BEHAVIOR OF THE PMD COEFFICIENT OF OPTICAL LINKS UNDER INFLUENCE OF MECHANICAL AND THERMAL EFFECTS

Transcription

1 71 BEHAVIOR OF THE PMD COEFFICIENT OF OPTICAL LINKS UNDER INFLUENCE OF MECHANICAL AND THERMAL EFFECTS Jeancarlo Dotto, Januncio A. de Medeiros Neto and Alexandre A. P. Pohl J. Dotto is with Furukawa Industrial S.A.- Produtos Elétricos, Curitiba; A.A.P. Pohl is with the Centro Federal de Educação Tecnológica do Paraná Curitiba and J.A. de Medeiros Neto is with the Universidade Estadual Paulista Júlio de Mesquita Filho (UNESP), Itapeva. Endereço para contato: CEFET-PR, Av. Sete de Setembro, 3165 CEP: Curitiba, Paraná. tel: / fax: Abstract This papers presents results on the variation of the PMD coefficient of optical links under influence of mechanical tests, such as tensile strength, bending and compression, and also during the application of a thermal cycle. Results revealed that the link coefficient is more influenced by the application of a tension load and also suffers significant variation under strong temperature changes. Index Terms Optical Cables, Polarization Mode Dispersion, PMD Link Coefficient, Environment Effects on Cables. I INTRODUCTION With increasing bit-rate, polarization mode dispersion (PMD) is considered to be the limiting mechanism in high-capacity single-mode fiber-optic transmission systems when material and waveguide dispersion effects are compensated for [1]. In an ideal single-mode fiber the fundamental propagation mode is indeed split into two polarization modes that propagate with the same group velocity along perpendicular axes. The dispersion of the polarization modes arises when circular symmetry is broken by the presence of an elliptical core or by non-circular symmetric stresses, which are induced during the fiber drawing process. The loss of circular symmetry removes the degeneracy and causes an intrinsic birefringence resulting in the two polarization modes propagating with different group velocities. Besides, fibers are submitted to tension, bending and compression during the cabling process and field installation, after which the optical cable is continuously exposed to harsh environment and strong temperature variation during its lifetime. All these mechanisms induce a complex and random variation of the birefringence. Therefore, the polarization mode dispersion is a parameter that must be treated statistically [2]. From the practical point of view, it is imperative that optical cable manufacturers be able to guarantee a low PMD in their products. This way, manufacturers are interested in understanding how the PMD value of different cable configurations are affected by

2 72 environmental conditions such as bending, compression and temperature variations. Above all, this behavior shall be understood for cables spliced together and building an optical link. And since the PMD value of a single piece of cable provides only a limited insight into the phenomenon, one should take into consideration the analysis of the PMD variation in the optical cable link as a whole. This approach may allow manufacturers in a feedback process to optimize cable parameters and improve design and fabrication conditions. With this purpose in mind, we report in this article a series of mechanical and thermal cycle tests applied to links built with different types of cables and the corresponding results. Following some recognized Brazilian (ABNT) test standards and pushing test variables beyond parameters specified in the norms, we were able to evaluate the PMD behavior under harsher conditions. With these results it is possible to specify an upper limit for the PMD link coefficient based on the cable configuration. This article is structured the following way: section II describes the types of cables used in the experimental tests, section III explains how the PMD coefficient of a link may be estimated departing from measured PMD values of single fibers in cables by using the Monte Carlo Method; section IV describes the mechanical and the thermal cycle tests applied to loose tube optical cable configurations and reports on the behavior of the PMD link coefficient. Finally, section V brings our comments on the obtained results. II CABLE DESIGN Three types of cables were used in the experiments, all with fibers placed in loose tubes stranded around the cable central member: a single jacket cable with 30 fibers used for inner duct installation, denominated here IDC, a single jacket cable with 24 fibers used for aerial installation, called DD-G, and an all-dielectric self-supported cable with 36 fibers for aerial installation, known as AS. Both single jacket cable differ only in the number of loose tubes inside the cable. On the other hand, the AS cable presents an extra external jacket. Fig. 01 shows a cross-section of such cables. All cables were fabricated with standard G652 single mode optical fiber. Fig. 01 Left side: cross-section of the single-jacket optical cable (IDC or DD-G type). Right side: crosssection of the all-dielectric self-supported (AS) loose tube cable. Table 1 describes the characteristic parameters of the IDC and the AS cables. The main differences in the design of the AS and the single-jacket cables are the lay length and the central member diameter. Due to this, fibers are submitted to different degrees of stress associated with the cable design type, which can induce a significant variation of the fiber

