Optical Fibers and Optical Fiber Cables for Consumer Electronic Devices (1)
Denis Molin , Brian G. Risch , Marianne Bigot-Astruc , Erin Bowman , Olivier Tatat , (1) (2) (1) Laurent Gasca , Jack Rosko , Pierre Sillard (1) Prysmian Group, Marcoussis, France, denis. [email protected]
com (2) Prysmian Group, Claremont, NC, USA (3) Prysmian Group, Calais, France
Abstract In this paper, we present the realization and the characterization of high-bandwidth 80µm-core MMFs. Insertion loss, modal bandwidth and its assessment by Differential Mode Delay (DMD) measurements as well as macro-bend-loss measurements will be particularly detailed. Fiber macro-bend losses are validated in cable bending loss measurements in extreme cable bending and abuse testing.
Keywords: Multimode Fiber, Consumer Electronic Devices, Macrobending Loss, Insertion Loss, Bandwidth.
1. Introduction Transmissions of 5 to 10 Gbps over few meters delivered by the recently introduced USB 3. 0  and Thunderbolt  interfaces for consumer electronic devices are close to the limit of Coppercable-based solutions. Low-cost optical interconnects are thus logically investigated as a next-generation solution to cope with the bandwidth demand. The assembly cost of optical interconnects, critical for consumer electronic device development, is mainly driven by alignment tolerances : relaxed alignment tolerance allows the use of cost efficient automatic pick-and-place equipment. An alignment tolerance of 10-20µm is expected to provide a significant cost reduction compared to the 1-5µm offered by conventional 50µmcore multimode fibers (MMFs). MMFs with larger cores and higher numerical apertures (NAs) can offer larger tolerances and allow for higher channel capacity and/or longer reach than Copper-cablebased solutions.
fiber coupling stage. It characterizes the ability of the fiber to catch the light of the source. IL depends on the incident field generated by a source positioned and eventually magnified at the input face of the fiber. Ideally, the source is well aligned, both angularly and laterally, to minimize the IL. However, the alignment tolerances drive the assembly cost: the ability to deal with loose alignment tolerances up to 20µm offset allows the use of automatic pick-andplace equipment that are key to lowering the cost of optical of optical interconnects . To numerically investigate the insertion loss of 80µm MMFs under various excitations, we have calculated the complete set of linearly polarized (LP) guided-modes by solving the scalar wave equation under the weakly guiding assumption. The set of guided modes sustained by the graded-index core defines an orthogonal basis upon which any guided wave can be univocally expanded onto. The light coupling into the fiber consists thus of a spatial filtering of the incident field that filters out the non-guided part, from which we can calculate the associated power loss, called insertion loss (IL). We have simulated the coupling of hundreds of VCSEL-based sources, yet widely used in high-speed optical data networks for a decade. The variety of launching conditions is directly inspired from , except that the lateral offset follows here a uniform distribution between 0 and 20µm. The launching conditions are practically characterized by the normalized Encircled Flux (EF) that is the circularly averaged and radially integrated output near field. The obtained EF coordinates, that are the 86% EF radius in µm and the EF at 4. 5µm in %, are reported in Figure 1 both for 50µm, 62. 5µm and 80µm MMFs. We note at first that the 86% EF radius does not exceed 30µm for the 80µm MMF.
Such MMFs have not been extensively investigated, especially when used with restricted launches that are typical of high-speed, VCSEL-based sources, and are certainly not all standardized. In this paper, we will present the realization and the characterization of laser-optimized, large-core (>50µm), high-NA (>0. 22) gradedindex MMFs designed for consumer electronic applications as well as the inclusion of these fiber types in interconnect cables with and without electrical power and ground wires. We will show that such geometrical features reduce the insertion losses compared to those obtained with conventional 50µm-core MMFs. We will also demonstrate how high bandwidths, that far exceed the needs for error-free transmissions at 20Gbps over tens of meters, can be obtained. Bending behaviors are critical in harsh consumer environments, we will thus compare the macro-bend losses of such MMFs to those of 50µm-core bend-insensitive MMFs and illustrate the realization in cable performance in multiple designs with and without electrical power and ground wires.
