Open Access
Add to BibSonomyAbstract
We describe a novel method of measuring PMD (polarization mode dispersion) of an in-service DWDM system by PMD compensation. We successfully demonstrate the method in a 1500-km ultra-long haul DWDM test bed. We further verify the feasibility of the method for in-service light path PMD monitoring in a field trial in a revenue-generating route in Verizon Network, and obtain an accurate PMD value without impacting live network traffic. The discrepancy between the measured and expected PMD values is less than 6% for all cases tested.
©2010 Optical Society of America
1. Introduction
As the bit rate of fiber optic communication systems increases from 10 Gbps to 40Gbps, 100 Gbps [1], and beyond, polarization mode dispersion (PMD) has more and more impact on signal transmission [2–5]. PMD generally causes two principle polarization components of a light signal to travel at different speeds and hence spreads the bit-width of the signal. Consequently, it causes the increase of bit-error rate (BER) and service outage.
Unlike system impairments caused by chromatic dispersion (CD), nonlinearity, or noise, the signal distortion caused by PMD is random in nature and changes with time. At any instant, it depends on not only the PMD value of the signal channel, but also the state of polarization (SOP) of the signal itself. For signal channel of a given PMD value, the signal distortion is maximum when the input SOP is either circular or linear with a 45-degree orientation from the principle state of polarization (PSP) of the signal channel [5,6]. On the other hand, the PMD has little effect on the signal when the input SOP is aligned or counter-aligned with PSP of the signal channel, assuming first order PMD is dominant. Although both are time varying, in general, SOP changes much faster than PMD value of the signal channel [2,7]. The rapid varying SOP is mainly due to external disturbances on the fiber, such as vibration, wind, and temperature changes [8,9].
On the other hand, the PMD value of a signal channel changes randomly when the temperature or mechanical stress on the fiber cable changes, with a probability following Maxwellian Distribution [2,10,11]. For a fiber link with multiple WDM signal channels, the instant PMD values of different signal channels (or wavelengths) may differ, however, they all follow the same Maxwellian Distribution with the same average PMD value if monitored for a sufficiently long period of time, generally tens of hours. Similarly, at a same instant of time, the PMD values of different WDM channels or wavelengths also follow the same Maxwellian Distribution with the same average PMD value. Therefore, one may obtain the average PMD of a fiber link either by averaging PMD values of a signal channel over time, or by averaging PMD values of different DWDM channels at about the same time [12–14].
Generally, the average PMD value reflects the mechanical stress condition of the fiber link itself, and it also changes when the mechanical stress redistributes along the fiber cable due to the changes in environmental conditions along the fiber route, such as snow, rain, flood, mud slide, or earthquakes. Therefore, the average PMD value of a fiber link obtained in prior times may not valid at present and future times. It needs to be monitored to ensure the healthy operation of all the WDM channels of the fiber link, because a WDM channel worked well before may not anymore due to increased PMD. In addition, the signal entering a receiver may come from different routes due to ROADM dynamic reconfiguration and hence experiences different PMD. Such PMD monitoring is useful for diagnosing channel problems and identifying its cause among several possibilities, including CD, PMD or signal to noise ratio degradations. Special measures must be taken to mitigate the PMD effect if identified.
Several methods [2] can be used to accurately measure the PMD of a fiber link, including Jones Matrix method [15,16], Muller Matrix method [16,17], fixed polarizer method [18], white light interferometer method [19,20], and polarization sensitive OTDR method [21,22]. The first three methods require precise polarization control of light signal inputting to the fiber and fast analysis of the received signal before light signal’s SOP changes caused by the external mechanical or temperature disturbances, and therefore not practical for field applications. In addition, the first four methods all require a large wavelength range of tens of nanometers [2,23,24] and not suitable for in-service WDM networks with a typical channel spacing of 50GHz, although they can generally be used to measure the PMD of installed dark fibers themselves without multiplexers, demultiplexers, add-drops multiplexers, and ROADMs. The polarization sensitive OTDR has the advantage of identifying distance resolved PMD of a fiber link, however, its application is limited because its measurement range of about 20 km is much less than a typical fiber span of 80 km. Therefore, no practical method exists for the PMD monitoring of an in-service WDM system.
