1. Introduction

Fiber-optic distributed acoustic sensing (DAS) systems have received a lot of research interest in the recent years [1,2]. Such systems are based on detecting the randomly distributed Rayleigh backscattering of light that appears in optical fibers using interferometric or coherent interrogation techniques. By measuring the changes in this backscattering versus fiber delay the spatial distribution of strain or refractive index changes in the fiber can be obtained. Since the strain is often caused by acoustic vibrations, the term DAS is normally used for these measurements.

DAS is currently widely used for oil and gas borehole monitoring [3], border control and security applications [4] and terrestrial pipeline monitoring [5]. Future applications are expected to include railway and highway monitoring [6], cable monitoring [7] (telecom and power) and various geophysical and seismic monitoring [8,9] where long range might be beneficial. Detection of trawler activity near a subsea telecommunication cable has already been demonstrated at locations more than 50 km away from the instrument [10].

DAS interrogators are commonly based on transmitting short pulses generated by a coherent laser into the fiber. These pulses are reflected by the distributed Rayleigh scatters in the fiber and the backscattered response is detected by the interrogator. It follows that such interrogators have a spatial resolution that are essentially limited to the pulse length. When strain is applied to a fiber section the interference between light components in the first and second half of the transmitted pulse that are scattered at different positions in the fiber. When strain is applied to a fiber section, the response will fluctuate due to changes in the phase delay between the interfering Rayleigh scatters. These techniques have been named phase-sensitive optical time-domain reflectometry which is abbreviated to Φ-OTDR or Phase-OTDR.

Early versions of Φ-OTDR measured the delay dependent fluctuations in the Rayleigh backscattering intensity [11]. The sensitivity to strain in such systems will vary randomly with fiber position and optical frequency, and the response will be highly nonlinear. Thus, while this kind of backscatter intensity monitoring has proven to be useful for detection of disturbances, it is not so useful for quantitative strain measurements.

More recently, quantitative DAS techniques have been demonstrated that provide quantitative measurement with well-defined and linear responsivity to fiber strain, refractive index change or temperature. One way to achieve this is to extend the concept of Φ-OTDR techniques to allow for demodulation of the delay dependent phase of the backscattering [12,13]. The phase between adjacent resolved scattering regions, separated by one gauge length (GL), is then taken to be proportional to the fiber strain. Consequently, the spatial resolution equals the gauge length. Coherent detection may be used to improve the optical detection signal to noise ratio (SNR), while facilitating phase detection by shifting the back reflected signal to a subcarrier frequency [14]. A different method for achieving a linear strain response is to chirp the interrogation pulse over several GHz and calculating the change in strain by correlating the detected response with previous responses [15].

The self-noise of a quantitative DAS interrogator may be defined as the power spectral density (PSD) of the demodulated phase signal without any strain modulation of the fiber. In many cases the square root of this PSD is stated, with unit rad/√Hz, thus representing the root mean square (RMS) deviation of the signal within a 1 Hz bandwidth. One may calculate the corresponding self-noise in terms of strain per √Hz by dividing with the strain sensitivity 4πnζGL/c [rad/ɛ], where n is the refractive index of the fiber, ζ=0.78 is the strain optic coefficient, c is the speed of light, and ɛ is used as a unit for strain. Note that the self-noise in ɛ/√Hz scales with 1/GL, such that self-noise in terms of rad/√Hz is a better measure for self-noise with less sensitivity to gauge length. To our knowledge the lowest self-noise noise reported is 398 µrad/√Hz with a gauge length (and spatial resolution) of 40 m [16], corresponding a minimum detectable strain modulation of 1.1 pɛ/√Hz.

The maximum range with an implementation of a quantitative DAS based short coherent pulse Φ-OTDR was investigated in [17]. They report 80 km range for ITU G.652.D compliant standard single mode fiber (SSMF) and 95 km with OFS TeraWave SCUBA ultra-low loss fiber. They further increased the range to 112 km by splicing in an enhanced backscattering fiber (ENHF) at end of the sensing fiber. Since Φ-OTDR systems rely on short pulses to obtain high spatial resolutions, optical nonlinearities, such as self-phase modulation and modulation instability [18], limits the interrogation energy that can be emitted per pulse, and thus the fiber length that these systems can interrogate.

