1. Introduction

Fiber sensors using stimulated Brillouin scattering have been good candidates for the distributed measurement of temperature or strain variation in structures and materials [1–4]. A number of successful measurements have been reported based on Brillouin optical time-domain technique, showing the advantages of long measurement range more than 100 km [4] and high accuracy of ~10 µε or sub K [3]. However, their long measurement time (~several minutes) and limited spatial resolution (>50 cm) are not sufficient for some applications like dynamic health monitoring of various civil structures. As another approach, a correlation-based Brillouin sensing system, called Brillouin optical correlation domain analysis (BOCDA), has been proposed a few years ago based on the synthesis of optical coherence function [5]. The BOCDA has been shown to provide unique advantages of high spatial resolution (~1 cm) and high speed (~57 Hz) measurements with the random accessibility of measuring position [5–7], which possibly meets all the requirements to replace FBG-based point sensors.

In the BOCDA system, Brillouin interaction is occurred by two counterpropagating continuous waves, called pump and probe waves. A frequency modulation is applied to both waves in order to localize the sensing position by generating a sharp correlation peak along fiber under test. The enlargement of measurement range under fixed spatial resolution requires the increase of the modulation amplitude (Δf) of light source [5], which leads to two significant problems in the measurement. One is the inability of optical filtering due to the spectral overlap between different wave components. When Δf is increased to be more than the half of the Brillouin frequency (ν_B) of fiber, the spectra of pump and probe waves start to overlap due to the modulation. The backward reflection of pump waves in the overlapped spectrum can not be distinguished from probe waves by any optical filter, which induces a considerable amount of noise to the signal. In case of using a typical electro-optic modulator (EOM) to generate frequency offset between pump and probe waves, the system additionally suffers the reduction of the signal when Δf is larger than ν_B. Since the spectral overlap takes place between the two sidebands of the EOM which contribute oppositely (gain and loss) to Brillouin measurements, this condition results in the strong decrease of the signal. These filtering problems have been recently solved by introducing double lock-in amplifiers and a single-sideband modulator to the system [8].

The other problem is more intrinsic and related to the structure of the measured Brillouin gain spectrum (BGS). In principle, the measured signal of the BOCDA system is the sum of all the local BGS’s along the fiber. While the local BGS is sharp and Lorentzian-shaped at the correlation peak position, those at all the other positions are spread out and accumulated to compose a background noise as the substructure of the signal. The background noise tends to stack higher with longer measurement range, which restricts the maximum measurable strain and the measurement range of the BOCDA system by leading to the failure of sensing over certain limits.

In this paper, we newly introduce intensity modulation scheme to the BOCDA system to suppress the substructure appearing as the background noise of the measurement. Different kinds of modulation schemes are tested, and an optimum configuration is experimentally demonstrated where the maximum measurable strain is increased as a result of suppression and proper modification of the background noise. The enhancement of the performance is confirmed by the successful distributed measurement of a short and heavily-strained section that is not detectable in former BOCDA configuration.

2. Principle

Figure 1(a) shows the schematic of the BOCDA system with a few local Brillouin gain spectra depicted near a correlation peak position. Continuous light waves are used for both pump and probe waves, and their frequencies are simultaneously modulated to generate a sharp and periodic correlation peak where the local BGS reflects a clear Lorentzian shape of intrinsic Brillouin gain (red curve). On the contrary, as the position is shifted from the correlation peak, the local BGS is broadened and spread out (gray curves). The measured BGS of the system is composed of the sum of the local BGS’s from all the positions of a fiber under test (FUT).

When a sinusoidal frequency modulation is applied to the light source, the spatial resolution Δz and the measurement range d_m (interval between nearby correlation peaks) of the BOCDA system is given by [6]:

where V_g, Δν_B, f_m and Δf are the speed of light in the fiber, the intrinsic linewidth of BGS (~30 MHz in fibers), the laser modulation frequency and the modulation amplitude, respectively.

The measurement is carried out by recording the Brillouin gain of the probe wave at the end of the FUT, while the frequency offset (Δν) between the pump and the probe waves is swept around νB. The upper graph of Fig. 1(b) shows the typical variation of the measured BGS when strain is applied to the sensing (correlation peak) position. Only the real signal (red) is shifted from the original position in response to the applied strain, while the background noise (black) is unchanged. Since the background noise is pyramid-shaped substructure centred on the initial ν_B of the fiber, the peak to peak ratio of the signal and the noise (i.e. SNR) is decreased with larger strain. Beyond certain strain limit (dashed line), the signal is decreased below the noise level, which results in the failure of strain measurement (gray). As is depicted in the lower graph, the strain limit tends to decrease in longer measurement range due to the increase of the background noise. Therefore, this feature restricts the maximum measurable strain under constant measurement range or the measurement range under fixed measurable strain limit, in addition to that given by Eq. (2).

