Optimization and Miniaturization of a Fiber-Optic φ-OTDR

(1)

Degree Project in Engineering Physics, Master Level SK202X TRITA-SCI-GRU 2018:278

Optimization and Miniaturization of a Fiber-Optic φ-OTDR

Distributed Vibration Sensor

Master’s Thesis in Applied Physics, KTH

Author:

Ola Sjölander

Supervisors:

Carolina Franciscangelis Walter Margulis

Examiner:

Fredrik Laurell

2018

(2)

Abstract

This thesis evaluated the performance of a portable Phase Optical Time-domain Reflectometer (φ-OTDR) system employed as a distributed vibration sensor. Firstly, the technique was studied and a literature re- view was performed. Then, a miniaturized φ-OTDR was built and tested, in order to be capable to perform on-field measurements. In order to test the capacity limits of the developed sensor, a poled fiber was inserted in the optical fiber under test and sinusoidal signals of different frequencies were applied, testifying a capability of the system to measure frequencies up to 80 kHz. Furthermore, a Fiber Bragg Grating (FBG) array was inserted in the fiber under test, aiming to investigate the interaction between both sensing techniques, which resulted in vibration measurements with higher signal-to-noise ratio and the capability to remove an amplifier from the miniaturized distributed sensor, reducing its total cost. Lastly, the portable sensor was evaluated on a field trial, where it was capable to detect vibrations in rotating bearings.

(3)

ACKNOWLEDGEMENT

Completing my master’s degree in engineering physics represents a momentous achievement and juncture in my life and I would like to sincerely thank every- one who contributed to this accomplishment and helped me along the way. In particular I would like to give a special thanks to:

Lars Kahlman, Olle Bankeström, and the whole SKF team, for providing me the opportunity to visit factory locations in order to gain valuable experiences in an application environment.

Carolina Franciscangelis and Walter Margulis, for taking on the role of being supervisors, consistently providing guidance and aid when necessary, and proofreading this thesis report.

Leif Kjellberg, for contributing with hardware solutions to practical difficulties encountered in the project.

Fredrik Laurell, for taking me on as a master’s thesis student and providing me with the contacts necessary to become involved in this project.

(4)

Acronyms

φ-OTDR Phase Optical Time-domain Reflectometer.

ADC Analogue-to-Digital Converter.

APC Angled Physical Contact.

BNC Bayonet Neill–Concelman.

BOA Booster Semiconductor Optical Amplifier.

CW Continuous Wave.

DFB Distributed Feedback.

EDFA Erbium-Doped Fiber Amplifier.

FBG Fiber Bragg Grating.

FFT Fast Fourier Transform.

FPGA Field-Programmable Gate Array.

FUT Fiber Under Test.

NA Numerical Aperture.

OTDR Optical Time-Domain Reflectometer.

RBS Rayleigh Backscattering.

SNR Signal-to-Noise Ratio.

UV Ultraviolet.

(6)

1 Introduction

Optical distributed sensing is a powerful tool in heavy industrial applications as the sensing range can be extended by many kilometers with minimal hardware modifications and the sensing element is capable of tolerating extremely harsh environments. A sensing fiber can be installed in an environment that experiences temperatures in excess of 1000^◦C and where powerful ambient electromagnetic fields are present. These are conditions not well suited for electrical point sensors.

Distributed fiber sensing has been implemented in the telecommunications industry for several decades in the form of an Optical Time-Domain Reflectome- ter (OTDR). The device exploits the Rayleigh scattering phenomenon to detect faults in an optical fiber by observing the backscattered light in a time-resolved measurement. This method can be modified by using a highly coherent light source capable of resolving the phase information contained in the reflected light.

This phase-OTDR or φ-OTDR is then sensitive to vibrations on a nanometer scale. This high-sensitivity technology can be used in a multitude of different applications such as acoustic surveillance [1] and intrusion detection [2].

This degree project investigates an existing φ-OTDR setup, evaluates its performance, and also proposes and builds a portable version of the setup.

Its capacity limits are studied through the use of a poled fiber, a component exclusively developed at RISE Acreo. Additionally, a comprehensive study of the interaction between the φ-OTDR and an FBG array is done.

Lastly, the system is evaluated based on a field trial conducted at SKF facilities in Göteborg, where it was shown that the system is capable of measuring acoustic signals in bearings.

(7)

2 Background

2.1 OTDR Working Principles

The classical OTDR utilizes an incoherent pulsed light source that is directed into the Fiber Under Test (FUT) as shown in Figure 1. The pulse is traveling at the group velocity vg in the fiber and the back end of the pulse is delayed by a time τ relative to the front end. Due to the amorphous nature of the silica glass fiber, the passing light experiences minor refractive index changes at randomly distributed locations along the fiber. One such fluctuation in refractive index occupies a very small fiber length in the order of 10 nm [3] causing a small portion of the light to be scattered due to the Rayleigh scattering phenomenon [4]. The scattered light propagates in all directions, some subset of these directions fulfill the conditions for wave guiding in the backward direction of the FUT and that portion of light will propagate back towards the input end. Pragmatically, these scattering centers may be viewed as extremely weak mirrors, and the low-intensity reflection is referred to as Rayleigh Backscattering (RBS).

FUT

Scattering Center

Figure 1: An illustration showing the basic principles of OTDR. Each scattering center produces a weak reflection of the incident pulse.

The obtained time resolved data translates to a spatially resolved signal according to the transformation

z = vgt

2 (2.1)

where z is the spatial coordinate along the fiber and t is the time after pulse launch at t = 0. By measuring the RBS one can determine the attenuation coefficient of an optical fiber as well as locate any faults such as breakpoints or bad splices.

