You are on page 1of 4

442 Vol. 42, No.

3 / February 1 2017 / Optics Letters Letter

Multi-event waveform-retrieved distributed


optical fiber acoustic sensor using dual-pulse
heterodyne phase-sensitive OTDR
XIANGGE HE,1 SHANGRAN XIE,1,2 FEI LIU,1 SHAN CAO,1 LIJUAN GU,3 XIAOPING ZHENG,1 AND MIN ZHANG4,*
1
Department of Electronic and Engineering, Tsinghua University, Beijing 100084, China
2
Current address: Max Planck Institute for the Science of Light, Staudtstr. 2, 91058 Erlangen, Germany
3
School of Physics, Peking University, Beijing 100871, China
4
Institute of Ocean Research, Peking University, Beijing 100871, China
*Corresponding author: 1606871001@pku.edu.cn

Received 3 November 2016; revised 18 December 2016; accepted 21 December 2016; posted 21 December 2016 (Doc. ID 279840);
published 19 January 2017

We demonstrate a novel type of distributed optical fiber to retrieve multiple events at the same time along the fiber is
acoustic sensor, with the ability to detect and retrieve actual also important for the practical applications of DOFAS. [7]
temporal waveforms of multiple vibration events that occur uses BOTDA system combined with a signal synthesizer to
simultaneously at different positions along the fiber. The realize the dynamic measurement for vibrations induced by
system is realized via a dual-pulse phase-sensitive optical audio speakers, while the need for averaging in the Brillouin
time-domain reflectometry, and the actual waveform is re- spectrum recovery process limits the maximum detection fre-
trieved by heterodyne phase demodulation. Experimental quency to hundreds of hertz on 100 m long fiber. The OFDR
results show that the system
p has a background noise level system is based on a wavelength scanning technique and is
as low as 8.91 × 10−4 rad∕ Hz with a demodulation signal- known for its high spatial resolution [10,11]. Although several
to-noise ratio of 49.17 dB at 1 kHz, and can achieve a techniques have been proposed for dynamic strain measure-
dynamic range of ∼60 dB at 1 kHz (0.1 to 104 rad) for phase ment [5,6], the measured strain accuracy is limited to the
demodulation, as well as a detection frequency range from με level, mainly due to the laser scanning wavelength uncer-
20 Hz to 25 kHz. © 2017 Optical Society of America tainty [8]; also, there is a lack of reports on the ability to resolve
multiple vibration events along the fiber using OFDR. For
OCIS codes: (060.2370) Fiber optics sensors; (060.2840)
Φ-OTDR systems, with the potential of waveform retrieval,
Heterodyne; (060.5060) Phase modulation; (280.1350) Backscattering. phase demodulation techniques using phase diversity [12],
https://doi.org/10.1364/OL.42.000442 additional unbalanced interferometry [13,14], and quadrature
detection [15] have been reported. The sensor demonstrated in
[13] can detect multiple dynamic perturbations within a fre-
quency range of 500 to 5000 Hz along 1 km sensing fiber, with
Distributed optical fiber acoustic sensor (DOFAS) has been a minimum detectable strain of 80 nε. [14] combines a phase-
widely used on vibration detections in several practical appli- generated carrier demodulation algorithm p with Φ-OTDR to
cations, such as intrusion detection, railway safety monitoring, achieve a noise level of 3 × 10−3 rad∕ Hz and a signal-to-noise
and seismic applications [1–3]. Several types of DOFAS ratio (SNR) of about 30.45 dB. Φ-OTDR is currently the most
have been reported, including fiber interferometry (e.g., dual promising technique to achieve multi-event waveform-
Mach–Zehnder interferometer [4]), phase-sensitive optical retrieved distributed vibration sensing; however, the relatively
time-domain reflectometry (Φ-OTDR) [1], optical frequency- low SNR at the detection end sets the bottleneck for its mea-
domain reflectometry (OFDR) [5,6] and Brillouin optical surement bandwidth and resolution.
time-domain analysis (BOTDA) [7]. A most recent compre- In this Letter, we report a dual-pulse Φ-OTDR system,
hensive review on DOFAS can be found in [8]. Currently, the which, together with a properly designed heterodyne phase
main purpose of DOFAS development is to further push demodulation algorithm, can simultaneously retrieve the actual
the limit of the parameters, including sensing length, spatial waveform of multiple vibration events with a high SNR along a
resolution, maximum detectable frequency, and minimum 470 m long single-mode fiber. The two pulses are offset in both
detectable dynamic strain. temporal and frequency domains, functioning similar as the
On the other hand, full retrieval of the waveform of the sensing and reference arms of an interferometer. Hence, the
acoustic signal can be used to gather more information on vibration-induced phase change is modulated onto the two
the temporal dynamics of the vibration events, and is poten- sidebands of the heterodyne frequency, which can be further
tially useful for several key applications [9]. Besides, the ability recovered using heterodyne demodulation. As the two pulses

