IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 53, NO. 1, FEBRUARY

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 53, NO. 1, FEBRUARY 2004 67 Chromatic Dispersion Measurement Using a Multiwavelength Frequency-Shifted Feedback Fiber Laser J.-N. Maran, Radan Slavík, Sophie LaRochelle, and Miroslav Karásek Abstract Although fiber lasers emitting at several wavelengths would find many applications in optical communications, their present performance, for example, in terms of wavelength coverage, is not sufficient to meet current requirements. In this paper, we present the realization of a fiber laser source emitting simultaneously over 17 wavelengths spread over the whole C-band. An acoustooptic frequency shifter is placed in the laser ring cavity to suppress the cross-gain saturation effects of the erbium-doped fiber. The emitted wavelengths are fixed by a set of fiber Bragg gratings (FBGs). A power uniformity reaching 6 db is achieved by inscribing the FBGs while monitoring the laser output. We demonstrate the reliability of this laser as a source for characterization of optical components and networks by the measurement of optical fiber chromatic dispersion. The measurement is performed over the whole telecommunication C-band (1530 1560 nm) using the time-of-flight method. We perform the measurement on three different fibers with different levels of dispersion, namely a standard fiber, a nonzero dispersion shifted fiber, and a dispersion compensating fiber. The results are compared with measurements obtained using a standard network analyzer. The agreement between the two methods is better than 1%, thus proving the suitability of the developed laser source for this application. Index Terms Optical fiber dispersion, optical fiber lasers, optical fiber measurement applications, ring lasers. I. INTRODUCTION IN RECENT years, the communication industry has developed and installed wavelength division multiplexed (WDM) optical systems to meet the growing bandwidth demand resulting from the expanding data traffic. The capacity of WDM networks can be further increased by adding extra wavelengths or by modulating the channels at higher rates. At high bit rates, 10 Gbit/s or 40 Gbit/s, optical communication systems now face limitations imposed by nonlinear effects and chromatic dispersion. As the optical signals propagate in the fiber, these two phenomena cause severe pulse distortions resulting in intersymbol or interchannel cross talk. These detrimental Manuscript received March 1, 2003; revised August 27, 2003. This work was supported in part by the Canadian Institute for Photonic Innovations, and its industrial affiliates, and in part by the Canada Research Chair Program. J.-N. Maran and S. LaRochelle are with the Centre d optique, photonique et laser, Department of Electrical and Computer Engineering, Université Laval, Québec, QC G1K 7P4 Canada (e-mail: larochel@gel.ulaval.ca). R. Slavík is on leave from the Institute of Radio Engineering and Electronics, Academy of Sciences of the Czech Republic, Prague, Czech Republic. M. Karásek is with the Institute of Radio Engineering and Electronics, Academy of Sciences of the Czech Republic, Prague, Czech Republic. Digital Object Identifier 10.1109/TIM.2003.822008 effects can, to a certain degree, be reversed or controlled using dispersion management techniques. The starting point of managing the chromatic dispersion is its accurate and reliable measurement. Ideally, the measurement should be fast and simple, over the whole spectral band, and it should offer the possibility to measure already installed long optical fiber links. Subsequently, the chromatic dispersion of each fiber span may be totally or partially compensated using a variety of techniques including, for example, dispersion compensating fiber (DCF) [1] or chirped fiber Bragg gratings (FBGs) [2]. There are three principal methods to measure chromatic dispersion: phase-shift, time-of-flight, and interferometric methods [3]. The recommended test method of the International Telecommunications Union (ITU) is the phase-shift technique. In the so-called modulated phase-shift implementation of the phase-shift method, the input signal is amplitude modulated and launched into the test fiber. At the output, the phase of the transmitted signal envelope is compared with a reference in order to get the group delay. This measurement is repeated while changing the wavelength of the signal. Chromatic dispersion is obtained by differentiating the group delay with respect to the wavelength. It