Slightly longer ranges have been achieved with inline amplification. 108 km range was demonstrated in [19] with bidirectional Raman amplification. Single-ended interrogation of 100 km fiber was demonstrated in [20] by remote pumping of sections of Erbium fiber within the sensing fiber.

One of the greatest advantages of DAS is the possibility to utilize existing telecom cables for measuring acoustic signals. In these applications, neither enhanced backscattering fiber, bi-directional Raman pumping with access to both ends of the fiber or remotely pumped Erbium fiber amplifiers are available. A better approach for increasing the range is therefore desirable.

One approach to further increase the range of quantitative DAS is to use coherent detection in combination with linear frequency modulated (LFM) optical pulses with much longer duration than used with Φ-OTDR, and thereby launching much more energy into the fiber per interrogation period [21]. Chirp compression techniques well known from radar applications [22] can be applied to the detected back reflected signal to compress the LFM signal into a short pulse. The achievable spatial resolution is then limited to ∼c/2n/SBW, where, SBW is the optical sweep bandwidth. This means that the interrogation pulse energy is no longer limited by the peak power. However, stimulated Brillouin scattering (SBS) will still limit the pulse energy. An evaluation of the long-range capabilities of an LFM-based interrogation technique was made in [23] reporting a range of 80 km with SSMF fiber without inline amplification.

In this paper we demonstrate the OptoDAS interrogator, based on LFM pulses with long duration and coherent detection in combination with a very low noise fiber DFB laser. A theoretical evaluation shows that this setup has a potential of demodulating the phase from backscattering levels that are at least 21.8 dB lower than that of Φ-OTDR, which corresponds to a 57 km increase in range for an SSMF fiber. The response fidelity and the self-noise are evaluated in a 10 km fiber with gauge lengths of 10 and 34 m. We further evaluate the maximum achievable range in both SSMF and low-loss SCUBA fiber. To our knowledge, the results show both a lower noise at short range and the longer maximum ranges than previously reported in the literature without the use of inline amplification.

2. Principle of operation and hardware setup

The OptoDAS interrogator is shown schematically in Fig. 1. Light from a low noise laser is split by a coupler into a local oscillator (LO) arm and a modulator arm. In the modulator arm, an acousto-optic modulator (AOM) shapes the light into apodised pulses with a FWHM duration T_sw that can be varied from a few µs to more than 100 µs, with a frequency shift that is linearly swept over a 90 MHz. The optical power is apodised with a lifted half-cosine at each end to produce an effective full width half maximum (FWHM) SBW of 80 MHz. The AOM is controlled from an FPGA via a DAC and an RF driver. The signal is boosted by an EDFA and transmitted via a 1 × 2 optical switch and a circulator into the sensing fiber. The shape of the generated pulse is illustrated in Fig. 2.

 

Fig. 1. The DAS experimental setup for the long range measurements in section 5. The setup for short range measurements in section 4 is similar except that the sensing fiber arrangement is replaced with that of Fig. 4. AOM: acousto-optic modulator; RF: radio frequency; EDFA: erbium doped fiber amplifier; LO: local oscillator; LB: loopback; PZT: piezoelectric transducer; VOA: Variable optical attenuator; DAC: Digital-to-analog converter; ADC: analog-to-digital converter; FPGA: field-programmable gate array.

Download Full Size | PDF

 

Fig. 2. Cosine-apodized LFM pulse used for interrogation and the resulting chirp compressed response from a discrete reflector. T_sw: FWHM sweep duration, SBW: FWHM sweep bandwidth.

Download Full Size | PDF

The switch can be set to a loopback mode where the output from the EDFA is returned directly to the receiver. This is used for sweep characterization and predistortion of the DAC signal to optimize the sweep waveform.

The Rayleigh backscattering from the sensor fiber can be described by the complex optical impulse response $r(\tau)$ vs the scattering delay τ. The shape of $r(\tau)$ originates from thermal fluctuations in the material density and structure that are frozen into the fiber during production. The backscattered signal from the sensor fiber equals the convolution $r(t )\ast {E_t}(t )$ of the optical field of the transmitted pulse ${E_t}(t )$ with the fiber response. When the fiber in a certain region is strained, $r(\tau)$ will be stretched in the corresponding region of τ. This will cause a change in the phase of the back scattered signal versus delay, which in the end should be demodulated by the DAS interrogator.

