1. Introduction

Distributed optical fiber sensing has been extensively investigated in past decades due to its superiority over other conventional sensors. The most widely used approaches are based on either the optical time domain reflectometer (OTDR) [1] or the optical frequency domain reflectometer (OFDR) [2], both of which locate external disturbances by optical characteristics of various backscattered light including Rayleigh, Brillouin, and Raman. Distributed Brillouin sensors offer truly distributed sensing capability for deformation and crack detection of steel and concrete structures for static measurement [3, 4]. They can be used to monitor pipelines and bridges, where large size structures have very low vibration frequency, usually of a few Hz. For crack and vibration detection on small structures, such as engines and airplane parts of aerospace structures, power cables, trolley cables and impact damage detection, hundreds to kHz frequency detection is required in addition to location information.

Up to now, distributed optical fiber sensors have been mainly studied for static measurements, i.e. no time-varying or slowly time-varying signals, such as, static strain or temperature. Dynamic measurements using the above techniques is difficult to achieve because of the large number of waveforms required to average out the polarization effect induced signal fluctuation or because of the large range of frequency scans that are needed in order to obtain a reasonable signal to noise ratio (SNR) and spatial resolution over a kilometer fiber length.

Hotate et al. recently presented a frequency modulation of the distributed Brillouin sensor by correlation-based continuous wave to implement vibration measurement without suffering weak backscattered signals [5]. However, each time only one sensing point is chosen by the correlation peak of pump and probe light, it is particularly suitable for material processing over a short fiber distance while it is not essentially a fully distributed sensor which should provide information for every point along the fiber under test simultaneously. Time domain analysis based on Brillouin gain dependence of the polarization change is proposed for the dynamic monitoring of highway concrete slab [6]. However, waveform averaging is required to reduce signal fluctuation, limiting the frequency response to 200Hz over 300m sensing fiber.

In this paper, for the first time, a spectrum density of Polarization-OTDR system is proposed to realize fully distributed dynamic measurement along a 1km fiber link with 10m spatial resolution. By carefully designing the transceiver module, high SNR is available to detect up to 5 kHz vibration signals without averaging. Polarization-OTDR was developed as the first fully distributed optical fiber measurement for static physical parameters in the earlier 80’s [1] and then adopted as a diagnostic tool in optical communication systems to identify high polarization mode dispersion (PMD) fiber sections [7]. Polarization-OTDR systems described in [1, 7] utilize a polarization analyzer containing three polarizers at the receiver end to extract full knowledge of the state of polarization (SOP) which consumes the testing time severely; based on the analysis of the rotation of polarization state one can get the distributed polarization state change [8]. Here we will show in section 3 that one polarizer is sufficient to identify dynamic events, through which the birefringence change along the fiber could be detected. Moreover, with a novel fast Fourier transform (FFT) spectrum analysis, multiple simultaneous events with different vibration frequencies or even with the same frequencies are able to be accurately located. The spectral density function of location change is equivalent to many variable narrowband filters with bandwidth of <1Hz to improve the SNR of multiple events detection, which allows the disturbance to be detected simultaneously at any location along the sensing fiber. This new technology could in a cost-effective manner provide intrusion sensing for perimeter security at various places or structure health monitoring for large structures, such as bridges, highway pavements, pipeline leakage, etc. with low fault rate due to the multiple frequency components discrimination at <1Hz narrow band.

2. Experimental configuration

2.1 Setup

The choice of the light source is critical for the Polarization-OTDR system. The laser linewidth should be narrow enough to prevent the depolarization in wavelength that is induced by different polarization transformations of various frequency components in the light source [7]; while the interference of backscattered components from different parts of the fiber due to the coherent light source should be avoided. Therefore, the laser linewidth of Polarization-OTDR usually falls between that of conventional OTDR and Coherent-OTDR which requires kHz bandwidth [9]. Here, a pulsed laser with 0.2nm spectral linewidth has been chosen as the light source.