2. Insertion loss The Insertion Loss (IL) exhibited by a MMF, also called coupling loss, is the optical power loss that occurs at the source-to-
b) c) Figure 1 – Encircled flux coordinates calculated under hundreds of launches for 50µm MMFs (a), 62. 5µm MMFs (b) and 80µm MMF (c) The corresponding ILs have then been calculated. The IL distribution and cumulative distribution are shown in Figure 2. The benefit of the larger core and numerical aperture of large core MMFs on insertion loss over traditional 50µm MMFs is clear: while more than 50% (resp. 20%) of the launches lead to more than 0. 1dB (resp. 1dB) loss with 50µm MMFs, these risks fall below 10% (resp. 1%) for 62. 5µm and 80µm MMFs. Insertion losses of 80µm MMFs are typically reduced by a 20-fold compared to that of traditional 50µm MMFs.
c) Figure 2 – Calculated insertion loss distributions obtained under hundreds of launches for a 50µm MMF (a), a 62. 5µm MMF (b) and a 80µm MMF (c); lateral offset follows a uniform distribution over the 0-20µm range.
3. Bandwidth The bandwidth is the most important key characteristic of multimode fiber dedicated to data communications because it quantifies the capabilities of the fiber to support high bit rate transmissions. The bandwidth of a MMF is basically limited by the chromatic and the modal dispersion. The first mainly depends on the material dispersion, while the second mainly depends on the refractive index profile. In consumer electronic devices or applications, the link lengths hardly exceed 10m or 30m. Therefore, the chromatic dispersion is not a real issue as soon as the spectral width of the sources is kept reasonably narrow. Therefore, only the modal dispersion, meaning that all the guided modes do not propagate at the same group velocity, may become limiting. In order to reduce its impact, fiber manufacturers control the refractive index profile deposition through advanced process tuning and accurate characterizations. The most well-known technique to minimize the modal dispersion of a MMF consists of depositing parabolic graded-index core. In practice, refractive index profiles exhibit deviations from the optimal shape, so that all modes generally do not travel at the same speed. Therefore, the modal bandwidth depends on the launching conditions. As a consequence, a MMF does not exhibit a unique modal bandwidth but a plurality of modal bandwidths that does not only depend on the source type (LEDs or VCSELs) but also on the source itself. While LEDs excite all the guided modes approximately equally, high-speed VCSELs excite only a restricted number of modes. The OverFilled Launch (OFL) modal bandwidth generally used for characterizing fiber dedicated to LEDs applications does not assess the capacity of fibers for consumer devices or applications typically used in restricted launch. Generally, the term of Effective Modal Bandwidth (EMB) is used to characterize the modal bandwidth of the fiber under a given VCSEL excitation. Because the bandwidth depends on the way the light is coupled into the MMF and because there is a large variety of an emission pattern provided by high-speed sources , a Differential Mode Delay (DMD) measurement has been standardized to accurately characterize the modal dispersion of the MMFs . It consists of recording pulse responses of the MMF for single-mode launches that radially scan the core. It provides a DMD plot that is then post-processed in order to assess the capability of the MMF by quantifying its minimum or "worst case" EMB.
GI-MMFs respectively. For that purpose, we have calculated the mode power distributions (MPDs) induced by each offset launches during a DMD measurements and the MPDs induced by a plurality of VCSEL launches using the same model that for the insertion loss study of paragraph 2. Then for each VCSEL launch, we have fitted the weights to apply to the first MPDs set to reconstruct the MPD induced by the VCSEL. Figure 3 and Figure 4 report normalized DMD plots of respectively 80µm and 62. 5µm GI-MMFs made by the versatile Plasma Chemical Vapor Deposition (PCVD), and the DMD obtained by modeling of the targeted refractive index profile that follows an alpha shaped profile.
