Item Subject: dh_0197.txt 2/5/97 9:57A
From: Del Hanson/ Hewlett-Packard Co.          02/05/97

 ASCII Version of Presentation, "Link Lengths For 850 nm & 1300 nm Optical PMDs With All Link Penalties"

Presented At IEEE 802.3z Interim Meeting, 01/27-29/97, San Diego

Objectives

* Meet Target link lengths from 11/96 IEEE 802.3z Plenary Meeting with Target specification parameters
* Develop a unified MMF interface specification methodology for both 850 nm and 1300 nm links
* Characterize worst case link length limits based on defined interface parameters
* Assess transceiver Target parameter modifications, if needed, to meet Target MMF link lengths

Link Optical Power Budget Definition

The link optical power budget defines the difference (in dB) between the:
* Transmitter minimum average power level
* Receiver maximum average sensitivity level
  - At the prescribed bit error ratio (BER) and eye window
    opening to support the clock recovery jitter
  - Using a short test fiber between the Tx and Rx
  - With TEST laser Tr and Tf not affecting sensitivity
  - Using a TEST laser whose Extinction Ratio (ER) is worst case
    (WC), i.e., 9 dB, and Relative Intensity Noise (RIN) is at
    least 5 dB better than WC, e.g., -127 dB/Hz

The Reference 1.063 GBd Fibre Channel SWL Link Specification

* Specifies a 6 dB optical power budget (OPB)
* Allocates 2 dB for all link optical power penalties, except for
  attenuation and in-line connectors
* Allocates 4 dB to the cable plant
  - Link attenuation
  - In-line connectors

Current GbE 850 nm & 1300 nm LD Link Length Projections

* GbE Link Length Limits
  - Assume all optical penalties are supported by the 7 dB OPB
  - Are defined by equating the WC fiber exit response time to
    the signal period
  - Do not explicitly account for details of mode partition 
    noise & other link penalties
* The following spread sheet analysis
  - Includes all link noise & power penalties
  - Allocates 2 dB for in-line connectors

Laser Link Power Penalties

* Non Link-Length Dependent                                               Penalty
  - RIN @ -121 dB/Hz @ Br=1000 MHz & BER=10^-12     =0.4 dB
  - Modal Noise, MN,                                                       = 0.5 to 1.5 dB
* In-Line connectors                                                                 2.0 dB
* Link-Length-Dependent Penalties
  - Fiber attenuation (see plots)
  - Inter-symbol Interference (see plots)
  - Mode Partition Noise (see plots)

WC Analysis Projections Using Target Parameters (Shown In Plots)

* 770 nm LD With 62MMF
  - Target Link Length = 300 m
  - Link Analysis (from plot) = 235 m
* 770 nm LD With 50MMF
  - Target Link Length = 550 m
  - Link Analysis (from plot) = 452 m
* 1270 nm LD With 62MMF
  - Target Link Length = 550 m
  - Link Analysis (from plot) = 677 m

Conclusions

* Using Target parameters, including all WC LD link OPB penalties, significantly lowers the projected WC link length below the Targeted lengths for 850 nm lasers
* Alternatives should be investigated to reduce the differences between Targeted link lengths & WC link length analysis limits

Some Alternatives For Reconciling Analysis & Target Link Lengths

* Follow FC SWL Link Specification Procedure
  - Define a 2 dB maximum for all link penalties
  - Don't separate out individual link penalties
* Adjust Link Interface Parameters
  - Increase link OPB or reduce connector loss
  - Modify ER, RIN, or C1 coefficient
* Assume all penalties are not WC Simultaneously
  - MN is worse for short links (a 2 dB connector loss could
    reduce MN penalty based on 3 dB connector loss)
  - MPN, attenuation & ISI are worse for long lengths
    the allowance)

* Increase OPB From 7 dB to 8 dB (see previous plots)

* 770 nm LD With 62MMF
  - Target Link Length = 300 m
  - Link length increases from 235 m to 268 m
* 770 nm LD With 50MMF
  - Target Link Length = 550 m
  - Link length increases from 452 m to 530 m
* 1270 nm LD With 62MMF
  - Target Link Length = 550 m
  - Link length increases from 677 m to 780 m

Plots Of ER & RIN Penalties Are Shown

Tentative Conclusions From Modifying C1 Coefficient

* 770-860 nm Transceiver, 7 dB OPB
  - Shifting C1 from 480 to 350
    + Increases 62MMF link length from 235 m to ~300 m
    + Increases 50MMF link length from 452 m to ~550 m
      (With shifting of some 50MMF parameters)
  - Issue: Can C1 of 350 (vs. 480 for gaussian) be justified?
* 1270 to 1355 nm Transceiver, 7 dB OPB
  - Maintaining C1= 480
  - Exceeds 62MMF 550 m Target; Link Length = 677 m