A distributed optical fiber sensing system merged Mach-Zehnder interferometer and phase sensitive optical time domain reflectometer (φ-OTDR) system for vibration measurement with high-frequency response and high spatial resolution is demonstrated, where modulated pulses are proposed to be used as sensing source. Frequency response and location information are obtained by Mach-Zehnder interferometer and φ-OTDR technology, respectively. In order to simulate high-frequency vibration of crack of cable and civil structure, experiments on detection of piezoelectric transducer and pencil-break are carried out. Spatial resolution of 5 m and the maximum frequency response of ~3 MHz are achieved in 1064 m fiber link when the narrow pulse width is 50 ns.
©2013 Optical Society of America
The distributed optical fiber sensing system, with a range of advantages over conventional sensors, has been extensively studied and adopted in industrial applications during the past decades. Owing to the huge number of processing waveforms to acquire a reasonable signal to noise ratio (SNR), distributed optical fiber sensors have been mainly investigated for static measurements, such as temperature or static strain. Dynamic process, i.e. time-varying signals, involving a walking person, leakage of pipelines, vibration of engines, and crack of bridges or cranes, requires not only precise location, but also frequency detection spanning from a few Hz to hundreds of KHz, or even to tens of MHz, in order to determine the event type.
Up to now, distributed vibration measurements mainly include optical fiber interferometer sensors and optical backscattering based sensors. Interferometer sensors acquire vibration information by integration of the phase-modulation signal, and usually two interferometers are used to determine position of the vibration. Sagnac interferometers have received significant interests, because of the high sensitivity of vibration along a long sensing range. In order to integrate this interferometer, a variety of configurations have been proposed, including combining the Sagnac to a Michelson , modified Sagnac/Mach-Zehnder interferometer , using twin Sagnac interferometers , and adopting a variable-loop Sagnac  . Due to the differences between two correlated signals caused by noises or changes of polarization state, the spatial resolution of these twin-interferometers is not satisfied. 40 m spatial resolution over 6 km sensing fiber has been achieved by using Sagnac-Michelson interferometer , and system based on Dual Michelson interferometers has realized 102 m spatial resolution in 4012 m fiber link . In addition, frequency response has been analyzed in some systems. A dual Mach-Zehnder interferometer with both location information and frequency response is demonstrated and the vibration event type is distinguished through frequency analysis . Also, KHz level of frequency response and location information are obtained by adding a sub-loop in Sagnac interferometer . A configuration based on Michelson interferometer has been reported, by using 3 × 3 coupler, the frequency response is obtained through integrating the interference signal, and the vibration point is located . Theoretically, the detected frequency range of interferometers could be very wide because it is mainly determined by sampling frequency of data acquisition card, but the spatial resolution should be further improved.
Another distinguished technique is the use of optical backscattering based sensors. A promising technique is phase sensitive optical time domain reflectometer (φ-OTDR), which improves the spatial resolution by using a narrow line-width laser, and a field test for intruders has been demonstrated, achieving 19 km sensing length with 100m spatial resolution  . In order to detect high vibration frequency in health monitoring of civil structure, system based on φ-OTDR, with 5 m resolution of location and 1 KHz detected frequency response, has been reported lately . Another system based on Polarization-OTDR (POTDR) has achieved 10m spatial resolution in 1 km sensing range, while the detected frequency response is 5 KHz . Brillouin based dynamic strain sensors have been researched recently  . Bernini et al. proposed a configuration based on stimulated Brillouin scattering, and obtained 98 Hz frequency response with spatial resolution of 3 m . These optical backscattering based sensors exploit the vibration sensitivity of backscattering light, which unfortunately is extremely low power values, thus requiring a great number of waveforms to obtain a high SNR through averaging, resulting in severely decreasing the detected frequency, and restrict detected vibration frequency ranges of the sensing system.