nr nco 1 2 r r 0 nr ncl
, r r0
, r r0
nco2 ncl2 2nco2
r0 is the core radius, α the power law parameter (typically around 2. ), and nco and ncl are the refractive indices of the core and the cladding respectively. The 62. 5µm MMF has been assisted by a trench for macrobending loss purpose (cf. paragraph 4). Both in modeling and measurements, the DMD is performed on a 750m long sample for the 80µm MMF and 550m for the 62. 5µm MMF with a pulse duration of ~85ps FWQM. The sample length has been carefully chosen to be long enough to separate temporally the different mode groups, and short enough to not attenuate too much the largest offset launches. The minimum EMB calculated from the measurements is about 1,8GHz-km at 850nm, and about 1. 2GHz-km for the simulation for the 62. 5µm MMF. The minimum EMB calculated from the measurements is about 1,7GHz-km at 850nm, and about 3GHz-km for the simulation for the 80µm MMF.
Each offset launch excites a restricted number of modes or mode groups: basically, centered launches excite mainly the lowest order mode groups, while large-offset launches excite the highest order mode groups . One common way to assess the minimum EMB of a MMF from its DMD measurements consists of calculating the bandwidths of a plurality of fiber responses generated as combinations of the different fiber responses recorded in the DMD plot . The minimum EMB is assessed by the minimum calculated bandwidth. The combinations are actually weighted sum of the DMD traces. The weights, also called weight functions since they are function of the offsets, correspond to different launching conditions and/or transmitters. A set of ten weight functions is available for 50µm MMFs within Standards but not for 62. 5µm or 80µm MMFs. We have thus calculated two sets of 150 weight functions dedicated to the minimum EMB assessment of 62. 5µm and 80µm
We note the fair model-measurement agreement; for the 80µm MMF, the DMD presents a bending at the largest offsets that is inherent to cores following equation (1). Based on those results, a second fiber has been realized to correct this effect. Its measured and expected DMD plots are reported in Figure 5. A minimum EMB of about 2. 2 GHz-km has been achieved. According to modeling, minimum EMB as large as 9GHz-km are feasible. The same correction can be applied to 62. 5µm MMF as well to achieve larger EMB.
b) Figure 3 – Measured (a) and simulated (b) DMD plots of 80µm MMFs exhibiting an alpha shaped graded-index core.
b) Figure 5 - Measured (a) and simulated (b) DMD plots of 80µm MMFs exhibiting an optimized graded-index core.
4. Macrobending loss
b) Figure 4 – Measured (a) and simulated (b) DMD plots of 62. 5µm MMFs exhibiting an alpha shaped graded-index core.
The macro-bending sensitivity of MMF is an important aspect for consumer electronic devices and applications since the fiber is likely to be subjected to sharp bends and other physical stresses in consumer environments. Basically, each guided mode exhibits its own macro-bending sensitivity, and bending losses are significantly different between modes. As a consequence, bends filter the most bend sensitive modes. Therefore, only the first and the tightest bends in a link lead to signal attenuation, the following and or larger bends do not induce losses: in other
words, bend losses do not linearly increase with the number of bends in MMF . Figure 6 reports the calculated bending losses of the guided modes belonging to the 15 last mode groups in an 80µm MMF for a bending radius of 5mm at 850nm. The case of a 50µm MMF has been added for comparison. The modes of the first thirty-four mode groups of the 80µm MMF exhibit bend losses below 1dB/m at 850nm for a 5mm bend radius. Actually, thanks to their larger numerical aperture, the number of modes that exhibit bend losses lower than 1dB/m is proportionally larger in a 80µm MMF. Therefore, an 80µm MMF is expected to be more bend resistant, meaning it would exhibit lower bend losses, than a conventional 50µm MMF. A 62. 5µm MMF would be in-between because of its intermediate numerical aperture and core size; therefore, it is worth using the trench assistance concept to 62. 5µm MMF to eventually improve the macro bending loss behavior at small bending radii.