The experimental setup constructed is shown in Fig. 1. A 100ns pulse of polarized light with 10kHz repetition rate is used. The input polarization direction is chosen to ensure zeros of the Rayleigh scattering to be minimized in the fibers, and the signal strength is maximized along the fiber length. As the pulse propagates along the fiber under test, the Rayleigh backscattering is continuously collected at the input end via a circulator. An avalanche photodiode (APD) detector and amplifiers of 20MHz bandwidth is designed to detect weak backscattered signals and to eliminate extra noises. As we know the SOP evolves in periodical oscillation due to the beat length of the single mode fiber.

The measurement range can be improved by increasing pump power or pulse width; while the spatial resolution is determined by the pulse width. The long sensing length means pulse takes longer time to travel in the fiber, which leads to the decrease of the maximum detectable frequency. Therefore, there is always a trade off between the sensing length, spatial resolution and vibration frequency range. Thus, all the parameters should be carefully chosen according to a specific application.

 

Fig. 1. Experimental setup of Polarization-OTDR system

Download Full Size | PDF

2.2 Data processing

In order to achieve a fully distributed vibration sensor, the signals collected by DAQ card (NI-PCI 5122) have been processed in the following way as illustrated in Fig. 2. First, one Polarization-OTDR curve is stored onto the on-board memory of the DAQ card every 0.1ms, corresponding to a 10kHz repetition rate. Once the memory is filled up, all of thousands of curves will be transferred to computer to be further analyzed. Then, the signals at a certain position will be extracted out and plotted verse time as in step ii. So if the fiber at z₁ is dynamically disturbed, the time trace at z₁ will vary from constant, indicating one disturbing point.

However, there was a problem remaining in the Polarization-OTDR system preventing its application in multiple events detection: the backscattered signals beyond the disturbing point z₁ would be affected as well. Rogers implemented a complicated algorithm [8] to correct the subsequent effects in static measurement; however, such method is not suitable for the purpose of dynamic measurement due to time requirements. Here, a novel FFT spectrum method is proposed for the first time to identify multiple events along a fiber link for the distributed vibration sensor. As shown in step iii, the FFT spectrum is computed for every point beyond the disturbing point z₁. Multiple events with different frequency components are able to be distinguished by the peak readings of the spectrums. As for multiple events with exactly the same frequency f₁, we will demonstrate that by plotting the intensity of f₁ along the fiber as in step iv, these events could be located successfully. Results and further discussions can be found in section 3.2 and 3.3.

 

Fig. 2. Schematic diagram for data processing

Download Full Size | PDF

3. Results and discussion

3.1 Single event detection

The fiber under test consists of a piezo fiber stretcher (OptiPhase PZ2) sandwiched between two spools of SMF with 500m and 300m long, respectively. And several small loops are made at the end of the fiber to avoid the end-reflection by bending loss. The piezo fiber stretcher is constructed by a 40m long single mode fiber (SMF) wounded on a piezo plate to produce continuous vibration by an applied periodic signal and with 1~5 milli-radians/V birefringence modulation. Although the piezo is able to stand 800V peak to peak input, only a 5Vpp driven signal is used here, which causes ~0.5 micro-strain equivalent stretch. The amount of birefringence modulation detected by spectral analyzed POTDR is not strain, rather the birefringence induced intensity change by the piezo fiber stretcher. A low drive voltage is used as we want to check the minimal detectable birefringence change caused by a disturbance. And the birefringence change is detectable for the driven signal reduces to 2Vpp, corresponding to 2~10 milli-radians change and it is equivalent to 0.2 micro-strain change for PZT stretcher.

