Polarization-OTDR: identifying high-PMD sections along installed fibers

Th 83235

SPECIAL REPORTS: Test and Measurement

A new technique for detecting high PMD allows network operators to maximize the reuse of existing physical plant.

MICHEL LEBLANC, EXFO, and ANDRÉ FOUGÈRES, Gap Optique S.A.

Polarization-mode dispersion (PMD) is a serious limitation in modern optical communications systems, particularly with the advent of 10-Gbit/sec and 40-Gbit/sec systems. That, coupled with the use of optical amplifiers to increase the distance between signal regenerators, has led to a significant decrease in the acceptable level of PMD. Typically, 10-Gbit/sec systems can tolerate up to 10 psec and 40-Gbit/sec systems up to 2.5 psec of end-to-end PMD.

The installation of new transmission systems on existing fiber plant can cause major difficulties because of the predominance of high-PMD fibers. Older fibers, for which PMD was not measured at the time of manufacturing, often exhibit variability in PMD even within a given cable. This variability between fibers suggests that PMD can change significantly along the length of any given fiber link. With the economic incentive to make use of existing installed fibers wherever possible, there is considerable interest in localizing "bad" sections in a fiber link exhibiting an unacceptably large PMD value. Two promising approaches to solving this problem have been purposed and demonstrated. Both solutions are based on polarization-optical time-domain reflectometry (P-OTDR).

End-to-end PMD measurement is done using a variety of techniques: interferometric testing, Jones matrix eigenanalysis, Poincare sphere analysis, and fixed analyzer. The fixed analyzer approach, which was the first technique developed to measure PMD, is based on the analysis of the transmission spectra measured through an analyzer, which is placed at the output of a fiber under test. The source and detection units consist of either a continuous-wave tunable laser/power meter or a broadband polarized source/optical spectrum-analyzer combination. The density of extremes in the transmission spectra (or density of mean-level crossing) is then used to determine the PMD of the fiber. The fixed analyzer technique can be modified to operate in a single-ended configuration. The source and detection units are connected to the same end of the fiber under test and a high-reflectance mirror is connected to the other end. A slight modification of the algorithm converting the detected spectra to PMD is then needed to provide the correct "one-way" PMD value.Th 83235

Figure 1. The wavelength-scanning polarization optical time-domain reflectometer (P-OTDR) uses a pulsed tunable-laser source. To improve the dynamic range of the instrument, an erbium-doped fiber amplifier and an acousto-optic modulator are used to boost the source power. A P-OTDR trace consists of the measurement of the backscattering power intensity (P) against distance (z). A map of P(z,λ) is obtained by taking a large number of P-OTDR traces at different wavelengths. For a given position (L), it is then possible to analyze the backscattering spectra P(L,λ) and deduce the accumulated polarization-mode dispersion up to this point.

An elegant P-OTDR approach is based on an extension of the single-ended fixed analyzer PMD measurement technique. A tunable high power pulsed laser is used as a source. The detection system consists of a high-sensitivity OTDR receiver. A polarizer is inserted just before the connection to the fiber under test (see Figure 1). A P-OTDR trace consists of the measurement of the backscattered power intensity (P) against distance (z). A map of P(z,λ) is obtained by taking a large number of P-OTDR traces at different wavelengths. For a given position L, it is then possible to analyze the backscattering spectra P(L,λ) and deduce the accumulated PMD up to this point.

This P-OTDR technique is conceptually simple but requires a complex hardware implementation. Moreover, measurement time is long because this technique requires a large number of traces. Finally, P-OTDR must provide sufficient spatial resolution to resolve oscillations in the backscattering power as a function of distance. This requirement is more difficult to meet for high-birefringence fibers. Th 83234

Figure 2. The DOP-statistics P-OTDR uses a pulsed distributed-feedback laser source and polarimeter before the receiver. Four P-OTDR traces are taken to measure the degree of dispersion (DOP) against distance. The width of the autocorrelation function of the DOP (hDOP) is calculated for each 250-m section. High value of hDOP indicates a section with poor mode coupling, indicating a high probability of large PMD.

The state of polarization (SOP) of the light transmitted through a fiber is known to rotate about the birefringence axis at a rate that depends on the local birefringence of the fiber. The beat length (Lb) represents the period of this rotation and is inversely proportional to the birefringence (β). For example, a birefringence of 10 psec/km corresponds to Lb=0.5 m for a signal at 1550 nm. If such a fiber were measured with a P-OTDR and if the fiber birefringence were only linear (no circular birefringence), the backscattering signature would oscillate with a period of 0.25 m, since the light travels forward and backward in the fiber. The beat length of the backscattered signal is therefore equal to Lb/2. Accordingly, a P-OTDR would need a spatial resolution (Lp) of roughly 0.1 m (or less) to discriminate the oscillations, requiring optical pulses of 1 nsec or less. Typically, OTDRs do not offer a useful wide dynamic range with such a high spatial resolution, unless a very complex hardware implementation is used. Consequently, the technique of wavelength scanning P-OTDR is normally usable only for low-to-medium birefringence fibers.