The back scattered signal is passed from the circulator to a commercially available bulk-optic optical hybrid for coherent detection. Inside the hybrid, the light is split by a polarization beam splitter into two orthogonal polarization components. Each polarization component is passed to a 2-by-4 splitter and is mixed with the LO signal ${E_{LO}}(t )$ to obtain four interference outputs with intensities that are phase shifted in steps of 90°. These interference signals are passed to detectors and converted to electrical signals using circuitry that is tailored to the bandwidth of the detected signal and optimized for high dynamic range. Electrical signals that are 180° separated in phase are then pairwise subtracted to provide balanced detection of the in-phase (I) and quadrature (Q) part of the optical field of each polarization component. The resulting signals are sampled with 300 MHz ADCs to provide a complex polarization resolved representation of the back reflected optical field.

The I and Q components for each polarization can be combined into a complex response on the form,

We choose to let the phasors in the above equation represent the magnitude and phase relative to a carrier at the center of the transmitted signal spectrum. When an AOM is used in this way, ${E_{LO}}(t )$ will typically have an offset of ∼100 MHz or more from ${E_t}(t )$. A first step in our processing of ${S_r}(t )$ is therefore to shift the signal by digital mixing such that it is centered at DC. This is followed by antialias filtering and decimation to 100 MHz sampling rate. Chirp compression [22] is then applied to the response by convolving with a matched filter, having an impulse response equal to the time reversed conjugate of the interrogation pulse $E_t^\ast ({ - \tau } )$, which is known from sweep characterization performed in the loopback mode,

For comparison, a Φ-OTDR interrogator that uses short constant-frequency interrogation pulses with duration T_p will provide a resolution equal to the pulse length, while the pulse occupies a spectral bandwidth of ∼1/T_p.

The matched filter response may contain more than 10 000 samples depending on the chosen sweep duration, and the chirp compression requires high processing capacity and efficient coding. The algorithm is therefore implemented on an FPGA to provide sustained real time output of ${S_c}(\tau )$ for the two separate polarization channels, representing the Rayleigh scattering with a spatial sampling period of ∼1 m (10 ns return delay) and ∼1.25 m resolution (1/SBW = 12.5 ns return delay).

By detecting and demodulating the backscattered signal in two polarization channels we avoid polarization fading, a problem that occurs when the polarization state of the scattered signal is close to orthogonal to the receiver channel polarization state. In our case, the sum of the signal power detected by the two polarization channels is independent on the input polarization state. Therefore, a polarization independent SNR can be obtained by combining the data from the two channels.

The FPGA extracts the phase difference of the scattering signals for every spatial sample location relative to the phase of the response from the previous pulse. Due to the random nature of the Rayleigh scattering, the magnitude of the back reflected signal will vary a lot with the sample delay. Locations where the back reflection is close to zero is said to be affected by Rayleigh fading, resulting in poor SNR of the demodulated phase at those locations. This is especially a problem near the end of long fibers, where the back scattered levels to the receiver is already marginal. To suppress this problem and improve the overall SNR we apply a spatial moving average on the phase data with a resolution that equals the wanted gauge length.

We then calculate the phase change per GL for each sample location as the difference between two samples separated by one GL. The signal is finally integrated in time to represent the spatially differentiated scattering phase, which is proportional to the fiber strain.

3. Range and performance limitations

The range limitation of the system comes essentially from the fact that we need to receive a minimum number of photons from a resolved backscattering section of the fiber per interrogation pulse to be able to detect the backscattered phase. This is a quantum noise limitation. To maximize the interrogation range we therefore want to maximize the transmitted optical energy per pulse.

For DAS interrogators that rely on short pulses to obtain a high spatial resolution the maximum pulse energy that can be transmitted into a long fiber will be limited by modulation instability (MI). This is a nonlinear process where the Kerr-induced self-phase modulation interacts with the fiber dispersion to severely deform the propagating pulses. MI occurs when the pulse peak power exceeds a threshold of approximately 23 dBm [24]. Consequently, an interrogator using 50 ns coherent pulses (providing 5 m resolution) may be able to launch ∼10 nJ per pulse before modulation instability starts to destroy the measurements.