(8)

2.1.1 Attenuation Coefficient

An optical signal is attenuated as it propagates through a fiber due to fiber loss.

The optical power at a point z along the fiber may be expressed as:

P (z) = P (0)e^−αz (2.2)

where P (0) is the launched optical power and α is the attenuation coefficient [5]. It is true that in general α is a function of z, but for simplicity the average attenuation along a spool of fiber shall be considered. As such, α is constant.

The power that is reflected via RBS from a scattering center is proportional to the forward propagating optical power at that point. Let the total attenuation coefficient be the sum of the scattering coefficient and the complementary attenuation coefficient, α = αRBS+ αC. Also assume the portion of scattered light that is coupled to the guided mode is equal to the capturing constant η which is related to the Numerical Aperture (NA) of the fiber. The detected optical power at the input end of the FUT originating from a point z along the fiber can now be described as [5]:

dP_RBS(z) = P (0)α_RBSηe^−2αzdz. (2.3) The factor 2 in the exponent is required as the launched pulse will have traveled twice the distance z to be detected at the input end. As is illustrated in Figure 2 the detected power is a sum of the optical power reflected along a fiber length equal to half the spatial pulse width, dz = ^∆z₂ =^v^g₂^τ.

A B

Probe

RBS A RBS B

Figure 2: An illustration displaying the interaction between multiple reflection sources. Reflections produced from all scattering centers within a fiber length equal to half the pulse width will partially overlap.

One now arrives at the power distribution P_RBS(z) = v_gτ

2 P (0)α_RBSηe^−2αz. (2.4)

(9)

The attenuation coefficient may now be deduced from (2.4) as the constant parameters allows one to fit the measured signal to the model PRBS(z) = Ce^−2αz where C is a collective constant.

2.1.2 Fault Detection

The issue of identifying faults in a length of fiber can often be managed by a more intuitive approach. Since the normal sources of attenuation are distributed along the entire fiber, the obtained signal has the characteristic shape of a continuous exponential decay curve. If a fault exists, such as a bad splice, the measurement will display discontinuity in the form of a sudden, sharp decline in power conceptualized in Figure 3.

Figure 3: A conceptualization of a measured OTDR signal examining a fiber containing a fault where Equation 2.1 has been used to transform the time t to the spatial coordinate z plotted on the x-axis.

By resolving the measurement in time, and thus in space, one has the ability to determine the location of the fault along a fiber. This is a very powerful tool when using many kilometers of fiber. Importantly, one must note that the accuracy with which one can locate such a fault is limited by the pulse width.

Since the received signal represents a sum over a fiber length equal to half of the pulse width, the spatial resolution is defined by half of the pulse width [6].

2.2 φ-OTDR Theory

The operational difference between a classical OTDR and a φ-OTDR lies in the coherence of the light source. In a φ-OTDR one uses a highly coherent pulsed laser light source with a coherence length at least as long as the pulse width.

This allows for the phase information of the backscattered light to be resolved.

By observing Figure 4 one understands that the overlapping coherent optical signals must not merely add in average intensity as with a classical OTDR, but instead the signal amplitude depends on the phase relation between multiple reflections originating from different scattering centers as they interfere.

(10)

A B

Probe

RBS A RBS B

Figure 4: An illustration showing how different reflections of a coherent probe pulse interfere with each other. The intensity detected at the input end is now sensitive to the phase relation between multiple reflections.

2.2.1 Basic Principles

Consider two scattering centers in a short length of fiber as previously shown in Figure 4. Reflections originating from scattering center B arrive at scattering center A while a portion of the pulse is still passing through scattering center A. In accordance with Liu et al. [7] the electric fields from the two scattering center sources can be expressed as:

E1= A1exp[−i(ωt + φ1)]

E₂= A₂exp[−i(ωt + φ₂)]

where A1,2 is the amplitude of the respective fields, ω is the angular frequency of the launched pulse, and φ1,2 is the initial phase of the respective fields. The intensity produced by interference is then

I = (E₁+ E₂)(E₁+ E₂)^†

= A²₁+ A²₂+ 2A1A2cos (φ1− φ2)

= A²₁+ A²₂+ 2A1A2cos δφ.

The complex conjugate is denoted by ^† and δφ represents the phase difference between the two fields. In a real implementation the pulse width will be far greater than the average separation of scattering centers resulting in a superpo- sition of many more than two reflection sources.

(11)

Since the scattering centers are randomly distributed along the fiber the phase relation between reflected pulses does not follow any particular pattern.

As such, the detected optical intensity also exhibits random behavior as shown in Figure 5. However, since the scattering centers are static in an unperturbed fiber, the detected signal is identical for each pulse launched into the FUT given that each pulse is identical. While it is not possible to produce exactly identical pulses, modern hardware is capable of achieving very low shot-to-shot variation. If an external disturbance causes material stress within the fiber, the scattering center separation L_AB will be affected causing the phase relation between interfering reflections to change. This process is illustrated in Figure 6.

5.21 5.22 5.23 5.24 5.25 5.26 5.27 5.28 5.29 5.3

Time [s] _×10^-5

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2

Voltage [V]

Figure 5: A measurement example showing a typical φ-OTDR response. Measurement points near the edges of the trace, corresponding to points at the beginning and end of the FUT, tend to be noisy as they are influenced by Fresnel reflections originating either from the Angled Physical Contact (APC) pigtail connector leading into the fiber or from the glass-to-air interface at the end of the fiber.

An external disturbance in the form of a vibration will impart a continuous set of stress states upon the fiber that repeat with a frequency equal to that of the vibration. Such a disturbance can therefore not only be located using the same method described in Section 2.1.2, but the frequency of the disturbance can also be measured by the temporal modulation of the φ-OTDR signal as demonstrated by a sample measurement in Figure 7.