0146-9592/17/030442-04 Journal © 2017 Optical Society of America


Letter Vol. 42, No. 3 / February 1 2017 / Optics Letters 443

propagate along the same fiber, they share the noise properties Ez m   S m · e jΦm · e j2πf 1 tφ1   S m−N d · e jΦm−N d
with each other. Therefore, the SNR of the demodulated signal
is greatly enhanced after heterodyne detection. Without the · e j2πf 2 tφ2  ; (3)
need for averaging, the reported
p system has a background noise
level of 8.91 × 10−4 rad∕ Hz with a SNR of 49.17 dB at where φ1 and φ2 represent the initial phases of the two pulses.
1 kHz, improved by more than one order of magnitude, The AC component of the interference intensity can be there-
compared with previous work [14]; the system can recover fore expressed as
the dynamic phase change from 0.1 to 104 rad (60 dB dynamic I s z m   jS m S m−N d j cos2πΔf t  Φt  Δφ0 ; (4)
range) with a frequency up to 25 kHz.
The working principle of the proposed DOFAS is illustrated where Δf  f 1 − f 2 is the heterodyne frequency, Φt 
in Fig. 1(a), which is based on dual-pulse heterodyne Φm − Φm−N d is the vibration-induced phase change, and
Φ-OTDR. Two pulses with the same duration, one with fre- Δφ0  φ1 − φ2 . A heterodyne demodulation algorithm shown
quency f 1 and the other with f 2 , are separated with Δt in in Fig. 1(b) [17] is then adopted to retrieve Φt.
time. The red and blue curves in Fig. 1(a) represent, respec- The experimental setup is shown in Fig. 2. A narrow line-
tively, the backscattered Rayleigh signal from pulse 1 and pulse width CW laser at 1550.12 nm (NKT Koheras BasiK E15,
2, and the detected signal is the interference between them. with a central frequency f 0 ) was separated into two paths by
Assume the sensing fiber is composed of cascaded slices with an optical coupler (OC1); then two acousto-optic modulators
thickness ΔL [16]. For each individual injection pulse, at a (AOM1 and AOM2), with frequency shifts f 1  100 MHz
distance z m  mΔL, the interference field of backscattered and f 2  100.05 MHz, respectively, were used to generate
light within a pulse can be generally written as the heterodyne pulses, with a pulse duration of 100 ns and
X−1
mN a repetition rate of f r  200 kHz. A Ld  36 m delay fiber
Ez m   E 0 γ k σ k e jθk e jφk · e −αkΔL ; (1) was placed after AOM2 to temporally offset the two pulses with
km each other. Ld was chosen to be long enough to separate the two
where E 0 is the field amplitude of the incident light; α is the pulses considering the rising time of each pulse. The spatial
fiber attenuation coefficient; γ k , σ k , and θk are, respectively, the resolution of our system is then given by w  Ld ∕2, where
polarization fading factor, the Rayleigh scattering cross section w is the pulse width. This value can be further pushed down
and the phase delay coming from the kth slice within the pulse. through reducing w and Ld by upgrading the optical compo-
φk is the dynamic phase change caused by the vibration in the nents in the setup (e.g., AOMs). The two pulses were then
kth slice, and N is the total number of slices within a pulse. combined via OC2 and were amplified by an erbium-doped
Given that all slices within a pulse share the same properties, fiber amplifier (EDFA) before injecting into a circulator. The
γ k , σ k , and θk can be considered as constants, and the dimen- fiber under test was 470 m long single-mode fiber (SMF-28).
sion of the vibration event is comparable with ΔL; Eq. (1) can Three piezoelectric ceramic transducers (PZTs) were placed at
be simplified as 170, 220, and 330 m positions, wrapped with 5, 7, and 6 m
long fiber, respectively. The PZTs were driven by signal gen-
Ez m   S m · e jΦm ; (2) erators with tunable amplitude and frequency to mimic vibra-
where S m  E 0 e −αmΔL γ
m σm and Φm represents the total
e jθm , tion events. The backscattered Rayleigh signal was amplified by
phase change induced by the vibration acting on the pulse another EDFA, and the amplified spontaneous emission noise
at position z m . When the separation between the two pulses was filtered out before being delivered into a photodetector.
(Ld , composing of N d slices) is longer than the length of each The data were sampled by a data acquisition card with a
pulse, the total field of backscattered light can be expressed as