 

Fig. 1. (a) Schematic of a Brillouin correlation domain analysis (BOCDA) system. Measured Brillouin gain spectrum (BGS) is the sum of local BGS’s (LBGS); Δν, frequency offset between pump and probe waves. (b) Variation of the BGS in response to the applied strain to the sensing (correlation peak) position. Note that the maximum measurable strain (dashed line) is limited by the peak of the background noise and that the measurable strain limit is decreased in longer measurement range (lower) than the shorter case (upper). Δν is the relative frequency offset with the initial value set to zero.

Download Full Size | PDF

In this work, an intensity modulator (IM) is newly inserted after the light source (laser diode) of the BOCDA system and the transmittance is controlled synchronously to the frequency modulation of the laser diode. A similar approach has already been applied to the reflectometry based on the synthesis of optical coherence function, where an intensity modulation scheme was introduced to modify the optical spectrum for the purpose of efficient suppression on the sidelobe in synthesized coherence function [10]. In the same manner, the optical spectra of pump and probe waves of this scheme are modified by way of the variation of the transmittance of the IM, which results in the suppression and the modification of the background noise of the BGS. We test different shapes of optical spectra using different intensity modulation and experimentally propose an optimum configuration which provides the best result in the measured BGS. The improvement is confirmed by the increase of the SNR more than 40% and the successful distributed measurement of a critical configuration which can not be detected using former BOCDA system.

3. Experiments

The experimental setup is depicted in Fig. 2(a). A 1555-nm DFB laser diode (LD) was used as a light source, and a sinusoidal frequency modulation was applied to generate a correlation peak within a FUT. An intensity modulator (IM) was connected to the output of the LD and the transmittance was controlled synchronously to the frequency modulation. The output from the IM was directly used as the Brillouin pump wave, after passing 3-km delay fibers to control the order of periodic correlation peaks and being amplified by an Er-doped fiber amplifier (EDFA).

 

Fig. 2. (a) Experimental setup of the BOCDA system with the intensity modulation scheme applied: LD, laser diode; FUT, fiber under test; PD, photodiode. (b) Structure of the fiber under test composed of several sections of dispersion shifted fiber (DSF) and standard single-mode fiber (SMF). Note that the length of the DSF section (30 cm) and the overall length (~305 m) were set to the nominal spatial resolution and the maximum range determined by the modulation parameters.

Download Full Size | PDF

The probe wave was generated through a single-sideband modulator (SSBM) using a microwave synthesizer and a proper DC bias control, so as to suppress the carrier and the anti-Stokes component of two first-order sidebands, and maintain a stable frequency difference Δν from the pump wave [8]. The suppression ratio of the other frequency components was kept more than 25 dB. A polarization switch, PSW, was inserted after the SSBM to apply the polarization diversity scheme [9], where two orthogonal polarizations were alternatively applied for the probe wave to suppress the polarization dependent fluctuation of the Brillouin gain by performing an averaging between them. Lock-in detection was applied twice to both the pump and the probe waves at the chopping frequencies of 5.3 MHz and 8.2 kHz respectively, which clearly extracted amplified probe component without using any optical filter [8]. A 125-MHz photodiode (PD) was used as a detector and the final data was acquired after serially connected two lock-in amplifiers.

The FUT was prepared by concatenating conventional single-mode fibers (SMF) and three pieces of 30 cm dispersion-shifted fibers (DSF) as shown in Fig. 2(b), where the overall length was about 305 m. The average ν_B’s of the fibers were about 10.5 GHz for DSF and 10.8 GHz for SMF, respectively. The ν_B difference of the two fibers corresponds to the strain amplitude of ~6000 µε. The modulation frequency f_m of the laser was 310~330 kHz depending on the measurement position along the fiber, which corresponds to the measurement range of about 310 m according to Eq. (2). The amplitude of the frequency modulation Δf was set to 9.5 GHz in order to set the nominal spatial resolution to 30 cm based on Eq. (1), which is the same as the length of DSF sections. The structure of the FUT can be regarded as the worst configuration for the BOCDA measurement, considering the long and uniform non-correlation portions (SMF) which will stack the highest background noise under given measurement range, and the strain section (DSF) of the minimum measurable length.

We applied three different intensity modulation waveforms to generate different optical power spectra as shown in Fig. 3(a). Each spectral shape can be characterized by different power distribution between the side and the center of the initial spectrum (No IM) of the sinusoidal frequency modulation. At first, a suitable intensity modulation waveform was calculated to make a flat-top output spectrum (IM 1) out of the initial spectrum. The calculated waveform was applied to the IM using an arbitrary function generator which was synchronized to the frequency modulation as depicted in Fig. 3(b). The other optical spectra (IM 2, IM 3) were produced by manipulating the offset and the amplitude of the modulation waveform of the transmittance used for the IM 1.

