; yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx; shawn.nicholl@xxxxxxxxxx; hyaku@xxxxxxxxxxxxx; john.yurtin@xxxxxxxxx; takeshin@xxxxxxx; derek.cassidy@xxxxxx; kpjackson@xxxxxxxx; scott.sommers@xxxxxxxxx; salvatore.rotolo@xxxxxx; vineets@xxxxxxxxx; kshrikhande@xxxxxxxxxxxx; ureddy@xxxxxxxxx; sam_sambasivan@xxxxxxxxxxxx; greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx; tommitcheltree@xxxxxxxxxxx; Indris <Iseman85@xxxxxxxxx>; Rick Pimpinella <Rick.Pimpinella@xxxxxxxxxxx>; Samamra Raed, US <raed.samamra@xxxxxxxxxxxxxxxxx>; Harris Dave, US <Dave.Harris@xxxxxxxxxxxxxxxxx>; Kobi Hasharoni <kobi.hasharoni@xxxxxxxxxxxxxxxxx>; Ramana Murty <ramana.murty@xxxxxxxxxxxx>; fymchang@xxxxxxxxx; Carlo Tosetti (ctosetti) <ctosetti@xxxxxxxxx>; Carlo Mariotti (cmariott) <cmariott@xxxxxxxxx>
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

External Email

Mike and John,

Lateral offset of the fiber cores, while a major cause, is not the only mechanism that creates mode selective loss at connections. The other major contributors are mismatches between the fiber core diameters and numerical apertures. So the answer to how much offset causes a certain loss can only be answered with constraints on the other contributors.

What follows are some thoughts on experimental approaches. To control the NA and CD, experimental apparatus with variable offset mechanics can be set up between fibers from the same spool, which have matched CD and NA, preferably near nominal values. Then, a source meeting the standard loss test Encircled Flux (which is defined to be near the maximum fill permitted of a transmitter (e.g. VCSEL) source) can be used to dial in the amount of offset needed to induce 0.75 dB loss. Modal noise can be measured as a function of offset above and below this “offset line in the sand” using VCSEL sources from different transmitters. Note that modal noise can’t be measured with a loss test source because they use LEDs with prescribed minimum spectral width, not VCSELs, and such LEDs don’t produce speckle.

This will allow assessment of a “worst-case” connection at the first connection point in a channel. But it won’t assess what happens downstream with additional connections. The VCSEL launch condition evolves, generally towards more filling conditions, as it passes thru connections. So it may be more meaningful to instead set up a concatenation of fibers, adding as many as needed to reach the 1.5 dB loss limit of the 802.3 standard when measured with the LED loss test set meeting the EF constraints. Then measure MN with a variety of VCSEL transmitters thru this concatenation series. This approach can be used with connections of various loss performance, attempting to find a sufficiently poor set of connections that reach 1.5 dB with just two or three. That’s one corner case we have been thinking will produce the worst MN. It would be of interest to see how that compares to a case where it takes several more connections to reach 1.5 dB. This would either confirm or refute the notion that connections near the 0.75 dB limit represent the worst case.