is possible to obtain the dispersion data directly by using the differential phase-shift technique. In this implementation, both the amplitude and the frequency of the laser source are modulated, which makes it possible to obtain the chromatic dispersion immediately [4]. The laser wavelength still has to be scanned to determine the chromatic dispersion over the entire spectral band. Phase-shift methods allow precise measurement of long, as well as short fiber links, or even the measurement of components such as FBGs. The accuracy of this method is typically 0.02 ps nm km and the resolution in the group delay of the order of 0.1 ps. Implementation of this technique requires the transmission of a reference signal, either a modulated optical channel or a radio frequency (RF) signal, to the analyzer located at the output of the fiber. The wavelength scanning over the whole C-band typically takes up to 30 min. Interferometric methods are one of two alternative techniques for the measurement of chromatic dispersion that are recognized by the ITU. The implementation of the interferometric technique is usually performed using an interferometer, either a Michelson [5] or, more often, a Mach Zehnder [3], and a wide-band source. One arm of the interferometer consists of the fiber under test and the other one of a reference fiber with a well-known group delay and/or an air gap with adjustable length. The measured cross-correlation signal depends on the time delay difference between the test and the reference arm. 0018-9456/04$20.00 2004 IEEE

68 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 53, NO. 1, FEBRUARY 2004 Knowing the group delay in the reference arm, the group delay of the measured optical link and consequently the chromatic dispersion can be determined. Due to the short optical length of the reference arms, this technique is suitable for short samples of the fiber under test. This method is very accurate, with the group delay resolution being as small as 0.1 ps. As with the phase-shift method, interferometric techniques also require the transmission of a reference signal. The other alternative test method recognized by the ITU is the pulse delay technique, also called time-of-flight. In this method, short pulses at different wavelengths are simultaneously launched into the fiber under test. Due to the chromatic dispersion, the pulses propagate with different group delay velocities. At the optical link output, the signal is captured with a fast photodetector and processed using a sampling oscilloscope. The time delay between the consecutive pulses, corresponding to the different wavelengths, is then determined from the recorded trace. Subsequently, the group delays are differentiated with respect to wavelength in order to obtain the chromatic dispersion spectral data. This method can only measure relatively large values of cumulative chromatic dispersion, typically corresponding to at least several kilometers of standard fiber, as the delay between the consecutive pulses needs to be measurable. In this work, we found that this delay can be measured with sufficient accuracy if the separation of the pulses is at least twice their FWHM. Advantages of this method are its speed, as no kind of scanning is used, and the fact that no reference is needed between the input and the output of the measured link. This method therefore appears particularly well suited for measurement of already installed fiber links. Previous implementation of this method can be found in [3], [6]. In [3], where a Raman laser source was used, the observed timing jitter and optical triggering instabilities degraded the accuracy of the measurement. In [6], a superluminescent diode filtered by a set of spectrally separated FBGs was used. The low power efficiency of this approach was compensated for by using very sensitive photon-counting detectors cooled to 220 K. In this paper, we propose to use a newly developed multiwavelength Erbium doped fiber laser [7] to perform chromatic dispersion measurement using the time-of-flight method. Usually, in a predominantly homogenously broadened active medium, such as in Erbium doped fiber (EDF), laser operation can occur only at the wavelength exhibiting the highest net gain [8]. Recently, a new type of EDF laser, which suppresses the cross-gain saturation effect, was proposed [7]. This laser is characterized by the use of a frequency shifter placed in the laser ring cavity, which modifies the laser build-up mechanism, and of the FBG filter with multiple pass