The BGS of each modulation scheme was measured at both the SMF and the DSF sections, where the section 2 of the DSF in Fig. 2(b) and the middle position of 200-m SMF section were selected as sensing points. The Δν of the SSBM was swept from 10.2 to 11.2 GHz, and the speed of the measurement for single position was about 0.5 Hz.

 

Fig. 3. (a) Power spectra measured by an optical spectrum analyzer with intensity modulation schemes (IM 1~3) applied in addition to the initial frequency modulation (No IM). (b) Time waveforms showing the synchronization between the frequency modulation of the LD (black) and the transmittance of the intensity modulator (red) applied to generate a flat-top spectrum (IM 1) shown in (a). Note that the other waveforms (IM 2~3) are synchronized in the same way.

Download Full Size | PDF

Figure 4 shows the measured BGS’s in different modulation schemes. As is clearly seen in the case of no intensity modulation (No IM), the real signal from the DSF section is lower than the noise peak, so can not be detected properly in this condition. When the intensity modulations are applied, strong suppression of the noise peak is observed on the DSF sections compared to the signal amplitude in all cases (IM 1~3). At the same time, a large dip is observed at the center of the BGS in the cases of IM 1 and IM 2, which might be the effect of over-suppression. This feature gives a problem in the peak detection of normal position by ‘absorbing’ the signal as shown in the BGS’s of their SMF sections.

 

Fig. 4. Brillouin gain spectra measured on the DSF and the SMF sections using several intensity modulation schemes shown in Fig. 3.

Download Full Size | PDF

 

Fig. 5. Comparison between the measured BGS’s of the DSF (a) and the SMF (b) sections in no intensity modulation (No IM) and the optimum modulation (IM 3) cases. Note that the BGS’s were rescaled to the same signal amplitudes in the DSF sections and the same scale factor was also applied for the comparison of the SMF sections.

Download Full Size | PDF

We found out that this problem could be avoided by properly modifying the optical spectrum with the offset and the amplitude of the intensity modulation. The graph of the IM 3 shows this optimum situation, where the background noise of the BGS remains low and flat, and the signal peaks are clearly distinguished in both DSF and SMF sections. The optical spectrum of the optimum condition is depicted as a blue line in Fig. 3(a).

The direct comparison between the BGS’s of no intensity modulation and the optimum situation (IM 3) is shown in Fig. 5. The signals in both DSF sections were rescaled to the same amplitude and the scale factor was also applied for the BGS’s in the SMF sections. The suppression of the noise peak is evidently seen in both sections, and the SNR is improved by 45% (from 0.82 to 1.27) when calculated from the BGS’s of the DSF sections. Considering the achieved SNR, it seems the system still has more rooms to extend the measurement range under current strain limit (~6000 µε).

The physical origin of the effect of the intensity modulation on the BGS, i.e., the dependency of the BGS on the optical spectrum, can be roughly explained by the fact that the amount of gain at each Δν in the measured BGS has different amount of contribution from each spectral component of the optical spectrum. A detailed theoretical analysis on the relation between the BGS and the spectrum is now under preparation and will be published soon.

In order to confirm the effect of the optimum modulation scheme, we performed distributed measurements with and without the intensity modulation by 10-cm step on the FUT using the same experimental parameters. The measured ν_B’s around the DSF sections are depicted in Fig. 6. As is clearly seen, the DSF sections are successfully measured using the optimum intensity modulation scheme (IM 3), while the positions are missed in the case of no intensity modulation (No IM).

 

Fig. 6. Result of distributed measurement on the fiber under test near the DSF sections. The DSF sections are properly detected only in the optimum intensity modulation (IM 3). The measurement inaccuracy of the ν_B at each position was about +/-3 MHz.

Download Full Size | PDF

4. Conclusion

We proposed a novel method to enhance the performance of Brillouin optical correlation domain analysis system based on intensity modulation scheme. We showed that the background noise of the Brillouin gain spectrum could be suppressed and modified by the control of the optical spectrum using the optimum intensity modulation, and demonstrated the substantial increase of the measurement range as a result of the enlargement of the measurable strain limit.

The performance enhancement was confirmed by the successful measurement of a critical configuration which could not be properly detected by former BOCDA scheme. The improvement of the signal to noise ratio at the sensing position was more than 40%. We believe further theoretical study will be able to provide more detailed analysis on the effect of the intensity modulation on the BOCDA system.

With this intensity modulation scheme and recently proposed double lock-in detection [8], it seems there is no other fundamental issue in the enlargement of the measurement range except the acquisition of a proper light source for fast and broad frequency modulation. We expect the BOCDA system to have more capability as a practical distributed sensor with this implementation.