1. Introduction

Thanks to the outstanding capability to measure distributed strain and temperature along tens of kilometers range with meter-scale spatial resolution, Brillouin distributed optical fiber sensors have shown great potential in industry fields, including monitoring of electrical power cables, pipelines, and civil infrastructures [1], etc. Despite the unique performance, Brillouin distributed sensors are however intrinsically suffering from the well-known temperature-strain cross-sensitivity issue, which has significantly degraded their reliability in real applications. In order to address this problem, several solutions have been reported, such as discriminating from multiple BFS peaks of a dispersion-shifted fiber [2], measuring both the BFS and Brillouin intensity variations [3–5], exploiting both the BFS and birefringence or both the Brillouin gain spectrum (BGS) and dynamic acoustic grating spectrum in a polarization-maintaining fiber [6, 7], utilize both the Brillouin scattering and fluorescence in an erbium-doped fiber [8], employ hybrid Brillouin-Rayleigh system in SMF [9], and separate from two spatial modes of a few-mode fiber based BOTDR [10], etc. Besides, the use of a multicore fiber (MCF) has also been proposed, but has not been thoroughly demonstrated [11].

In addition, hybrid Raman-Brillouin distributed sensor based on a single sensing fiber has also been proved to be a good scheme for temperature and strain discrimination [12]. However, as the intensity of anti-Stokes spontaneous Raman scattering (SpRS) signal is about 60 dB below the launching pump peak power [13], very high incident pump power is normally used to make sure that the anti-Stokes SpRS signal can be detected. On the contrary, the threshold of stimulated Brillouin scattering (SBS) is actually low, typically 6 dBm in 40 km long pure silica core fiber [14], which results in a constraint in the maximum usable input power level. Consequently, in that system [12], the high pump power required for the measurement of SpRS has dramatically lead to the appearing of nonlinear effects, such as modulation instability, SBS and so on; a second-time measurement with decreased pump power was thus necessary for measuring the BFS. In order to enhance the signal-to-noise ratio (SNR) in these hybrid Raman-Brillouin distributed sensors while keeping the input pump power below the onset of nonlinear effects, Fabry–Pérot laser containing multiple narrowband longitudinal modes [15, 16] and optical pulse coding technique have been proposed to mitigate this limitation [13, 16, 17].

Here, we propose an alternative solution enabled by MCF based spatial-division multiplexed hybrid Raman and Brillouin scattering interrogation configuration for temperature and strain discriminative measurement. Thanks to the separate spatial channels embedded in a single fiber used for ROTDR and BOTDR, the incompatible of distinct input pump threshold can be elegantly addressed, allowing for simultaneous measurement of Raman and Brillouin scattering signal. The temperature information is acquired from the normalization of anti-Stokes to Stokes components of SpRS, and then strain can be determined from BFS with temperature profile extracted by Raman measurement. The proposed system employs a single laser source, generating identical pump pulse for both Raman and Brillouin which ensures the same spatial resolution. Distinct spatial interrogation channels are used to measure SpRS and BFS simultaneously.

2. Working principle

It’s well known that Raman scattering signal is only sensitive to temperature, but not to strain; while Brillouin scattering signal (i.e. BFS) is sensitive to both temperature and strain. Specifically, in the case of a MCF, the BFS in off-center cores is not only temperature and strain dependent, but also bending-sensitive [18]. Employing the parameters in reference 18, i.e. $\alpha =4.88$, ${\nu }_B=10.735GHz$, $d_i=40\text{um}$, additionally let’s assume that the off-center core is located in the bending plane, then 1 m bending radius is calculated to create about 2.1 MHz BFS variation. However the BFS is not sensitive to bending in the central core, because the central core is always located in the neutral axis of the fiber geometry. It’s therefore not preferred to implement the measurement of BFS in outer cores in this MCF based SDM hybrid Raman-Brillouin distributed sensing system, but instead the central core should be chosen to conduct Brillouin distributed sensing, meanwhile the outer core can be employed to perform the measurement of SpRS. In this optimized configuration, the detrimental response of BFS imposed by bending in off-center cores can be completely eliminated, thus ensuring strict quantitative discrimination between just temperature and strain.