A recent demonstration of an alternative technique in which the distributed PMD is not measured directly but inferred from the detection of weak mode coupling also shows promise. This technique can be used even when the resolution of the P-OTDR is not sufficient to resolve oscillations on the P-OTDR traces (e.g., Lp is much greater than Lb, shown as Lp >> Lb). Very-high-PMD fibers can therefore be detected. To understand how the technique works, it helps to review some basic concepts of PMD and fibers.

The PMD of a fiber is known to depend on both the birefringence (β) (relative delay between the fast and slow axis of the fiber), the length of fiber, (L) and the coupling length (h) through the following approximation:Th 83237

Loosely speaking, h can be defined as the distance required for a significant portion of energy in one mode (fast or slow) to be transferred to the other mode. When h is short, there is a considerable amount of "scrambling" between the fast and slow axis, and the total PMD for a fiber increases proportionally to the squareroot of the fiber length (√L ). In contrast, if h is very long, there is very little coupling between the fast and slow axis and the PMD increases linearly with distance (L).Th 83236

Figure 3. The top graph shows the OTDR trace of a link-build from a concatenation of six fibers. The polarization-mode dispersion (PMD) value of each individual fiber was measured initially with an interferometric test set. The degree of polarization (DOP) against distance is measured by the P-OTDR (middle graph), and its statistics are analyzed. The width of the DOP autocorrelation function (hDOP) is plotted on the lower graph. By applying simple threshold rules on DOP and hDOP, it's possible to classify the fibers as low-PMD, medium-PMD, or high-PMD (respectively, light green, orange, and dark green stars on the top graph).

Fibers with very little coupling between the fast and slow axis (long h) most likely will exhibit high-PMD values, because PMD accumulates more rapidly with distance. Therefore, the detection of a long coupling length should allow the identification of most of the high-PMD sections in a fiber link.

To estimate the distribution of h values along the fiber, it is necessary to undertake a fully polarimetric measurement of the SOP as a function of distance. Although that can be achieved via several different P-OTDR implementations, a simple approach is to use a rotating quarter-wave plate followed by a polarizer prior to the P-OTDR detector (see Figure 2). By taking four different P-OTDR traces, with an appropriate orientation for the quarter-wave plate and the polarizer for each trace, it is possible to retrieve the polarimetric SOP information (the four Stokes parameters S0, S1, S2, and S3). It is S0, the degree of polarization (DOP) that contains the critical information needed to estimate h.

The degree of polarization of the light launched by the P-OTDR source can be considered as being 100% (DOP=1) to a first approximation, since the light source is a laser. The DOP of the backscattered light from a specific position along the fiber is also equal to 1. However, the measured DOP (by the P-OTDR) will diminish if the SOP of the backscattered signal against distance varies significantly within the P-OTDR resolution (i.e., if Lb < Lp). The measured DOP against distance will therefore vary depending on the ratio between Lb and Lp. For long P-OTDR pulses (Lp >> Lb), a strong depolarization will occur, but it's possible to distinguish between short and long coupling length (h).

When Lp >> Lb and Lp >> h, the orientation of the fast and slow axis changes rapidly within the P-OTDR resolution. That makes the SOP along the pulse completely random, and the measured DOP collapses.

When Lp >> Lb but Lp << h, the orientation of the fast and slow axis does not change within the P-OTDR resolution and the SOP rotates rapidly around the birefringence axis. The measured DOP will then depend on the angle between the SOP and the birefringence axis. It is therefore expected that the measured DOP will be anywhere between 0 and 1. Slow fluctuations with large amplitude of the measured DOP are therefore expected on fibers with a very long coupling length (h).

The rate of change of the measured DOP can be quantified through calculation of the width of its autocorrelation function (hDOP). That can be thought of as the distance required for the value of the measured DOP to vary significantly. With the aid of numerical simulation, it was demonstrated that hDOP is directly proportional to the coupling length (h) of the fiber. Therefore, by studying the statistics on the DOP, it's possible to localize portions of a link that exhibit large h values, which indicate a high probability of large PMD.

DOP-statistics P-OTDR measurements were taken on deployed legacy fibers at various sites belonging to different carriers. The fibers originated from several different manufacturers and varied in age from 0 to 15 years old.

The first site had a 2.4-km buried cable with multiple fibers. Each fiber was initially characterized with a standard interferometric test set to measure its end-to-end PMD. The measured PMD values ranged from 0.8 psec to 10.3 psec. The different fibers in the cable were then spliced together in a loopback configuration to build a longer link for which the PMD values of each section were known. Figure 3 shows the corresponding P-OTDR measurements, demonstrating how this instrument can differentiate between good and bad fibers. There is a strong relationship between the measured value of the local hDOP and the PMD of each individual section, which indicates that for these fibers, the large PMD originates mostly from weak mode coupling and that the technique used was appropriate to localize high-PMD sections. Measurements taken at other sites confirmed the capacity of the instrument.


Michel Leblanc is a research manager at EXFO (Vanier, Quebec) and AndrÉ FougÈres is a marketing manager at GAP Optique S.A. (Geneva, Switzerland). Their e-mails are michel.leblanc@exfo.com and andre.fougeres@physics.unige.ch.

More in Transmission