In high-frequency measurements, like crack of cables and civil structures, both high frequency response and accurate location are required, considering the advantages of interferometer and OTDR system, it could be an effective solution to realize these two sensors in one configuration. To achieve that, we proposed that modulated laser pulses, with ultra-narrow line width as well as minimal frequency drift, are injected into the sensing fiber of a Mach-Zehnder interferometer (MZI) and interfere with the reference light, while the Rayleigh backscattering light in sensing fiber is collected through a circulator. Frequency information is obtained by demodulating the interference signal, and vibration points are located by φ-OTDR system. The combined benefits of MZI and φ-OTDR allow this new scheme for high-frequency responses with satisfied spatial resolution. Experimental results show that maximum detected frequency response of piezoelectric transducer (PZT) is 25 KHz with 100 MHz sampling rate and up to 3 MHz when pencil-break is utilized as a vibration source. At the same time, tests of 50 ns and 100 ns pulse width have been carried out, corresponding to 5 m and 10 m of spatial resolution, respectively. Hence, this new sensing method could realize high frequency response and accurate location detection at the same time in vibration measurements.
2. Principle and signal processing
CW light is used in conventional M-Z interferometer, and data is continuously acquired, while in φ-OTDR system, light pulses are used to obtain high spatial resolution. The nanosecond pulses adopted in φ-OTDR system, are narrow enough that the wide pulses I2 can be treated as continuous light, so that the measurement of MZI is still continuous with narrow pulse I1. In order to combine these two systems in one configuration, a modulated light pulses is developed in our work.
2.1 Modulated pulses
The modulated pulses used in our experiment are shown in Fig. 1(a) . The ultra-narrow line width laser light is intensity-modulated by an AOM, and sequences of pulses are continuously generated. Where one interval of the modulated pulses consists of a narrow pulse I1 (red in Fig. 1(a)) with high intensity and a wide pulse I2 (dark blue in Fig. 1(a)) with low intensity, whose pulse width is τ1 and τ2, respectively. Normally, τ1 is about several to tens of nanoseconds. The pulse pattern can be written as:
In our experiments, the acoustic power IS of the AOM, corresponding to the modulated offset value, is relatively low, so the approximate expression between the optical output and input light intensity is , where Ii is the incident light intensity, Io is the output light power,θB is the Bragg diffraction angle, L is the length of the acoustic beam, M2 is the acousto-optic figure of merit for the crystal, λ is the wavelength of the incident beam and IS is the acoustic intensity. Therefore the output light power of AOM is linear to IS, as shown in Fig. 1(b). Because the range of the driven current is small in our experiment, the linear relationship could be maintained for the small driven current range when the X axis is linear scale and Y is log scale.
Noticed that the intensity of I2 is much smaller than that of I1, because the intensity of I2 determines not only the SNR of interference signal, it also generates backscattering light in sensing fiber, which disturbs backscattering signal of I1, thus decreasing the SNR of location signal. So, experiments are carried out under different intensity value of I2 to optimize the parameters.
2.2 Principle of operation & data processing
The experimental setup is shown in Fig. 2 . A CW light with frequency f is injected into sensing and reference fiber (Corning SM-28e) through a 50:50 fiber coupler, and then an acoustic-optic modulator (AOM) is used to generate modulated pulses with a frequency shift of ∆f in the sensing fiber. Through a circulator, backscattering traces of modulated pulses are acquired by an InGaAs photo detector. To clearly show the amplitude changes at particular vibration points, we adopt conventional adjacent differential method in detection of PZT, while in the case of pencil-break detection, considering the vibration is a sudden event and relatively weak, moving averaging and moving differential method is used to enhance the SNR of location signal .
While Rayleigh signal travels backward, modulated optical pulses and CW reference light are coherent at coupler 2, and the interference signal is received by a photodetector, where the detected current can be described as
Where θ(t) and φ(t) are the relative polarization angle and phase between the modulated pulse light and the CW reference light, respectively, c(t) means the amplitude of pulses sequences. The AC component is obtained through a high pass filter, and it is mixed with the signal to realize the frequency down converting transform. The integrated coherent signal is continuously sampled by the acquisition card, and frequency information can be obtained by a fast Fourier transform spectrum analysis.
