This paper presents a high bend tolerant multimode optical fiber transmission system that is compatible with standard 50 µm graded index multimode fiber, in terms of achievable bandwidth and interconnectivity losses. When the 10 loops of the proposed bend resistive multimode fiber were wrapped around a cylinder of 1.5 mm radius, bend losses below -0.2 dB were achieved in case of experimentally produced fiber. Furthermore, when the section of the proposed bend resistive fiber was inserted between two sections of a standard 50 µm graded index multimode fiber, the total experimental measured loss proved to be below -0.15 dB.
©2009 Optical Society of America
Bend loss has been one of the major concerns when manufacturing fiber, cable and photonic devices. There are increasing numbers of applications where optical fibers need to be routed through constrained spaces or where fibers or fiber cables need to be matched, packaged or mounted tightly to structures having arbitrary shapes and forms. In such environments, a significant number of tight bends can be expected along the fiber’s path. Fibers with high bend tolerance are therefore required in such environments. Typical examples are house, building or various vehicles wiring systems. The packaging size, and thereby cost of many practical photonic devices, is often limited by allowable fiber bend diameters.
To date, significant works relating to the understanding and improvement of bend-loss sensitivity have been carried on for single-mode fibers and fiber systems [1–12] and multimode fibers [13–18]. Achieving high bend tolerance in multimode fibers without compromising other properties such as bandwidth, and other compatibility with existing standard telecommunication fibers has however, proved to be a challenging problem.
There are several serious limitations that restrict design opportunities for bend resistive multimode fibers. Since multimode fibers support a large number of modes and each mode bears individual bend-loss characteristics, it is difficult to control bend-loss and other waveguide properties of all propagating modes simultaneously. This becomes a particularly challenging problem in high bandwidth transmission multimode fibers, where even minor intervention into the optimum shape of the graded index profile inevitably leads to serious degradation of transmission fiber bandwidth. Finally, bend resistive multimode fibers should also exhibit good compatibility with existing standard multimode fibers and terminal equipment, to allow for effective and economical interconnectivity.
This paper presents an optical fiber transmission system that exhibits very high tolerance to bend induced losses, while achieving low-loss connectivity with standard 50 µm multimode fiber. Modeling also shows that the proposed system can achieve and even exceed the bandwidth of the standard 50 µm multimode fiber.
2. The design of bend resistive multimode fiber transmission system
2.1 Identifying major parameters that affect bend-loss characteristics in multimode fibers
When an optical fiber is exposed to bending, two types of bend related losses can be observed that possess distinctive sets of characteristics and should be treated separately when designing bend resistive multimode fibers. The first is the microbend-loss, and occurs as a result of coupling among guided and cladding modes. It is caused by small, usually random, fiber core displacements along the fiber axis. The microbend loss in randomly perturbed fiber is proportional to the coupled power coefficients between guided and cladding modes . The coupled power coefficients depend on several parameters however, as shown in , the difference in propagation constants between observed modes (Δβ) has a profound impact on coupling strength. The coupled power coefficient among two modes is a highly non-linear function of Δβ, proportional to 1/Δβ4 . Coupling, therefore, predominately occurs between modes (or modal groups) that are close (e.g. neighbors) in β space  and the optical power in the multimode fiber due to microbending is lost through coupling between the highest order guided modes and the cladding modes. The strong dependence of the mode coupling on Δβ was also demonstrated experimentally in several references [22, 23]. For example, in reference  examples of nearly parabolic fibers were produced that had different values of Δβ between modal groups. The experiments clearly showed strong increase in the mode coupling coefficients (microbend loss) when Δβ was low.
The modes in nearly parabolic multimode graded index fibers (GI-MMF) are grouped in modal groups (a modal group consists of degenerate modes). These modal groups are well spaced within the phase constant space (β space), in standard telecommunication fibers. For example, the modal group spacing corresponds to the spatial mode beat length period between 1 to 1.2 mm in both 50 µm and 62.5 µm standard fibers (at all wavelengths of interest) that corresponds to Δβ≈5700 m-1 (at 1310 nm). This is considerably larger spacing than encountered, for example, in most practical polarization maintaining fibers (beat lengths of a few mm), but still less than in standard singlemode fibers (the beat length between fundamental mode and closest cladding modes is roughly 0.62 mm or Δβ≈10000 m-1at 1310 nm). This relatively large spacing of modal groups in the phase constant space encountered in standard multimode fibers, prevents significant mode coupling losses when the fiber is exposed to shorter sharply curved sections such as, for example, those encountered during routing of the fiber through constrain spaces with multiple 90 degree bends. This assumption was also confirmed, for example, in , where experimental evidence shows that bends with a few millimeters radii do not cause mode coupling and, consequently, microbend loss. The given separation of modes in β space in standard MMF is also sufficient to prevent any significant mode coupling in practical multimode telecommunication optical link lengths and, thus, allows for the operation of multimode fibers under restricted mode launch conditions that are routinely used today to enhance MMF link performance [25,26].