bands, which selects the laser lines. In order to get laser lines with equal power, it is nonetheless necessary to equalize the net cavity gain with rather high precision. For example, in [9] it was observed that variations as small as 0.4 db in the net gain lead to variations in the output power among the laser lines as large as 10 db. In Section II of this paper, we present the realization of the fiber laser. The source is designed to achieve cw operation of 15 to 20 laser lines covering at least the C-band and equally spaced by 2 3 nm. The laser is optimized to reach a uniform power distribution among all the emitting wavelengths. Output power flatness better than 6 db is achieved using a new technique based on the inscription of the FBG array during the operation of the laser. The total laser power is greater than 0 dbm. A Mach Zehneder modulator is then inserted at the laser output to produce low-duty cycle short pulses from this multiwavelength cw source. Results of chromatic dispersion measurement based on the time-of-flight technique are presented in Section III. Measurements performed on standard, dispersion shifted, and dispersion compensating fibers are compared with the results obtained with the phase-shift technique. II. MULTIWAVELENGTH LASER REALIZATION The multiwavelength fiber laser setup is shown in Fig. 1. A 14-m-long EDF (HP800, Lucent Technologies) is pumped by a 120-mW laser diode at 980 mw through a 980/1550-nm WDM coupler. We introduced isolators inside the cavity to induce unidirectional propagation. The light is frequency down-shifted at each round trip by 80 MHz using a fiber coupled acoustooptic device (AOFS) (Gooch and Housego). We also use a variable attenuator to control the total cavity loss. A filter with multiple pass bands is inserted into the cavity to determine the lasing wavelengths. The filter is a succession of FBGs written in a single piece of a photosensitive fiber with Sagnac interferometric writing set-up [10]. It was previously demonstrated that this type of laser emits radiation in different regimes: cw, selfpulsed and mode-locked, depending on the pump power and the cavity configuration [11], [12]. The cw regime can usually be favored by increasing the cavity loss and the number of simultaneously lasing wavelengths to prevent the onset of nonlinear effects in the cavity. In this work, we operated the laser exclusively in the cw regime. To obtain a uniform output spectrum, it is necessary to spectrally flatten the net gain. Among the proposed methods, it was shown in [13] that adjusting the EDF length and ring losses might lead to a very flat EDF gain in the limited spectral region of 1543 1562 nm. In this work, the best results were obtained with a Fabry-Perot etalon as the multiple pass band filter to select the wavelength bands [7], [9]. In [9], a gain-flattening filter was used to equalize the EDF gain over the entire C-band. However, the transmission loss ripples of the gain flattening filter of 0.2 db resulted in an output spectrum flatness of only 10 db. Here, we propose another method of equalizing the net laser gain using an array of FBGs. The reflectivity of each individual grating is adjusted to compensate for the uneven EDF gain in the C-band. Furthermore, while performing the EDF gain equalization, this filter simultaneously selects the laser lines. It, therefore, leads to quite simple laser configuration. Exact compensation of the EDF uneven gain poses a challenge as its spectral shape depends on the EDF population inversion [14]. In a ring laser, the EDF saturation depends on the laser feedback, which varies with the parameters of the set of FBGs. To get the output power spectrum as even as possible, we used the following procedure. First, we inscribed a set of 18 FBGs with equal strength (transmission loss of 16 db) and with 3-dB FWHM of 25 GHz. Each FBG was 20-mm long and the longitudinal spacing between the consecutive FBGs was 1.5 mm. The spectral separation between the reflection peaks of the neighboring FBGs was 2.2 nm. After the writing