b Figure 6 – Modal bend losses with respect to the principal mode number at 5mm bend radius at 850nm for a 50µm MMF (a) and an 80µm MMF (b)
Because of the large variability of bend sensitivity of the modes within a GI-MMF, the observed macro bending loss strongly depends on the launching conditions. Launching conditions that only excite the lowest order modes lead to extremely low bending losses. For 50µm and also 62. 5µm conventional GI-MMFs, the EFs of the respective launching conditions for macro bending loss characterization are standardized in IEC 61280-4-1. They defined worst-case launches that reasonably assess the maximum macro bending loss a fiber would exhibit in the field. Knowing that worst case launches are those with the largest 86% EF radius , we have defined a reasonable worst case launch characterized by an 86% EF radius of 28µm and a EF of 2% at 4. 5µm according to the EF distribution reported in Figure 1. In practice, we realized the worst-case launch using a fiber with a core diameter of 23.5 mm, and a numerical aperture of 0.208 under overfilled launch, mechanically spliced to the fiber under test, with an adequate lateral offset so that the EF coordinates match our requirements. The macro bending loss measured on the 80µm MMF after two turns at different bending radius down to 3mm are reported in the left plot of Figure 7. The macro bending loss of conventional and bend-resistant 50µm and 62. 5µm MMFs, measured according to the Standards IEC 60793-1-47, are also reported. The results show that the macrobending losses exhibited by an 80µm MMF are equivalent to those of bend-resistant 50µm MMFs. While for 50µm bend resistant MMF, the bend resistance is brought by the trench-assistance concept , the bend resistance of the 80µm MMF is brought by its large numerical aperture. Finally, the 62. 5µm MMF takes benefit from the trench-assistance concept since the trench assistance leads to a 10-fold macrobending loss improvement at 3mm for 2 turns. The macrobending losses for a bending radius of 3mm with respect to the number of turns are reported in the right plot of Figure 7 for 80µm BI-MMF only. As expected, the macrobending losses do not vary linearly with the number of turns: if the first turn at 3mm leads to 0. 5dB loss, the macrobending losses do not exceed 1. 5dB after 5 turns.
office chair roller, and a cable pinch test at the cable self-limiting radius.
b) Figure 8– Cable designs used to test cabled fiber macrobending performance: (a) 4 x 80µm MM fibers and two electrical conductors. (b) 4 x 80µm MM fibers and no electrical conductors.
b) Figure 7– Macrobending loss after 2 turns in dB with respect to the bend radius in mm (a) for a standard 50µm MMF (dark diamonds), a bend insensitive 50µm MMF (light diamonds) and a 80µm MMF (circles); macrobending loss at 3mm bending radius with respect to the number of turns for a 80µm MMF (b)
5. Cable Performance To verify that the fiber macrobending performance will be realized in actual consumer electronic interconnect cables, cables containing 80µm MMF with and without electrical conductors were manufactured and tested to various abuse situations which may be encountered in office environments. Figure 8A illustrates the cable design including 4 x 80µm MM fibers and two electrical conductors and Figure 8B illustrates the cable design including 4 x 80µm MM fibers and no electrical conductors. To assess suitability of these cables for abuse in office environments abuse simulation testing was completed on these optical fiber cables. The cables were subjected to a simulated crushing event simulating having the cable stepped on by a high heel, a crushing event simulating having the cable crushed by an
Cable crushing tests were performed using an Instron model 4468 mechanical tester. The high heel simulation test involved driving a 9mm diameter steel pin into the cable at a load of 150lbf. (667N) and holding the load for 10 minutes. The cable attenuation was measured at 850nm and 1310nm during the testing and 5 minutes after the load was released. An image of the high heel simulation test being performed on a cable is shown in Figure 9a. Chair roller crushing simulations were performed on rollers with a load rating of 80lb (356N) and 100lb (445N). In the chair roller simulations, the roller was placed over the cable and brought to the maximum rated roller rating and held for 10 minutes. The cable attenuation was measured at 850nm and 1310nm during the testing and 5 minutes after the load was released. Images of chair roller simulation tests being performed on a cable is shown in Figure 9b &c. Table 1 summarized the test results attained during testing of the cable with 4x 80micron bend insensitive fibers and no electrical conductors. In the testing the only situation that created any significant cabled fiber attenuation was the high heel simulation test. The attenuation was almost completely reversible after the load was removed. Cable pinch testing was performed at the cable self-limiting bending radius. Cable designs 1 and 2 were the cables with electrical conductors and cable design 3 was the cable design without electrical conductors. Cables with standard 80 micron fibers and with bend insensitive 80 micron fibers were used. Cables were placed in a fixture where the cable was brought to a pinched configuration between parallel plates fixed at displacements of 9.6mm and 7.6mm. These distances correspond to the self-limiting bend radius of the two cable designs studied.