Microbend loss in standard multimode fibers, therefore, does not present usually a significant concern unless the fiber core displacement amplitude is large over longer spans of fiber (for example in squeezed or improperly designed cable).
The second and dominant loss mechanism encountered during practical bending of the fiber to small radius is caused by continuous loss of the modal wavefront guidance at the outer side of a curved waveguide (e.g. macrobend loss). This mechanism is graphically depicted in Fig. 1 and was described by many references, for example in [4,6] (both references concern singlemode fibers, however the principles describe in references are applicable to individual modes of MMFs as the macro bend loss is typical mode attribute), and can be summarized in the following way: when modal wavefront propagates through the curved waveguide, the phase velocity at the outer side of the curvature increases proportionally to the radial distance measured from the curvature center of the waveguide, in order to compensate for the longer path length associated with the increasing local radius. Since the average phase velocity of mode wavefront for curved and straight fibers is about the same, the local value of the mode phase constant βc(r) in the curved fiber can be approximated as (1):
Where r is the radial distance measured from the curvature center, R is fiber core curvature radius and β is the phase constant of the mode in a straight fiber. The increase in local phase velocity of the mode wavefront is, however, limited by the speed of light in the cladding. When local phase velocity reaches the speed of light in the cladding, the part of the wavefront beyond this point detaches itself from the original mode wavefront and propagates in a radial direction. The radial position of the dissociation caustic, where the fraction of propagating modal power starts to irreversibly detach itself from the original wave front, is thus located at radial distance rc at which the expression (1) equals the free space propagation constant in the fiber cladding (2):
where k0 is the free space propagation number and nc is the refractive index of the cladding. The loss per unit length in the curved fiber is thus determined by mode field amplitude at dissociation caustic, e.g. by mode field amplitude at radial distance rc. The rc can be expressed from (2) as (3):
where neff is effective refractive index of the observed mode. Since the mode field roll-off is approximately exponential in the cladding, the loss per unit is fast (non-linear) function of the curvature radius R and neff/nc ratio. From this simple qualitative description, one can see that neff/nc ratio plays a profound role in fiber mode bend-loss sensitivity. A straightforward way for increasing bend-loss tolerance of a fiber mode is to increase its neff/nc ratio.
The experimental support for the above conclusion can be also found for example in reference  (in Fig. 12), where standard GI MMF was selectively excited by the single mode fiber that was offset across the GI MMF fiber core. When the lowest order modes were excited (e.g. modes with high neff/nc ratio) the bend loss tolerance of tested GI MMF fiber was also high. The bend loss however quickly increased when the significant offset of the launching fiber was introduced and the modes with low neff/nc were selectively launched into the tested GI MMF.
2.2 A general approach to multimode bend resistive fiber design
An ideal multimode bend resistive fiber profile would only support the existence of modes that have high values of neff/nc ratio. In such a hypothetical case, all the modes would be well separated from the cladding level and would, therefore, exhibit high macrobend tolerance – Fig. 2(a). Unfortunately, such profiles are unknown in practice (at least not in terms of conventional index graded fibers) as the modes in multimode fibers always (more or less) evenly fill-up the entire available phase constant space that is determined by cladding and maximum core refractive indexes – Fig. 2(b).
The following approach can be applied to overcome this limitation: a multimode bend resistive transmission fiber is designed in such a way that the excitation system like, for example, standard multimode fiber or optical source selectively only excites those modes in the transmission fiber that have high effective refractive index, while the modes with lower values of propagation constants remain unexcited and thus do not participate in signal transmission - Fig. 3.
The transmission bend resistive fiber will therefore support a large set of fiber modes, for example significantly larger set of modes than encountered in exciting fiber, while only part of these modes is used for signal transmission. The basic configuration of the proposed multimode bend resistive system is shown in Fig. 4 and consists of an optical source coupled to lead-in excitation fiber that can be of any desired length (for example standard GI MMF that is bend intolerant), arbitrary length of multimode transmission fiber that is bend tolerant and can be subjected to severe bending while exhibiting low optical loss, and optical receiver.