MARAN et al.: CHROMATIC DISPERSION MEASUREMENT 69 Fig. 2. Measured laser output spectrum observed with an OSA with resolution of 0.5 nm. Fig. 1. Multiwavelength frequency-shifted feedback fiber laser setup. of the FBGs, we turned the laser on. As expected, the laser emitted light predominantly over a few laser lines, as the spectral variations of the net gain were too high. Then, we decreased the reflectivity of the FBGs corresponding to the most powerful laser lines, while the laser was in operation, in order to get an even distribution of the laser energy over all the lasing lines. The reduction of a FBG reflectivity was realized by uniform UV exposure of the FBG. It was an iterative process using 50 mw of UV power. Each exposure lasted tens of millisecond or more depending of the grating strength. This increases the average refractive index of the fiber core and, at the same time, shallows the FBGs index modulation depth, as the photoinduced refractive index change gradually saturates. The laser output was monitored during this intra-cavity FBGs adjustment using an optical spectrum analyzer (OSA). Adjusting the reflectivity of the FBGs, however, also changed the laser feedback. These variations in the cavity loss were compensated for by the variable attenuator. Slight modifications of the output spectrum were also possible using the polarization controller, since the cavity had polarization dependent losses (PDL) originating from the PDL of individual cavity components. Fig. 2 shows the final laser output spectrum measured by OSA with a resolution of 0.5 nm. We were able to equalize 17 out of 18 laser lines. The difference in the EDF gain spectrum between the laser line at 1564 nm and the other lines was too large to properly amplify this laser line. The realized laser source had therefore 17 lines with an optical signal to noise ratio of 45 db. Each line had an output power of approximately 12 dbm and the total output power of the laser was 0.4 dbm. Using a fast photodiode (bandwidth of 25 GHz) and a RF spectrum analyzer, the 3-dB bandwidth of the RF spectrum of the entire laser was measured to be 2.5 3 GHz. The power spectral flatness, i.e., the difference between the weakest and the strongest laser lines, was 6 db. Fig. 3. Experimental setup for chromatic dispersion measurement. III. CHROMATIC DISPERSION MEASUREMENT The experimental setup for chromatic dispersion measurement is shown in Fig. 3. The light from the multiwavelength laser is modulated by a Mach Zehnder element driven by a pulse generator producing optical pulses with FWHM of 83 ps with a repetition rate of 93.75 MHz. The polarization controller is used to optimize the modulator throughput and extinction ratio. The modulated signal is launched into the test fiber and the transmitted signal is acquired with a digital sampling oscilloscope with an optical input having a bandwidth of 30 GHz (Agilent, model 83 480A). Fig. 4 presents the temporal response traces measured for different fibers including a standard single mode fiber (Corning, SMF-28), a nonzero dispersion shifted fiber (Corning, LEAF), and a dispersion compensating fiber (DCF) module. After propagation through the test fiber, the pulses at the different wavelengths are separated in time and we are able to measure the delay between the neighboring pulses (Fig. 4). The traces displayed in Fig. 4 were obtained by averaging the signal over 256 measurements. Note that for the DCF fiber, the order of the wavelengths are reversed due to the negative dispersion. The lengths of the fiber samples were as follows: 10 km of SMF, 40 km of LEAF, and 14 km of DCF. To determine the arrival time of the individual pulses with different wavelengths, the temporal trace of Fig. 4 was fitted with a set of 17 Gaussian functions. Knowing the laser wavelengths, we thus obtain the variation of group delay as a function of the wavelength. We fitted this data with second order and fourth order polynomial functions. We did not note any differences between the two fitting functions. Thus, we used a second order polynomial fitting function and differentiated this

70 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 53, NO. 1, FEBRUARY 2004 (a) (a) (b) (b) (c) Fig. 4. Measured temporal responses of (a) SMF, (b) LEAF, and (c) DCF fibers to the input multiwavelength pulse. function with respect to wavelength in order to obtain the chromatic dispersion. We repeated the acquisition 10 times for each fiber sample to evaluate the measurement repeatability. We also measured the chromatic dispersion of the fiber samples using the phase-shift method implemented with a lightwave component analyzer (HP-8703, Hewlet Packard). The results from both methods are shown in Fig. 5. The dispersion slope determined by the time-of-flight method was ps nm km, ps nm km, and ps nm km for the SMF, LEAF and DCF module, respectively. With the HP-8703 analyzer, we found values of ps nm km, ps nm km, and ps nm km for the SMF, LEAF and DCF module, respectively. The above errors on the dispersion