In this paper, we describe and demonstrate a novel method for measuring PMD in an in-service fiber network by PMD compensation. The basic concept was first suggested by D. Chen [25] and implemented by General Photonics. Using filtered and polarized ASE source with a bandwidth of 0.3 nm as the input light, we successfully demonstrate the method in Verizon’s 1500-km ultra-long haul DWDM test bed over 10 DWDM channels. We further verify the feasibility of the method using an in-service 40Gbps signal itself in the test bed. Finally, we demonstrate the effectiveness of the method in a field trial with a revenue-generating route in Verizon Network. The trial is performed on a single fiber in a selected long-haul fiber route with 23 revenue-generating DWDM channels. No impact of the in-service PMD measurement to the live traffic in the neighboring channels is observed by the network operation center (NOC). The relative error between the measured and expected PMD values is less than 6% for all the cases tested. We anticipate that this method is effective in monitoring PMD in a fiber link having an average PMD larger than 5 ps.
The method has the following advantages: 1) It accurately measures the cascaded PMD of all passive and active components and mixed fiber sections in the fiber route, including EDFAs, multiplexers, demultiplexers, ROADMs, optical fibers. 2) It can obtain the instantaneous effective PMD of each DWDM channel, without requiring a large bandwidth. Even an in-service 40Gbps signal itself can be used as the light source for the measurement. The PMD obtained reflects the effective PMD for the particular modulation format used. 3) It can be used to obtain the average PMD of the whole fiber link by long term monitoring a single DWDM channel. 4) Finally, with proper under and over compensation adjustment, we can see the real time BER impact on a specific channel with any bit rate and modulation format.
2. Concept and system description
The basic concept of PMD measurement by PMD compensation is illustrated in Fig. 1 , where an optical signal passes through a fiber link with a certain amount of PMD (labeled as PMDf herein). At the receiving end, a PMD compensator, consisting of a polarization controller, a variable PMD generator, and a PMD effect monitor, is used to compensate the signal distortion caused by PMDf. The PMD effect monitor is used to provide the feedback signal to the polarization controller to adjust the polarization of the input signal as to minimize the PMD effect. When properly adjusted, signal’s slow polarization component is aligned with the fast axis of the PMD generator, while its fast component is aligned with the slow axis. At the output end of the compensator, the two relatively delayed polarization components are brought closer together. When performing PMD compensation, the compensator will vary the PMD generator and look for a PMD value, PMDb, which can best compensate the signal distortion caused by PMDf. Assuming the resolution of the internal PMD generator is sufficiently high, the PMD value for the best PMD compensation, PMDb, must be the PMD value of the fiber link, PMDf. Such a compensator is sometimes called PMD nulling compensator [2].
Fig. 1 The concept illustration of PMD monitoring by PMD compensation. PMDf is the PMD value of the fiber link and PMDb is the PMD setting of the compensator for best compensation.
Figure 2 shows the system diagram of our implementation of the PMD compensator (a function in General Photonics PMD-1000). Our variable PMD generator is made of 9 pieces of birefringence crystals, with a magneto-optic polarization rotator sandwiched between two adjacent crystals [26], as shown in the insert of Fig. 2. The lengths of the crystals are decreasing successively by a factor of two. Different PMD values can be generated when the solid state polarization rotators rotate signal’s polarization between the crystals. The PMD generator in our system can generate 256 precise DGD values with a resolution of 0.35 ps and a range of 90 ps.
Fig. 2 Illustration of PMD compensator construction. PC: polarization controller. The PMD generator is made of 9 birefringence crystals, sandwiched with 8 MO polarization rotators. The PMD deterministically generator can generate 256 precise DGD values with a resolution of 0.35 ps and a range of 90 ps. The insert at right shows a typical DOP vs. DGD curve obtained by performing PMD compensation at each DGD setting.