In the proposed MCF based SDM hybrid Raman-Brillouin distributed sensing system, the measurement of SpRS and BFS is performed with the optical time-domain reflectometry technique, i.e. RODTR and BODTR, and the measurements are implemented simultaneously.

The intensity of anti-Stokes SpRS signal is temperature dependent, while the Stokes SpRS signal is however not temperature-sensitive. The temperature profile is then normally derived from the anti-Stokes SpRS ($I_{as}(z)$) to Stokes SpRS ($I_s(z)$) light power ratio [19], according to

On the other hand, the BFS is sensitive to both temperature and strain, giving by

The major limitation factor in the SMF based hybrid Raman-Brillouin distributed sensor is given by the noise in low-power SpRS based temperature measurements. A straightforward way is to use high input power, however it may lead to the appearing of nonlinear affects, such as SBS, modulation instability and so on, which will restrict the interrogation of BFS, and degrade the feasibility of the system [12]. In the proposed MCF based SDM system configuration, the interrogation of SpRS and BFS are implemented in different fiber cores, allowing for independent control of the input power for ROTDR and BOTDR, while the two sensing channels share the same pump source, which ensures identical spatial resolution.

3. Experimental setup and results

To evaluate the performance of the proposed MCF based SDM hybrid Raman-Brillouin distributed sensing system, the experimental setup shown in Fig. 1 has been implemented. A single distributed-feedback (DFB) laser with ~1 MHz linewidth working at 1550nm is used in the system; the CW light is then split into two branches through a 3 dB optical coupler. One branch will be served as the optical local oscillator (OLO) in the heterodyne detection for reconstructing Brillouin gain spectrum (BGS), while the other branch is used for generating pump pulse through a high extinction ratio semiconductor optical amplifier (SOA) based pulse carver, which is driven by an electrical pulse generator with 30 ns pulse width, allowing for 3 m spatial resolution. Subsequently, the pulse is boosted by an erbium-doped fiber amplifier (EDFA), and followed by an optical band-pass filter to filter out the amplified spontaneous emission (ASE) noise. The amplified pulse is then split into two paths by a 3 dB optical coupler, and after an optical circulator the pulses are launched into different cores through the MCF fan-in coupler for Raman and Brillouin pump, respectively. It should be pointed out that a variable optical attenuator is assigned before the input port of the BOTDR-path, which allows for flexible control of the input power for Brillouin sensing. The peak optical input powers for the Brillouin and Raman sensing are about 20 dBm and 33 dBm, respectively.

 

Fig. 1 Experimental setup of the MCF based SDM hybrid ROTDR and BOTDR system; LD: Laser diode; PC: polarization controller; SOA: semiconductor optical amplifier; MZM: Mach-Zehnder modulator; EDFA: erbium-doped fiber amplifier; PS: polarization switch; BPF: band-pass filter; BPD: balanced photodetector; Att.: tunable attenuator; APD: avalanche photodiode; ESA: electrical spectrum analyzer; OSc.: oscilloscope;

Download Full Size | PDF

As has been explained, in our experiment the central core is used for BOTDR, and one outer (off-center) core is chosen to implement ROTDR, therefore the receivers are also separated correspondingly. For the ROTDR-path, the backscattered Stokes and anti-Stokes SpRS components are separated by a single mode fiber based Raman filter, and measured by two avalanche photodiodes, which are then connected to an oscilloscope for data acquisition.

On the other hand, light at the OLO-branch is firstly modulated through a Mach-Zehnder modulator driven by microwave generator with a frequency shift ${\nu }_L$that is approximately equal to BFS ${\nu }_B$, in order to down-convert the beat frequency to intermediated frequency (IF), therefore a low speed photodiode could be employed for detection. Additionally, a polarization switch has been placed at the OLO-branch to reduce polarization-fading noise produced in coherent detectors. At the receiver of BOTDR-path, heterodyne detection is employed, where the backscattered Brillouin signal is mixed with the OLO, and a 400 MHz balanced photodiode (BPD) is used for detecting the beating signal. The BPD is then connected to an electrical spectrum analyzer (ESA), which is working in zero-span mode for time trace acquisition, and by scanning the central frequency of ESA, the BGS can be reconstructed. Both the ESA and oscilloscope are controlled by a computer for frequency scanning and/or data acquisition.

