| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
The HSSG has adopted 100m over OM3 MMF reach objectives for
both 40GE and 100GE. These are anticipated to be supported with 4x10G and
10x10G parallel fiber approaches, and have been subject of a number of HSSG
presentations, most recently for example ghiasi_01_1107 and cole_02_1107. To date, these presentations have focused on the single channel
aspects of the Nx10G MMF parallel approaches. This is not surprising since this
aspect is well understood within the industry, and has been extensively studied
during standards activities such as 10GBASE-SR or SFP+ MSA. Excellent tools
exist, such as the IEEE 10GE modeling spread-sheet that given an agreement on
assumptions, generate a link budget. What has been missing is a similarly comprehensive treatment
of the crosstalk effects. During question and answer exchanges, several HSSG speakers
commented that their experience with higher speed multi-channel applications
showed that cross-talk rather then channel loss/distortion is the dominant
factor. So while the debate about the single channel 100m OM3 link budget has
been entertaining, it is not where the challenge lies in moving forward on
developing a standard for the 100m reach objectives. A number of individuals asked me questions after the
presentations as to what cross-talk data I was requesting in my presentation that
will be required for the 40GE and 100GE MMF link budgets. Since we have
extensive experience in the IEEE in developing standards for multi-lane copper
based links such as for backplane or cable, we should look to these successful efforts
to what is required to properly characterize the electrical portion of the MMF
links. A common characteristic of these efforts has been the development of comprehensive
multi-channel models and agreement on methodology. The following areas will need to be quantified for
cross-talk effects, ideally based both on analysis/simulation and measurements.
Some maybe found to be insignificant contributors, but that will be decided on the
basis of data rather then assumptions. - Host ASIC TX Output (IC and package) - Host - Host to - Module TX traces - Module Laser Driver Array (IC and package) - VCSEL Array (die and flex) - PIN Array (die and flex) - TIA/LA or TIA/AGC Array (IC and package) - Module RX traces - Module RX connector pins - Module RX PCB traces - HOST ASIC RX Input (IC and package) Contributions bringing in data on 4x10G and 10x10G cross-talk
effects for the above will be very much appreciated. Chris |