slope were determined as standard deviations. IV. DISCUSSION The repeatability of the measurement of the dispersion slope was 0.5% for SMF, 0.1% for LEAF, and 0.003% for DCF module. The variations in the precision of this measurement are related to the difference in the group delay experienced by successive pulses corresponding to neighboring wavelengths. The larger the temporal separation of the pulses for a given fiber length, the smaller is the impact of the error on the measured temporal position on the resulting chromatic dispersion. We found that the difference between the maximum of the measured pulses and the central positions of their Gaussian fits was smaller than 10 ps. The temporal separation of the pulses is influenced by the chromatic dispersion, as well as by the actual (c) Fig. 5. Chromatic dispersion derived from data shown in Fig. 4: (a) 10 km of SMF, (b) 40 km of LEAF, and (c) DCF module. The symbols and the solid lines show the results obtained with the time-of-flight method and the phase shift method, respectively. The dotted line represents the difference between the two measurements. length of the measured sample. Therefore, the use of longer samples improves the precision of the measurement. On the other hand, there is a minimum fiber length that can be measured with acceptable precision since the temporal separation of the pulses must be sufficient to enable the determination of their arrival time. Experimentally, we found that we need more than 10 km of the SMF and more than 30 km of the LEAF fiber to get sufficient temporal separation between the neighboring peaks (twice the measured FWHM of the peaks). Comparison of the chromatic dispersion measurements obtained with the time-of-flight and the phase-shift methods shows an agreement better than 1% as indicated on Fig. 5. The sign of the dispersion is not a limiting factor for the presented measurement technique, which we demonstrated by measuring the dispersion of the DCF module. However, we would not be able to evaluate the chromatic dispersion of a fiber with a zero dispersion wavelength located in the C-band, i.e. dispersion-shifted fiber (DSF). This is due to the fact that we cannot determine the

MARAN et al.: CHROMATIC DISPERSION MEASUREMENT 71 order in which the pulses at different wavelengths arrive at the detector. This method is also not adequate for measuring the chromatic dispersion of optical fiber components, because their chromatic dispersion is usually relatively small and does not result in sufficient temporal separation of the pulses. However, this technique is well adapted to chromatic dispersion measurements in installed long optical fiber link. Due to its inherent fast speed, the measurement is independent of temperature drift, because the temperature variations are slow compared to the speed of the light. V. CONCLUSION We proposed and successfully demonstrated the application of a newly developed multiwavelength fiber laser for chromatic dispersion measurement using the time-of-flight technique. This measurement system was tested by measuring the chromatic dispersion of three different optical fibers: SMF and LEAF Corning fibers, and a DCF module. The experimental results were in agreement with those obtained by the standard phase-shift method using a commercial component analyzer. With the input pulse width of 83 ps, the current implementation of this time-of-flight technique could precisely resolve differences in the group delay between the adjacent peaks of about 300 ps. The precision in the evaluation of pulse arrival time was 10 ps, and the repeatability of the chromatic dispersion measurement was better than 0.5%. The experimental results are in agreement with the measurement performed with the standard phase-shift technique within 1%. Since the time-of-flight method typically requires very stable sources, these results prove that the stability of the laser is adequate for characterization applications. The proposed measuring system seems particularly suitable for fast chromatic dispersion measurement of long and already installed fiber links. [9] S. K. Kim, M. J. Chu, and J. H. Lee, Wideband multi-wavelength erbium-doped fiber ring laser with frequency shifted feedback, Opt. Commun., vol. 190, pp. 291 302, 2001. [10] P.