(12)

L_AB

A B

Figure 6: An illustration of a fiber experiencing a vibrating disturbance. The disturbance is periodically changing the distance between adjacent scattering centers.

3.75 3.76 3.77 3.78 3.79 3.8 3.81 3.82 3.83 3.84 3.85

Time [s] _×10^-7

0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28

Voltage [V]

Stress State 1 Stress State 2 Stress State 3

Figure 7: A sample measurement showing a short segment of the φ-OTDR signal for three different arbitrary stress states imparted on the corresponding location in the fiber by a mechanical vibration.

2.2.2 Light Source Coherence

As previously mentioned, the coherence length of the light source must exceed the launched pulse width. While it is practical to impose a precise quantitative condition for the quality of the light source in this way, it does not accurately describe the dependency. Shi et al. explains that a narrow laser line width contributes significantly to the signal stability, and a coherence length far greater than the pulse width continues to increase the Signal-to-Noise Ratio (SNR) due to improved spatial coherence [8].

(13)

Importantly, a subtly related attribute is the temperature dependence of the peak wavelength emitted from the light source. A particular laser may be capable of maintaining an exceptionally narrow line width, but if temperature fluctuations or other disturbances cause the laser emission line to move in frequency-domain the detriment to the spatial coherence is equivalent to that caused by having a wider laser emission line.

2.2.3 Pulse Width

A trade-off evaluation must be made when determining the desired pulse width.

As described by Equation 2.4, the signal amplitude increases as the pulse width increases. This is an important quantity to ensure a sufficient SNR. However, section 2.1.2 explains that the spatial resolution is limited by the pulse width, meaning that an improvement in spatial resolution comes at the cost of the SNR.

2.2.4 Pulse Repetition Rate

The Nyquist-Shannon sampling theorem states that a sampling rate S is capable of resolving a maximum frequency of^S₂ [9]. In a φ-OTDR the state of the FUT is sampled with a frequency equal to the pulse repetition rate. In order to expand the detection bandwidth one must then increase the pulse repetition rate. Here the upper limit is determined by the length of the FUT. In order to have a true spatial resolution, the source of a detected signal must be unambiguous despite the lack of knowledge about the scattering center distribution.

Suppose that two short pulses enter the FUT starting at t = 0 and they are separated by a delay ∆t. A scattering center located at z = v_gT sends a reflection of the first pulse back towards the input that is detected at time t = 2T . Let a second scattering center exist at z = v_g(T − ^∆t₂ ). The second pulse will arrive at this scattering center at

t = ∆t + T −∆t

2 = T +∆t 2 , thus being detected at

t = (T +∆t

2 ) + (T −∆t 2 ) = 2T.

To avoid successive pulses interfering in this way one must have v_g(T −∆t

2 ) < 0 ⇒ ∆t > 2T, ∀T (2.5) The largest possible T is Tmax= _v^L

g where L is the length of the FUT. In other words, the maximum allowed pulse repetition rate is

rmax= vg

2L (2.6)

(14)

and therefore the maximum resolvable frequency is fmax= rmax

2 = vg

4L. (2.7)

2.3 Project Starting Point

2.3.1 Experimental Setup

The pre-existing φ-OTDR uses the core setup as shown in Figure 8. The tunable Continuous Wave (CW) laser source output passes through a polarization controller and is then modulated by a Booster Semiconductor Optical Ampli- fier (BOA) with the help of an electric function generator in order to produce square optical pulses. The pulses are then amplified via an Erbium-Doped Fiber Amplifier (EDFA) and launched through input channel 1 of the circulator, then exiting into the FUT from channel 2. The RBS traveling back into the channel 2 input is then outputted in channel 3 before being amplified using a second EDFA. The amplified signal is then filtered to isolate the source wavelength before finally reaching the photodetector. A specially designed high NA fiber is used to maximize the RBS intensity in the FUT.

Tunable Laser Source BOA EDFA

EDFA

Circulator

Tunable Bandpass

Filter Avalanche Photodetector Function

Generator

SMF-28 Optical Fiber RBS

Coaxial Cable Polarization

Controller

1 2

3

ACREO High NA Single Mode Fiber

Figure 8: A schematic illustrating the equipment used in the initial φ-OTDR setup as well as the most basic principles of operation.

In order to perform quantitative real-time frequency measurements a sample and hold device is introduced to the setup and synchronized with the function generator. The device is shown in Figure 9. With it, one can examine a short segment of the recorded φ-OTDR trace corresponding to a particular length of fiber that may be controlled by adjusting the launched pulse width or trigger delay of the sample and hold. One may then observe the signal oscillations at that specific location using an oscilloscope.

An alternative way to obtain quantitative measurements is to post-process the signal using an Analogue-to-Digital Converter (ADC) [10] with an integrated Field-Programmable Gate Array (FPGA) evaluation board [11] shown in Figure 10. The 500 MHz sampling rate and 2 GB memory allows the user to store the collected data and to favorably analyze the frequency response using the Fast Fourier Transform (FFT) [12] technique.

(15)

Figure 9: A photo of the sample and hold device used to observe the signal response from a short section of fiber in real-time.

Figure 10: A photo of the ADC and FPGA device used to record high sample rate data for post- processing.

2.3.2 Performance

Evaluating the performance of the φ-OTDR is done by determining the maximum resolvable frequency. Such a measurement is performed by applying a controlled disturbance of a known frequency to the fiber using a mini-shaker device [13] connected to a function generator. The fiber is attached to the mini- shaker by the use of tape. Using the initial setup together with the sample and hold device, frequencies of a few hundred hertz can be detected with high con- sistency and contrast. Figure 11 displays a measurement of an applied 400 Hz disturbance. During these measurements a pulse width of 10 ns and a pulse repetition rate of 500 kHz was used. This implies that one theoretically expects to be capable of detecting frequencies of up to 250 kHz according to Equation 2.7.