Using LFM pulses allows the transmitted energy to be spread out on tens and hundreds of µs, and peak power limitations such as MI will no longer limit the pulse energy. The next nonlinear mechanism that will limit the pulse energy is stimulated Brillouin scattering (SBS).

For sensor fiber lengths > ∼20 km and pulse bandwidths SBW > B_SBS ≈ 20 MHz (the Brillouin gain bandwidth) our investigations indicate that the pulse power limit can be approximated as 375 nJ · SBW/B_SBS. If we assume SBW = 80 MHz (providing 1.25 m resolution) the SBS limit is ∼1500 nJ per pulse, which is 21.8 dB above the estimated maximum energy of a 50 ns pulse. This means DAS interrogators employing LFM and chirp compression should be able to interrogate through a two-way fiber loss that is at least ∼21.8 dB higher than Φ-OTDR and other interrogators that use short coherent pulses. For a typical SSMF with 0.19 dB/km one-way loss this corresponds to a 57 km increase in interrogation range.

As the difference in propagation loss to the start and the end of the sensor fiber increases, the requirements on the optical dynamic range, i.e. the ratio of the clipping power to the minimum detectable power, of the detector electronics and the ADCs increases. Coherent detection allows for simple adjustment of the dynamic range of the receiver by changing the local oscillator power. More important, the signal amplitudes on the detectors are proportional to the optical amplitudes received from the sensor fiber rather than to the optical power, as it would be with direct detection. Therefore, the electrical signal amplitude increases with 1 dB for each dB increase in optical power, as opposed to 2 dB electrical per dB optical for direct detection. This essentially allows for a doubling of the number of dBs of dynamic range for a coherent receiver system compared to direct detection.

At long interrogation ranges with low backscattered power, detector shot noise becomes the dominating noise source, and we will see that the demodulated phase noise amplitude increases with 1 dB for each dB increase in the 2-way fiber loss. Due to the 2π phase ambiguity, the tracking of the phase is based on the assumption that the phase difference between consecutive samples is less than π. As the detection SNR is reduced, the phase noise will eventually be so large that this requirement is no long fulfilled and the signal becomes useless.

At short ranges the laser phase noise will limit the demodulated phase noise performance. Figure 3 shows a schematic time-frequency representation of the reflected signal from four resolved scattering regions. For simplicity the signals are shown as if the scatters were discrete reflectors. As discussed above the delay resolution, and thus the delay spacing of the resolved channels, is ${\tau _d} = 1/SBW$. From the geometry in the figure we see that the frequency spacing is ${\nu _d} = {\tau _d} \cdot SBW/{T_{sw}} = 1/{T_{sw}}$, i.e. the inverse of the sweep duration.

 

Fig. 3. Time-frequency representation of the signal from four resolved reflectors.

Download Full Size | PDF

Any broadening of the laser spectrum will result in a corresponding vertical broadening of the channel lines in Fig. 3. To allow the reflector channels to be separated at all, the laser FWHM linewidth must be < $1/{T_{sw}}$. However, a laser with a bandwidth at this upper limit will cause a lot of noise on the demodulated data.

Parts of the laser spectrum that are separated from the center frequency by more than $1/({{T_{sw}}} )$ will cause cross talk to the neighboring channels. Noise in the laser spectrum separated by a frequency f from the center frequency will be proportional to the laser phase noise at frequency f. Cross talk from the neighboring channels will therefore cause random noise on the chirp compressed pulses that is essentially proportional to the laser phase noise at frequencies close to multiples of $1/{T_{sw}}$. The magnitude of the resulting demodulated phase noise depends on the ratio of the signal magnitude of the demodulated channel to that of the neighboring channels. Since both phase and magnitude varies randomly between the resolved reflectors, the sign and magnitude of the demodulated phase noise will also vary randomly between the channels.

The above discussion illustrates that DAS interrogation techniques employing very long LFM pulses are highly sensitive to laser phase noise, and very low noise laser sources are therefore needed to achieve low noise operation.

For the experiments presented in this paper we have used a custom built low-noise fiber DFB laser operating at 1546 nm with laser intensity noise suppression [25,26] providing phase noise < [1.8, 0.6, 2.9, 0.8] µrad/√Hz at [1,10, 100, 1000] kHz. This allows us to use long interrogation pulses without compromising the demodulated noise performance at short to medium fiber distances.

