1. Introduction

Distributed fiber sensors based on Brillouin scattering have been studied for over two decades, and several interesting schemes have been developed in the form of reflectometry or analysis for the measurement of local Brillouin frequency (ν_B) [1–5]. In particular, recent progresses in the Brillouin analysis such as Brillouin optical time-domain analysis (BOTDA) and Brillouin optical correlation-domain analysis (BOCDA) are remarkable, where demonstrations of a measurement range over 50 km with sub-m spatial resolution [6,7], an extreme spatial resolution of cm or mm [8–12], and high-speed measurements (1 kHz for single point and several Hz for the distribution) [13,14] have been reported. One of requisite functions in the Brillouin analysis is to sweep the frequency offset between the pump and the probe waves, Δν, in the vicinity of ν_B of an optical fiber (10 ~11 GHz at 1550 nm wavelength) for the acquisition of the Brillouin gain spectrum (BGS). In general, an optical phase-lock with a frequency counter [15] or a sideband-generation technique with a microwave generator and an electro-optic modulator [3,16] is adopted for the control of Δν, which contributes to the advance in the cost and deteriorates the practicality of the Brillouin sensors.

In former works, there has been a noticeable report on the use of direct modulation of a laser diode for the measurement of the Brillouin frequency of optical fibers [17] which is currently called “time-division pump-probe generation scheme” [18]. The basic principle is to generate the pump and the probe waves periodically at different time and control the frequency offset by the amplitude of the modulation. This method has been successfully applied to build correlation-domain Brillouin sensors, called “simplified BOCDA systems” [13,18].

In this paper, we newly apply the time-division pump-probe generation method to build a time-domain Brillouin sensor and perform a feasibility research on a simplified BOTDA system which does not require any microwave component for the operation. The basic principle, the operational characteristics, and the experimental confirmations are presented. In the experiments, distributed measurements of the ν_B-variation in a 2 km optical fiber are demonstrated with a spatial resolution of 2 m and an accuracy of ± 2 MHz. We think our scheme shows a large potential to provide a low-cost and practical solution for structural health monitoring systems.

2. Principle

The basic principle of the time-domain pump-probe generation for the simplified BOTDA is to control the optical frequency of a LD by the current modulation so that the pump and the probe parts are alternately generated with a period of τ₀ as depicted in Fig. 1(a) . The constructed waves are counter-propagated as shown in Fig. 1(b) with certain time delay for proper interaction of the pump and the probe part within a fiber under test (FUT). The maximum measurement range, R, is determined by R = c (τ₀ –τ₁) / 2 n_eff, where τ₁, n_eff, c are the duration of the pump section, the effective index of mode, and the speed of light, respectively. The measurement range is easily controllable matching to the length of the FUT (L) by changing τ₀, and the required time delay is given by L + cτ₁ / n. The sweep of Δν between the pump and the probe waves is carried out by controlling the amplitude of the LD modulation.

 

Fig. 1 (a) Time-division pump-probe generation and (b) schematic view of the simplified BOTDA system.

Download Full Size | PDF

Figure 2 shows the generation process and the resultant shape of the time-division pump-probe wave adopted in our experiments, which is similar to the waveform applied in [18] with a different time ratio of the pump to the probe section (5:5 in [18] and 2:8 in current setup). The transfer function of a LD is calculated by comparing the Fourier components of the applied ideal RF wave (top) and the resultant variation of the optical frequency of the LD (bottom) as depicted in Fig. 2(a), where the rapid variation of the optical frequency is measured by using the slope-filter method [18]. The transfer function is used to construct a compensated RF wave (top of Fig. 2(b)) which is applied to generate a proper pump-probe wave (bottom of Fig. 2(b). The red curves in Fig. 2(c) are the zoomed views of the generated pump (top) and the probe (bottom) sections, where the frequency fluctuation over 30 MHz is seen in each part due to imperfect compensation mainly coming from the nonlinearity of the transfer function of the LD.