As one attempts to measure higher frequencies, the signal amplitude decreases significantly and the signal stability deteriorates because other weak perturba- tion sources now heavily modulate the detected signal. Shown in Figure 12 is a

(16)

successful measurement of a 2 kHz vibration, and this is the highest frequency accurately detected using this setup.

-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05

Time [s]

-0.3 -0.2 -0.1 0 0.1 0.2

Voltage [V]

Figure 11: A measurement showing the sample and hold response obtained by attaching the sensing fiber to the mini-shaker applying a 400 Hz vibration.

-0.01 -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 0.01 Time [s]

-0.03 -0.02 -0.01 0 0.01 0.02 0.03

Voltage [V]

Figure 12: A measurement showing the sample and hold response obtained by attaching the sensing fiber to the mini-shaker applying a 2 kHz vibration.

The discrepancy between the highest detected frequency and the theoretical detection bandwidth is likely not due to a design flaw or use of improper technique, but rather the mechanical issue of vibration coupling between the shaker and the sensing fiber. This is investigated further in Section 4.

3 Miniaturization

One significant limitation of the initial φ-OTDR setup is that it utilizes overly sophisticated and bulky equipment making it difficult to move. This is an important issue as performing field tests are essential to the development and

(17)

implementation of a device such as this. In Figure 13, an early version of the φ-OTDR setup is shown, depicting the approximate size of the setup before miniaturization. The main product of this work is to replace some of the existing components with smaller, specialized components to allow a single person to move the setup while at least retaining equal performance.

Figure 13: A photo conveying the large size of the pre-miniaturized setup.

3.1 Laser Source

The initial setup utilizes the Ando AQ4321A Tunable Laser Source [14] viewed in Figure 14.

Figure 14: A photo of the tunable laser source used in the initial setup.

This piece of equipment overperforms in the context of a typical φ-OTDR as the

(18)

operating wavelength does not have to be chosen with such accuracy. The sophisticated functionality of this instrument comes at the cost of its size, therefore a smaller Greatway Technology GLD200 Coaxial Pigtail 1550 nm Distributed Feedback (DFB) Laser [15] may be used instead. The laser is powered by a 15 mA input provided using a 9 V battery to achieve an approximate output power of 1 dBm. A plastic case houses the laser for thermal isolation purposes, thus increasing the stability of the output power and peak emission wavelength.

Figure 15 demonstrates the miniaturized configuration.

Figure 15: A photo of the DFB laser source used in the miniaturized setup.

3.2 BOA

The Thorlabs BOA1004PXS [16] is responsible for the generation of optical pulses and can be seen in Figure 16. This component performs sufficiently well with respect to both the φ-OTDR technical requirements as well as the setup size.

(19)

Figure 16: A photo of the BOA used in the initial setup.

3.3 EDFA

At both the transmission and receiver end of the setup a Fiberamp-BT 17 EDFA [17] is used. The device may be seen in Figure 17.

Figure 17: A photo of the EDFA device used in the initial setup.

This part of the setup may be reduced in size significantly by replacing the BT 17 amplifiers with two copies of the Optilab 25 dB Gain Pre-Amp EDFA Module [18] shown in Figure 18.

(20)

Figure 18: A photo of the EDFA device used in the miniaturized setup.

3.4 Tunable Bandpass Filter

The DiCon TF-1550-0.8 Manually Tunable Bandpass Filter [19] is implemented in the original setup and can be viewed in Figure 19. A filter is necessary to counteract the noise amplification caused by the EDFAs due to the received RBS signal being very small. This component is not a concern with respect to the miniaturization process.

Figure 19: A photo of the tunable bandpass filter used in the initial setup.

3.5 Avalanche Photodetector

Detection of the optical signal is handled by a Thorlabs APD430C/M-InGaAs Variable-Gain Avalanche Photodetector [20] shown in Figure 20. This detector is of high quality and is sufficiently small such that it needs not be replaced.

(21)

Figure 20: A photo of the avalanche photodetector used in the initial setup.

3.6 Function Generator

Modulation of the optical signal is controlled by the Tektronix AFG3252 Func- tion Generator [21] viewed in Figure 21. While this component is a significant contributor to the large size of the setup, the customization options and flexibil- ity of the device justifies its inclusion in the system at its current state. As more development progress is made and the optimal configuration is approached, the device may be replaced by a specialized circuit, however that is beyond the scope of this thesis project.

Figure 21: A photo of the electric function generator used in the initial setup.

(22)

3.7 End Result

In addition to replacing individual components, the miniaturized setup is collected and fastened onto a two-level rack as shown in Figure 22.

Figure 22: A photo of the miniaturized setup mounted on the two-level rack. On top is the DFB laser and battery, the BOA and BOA driver, photodetector, tunable bandpass filter, and polarization controller. On the bottom are both mini-EDFAs and enough free space to include either the sample and hold device or FPGA evaluation board. This image does not include the function generator.

The composed piece of equipment has the approximate dimensions 40 cm × 30 cm × 40 cm and weighs less than 10 kg.

4 Exploration

4.1 Fiber Bragg Gratings

In place of the naturally occurring scattering centers that exist in a fiber, it is possible to incur similar backreflections by locally modifying the refractive index inside the fiber core. One method of exploiting this in a φ-OTDR is to introduce an array of FBGs.

An FBG is a series of small refractive index changes in the order of 10⁻⁴ [22, 23] etched in a fiber via UV-radiation treatment as illustrated in Figure 23.