3. Experimental results and discussion
The setup of merged MZI and φ-OTDR is shown in Fig. 2. An external cavity laser with ultra-narrow line width of less than 50 KHz is used as light source, which generates CW light with wavelength of 1550.5nm and the maximum optical power is ~10mW. The light is then split into two parts by using 3dB fiber coupler 1 (50:50). One is utilized as reference light to interfere with sensing light through the 3dB coupler 2, while another part is modulated into sequence of pulses with repetition rate of 5KHz and 80MHz frequency shift induced by an acoustic-optic modulator (AOM), where AOM is modulated by an arbitrary function generator (AFG). As demonstrated before, I2 may cause decrease of SNR of φ-OTDR signal, so as to reduce the location accuracy. In experiment, by adjusting the offset (mV level) value of pulses signal in AFG, we can acquire modulated pulses with different I2, and the relationship between I2 and offset value is shown in Fig. 1(b) when I1 is 6.5dBm, thus confirming an appropriate I2 for the system. After that, modulated pulses are amplified through an Erbium-doped fiber amplifier (EDFA), the output of EDFA is filtered by a FBG to remove spontaneous emission. Modulated pulses are injected into the sensing fiber of 1064m by a circulator and the backscattering light from the sensing fiber is detected by an InGaAs photodetector.
Modulated light pulses in sensing fiber interfere with reference light at 3dB coupler 2, and received by a balanced detector to enhance SNR. Filtered by a high-pass filter in which the cut-off frequency is ~75MHz, interference signal is mixed with a sine waveform of 80MHz produced by AFG, and then it passes through a low pass filter with cut-off frequency of 22MHz. The interference waveforms and the Rayleigh backscattering traces are acquired with a data acquisition (DAQ) card with 100MHz sampling rate and 20000 sampling points within one interval, and the data process software is Labview system.
In our experiments, a PZT is used as the vibration source, whose maximum vibration frequency is 28.5 KHz. Also, in order to simulate the crack of civil structure and cables, whose vibration frequency is up to MHz, we have finished the tests on pencil-break detection.
3.1 Detection of PZT
A PZT vibration source is at the position of ~390m of sensing fiber. The sensing fiber is wounded about 0.85m length on PZT, where the loop diameter and number of loops are 34mm and 8, respectively.
Figure 3(a) shows the original interference waveform recorded by one channel of DAQ. There is a pulse in the curve shown in Fig. 3(a), circled in red dashed line, which corresponds to the narrow pulse I1 of modulated pulses reaching the end of sensing fiber. The sampling process and the generation of narrow pulses I1 are synchronized, thus the delay time between generated by AOM and sampled by DAQ card is t = L/v, where L is the sensing range, v is the group velocity, here, L is 1064m and v is 2*10^8m/s, adding the delay of electric signal, the delay time is around 5 to 10 microsecond. It could be removed before FFT transform to avoid generating unexpected frequency components at the frequency of repetition rate in frequency spectrum. 100 consecutive Rayleigh backscattering traces were recorded by another channel of DAQ, as shown in Fig. 3(b), superimposing those traces to show the amplitude changes caused by PZT at 390m location (also circled in red dashed line). The zoomed in vibration position is shown in inset of Fig. 3(b).
Figure 4 to Fig. 5 gives the experimental results of 50ns modulated pulses. Given offset value is the vital factor to the SNR of interference and backscattering signals, experiments, consisting of vibration location with pulse width of 50ns and frequency analysis of 10 Hz/25kHz, were carried out with different offset values.
Restricted by the 5 kHz of pulse repetition rate, the lowest frequency response of MZI is 5 kHz. However, we can adopt φ-OTDR system to obtain the true detectable lower frequency response of this system , while the higher test frequency response could be acquired by MZI. This is the outstanding merit of our system. Here we define frequency resolution (FR) as FR = sampling rate/total sampling points. During data processing of MZI, 10 sampling intervals with 20000 sampling points within one interval were adopted for FFT transformed, so the frequency resolution is 100MHz/200000 = 500Hz, and for φ-OTDR system, FR = 5kHz/1000 = 5Hz. As shown in Fig. 4, the highest test frequency response is 25 kHz, and the lowest frequency response is 10Hz. The total amplitude of power spectrum increases with the increasing of offset level, because enhancement of optical intensity in sensing fiber directly improved the SNR of interference signal. The power of spectrum was relatively low, however, the main frequency could be easily recognized, because SNR of the spectrum ranges from 20dB to 30dB (shown in Fig. 4).