An image of the optical cable without electrical conductors being tested with a plate displacement of 7.6mm is shown in figure 10. Tabulated cable attenuation performance during and after pinch testing is summarized in Table 2. From the data in table 2 it is evident that the bend insensitive fibers have substantially improved performance in the pinch testing. Additionally the cables without the electrical conductors performed better in pinch testing at a common bend radius and could also be brought to a lower minimum bend radius due to their smaller self-limiting bend radius. The improvement in the performance of the cables without conductors is probably due to the lack of the rigid conductors within the core which can create additional stresses on the fibers during testing. Test data indicates that cable functionality can be obtained at both 850 and 1310nm even under extreme pinching conditions. Figure 10– Cable Pinch Testing: Table 2. Cable Attenuation Summary at 850nm and 1310nm for Cable Pinch Testing.
Fiber Type Cable O.D (mm): Wall Thickness (mm):
80μm NonTrench Assisted
80μm Trench Assisted
80μm NonTrench Assisted
0.16 Below selflimiting radius
0.13 Below selflimiting radius
Below selflimiting radius
Below selflimiting radius
10mm Kink Attenuatoin @ 850nm (dB)
Figure 9– Cable crush testing: High heel crush simulation. (b) 80lb chair roller crush simulation. (c) 100lb chair roller crush simulation. Table 1. Cable Attenuation Summary at 850nm and 1310nm for Cable Crush Testing 4 x 80micron Bend Insensitive Optical Fibers. Test Load
150 lb. (667N) High Heel 80 lb. (356N) Roller 100 lb. (445N) Roller
During Test Relaxed 5 minutes
10mm Kink Attenuatoin @ 850nm (dB), Relaxed 8mm Kink Attenuation @ 850nm (dB) 8mm Kink Attenuation @ 850nm (dB), Rleaxed 10mm Kink Attenuatoin @ 1350nm (dB) 10mm Kink Attenuatoin @ 1350nm (dB), Relaxed 8mm Kink Attenuation @ 1350nm (dB) 8mm Kink Attenuation @ 1350nm (dB), Relaxed
0.00 Below selflimiting radius
0.00 Below selflimiting radius
Below selflimiting radius
Below selflimiting radius
6. Conclusions In conclusion, 62.5µm and 80µm MMF are attractive alternatives to copper-based solutions for consumer electronic devices. First, the large core diameter and numerical aperture reduce the insertion losses by a twenty fold compared to conventional high bandwidth graded-index multimode fiber allowing the use of automatic pickand-place equipment that are key to lowering the assembly cost of optical of optical interconnects. Second, the large bandwidth
provided by graded-index core design far exceeds the needs for 20Gbps and even more error-free transmissions over tens of meters. The feasibility of fiber bandwidth exceeding 2GHz-km with such large cores has been demonstrated. Finally, thanks to their large numerical aperture, 80µm MMF and trench-assisted 62. 5µm MMF exhibit impressive bend resistance, comparable to recently introduced to the market bend insensitive 50µm MMFs, that makes it adapted to harsh consumer environment. The suitability of bend performance of 80µm MMF for crushing and pinching forces that may be experienced in office environments has been demonstrated in cable testing.
7. References  http://www. usb. org/developers/docs/
the interaction between VCSEL sources and multi-mode fibers. He has authored and co-authored numbers of papers and holds more than 17 patents.
