BROAD-BAND rare-earth-doped fiber sources have been

JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 15, NO. 8, AUGUST 1997 1587 Feedback Effects in Erbium-Doped Fiber Amplifier/Source for Open-Loop Fiber-Optic Gyroscope Hee Gap Park, Kyoung Ah Lim, Young-Jun Chin, and Byoung Yoon Kim Abstract We have investigated rotation-dependent feedback effects in an erbium-doped fiber source for an open-loop fiberoptic gyroscope with fiber amplifier/source configuration. Average power and optical spectrum of the source output as well as of the amplified signal output were found to vary depending on the rotation rate, resulting in the distortion of the gyroscope output. The distorted gyroscope output could be quantitatively explained by a theoretical model with the assumption that the source power and the gain variation due to feedback depends primarily on the average feedback signal power. It was found that the distortion in gyroscope output due to rotation-dependent feedback can be minimized when the phase difference is modulated with an amplitude of 2.405 rad where the average feedback signal power becomes independent of the rotation rate. Index Terms Erbium-doped fiber, gyroscope, optical feedback, optical fiber, optical fiber amplifier. I. INTRODUCTION BROAD-BAND rare-earth-doped fiber sources have been considered as the sources of choice for high-performance fiber-optic gyroscope, [1] [6] owing to their excellent spectral stability as well as the high output power. Both the Nd-doped fiber sources at 1.06 m wavelength [1], [2] and the Er-doped fiber sources around 1.55 m wavelength [3] [6] have been extensively studied. Various source configurations with doped fibers have also been investigated for higher output power, broader bandwidth and more stable mean wavelength [5], [7]. Among them, fiber amplifier/source (FAS) configuration [7] is worthy of note. In this configuration, the doped fiber acts not only as a source but also as an optical amplifier for the returning gyroscope signal, resulting in two to three orders of magnitude increase in detected optical power. With high detected power, it becomes trivial to overide electronic noise. This FAS configuration, however, has an apparent disadvantage in that the source characteristics are susceptible to feedback from the gyroscope since the source is not isolated. If the feedback is time-varying or dependent on the rotation rate, it may lead to a scale factor error or instability [8]. The Manuscript received November 25, 1996; revised March 28, 1997. This work was supported by Center for Electro-Optics and Opto-Electronics Research Center at KAIST in Korea. H. G. Park, K. A. Lim, and Y.-J. Chin are with the Department of Physics, Chonbuk National University, Dukjin-dong, Dukjin-gu, Chonju 561-756 Korea. B. Y. Kim is with the Department of Physics and Center for Electro-Optics, Korea Advanced Institute of Science and Technology, Taejon 305-701 Korea. Publisher Item Identifier S 0733-8724(97)05916-1. fiber gyroscopes with semiconductor sources also have the same problem, but it is much less serious since the coupling between the source and the fiber is much smaller than in the FAS gyroscope. In this paper, we report, for the first time to our knowledge, the experimental analyses of the feedback effects on the optical power, spectrum and scale factor of the gyroscope with an erbium-doped fiber amplifier/source as a function of rotation rate. We have also found the proper phase modulation amplitude for minimum feedback sensitivity of the FAS configuration gyroscope. In most open-loop fiber optic gyroscopes, phase modulation is used to dynamically bias the gyroscope. When a sinusoidal waveform is used, the gyroscope output signal contains frequency components at the modulation frequency and its harmonics along with a quasi-dc component that also depends on the rotation rate. If a portion of the gyroscope output signal is fedback into the source, this time-varying and rotation-dependent optical feedback causes changes in source characteristics. In case of an erbium-doped fiber source, however, the dynamic gain response to the high frequency feedback is substantially suppressed owing to the slow response of the erbium-doped fiber gain [9]. It has also been reported that the gain response to ac feedback begins to decrease at around 100 Hz and falls down to around 25 db at 100 khz with reference to the dc feedback response [5]. Therefore, the quasi-dc component of the feedback signal is the major source that affects the spectral gain of the erbiumdoped fiber. If the gyroscope is operated such that the quasi-dc component of the feedback signal is kept constant for varying rotation rate, as in the case of closed-loop operation, the source spectral gain will remain unaffected. We point out that, for an open-loop operation, a stable scale factor can be achieved independently of rotation rate by keeping the average feedback signal power constant. II. THEORY Suppose a fiber gyroscope that incorporates a Sagnac loop where the phase difference between counter-propagating waves is modulated at a frequency with an amplitude. Then the phase difference between counterpropagating waves,, is expressed as (1) 0733 8724/97$10.00 1997 IEEE