We use degree of polarization (DOP) of the signal as the PMD effect monitor and use an in-line polarimeter to obtain the DOP [27]. A polarimeter (Polarimeter 1) before the PMD generator was used to measure the DOP before PMD compensation and a second polarimeter (polarimeter 2) after the PMD generator was used to provide the feedback DOP signal to the digital signal processing (DSP) circuit. The DSP circuit changes the PMD setting step by step and controls the polarization controller to maximize the DOP received by polarimeter 2 for each PMD setting. A plot of DOP vs. PMD shown in the insert of Fig. 2 can be obtained and the PMD setting corresponding to the maximum DOP value is the PMD of the fiber link at the measured wavelength. It is important to note that the comparison between the two DOP values reveals the effectiveness of PMD compensation and can be used for data analysis, as will be discussed below. Note that with the PMD compensator of Fig. 2, one may vary both 1st and 2nd order PMD during the optimization process of PMD compensation. However, because a large number of 1st and 2nd order PMD values to scan through, the time for finding the optimized PMD is much too long to be practical. Therefore we choose to only vary 1st order PMD (DGD) during the PMD optimization process, even though the PMD in the DWDM channel have both 1st and 2nd order PMD components. The obtained optimized PMD (DGD) value actually contains the contribution of both the PMD components from the fiber because the PMD nulling compensator can also mitigate some higher order PMD [28–30], although only 1st order PMD (DGD) is used in compensation. We therefore call it the effective PMD and anticipate that the average PMD obtained using this method through either wavelength average or time average is sufficiently close to the real PMD of the fiber link. In our PMD compensator of Fig. 2, the time for obtaining the optimized PMD value among 256 DGD choices is about 2 seconds.
In operation, The DOP measured by Polarimeter 1 in the compensator, DOP1, is an indication whether the SOP of the signal is aligned with PSP of the transmission fiber or not. A value of DOP1 close to 100% indicate that the input signal’s SOP is nearly aligned with fiber’s PSP and very little PMD distortion is caused by PMD in the fiber. As a result, the DOP after the compensator measured by Polarimeter 2, DOP2, is always close to 100% with and without the compensation action. Consequently, the obtained PMD value for the best compensation is less accurate and the corresponding data points can be removed in post data processing.
On the other hand, a small (less than 90%) value of DOP1 indicates a sufficiently large PMD induced signal distortion and the PMD compensation action is effectively in restoring the signal and bringing the DOP2 back to its maximum. The PMD values obtained for the best compensation for these data points are therefore accurate and can be kept in the data processing.
As expected, this PMD measurement method is less accurate if the PMD in the transmission system is sufficiently small and the corresponding DOP is always larger than 90%, regardless of the polarization launching condition. On the other hand, one usually does not care about small PMD values in the system because their impact on signal transmission is negligible. Therefore this method is very practical for the deployed Long-Haul and Ultra-Long-Haul systems.
3. Laboratory experiments with a channelized ASE source
PMD only impacts signals with a sufficient large bandwidth. For a 40 Gbps signal, the typical bandwidth is around 0.2 to 0.3 nm. To evaluate the feasibility of PMD measurement by PMD compensation for DWDM systems, we first made a tunable channelized ASE (TCA) source to emulate a 40 Gbps signal, as shown in Fig. 3 . The TCA source (General Photonics TCA-1000) consists of an ASE source, a tunable band-pass filter with a 3-dB bandwidth of 0.3 nm, an Er + doped fiber amplifier (EDFA), a polarizer, and a binary polarization rotator with +/− 22.5 degree rotations. The tunable filter can be precisely tuned to WDM ITU grid to simulate WDM channels.
Fig. 3 System configuration for in-service PMD measurement by PMD compensation with a compensator shown in Fig. 2. The tunable channelized ASE (TCA) source with a 3-dB bandwidth around 0.3 nm can be precisely tuned to WDM ITU grid and is used to emulate 40Gbps WDM channels.