Post-signal processing as shown in Fig. 2, step i to iii, is employed here by taking an average every 100 Polarization-OTDR curves in step ii. Considering a 10kHz repetition rate of the pulsed light, the effective sampling rate becomes 100Hz, which has set the limitation for impact wave detection. Figure 3(a) plots the FFT spectrum of 1.5 seconds time domain data at 550m with a peak at 22Hz when the piezo is driven by 5Vpp, 22Hz square wave. It is worthy to mention that the on-board memory of the DAQ card would be full with 1.5 seconds of raw data, and after transferring and analyzing the data, it is able to collect data again. Benefited to its high sensitivity, this Polarization-OTDR system makes it possible to measure higher frequency disturbance without any averaging in step ii. Hence, the maximum detectable frequency is 5kHz using a 10kHz sampling rate. As plotted in Fig. 3(b), we demonstrate that when driven frequency of the piezo is set to 4234Hz, this peak frequency is clearly shown in the FFT spectrum at 550m.

Furthermore, in real field applications, a vibrating event usually contains a frequency spectrum rather than one distinct frequency. In order to stimulate events with multi-frequencies, a sweep function is used as the driven signal for the piezo fiber stretcher. Figure 3(c) plots a FFT spectrum of the P-OTDR time trace at 550m (black solid line) compared to a spectrum of the driven signal (red dotted line), which sweeps from 15Hz to 30Hz in 1.5 seconds. This result is obtained by averaging 100 P-OTDR waveforms. If no average is taken, higher frequency components are detectable. Figure 3(d) shows the spectrum of a signal at 550m at a higher vibrating frequency range of 4600 to 4680Hz, the measured vibration spectrum agrees well with its corresponding driven signal over 1.5 seconds. For the same driving power the wider the spectrum is equivalent to lower the energy of each frequency component, hence the detected spectrum range is limited by signal to noise ratio.

 

Fig. 3. Piezo fiber stretcher driven by 5Vpp square wave, FFT spectrum of time trace signal at 550m of (a) 22Hz driven signal; (b) 4234Hz driven signal; (c) FFT spectrum of P-OTDR signal (black solid line) comparing to the spectrum of sweep driven signal (red dotted line), from 15Hz to 30Hz; (d) FFT spectrum of P-OTDR signal (black solid line) comparing to the spectrum of sweep driven signal (red dotted line), from 4600 to 4680Hz.

Download Full Size | PDF

3.2 Double simultaneous events with different vibration frequencies

In order to identify two simultaneous events with different vibration frequencies (f₁ and f₂), an electrically driven polarization controller (EO Space) and another 200m spool of fiber are added into the fiber link. Since the polarization controller is made of LiNO₃, it has 3dB loss one way, leading to 6dB total loss for backscattered light. As shown in Fig. 4(a), before the polarization controller (at 850m), the amplitude (peak to peak voltage) of the oscillating curve is about 80mV, compared to the 20mV afterwards. Although the backscattered light at the detection end has been significantly attenuated, the second disturbing point still could be located. Due to the speed limitation of our analog output driven card to polarization controller, the SOP of light is tuned at 25Hz (f₂=25Hz). While the driven frequency of the piezo fiber stretcher is set to 37Hz (f₁=37Hz). By processing data from step i to step iii as in Fig. 2 with 100 curves averaging, we obtain the following results from the FFT spectrums: any point before the first disturbing point, its FFT spectrum doesn’t have any obvious peak; any point between the first and the second disturbing points, it has a peak at 37Hz; and any point beyond the second disturbing points, its spectrum has two peaks, one at 37Hz and the other at 25Hz. The typical FFT spectrums of the above three cases are plotted in Figs. 4(b), 4(c), 4(d), respectively.

Thi multiple simultaneous events detection is valuable and crucial for distributed vibration sensor. Taking fence intrusion sensor for example, different intruder would cause different disturbing frequencies due to all facts of the intruder, like size, weight, way of climbing, etc. Hence, from the FFT spectrum analysis, multiple intruders could be easily distinguished and located.