1588 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 15, NO. 8, AUGUST 1997 Here is the phase shift due to rotation and is given by where is the optical scale factor and is the rotation rate. When an optical wave from a source with power is launched into the gyroscope, the optical power returning to the source is where is the transit time of the light in the gyroscope, and is the optical loss factor through the gyroscope. If we ignore the ac feedback response of the source as an approximation, we can set that is independent of time. Then the feedback power, which is proportional to the gyroscope output power in the conventional gyroscope configuration, can be expanded in a series as follows: (2) Addition of the rotation-dependent portion in a new source output power as will result where is the change in source output power due to rotation-dependent portion of the average feedback power. In the region where the variation of is small compared with, we can approximate as where is a proportionality constant. Then the new feedback power becomes The feedback power and the source output power under the feedback can be obtained by the iterative steps. As the first approximation, the feedback power can be calculated by substituting (5) into (6), and we can obtain the frequency component as (5) (6) where is the first kind Bessel function of order. The quasidc component of the feedback power, that is, the feedback power time-averaged over cycles of the phase modulation signal, becomes where the bracket denotes the time average. When the phase modulation period is much shorter than the response time of the gain medium, will be the major cause of change in the source characteristics. As expressed in (4), the average feedback power consists of a constant term and a rotation-dependent term. Hence, the spectral gain of erbium-doped fiber can vary as a function of rotation rate. However, if the modulation amplitude of the phase difference is properly chosen so that it satisfies, the average feedback power becomes independent of the rotation rate. For instance, if we take rad, the output power, the gain, and the spectrum of the erbium-doped fiber amplifier/source become insensitive to rotation rate, resulting in constant scale factor. To see it more quantitatively, we follow the steps similar to the treatments of Moeller [8] and Blake [10]. In their treatment, the change in source power due to feedback was assumed to be proportional to the instantaneous feedback power. Their results can be directly applicable to diode sources. In case of an erbium-doped fiber source, however, the average feedback, not the instantaneous feedback, is responsible for the change in source power. In addition, the source power change is quite nonlinear to the average feedback power owing to the high coupling efficiency between the source and the gyroscope. Henceforth, we take as the source output power under the constant average feedback that is not rotation dependent. (3) (4) As shown in (7), the frequency component of the feedback power, or the gyroscope output in the conventional configuration, contains a term that is proportional to which is a deviation from an ideal gyroscope response resulting in a scale factor error. This term can be removed by making. For fiber sources in a low loss fiber gyroscope, adjusting to make may be the only option to eliminate the feedback sensitivity. Next, the feedback signal is amplified with a gain through the erbium-doped fiber and directed to a photodetector. Since the erbium-doped fiber gain is assumed to be saturated on a time-averaged power basis, the gain is a function of the average feedback power that varies with the rotation rate unless. Thus, the rotation-rate dependent gain variation may lead to an additional scale factor error except the case of. The frequency component of the amplified output, can be expressed as where is the amplifier gain as a function of the average feedback power. The analytic expression for the saturated gain cannot be expressed simply since many other parameters should be involved. For simplicity, however, we used an empirical formula as below as in the case of a lumped amplifier where is the unsaturated gain and is the saturation input power of the amplifier. (7) (8) (9)