Li et.al. argued in [27] that there will be a reduction of the visibility with coherence detection when the distance exceeds the coherence length of the laser. However, this is only the case if the responses are averaged over a time period that is longer than the source coherence time. When the phase is calculated from the response of individual interrogation pulses, the only requirement on the coherence length to achieve proper interference visibility is that it must be larger than the pulse duration.

One may think that it is a requirement that the phase of the interferometer comprised by the local oscillator arm and scattering from near the end of the fiber is stable from one interrogation pulse to the next. However, this is not really a requirement, since we are only recording the phase delay of each GL, and how this phase changes with time.

While the phase noise variance per sample depends on the detected pulse energy and the laser phase noise spectrum vs. the inverse sweep duration, the demodulated noise spectral density also depends on the pulse repetition rate. In general, we may expect the demodulated noise with a given transmission loss increase by 3 dB each time the repetition rate is increased by a factor of 2. The OptoDAS interrogator can only detect scattering from one region of the fiber at the time, so the pulse repetition period should always be slightly longer than the dual pass delay of the sensor fiber, i.e. >100 µs for 10 km and >1.5 ms for 150 km sensor fiber.

4. Short range performance results

In a first experiment we demonstrate the short range performance with 10 km of SSMF fiber according to the SEAFOM industry standard procedure described in [28]. In this setup, three sections of 100 m fiber each were wound onto PZT ring transducers at the start, center and end of the fiber as shown in Fig. 4. The pulse repetition period was set to 100 µs and the pulse duration to T_sw = 5 µs.

 

Fig. 4. Fiber arrangement for short range measurements. The boxes indicate sealed containers.

Download Full Size | PDF

The response linearity was investigated by straining the PZT rings with a 500 Hz modulation signal applied from a signal generator. The strain amplitude was 25 nɛ, giving 2.3 rad phase modulation in a gauge length of 10 m. Figure 5 shows the recorded phase demodulated with 10 m gauge length at 9 positions separated by 10 m in the first PZT section. The corresponding spectra are shown in Fig. 6. The first harmonic averages to 2.3 rad with a standard deviation of 0.2 rad. The total harmonic distortion is −42 dB on average, with the worst case at −33 dB. Measurements in the remaining two PZT sections showed similar performance.

 

Fig. 5. Response from 9 positions with 25 nɛ modulation. The phase unit of the vertical axis is 5 rad.

Download Full Size | PDF

 

Fig. 6. FFT of demodulated phase at 9 positions with 25 nɛ@500 Hz modulation.

Download Full Size | PDF

The observed harmonic distortion is attributed to detuning of the reflection spectrum of each scattering region resolved by the chirp compression, which is caused by the strain modulation. This detuning modulates both the phase and amplitude of the scatter, and therefore contributes to distortion of the phase response. The fact that the processing applies an average over all resolved reflectors in the GL helps to suppress this distortion.

The instrument self-noise vs distance is shown for gauge length GL = 10 and 34 m in Figs. 7 and 8, respectively. The root-mean spectral density, i.e. square root of (variance / bandwidth), is plotted for three frequency bands: 10–50 Hz, 50–200 Hz and 200–5000 Hz. Transparent colors show the noise of the individual channels, while solid colors show the result after applying a 50 m moving average on the variance data (moving RMS). The noise in the 50–200 Hz and the 200–5000 Hz bands are quite similar in both figures, indicating a flat noise spectrum above 50 Hz at all positions. By averaging the variance in the full bandwidth above 50 Hz over the full fiber length we obtain a noise level of 134 and 89 µrad/√Hz for GL = 10 and 34 m, respectively. The corresponding strain spectral noise densities are 1.4 and 0.3 pɛ/√Hz, respectively.

 

Fig. 7. Instrument self-noise for GL=10 m in 3 frequency bands for 10 km fiber. The self-noise of individual channels are shown in transparent colors, while solid colors show a 50 m moving RMS.

Download Full Size | PDF

 

Fig. 8. Instrument self-noise for GL=34 m in 3 frequency bands for 10 km fiber. The self-noise individual channels are shown in transparent colors, while solid colors show a 50 m moving RMS.