 

Fig. 4. (a). a typical Polarization-OTDR curve for 10m lead fiber+500m SMF+40m piezo fiber stretcher+300m SMF+ polarization controller+200m SMF; (b). FFT spectrum for any point before the piezo fiber stretcher: no obvious peaks; (c). FFT spectrum for any point between the piezo fiber stretcher and polarization controller: one peak at 37Hz; (d). FFT spectrum for any point beyond the polarization controller: one peak at 37Hz and the other at 25Hz

Download Full Size | PDF

3.3 Double simultaneous events with the same vibration frequency

In section 3.2, we demonstrate two simultaneous events with different vibration frequencies, which should be sufficient for most applications. However, in some rare situations, it might happen that two simultaneous events have exactly the same vibration frequency. In order to capture every disturbing event, as shown in step iv in Fig. 2, we propose to utilize the magnitude of the peak frequency along the fiber to identify the second event. In other words, a narrow filter with bandwidth <1Hz has been put in the vicinity of the peak frequency, so that the SNR would be tremendously improved.

First, when the piezo fiber stretcher vibrates at 25Hz (f1=25Hz) while the polarization controller remains stable, the magnitude of 25Hz in the FFT spectrum has been plotted as in Fig. 5(a). As it has been mentioned in section 2 that the SOP evolves through the fiber and only one polarizer is placed at the detection end, the magnitude of 25Hz oscillates along the fiber link. Due to the large insertion loss (6dB) of the polarization controller for the same birefringence change amount, the signal variation from the second half (beyond 850m) is much smaller than that from the first half (before 850m). If the second polarization controller has negligible insertion loss, and then the noise floor will be the same for the first and second half with the same environmental and electrical noises. Nevertheless, this phenomenon won’t affect our results of two events detection and could be avoided by using a zero insertion loss vibrating component, like an optical fiber based polarization controller. In Fig. 5(a), at certain locations, since the SOP change is perpendicular to the direction of the polarizer, no intensity change appears at the detector, leading to the magnitude of 25Hz reaches the noise floor (-70dB for the first half and -75dB for the second half). This effect would have an influence upon accuracy of the event’s location decision. However, since the uncertainty is in the order of a few meters, which is less than the 10m spatial resolution equivalent to 100ns pulse width. The oscillation amplitude would be reduced when two or more polarizers and detectors are used, but yet it would sacrifice the maximum detectable frequency.

 

Fig. 5. Magnitude of 25Hz along the fiber link for (a) only the piezo fiber stretcher vibrating at 25Hz; (b) both the piezo fiber stretcher and the polarization controller vibrating at 25Hz.

Download Full Size | PDF

If the piezo fiber stretcher is the only disturbing event, the maximum magnitude of the second half will be remain around -57dB. However, once the polarization controller starts to vibrate at the same frequency (f₁=f₂=25Hz), the maximum of the second half increases 16dB to about -41dB even though the time domain backscattered signal beyond the controller is very weak. As in Fig. 5(b), this sudden “jump” indicates the second event with the same frequency as the first one. Since at some locations, the SOP change of two events adds up together for a certain polarizer alignment, the maximum of the second half would always be larger than that of the first half regardless of the SOP at the second event. Without 1Hz frequency filter this “jump” would not be observable.

Using micro-processor can reduce the signal processing time significantly without going through computer for digitization and programming time, making current system response in the ms time frame, as the FFT signal processing and averaging are conducted by electronic circuits directly.

4. Conclusion

We have presented a spectral Polarization-OTDR system that provides fully distributed vibration information. Benefited to its high sensitivity, our current configuration has been demonstrated to detect up to 5kHz vibrating events without any data averaging in a 1km fiber link with 10m spatial resolution. Also, by plotting the magnitude of a certain frequency component from a FFT spectrum along the fiber position, two simultaneous disturbances with the same frequency can be successfully identified. The obtained results imply that the proposed distributed vibration sensor would find numerous applications on intrusion sensing and structure health monitoring.