Measuring bend losses in large-mode-area fibers

Measuring bend losses in large-mode-area fibers Changgeng Ye,* Joona Koponen, Ville Aallos, Teemu Kokki, Laeticia Petit, Ossi Kimmelma nlght Corporation, Sorronrinne 9, 08500 Lohja, Finland ABSTRACT We investigate the measurement of bend losses in few-mode large-mode-area (LMA) fibers. The influence of the light source spectral characteristics, modal power content and cladding light on the measurement accuracy and precision is studied experimentally. Monte-Carlo simulations are performed to understand the distribution of the variations. This study provides practical guidelines for bend loss measurements. Keywords: bend loss, large-mode-area fiber, few-mode, silica optical fiber, fiber lasers, fiber amplifiers 1. NTRODUCTON Large-mode-area (LMA) fibers are widely used in high power fiber lasers and amplifiers. 1 Those fibers typically have a core diameter of 20-30µm and a core NA of 0.06-0.08, which can provide relatively high nonlinear thresholds and power handling capabilities while maintaining good beam quality. The LMA fibers are often coiled with relatively small diameters, mainly for two reasons. The first reason is to obtain a compact package size, which is always desirable for a commercial product. However, the smallest coiling diameter of the fiber is often determined by the bend loss induced laser efficiency reduction. One has to know the bend loss characteristics of the fiber to design the fiber packaging and the related mechanics. Secondly, fiber is coiled to improve the output beam quality. Koplow et al. demonstrated that fundamental mode operation can be achieved in multi-mode fibers by properly choosing the bending diameter of the fiber to introduce differential bend losses to different fiber modes. 2 To do so, the bending diameter of the fiber has to be in a narrow window, so that the higher-order modes (HOMs) are well suppressed with high enough bend losses (at least few db/m), while the fundamental mode (LP01) is with low enough bend loss (< ~0.1dB/m) to maintain the laser efficiency. Thus, the bend loss, or more specifically the bend loss of the LP01 mode, is a basic but important parameter, when using the LMA fibers, especially with compact packages. The measurement of the bend loss seems to be trivial, by taking the transmitted power at a series of bend diameters and processing/fitting the data. However, we found that some factors can significantly affect the accuracy and precision of the measurement results. n this paper, we investigate the bend loss measurement from a practical and engineering perspective, and analyze the influence of the major factors. We also perform Monte-Carlo simulations to understand the distribution of the variations. 2. BEND LOSSES N LMA FBERS The LMA fibers are typically few-moded at their operational wavelength. The eigenmodes of the optical fiber could be solved either analytically under the step-index approximation, or numerically by using the finite difference method (FDM) or the finite element method (FEM). To illustrate the mode structure of a typical LMA fiber within the scope of this study, the eigenmodes of an LMA fiber with a 25µm core diameter and a step-index core NA of 0.07 under an 8cm bending diameter are shown in Fig. 1. *Email: changgeng.ye@nlight.net Fiber Lasers X: Technology, Systems, and Applications, edited by L. Brandon Shaw, John Ballato, Proc. of SPE Vol. 9344, 934425 2015 SPE CCC code: 0277-786X/15/$18 doi: 10.1117/12.2076813 Proc. of SPE Vol. 9344 934425-1

0E 1:1 0Q 50 #1;dn'1e3=1.4875 BL=0.00dB/m: KA D= 19.44 #2;dn'1e3=1.0898 BL=0.00dBm; MFD= 19.90 Sc #3;dn'1e3=1.0601 BL=0.13dBlm; MFD= 20.62um -5u U bu -bu bu 50 0 50 #4; dn'1 e3=0.5925 #5; dn'1 e 3=0.5876 BL=68.98dBlm; MFD= 22.35um BL=37.02dBlm; MFD= 22.25um 50 50 E 5( #5;dn'1e3=0.4642 L=617.74dBlm; MFD= 26.56um 0 50 50 0 50 Fig. 1. Eigenmodes of an LMA fiber with a 25µm core diameter and a step-index core NA of 0.07 under an 8cm bending diameter. The bend losses of different modes can be theoretically calculated using different methods. The analytical approach given by Marcuse 3 is considered to be valid for conventional telecom fibers but inaccurate for LMA fibers. 4 Modified analytical models for LMA fibers exist. 4 On the other hand, numerical methods, such as FDM, FEM and BPM incorporating with perfectly matched layers (PMLs), can provide reasonable results with arbitrary refractive index profiles (RPs). Fig. 2 shows the simulated bend loss as a function of bending diameter for different modes in an LMA fiber with a 25µm core diameter and a step-index core NA of 0.07. The bending is applied to one plane, which makes the two components of the modes (e.g. LP11 and LP11 ) degenerate and have significant differences in bend losses. 100 10 1 0.1 0.01 -#1 LP01 #2 LP11 ; #3 LP11' #4LP21 #5 LP21' #6 LP02 2 3 4 5 6 7 8 9 10 Fig. 2. Simulated bend loss as a function of bending diameter for different modes in an LMA fiber with a 25µm core diameter and a step-index core NA of 0.07. Proc. of SPE Vol. 9344 934425-2