PARK et al.: FEEDBACK EFFECTS IN EDFAS 1589 Fig. 1. Experimental setup for FAS gyroscope configuration. III. EXPERIMENTAL The schematic diagram of an experimental setup is shown in Fig. 1. The testbed gyroscope consisted of a 3.4 km-long single-mode fiber loop and a 35 m-long erbium-doped fiber to form a FAS configuration. The erbium-doped fiber used in the experiment had an erbium concentration of around 300 ppm with Al 2 O 3 codoping, the cutoff wavelength of 1.02 m, core diameter of 3.0 m, and core-clad index difference of 0.023. Pump light at 1.48 m wavelength from a laser diode was coupled into the erbium-doped fiber via a wavelength division multiplexer (WDM). In this configuration, backward amplified spontaneous emission (ASE) from the erbium-doped fiber was the light source of the gyroscope. A piezoelectric transducer (PZT) phase modulator was placed near the loop coupler inside the fiber loop. The phase modulation frequency was 25 khz, which was not high enough for the total suppression of the source response to timevarying feedback, but was sufficient for the demonstration of the principle. The returning gyroscope signal was amplified while transmitting through the erbium-doped fiber, and was detected using an InGaAs photodiode. Another WDM was used to filter out the residual pump light. The detected signal at the applied phase modulation frequency was measured by using a lock-in amplifier. A 10% tapping coupler was placed between the erbium-doped fiber source and the gyroscope to measure the source output power and spectrum at the monitor port 1, as well as the feedback signal and its average power at the monitor port 2. All the unused fiber ends were anglepolished to eliminate an additional feedback due to reflection. In the following description, both the source output power and the feedback power are the values at a point between the pump WDM and the 10% coupler, as shown in Fig. 1. Overall loss of the gyroscope was 9.8 db, which was determined from the power ratio at the monitor port 1 and 2 in the absence of phase modulation. The pump power coupled into the erbium-doped fiber was 15 mw throughout Fig. 2. Average feedback signal power versus rotation rate for various amplitudes of the phase difference modulation, m s. the experiment. The peak optical power as high as 7.4 mw was detected at the monitor port 3 when the phase modulation amplitude was 2.4 rad. Note that not all the detected power carries the rotation rate information, and background ASE can cause an increased noise level. The background ASE level was below 5% of the peak power for the phase modulation amplitudes chosen in the experiment. If the optical loss in the gyroscope increases, then the feedback signal power decreases and the background ASE level increases. The average power of the returning gyroscope signal was measured at the monitor port 2 as a function of rotation rate for various modulation amplitudes. Fig. 2 shows that the measured average feedback power varied significantly depending on the rotation rate except for the modulation amplitude of rad. Source output spectra measured at the monitor port 1 are shown in Fig. 3, for the two rotation rates of 0 and 24 /s that induces. When the modulation amplitude was set

1590 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 15, NO. 8, AUGUST 1997 (a) Fig. 4. Source output power versus rotation rate for various modulation amplitudes, m s. Fig. 3. (b) Source output spectra (linear vertical scale) measured with and without rotation. Rotation rates are 0 and 24 /s. (resolution: 0.2 nm) (a) m =1:8rad and (b) m =2:4rad. at rad, not only the total power but also the mean wavelength changed as a function of rotation rate [Fig. 3(a)]. In contrast, for rad, the output spectrum remained nearly unchanged for the two rotation rates [Fig. 3(b)]. The source output power was measured as a function of rotation rate for different modulation amplitudes, and are shown in Fig. 4. For rad, the output power varied by about 16% over the range of Sagnac phase shift. For rad, the output power remained nearly constant independently of the rotation rate. By comparing Figs. 2 and 4, one can notice that, as the average feedback power increased, source output power decreased due to the gain saturation. This means that is negative. The first harmonic component of the feedback signal, which corresponds to the gyroscope output in case of the conventional configuration, was measured at the monitor port 2 by using a lock-in amplifier. The measured outputs versus rotation rate were shown in Fig. 5 where the outputs for different s were normalized so that they have the same maximum of unity. It clearly shows that the deviation from an ideal sine curve ( ) except for the case of rad. Fig. 5. Normalized first harmonic of the feedback signal versus rotation rate for m =1:8,2:4, and 3:2 rad. The solid line denotes the ideal sine curve sin 1R. The measured deviations of the normalized results from an ideal sine curve were plotted together with calculated ones for and rad in Fig. 6. For the calculation, we used (7) with, which was obtained from the dependence of the source output power on the average feedback power that can be obtained from the data in Figs. 2 and 4. The measured deviations agreed reasonably well with the calculation, showing nearly dependence as predicted in the theory. We should mention that the calculated deviations are not exactly proportional to because the second term in (7) is included in the normalization factor. For rad, we could only observe small deviations within our experimental accuracy. The small deviation in the data are mainly due to the resolution limit in the amplitude setting of the function generator used for the experiment. The spectra of the amplified output were measured at the monitor port 3. The results for the two different rotation