The transmission fiber profile shall be tailored to the excitation fiber (or source) in order to achieve selective excitation of modes that have high effective refractive index values in the bend resistive transmission fiber. The transversal distribution of electric field for each individual mode of the excitation (lead-in) fiber is closely matched to the transversal electric field distribution of the mode having high value of the phase constant in the transmission fiber. Each individual mode of excitation fiber, thereby, obtains a matching pair with high value of effective refractive index in the transmission fiber. In an optimum case, the number of modes with high values of effective refractive index in transmission fiber matches the total number of modes in the lead-in (excitation) fiber.
The transversal field distributions of lead-in (excitation) fiber modes and high effective index modes of transmission fiber can be matched by implementing the same relative refractive index profile shape in both fibers within the region that has the same radial dimensions as the lead-in (excitation) fiber core (this assumption is valid for weekly guided approximation). However, to achieve high neff/nc ratio of the launched modes in the transmission fiber, the absolute index values of the transmission fiber core, cladding or both should be substantially different when compared to the excitation fiber (e.g. the absolute core index shuld be higher and/or the cladding index should be lower than in the lead-in fiber). An example of matching lead-in and transmission fiber profile pair is shown in Fig. 5.
The fiber with the profile shown in Fig. 5(b) will already provide high bend tolerance if excited by the fiber having the profile shown in Fig. 5(a). However, additional consideration should be given to the possible adverse effects of additional modes, which are created by incensed core to cladding index difference in such profile. Besides the high effective refractive index modes, the transmission fiber described above will support propagation of other modes having lower values of refractive index with reduced or even high bend loss sensitivity. Depending on the fiber profile, this set of modes can be large and is usually confined within an entire transversal plane of transmission fiber core. Therefore, the phase constant differences (Δβ) among those modes can be quite small, resulting in a high susceptibility of these modes to coupling by relatively mild microbending. It is, therefore, necessary to prevent coupling of these modes with modes excited by the lead-in fiber (e.g. high effective refractive index modes), otherwise the optical power coupled in the transmission fiber can leak out from the fiber through combined micro and macro-bend effects.
This may be achieved by proper design of the transmission fiber profile. Of particular importance is the shape of the transmission fiber refractive index profile immediately beyond the radial dimension that corresponds to the lead-in fiber core radius. In nearly parabolic profiles, the grouping of modes in modal groups can be used to assure the required separation among the last excited mode and the first unexcited mode of the transmission fiber. This may be achieved by maintenance of proper index gradient in the region of the transmission fiber profile that lies beyond a radial dimension that corresponds to lead-in (excitation) fiber core radius.
The proposed bend resistive transmission fiber profile, therefore, consists of two regions. In the inner region, the transmission fiber profile has the same relative shape as excitation (lead-in fiber). In the outer region (e.g. region beyond lead-in fiber core radius), the transmission fiber profile is graded, preferably parabolic or nearly parabolic. The gradient of this extended region can provide the desired modal group separation in phase constant space and, consequently provide desired fiber microbend tolerance. Modes in the extended region continue to be grouped in modal groups that are separated in phase constant space by the same value as in the lead-in fiber core. While the application of 62.5 µm standard MMF in 50 µm transmission system would already lead to improved bend loss performance, the significant improvement can be achieved by optimized fiber design as explained further below. The maximum size of this extended graded region is not limited and can range up to the fiber edge. However available production process, stress build up caused by doping and maximum achievable refractive index difference limits the size of this region and determines the maximum size of the transmission fiber core. Furthermore, to simplify the production process and to allow for even greater achievable refractive index difference between the core and the surrounding cladding region, the size of the extended graded region can be reduced or truncated. This forms a step transition in the refractive index profile of the transmission fiber. The minimum required size of the graded core extension depends on the desired microbend tolerance of the the transmission fiber, transmission fiber’s outer diameter, operating wavelength, and lead-in (excitation) fiber profile parameters. It is reasonable to require sufficient core extension size to allow for formation of at least one additional modal group in the transmission fiber that will not be excited by the lead-in fiber. In the later case, the separation in β space between the last excited and first unexcited modes (modal groups) is guaranteed to be the same as between the lower order (excited) modal groups. In most practical telecommunication GI fiber cases, the radial dimension of a transmission fiber graded index core extension of less than 5 µm is sufficient to provide the desired modal grouping/separation. Finally, the transmission fiber core is surrounded by cladding. This cladding can be depressed by fluorine doping to achieve additional separation between effective indices of the excited modes and the index of the fiber cladding. Table 1 summarizes design parameters and corresponding criteria that effect bend losses in proposed GI-MMF.