Figure 5 gives experimental results of vibration location with 50ns modulated pulses with different offset values. 100 consecutive traces of sampled backscattering light were recorded, then, we can obtain 99 differential traces by subtracting 2 adjacent traces. The vibration point is evidently shown in these superimposing differential signals, which was a peak around 390m. Noticed that voltage is proportional to light intensity, here we define SNR as the ratio between the signal peak-to-peak voltage and the background noise peak-to-peak voltage, thus, SNR = 10log (Vsignal/Vnoise). In Fig. 5(a), when offset value is 50mV, Vsignal = 0.0123V, Vnoise = 0.0029V, so SNR is ~6.2dB. When increasing the offset value to 100mV, SNR in Fig. 5(b) decreases to ~5.3dB. In Fig. 5(c), the SNR decreases to ~4.1dB when offset value is 150mV. Then, when offset value is 200mV, SNR in Fig. 5(d) is too bad to locate the vibration point for modulated pulses with 50ns pulse width. It is obviously that SNR of location signal is decreased as offset value is increased, on the contrary, the amplitude of spectrum is increased with the increase of offset value, thus making it necessary to confirm the optimized offset value to obtain a satisfied SNR of location signal and also fine spectrum intensity in applications.
3.2 Detection of pencil-break
In order to test the system performance on broad band signals, pencil-break is adopted to simulate the crack of civil structure, considering the frequency spectrum of pencil-break is broad band from few Hz to tens of MHz. 9 loops fiber with 6.5cm diameter was glued to a thin aluminium board of 2mm thickness by silicone. The length of fiber glued to board is about 1.8m, adding in 1064m sensing fiber link at around 610m. In experiment, we break the pencil adjacent to the edge of the fiber loop instead of directly on the fiber, that is, the fiber senses vibration propagating through the aluminum board, which means sensing fiber can detect crack with no damage itself in applications.
The repetition rate of modulated light pulses is 5kHz and the pulse width is 50ns, considering the frequency components are broad as high as 5MHz and vibration lasts about 20ms-200ms , 100 waveforms were recorded. Like the tests on detection of PZT before, experiments were carried out under different offset values. Figure 6(a) shows the integrated interference waveform of MZI, due to pencil-break is a sudden event, the red dashed line indicates the happen of pencil-break. The thin line on the left of Fig. 6(a) means the high pulse with very short pulse width. We adopted moving averaging and moving differential method  to process 100 backscattering traces, and the averaging time was 50. Superimposing these 50 traces shows the amplitude changes around 610m (circled in red dash line in Fig. 6(b)).
Frequency response with different offset values under 50ns modulated pulses was obtained by FFT of integrated interference signal (shown in Fig. 7 ). We observed that as the increasing of offset value, the detectable maximum frequency response and the total power of spectrum increases. Because when offset value is relatively low, signal is strongly disturbed by noise, resulting in a low SNR. Also, considering the pencil-break experiments are carried out by manual work, therefore the spectrum power of each frequency components are different when we break the pencil each time, leading to the spectrum differences. And the maximum detectable frequency response is 3MHz with 200mV offset value.
Figure 8 shows the results of location with 50ns modulated pulses with different offset values, in which an evident peak is observed around 610m, which means the location of pencil-break event. The SNR, shown in Fig. 8, decreases as the increase of offset value, which makes a good agreement with the experimental results of location of PZT. 5m of spatial resolution with 50ns pulse width is acquired, and the similar results are also achieved.
Figure 9 shows the relationship between frequency range and the SNR of location information for 50ns/100ns modulated pulses under different offset values. It is obvious that the SNR of location information decreases with the increase of offset value under different pulse width of modulated pulses. As a whole, SNR of 100ns modulated pulses is slightly larger than that of 50ns modulated pulses, though the SNR of pencil break location information of 100ns modulated pulses is smaller than that of 50ns modulated pulses under 175mV or 200mV offset value, it could result from the backscattered light of I2 interfering with itself and the backscattered light of I1.Frequency range of PZT with 50ns or 100ns modulated pulses under different offset values remains the same of ~25KHz, which depends on the maximum vibration frequency of PZT shown in Fig. 9(a). While frequency range of pencil break increases with increase of offset value, owing to increase of power of I2, shown in Fig. 9(b). To obtain reasonable SNR of location information and frequency range, an appropriate value of I2 should be selected.