As illustrated in Fig. 3, the light signal from the TCA source first goes through a PMD emulator before entering the PMD compensator. The SOP of the light input to the emulator is uncertain and changes with time with respect to fiber’s PSP, causing varied PMD distortion on the light signal, which can be measured by its DOP (DOP1). In operation, the polarization rotator is used to switch the polarization by 45 degrees periodically with a periodicity more than twice longer than the time required for the compensator to obtain best PMD value for compensation (about 2s), and it is not necessary to be synchronized with the PMD compensator. If the input SOP to the transmission fiber is close to the fiber’s PSP, as indicated by a DOP1 close to 100%, a 45 degree rotation will cause it to be at the worst launching condition for PMD distortion, indicated by a DOP1 of a small value. Therefore, the periodical 45 degree polarization rotations assure that a sufficient large number of PMD data points obtained are valid, corresponding to the case of sufficiently small DOP1 values. As describe previously, we will remove the invalid data points with DOP1 close to 100% in post data processing.
Figure 4a shows a typical curve of compensator’s output DOP (DOP2) vs. compensator’s DGD. The compensator has an automatic search algorithm for obtaining the PMD value based on maximizing DOP. The average time for taking each curve is about 2 seconds. We use a piece of PM fiber as the PMD emulator. It is evidence that the DGD value corresponding to the peak DOP is 17.9 ps, the same as that measured with a high precision polarization analysis system [16]. We made 10 tests at different times at each wavelength to obtain the repeatability of our method, and the result is shown in Fig. 4b. The corresponding standard deviation is less than 0.6 ps.
Fig. 4 a) A typical DOP vs. DGD curve of PMD compensation at different DGD values. The DGD value at the peak DOP location is the effective PMD of the DUT (device under test) or SUT (system under test). In this case, the PMD of DUT is 17.9 ps. b) The PMD of DUT measured at different wavelength channels. The standard deviation of measurement is 0.6 ps.
We also made measurements with PMD emulators of different PMD values and found that this method is sufficiently accurate in measuring PMD in a fiber link having an average PMD larger than 5 ps.
4. Experimental results with Verizon 1500 km ultra-long haul test bed
In order to further confirm its feasibility for field applications, we tested our PMD monitoring method with Verizon 1500-km ultra-long haul test bed in Richardson, Texas, as shown in Fig. 5 . The system has a residue PMD on the order of 5 ps. We inserted two PMD emulators of different average values, one at a time, in the system at a location around 700 km from the transmitter side, as shown in Fig. 5. Each PMD emulator is made with multiple sections of PM fibers of different lengths fusion-spliced together with randomly relative orientations. The instantaneous PMD value of each emulator at each wavelength channel is expected to change with time due to temperature variations and the changing rate is on the order of several ps per minutes in an air-conditioned laboratory environment. We measured PMD of both the emulators using a highly accurate PMD analyzer, PSGA-101 from General Photonics [16], and the results are shown in Figs. 6a and 6b. The first emulator, purchased from EXFO, has an average PMD of 11.2 ps, and the second PMD emulator, made by General Photonics, has an average PMD value of 21.6 ps. Considering the residue PMD value around 5 ps in the system, the expected PMD values for the two cases are 12.3 ps and 22.2 ps respectively, following
where PMD1 and PMD2 are the emulator PMD and system residual PMD respectively.Fig. 5 Demonstration setup at Verizon’s 1500 km ultra-long haul test bed in Richardson, Texas. Top portion: Using a TCA source to measure PMD in the system. Bottom portion: using an in-service 40Gbps signal to measure the PMD in the system.
Fig. 6 Experimental results of in-service link PMD measurement by PMD compensation. a) and b): PMD measurement of two emulators using a highly accurate PMD analyzer. c) and d): Instantaneous PMD (dots with error bars) and average PMD (line) obtained by PMD compensation method. The average PMD values obtained for the two different cases are close to those expected based on off-line measurement using a high precision PMD analyzer, considering that the PMD in the system include the contributions from the PMD emulator and the residue PMD of the system itself.