3 73 birefringence along the cable length. The theoretical bending strain (see Table 1) associated with the cables was estimated considering the fibers inside the loose tube and stranded around the strength member and a cable without bending. Table 1- Summary of Characteristics IDC-30 fibers AS - 36 fibers Fiber per tube 6 6 Quantity of tubes 5 6 Maximum fiber 4 mm 10 mm overlength in 10 m of loose tube Internal diameter of loose tube 1.7 mm 1.7 mm External diameter of loose tube 2.5 mm 2.5 mm Lay length 80 mm 90 mm Bending Strain 0.11% 0.20% III ESTIMATION OF THE PMD LINK COEFFICIENT The polarization mode dispersion arises in single mode fiber when circular symmetry is broken by the presence of an elliptical core or by non-circularly symmetric stresses. The deviation from the geometrical symmetry breaks the mode degeneracy and causes a birefringence resulting in the polarization modes propagating with different group velocities along two orthogonal axes. The birefringence in optical fiber is also perturbed by microbends, twists and temperature change, which are randomly distributed along the fiber. The time delay between the two polarization modes at a given wavelength is known as the differential group delay (DGD). The variation of DGD in time and wavelength follows a Maxwell distribution [3]. It is usually measured in a picosecond scale and it represents a singular result of a statistical process. The average of DGD values across wavelengths and time yields the PMD value, which represents the expected value (mean) of the random process. In a practical way, a parameter denominated PMD coefficient is defined as the mean value of DGD divided by the square root of the length (ps / km) of a cable section or link. This definition is consistent with results observed experimentally, where PMD assumes a dependence with the square root of the distance [4] for long fiber lengths. Usually a long optical link has hundreds of kilometers. However, the fabrication process limits optical cable length to 5 or 6 km. This way, an optical link is formed by several sections of optical cable spliced together. The PMD coefficient measured on single cable sections is not enough to provide information on the PMD of the link. So, one needs to define a PMD coefficient for the optical link. This may be done by taking the square root of the sum of squares of the PMD coefficients of the fibers in each cable section that is used to form the link: X M = M i= 1 M x 2 i (1)