1) In theory, the maximum detectable frequency of conventional MZI could be tens of MHz. However, the intensity of I2 in sensing fiber is very low compared to that of conventional MZI, therefore, signal is strongly disturbed by noise, and resulting in a low SNR, limiting the maximum detectable frequency response in our experiments. Also, considering that fiber senses vibration by acoustic wave propagating through a metal board, the frequency range of pencil-break is dependent on the ability of transforming acoustic wave of metal board. Restricted by the lab condition, calibration of frequency response of pencil-break on an aluminium board is not carried out. Thus, the maximum frequency response of this system in experiments is mainly limited by the above two reasons.
2) Noticed that there is a gap between the highest frequency response of φ-OTDR system as fmax (restricted by the repetition rate) and the lowest frequency response fmin of MZI (related to the repetition rate). In application, normally there is a characteristic frequency range, therefore one can adjust the repetition rate to adapt to the measurement occasion, moreover, the sampling ends of φ-OTDR system and MZI are separate, so two DAQ cards with different sampling rate can be adopted in the scheme to avoid the problem.
3) Experiments were carried out with different intensity of I2 (corresponding to offset level in AFG). I2 interferes with reference light as a signal light to form M-Z interferometer, at the same time, although the intensity of I2 is much lower than that of I1, the generated Rayleigh backscattering light can disturb the location signal in two ways: (I) the backscattering light interferes with I1. (II) The backscattering light becomes the background noise, so that the SNR of location signal is decreases. In order to prove the backscattering of I2 would not deteriorate the spatial resolution, system spatial resolution with no offset level and 200mV offset level were analyzed (shown in Fig. 10 ). The spatial resolution remains 5.0m (10%-90% response distance to a power step) in Fig. 10(a) and (b), indicating that I2 only decreases the SNR of location signal. Also, modulated pulses with wider pulse width of τ1, which means higher peak power of light pulses, can adopt wider range of I2. In our experiments, by using conventional differential methods, 50ns pulses operate well when offset level is under 150mV, while 100ns pulses can make it at 200mV.
4) The theoretical sensing range of MZI is L = (P0-PD)/ α, where P0 is the injected light power, PD is the minimum received light power, and α is the fiber loss at 1550nm, therefore the theoretical sensing range of MZI could be hundreds of kilometer. As to φ-OTDR system, the theoretical sensing range is L = (PR-PD)/ α, where PR is the received light power of Rayleigh backscattering light which is related to the injected light power P0, PD is the minimum received light power, and α is the fiber loss at 1550nm. Because the Rayleigh backscattering light is very low, the sensing range of this scheme mainly depends on the φ-OTDR system, which is commonly less than 20km without distributed optical amplifier.
5) In theory, two vibration points can be monitored in our system. High frequency response could be obtained by MZI, while low frequency components are acquired from φ-OTDR system, here we define the highest frequency response of φ-OTDR system as f0 (restricted by the trigger frequency), f1 is the vibration frequency of point 1, and f2 is of point 2. The two vibration events can be monitored in these 3 states: (I) f1 or f2 is lower than f0. Then the lower frequency could be integrated form φ-OTDR system, and the higher one could be obtained by MZI. (II) f1 and f2 are both lower than f0. These two frequency responses can be acquired by φ-OTDR system. (III) f1 and f2 are both higher than f0. If f1 and f2 are frequency independent, they can be acquired by MZI, and if not, they cannot be distinguished from each other in frequency domain.
6) By now, experiments are all carried out in 1064m fiber link, in order to increase the sensing distance, more homogeneous fiber with high ability of anti-noise could be utilized, and vibration-sensitive material can be coated to sensing fiber, which make it more effective for field measurement.