Regards,

Paul

From:John F Petrilla <john.petrilla@xxxxxxxxxxxxxxxx>
Sent: Thursday, September 27, 2018 9:42 PM
To: Dudek, Mike <Mike.Dudek@xxxxxxxxxx>; Lingle, Robert L (Robert) <rlingle@xxxxxxxxxxxxx>; Jeffery Maki <jmaki@xxxxxxxxxxx>; zhongqiwen@xxxxxxxxxx; petejone@xxxxxxxxx; Mark.ives@xxxxxxxxxxx; McCurdy, Alan H (Alan) <mccurdy@xxxxxxxxxxxxx>; Kamino, John T (John) <jkamino@xxxxxxxxxxxxx>; Choudhury, Mabud (Mabud) <MChoudhury@xxxxxxxxxxxxx>; Vipul Bhatt <vipul.bhatt@xxxxxxxxxxx>; Mark Nowell (Cisco) <mnowell@xxxxxxxxx>; John Johnson (Broadcom) <john.johnson@xxxxxxxxxxxx>; Rakesh Sambaraju Ph. D <rakesh.sambaraju@xxxxxxxxxx>; David_Piehler@xxxxxxxx; Petar Pepeljugoski <petarp@xxxxxxxxxx>; Dale Murray - LightCounting <dale@xxxxxxxxxxxxxxxxx>; Paul E. Neveaux Jr. Ph. D. <paul.neveaux@xxxxxxxx>; Jonathan David Ingham <jonathan.ingham@xxxxxxxxxxxxxxxx>; Darryl Heckle <HeckleDC@xxxxxxxxxxx>; Jonathan King <jonathan.king@xxxxxxxxxxx>; Piers Dawe <piersd@xxxxxxxxxxxx>; Pondillo, Peter (Corning) <PondilloPL@xxxxxxxxxxx>; Ted Sprague <tsprague@xxxxxxxxxxxx>; David Lewis (Lumentum) <David.Lewis@xxxxxxxxxxxx>; abbottjs@xxxxxxxxxxx; Tellas, Ronald <Ronald.Tellas@xxxxxxxxxx>; Gary Nicholl (Cisco) <gnicholl@xxxxxxxxx>; Steffen Koehler (Finisar) <steffen.koehler@xxxxxxxxxxx>; Bulent Kose (Panduit) <Bulent.Kose@xxxxxxxxxxx>; Brett Lane (Panduit) <Brett.Lane@xxxxxxxxxxx>; swansonse@xxxxxxxxxxx; Xu, Yang (Sunny) <Sunny.Xu@xxxxxxxxxxxxx>; Jose Castro (Panduit) <Jose.Castro@xxxxxxxxxxx>; Adrian Amezcua (Prysmian) <Adrian.Amezcua@xxxxxxxxxxxxxxxxx>; Brad Booth <brbooth@xxxxxxxxxxxxx>; mgustli@xxxxxxxxxx; Kolesar, Paul <PKOLESAR@xxxxxxxxxxxxx>; George Zimmerman <george@xxxxxxxxxxxxxxxxxxxx>; David.Piehler@xxxxxxxx; Paul Vanderlaan (Nexans) <paul.vanderlaan@xxxxxxxxxx>; phongpham@xxxxxxxxxxx; Kenneth Jackson (Sumitomo) <kjackson@xxxxxxxxxxxxxx>; Joost.grillaert@xxxxxxxxxx; Chow, Bruce (Corning) <chowbc@xxxxxxxxxxx>; Chris Cole (Finisar) <chris.cole@xxxxxxxxxxx>; Rick Pimpinella (Panduit) <rjpi@xxxxxxxxxxx>; james.withey@xxxxxxxxxxxxxxxxx; liyongyao@xxxxxxxxxx; zhang.yuanbin@xxxxxxxxxx; yang.zhiwei2@xxxxxxxxxx; fmarques@xxxxxxxxxxxxxx; wu.chengbin@xxxxxxxxxx; zshen@xxxxxxxxxx; rhorner@xxxxxxxxxxxx; vbala@xxxxxxxxxxxx; yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx; shawn.nicholl@xxxxxxxxxx; hyaku@xxxxxxxxxxxxx; john.yurtin@xxxxxxxxx; takeshin@xxxxxxx; derek.cassidy@xxxxxx; kpjackson@xxxxxxxx; scott.sommers@xxxxxxxxx; salvatore.rotolo@xxxxxx; vineets@xxxxxxxxx; kshrikhande@xxxxxxxxxxxx; ureddy@xxxxxxxxx; sam_sambasivan@xxxxxxxxxxxx; greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx; tommitcheltree@xxxxxxxxxxx; Indris <Iseman85@xxxxxxxxx>; Rick Pimpinella <Rick.Pimpinella@xxxxxxxxxxx>; Samamra Raed, US <raed.samamra@xxxxxxxxxxxxxxxxx>; Harris Dave, US <Dave.Harris@xxxxxxxxxxxxxxxxx>; Kobi Hasharoni <kobi.hasharoni@xxxxxxxxxxxxxxxxx>; Ramana Murty <ramana.murty@xxxxxxxxxxxx>; fymchang@xxxxxxxxx; Carlo Tosetti (ctosetti) <ctosetti@xxxxxxxxx>; Carlo Mariotti (cmariott) <cmariott@xxxxxxxxx>
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

Hello Mike

It seems your question may be better stated as “What is the max offset that guarantees less than 0.75 dB loss?” or “What is the min offset that can cause 0.75 dB loss?” However, is offset an attribute that is normally measured? If not, why try to establish a correlation between attributes where one of the attributes may not be well defined, measured or (when the connector loss is for an overfilled launch) relevant? Perhaps the correlation of interest is between connector loss and mode selective loss noise for a launch that meets the encircled flux criteria. One may expect that the connector loss would be dependent on the laser launch condition. TIA folks currently use 10 weight factors to cover the range of launch conditions when estimating effective modal BW. Then, there’s the issue of multiple connectors in series? In summary, I don’t expect a simple answer will be useful but worry that it might be misleading.