We performed three different types of test to cross check the measurement results. The first is to measure the PMD values of the system over 10 DWDM channels in a short period of time, using a TCA source described in Fig. 3. The average PMD across the 10 channels over 27 nm wavelength range is a good estimation of the average PMD of the link. Table 1 lists detailed experimental conditions of each channel, where TCA Output is the center wavelength and bandwidth of each TCA channel input to the system and Verizon System Output is the corresponding center wavelength and bandwidth measured at the output end of the system. One may notice that although the bandwidth of the input signal (TCA output) is about 0.3 nm, the final output bandwidth is around only 0.2 nm, narrowed by the ROADMs in the system by about 30%. Our PMD monitoring system actually compensates the PMD of a signal of 0.2 nm bandwidth, although the signal inputting at the transmitter end has a bandwidth of 0.3 nm.
Table 1. Experimental conditions of the 10 DWDM channels
Figure 6c and Fig. 6d show the experimental results of the PMD measurements of the 10 DWDM channels. The average PMD values over the wavelength channels are 12.6 ps, and 23.4 ps for the cases of inserting 11.2 ps, and 21.6 ps PMD emulators respectively. The PMD values obtained by the average of 10 DWDM channels are sufficient close to those expected values of 12.3 and 22.2 ps respectively. The corresponding errors are 2.4% and 5.4% respectively.
The second test is to monitor the PMD of a single WDM channel (#50) over a long period of time (over 12 hours) using the channelized ASE source, and obtain the average PMD of the system. We inserted a PMD emulator with an average PMD of 11.3 ps. The total expected PMD in the system is about 12.3 ps, considering the residue PMD in the system. In the measurement, the polarization input to the fiber link was periodically rotated 45 degrees every 30 seconds to make sure sufficient data points are acceptable, as described previously. Figure 7a is the processed PMD data points of a DWDM channel, as a function of time, in which bad data points corresponding to bad SOP launching condition (indicated by high DOP1 values of larger than 90%) were removed. Figure 7b shows the average PMD of the DWDM channel as the sample points increase with time. The average PMD gradually stabilizes with time toward the expected value of 12.3 ps of the fiber link. Longer time may be required for the measured average PMD to have an even better agreement with the expected value. The probability density of the measured PMD values is shown in Fig. 7c and its distribution resembles that of a Maxwellian, with an average value of 12.6 ps, sufficiently close to fiber link’s expected value of 12.3 ps. In the measurement, the polarization rotator in the TCA source is not synchronized with the PMD compensator in order to simulate the real field test situation.
Fig. 7 (a)-(c): experimental results of long term PMD monitoring using TCA source. (d)-(f): experimental results of long term monitoring using 40Gbps signal itself. (a) & (d): instant PMD vs. time; (b) & (e): average PMD vs. time; (c) & (f) probability density function of PMD of all PMD values in (a) and (d) respectively. Note that in the test of using TCA source, there are much more valid data points available because of the periodic 45 degree polarization rotation capability of the TCA source.
Finally, we use a 40Gbps signal with advanced DPSK modulation format itself to perform long term monitoring of the PMD in a system. The test setup is shown in the lower half of Fig. 5 and the results are shown in Figs. 7d-7f. Here we use a different DWDM channel to perform PMD measurement from that using a TCA source shown in Figs. 7a-7c. Because we have no freedom of rotating the input SOP, much fewer data points are valid. As discussed previously, all data points with a DOP1 above 90% were removed during data processing and the processed data are shown in Fig. 7d. The data discontinuities in Fig. 7d are the results of removed invalid data points with DOP1 larger than 90%. As shown in Fig. 7e, the average PMD stabilizes with time, as the data points increase, towards the expected value of 12.3 ps. The probability density as a function of PMD is shown in Fig. 7f and the resulting average PMD is 13.7 ps. Because much fewer data points are valid for a certain period of time than in the case of Figs. 7a-7c, much longer monitoring time is required for getting more accurate result than for the case of using TCA source. Unfortunately, we were only allocated 12 hours for the test. Nevertheless, the obtained 13.7 ps is already sufficiently close to the expected value of 12.3 ps. Note that because measurements using TCA source and 40Gbps signal are performed with different DWDM channels at different times, their instantaneous PMD values are expected to be different, as shown in Figs. 7a and 7d. The results of PMD long-term monitoring at Veriizon 1500km test bed is summarized in Table 2 below. In the table, the test on channel #35 using TCA is also included.