4 74 Where: M number of cable sections of equal length that constitutes the link x i The PMD coefficient of a fiber in an cable section ( ps/sqrt(km) ) X M The PMD coefficient of a link with M cable sections ( ps/sqrt(km) ) As the PMD is a random process, one has to calculate the probability distribution function of the link coefficients, which depends on M and on the distribution of the PMD coefficients of the fibers in each cable section. There are some methods to evaluate the link coefficient [5]. In this paper, the process for calculating it applies the Monte Carlo technique [6], explained below. In this method, the maximum link coefficient for a given distribution (PMD max ) is determined in terms of a very small value, Q, which gives the probability for a PMD link coefficient to exceed the maximum value. III.1 Monte Carlo Technique In order to build the probability distribution using the Monte Carlo Method, the PMD coefficient of a fiber in each cable section, x i, is picked up randomly from a set of measured PMD values belonging to that cable section. These are then added in quadrature according to equation (1), in order to calculate one value, X M, of the distribution for a link comprised of M cable sections. This same process is repeated k times, where k 10 / Q. If Q is equal to 10-4, one calculates possible values for the link coefficients. The maximum PMD coefficient is then determined from the accumulated probability density function given by m c m = p k k = 1 (2) where p k represents the normalized relative frequency in which the values of the link PMD coefficient appear in the distribution. The maximum PMD coefficient of the link, X Q, is defined as the first value of X M that satisfies the condition c m 1 Q. III.2 The PMD Link Coefficient Distribution of AS and IDC Cables In order to build the probability distribution function, the PMD coefficient of fibers was measured by the interferometric technique with a 1.55 µm light source [7]. Measurements were performed in one hundred cabled fibers of the IDC and AS cable type. All measurements were made in a room where temperature was kept constant at 21 C. Examples of the measured values and the calculated PMD link coefficient distribution are seen in figures 02 and 03. Fig. 02 shows on its left side the distribution of the measured values in fibers of an all-dielectric self-supported (AS) loose tube cable and on its right side the calculated PMD link coefficient distribution using the Monte Carlo Technique. Fig. 03 shows on its left side the distribution of the measured values in fibers of a single jacket (IDC) loose tube cable and on its right side the calculated PMD link coefficient distribution. Both link coefficients were calculated considering a link made of 20 cable sections of equal lengths (the AS cable presented a section length of 3,25 km and the IDC a length of 4 km). For the AS cable the mean PMD, considering the measured values in the fibers, was ps/sqrt(km) and the maximum measured PMD value obtained for one of the fibers in the cable was ps /sqrt(km). For the IDC cable the mean PMD calculated using the measured values was ps/sqrt(km) and the maximum PMD value obtained was ps/sqrt(km). The

5 75 maximum link coefficient was calculated according to Monte Carlo as 0.30 ps/sqrt(km) for the AS cable and 0.47 ps/sqrt(km) for the IDC cable. The calculated Monte Carlo distributions are strongly influenced by the distribution of the measured values in each cable type. For instance, the measured PMD values seen in Fig. 03 (left side) are much more spread than the values appearing in Fig. 02. This spreading is a consequence of the number of fibers in the cable and also depends on the way fibers are stranded around the cable central member, which is mainly dictated by the lay length or, in another words, the fabricating process. Figure 2 The left side shows the distribution of the measured values in fibers of an all-dielectric self-supported (AS) loose tube cable and the right side shows the calculated PMD link coefficient distribution. Figure 3 The left side shows the distribution of the measured values in fibers of a single jacket (IDC) loose tube cable and the right side shows the calculated PMD link coefficient distribution. IV ENVIRONMENTAL TESTS Many of the mechanical and thermal tests applied today deals with experiments on cable spools. These tests follow well established standards and apply to short pieces of cable. However, it is difficult to simulate in the laboratory an environment such as the field where optical cable links are installed. Despite such difficulties, one may devise some tests in the lab that can translate the external physical conditions to the optical link. In this process, some simplifications need to be taken into account. The most important of them is that fibers of the