The FBG period Λ is such that reflections originating from successive interfaces

(23)

interfere constructively in the backward propagating direction, this occurs for the wavelength λ = 2n0Λ [23] and results in a very high reflectivity for the resonant wavelength to be achieved. The reflectivity profile of a typical FBG can be visualized in Figure 24. This example displays a maximum reflection for approximately 1550 nm implying a separation Λ of roughly 500 nm.

δ Λ

n

z

n0

n0+∆n

n₀+∆n n₀

Figure 23: An illustration showing the basic structure of an FBG and how the produced refractive index variation is designed.

In this case the refractive index variation occurs over a length δ = ^Λ₂ which is comparable to the wavelength of the incident pulse, thus Rayleigh scattering is no longer a prevalent phenomenon and instead light is reflected via Fresnel reflection.

Figure 24: The reflection spectrum of an arbitrarily chosen FBG serving as an example to show the typical behavior of an FBG.

By using a probing wavelength that is far away from the resonant wave-

(24)

length one obtains reflections that are still of very low intensity relative to the launched intensity, allowing for the launched pulse to pass through many FBGs with approximately equal intensity, but of significantly higher intensity relative to the reflections produced via Rayleigh scattering in an untreated fiber. The higher signal intensity reduces the need for amplification. This method is still viable when using an FBG array consisting of gratings with unidentical reflection spectra. One must simply choose a wavelength that avoids the collective reflection peaks of the FBG array. The following text will present two different probing methods given an array of N equidistant FBGs.

4.1.1 Single Pulse Theory

An FBG array can be probed with a single pulse. The pulse is launched into the fiber as shown in Figure 25 and is partially reflected at each FBG location. In this context, since the grating length Lg∼ 1000Λ is so short relative to the FBG spacing ∆L, ∆L being in the order of 1 m, each grating can be seen as a point reflection site. One understands that if the incoming pulse width is greater than 2∆L, the reflection of the front end of the pulse produced by F BG_n+1 will arrive at F BG_n while the back end of the pulse is still passing through F BG_n. This produces an interference signal modulated by perturbations of the fiber section between the interfering gratings. As the signal is modulated by the phase transformation that occurs along the entire fiber length between each FBG, it implies that the spatial resolution is equal to ∆L.

It is important to note that the pulse width may be a maximum of 3∆L in order for the interference signal to exclusively contain information about the fiber length between two successive FBGs. Having a wider pulse will result in the interference of reflections originating from three or more sources. Such a configuration is still viable for measuring disturbances in the relevant fiber length and the wider pulse still increases the signal amplitude in the way described by Equation 2.4, however the spatial resolution is once again limited by the pulse width rather than the FBG spacing.

(25)

Reflection Reflection Probe

I

t

Detector Response

Interference Signal

Figure 25: An illustration showing how segments of the single pulse probe overlap and interfere after reflection upon successive FBGs. For each grating pair the pulse reflections will self-interfere for a length equal to the margin with which the pulse width exceeds 2∆L.

4.1.2 Double Pulse Theory

The concept can be developed further by using a carefully designed double pulse described by Liu et al.[7], instead of a wide single pulse. Figure 26 shows that a double pulse with a back-end to back-end spacing of 2∆L and a pulse width smaller than 2∆L produces a familiar interference signal between two adjacent FBGs, however it achieves a superior contrast due to the removal of signal obtained from single source reflections. The single pulse scheme periodically produces an incoherent signal as the back end of the pulse passes through an FBG while the reflection of the front end of the pulse upon the next FBG has not yet reached that location, whereas the double pulse scheme in principle achieves a perfect overlap of reflecting signals.

The double pulse scheme results in a collected signal displaying N + 1 peaks, the first and last peak represent single source reflections. The first originates from the front pulse reflecting off of the first FBG and the last peak originates from the back pulse reflecting off of the last FBG. In between are N − 1 interference peaks, each representing the activity in the length of fiber between two

(26)

adjacent FBGs.

Reflection Reflection Probe

I

t

Interference Signal

Detector Response

Figure 26: An illustration showing how the front and back pulses interfere after reflection upon successive FBGs in a double pulse setup. The spacing of the two pulses is designed such that the successive pulses reflecting off of a grating pair will overlap perfectly.

4.1.3 Single Pulse Experiment

An experiment is conducted using a spliced array of three different FBGs spaced 1 m apart. The reflection peaks are located at 1550.0 nm, 1558.8 nm, and 1563.3 nm while the probing wavelength is selected as 1555.0 nm. A pulse width of 50 ns is used, corresponding to approximately 10 m, and a pulse repetition rate of 500 kHz is used. The experimental setup is identical to the one displayed in Figure 8 with the omission of the receiver end EDFA and the tunable bandpass filter. The EDFA is removed due to the reduced amplification requirement when using FBGs and the tunable bandpass filter is removed as it is no longer necessary considering the reduced noise amplification.

The maximum detectable frequency is tested by applying a controlled disturbance in the same way as explained in Section 2.3.2. The measured signal using the sample and hold device of a 400 Hz and 2 kHz disturbance is shown in Figure 27. Once more the 2 kHz measurement represents the limit of the system despite attaining a higher SNR than in the experiment presented in Sec-

(27)

tion 2.3.2. This further reinforces the idea that there is an issue in the coupling between the vibration source and the fiber sensor.

1

-4 -1

-3 0

-2 2 3 4

-25 -20 -15 -10 -5 0 5 10 15 20 25

Time [ms]

Voltage [V]

100

-400 -100

-300 0

-200 200 300 400

-2.5 -2.0 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5

Time [ms]

Voltage [mV]

Figure 27: Two frequency measurements, 400 Hz (left) and 2 kHz (right), using the sample and hold device demonstrating the deterioration in signal quality with increasing frequency.