Fig. 2. Experimental setup of a dual-pulse heterodyne Φ-OTDR


Fig. 1. (a) Working principle of the dual-pulse heterodyne system. OC, optical coupler; AOM, acousto-optic modulator; EDFA,
Φ-OTDR. (b) Heterodyne demodulation algorithm. LPF, low pass erbium-doped fiber amplifier; Cir, circulator; SMF, single-mode fiber;
filter. PZT, piezoelectric ceramic transducer; DAQ, data acquisition card.
444 Vol. 42, No. 3 / February 1 2017 / Optics Letters Letter

sampling rate of f s  20 MS∕s before being sent to phase


demodulation.
Figure 3 plots 10 consecutive typical spatial backscattered
traces (blue curves) collected in the experiment, with the rep-
etition rate of 200 kHz, and with the length equal to the fiber
length. In each measurement, M  104 traces were collected,
and then the temporal domain interference signal at position z
can be reconstructed by extracting the corresponding data point
from the spatial traces (red curve), which has the form of
Eq. (4). Finally, the heterodyne demodulation algorithm was
further applied to retrieve Φt. Actually, each channel around
position z with a length given by the spatial resolution can be
considered as an “independent sensor” for multi-event detec-
tion, with their total number estimated as a sensing fiber length
divided by the spatial resolution.
To demonstrate the performance of our system, three
different types of vibration events, with waveforms of cardinal
sinusoidal (sinc) function, burst signal, and frequency-swept
signal, respectively, were applied on PZT1, PZT2, and PZT3
simultaneously. Specifically, on PZT1, a sinc waveform with a
frequency of 300 Hz and peak-to-peak voltage (Vpp) of
500 mV was applied. The signal applied on PZT2 was a burst
waveform with 100 Hz burst rate and one-cycle sine wave of
1 kHz frequency and 500 mVpp. On PZT3, a frequency-swept
sine waveform from 300 Hz to 2 kHz, with a 20 ms sweep time
and 500 mVpp was applied.
Figure 4(a) shows the 3D spatial-temporal domain demodu-
lation results, from 100 to 400 m positions along the fiber. It
can be seen that the vibrations at 170, 220, and 330 m, cor-
responding to the positions of PZT1, PZT2, and PZT3, can be Fig. 4. Demodulated waveform when different types of vibration
clearly resolved. Figures 4(b)–4(d) compare the demodulated events were applied simultaneously on PZT1, PZT2, and PZT3.
(a) 3D plot of spatial-temporal domain demodulated signals from
signal (red curves) with the actual waveform (blue curves) ap-
100 to 400 m positions along the fiber. A comparison between the
plied on the three PZTs, respectively. The results show that the demodulated signal (red) and the actual waveforms (blue) at a position
actual signal at all positions can be well detected and correctly of (b) 170, (c) 220, and (d) 330 m.
retrieved simultaneously with almost no distortion. (The cor-
relation coefficient is 0.9976).
A sinusoidal signal with 300 mVpp and 1 kHz frequency
was then applied on PZT1 to calibrate the demodulation the waveform can be demodulated with a backgroundp noise
SNR. Figure 5(a) plots the reconstructed temporal domain level of ∼ − 61 dB rad2 ∕Hz, i.e., 8.91 × 10−4 rad∕ Hz.
signal with its power spectrum shown in Fig. 5(b). The inset Moreover a SNR of 49.17 dB is achieved. Here the SNR is