In the previous description we assumed that the launching fiber is positioned between the source and transmission fiber. However the launching fiber can be inserted anywhere down the fiber link as it only acts as a mode filter for removing potentially excessive launched bend sensitive modes with potentially high differential mode delays. More than one launching/filtering fiber can also be used without inducing a significant additional loss in system. For example, launch/filtering fibers can be added on both sides of the transmission fiber to maintain strict compatibility with standard fibers on both sides of fiber link, or a single short section for the launching/filtering fiber can be inserted anywhere down the transmission fiber. Finally, the launching fiber can be completely absent from the system if the source provides sufficient selective excitation of the transmission fiber modes as, for example, properly fiber coupled vertical cavity surface emitting lasers (VCSEL).
2.3 Bend resistive fiber profile design that is fully compatible with 50 µm standard telecommunication multimode fibers
The design of a bend-resistive fiber profile compatible with standard 50 um fiber is presented in this section, in order to illustrate the approach proposed in previous section. This design was also used to produce and experimentally evaluate bend resistive fiber (BRF).
Figure 6 shows a typical profile of standard 50 µm core telecommunication graded index multimode fiber at 1300 nm. The refractive index in the core’s center is typical around 1.4615. The cladding is made of pure silica with a refractive index of about 1.4469 at 1300 nm. The relative difference between the core and the cladding is 1%, and the alpha profile parameter is around α=2.08 (optimized for operation at 850 nm).
The transmission bend resistive fiber profile is tailored to 50 µm standard fiber using the following steps: The standard 50 µm multimode fiber profile, shown in Fig. 6, is used as a template for determining the transmission fiber’s profile shape in the region between the fiber center and the radial distance of 25 µm (e.g. in the area that corresponds to 50 µm multimode fiber core). In order to obtain the desired profile shape of the transmission fiber, the original 50 µm multimode fiber profile is shifted “upwards”, as shown in Fig. 5, until the core center reaches maximum practical refractive index value. In the particular design example, we chose this as 1.4765, which corresponds to about 2% relative difference when compared to the pure silica level (higher index is also possible with germanium doping, but the production process difficulties and stress build-up in the preform makes a higher doping level less convenient for practical fiber production). The graded profile is also extended in the area that is larger than lead-in core size, e.g. in the area beyond 25 µm measured from the fiber core center. In this region the refractive index profile shape is a continuation of the graded (nearly parabolic) profile that extends between 0 and 25 µm. This graded profile extension can continue until the lowest practical value for refractive index is achieved (for example -1%), as indicated by “tailored profile” in Fig. 6.
While such a profile could already provide high bend tolerance in the case of excitation by standard 50 µm fiber, the transmission fiber core would need to be large if the graded profile were to be maintained until the minimum refractive index is reached (e.g. over 80 µm), which would be impractical for the reasons discussed in previous section. The fiber profile is, therefore, truncated at a radial position that corresponds to approximately 32.5 µm (e.g. the truncated core size is approximately 65 µm in diameter) as shown in Fig. 7. The cladding surrounding the core is chosen to have a refractive index about -1% below pure silica level. In this particular design the graded region can only be germanium doped within easily-achievable manufacturing doping level limits, while the cladding can be produced by strong fluorine doping but within the limits routinely achieved in practice by the outside plasma depositions process. Such a combination is, therefore, convenient for practical fiber production.
The transmission fiber profile obtained by the procedure described above, provides a set of modes with high effective indices that closely match the transversal mode field distributions of all propagating modes of standard 50 µm multimode fiber. As an example, the upper row in Fig. 8 shows calculated intensity distribution for modes LP(0,1), LP(5,4) and LP(9,2) of the transmission fiber profile described in Fig. 7. For comparison, intensity distributions for the same modes of the standard 50 µm multimode profile are shown in the bottom row. Figure 8 clearly shows the identical transversal mode field distribution in both cases. When standard 50 µm fiber is spliced to the transmission fiber, the lowest order mode of the 50 µm standard GI MMF very selectively excites the lowest order mode in transmission fiber. The same applies for all other higher-order modes of the 50 µm standard fiber that excite highly selectively their counterparts in the transmission fiber. Since the number of modes supported by standard 50 µm fiber is considerably lower than in the transmission fiber, only the limited set of modes having the highest effective indices is excited in the transmission fiber.
Figure 9 illustrates the values of effective refractive indices for all propagating modes in both transmission fiber having a profile, as shown in Fig. 7, and standard 50 µm GI MMF fiber. Each dot in the graph presents one LP mode. Modes belonging to standard 50 µm multimode fiber are indicated by parenthesis (a) and modes belonging to transmission fiber are marked by parenthesis (b). Parenthesis (c) shows those modes of transmission fiber that are grouped in modal groups and are separated considerably within the effective refractive index domain. Thus the mode coupling between these well-separated modal groups is less likely. Parenthesis (d) indicate those modes of the transmission fiber that are close to one to another in β space (neff space) and are, therefore, more susceptible to microbend induced coupling.