-Y. Cortes, H. Fathallah, S. LaRochelle, L. A. Rusch, and P. Loiselle, Writing of bragg gratings with wavelength flexibility using a Sagnac type interferometer and application to FFH-CDMA, in European Conf. Optical Communication, Madrid, Spain, 1998, pp. 411 412. [11] J. N. Maran and S. LaRochelle, Temporal characterization for a multifrequency erbium-doped fiber laser with frequency shifted feedback, Proc. SPIE, vol. 4833, pp. 855 861, 2002. [12] H. Sabert and E. Brinkmeyer, Pulse generation in fiber lasers with frequency shifted feedback, J. Lightwave Technol., vol. 12, pp. 1360 1368, 1994. [13] M. Karásek and A. Bellemare, Numerical analysis of multifrequency erbium-doped fiber ring laser employing a periodic filter and a frequency shifter, in Proc. Inst. Elect. Eng., Optoelectronics, vol. 147, 2000, pp. 115 119. [14] E. Desurvire, Erbium-Doped Fiber Amplifier. Principles and Applications. New York: Wiley, 1994. J.-N. Maran was born on May 17, 1975. He received the B.S. degree in engineering optics from the Ecole Nationale Supérieure de Science Appliquée et Technologie (ENSSAT), France. He is currently pursuing the Ph.D. degree in electrical engineering, with a dissertation on multiwavelength erbium-doped fiber lasers, at the Université Laval, Québec, QC, Canada. His present research interests include fiber lasers, modelocked fiber lasers, fiber Bragg gratings, and laser physics. Radan Slavík was born on August 21, 1973, in the Czech Republic. He received the M.A.Sc. and Ph.D. degrees in optics and optoelectronics from the Faculty of Mathematics and Physics, Charles University, Prague, Czech Republic, in 1996 and 2000, respectively. From 1995 to 2000, he was with the Institute of Radio Engineering and Electronics, Czech Academy of Sciences, Prague. In November 2000, he joined the Centre d optique, photonique et lasers, Department of Electrical and Computer Engineering, Université Laval, Québec, QC, Canada, as a Postdoctoral Research Fellow. His research interests focus on all-fiber technologies, including fiber gratings, fiber amplifiers and lasers, and fiber-optic sensors. REFERENCES [1] L. Grüner-Nielsen, S. N. Knudsen, B. Edvold, T. Veng, D. Magnussen, C. C. Larsen, and H. Damsg, Dispersion compensating fibers, Opt. Fiber Technol., vol. 6, pp. 164 180, 2000. [2] B. J. Eggleton, A. Ahuja, P. S. Westbrook, J. A. Rogers, P. Kuo, T. N. Nielsen, and B. Mikkelsen, Integrated tunable fiber gratings for dispersion management in high-bit rate systems, J. Lightwave Technol., vol. 18, pp. 1418 1432, 2000. [3] L. G. Cohen, Comparison of single mode fiber dispersion measurement techniques, J. Lightwave Technol., vol. 5, pp. 958 966, 1985. [4] S. Ryu, Y. Horiuchi, and K. Mochizuki, Novel chromatic dispersion measurement method over continuous gigahertz tuning range, J. Lightwave Technol., vol. 7, pp. 1177 1180, 1989. [5] L. Thevenaz, J. P. Pellaux, and J. P. Von der Weid, All-fiber interferometer for chromatic dispersion measurements, J. Lightwave Technol., vol. 6, pp. 1 7, 1988. [6] H. Riedmatten, M. Wegmüller, H. Zbinden, and N. Gisin, Group delay analysis of chirped fiber Bragg gratings using photon counting, IEEE Photon. Technol. Lett., vol. 13, pp. 615 617, 2001. [7] A. Bellemare, M. Karásek, M. Rochette, S. LaRochelle, and M. Tetu, Room temperature multifrequency erbium-doped fiber lasers anchored on ITU frequency grid, J. Lightwave Technol., vol. 18, pp. 825 831, 2000. [8] M. J. F. Digonnet, Rare-Earth-Doped Fiber Lasers and Amplifiers. New York: Marcel-Dekker, 2001. Sophie LaRochelle received the B.S. degree in engineering physics from the Université Laval, Québec, QC, Canada, in 1987 and the Ph.D. degree in optics from the University of Arizona, Tucson, in 1992. From 1992 to 1996, she was a Research Scientist at the Defense Research Establishment Valcartier, where she worked on electrooptical systems. She is now a Professor with the Department of Electrical and Computer Engineering, Université Laval, where she holds the Canada Research Chair in Optical Fiber Communications and Components. Her current research activities are focused on active and passive fiber optics components for optical communication systems including Bragg gratings, multiwavelength lasers, and amplifiers. Dr. LaRochelle is a member of OSA. Miroslav Karásek received the Ing. degree with honors from the Prague Technical University, Prague, Czechoslovakia, in 1969, the Ph.D. degree in the area of microwave semiconductor devices from the Institute of Radio Engineering and Electronics (IREE), Czechoslovak Academy of Sciences, Prague, in 1974, and the Doctor of Sciences degree in 1990. He is currently a Senior Researcher at the IREE. His research activities are in the area of measurement and computer modeling of active fibers for fiber lasers and fiber amplifier applications.