in Fig. 5(a) clearly shows the interference pattern. It can be seen calculated as SNR  10 × log10 P signal ∕P noise , where P signal
that the power spectrum has a carrier frequency of 50 kHz and P noise is the power density of the 1 kHz sinusoidal signal
(heterodyne frequency), and the vibration-induced phase and the background noise, respectively.
change is modulated to the two sidebands. Figures 5(c) and The dynamic range of our system was further measured by
5(d) show, respectively, the demodulated phase signal [Φt] applying on each PZT a sinusoidal voltage (at 1 kHz frequency)
in temporal and frequency domain. Figure 5(d) shows that with voltage varying from 5 mVpp to 7 Vpp. The blue dia-
monds, red circles, and black triangles in Fig. 6(a) show the
measurement data on PZT1, PZT2, and PZT3, respectively,
with solid lines their linear fitting. For all PZTs, the minimum
detectable amplitude of the phase change (defined by
SNR  1) is ∼0.1 rad (corresponding to ∼2.5 nε strain mea-
surement resolution, some 10 times higher than previous work
[13]), and the maximum detectable amplitude is ∼104 rad
(limited by the sampling rate of the heterodyne signal, i.e., the
pulse repetition rate f r [17]), showing a dynamic range of
60 dB. It can also be seen that for all PZTs that the measure-
ment data fit excellently with a linear function (with R 2 
0.9998, 0.9996, and 0.9999, respectively, for each PZT), with
the slope obtained as 14.73, 20.86, and 17.86 rad/V, approx-
Fig. 3. Experimentally collected backscattered traces (blue) and re- imately agreeing with the 5∶7∶6 ratio of the fiber length
constructed interference signal (red) in temporal domain. wrapped on the PZTs.
Letter Vol. 42, No. 3 / February 1 2017 / Optics Letters 445

in our system, the maximum detectable frequency is given by


half of the heterodyne frequency (50 kHz). The demodulation
algorithm requires the sampling rate of the heterodyne signal
(i.e., the pulse repetition rate f r ) to be more than four times
the heterodyne frequency, in order to ensure that both the sine
and cosine carrier waves in the demodulation algorithm [see
Fig. 1(b)] are correctly constructed. The upper range of the
heterodyne frequency is therefore limited by the pulse repeti-
tion rate, which is determined by the fiber length L via the
relation f r < c∕2nL (∼220 kHz in our case), where c is the
speed of light in vacuum, and n is the mode index of SMF.
Given f r , the minimum detection frequency is determined
by the longest time duration of data collection, thus, the num-
ber of spatial traces (M) collected in one experiment. In our
case, M  104 (limited by the memory size of DAQ), giving
a minimum detection frequency of f r ∕M  20 Hz.
In summary, we report a waveform-retrieved Φ-OTDR
acoustic sensor using double pulses and heterodyne demodu-
lation. Without any type of averaging, the system can simulta-
neously retrieve the waveform of multiple acoustic signalspwith
a background noise level as low as 8.91 × 10−4 rad∕ Hz,
offering a SNR of 49.17 dB at 1 kHz. The frequency range
of the system is measured from 20 Hz to 25 kHz, and the
dynamic range is ∼60 dB. The reported system is expected
to greatly benefit the practical applications of DOFAS.