Marianne Bigot-Astruc received the Ph. D. degree in Materials Science and Engineering from Mines ParisTech school in 2001. The same year she joined Alcatel in the Fiber Optic R&D Unit, Conflans Sainte-Honorine (France) and is now a R&D Scientist within Prysmian Group in the R&D Fiber Product group based in Marcoussis, France.
 http://www. intel. com/technology/io/thunderbolt/index. htm  E. Palen, "Low cost optical interconnects," Proc. of SPIE Vol. 6478 647804-1 (2007)  P. Pepeljugoski, et al. , “Development of System Specification for Laser-Optimized 50-µm Multimode Fiber for Multigigabit Short-Wavelength LANs,” Journal of Light. Tech. , Vol. 21, No. 5, May 2003  C. Caspar et al. , “Impact of transceiver characteristics on the performance of 10 GbE links applying OM-4 multimode fibers,” Proc. 57th IWCS, Providence, RI, USA, p. 293 (2008)  Differential Mode Delay Measurement of Multimode Fiber in the Time Domain, TIA Fiber Optic Test Procedure (FOTP) 220, January 1, 2003  R. E. Freund et al. , "High-Speed Transmission in Multimode Fibers," J. of Lightwave Technol. 28, 569-586 (2010).  D. Molin, et al. , “Multimode Fiber Bending Sensitivity for Multi-Gigabit Ethernet Application,” Proc. 58th International Wire & Cable Symposium, Charlotte, NC, USA, p. 442 (2009)  D. Molin, L. -A. de Montmorillon, P. Sillard, “Low Bending Sensitivity of Regular OM3/OM4 Fibers in 10GbE Applications,” Proc. OFC, San Diego, CA, USA, paper JThA55 (2010)  D. Molin, M. Bigot-Astruc, K. de Jongh, P. Sillard, “TrenchAssisted Bend-Resistant OM4 Multi-Mode Fibers,” Proc. ECOC, Torino, Italy, paper P1. 12, p. 1061(2010)
8. Pictures of Authors Denis Molin received the engineering diploma of the Ecole Supérieure d’Optique, now Institut d’Optique Graduate School, in 2000. He joined the System Design Team in Corvis-Algety the same year and the Fiber Optic R&D Unit in Alcatel in 2001 where he worked on modeling long-haul WDM transmissions. He’s now with Prysmian Group, Marcoussis, France, working on single-mode & multi-mode fibers and systems modeling for communications and data applications. His activities particularly include the design of new multi-mode fibers for high-speed data applications with improved bending and bandwidth performances, and the support of new developments in International Standardization bodies. He also works on multigigabit Ethernet system modeling with a particular attention paid to
Pierre Sillard received the engineering diploma of Telecom ParisTech ENST, in 1994 and the Ph. D. degree from the University of Paris VI in 1998, in collaboration with the Laboratoire Central de Recherches of Thomson-CSF (now Thales Research & Technology). He has been working in the field of optical fibers since 1999, and he's now with Prysmian Group in Marcoussis, France. His activities include modeling and characterizing optical fibers and systems for communications, data and specialty applications. He has authored and co-authored more than 110 papers and holds more than 40 patents. He is a member of the OSA and IEEE communications societies. Brian G. Risch is the Materials Technology Manager at the Prysmian fiber optic cable plant in Claremont, NC. He holds a B. A. degree in physics from Carleton College and a Ph. D. in Materials Science and Engineering from Virginia Polytechnic Institute and State University. Brian has studied structure property relationships in polymeric materials and materials reliability for the last 20 years. Prior to working for Draka, Brian worked for Alcatel for 7 years at Alcatel’s Optical Fiber Cable R&D center specializing in cable materials and fundamental material reliability and then for 6 years at Hewlett Packard as a failure analysis engineer. Erin Bowman is the Senior Materials Engineer at the fiber-optic cable plant in Claremont, NC. She holds a B. S. in Chemistry and a B. S. in Applied Mathematics from North Carolina State University. She has more than 10 years experience in the areas of analytical chemistry and polymer characterization in the fiber-optic, aerospace, and specialty textiles industries. Before working for the Prysmian Group Erin worked on Lockheed Martin Space Shuttle Materials for NASA.