Regards,

John

From:Dudek, Mike [mailto:Mike.Dudek@xxxxxxxxxx]
Sent: Thursday, September 27, 2018 5:16 PM
To: Lingle, Robert L (Robert); 'John F. Petrilla (FOIT)'; 'Jeffery Maki'; 'zhongqiwen@xxxxxxxxxx'; 'petejone@xxxxxxxxx'; 'Mark.ives@xxxxxxxxxxx'; McCurdy, Alan H (Alan); Kamino, John T (John); Choudhury, Mabud (Mabud); 'Vipul Bhatt (vipul.bhatt@xxxxxxxxxxx)'; 'Mark Nowell (Cisco)'; 'John Johnson (Broadcom)'; 'Rakesh Sambaraju Ph. D (rakesh.sambaraju@xxxxxxxxxx)'; 'David_Piehler@xxxxxxxx'; 'Petar Pepeljugoski'; Dale Murray - LightCounting (dale@xxxxxxxxxxxxxxxxx); 'Paul E. Neveaux Jr. Ph. D. (paul.neveaux@xxxxxxxx)'; 'Jonathan Ingham (FOIT)'; 'Darryl Heckle'; 'Jonathan King'; 'Piers Dawe'; 'Pondillo, Peter (Corning)'; 'Ted Sprague'; 'David Lewis (Lumentum)'; 'abbottjs@xxxxxxxxxxx'; 'Tellas, Ronald'; 'Gary Nicholl (Cisco)'; 'Steffen Koehler (Finisar)'; 'Bulent Kose (Panduit)'; 'Brett Lane (Panduit)'; 'swansonse@xxxxxxxxxxx'; 'Xu, Yang (Sunny)'; 'Jose Castro (Panduit)'; 'Adrian Amezcua (Prysmian)'; Brad Booth (brbooth@xxxxxxxxxxxxx); 'mgustli@xxxxxxxxxx'; 'Paul Kolesar (CommScope) (pkolesar@xxxxxxxxxxxxx)'; 'George Zimmerman'; 'David.Piehler@xxxxxxxx'; 'Paul Vanderlaan (Nexans)'; phongpham@xxxxxxxxxxx; 'Kenneth Jackson (Sumitomo)'; Joost.grillaert@xxxxxxxxxx; 'Chow, Bruce (Corning)'; 'Chris Cole (Finisar)'; 'Rick Pimpinella (Panduit)'; 'james.withey@xxxxxxxxxxxxxxxxx'; 'liyongyao@xxxxxxxxxx'; 'zhang.yuanbin@xxxxxxxxxx'; 'yang.zhiwei2@xxxxxxxxxx'; 'fmarques@xxxxxxxxxxxxxx'; 'wu.chengbin@xxxxxxxxxx'; 'zshen@xxxxxxxxxx'; 'rhorner@xxxxxxxxxxxx'; 'vbala@xxxxxxxxxxxx'; 'yasuaki.kawatsu.sv@xxxxxxxxxxxxxxxxxxxx'; 'shawn.nicholl@xxxxxxxxxx'; 'hyaku@xxxxxxxxxxxxx'; 'john.yurtin@xxxxxxxxx'; 'takeshin@xxxxxxx'; 'derek.cassidy@xxxxxx'; 'kpjackson@xxxxxxxx'; 'scott.sommers@xxxxxxxxx'; 'salvatore.rotolo@xxxxxx'; 'vineets@xxxxxxxxx'; 'kshrikhande@xxxxxxxxxxxx'; 'ureddy@xxxxxxxxx'; 'sam_sambasivan@xxxxxxxxxxxx'; 'greg.mcsorley@xxxxxxxxxxxxxxxxxxxxxx'; 'tommitcheltree@xxxxxxxxxxx'; Indris (Iseman85@xxxxxxxxx); 'Rick Pimpinella'; 'Samamra Raed, US'; 'Harris Dave, US'; 'Kobi Hasharoni'; 'Ramana Murty'; 'fymchang@xxxxxxxxx'; Carlo Tosetti (ctosetti) (ctosetti@xxxxxxxxx); Carlo Mariotti (cmariott) (cmariott@xxxxxxxxx)
Subject: RE: 400G Ethernet over MMF Task Force Ad Hoc Call

On today’s call I was asked to send an e-mail to detail a question I had. It is related to simulations of modal noise generated at connectors. Offsets in the connector create mode selective loss which along with the VCSEL’s energy moving between modes creates the modal noise. My question was what is the offset we can get in connectors that are compliant to our specification. In 802.3cd 138.10.2.2.1 it is stated that “the loss of a single connection shall not exceed 0.75dB”. My understanding was that the connector loss used to be measured with over-filled launch but this has been somewhat changed but I don’t know the details. The question is when measuring connector loss with the over-filled launch or with the new changed method what is the possible offset while still meeting the requirement that the loss is less than 0.75dB?

To unsubscribe from the STDS-802-3-NGMMF list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-NGMMF&A=1

Re: [802.3_NGMMF] what offsets occur for connectors which meet our specifications?