Identical field distributions (as shown in examples of Fig. 8) and mode distribution in refractive index space (Fig. 9) indicate that when the 50 µm GI MMF is connected (for example spliced) to the bend resistive fiber, all modes of 50 µm GI MMF will be “up-converted” into the highest order modes of bend resistive fiber, thus assuring high neff/nc ratio for all excited modes in the transmission fiber. This mode conversion is indicated by the arrow (f) in Fig. 9 and is fully reversible, e.g. when selectively-excited bend resistve transmission fiber is connected to standard fiber, a lossless “down-conversion” of modes can be expected. At the same instance, separation in β space (neff space) between the last excited and the first unexcited mode (modal group) of the transmission fiber is maintained at the same level as in 50 µm GI MMF. A high neff/nc ratio of excited modes and considerable separation between excited and unexcited modes in the β space thus assures low macro bend loss susceptibility and acceptable micro bend performance.
Furthermore, since all the excited modes in the transmission fiber are confined and guided in nearly parabolic region, their modal group velocities are well-balanced, and the presented profile design can provide even higher bandwidth than standard telecommunication GI-MMF. Figure 10 shows simulated modal group delay for LP modes of bend-resistive fiber profile, as shown in Fig. 7 at 850 nm (α=2.087, GeO2 doping is assumed; simulation was performed by OptiFiber from Optiwave Inc.). The differential mode delay (DMD) for approximately first one hundred LP modes that are excited by standard 50 µm GI-MMF proves to be as low as 50 ps/km. This is a considerably lower value than theoretically achievable DMD in standard GI-MMF under overfilled launch conditions. The low DMD of the proposed profile arises from the fact that the nearly parabolic profile extension (used to maintain separation of modes in phase constant space) balances well group delays of the modes of the transmission fiber that are not otherwise well balanced in standard fiber due to the presence of cladding with constant refractive index. This phenomenon was investigated before, and is discussed in detail in ref .
3. Experimental results
Experimental transmission bend resistive fiber (BRF), having the profile shown in Fig. 7, was produced and tested for bend loss performance. Modified chemical vapor deposition (MCVD) was used to cerate germanium doped nearly parabolic region core and plasma-assisted vapor deposition process (PCVD) to form fluorinated low refractive index cladding (core was deposited into low refractive index substrate tube prepared by PCVD by independent supplier). In total, a refractive index difference close to 2.7% was achieved. The profile was measured by the preform analyzer, as shown in Fig. 11. Fine tuning of the fiber profile for the purpose of achieving high bandwidth was not performed due to limited access to fiber production and a need for the extensive production process optimization usually required to achieve high performance MMF fiber production.
The macrobend performance of the proposed system was extensively tested over various experimental configurations as shown in Fig. 12. In all cases 1 and 10 loops of fiber were warped around smooth cylinders having different diameters. Optical powers were measured at the output of all setups when the fiber was straight (Pstright) and wrapped around the cylinders (Pwraped). The ratio of these powers Pwraped/Pstright (i.e. transmission) versus cylinder diameter is reported in Figs. 13 and 14.
Standard 50 µm GI-MMF was used in configurations A and B in order to launch light into BRF. The launching was performed by 850 nm and 1300 nm LED sources. Offset splice was used to perform mode mixing and to create very rich overfill launch conditions in lead-in GI-MMF, thus creating the worst case bend-loss conditions that could be expected in practice. In Configuration A, bend-resistive fiber output was directly coupled to the power meter, while Configuration B used another short section of standard 50 µm GI-MMF at the output, acting as a mode filter. Practically identical bend loss characteristics were obtained in both cases. This indicates absence of the significant mode mixing that would be caused by (macro) bending, as predicted in previous sections. In both configurations and at both wavelengths the loss remained below -0.2 dB for bend-diameters larger or equal to 3 mm in single and ten loop tests. In the single loop test the transmission loss remained below -0.25 dB even at 2.5 mm bend-diameter. A detailed examination of Figs. 13 and 14, however, does show some minor differences between Configurations A and B. Configuration B showed a minor increase in transmission loss (about 1% of the total transmission power) at relatively large bend diameters (around 20-40 mm), but this loss remained stable at lower bend-diameters, as long as the bend diameter did not drop below 3 mm. This minor increase can likely be attributed to the minute presence of mode coupling that is caused by the transition from straight to curved regions of the fiber or by an imperfect launch (launching GI-MMF and practically produced BRF had only approximately matched profiles). The transition from straight to curved fiber at low radii bends can be quite rapid, below or around the beat length (roughly 1.1 mm) of excited modes in BRF. Under such non-adiabatic conditions the minor presence of mode coupling is not unexpected. Further support for this assumption can also be found in the fact that this extra loss is about the same in the both 1 and 10 loops test configurations, and points towards a single event rather than to continuous loss caused by curvature. Reference measurement was also performed using 850 nm LED where standard 50 µm GI-MMF was tested in the same mechanical setup as BRF (Configuration G). In this particular case, no extra mode mixing was used and a loss of optical power of -0.3 dB was obtained for the single loop at diameter of 45 mm. It is apparent that the presented bend resistive system can tolerate at least a 20 times smaller bend diameter than standard 50 um GI-MMF.