4.1.4 Double Pulse Experiment

The same FBG array is examined using a double pulse that is created by the function generator using the scheme viewed in Figure 28. Both available channels are set to generate identical pulses. The channel 1 output is set as a trigger for channel 2 via a BNC T-splitter allowing the channel 1 output to also be fed to the BOA. The channel 2 output is added to the channel 1 signal using the "Add Input" connection. Adjusting the delay of either channel output also allows one to control the pulse separation.

Figure 28: An illustration showing the method used for generating a double pulse by combining the output signals of 2 channels.

Using a pulse width of about 8 ns, corresponding to approximately 1.6 m, and a pulse separation of about 10 ns, corresponding to twice the FBG spacing, one is able to produce the four expected peaks that may be seen in Figure 29.

(28)

The two center peaks are sensitive to disturbances in the fiber length between the first and the second FBG, and the second and third FBG respectively.

-25

Time [ns]

-20 -15 -10 -5 0 5 10 15 20 25

400

-100 200

0 300

100 500 600 700

Voltage [mV]

Interference Peaks

Pure First Pulse Reflection

Pure Second Pulse Reflection

Figure 29: A measurement displaying the four distinguished peaks produced when probing the triple FBG array with an accurately designed double pulse.

Due to unexpected difficulties regarding the technical equipment, a quantitative measurement of the detection bandwidth is not made using the double pulse. However, the previously presented experiments represent a strong argu- ment for the result of such a measurement to show the frequency limit 2 kHz.

4.2 Poled Fiber Optical Modulator

To gain a greater understanding of the limited detection bandwidth demonstrated in the preceding experiments, one wishes to remove the mechanical ob- stacle of ensuring strong coupling between the sensing fiber and the vibration source. One way of doing this is to use a poled fiber optical modulator to inject a detectable signal directly into the fiber.

4.2.1 Poled Fiber Theory

A poled fiber is an optical fiber in which the charged ions within the glass material have been redistributed to form a permanent internal electric field. The most prominent method used to achieve this is referred to as thermal poling, Figure 30 illustrates the process. Here, the silica glass fiber is heated, often approaching 300^◦C, in order to grant the positively charged ions high mobility within the material [24]. A powerful voltage in the order of 1 kV may then be applied to produce a new charge distribution, where positive and negative charges are separated. By subsequently cooling the glass back to room temperature while maintaining the applied voltage, the charged ions are frozen in place, yielding a permanent electric field [24, 25].

Generally the electric field dependent refractive index of a material may be expressed as:

(29)

V

Figure 30: An illustration depicting the process of ion separation in the thermal poling of an optical fiber.

n = n₀+ n_pE + n_kE² (4.1)

where n₀is the linear refractive index, n_pis the Pockels coefficient, and n_k is the Kerr coefficient [26, 27]. Isotropic materials such as glass typically have a negli- gible Pockels effect, but the presence of an internal field eradicates this isotropy and causes n_p to become far greater than n_k resulting in a linear modulation of the refractive index with respect to an applied electric field E [28, 26]. The change in refractive index changes the optical path between adjacent scattering centers, precisely like the mechanical stress caused by a vibration does. A poled fiber can then be transformed into an optical phase modulator by applying a sine wave voltage of desired frequency using a function generator.

4.2.2 Poled Fiber Experiment with High Voltage Amplifier

The φ-OTDR setup together with the sample and hold device is used to probe a length of fiber containing a poled fiber segment that is approximately 1 m long using a pulse repetition rate of 500 kHz. In order to incur a sufficiently large refractive index modulation, the poled fiber must be driven by a voltage in the order of 100 V. The output signal from the function generator is therefore routed through a high voltage amplifier shown in Figure 31 before being fed to the poled fiber modulator.

Using the setup described above, one is capable of detecting significantly

(30)

Figure 31: A photo of the high voltage amplifier used to drive the poled fiber. The input signal is amplified by a factor of 1000 for low frequencies.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Time [s] _×10^-3

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07

Voltage [V]

Figure 32: A measurement showing the sample and hold response obtained by modulating the refractive index of the sensing fiber using the poled fiber optical modulator at 20 kHz.

higher frequencies than in the measurements involving the mini-shaker. Figure 32 displays the sample and hold response to a 20 kHz modulation. Although other frequency components clearly exist, the 20 kHz signal is distinguishable.

This can be made more apparent by processing the data using the FFT technique, the result of which may be viewed in Figure 33. The noise in this type of measurement is commonly known as ’pink noise’ and exhibits a characteristic_f¹_α dependence explaining the high amplitude peaks at low frequencies [29]. With

(31)

this in mind a local maximum is readily identified at 20 kHz as this peak is not consistent with the _f¹α envelope curve. This measurement has a high contrast and allows the φ-OTDR bandwidth to be pushed further. The detection limit is found to be at 50 kHz, and this FFT is plotted in Figure 34.

0 1 2 3 4 5 6 7

Frequency [Hz] _×10⁴

0 2 4 6 8 10 12 14 16

Amplitude [arb. unit] ^1.6 ^1.8 ² ^2.2 ^2.4_×104

0 1 2 3 4 5 6 7 8 9

Figure 33: The FFT of a 20 kHz measurement. The detection peak is at a slightly higher frequency than the intended 20 kHz due to a calibration issue regarding the function generator.

0 5 10 15

0 10 20 30 40 50 60 70

Amplitude [arb. unit]

4.5 5 5.5 6

×10⁴ 0

5 10 15 20

Figure 34: The FFT of a 50 kHz measurement. Beyond this frequency the detection peak can no longer consistently be distinguished from the noise.

While a marked improvement has been demonstrated, the detection bandwidth has not begun to approach the near vicinity of the theoretical maximum.