Table 2. Summary of demonstration at Verizon 1500-km test bed
5. Experimental results of field trial
To further test the feasibility of the method for a field technician to measure and identify the PMD of an installed link, we performed the in-service PMD measurement field trial in a long-haul route in an operational network. The expected mean DGD of the route is 19.77 ps, calculated from the mean DGD values of the individual fiber sections, each measured using commercially available PMD measurement equipment before the long-haul system was installed. The length of the route is 414 km with a ROADM at each end, as illustrated in Fig. 8 . There are four in-line optical amplifiers in the route. Add/drop ports are accessible in both ROADMs. In this trial a TCA signal shown in Fig. 3 is injected at Node A and is dropped at Node B to be used for PMD measurement. There are 23 traffic-carrying channels along the route, most having a data rate of 10 Gb/s. 16 idle DWDM channels, from 194.20 to 195.70 THz with a channel spacing of 100 GHz, are used for this trial. Another key objective of the trial is to confirm that our in-service PMD measurement is safe to the live traffic. To this end, the performance of all the working channels is monitored by the NOC during the trial. To ensure that the test signal passes through the route, the idle channels are maintained open temporarily by the NOC during the trial. When the test signal is injected into an idle port, the DWDM system adjusts its optical gain in each amplifier to accommodate the added signal.
We performed two different tests of PMD monitoring: one is to measure the PMD of the route over the 16 WDM channels sequentially in a short period of time (around 15 minutes per channel, including the setup time for each channel) and the other one is to perform long term monitoring for about 3 hours. For the short term PMD test, at least 20 valid data points were taken for each channel and the polarization from TCA source was periodically rotated by 45 degrees every 90 seconds to avoid invalid data points. The average value and its standard deviation of each channel are shown in Fig. 9a . The average PMD over all 16 wavelength channels is 18.57 ps, sufficiently close to the expected value of 19.77 ps. The corresponding error is 6%. The average of more wavelength channels should give even better accuracy. It should be noted that the possible impact of the in-service measurement on the live traffic was monitored very closely with NOC. The measurements triggered no minor or major alarms, thereby demonstrating that in-service PMD measurement methods may safely be applied to traffic-carrying networks.
Fig. 9 Field trial results. a) PMD measured over 16 DWDM empty channels. b) PMD monitoring over time of a single channel. DOP1 is the DOP measured before PMD compensator and DOP2 is the DOP after PMD compensator. The PMD in a real system is much more stable than that in a test bed.
We used a DWDM channel at 195.7 THz for long term monitoring experiment and the results are shown in Fig. 9b. Compared with the test at 1500-km test bed, the PMD value is much more stable with time for the selected DWDM channel, especially after 11:00 PM at night. It is likely because the fiber is buried underground. In order to obtain the statistical average PMD of the fiber route, much longer monitoring time (on the order of 10 days) may be required.
6. Conclusion
In summary, we describe a novel method of using PMD compensation to obtain the PMD value of a fiber DWDM channel of an in-service multiple mixed fiber optic network. The method is simple, fast and does not require a tunable or wide bandwidth light source. There is absolutely no impact to the live long-haul and ultra-long-haul traffic to service provider. Only 2 seconds are required to make a measurement. Either a channelized ASE source with a bandwidth around 0.2 nm or a 40Gbps signal itself of any modulation format can be used as the signal source to the PMD monitoring. We successfully demonstrated the method in Verizon’s 1500-km ultra-long haul test bed in Richardson, Texas and in a field trial in a 414-km revenue-generating fiber route using both channel (wavelength) averaging and time averaging methods. The differences between our measured average PMD values and expected PMD values are less than 6% for the cases tested. We anticipate that the method is accurate for a link with an average PMD of 5 ps or more.