The narrow spectrum width of BGS (~25 MHz in silica SMFs) requires the use of narrowband laser diode for accurate BFS measurement, which however will create coherent Rayleigh noise (CRN) at the coherent receiver [15, 16]. In this case, the intermediated frequency should be chosen properly in order to prevent the CNR from distorting the measured BGS [20, 21]. Therefore here we firstly measured the CRN gain spectrum by acquiring the time-domain power trace when scanning the frequency from 1 MHz to 50 MHz with a step of 1 MHz. Note that the beating signal is only from the backscattered light of the sensing fiber and the OLO-branch is disconnected in the measurement of CRN spectrum. Figure 2(a) presents the measured coherent Rayleigh time-domain trace at a specific frequency of 10 MHz, showing typical temporal amplitude fluctuation on the backscattered signal. Figure 2(b) depicts the measured CRN spectrum by scanning the frequency of ESA, and Fig. 2(c) exhibits the CRN gain spectrum at fiber locations of 1 km, 2 km and 3 km, respectively. The measured CRN spectrum indicates that the CRN is a kind of low frequency noise, and it becomes very weak when it’s higher than 30 MHz, thus the intermediated frequency should be located out of this frequency range.

 

Fig. 2 (a) The measured CRN time-domain trace at 10 MHz; (b) the measured CRN spectrum; (c) CRN gain spectrum at fiber locations of 1km, 2km and 3km, respectively.

Download Full Size | PDF

For proof of concept, a 7-core ~6 km long MCF has been used in the experiment, however only the central core and one outer core have been employed to implement BOTDR and ROTDR, respectively. Figure 3(a) shows the cross sectional view of the MCF used in the experiment, which has 150 um cladding diameter and 42 um core-core pitch; the outer six cores are arranged hexagonally. All the cores are surrounded by deep trench. As a result, the crosstalk between adjacent cores has been suppressed to be as low as −45dB/100km.

 

Fig. 3 (a) Cross sectional view of the 7-core MCF; (b) The measured Brillouin gain spectrum as a function of fiber distance.

Download Full Size | PDF

The temperature and stain dependences of BFS have been calibrated in order to determine the coefficients of Eq. (3). The modulation frequency applied to the MZM modulator at the OLO-branch is 10.45 GHz; in this case the intermediated frequency is found to be located around 240 MHz, which is out of the CRN spectrum range. The BGS is then obtained by scanning the frequency of ESA from 150 MHz to 350 MHz with a step of 2 MHz, as shown in Fig. 3(b). A short segment of the sensing fiber at far end was immersed in a water bath to apply different temperature. The measured BFS under distinct temperature has been presented in Fig. 4(a); the temperature sensitivity of BFS ($C_T$) is linearly fitted to be 1.126 ± 0.014MHz/°C. For the calibration of strain response coefficient, the BFS distributions under different strain have also been recorded; the peak frequency shift of BGS as a function of strain is given in Fig. 4(c), indicating that the linearly fitted strain sensitivity ($C_{\epsilon }$) is 495.6 ± 1.05 MHz/%. The frequency accuracy of the measured BFS can be estimated by calculating the standard deviation of the BFS along the fiber [22], as shown in Fig. 4(d).

 

Fig. 4 The calibration of temperature and strain sensitivity. (a) The measured BFS along the whole sensing fiber with different temperature; the inset shows the BFS distribution near the hot-spot at the far end of the sensing fiber; (b) the peak frequency shift of BGS as a function of temperature; (c) the peak frequency shift of BGS as a function of strain; (d) error in BFS measurement versus fiber length.