This is in part due to new complications introduced when working with high voltage components, mainly the large amount of noise produced by the amplifier.

Figure 35 shows the contrast in quality between the function generator signal output and the high voltage monitoring signal. A related issue is the frequency dependence of the signal amplitude, higher frequencies yield a smaller signal.

This behavior is characterized by a measurement in Figure 36 and is problem- atic as it significantly reduces the SNR for high frequencies. The smaller signal necessitates higher amplification to continue to drive the poled fiber but due to noise amplification this does nothing to aid the SNR.

(32)

-1 0 1 2

Time [s] _×10^-4

-0.2 0 0.2

Voltage [V]

Function Generator Signal

-1 0 1 2

Time [s] _×10^-4

-0.2 0 0.2

Voltage [V]

High Voltage Monitoring Signal

Figure 35: A comparison between the function generator output signal and the high voltage monitoring signal for a frequency of 10 kHz. The monitoring signal is an attenuated copy of the high voltage output signal driving the poled fiber.

10 20 30 40 50 60 70 80 90 100

Frequency [kHz]

0 0.05 0.1 0.15 0.2 0.25

RMS Voltage [V]

Figure 36: A measurement displaying the frequency dependence of the high voltage signal amplitude.

4.2.3 Poled Fiber Experiment with Step-up Transformer

The above investigation may be extended by replacing the high voltage amplifier with a step-up transformer shown in Figure 37. This particular device replicates the input signal with high accuracy and produces far less noise than the high voltage amplifier. The advantage of using this device is made clear when observing the FFT of a 50 kHz signal measurement, representing the maximum detectable frequency of the previous setup. This may be viewed in Figure 38.

A choice is made to increase the pulse repetition rate to 600 kHz to increase the oversampling of the signal, thus increasing the SNR slightly. This, together with the reduced noise of the modulating signal driving the poled fiber, allows the frequency bandwidth to be extended up to 80 kHz as shown by the measurement in Figure 39.

At this point, many external sources of error have been eliminated and it

(33)

Figure 37: A photo of the step-up transformer used to drive the poled fiber. The transformer amplifies the input signal by a factor of 25 for low frequencies.

0 5 10 15

0 50 100 150 200 250

Figure 38: The FFT of a 50 kHz measurement using the step-up transformer. The contrast displayed in this measurement is noteworthy when comparing to the measurement represented in Figure 34.

seems that the 80 kHz limit is related to electrical and optical noise, such as that contributed by the instability of the laser emission line, or that generated by the 600 kHz trigger signal being fed to the sample and hold device.

(34)

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Time [s] _×10^-4

0 0.1 0.2

Voltage [V]

0 0.5 1 1.5 2 2.5

Frequency [Hz] _×10⁵

0 100 200

Figure 39: On top is the sample and hold response to an 80 kHz modulation post-processed using a digital smoothing function. On the bottom is the corresponding FFT. This measurement represents the frequency limit of the current detection setup.

5 SKF Field Trial

The portability and technical performance of the miniaturized φ-OTDR setup is tested by a field trial taking place at SKF facilities in Göteborg. The setup is transported and unpacked on site in working condition as seen in Figure 40.

Figure 40: A photo showing the portable setup installed at SKF facilities. The oscilloscope displays the φ-OTDR trace of the sensing fiber.

(35)

5.1 Field Measurements

The subject of measurement is a large steel bearing shown in Figure 41. The bearing is installed in a motor shaft designed for stress testing which is capable of controlling the rotation speed as well as both the radial and axial load applied to the bearing. The sensing fiber is attached to the bearing along the middle groove and passes over a number of holes that exist in this groove. One such hole may be seen in Figure 42. Before a rotation is applied to the bearing, the sample and hold is tuned to center around the length of fiber passing over the hole. This configuration is used because vibrations of larger amplitude will originate from the edges of a hole due to the geometric discontinuity. Additionally, the length of fiber passing over the hole is less restricted in movement and therefore has higher sensitivity.

Figure 41: A photo of the examined steel bearing installed in an SKF stress test machine. The sensing fiber is attached to the bearing under strain with the use of tape.

An electronic accelerometer is already installed and represents the current method of vibration monitoring used by SKF. The simultaneous use of these sensing methods on a single bearing allows the φ-OTDR to be evaluated by

(36)

Figure 42: A photo of one of the holes located in the middle groove of the bearing. The sensing fiber passes over the hole.

comparison in real-time. As the bearing is slowly rotated with a small axial load, the output signals from both the accelerometer and the sample and hold device can be observed and recorded on the oscilloscope. As shown in Figure 43, the correlation of the two recorded wave-forms is strong and the φ-OTDR is clearly able to monitor the relatively gentle mechanical activity. Multiple additional measurements were made on site that support the repeatability of this sensing technique, one such measurement may be viewed in Figure 44.

(37)

Finally, Figure 45 displays a measurement for the case of an increased axial load.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Time [s]

-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Amplitude [arb. unit]

Electronic Accelerometer Envelope

Figure 43: An initial comparison of the φ-OTDR sample and hold signal and the envelope of the accelerometer signal. Note that the detected peaks coincide in time-domain. The third accelerometer peak differs in character from the corresponding φ-OTDR peak, this may be due to saturation-induced ringing in the accelerometer.

0 0.5 1 1.5 2 2.5 3 3.5 4

Time [s]

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Electronic Accelerometer Envelope

Figure 44: An additional measurement under equivalent circumstances to that represented in Figure 43.

The result of this field trial proves not only that the miniaturized setup is indeed portable for the purpose of conducting field tests, but also that the φ-OTDR is capable of detecting mechanical vibrations in steel bearings.