This method can be used 1) for accurately measuring the cascaded PMD of all active and passive components and mixed fiber sections in a fiber route, including EDFAs, multiplexers, demultiplexers, ROADMs, optical fibers; 2) for obtaining the instantaneous effective PMD of each DWDM channel, without requiring a large bandwidth scanning and without interrupting in-service 40G or 100Gbps traffic. The PMD obtained reflects the effective PMD for the particular modulation format used and can be used to determine whether PMD compensation is required for the channel; 3) for obtaining the average PMD of the whole fiber link by long term monitoring a single DWDM channel and gaining knowledge about the PMD degradation of a fiber route. 4) We can also use under or over compensation to reflect the impact on the BER for a specific channel with any type of modulation.
Acknowledgements
We thank Shoa-kai Liu for his great support and Lynn Lin, Hemin Zhang, Zhanfeng Wang, and Jianwei Ma of General Photonics for their design and engineering contributions to the equipment used in this work.
References and links
1. T. J. Xia, G. Wellbrock, W. Lee, G. Lyons, P. Hofmann, T. Fisk, B. Basch, W. Kluge, J. Gatewood, P. J. Winzer, G. Raybon, T. Kissel, T. Carenza, A. H. Gnauck, A. Adamiecki, D. A. Fishman, N. M. Denkin, C. R. Doerr, M. Duelk T. Kawanishi K. Higuma Y. Painchaud, and C. Paquet, “Transmission of 107-Gb/s DQPSK over Verizon 504-km Commerical LambdaXtreme Transport System,”OFC’2008, paper NMC2.
2. S. H. Kogelnik, and R. Jopson, “Polarization-mode dispersion”, in Optical Fiber Telecommunications, IVB, Edited by I. Kaminow and T. Li, Academic Press, ISBN 0–12–395173–9.
3. C. D. Poole, and J. A. Nagel, “Polarization effects in Lightwave systems, in Optical Fiber Communications IIIA, I. P. Kaminnow and T. L. Koch, eds., Academic Press, CA. pp. 114-161.
4. H. Bulow, “System outage probability due to first- and second-order PMD,” IEEE Photon. Technol. Lett. 10(5), 696–698 (1998). [CrossRef]
5. R. Jopson, L. Nelson, G. Pendock, and A. Gnauck, “Polarization-mode dispersion impairment in return-to-zero and nonreturn-to-zero systems,” Proc. Optical Fiber Communication Conference, OFC’99, paper WE3 (1999).
6. C. D. Poole, J. H. Winters, and J. A. Nagel, “Dynamical equation for polarization dispersion,” Opt. Lett. 16(6), 372–374 (1991). [CrossRef] [PubMed]
7. H. Bulow, W. Baumert, H. Schmuck, F. Mohr, T. Schulz, F. Kuppers, and W. Weiershausen, “Measurement of the maximum speed of PMD fluctuation in installed field fiber,” Proc. OFC ’99, Technical Digest, W: 83–85 (1999)].
8. J. Cameron, L. Chen, X. Bao, and J. Stears, “Time evolution of polarization mode dispersion in optical fibers,” IEEE Photon. Technol. Lett. 10(9), 1265–1267 (1998). [CrossRef]
9. D. Waddy, P. Lu, L. Chen, and X. Bao, “The measurement of fast state of polarization changes in aerial fiber,” Proc.OFC ’01, paper ThA3 (2001).