Configurations C through D were used to test the behavior of experimentally produced BRF when excited with VCSELs. Two 850 nm VCSELs were used when testing: Finisar 8HFE4391-541 (TOSA package) and Optek OPV214AT (ST receptacle). Both VCSELs are intended for use in gigabit performance multimode links and should, therefore, perform restricted mode launch. The restricted lunching conditions should reflect in the bend loss behavior of the proposed system. The measured and reported  mean value of the encircled flux (EF) for Finisar HFE4391 family of VCSELs at 9 µm diameter is 9.8%, with a standard deviation of 2.2%. At the 38 µm diameter, the mean value for EF is 93.4%, with a standard deviation of only 3.2%. The Finsar VCSEL complies with IEEE 802.3ae requirements. No data was available on EF for Optek OPV214 and is likely not IEEE 802.3ae compliant.
In Configuration C, a Finisar VCSEL was connected through a standard 50 µm GI-MMF to BRF. In Configuration D Optek VCSEL was directly connected (no launching fiber used) to the BRF, but the standard 50 µm MMF was added between the BRF and power meter to perform mode filtering at the output. In configuration D, BRF was connectorized with an ST connector to allow direct coupling of BRF to OPV214AT ST receptacle package.
In Configuration E both Optek and Finisar VCSEL were directly (no launching fiber) connected to the BRF. Again BRF was connectorized with an ST connector to allow direct connection to OPV214AT ST receptacle and 8HFE4391-541 TOSA case. Reference measurements using 50 µm standard GI-MMF were also performed for both VCSEL’s (Configuration F).
Direct coupling of Optek OPV214AT VCSEL to BRF (Configuration E) showed considerably improved bend loss performance of BRF when compared to bend loss obtained in standard fiber, but it is also apparent that this loss is significantly higher than in the case of configurations using launching fibers. It can be concluded that Optek VCSEL launched a significant fraction of the total power into the higher order modes of BRF. When filtering or launching 50 µm standard GI-MMF fiber was added in-between Optek OPV214AT VCSEL and BRF or between BRF and detector (Configuration E), the bend loss characteristics became even better than when LED sources were used in rich over field launches. Ten loop Configuration D also exhibited minor transmission fluctuations over 100%, which can be attributed to mode interference effects at BRF to standard fiber splice. These fluctuations, however, did not exceed 2.5% of total transmitted power.
The behavior of Finisar 8HFE4391-541 VCSEL was quite different from the Optek OPV214AT VCSEL. Direct coupling of Finisar VCSEL to BRF resulted immediately in a high-bend tolerant system with performance very close to those cases using launching GI-MMF fiber. While minute (few percent) power drop can be observed at bend diameters below 8 mm, the bend loss is not significantly higher from configurations using lunching fibers at low-bending diameters (e.g. at a bend-diameter of 3 mm BRF launched by Finisar VCSEL demonstrated about the same loss as in the case of an LED using launching fiber). Apparently Finisar VCSEL does not launch any significant power outside 50 µm diameter spot and therefore could likely be directly coupled to BRF without serious bandwidth and bend-loss impairments. Properly design VCSEL therefore, eliminates the need for launching/filtering fiber.
The highest bend loss tolerance was, however, achieved when launching or filtering fiber was used in combination with VCSEL (Configurations E or D). In the latter cases no bend loss was observed in single loop tests and maximum one percent (-0.05 dB) of optical power loss was detected at 2.5 mm loop diameter in ten loop test configuration.
