9/17/04IEEE 802.3aq Ad-hoc task 4 Comparison of some Cambridge model results with simulation results of RSoft and Optium in-house simulators Gary Shaulov, RSoft Design Group Yu Sun, Optium

9/17/04 Outline The purpose of these preliminary studies was to check the results of Cambridge model for scaled/non- scaled index profiles against results by other mode solvers (both commercial and proprietary); Just two fiber models were studies – fibers #23 and #29 – before and after scaling; Optium in-house simulator results were compared with RSoft commercial multimode simulator (ModeSYS™ ) results Optium and RSoft simulation results were compared with results from Cambridge model for output pulse shapes, DMD evolution, scaling factors, and OFL bandwidth.

9/17/04 Parameters: 1.Input beam: Gaussian with 7 μm FWHM 2.Input pulse: Gaussian with 50 ps FWHM 3.Wavelength: 1300 nm 4.Fiber interested: Cambridge #23 and #29 5.Offset considered: 17, 20 and 23 μm 6.Fiber length: 300m Comparison of RSoft and Optium in-house simulators Pulse shape of output from multimode fibers were studied for new scaled Cambridge index profiles

9/17/04 Comparison of Rsoft and Optium in house simulator Fiber #23 with offset of 17 μm Optium RSoft

9/17/04 Comparison of Rsoft and Optium in house simulator Fiber #23 with offset of 23 μm Optium RSoft

9/17/04 Comparison of Rsoft and Optium in house simulator Fiber #23 with offset of 20 μm Optium RSoft

9/17/04 Comparison of Rsoft and Optium in house simulator Fiber #29 with offset of 17 μm Optium RSoft

9/17/04 Comparison of Rsoft and Optium in house simulator Fiber #29 with offset of 20 μm Optium RSoft

9/17/04 Comparison of Rsoft and Optium in house simulator Fiber #29 with offset of 23 μm RSoftOptium

9/17/04 Fiber # 29, offset 17 μmFiber # 23, offset 17 μm Comparison of pulse shapes: Optium vs Rsoft Very good agreement for both fibers Blue curves: Optium Red curves: RSoft

9/17/04 Blue curves: Optium Red curves: Cambridge Fiber #23 Offset 17 μm Fiber #29 Offset 17 μm Comparison of pulse shapes: Optium vs Cambridge Good agreement for fiber #29 but not so good for fiber #23

9/17/04 DMD study for scaled and non-scaled index profiles – comparison of Rsoft and Cambridge results Two fiber index profiles were studied - fiber 23 and 29 (according latest to Cambridge release) – first non-scaled and then scaled profiles.

9/17/04 Fiber # 23 Note: fiber #23 in previous Cambridge release for 65 fibers is fiber #18

9/17/04 Fiber # 23 – cont’d Cambridge model gives scaling factor for fiber #23 f= (r.3.0) According to J.Abbott’s presentation OFL_BW=600 MHz km, DMD mean =1.98 ns/km RSoft simulations give in case of non-scaled index DMD=3.3 ps/m, in case of scaled even higher DMD = 5.9 ps/m. The difference in results may be due also to the difference in calculation methods for DMM Calculated scaling factor = 2/3.3=0.6 Calculated OFL_BW =480 MHz km

9/17/04 Fiber # 29 Note: fiber #29 in previous Cambridge release for 65 fiber is fiber #24

9/17/04 Fiber # 29 – cont’d Cambridge model gives scaling factor for fiber #29 f= (r.3.0) According to J.Abbott’s presentation OFL_BW=660 MHz km, DMD mean=2.18 ns/km RSoft simulations give in case of non-scaled index DMD=5.2 ps/m, in case of scaled index DMD = 4.7 ps/m. The difference in results may be due also to the difference in calculation methods for DMD Calculated scaling factor f=2/5.2=0.38 Calculated OFL_BW =480 MHz km

9/17/04 * See John Abbott’s presentation at telecon 7/28/2004 Corning/Cambridge (DMD scaled) * RSoft (non-scaled) DMD results comparison

9/17/04 Preliminary Conclusions The agreement between RSoft and Optium in-house simulators is excellent. Both RSoft and Optium results are in good agreement with Cambridge results for fiber #29, however for fiber #23 there is discrepancies between Cambridge and RSoft/Optium The agreement between RSoft and Cambridge models results in DMD evolution behavior is good except points at very high values of offsets More studies needed to identify if the difference comes from difference in approaches to calculate mean DMD Cambridge model prescribes to ignore degenerate mode groups with g>20 (or g>18?). It may be not always the case for RSoft and Optium mode solvers – can that be the cause of discrepancies, and should we worry about higher order modes?