[10GMMF] AW: [10GMMF] Connector offsets in Worst case model
Jonathan,
Do you mean we should therefore use 2 connectors with each 6µ offset as the worst case model?
Jens Fiedler
-----Ursprüngliche Nachricht-----
Von: owner-stds-802-3-10gmmf@listserv.ieee.org [mailto:owner-stds-802-3-10gmmf@listserv.ieee.org] Im Auftrag von Jonathan King
Gesendet: Mittwoch, 11. August 2004 05:58
An: STDS-802-3-10GMMF@listserv.ieee.org
Betreff: Re: [10GMMF] Connector offsets in Worst case model
Following the e-mail threads on connector I think I conclude that
- Jonathan Thatchers observations: imply that connector offset is truncated beyond spec (i.e. Raleigh stats won't be realistic, so they shouldn't drive a standard link model ?)
- David Cunningham: Gigabit Ethernet specs a max 0.75dB OFL loss per connector, and 0.75dB OFL loss = 6 micron offset. Together, these would suggest that we should not consider more than 6 micron offset for any individual connector - since it would be much less likely to occur in real life and also be out of spec.
jonathan king
tel: 1 408 524 5110
e-mail: jking@bigbearnetworks.com
-----Original Message-----
From: Jonathan Thatcher [mailto:jonathan.thatcher@IEEE.ORG]
Sent: Monday, August 09, 2004 10:16 AM
To: STDS-802-3-10GMMF@listserv.ieee.org
Subject: Re: [10GMMF] Connector offsets in Worst case model
It will be important for you to decide on the importance of the precision of this data. During 10 Gig, we had a discussion about the relative error implicit in this model. It was deemed to be sufficiently close as to not require additional work.
More specifically, during 10G, I did an informal survey of two cable manufacturers to see what their process was for "connectorizing" cables. The basic process was that an operator "dialed in" the offset until it met specification. A variant of the process was that the operator only corrected when the offset was out of specification.
Both cases yield manufactured offsets which were not standard distributions, the distributions where strongly skewed towards the specified limit. If you think about it, in the second case, the distribution would have been close to a standard distribution if the number of corrections was small. In fact, the number of corrections approached 40%, meaning that without correction only 60% met specification. You get the idea.
In these informal surveys, I did not take data, as I promised to not interfere with the manufacturing flow; I simply watched the method.
If this is critical to the modeling you are doing, it is probably the case that you need a more rigorous survey and statistics.
jonathan
> -----Original Message-----
> From: owner-stds-802-3-10gmmf@listserv.ieee.org
> [mailto:owner-stds-802-3-10gmmf@listserv.ieee.org]On Behalf Of
> <Joerg-R.> <Kropp>
> Sent: Monday, August 09, 2004 2:38 AM
> To: STDS-802-3-10GMMF@listserv.ieee.org
> Subject: [10GMMF] Connector offsets in Worst case model
>
>
> Jonathan and all,
> in the conference call I stated that I will give a detailed
> explaination why we proposed this special set of offsets. We took the
> same Rayleigh distribution as it was done for the 10GbE -50µm
> multimode simulation (mean value about 3,6µm and 7µm offset with a
> probability of 5%). For the second offset we used the value which we
> get for two connectors (with the same offset) with a combined
> probability of 5%. For the third offset we used the value which we get
> for three connectors (with the same offset) with a combined
> probability of 5%. And similar for the fourth connector.
> See the short presentation encl.
> The resulting set of offsets is 7µm, 5µm, 4µm (not 4,5µm!)
> and 3,5µm (not 3,0µm!) and the overall probability (for
> exceeding this set of 4 offsets) is p=0,04%, which seems to
> us as a resonable value for a worst case scenario.
> Regards,
> Joerg
> <<Connector Scenario for Worst Case proposal 2.ppt>>
>
>
> Infineon Technologies Fiber Optics GmbH
> MOD CE
> Dr. Joerg-R. Kropp
> Senior Staff Expert
> Wernerwerkdamm 16
> D-13629 Berlin / Germany
> Phone -49-30-85400-4918
> Fax -49-30-85400-4900
> Mail joerg.kropp@infineon.com
>
>
>
>