6 76 same cable spool are spliced together to build up a link of equal cable lengths. The so called cable link is then submitted to mechanical and thermal tests. The standards used in this work, as reference for the tests, belong to norms issued by the Associação Brasileira de Normas Técnicas (ABNT) [8,9,10,11]. IV.1 MECHANICAL TESTS The relevant mechanical tests handle situations where fibers are submitted to strain, compression and bending. Specifically, these tests are denominated tensile strength, compression and macrocurvature tests. They have the purpose of simulating field situations during cable installation and during the exposure of cables to harsh environment over their lifetime. Particularly in this work, some physical parameters were pushed beyond the limits established in the norms with the purpose of checking the behavior of the PMD link coefficient in extreme situations. The interferometric technique was used again, but now to monitor the PMD coefficient of the link. For the tensile strength test the load is applied at an interval of 4 km, in a section of 150 meters of fiber. As the fibers in the spool are spliced together, this corresponds to applying the load to several paths of fiber along the optical link. Although this may not correspond to a truly field condition, the PMD link parameter will vary as a function of the different loads applied to the cable. The load is applied taking into account the cable's weight per its length [8]. During the test, the PMD parameter was monitored during one hour for each applied load. Fig. 04 shows the PMD link coefficient behavior for the AS and DD-G cable. A closer look at the curves, indicate two behavior regions for both cable types as the load is increased. The first region shows that the PMD link coefficient goes slightly down as the load is increased from the reference point (no load). The second region appears after a certain load is reached (approximately 300 Kgf for the DD-G and 750 Kgf for the AS cable) and marks the increase of the PMD link coefficient. A possible explanation for this behavior lies on the fact that fibers inside the cable have more room, that means, the length of fiber is longer than the length of cable. This fact arises from the cabling process and the cable design. As the load is applied and the cable is stretched, fiber does not feel the tension load immediately. Fibers seem to relax and occupy the available room inside the loose tubes. After the limit is reached, fiber feels the tension load applied to itself and, as a result, the PMD link coefficient starts rising. This limit occurs for tension loads corresponding to several times the cable s weight (approximately 3 times for the DD-G and 5 times for the AS). Macrobending effects were analyzed submitting the link of a DD-G cable to different bending radius. The tests were made simulating the bending according to limits established in the norm for the specific cable type [9]. After that, the bending radius was reduced for simulating a critical situation on the link. For each bending radius, the PMD link coefficient was monitored during one hour, in which 30 measurements were taken every 2 minutes. Fig. 05 shows two curves for the behavior of the link coefficient. The curve with dotted points shows the behavior after a retest of the cable link. For the retest, splices were broken, a piece of fiber cut off the cable ( 200 m) and fibers spliced again to form a link of a slightly shorter length. The curves show first a small decrease of the PMD link coefficient in both tests. This coefficient decrease manifests itself only in the third digit after the coma (see vertical axis of Fig. 05), which may be considered of no relevance. Important is the fact that the values of the link coefficient after retest (see solid circles curve) is lower than the values obtained during the first test. This is attributed to the piece of fiber that was cut out of the cable. In this case,

7 77 the dispersion for the shorter link should be lower than in the longer link (see solid squares curve). Fig. 04 Variation of the PMD link coefficient as a function of the tension load applied to the cable. The solid circles represent measured values for the DD-G cable and the solid squares represent measured results for the AS cable. Fig. 05 Variation of the PMD link coefficient of a DD-G 24 fiber cable as a function of the cable bending radius. The effects of compression on the PMD link coefficient is shown in Fig. 06. Two cable links of the same type (DD-G), but with different amount of fibers in each one were

8 78 submitted to a test, where a compressive load is applied in the vertical direction across the cable cross-section. The compression load was applied taking into account the maximum load defined in Standards [10]. One sees that for both cables the load increase does not affect the PMD link coefficient. Only after a load of 2000 N is applied, the coefficient suffers a very small increase. It is important to observe that the PMD link coefficient differs much for both cables. The cable with 30 fibers shows a coefficient of 0.08 ps / sqrt(km), while the cable with 24 fibers presents a value of ps / sqrt (km). A cable with 30 fibers has on more loose tube, where fibers are inserted, than its 24 fibers counterpart. This packing of more loose tubes inside the cable is responsible for the strain that provokes the PMD increase. IV THERMAL CYCLE TESTS An AS cable was submitted to temperature cycling, in order to verify the PMD link coefficient behavior under extreme temperature variation. The tests were made building a link by splicing 36 fibers together, which were chosen randomly from the AS cable. The total measured optical loss for the link was 15 db. The PMD link coefficient was observed submitting the spliced cable to a temperature cycle, where temperature varied from 20 o C to +65 o C, according to the NBR Standard [11]. Fig. 07 shows the results for a cycle with a duration of about 200 hours. During a single temperature step, positive or negative, it was observed that the PMD link coefficient varied strongly, where variations of more than 50% were noted. Table 2 shows the average coefficient calculated in each temperature step. It is also important to note that as cable goes through the different temperature steps, which simulate the aging process, the link coefficient steps up to a higher value. After the test is finished the link coefficient remains at a higher value. This behavior may suggest that under extreme temperatures, which may indeed occur in the field during the cable lifetime, the PMD link coefficient gets higher and should be compensated for in order to avoid the performance degradation, specially in long links. Studies relating extreme temperature and the PMD variation are in progress in order to better understand this issue. Fig Variation of the PMD link coefficient of a DD-G 24 fiber and a DD-G 30 fiber cable as a function of the compression load in Newtons.