The microbend performance of the proposed system was evaluated using the sandpaper test. The experimental setup used in the tests is shown in Fig. 15(a) and includes 850 nm and 1300 nm LED sources, launching fiber (standard 50 µm GI MMF), mode mixer, BRF under test, and filtering fiber at the output. The filtering fiber at the output removed power coupled into the modes that are not bend resistive and not well-balanced in terms of group delays. Four loops of fiber under test were placed between 35 cm long and 5 cm wide flat steel plates covered by sandpaper with grades P600 and P1000. Both sandpapers assert a broad spectrum of spatial perturbation frequencies to the tested fibers. The steel plate set-up was gradually loaded by weights while observing optical power transmission through the fiber. A similar test was also performed (Configuration B in Fig. 12) on the standard 50 µm GI MMF for comparison purposes. The results are shown in Fig. 15(b). As expected, both fibers demonstrated practically identical microbend response. Small difference among both fibers (bend resistive transmission fiber appears slightly less microbend sensitive) probably arises from the differences in coatings (different manufacturers and coating conditions) and differences in local numerical apertures (e.g. the refractive index difference of the core that is excited in bend resistive fiber and delta of standard 50 µm fiber were not perfectly matched, and as we used for the reference off the shelf standard 50 µm fiber multimode - Δβ could be slightly different in both fibers).
The total splice losses between standard and BRF fiber were also investigated. Configuration B in Fig. 12 was used to perform loss measurements by the cutback method. The total loss of the structure following the mode mixer (e.g. GI MMF-BRF-GI MMF structure) was lower than -0.15 dB while standard parameters for multimode to multimode fiber splicing were used to obtain splices between BRF and GI MMF (the first splice loss was not measurable, while the second splice loss corresponded to less than -0.15 dB). This indicates that BRF can be inserted anywhere down the standard 50 µm GI MMF link while not causing any excessive optical loses.
Finally it should be stressed that extensive optimization of the experimental BRF was not performed and the experimental BRF profile only approximately matched the profile of standard GI MMF. The profile mismatch should result in a certain degree of mode conversion at the splices and shall be particularly reflected in the second splice losses (e.g. in transition from BRF to GI MMF). Further optimization of BRF profile would likely lead to even lower GI MMF-BRF-GI MMF structure losses.
A high-bend resistive multimode fiber transmission system was presented. The system is based on properly designed bend resistive multimode fiber that supports a set of high-bend resistive modes that are closely matched to all existing modes of standard 50 µm multimode fiber. The experimentally-produced fiber system under worst case conditions demonstrated (LED excitation using additional mode mixing) losses of -0.2 dB when 10 loops of the bend-resistive fiber were wrapped around a cylinder with radius of only 1.25 mm. Even lower bend loss was achieved by combined application of VCSEL and launching fiber where loss as low as -0.05 dB was achieved when 10 loops of fiber were wrapped around cylinder with 1.25 mm radius. This is considerably better bend-loss performance that previously reported even for most single-mode fibers.
The microbend sensitivity of the proposed transmission fiber is comparable to standard 50 µm GI MMF. Furthermore, the proposed fiber transmission system can, theoretically, achieve comparable or even higher bandwidth than standard 50 µm GI MMF, however this was verified only by numerical simulations due to the limited access to the fiber production and ability to perform lengthily and expensive fine-tuning of the production process required for high bandwidth MMF production (however extensive experimental investigation was performed to confirm predicted absence of any significant mode coupling due to the macrobending that could eventually compromise presented modeling results).
While sources like LED’s require a short section of standard 50 µm multimode fiber to launch bend resistive and in terms of group delay well-balanced modes, direct launching by already commercially existing VCSELs of the transmission BRF can also be achieved.
The experimentally produced fiber showed losses below -0.15 dB when interconnected with standard 50 µm GI MMF, however further optimization of the BRF profile is likely to additionally reduce this loss. The insertion of BRF in a 50 µm GI-MMF link is, therefore, completely transparent from the system design perspective.
The proposed bend resistive multimode fiber is compatible with current fiber manufacturing technology and it is easy to splice or otherwise using with standard GI-MMF.
The potential drawback of the proposed approach might be however in stricter requirements for launching conditions and microbend environment. Inappropriate direct launch by source, bad splice or bad connector might lead to the excitation of modes having high DMD that can lead to system bandwidth degradation. Similarly, the system bandwidth might be degraded by pronounced microbend events, like for example local, strong lateral depression of optical cable.
The proposed concept overcomes limitations imposed by the bend loss sensitivity of standard MMFs and allows the use of cost-effective multimode transmission systems in constrained spaces and similar applications requiring tight fiber bending.
I would like to thank Borut Lenardic from OptaCore d.o.o. for producing an experimental BRF prototype and to Marko Kezmah for his help during BRF testing.