Download Full Size | PDF

There is a significant variation in the phase noise with position, which can be explained by the stochastic nature of the Rayleigh backscattering, which affects both the magnitude of the detected signal and the sensitivity to laser noise for the individual channels. The 90% percentile of the spectral noise density is 193 µrad/√Hz for GL=10 and 120 µrad/√Hz for GL=34 m.

For the two upper bands the phase noise is lower with GL=34 m than with GL=10 m, and the reduction in strain noise levels is even larger. This reduction is attributed to the increased number of averaged spatial samples.

The noise levels are significantly higher in the low frequency 10–50 Hz band, and in contrast to the higher frequency bands, the phase noise at the lowest frequencies is generally higher for GL=34 m than for GL=10 m. However, the noise is significantly lower for the parts of the fiber inside the sealed containers containing the PZT stretchers at the start, center and end of the fiber. This is especially pronounced for the measurements with GL=34 m. It also appears that the noise depends on the position of the fiber on the spools. We therefore believe that the apparent increase in noise is mainly caused by ambient vibrations and acoustic signals acting on the sensor fiber. The increased noise for GL=34 m also supports this hypothesis, since the sensitivity to acoustic noise is proportional to GL. If we convert the low frequency noise to strain we see that the magnitude is actually either similar or lower for the longest GL.

5. Long range performance results

The long range experiment was set up as shown in Fig. 1. The noise performance was tested in two sets of experiments using 156 km of ITU G.652.D compliant standard single mode fiber (SSMF) and 198 km of OFS TeraWave SCUBA 125 fiber, the latter with specified loss of 0.155 dB/km@1550 nm. In these experiments we used a pulse repetition period of 2.1 ms, pulse duration T_sw = 100 µs, and gauge length 10 m.

The SBS threshold was determined by increasing the output power until we could not see any further increase in the back reflected power from far the end of the fiber. Above this threshold we could also observe increased demodulated noise from the first 10 to 20 km of the fiber. The time averaged output power at SBS threshold for the SSMF fiber was found to be slightly below −1 dBm, corresponding to a pulse power of 1.7 µJ. The threshold for the Scuba fiber was roughly 1 dB higher.

Figure 9 shows the reflected Rayleigh power per meter vs distance measured from the two fibers. A 200 m moving average is applied to clearly show the trend in the reflected power. The fitted trend curves (black dashed lines) assume losses of 0.188 and 0.156 dB/km for SSMF and SCUBA, respectively, and a detection noise floor at −127 dBm/m. The backscattered level from the start of the fibers are −71 and −73 dBm/m for SSMF and SCUBA, respectively. The optical dynamic range of the two measurements is therefore 57 and 55 dB. The Rayleigh coefficient in the SCUBA fiber was reported in [17] to be 3.2 dB below that of SSMF. On the other hand, the SCUBA fiber has a higher mode field diameter, leading to an SBS threshold that is ∼1 dB higher. At distances beyond 32 km the SCUBA fiber provides more reflected power than SSMF due to the lower loss.

 

Fig. 9. 200 m moving average of the reflected power from 156 km SSMF with input powers −1,0,1 dBm and 198 km SCUBA fiber with 0 dBm input power.

Download Full Size | PDF

The SSMF fiber was interrogated with −1, 0 and 1 dBm average power. The reflected Rayleigh power does not increase when the power is increased from 0 to 1 dBm, indicating that the SBS limit is reached.

The reflected power stays >3 dB above the detection noise floor up to 148 and 171 km for SSMF and SCUBA, respectively. We define this 3 dB point as the range limit for the system. Consequently, the interrogation range is increased by 23 km when using the low-loss SCUBA fiber compared to SSMF.

Figure 10 shows the instrument self-noise averaged over the band from 10 to 238 Hz vs distance for the two fibers. Again, results with the SSMF fiber are shown for three power levels. We observe that the noise increases with the power in the first 10 km due to SBS. The lowest noise was achieved near 12 km with 800 µrad/√Hz (9 pɛ/√Hz) for both fibers. The two fibers show similar noise levels below 4.5 mrad/√Hz up to 100 km, limited mainly by the laser frequency noise. Beyond 100 km shot noise starts to dominate and the low-loss SCUBA fiber shows a clear advantage.