 

Fig. 2 The generation process and the resultant shape of the pump-probe wave used for the simplified BOTDA system. (a) Ideal RF wave (top) applied to the LD and the resultant variation of the optical frequency (bottom). (b) Compensated RF wave (top) applied to the LD and the final pump-probe wave (bottom). (c) Zoomed views of the generated pump (top) and probe (bottom) sections. Note that the red curves are the averaged results suppressing the dark current noise in the slope-filter method.

Download Full Size | PDF

There is an important feature in our system related to the frequency fluctuation and the accuracy of the measurement. The intrinsic fluctuations of the optical frequencies due to the imperfect compensation shown in Fig. 2(c) leads to measurement errors in the absolute values of local ν_B’s of an optical fiber. Meanwhile, the variation of ν_B, Δν_B, that is induced by environmental changes and more important in distributed sensing, is not influenced by these intrinsic fluctuations. This is due to the facts that the shapes of the intrinsic fluctuations are not randomly given but constant in repeated waveforms, so that their effect is simply to provide a position-dependent constant offset to the local ν_B’s and cancelled out in Δν_B-calculation. In addition, the effect of the frequency fluctuation in the pump section is expected to be even smaller since the frequency variation within a very short (~20 ns) pump pulse could be almost negligible. Therefore, one may expect a high accuracy measurement of local Δν_B regardless of large errors in the absolute ν_B’s. This is a unique feature of our simplified BOTDA system, which is clearly different from and advantageous over the case of simplified BOCDA systems where the optical frequency fluctuations in the pump-probe wave directly leads to the drop of accuracy in the Δν_B-measurement.

3. Experimental results

We composed an experimental setup for the simplified BOTDA as depicted in Fig. 3 , where a 1550 nm distributed feedback (DFB) LD was used as a light source being modulated by the pre-generated pump-probe wave at 20 kHz using a 20 MHz arbitrary waveform generator (Agilent 33220A). The output from the LD was divided by a 50/50 coupler to be propagated in both the pump and the probe directions. The direct output from the LD was launched into a fiber under test (FUT) as the probe wave, and a polarization switch (PSW) was used to average out the polarization-dependent variation of the Brillouin gain. A bell-shaped pump pulse with a width of 20 ns (full width at half maximum, FWHM) was generated by gating the middle of the pump section using an electro-optic modulator (EOM) and a pulse generator synchronized to the modulation of the LD. A fiber Bragg grating (FBG) was used to further suppress the leakage probe wave in the direction of the pump, and a 5 km delay fiber (~25 μs) was inserted to control the propagation timing of the pump and the probe waves. The pump pulse was amplified by an Er-doped fiber amplifier (EDFA) and launched into the FUT in the opposite direction to the probe wave. A 125 MHz Photo detector (PD) and a 250 MSa/s data acquisition card (DAQ) were used for the acquisition of time traces. The structure of the FUT, a 2,130 m single-mode fiber, is shown in the inset of Fig. 3, where a couple of 2 m strain-applied sections are located with 2 m separation at the end of the fiber for the test measurement.

 

Fig. 3 Experimental setup of the simplified BOTDA system: VOA, variable optical attenuator; FUT, fiber under test; PD, Photo detector; DAQ, data acquisition card; EOM, electro-optic modulator; FBG, fiber Bragg grating; EDFA, Er-doped fiber amplifier. The inset shows the structure of the FUT.

Download Full Size | PDF

In the distributed measurements of ν_B, the amplitude of the pump-probe wave was externally modulated with an up-ramp waveform by another function generator, so that Δν between the pump and the probe sections was linearly and continuously swept from 10.7 GHz to 11 GHz. The period of the single sweep was 300 second, and the acquisition of the time trace was carried out every 2 second which corresponds to a frequency step of 2 MHz. The start of the Δν-sweep and the data acquisition was synchronized, and each time trace was averaged 2048 times for noise suppression.

Figure 4(a) depicts some of the measured time traces, basically similar to those of common BOTDA systems, which were normalized to the initial amplitude to show the Brillouin gain along the fiber with Δν’s of 10.78, 10.86, and 10.92 GHz, respectively. Figure 4(b) shows an example of local BGS measured at the middle position (dashed line in Fig. 4(a)) of the FUT, which fits well to a Lorentzian curve with a FWHM of ~60 MHz (red curve). It is noticeable that the obtained BGS is fairly clear although large noise-like ripples are observed in the time traces.