,....... )ss (db/m) ó ó ó 1 3. MEASUREMENT SETUP t is obvious that the bend loss can be measured by recording the transmitted power when fiber is straight (or loosely coiled in practice) and then coiled at a series of bending diameters. A basic setup schematic is shown in Fig. 3. Light Power Fig. 3. Schematic of a simple bend loss measurement setup. t has to be noted that the bend loss measured in this way is actually the combination of bend losses of all propagating modes, since the LMA fiber is multi-moded. The contribution of the different modes cannot be de-coupled from the simple bending test. The results could have significant variations, both during a single test and between multiple tests, because the modal power contents vary. This makes the result interpretation difficult, and the measurement less meaningful. Although it is possible to measure the bend losses of separate modes, for example based on modal decomposition through computer generated holograms (CGH) 5, the technique is rather complex. From the engineering perspective, the bend loss of LP01 mode is the most important characteristics and is already essential for designing fiber lasers. Thus, in this study, effort is made to understand of sources of variations and to improve this setup in order to give more reliable results on the bend loss of LP01 mode. 4. ANALYSS AND DSCUSSON 4.1 nfluence of the modal content LMA fibers support a few modes, and typical input launching conditions cannot ensure a pure LP01 mode in the fiber under test. To have some quantitative understanding on how much the modal content will affect the measurement results, Monte-Carlo simulation was performed. Assuming only LP01, LP11 and LP11 modes exist in the fiber, the simulated bend losses of those modes at different bending diameters are shown as blue, green and red dashed lines in Fig. 4, respectively. f the modal power content of LP01 mode is at 60% and 90%, with a variation of ± 2% standard deviation, and the rest of the power is equally occupied by LP11 and LP11 modes, the measurement results will follow the distributions as scattered blue dots shown in Fig. 4, according to Monte-Carlo simulation with N=10000. : ì ì ì s: ì ì ì ì ì -6 100 r C N CO 101 102 BL - Monte-Carlo - - - LP01 - sim --- LP11 -sim -- - LP11'-sim 2 4 6 8 10 m 10 101 102 BL - Monte-Carlo -- - LP01 - sim --- LP11 -sim -- - LP11'-sim 2 4 6 8 10 (a) LP01 mode 60% ± 2%; (b) LP01 mode 90% ± 2%. Fig. 4 nfluence of the modal content. Proc. of SPE Vol. 9344 934425-3

t can be seen that when LP01 modal content is low, the measured bend loss is significantly biased, giving overestimated bend losses. The variation of the modal content leads to wide spread, especially at large bending diameters. This is a serious issue both for the measurement accuracy and precision. Experimental results also proved this. Fig. 5 shows the comparison of the measurement results when the input fiber is coiled onto fiber mandrels with different diameters, which provide different levels of HOMs suppression (i.e. different LP01 modal power content). 1000 10cm mandrel 100 i 8cm mandrel 7cm mandrel A 0.1 0.01 3 4 5 6 7 8 Fig. 5 Comparison of measurement results with different input fiber mandrel diameters. To increase the LP01 mode content, the input beam should either be properly launched or the HOMs should be filtered out from the input fiber. 4.2 nfluence of the light source t is obvious that any power variation of the light source leads to bend loss measurement errors. So, the light source needs to be as stable as possible. What is not as obvious is that, the spectral characteristics of the light source play an even more important role due to the interference between different modes. To illustrate this, two light sources, including a 1070nm FBG-based linear-cavity fiber laser, a 1064nm laser diode (LD) are tested. Fig. 6(a) shows the output power stability of the two lasers themselves over 2h. t can be seen that the 1064nm LD has much better power stability, while the 1070nm diode pumped fiber laser (air cooled bread-board system) undergoes a thermal stabilization period. Once the whole setup, which consists of passive fibers and the fiber under test, is spliced, the observed transmitted power stability is quite different. n particular, the 1064nm LD power fluctuates significantly, having both for short-term and long-term variation, as shown in Fig. 6(b). 1.00 1.00 0.99 0.95 Normalized Power 0.98 0.97 0.96 Normalized Power 0.90 0.85 0.95 1070nm Fiber Laser 1064nm LD 0.80 FL1070nm LD1064nm ASE1070nm 0.94 0.75 0 20 40 60 80 Time (m) 100 120 140 0 10 20 Time (m) 30 40 (a) stability of the output power of the source; (b) stability of the transmitted power through whole system Fig. 6 Comparison of the power stability of the sources themselves and through the fiber link. Proc. of SPE Vol. 9344 934425-4