Download Full Size | PDF

In conventional Brillouin distributed sensor, the variation of BFS could be induced by temperature and/or strain, thus giving rise to ambiguity. However in the proposed solution, this temperature and strain cross-sensitivity issue can be addressed with the help of Raman measurement, since Raman signal is only sensitive to temperature. Therefore here the temperature sensing performance of SpRS based on the MCF SDM configuration has also been evaluated. Several different temperatures have been applied to a short fiber segment at the far end for testing, and each trace has been averaged for 20480 times for the measurement of Stokes and anti-Stokes SpRS signals, among which the measured Raman Anti-Stokes time-domain traces have been presented in Fig. 5(a). It’s found that the measured intensity at the near-end up to about 1200 m exhibits random fluctuation. This is because the Raman filter used in our experiment has low extinction ratio, while the backscattered light at the near-end is high, thus a small portion of the Rayleigh scattering light is leaked into the photodiodes, creating coherent Rayleigh interference (CRI) and resulting in the intensity fluctuation. The profiles of temperature distribution are then calculated based on Eq. (1) with respect to a reference at 29 °C, as shown in Fig. 5(b), where the inset shows the local temperature profiles around the heated fiber segment with different temperatures applied. Due to the CRI induced random intensity fluctuation added at the measured Raman scattering signals at the near-end of the sensing fiber, the resolved temperature at the corresponding length also show unusual large error, as can be clearly observed at the near-end of the sensing fiber in Fig. 5(b).

 

Fig. 5 (a) Raman Anti-Stokes traces with different temperature applied at the heated section; the inset shows the local view of intensity around the heated section. (b) Resolved temperature distribution profiles along the sensing fiber based on Raman measurement; the inset shows the enlarged view around the hot-spot.

Download Full Size | PDF

Once the temperature profile is retrieved from Raman measurement, strain can be derived from Eq. (3) with the measured BFS information. In this way, temperature and strain discriminative measurement is achieved. The temperature resolution has been estimated by calculating the standard deviation of the retrieved temperature profile along a window of 5 m [13], as shown in Fig. 6(a), where the red dot line is the fitting of the entire fiber length, and it indicates that the calculated temperature has abnormal large uncertainty at the near end of the sensing fiber due to CRI induced random intensity fluctuation. While the black dot line is a fitting for partial fiber length ranging from 1200 m to the fiber end, which is out of the CRI region, and it reveals that the trend of this fitting is similar to any other normal systems, thus confirming the reliability of the system. On the other hand, derived from Eq. (3), the strain resolution has also been estimated by calculating the propagation of uncertainty of Raman measurement deduced temperature ($\Delta T$) and uncertainty of BFS ($\Delta {\nu }_B$), assuming that the two kinds of noise are not correlated [22]. The calculated strain resolution is shown in Fig. 6(b). The worst temperature and strain resolutions (fitted standard deviation) are estimated to be about 2.2 °C and 40με respectively, appearing at the end of the sensing fiber.

 

Fig. 6 (a) Estimated temperature resolution along the sensing fiber from Raman measurement; (b) Estimated strain resolution along the sensing fiber derived from both the measurements of Raman and Brillouin.

Download Full Size | PDF

4. Conclusion

In conclusion, we propose and experimentally demonstrated a multi-core fiber enabled spatial-division multiplexed hybrid Raman and Brillouin distributed sensor, based on optical time-domain reflectometry technology, achieving temperature and strain discriminative measurement. The system employs a single laser source, generates shared pulse for both Raman pump and Brillouin pump, which ensures identical spatial resolution for the two measurements. The interrogation of Raman and Brillouin scattering signals are implemented through different fiber cores, therefore it allows for flexible control of the input pulse power for Raman and Brillouin measurement, and this can effectively overcome the incompatible input pump power level required for hybrid ROTDR and BOTDR in SMFs. The worst temperature and strain resolutions are estimated to be about 2.2 °C and 40 με respectively in 6 km sensing range with 3 m spatial resolution. Moreover, it’s believed that the pulse coding technique can effectively extend the sensing distance in this system. The benefit of using MCF instead of a bundle of multiple SMFs is to ensure that every core at the same position is undergoing strictly identical longitudinal strain and temperature.