1. Introduction

Brillouin optical time-domain analyzer (BOTDA) sensors have attracted increasing interest in recent years for their ability to monitor temperature and (or) strain distribution over long distance as well as applications in large civil engineering [1–7]. In BOTDA systems, a continuous-wave (CW) probe signal interacts with a counter-propagating optical pump pulse through an acoustic wave in the fiber due to stimulated Brillouin scattering (SBS), leading to power transfer between the two optical signals. The CW probe light is locally amplified or reduced via SBS interaction when the frequency offset of these two optical beams is within the fiber Brillouin gain spectrum (BGS). The maximum SBS interaction occurs when the frequency difference equals to the Brillouin frequency shift (BFS) of the fiber. Exploiting the dependence of the BFS parameter on strain and temperature, BOTDA could realize accurately distributed measurements along the fiber under test (FUT) [3, 4].

The sensing length of BOTDA is mainly limited by the maximum allowed input pump power before onset of nonlinear effects in fiber and the nonlocal effects induced by the pump depletion. These effects would distort the measured BGS, which finally lead to the BFS errors, and thus the temperature and strain measurement errors [8, 9]. Various techniques have been proposed to extend the dynamic range of BOTDA sensors [10–16]. For example, the method using double sideband (DSB) probe beam consisting of the Stokes and the anti-Stokes component has been proposed and demonstrated to overcome the nonlocal effects [10]. The self-heterodyne detection has been employed to enhance the signal-to-noise ratio (SNR) of BOTDA [11]. In particular, the sensing lengths of BOTDA systems have been remarkably enhanced by the use of pulse coding combined optical pre-amplification [12] and distributed Raman amplification [13–15].

However, the pulse coding scheme increases the complexity of subsequent data processing and thus increases the measurement time. The schemes using Raman amplification may involve in the relative intensity noise (RIN) transferring from the Raman pumps to the CW Brillouin probe light, which would result in large fluctuation on the BOTDA trace and thus large measurement error [13]. Additionally, the optical amplification methods require high Raman powers and a more complex interrogation system with respect to conventional BOTDA systems. Although BOTDA sensors can provide one of the most attractive schemes, allowing for high-performance sensing over long fiber ranges, a fatal drawback in BOTDA is the requisite of access to both fiber-ends. This implies that the BOTDA doesn’t work when a break occurs along the FUT. Even worse, the location of the break cannot be traced by the system itself in this case.

In this paper, for the first time, we have proposed and experimentally demonstrate a BOTDA with the capacity of break interrogation. In this sensor system, the coherent detection technique is applied to improve the system’s SNR, and the DSB method is used to mitigate nonlocal effects. Without amplification, the sensor achieves a 72 km sensing range with 5 m resolution and an estimated temperature uncertainty of 1.8 °C. When the CW probe input-end of the FUT is disconnected, the sensor system is capable of locating the break along the FUT as coherent optical time domain reflectometry (COTDR), and achieves a peak dynamic range of 36 dB with 100 m spatial resolution.

2. Experiment setup and operation

The experiment setup of the proposed optical fiber sensor is shown in Fig. 1 . A tunable laser source (Agilent N7714A) with the linewidth of ~100 kHz and maximum power of 16dBm operates at 1549.960 nm. The output of the laser is divided into two branches by a 90:10 polarization-maintaining coupler (PMC1). The lower branch, with 90 percent of the signal, is modulated by a pulse generator to generate pump pulses via an electro-optic modulator (EOM1). The generated pump pulses have a high extinction ratio of 40 dB. Then the pump pulses are amplified at a required optical power level by an Erbium doped fiber amplifier (EDFA). Before the pump pulse is directed into the FUT through a circulator, a polarization scrambler is employed to compensate the polarization sensitivity of SBS.

 

Fig. 1 The experiment setup of BOTDA with break interrogation based on coherent detection. PMC: polarization-maintaining coupler, EOM: electro-optic modulator, PS: polarization scrambler, EDFA: Erbium-doped fiber amplifier, ISO: optical isolator FUT: fiber under test, OLO: optical local oscillator, ESA: electrical spectrum analyzer, AOM: acousto-optic modulator, PD: photodetector.