 

Fig. 10. Instrument self-noise in the range 10–238 Hz with 200 m moving average.

Download Full Size | PDF

Within the range limits defined in Fig. 9, i.e. below 148 and 171 km, the self-noise is less than 50 mrad/√Hz.

To demonstrate demodulation of an applied signal near the range limit we removed the last spool with SCUBA fiber and added a variable optical attenuator (VOA), followed by a PZT ring with 100 m of SSMF for strain modulation at 147 km and another 5 km of SSMF. In this experiment the pulse repetition period was reduced to 1.9 ms, while the pulse duration and the gauge length were kept at 100 µs and 10 m, respectively. The VOA was first set to equalize the backscattered level from the SSMF with the end of the SCUBA fiber. Measurements were performed with this setting, and with the VOA losses further increased by 2, 3 and 4 dB. The corresponding back reflected power is shown in Fig. 11 together with the results from the full length of the SCUBA fiber. The reflected power received from SSMF at the PZT position with added VOA attenuations of 2, 3 and 4 dB correspond to the levels received from the SCUBA fiber at 160, 165 and 171 km, respectively, as indicated by the arrows.

 

Fig. 11. Reflected power vs distance with 147 km SCUBA fiber followed by a VOA and 5 km SSMF fiber. Reflected power shown for VOA settings 0, 2, 3 and 4 dB.

Download Full Size | PDF

Figure 12 shows the measured response to modulation at 12, 30 and 60 Hz with durations of 0.5 s and an amplitude of 2 rad/GL of the 100 m fiber section on the PZT stretcher positioned at 147.31 km with VOA attenuations of 0, 2, 3 and 4 dB. The responses are filtered with a high pass filter with 10 Hz cutoff. As expected, the noise in the measurements increases with the added loss, but we are still able to detect the applied fiber strain even with 4 dB VOA attenuation, corresponding to 171 km of SCUBA fiber.

 

Fig. 12. Measured time response. The PZT simulator is placed after 147 km SCUBA fiber and modulated with a sequence of frequencies 12, 30 and 60 Hz with 0.5 s durations. Data is shown in the different plots with additional VOA attenuation of 0, 2, 3 and 4 dB.

Download Full Size | PDF

The spectral responses for 9 different positions within the modulated section are shown in Fig. 13 for VOA attenuations of 0 and 4 dB. We find that there is an increase in the variation in responsivity at the modulated frequencies both with position and frequency when the VOA attenuation increases. The variation with position is about 2 dB and 7dB for 0 and 4 dB attenuation, respectively. This increased variation can be attributed to the increased number of fading spatial samples before spatial moving average is applied.

 

Fig. 13. Spectral response of the 9 PZT modulated channels separated by 10 m with VOA setting 0 dB (left) and 4 dB (right).

Download Full Size | PDF

The noise floor at the highest frequencies increases with 8 dB with 4 dB increased attenuation, which is in agreement with the added 8 dB two-pass loss through the VOA. The noise floor follows a 1/f-trend at the low frequencies, especially in the case with the highest attenuation. This increase in 1/f-trend can also be attributed to the increased fading of the detected interference signals included in the spatial averaging, since faded channels will typically have a 1/f response.

The results in Figs. 12 and 13 clearly demonstrate that the interrogator is capable of phase demodulation of real signals with a backscattering level equivalent to that of 171 km in the SCUBA fiber.

6. Conclusion

We have demonstrated a DAS instrument capable of real time interrogation of fiber strain out to 148 km in standard SMF and 171 km in low-loss SCUBA fiber with a spatial resolution of 10 m. The interrogator self-noise stays below 4.5 mrad/√Hz out to 100 km for both fibers, and then increases to ∼50 mrad/√Hz at range limit. When interrogating a 10 km fiber we achieved an rms averaged self-noise above 50 Hz of 134 and 89 µrad/√Hz with gauge length settings of 10 and 34 m, respectively. This corresponds to strain resolutions of 1.4 and 0.3 pɛ/√Hz, respectively.

These performances are achieved by using an ultra-low noise fiber laser, long linear frequency modulated pulses, coherent detection and maximizing the energy per interrogation period launched into the fiber limited by Stimulated Brillouin scattering. To our knowledge, these are the longest DAS interrogation ranges reported in the literature without the use of inline amplification.