Funding. National Natural Science Foundation of China


Fig. 5. (a) Reconstructed temporal domain interference signal and (NSFC) (61327812).
(b) its power spectrum when a sinusoidal signal with 1 kHz frequency
and 300 mVpp was applied on PZT1. (c) Corresponding demodulated
temporal domain signal and (d) its power spectrum. PSD, power
REFERENCES
spectral density.
1. J. C. Juarez, E. W. Maier, K. N. Choi, and H. F. Taylor, J. Lightwave
Technol. 23, 2081 (2005).
2. F. Peng, N. Duan, Y. J. Rao, and J. Li, IEEE Photon. Technol. Lett. 26,
2055 (2014).
3. T. Parker, S. Shatalin, and M. Farhadiroushan, First Break 32, 61
(2014).
4. S. Liang, C. Zhang, W. Lin, L. Li, C. Li, X. Feng, and B. Lin, Opt. Lett.
34, 1858 (2009).
5. H. Gabai, Y. Botsev, M. Hahami, and A. Eyal, Opt. Lett. 40, 1725
(2015).
6. S. Wang, X. Fan, Q. Liu, and Z. He, Opt. Express 23, 33301 (2015).
7. Y. Peled, A. Motil, and M. Tur, Opt. Express 20, 8584 (2012).
8. X. Bao, D. P. Zhou, C. Baker, and L. Chen, Recent development in the
distributed fiber optic acoustic and ultrasonic detection,” J. Lightwave
Technol. (to be published).
Fig. 6. (a) Demodulated amplitude versus voltage for 1 kHz fre-
9. Y. Lu, T. Zhu, L. Chen, and X. Bao, J. Lightwave Technol. 28, 3243
quency sinusoidal wave. (b) Demodulated amplitude versus driving (2010).
frequency for 100 mVpp. 10. N. Riesen, T. T. Y. Lam, and J. H. Chow, Opt. Express 21, 4017
(2013).
11. N. Riesen, T. T. Y. Lam, and J. H. Chow, Meas. Sci. Technol. 25,
125102 (2014).
Finally, we measured the frequency response of the sensor by 12. A. E. Alekseev, V. S. Vdovenko, B. G. Gorshkov, V. T. Potapov, and
fixing the sinusoidal voltage acting on each PZT at 100 mVpp D. E. Simikin, Laser Phys. 24, 115106 (2014).
and varying the driving frequency. Figure 6(b) plots the 13. A. Masoudi, M. Belal, and T. P. Newson, Meas. Sci. Technol. 24,
demodulated amplitude versus frequency. Again, the blue 85204 (2013).
diamonds, red circles, and black triangles represent the exper- 14. G. Fang, T. Xu, S. Feng, and F. Li, J. Lightwave Technol. 33, 2811
(2015).
imental data from PZT1, PZT2, and PZT3, respectively. The
15. Z. Wang, L. Zhang, S. Wang, N. Xue, F. Peng, M. Fan, W. Sun,
increase of amplitude starting from ∼10 kHz arises from X. Qian, J. Rao, and Y. Rao, Opt. Express 24, 853 (2016).
resonance peak of the PZTs we used in the experiment 16. J. Park, W. Lee, and H. F. Taylor, Proc. SPIE 3555, 49 (1998).
(∼24 kHz). It can be seen that our system can properly 17. L. Wang, M. Zhang, X. Mao, and Y. Liao, Proc. SPIE 6292, 62921E
demodulate the signal from 20 Hz to 25 kHz. As a discussion, (2006).

You might also like