 

Fig. 4 (a) Time-traces obtained by the simplified BOTDA system with Δν of 10.78, 10.86, and 10.92 GHz. (b) BGS measured at the middle position of the FUT (dashed line in (a)).

Download Full Size | PDF

The result of a distributed measurement of ν_B’s along the 2 km test fiber is shown in the lower of Fig. 5(a) . A gradual increase of ν_B is seen in the overall distribution together with long-term (a few hundred meters) and short-term (several meters) fluctuations of several tens of MHz (bottom). These fluctuations can be attributed mainly to the intrinsic optical frequency fluctuations in the generated pump-probe wave (see Fig. 2(c)) and partly to the residual strain in the winding process of the spool of FUT.

 

Fig. 5 (a) A distribution map of ν_B obtained by the simplified BOTDA system (bottom) with a zoomed view of the dash-boxed position (top) which shows the results of two separate measurements (red dot and black curve) for comparison (b) Distributed measurements of the ν_B-variation (Δν_B) with strains of 200, 400, 600, and 800 με applied to the test sections near the end of the FUT.

Download Full Size | PDF

One can see a very interesting result in the zoomed view of the distribution maps in two separate measurements (red dot and black curve) depicted together for comparison in the upper of Fig. 5(a). It is remarkable that the results of two independent measurements are highly consistent regardless of the ν_B-fluctuations. This confirms the important feature of our simplified BOTDA system that the accuracy in the measurement of Δν_B can be much higher than that of absolute ν_B’s as explained in detail in the principle section.

Figure 5(b) shows the results of distributed measurements with axial strains of 200, 400, 600 and 800 με applied to the 2 m test sections of the FUT. The measurement error (i.e. the thickness of the base line) was about ± 2 MHz as seen in the maps of the Δν_B-distribution (bottom), which also confirms much higher accuracy of Δν_B than that of absolute ν_B in our system. The change of Δν_B at the 2 m strain-applied sections were clearly observed in the zoomed view near the strain-applied sections (top), which confirms the validity of this measurement scheme and the spatial resolution of 2 m. Currently, the accuracy of ± 2 MHz is poorer than that of conventional BOTDA systems (~ ± 0.5 MHz) and similar to that of BOCDA systems [16]. We think this error is originated from the amplitude jitter of the generated pump-probe wave, however further research is needed on this point.

The variation of Δν_B as a function of the applied strain is depicted in Fig. 6(a) . The result fits well to a line with a slope of 0.0454 MHz/με, slightly smaller than the reported value of ~0.05 MHz/με. This discrepancy seems to be attributed to the mitigation effect of the fiber jacket. Figure 6(b) shows the acquired 3D BGS near the strain applied sections when a strain of 800 με was applied. One can see the clear shift of the BGS at the 2 m strain-applied sections while those at the 2 m in-between section remains unchanged with ν_B of ~10.9 GHz.

 

Fig. 6 (a) Brillouin frequency shift (Δν_B) of one of the test sections as a function of the applied strain measured by the simplified BOTDA system. (b) Acquired 3D BGS near the test sections with a strain of 800 με.

Download Full Size | PDF

4. Conclusion

We have proposed and demonstrated a simplified BOTDA system based on the time-division pump-probe generation by direct modulation of a laser diode. In our system, a low-cost arbitrary function generator has replaced expensive microwave devices needed for sweeping the frequency offset between the pump and the probe waves in common BOTDA systems. The accuracy in the measurement of the Brillouin frequency variation is about ± 2 MHz, and we believe there is still room for the improvement by finding the optimum condition for the LD modulation. We think the measurement time and the spatial resolution of the simplified BOTDA system might be similar to those of conventional BOTDA systems.

Current measurement range of 2 km is related to the LD modulation frequency of 20 kHz, which is limited by the cut-off frequency of the bias-tee used for the LD modulation. We think a longer range would be easily obtainable by using a lower frequency for the LD modulation. For example, the measurement range over 50 km is theoretically possible if the modulation frequency is decreased to 1 kHz.