7) In application, the recognition of vibration events is very important, which directly shows what is happening around the sensing area, unfortunately, sensors with mono parameter measurement usually have a high misjudgment rate. So cameras are often used to determine the vibration event when vibration is located, thus making a high cost in monitor. Also, there could be multiple vibration points along the sensing fiber, by now, measurement and recognition of multiple vibration points is rarely reported. By using the method of combining interferometer and OTDR system could achieve wide detected frequency range and high spatial resolution at the same time in vibration measurement, therefore, location information, frequency response and even vibration mode can be monitored at the same time, which means there are three parameters can be obtained to be the decision criterion in recognition of vibration events. So, this low-cost system is potential of recognition of vibration events.
In this paper, a distributed optical fiber vibration sensing technology based on merged Mach-Zehnder interferometer and φ-OTDR system for high-frequency response and spatial resolution is demonstrated, and a modulated pulse is adopted to realize these two systems in one configuration. In our experiments, frequency response and location of vibration points are investigated with different offset level of modulated pulses. 5m of spatial resolution is achieved over 1064m sensing fiber by using 50ns pulses, also the lowest frequency response of 10Hz and highest frequency response of 3MHz are realized with 100MHz sampling rate. This new distributed vibration sensor is potential in large range frequency measurements.
This work was supported by the Fundamental Research Funds for the Central Universities (Project No. CDJZR12125502), and the Program for NCET (Grant No. NCET-08-0602).
References and links
1. S. J. Spammer, P. L. Swart, and A. A. Chtcherbakov, “Merged Sagnac-Michelson interferometer for distributed disturbance detection,” J. Lightwave Technol. 15(6), 972–976 (1997). [CrossRef]
2. A. A. Chtcherbakov, P. L. Swart, S. J. Spammer, and B. M. Lacquet, “Modified Sagnac/Mach-Zehnder interferometer for distributed disturbance sensing,” Microw. Opt. Technol. Lett. 20(1), 34–36 (1999). [CrossRef]
3. S. J. Russell, K. R. C. Brady, and J. P. Dakin, “Real-time location of multiple time-varying strain disturbances acting over a 40 km fiber section using a novel dual-Sagnac interferometer,” J. Lightwave Technol. 19(2), 205–213 (2001). [CrossRef]
4. X. Fang, “A variable-loop Sagnac interferometer for distributed impact sensing,” J. Lightwave Technol. 14(10), 2250–2254 (1996). [CrossRef]
6. M. Kondrat, M. Szustakowski, N. Pałka, W. Ciurapiński, and M. Życzkowski, “A Sagnac-Michelson fiber optic interferometer: signal processing for disturbance localization,” Opto-Electron. Rev. 15(3), 127–132 (2007). [CrossRef]
8. B. Kizlik, “Fibre optic distributed sensor in Mach-Zehnder interferometer configuration,” in Modern Problems of Radio Engineering, TCSET, Proceedings of the International Conference (2002), pp. 128–130.
9. T. Omori, K. Y. Hashimoto, and M. Yamaguchi, “Position-detectable optical distributed vibration sensor using an additional sub-loop,” Sensors, 2004. Proc. IEEE 2, 583–586 (2004).
10. M. Chojnacki, B. Kizlik, and W. Ciurapinski, “Distributed sensor of vibration in fiber optic Michelson interferometer configuration,” in The Experience of Designing and Application of CAD Systems in Microelectronics. CADSM 2001. Proceedings of the 6th International Conference (2001), pp. 183–186.
11. J. C. Juarez, E. W. Maier, K. N. Choi, and H. F. Taylor, “Distributed fiber-optic intrusion sensor system,” J. Lightwave Technol. 23(6), 2081–2087 (2005). [CrossRef]
13. Y. Lu, T. Zhu, L. Chen, and X. Bao, “Distributed vibration sensor based on coherent detection of phase-OTDR,” J. Lightwave Technol. 28, 3243–3249 (2010).
15. K. Hotate and S. S. L. Ong, “Distributed dynamic strain measurement using a correlation-based Brillouin sensing system,” IEEE Photon. Technol. Lett. 15(2), 272–274 (2003). [CrossRef]