References and Links
1. D. Marcuse, “Curvature loss formula for optical fibers,” J. Opt. Soc. Am. 66, 216–220 ( 1976). [CrossRef]
2. D. Marcuse, “Field deformation and loss caused by curvature of optical fibres,” J. Opt. Soc. Am. 66, 311–320 ( 1976). [CrossRef]
3. W. A. Gambling, H. Matsumura, C. M. Ragdale, and R. A. Sammut, “Measurement of radiation loss in curved singlemode fibres,” Microwaves, Opt. Acoust. 2, 134–140 ( 1978). [CrossRef]
5. A. J. Harris and P. F. Castle, “Bend Loss Measurements on High Numerical Aperture Single-Mode Fibres as a Function of Wavelength and Bend Radius,” J. Lightwave Technol. 4, 34–40 ( 1986). [CrossRef]
6. R. C. Gauthier and C. Ross, “Theoretical and experimental consideration for single-mode fibre optic bend-type sensors,” Appl. Opt. 36, 6264–6273 ( 1997). [CrossRef]
7. L. Faustini and G. Martini, “Bend loss in single-mode fibers,” J. Lightwave Technol. 15, 671–679 ( 1997). [CrossRef]
8. D. Donlagic and B. Culshaw, “Low-loss transmission through tightly bent standard telecommunication fibers,” Appl. Phys. Lett. 77, 3911–3913 ( 2000). [CrossRef]
10. G. B. Ren, P. Shum, P, L. R. Zhang, M. Yan, X. Yu, W. Tong, and J. Luo, “Design of all-solid bandgap fiber with improved confinement and bend losses,” Photon. Technol. Lett. 18, 2560–2562 ( 2006). [CrossRef]
12. P. R. Watekar, S. Ju, Y. S. Yoon, Y. S. Lee, and W. T. Han, “Design of a trenched bend insensitive single mode optical fiber using spot size definitions,, Opt. Express 16, 13545–13551 ( 2008). [CrossRef] [PubMed]
14. M. Y. Loke and J. N. McMullin, “Simulation and measurement of radiation loss at multimode fiber macrobends,” J. Lightwave Technol. 8, 1250–1256 ( 1990). [CrossRef]
15. M. Skorobogatiy, K. Saitoh, and M. Koshiba, “Full-vectorial coupled mode theory for the evaluation of macro-bending loss in multimode fibers. application to the hollow-core photonic bandgap fibers,” Opt. Express 16, 14945–14953 ( 2008). [CrossRef] [PubMed]
16. G. Ning, T. Katsuhiro, I. Katsuaki, A. Kazuhiko, and H. Kuniharu, “Hole-assisted holey fiber and low bending loss multimode holey fiber,” US Patent. 7,292,762 ( 2007).
17. K. Yasushi, T. Katsuhiro, and H. Kuniharu, “Low bending loss multimode fiber,” JP Patent Appl. Publ. JP2006047719 ( 2006).
19. D. Marcuse, “Derivation of Coupled Power Equations,” Bell Syst. Tech. J. 51, 229–237, ( 1972).
20. D. Marcuse, “Coupled Mode Theory of Round Optical Fiber,” Bell Syst. Tech. J. 52, 817–842 ( 1973).
21. R. Olshansky, “Mode coupling effects in graded-index optical fibers,” Appl. Opt. 14, 935–945 ( 1975). [PubMed]
23. D. Donlagic and B. Culshaw, “Microbend sensor structure for use in distributed and quasi-distributed sensor systems based on selective launching and filtering of the modes in graded index multimode fiber,” J. Lightwave Technol. 17, 1856–1868 ( 1999). [CrossRef]
24. D. Donlagic and B. Culshaw, “Propagation of the fundamental mode in curved graded index multimode fiber and its application in sensor systems”, J. Lightwave Technol. 18, 334–342 ( 2000). [CrossRef]
25. P. Pepeljugoski, M. J. Hackert, J. S. Abbott, S. E. Swanson, S. E. Golowich, A. J. Ritger, P. Kolesar, Y. C. Chen, and P. Pleunis “Development of system specification for laser-optimized 50-um multimode fiber for multigigabit short-wavelength LANs,” J. Lightwave Technol. 21, 1256–1275 ( 2003). [CrossRef]
26. P. Pepeljugoski, S. E. Golowich, A. J. Ritger, P. Kolesar, and A. Risteski, “Modeling and simulation of next-generation multimode fiber links,” J. Lightwave Technol. 21, 1242–1255 ( 2003). [CrossRef]
27. D. Donlagic, “Opportunities to enhance multimode fiber links by application of overfilled launch, Journal of Lightwave Technology,” 23, 2526–3540 ( 2005).
28. “Encircled Flux Testing,” Advanced Optical Components (Finisra), Internal Technical Report, 8/29/2000.