The reason is that the 1064nm LD is not spectrally stabilized. The output has very narrow longitudinal mode linewidth, but it hops around within a ~2nm spectral envelope rapidly. The few-moded LMA fiber acts like a Mach-Zehnder interferometer. The transmitted power is strongly modulated by the spectral variations of the source. To avoid the influence of such interference, low coherence sources should be used. We used an ASE source with a 1070nm band-pass spectral filter. The power stability is greatly improved, as shown as the green line in Fig. 6(b). Fig. 7 shows the Monte-Carlo simulation on the influence of the power stability. t can be seen that slight power variation results in significant BL measurement errors at large bending diameters. This effectively sets a lower limit to the measureable range. o m 10 10' 102 BL - Monte-Carlo -- - LP01 - sim --- LP11 -sim -- - LP11'-sim 6 10 Fig. 7 Monte-Carlo simulation on the influence of the power stability. 4.3 nfluence of the cladding light The LMA fibers used in high-power fiber lasers are often with low index coatings (i.e. they are double clad fibers). The light that escapes from the core due to bending is well confined in the cladding. We found that a certain amount of cladding light is coupled back to the core. The coupling coefficient is on the same level of the core-to-clad ratio. Such coupled-back core light is not able to be stripped off by a cladding mode stripper after the coiling section. This makes the measured bend loss underestimated, especially at small coiling diameters and/or when the coiling section is too long. For example, the measurement results at diameter 2-3cm are underestimated due to the influence of the cladding light. This sets an upper limit to the measureable range. 4.4 Measureable range As discussed previously, several factors introduce measurement errors in different ways. And thus, they, altogether, limit the measureable range. n general, the HOM content introduces bias at larger diameters, and sets a lower limit for the bend loss. The power instability introduces measurement variation, and similarly sets a lower limit for the bend loss. The cladding light, however, introduces bias at smaller diameters, and sets an upper limit for the measurable bend loss. Proc. of SPE Vol. 9344 934425-5

)ss (db/m) ó ó ó.... / ^ r ' A l :... t / : 0-1 LP01: 70%. LP01 : 90% - - - LP01 - sim LP11 - sim LP11' - sim 2 4 6 \ fitting' 8 10 Fig. 8 Monte-Carlo simulation taking all factors into account. Fig. 8 shows the Monte-Carlo simulation results and the upper/lower limits of the measureable range, by taking all those factors into account. t can be seen that, the reliable measurement data are within a relatively small range. The data processing/fitting should be done based on such a prediction of limits to get more reasonable results. To improve the reliability of the measurement, more data points can be taken within the measureable range. Any improvements on the above-mentioned factors will broaden the measureable range. 5. CONCLUSON We studied the bend loss measurement of few-mode LMA fibers from an engineering perspective. The influence of the light source spectral characteristics, modal power content, power stability and cladding light on the measurement accuracy and precision is studied experimentally and theoretically through Monte-Carlo simulations. This study provides more understanding and some practical guidelines for the bend loss measurement in few-mode LMA fibers. REFERENCES [1] Richardson, D.J., Nilsson, J., and Clarkson, W.A., High power fiber lasers: current status and future perspectives [nvited], J. Opt. Soc. Am. B 27(11), B63 B92 (2010). [2] Koplow, J.P., Kliner, D.A., and Goldberg, L., Single-mode operation of a coiled multimode fiber amplifier., Optics letters 25, 442 444 (2000). [3] Marcuse, D., Curvature loss formula for optical fibers, Journal of the Optical Society of America 66(3), 216 (1976). [4] Cole, J.H., and Schermer, R.T., mproved Bend Loss Formula Verified for Optical Fiber by Simulation and Experiment, EEE Journal of Quantum Electronics 43(10), 899 909 (2007). [5] Schulze, C., Lorenz, A., and Flamm, D., Mode resolved bend loss in few-mode optical fibers, Optics Express 21, 75 83 (2013). Proc. of SPE Vol. 9344 934425-6