9 79 PMD Coef. (ps/sqrt(km)) Temperature PMD Coefficient ( ps / (sqrt km) ) 0,3 0,25 0,2 0,15 0,1 0, Time ( hours ) Temperature ( Celsius ) Fig PMD link coefficient of an As fiber cable as a function of temperature and time. Table 2- Average values of the PMD link coefficient in the thermal cycle test Temperature ( o C) in each step Average PMD link coefficient (ps / sqrt (km) ) ,152 0,193 0,174 0,196 0,185 0,203 0,212 V CONCLUSION The influence of mechanical and thermal effects on the PMD link coefficient using different types of optical cable was studied. For estimating the maximum PMD coefficient of a link, the measured PMD coefficient of single fibers in a cable was used to build up the link probability distribution function by using the Monte Carlo method. The maximum link coefficient was calculated as 0.30 ps/sqrt(km) for the AS cable and 0.47 ps/sqrt(km) for the single jacket (IDC) cable. These results are strongly influenced by the distribution of the measured values in each type of cable. Comparison of results concerning the mechanical tests (tensile strength, macrocurvature and compression) shows that the PMD link coefficient is more influenced by the tensile strength test. After a certain load is reached the coefficient starts to increase. In this case a variation of about 30% was noted in both types of cable (AS and DD-G). Results of the bending and compression tests showed no significant variation in the PMD coefficient. This may be due to the fact that in such tests the fiber suffers the bending and the load only at

10 80 single points along the link, while for the tensile strength test, the load is applied to sections of cable as explained before. The PMD link coefficient behavior of an AS optical cable was studied considering a thermal cycle with strong temperature variation. Results showed that the coefficient suffers stronger variation at each temperature step and that the link coefficient increases as more cycles occur. This suggests that in an aging process the PMD link coefficient degrades over time. When comparing results of mechanical and thermal cycling tests, the later have a stronger influence on the PMD link coefficient than the former. REFERENCES [1] P.A.Andrekson, High-Speed Soliton Transmission on Installed Fibers, presented at the 2000 Optical Fiber Conference, Baltimore, USA, March 2000, paper TuP2. [2] G. J. Foschini and C. D. Poole, Statistical Theory of Polarization Dispersion in Single- Mode Fibers, J. Lightwave Technol., vol. 9, pp , Agosto [3] N. Gisin, R. Passy, J. C. Bishoff and B. Perny, Experimental Investigations of the Statistical Properties Polarization Mode Dispersion in Single Mode Fibers, IEEE Photonics Techn. Lett., vol. 5, no. 7, pp , [4] G. P. Agrawal, Fiber-Optic Communication Systems, John Wiley & Sons, Inc., 2 nd edition, chapter 2, [5] C.E.P. Blume, A.A.P. Pohl and J.A.Medeiros Neto, Modelos Estatísticos e sua Aplicação na Estimativa do Coeficiente de PMD Máximo em Enlaces Ópticos, Rev. Soc. Brasileira de Telecomunicações, vol. 17, no. 1, pp , [6] L.F. Marques, A. M. Simião, R. F. Cruz, J.A. Medeiros e M.R. Boulos, Statistical Analyses of PMD Using Monte Carlo Method for Different Configuration of Loose Tube Optical Cable, in Proceedings of OFC 2001, 2001, paper WDD12. [7] R. Cross, PMD Measurements Methods, in Fiberoptic Product News, pp.27-29, Junho [8] ABNT NBR Cabos ópticos - ensaio de tração em cabos ópticos e determinação da deformação da fibra óptica - out [9] ABNT NBR Cabos ópticos - ensaio de curvatura - out [10] ABNT NBR Cabos ópticos - ensaio de compressão - out [11] ABNT NBR Cabos ópticos - ensaio de ciclo térmico - out 1995.