10. G. J. Foschini and C. D. Poole, “Statistical theory of polarization dispersion in single mode fibers,” J. Lightwave Technol. 9(11), 1439–1456 (1991). [CrossRef]
11. F. Curti, B. Daino, D. De Marchis, and F. Matera, “Statistical treatment of the evolution of the principle state of polarization in single mode fibers,” J. Lightwave Technol. 8(8), 1162–1166 (1990). [CrossRef]
12. M. Karlsson, J. Brentel, and P. Andrekson, “Long term measurement of PMD and polarization drift in installed fibers,” J. Lightwave Technol. 18(7), 941–951 (2000). [CrossRef]
13. C. Poole and R. Wagner, “Phenomenological approach to polarization dispersion in long single-mode fibers,” Electron. Lett. 22(19), 1029–1030 (1986). [CrossRef]
14. N. Gison and J. Pellaux, “Polarization mode dispersion: Time versus frequency domains,” Opt. Commun. 89(2-4), 316–323 (1992). [CrossRef]
15. B. Heffner, “Automated measurement of polarization mode dispersion using Jones matrix eignenanalysis,” IEEE Photon. Technol. Lett. 4(9), 1066–1069 (1992). [CrossRef]
16. X. S. Yao, X. Chen, and T. Liu, “High accuracy polarization measurements using binary polarization rotators,” Opt. Express 18(7), 6667–6685 (2010). [CrossRef] [PubMed]
17. L. Nelson, R. Jopson, and H. Kogelnik, “Muller matrix method for determining polarization-mode dispersion vectors,” Proc. European Conference on Optical Communications,” ECOC’99, Vol. 2, p. 10 (1999).
18. C. Poole and D. Favin, “Polarization-mode dispersion measurements based on transmission spectra through a polarizer,” J. Lightwave Technol. 12(6), 917–929 (1994). [CrossRef]
19. K. Mochizuki, Y. Namihira, and H. Wakabayashi, “Polarization mode dispersion measurement in long single mode fibers,” Electron. Lett. 17(4), 153–154 (1981). [CrossRef]
20. L. Thevenaz, J. Pellaux, N. Gisin, and J. von der Weid, “Birefringence measurement in fibers without polarizer,” J. Lightwave Technol. 7(8), 1207–1212 (1989). [CrossRef]
21. D. Bebbington, J. Ellison, R. Schuh, X. Shan, A. Siddiqui, and S. Walker, “Fully polarimetric optical time-domain reflectometer with 1-m spatial resolution,” Proc. Optical Fiber Communication Conference, OFC’97, Technical Digest, pp. 185–186 (1997).
22. H. Sunnerud, B. Olsson, and P. Andrekson, “Technique for characterization of polarization mode dispersion accumulation along optical fibers,” Electron. Lett. 34(4), 397–398 (1998). [CrossRef]
23. N. Gisin, B. Gisin, J. von der Weid, and R. Passy, “How accurately can one measure a statistical quantity like polarization-mode dispersion?” IEEE Photon. Technol. Lett. 8, 1671–1673 (1996). [CrossRef]
24. J. Cameron, L. Chen, X. Bao, and J. Stears, “Proc. European Conference on Optical Communication, ECOC’99, Vol. 1, p. 308 (1999).
25. David.Chen, private communications.
26. L. Yan, C. Yeh, G. Yang, L. Lin, Z. Chen, Y. Shi, A. Willner, and X. Steve Yao, “Programmable Group-Delay Module Using Binary Polarization Switching,” J. Lightwave Technol. 21(7), 1676–1684 (2003). [CrossRef]
27. N. Kikuchi, “Analysis of signal degree of polarization degradation used as control signal for optical polarization mode dispersion compensation,” J. Lightwave Technol. 19(4), 480–486 (2001). [CrossRef]
28. M. Karlsson, C. Xie, H. Sunnerud, and P. Andrekson, “Higher-order polarization mode dispersion compensator with three degrees of freedom,” Proc. Optical Fiber Communication Conference, OFC’01, Paper MO1 (2001).
29. S. Lanne, W. Idler, J.-P. Thiery, and J.-P. Hamaide, “Demonstration of adaptive PMD compensation at 40Gb/s,” Proc. Optical Fiber Communication Conference, OFC’01, Paper TuP3 (2001).
30. J. Nagel, M. Chbat, L. Garrett, J. Soigne, N. Weaver, B. Desthieux, H. Bulow, A. McCormick, and R. Derosier, “Long-term PMD mitigation at 10Gb/s and time dynamics over high-PMD installed fiber, Proc. European Conference on Optical Communication, ECOC’2000, Vol. 2, p. 31(2000).