PARK et al.: FEEDBACK EFFECTS IN EDFAS 1591 (a) Fig. 6. Deviations of the normalized first harmonic of the feedback signal from an ideal sine curve for various m s. The solid lines represent the calculated curves for m = 1:8 and 3:2 rad. rates of 0 and 24 /s are shown in Fig. 7. For rad, we can observe only a very small change in spectrum between the two cases, while there is appreciable change for rad. In comparison with the results in Fig. 3, we can observe a line narrowing and a mean wavelength shift to longer wavelength due to the amplification process. The shift of mean wavelength results in the scale factor difference. It should be mentioned that the scale factor of gyroscope with FAS configuration is determined by the optical spectrum of the amplified optical signal, since the scale factor of a gyroscope is dependent on the optical signal reaching the detector [11]. The mean wavelength of the amplified output was measured as a function of rotation rate and were plotted in Fig. 8 for different s. Here, the mean wavelength was calculated from the measured spectral power density by using where is the wavelength. For rad, the mean wavelength remained nearly constant for different values of rotation rate within the spectral resolution of 0.2 nm used in this measurement. For rad, it changed by 0.5 nm over the range of Sagnac phase shift. The modulation frequency component of the amplified output was measured for different modulation amplitudes. Their normalized results versus rotation rate were shown in Fig. 9, showing sizable deviation from an ideal sine curve except for rad. The deviations from an ideal sine curve were plotted in Fig. 10. Compared with the results in Fig. 6, the deviations increased due to the amplification. The calculated deviations for and rad were also plotted. For the calculation, we used (7) (9) with the data in Fig. 2 for. To obtain and in (9), we measured the fiber amplifier gain as a function of input power in a separate experiment. For the gain measurement, light from an 1551 nm DFB laser diode was used as an input and was launched in the same direction as the feedback signal. By fitting the measured gain into (9), we obtained and W. Fig. 7. (b) Spectra of the amplified output (linear vertical scale) measured with and without rotation. Rotation rates are 0 and 24 /s (resolution: 0.2 nm) (a) m =1:8rad and (b) m =2:4rad. Fig. 8. Mean wavelength of the amplified output versus rotation rate for various m s. The calculated curves in Fig. 10 matched approximately with the experimental data, which shows that the saturated gain

1592 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 15, NO. 8, AUGUST 1997 the optical scale factor error due to the mean wavelength change should be included for a more accurate predictions. Actually, about 1 2% of the cw output from the source was found to be modulated owing to the ac feedback. Residual ac feedback effects can be further suppressed by using higher frequency modulation since the erbium-doped fiber gain response to feedback rolls off at approximately 10 db/decade [5]. Fig. 9. Normalized FAS gyroscope output versus rotation rate for m =1:8, 2:4, and 3:2 rad. The solid line denotes the ideal sine curve sin 1R. IV. CONCLUSION We have investigated the rotation-dependent feedback effects in erbium-doped fiber amplifier/source and the distortion in gyroscope output owing to the feedback effects. The output power and the mean wavelength were found to change depending on the rotation rate, which caused the distortion in gyroscope output. It was possible to explain quantitatively the distorted gyroscope output by theoretical treatment with assumption that the gain variation primarily depends on the average of the feedback signal power. It was also found that the distortion in gyroscope output due to rotation-dependent feedback can be minimized for the phase modulation amplitude that makes where the average feedback signal power is kept constant independently of the rotation rate. The performance of the FAS gyroscope can also be improved by optimizing gyroscope parameters such as doped fiber length, pump power, and signal processing