Download Full Size | PDF

The branch with 10 percent of the light is again separated into two branches by a 3-dB polarization-maintaining coupler (PMC2). In the lower branch, the CW light from PMC2 is used as optical local oscillator (OLO) to perform self-heterodyne coherent detection. The frequency of OLO light is up-shift 80 MHz by an acousto-optic modulator (AOM). In the upper branch, two probe sidebands (namely Stokes line and anti-Stokes line) around the laser wavelength are generated by the electro-optic modulator (EOM2), operating in the suppressed carrier regime, driven by a microwave generator. High carrier suppression is achieved by properly adjusting the bias of EOM2. Then the two sidebands are launched into the FUT. An isolator is used between the EOM2 and the FUT to avoid the interference of pump pulses with the EOM2. After propagating through the FUT, the probe signals passing through the circulator are directed to a 3-dB coupler to mix with the OLO light, and then are detected by a 12 GHz photodetector (PD). An electrical spectrum analyzer (ESA, Agilent E4440A) is employed to observe the beat frequency signals. The ESA in zero-span mode is used to acquire time-domain traces centered at the desired beat frequency, and the Brillouin gain spectrum can be measured by scanning the frequency offset of the pump pulse and the probe beam. The FUT comprises two sections of standard single-mode silica fiber (SMF) with the length of 48 km and 24 km, which are fusion spliced together. Their BFSs are slightly different, with a difference of ~12 MHz. The nominal loss coefficients of the two sections of fiber are both 0.196 dB/km. The last ~52 m fiber is placed in an oven. Note that, in our configuration, both Stokes and anti-Stokes sidebands could be applied to perform the self-heterodyne detection with the OLO light, which are corresponding to the Brillouin-gain and Brillouin-loss processes in conventional BOTDA, respectively.

Assuming the optical frequency of laser is f₀ and the modulated frequency of EOM2 is f, then the frequencies of the OLO and the two probe sidebands are f₀ + 80 MHz, f₀-f (Stokes line)and f₀ + f (anti-Stokes line), respectively. According to heterodyne theory, the detected alternating photocurrent $i_{\det }$ is [11]:

Since two probe sidebands are used in our experiment, two beat frequency signals output from PD would coexist. When the Stokes probe sideband is to be detected, the frequency of beat signal is:

If the anti-Stokes probe sideband is to be detected, the frequency of beat signal is:

The choice of signals with the frequency f_IF described in Eq. (2) and Eq. (3) corresponds to the Brillouin-gain and Brillouin-loss configurations in conventional BOTDA setup, respectively. It must be pointed out that the frequency difference of the pump pulse and the probe beam is f rather than f_IF.

The energy transfer mechanism of DSB configuration is schematized in Fig. 2(a) , in which the top diagram shows the probe light spectrum and the bottom one shows the pulse pump wave spectrum. When the modulation frequency f approaches to the Brillouin frequency of the FUT, the upper probe sideband transfers its energy to the pump pulse through SBS interaction, and at the same time, the lower probe sideband is amplified by pump pulse. In this way, the power of pump pulse can be continuously compensated by upper probe sideband when propagating through the sensing fiber, which could effectively reduce the nonlocal effects and therefore improves the SNR of Brillouin gain signal [10, 17, 18].

 

Fig. 2 Energy transfer mechanism between pump and probe light (a) and the frequencies of beat signals (b).

Download Full Size | PDF

In conventional BOTDA configurations, the Brillouin-loss based BOTDA gives a larger measurement range than that of Brillouin-gain configuration. However, the exact opposite is true in our case. The Stokes probe sideband is increased as a result of the SBS interaction, resulting in a much stronger signal for detection and thus the Brillouin-gain configuration presents larger measurement range than the Brillouin-loss one. Therefore, the beat signal with the frequency determined by Eq. (2) is chosen for observation and measurement. The beat signals from the PD are illustrated schematically in Fig. 2(b). Instead of using optical filter to remove the unwanted probe sideband, the coherent detection BOTDA can be easily implemented by choosing the corresponding beat signal, which simplifies and stabilizes the optical structure of the system. Another advantage of our configuration is that the sensor has the capacity of break interrogation as COTDR when the break occurs in sensing fiber.

3. Results and discussion

In order to evaluate the performance of the proposed distributed optical fiber sensor. We firstly performed the coherent detection BOTDA in Brillouin-gain configuration. The experiment setup is shown in Fig. 1. The duration of the pump optical pulse is 50 ns, which corresponds to a 5 m spatial resolution. Its peak power limited by the nonlinear effects such as modulation instability is fixed to 20.5 dBm. Two probe sidebands modulated by EOM2 are set at the same optical power level of −20.8 dBm so as to avoid nonlocal effects. The length of the sensing fiber is:72 km, and the last 52 m fiber at the end of FUT is heated to 60 °C by an oven. The ambient temperature is 28.3 °C.

The BOTDA trace obtained with the modulation frequency f = 10.875 GHz after averaging 4000 times and the Brillouin gain spectra at different distance are illustrated in Fig. 3 . From the Fig. 3(a), it can be seen that the BOTDA traces exhibits a linear attenuation, which indicates that the nonlocal effect of pump depletion is negligible in the system [12]. The large drop in power around 48km is caused by the slightly different BFS values (:12 MHz difference) of the two segments fiber used. The Brillouin gain spectra at 10 km, 54 km and 72.17 km are shown in Fig. 3(b). The undistorted BGSs also confirm that nonlocal effects can be ignored in the experiment [12].

 

Fig. 3 The BOTDA trace at f = 10.875GHz (a) and the obtained BGS at 10km, 54km, 72.50km (b).

Download Full Size | PDF

The BGS is reconstructed by sweeping the frequency difference between the Brillouin pump and probe signal over 150 MHz range centered on the expected Brillouin frequency shift of the fiber with steps of 5 MHz. The time-domain traces along the fiber at each beat frequency are collected after 2000 averages. The scanned frequency ranges from 10.800 GHz to 10.950 GHz in our case, and the corresponding frequency of beat signals observed on ESA is from 10.880 to 11.030 GHz. By fitting the measured BGS at each position along the FUT with Lorentzian curves, the BFS along the fiber can be obtained. The measured BFS is plotted in Fig. 4 . The standard deviations of the measured BFS for the two sections sensing fiber (not including the heated 52 m fiber) are ± 1.6 MHz and ± 1.8 MHz, which correspond to the temperature accuracy of ± 1.6 °C and ± 1.8 °C, respectively. The mean values of BFS for the first and the second segments are 10.8748 GHz and 10.8616 GHz, respectively. The difference of the measured mean BFS is about 13.2 MHz, which is in agreement with the expected value (:12 MHz). However, the maximum fluctuation on the BFS curve is up to 5 MHz. It is probably caused by the polarization-induced and phase noise that would appear in the conventional self-heterodyne scheme [11].

 

Fig. 4 The measured Brillouin frequency shift versus the fiber length.

Download Full Size | PDF

We perform the measurement that the last 52 m length fiber is heated to 60 °C by an oven, while keeping the rest of the fiber at ambient temperature (28.3 °C). The measurement length ranges from 60 to 72.5 km and the sampling number of the ESA is set to 4000 points. The three-dimensional BGS and the BFS parameter at range of 60-72.5 km are shown in Fig. 5 . It is visible that the BFS of heated fiber is separated from the rest one. The frequency difference between the heated and the unheated sections is approximately 34.2 MHz, considering the sensitivity of 1.1 MHz/°C in the BFS, this gives us a temperature variation of 31.1 °C, which is in good agreement with the expected temperature difference (31.7 °C). The achieved spatial resolution is about 5 m as shown in Fig. 5(b). Note that the 5 m spatial resolution obtained is slightly worse than the resolution achieved by similar Raman-amplified BOTDA sensors [13–15]. However, we reach a significantly long distance without amplification.

 

Fig. 5 The three-dimensional BGS (a) at the range of 60-72.5 km and the BFS versus distance near the end of FUT(b).

Download Full Size | PDF

To demonstrate that the proposed sensor system has the capacity of break interrogation, we disconnect the probe input-end of the FUT shown in Fig. 1 and perform the COTDR measurement. The pump pulse in the BOTDA is now used as the probe pulse when implementing COTDR measurement and the signals to be detected are the beat signals heterodyned the OLO with the Rayleigh backscatter from the FUT. Obviously, the f_IF of beat signals is equal to 80 MHz in this case. The COTDR trace is plotted in Fig. 6 . The obtained peak dynamic range is more than 36 dB for 1us probe pulse (corresponding to 90 km sensing length with 100 m spatial resolution) after 8192 averages (the number of average is limited by the ESA used). A linear fit of the data gives the value 0.194dB/km for the single way loss coefficient of the fiber, which is in good agreement with the nominal value of 0.196dB/km. The fluctuation of the measured OTDR trace is slightly large, which may be attributed to the coherent Rayleigh noise (CRN), polarization noise and phase noise [19, 20]. The CRN is due to the narrow linewidth laser we used, the polarization noise derives from the mismatching of polarization state between backscattered signals and LO, and the phase noise is because of the variation of phase of each intra Rayleigh scattering cell. As noise obeys statistic theory, the fluctuation could be reduced by increasing measurement and average times [20]. Nevertheless, benefiting from the convenient optical arrangement, the sensor can easily locate the breaks along the fiber link by setting the f_IF to 80 MHz and measuring the COTDR trace.

 

Fig. 6 COTDR measurement result for the links of 72 km with 1μs pulse, corresponding to 100m spatial resolution.

Download Full Size | PDF

4. Conclusion

In conclusion, we have demonstrated a BOTDA with the capacity of break interrogation. Benefiting from the proposed optical structure and the coherent detection technique, we successfully presented a 72 km sensing range with 5 m resolution and with temperature uncertainty of 1.8 °C without any amplification. Although the DSB probe beam was used to reduce the nonlocal effects, the coherent detection BOTDA can be easily implemented by choosing the corresponding beat signal instead of using optical filter to remove the unwanted probe sideband, which can stabilize the optical structure of system. We also showed that the system could be used to fast locate the break in FUT as a COTDR. It has achieved a peak dynamic range of 36 dB with a 100 m spatial resolution when the CW probe input-end of FUT was disconnected, which would considerably enhance the robustness of the sensor system.