scheme, which are under investigation. We believe that the erbium-doped fiber amplifier/source has a potential to be a stable source for open-loop gyroscopes with high-frequency modulation and properly chosen amplitude. ACKNOWLEDGMENT The authors would like to thank D. H. Lee in Chungnam National University for helpful discussions. They would also like to thank M. J. Chu and C. H. Lee in ETRI for providing a pump laser diode. Fig. 10. Deviations of the normalized FAS gyroscope output from an ideal sine curve for various m s. The solid lines represent the calculated curves for m = 1:8and 3:2 rad. behavior depending on the average feedback power plays the major role in the distortion of the gyroscope output. In the same experimental setup, we measured the normalized gyroscope output for various s with the phase modulation at 30 khz which is the proper frequency of the gyroscope. Even at the proper frequency operation, the gyroscope output showed almost the same deviation as in case of the 25 khz operation. We could not see the recognizable difference in the distortion between 25 and 30 khz within our experimental accuracy. This is in contrast with the fact that a proper frequency modulation can suppress the distortion in the first harmonic output when the source response to feedback is instantaneous as in the semiconductor sources [10]. In our theoretical treatment, only the source power and gain variation due to the quasi-dc feedback was considered, as they were dominant over the other possible causes in our experiment. However, the remaining ac feedback effects and REFERENCES [1] K. Liu, M. Digonnet, H. J. Shaw, B. J. Ainsile, and S. P. Craig, 10 mw superfluorescent single-mode fiber source at 1060 nm, Electron. Lett., vol. 23, pp. 1320 1321, 1987. [2] I. N. Duling, W. K. Burns, and L. Goldberg, High power superfluorescent fiber source, Opt. Lett., vol. 15, pp. 33 35, 1990. [3] P. R. Morkel, Erbium-doped fiber superfluorescent source for the fiber gyroscope, in Springer Proc. Physics, Optical Fiber Sensors, Paris, France, 1989, vol. 44. [4] K. Iwatsuki, Erbium-doped superfluorescent fiber laser pumped by 1.48 m laser diode, IEEE Photon. Technol. Lett., vol. 2, pp. 237 239, 1990. [5] P. F. Wysocki, M. J. F. Digonnet, B. Y. Kim, and H. J. Shaw, Characteristics of erbium-doped superfluorescent fiber sources for interferometric sensor applications, J. Lightwave Technol., vol. 12, pp. 550 567, 1994. [6] D. C. Hall, W. K. Burns, and R. P. Moeller, High-stability Er 3+ - doped superfluorescent fiber sources, J. Lightwave Technol., vol. 13, pp. 1452 1460, 1995. [7] K. A. Fesler, M. J. F. Digonnet, B. Y. Kim, and H. J. Shaw, Stable fiber-source gyroscope, Opt. Lett., vol. 15, pp. 1321 1323, 1990. [8] R. P. Moeller, W. K. Burns, and N. J. Frigo, Open-loop output and scale factor stability in a fiber-optic gyroscope, J. Lightwave Technol., vol. 7, pp. 262 267, 1989. [9] E. Desurvire, C. R. Giles, and J. R. Simpson, Gain saturation effects in high-speed, multichannel erbium-doped fiber amplifiers at =1:53 m, J. Lightwave Technol., vol. 7, pp. 2095 2104, 1989. [10] J. Blake and S. Navidi, Second harmonic detection for rotation sensing in fiber-optic gyros, Proc. SPIE, vol. 1795, pp. 69 73, 1992.

PARK et al.: FEEDBACK EFFECTS IN EDFAS 1593 [11] H. Lefevre, The Fiber-Optic Gyroscope. Boston, MA: Artech House, 1993, pp. 130 131. Hee Gap Park was born in Seoul, Korea, on October 13, 1955. He received the B.S. degree in physics from Seoul National University, Seoul, Korea, in 1978, and the M.S. and Ph.D. degrees in physics from Korea Advanced Institute of Science and Technology, Taejon, Korea, in 1980 and 1987, respectively. His thesis research involved the nonlinear fiber optics. In 1980, he joined Electronics and Telecommunications Research Institute, Taejon, Korea, where he was involved in optical communication systems research and development. He has also worked on nonlinear fiber switch and grating filter based on two-mode fibers at Ginzton Laboratory, Stanford University, Stanford, CA, as a Visiting Scholar from 1987 to 1988. Since 1991, he has been with Chonbuk National University, Chonju, Korea, and is currently an Associate Professor of Department of Physics. His research interests include fiber devices using active fibers, sources for gyroscope, and mode-locked fiber lasers. Kyoung Ah Lim received the M.S. degree in physics from Chonbuk National University, Chonju, Korea, in 1996. Since 1996, she has been with the Center for Dual Technology Laboratory at the Institute for Advanced Engineering, Yougin, Korea. Her current research area is fiber-optic gyroscope. Her research interests includes optical amplifier and fiber sources for optical fiber sensors. Young-Jun Chin, photograph and biography not available at the time of publication. Byoung Yoon Kim, photograph and biography not available at the time of publication.