(38)

0 0.5 1 1.5 2 2.5 3 3.5 4 Time [s]

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Electronic Accelerometer Envelope

Figure 45: This measurement displays many of he same qualities as shown in Figure 43 and 44 but for a larger axial load.

6 Conclusion

In this degree project, components of a φ-OTDR setup have been exchanged for corresponding parts of smaller size and mounted upon a single carrying unit weighing less than 10 kg granting the device portability for a single person without aid. This contributes greatly to the feasibility of performing practical measurements with the device outside of a lab environment. This prospect was also realized in the SKF field trial in addition to being able to monitor vibrations in bearings.

Furthermore, several options have been explored to evaluate the detection bandwidth of the φ-OTDR including the single and double pulse probe FBG array techniques and using a poled fiber optical modulator. The FBG techniques have been shown to work and in theory they achieve a very high SNR, in particular the double pulse scheme. However, frequencies higher than 2 kHz could not be detected. The poled fiber experiment detailed in Section 4.2.2 proves that this limitation is not due to a flaw in the sensing method, but rather the poor mechanical coupling between the vibration source and the sensing fiber.

Using a poled fiber, the setup is capable of detecting a signal of up to 80 kHz before the low-amplitude signal is masked by the present noise. The demonstrated bandwidth is more than sufficient for the detection of audible acoustic emissions such as those generated from speech. However, high-speed mechanical systems found in industrial environments commonly exhibit vibrations up into the MHz regime, thus demanding an improved performance.

Having that said, electrical vibration sensors struggle to accommodate this high frequency region as well and are otherwise significantly disadvantaged as compared to the studied distributed sensing technique. Independent electrical point sensors must be installed with appropriate spacing across the entire system being monitored. This makes the cost of such a project proportional to the size of the system and it may become exceedingly expensive. In a φ-OTDR the sensing element is distributed along an optical fiber, thus allowing the range

(39)

of the device to be increased by simply using a longer fiber, up to a maximum described by Equation 2.7. In addition to this, an optical fiber is much smaller in size and capable of tolerating harsh environments evincing high temperatures and powerful ambient electromagnetic fields. These are conditions that electrical sensors are very sensitive to.

The potential of this sensing technology is clearly valuable and the pragmatic development shows promise. Continued work on this project is certainly justified and it is only reasonable to believe in its success.

7 Future Work

The most pressing issue concerning the detection bandwidth of the φ-OTDR is the mechanical coupling between source and sensor. Although an 80 kHz signal could be detected using a poled fiber, a practical industry implementation of this sensing technology requires the ability to measure mechanical vibrations.

Alternative methods for attaching a fiber to a vibration source can therefore be explored. These methods may include using various types of glue, polymer- based lacquers, or using high-pressure direct contact. The performance of the φ-OTDR, were this issue to be resolved, could also be evaluated using a piezo- electric fiber stretcher [30]. Such a device is likely to outperform the poled fiber at high frequencies since it can be controlled using a low voltage driver.

More work may also be done with respect to the miniaturization process.

The most significant targets for future modification are the function generator and power supplies. Once the customizability of the current function generator is deemed superfluous due to the optimal configuration being approached, it may be substituted by a specialized circuit capable of generating electrical pulses according to the desired parameters. The driving voltage for each individual component may also be generated by a central power supply and distribution system, rather than having a multitude of independent power supplies. Further- more, the entire setup could be encapsulated within a case and equipped with exterior power switches in order to act as a single unit and further contribute to the manageability of the device.

Finally, FBG-based methods may be explored more thoroughly. The performance can be optimized by using an array of specially designed FBGs written into a single uninterrupted fiber. This way, one ensures that each FBG interacts with the incoming pulse in an equivalent way. Additionally, the FBG spacing may be determined with far higher accuracy and repeatability.

(40)

Bibliography

[1] Carolina Franciscangelis et al. “Real-time distributed fiber microphone based on phase-OTDR”. In: Opt. Express 24.26 (Dec. 2016), pp. 29597–

29602. doi: 10.1364/OE.24.029597. url: http://www.opticsexpress.

org/abstract.cfm?URI=oe-24-26-29597.

[2] Fei Peng et al. “Ultra-long high-sensitivity φ-OTDR for high spatial resolution intrusion detection of pipelines”. eng. In: Optics express 22.11 (June 2014). issn: 1094-4087.

[3] Carl R. Nave. Blue Sky and Rayleigh Scattering. url: http://hyperphysics.

phy-astr.gsu.edu/hbase/atmos/blusky.html. (accessed: 23.04.2018).

[4] Rüdiger Paschotta. Rayleigh Scattering. url: https://www.rp-photonics.

com/rayleigh_scattering.html. (accessed: 12.04.2018).

[5] Rongqing Hui and Maurice O’Sullivan. “Fiber Attenuation Measurement and OTDR”. In: Fiber Optic Measurement Techniques. Elsevier Academic Press, 2009. Chap. 4.3.

[6] Rongqing Hui and Maurice O’Sullivan. “Optical Low-Coherent Interferom- etry”. In: Fiber Optic Measurement Techniques. Elsevier Academic Press, 2009. Chap. 2.9.

[7] Tao Liu et al. “Phase sensitive distributed vibration sensing based on ultra- weak fiber Bragg grating array using doublepulse”. In: Optical Engineering 56.8 (Aug. 2017), p. 084104. doi: 10.1117/1.OE.56.8.084104.

[8] Leilei Shi et al. “Effect of laser linewidth on phase-OTDR based distributed vibration sensing regime”. In: vol. 9157. SPIE, June 2014. isbn:

9781628411751. doi: 10.1117/12.2059448.

[9] C.E. Shannon. “Communication in the Presence of Noise”. eng. In: Pro- ceedings of the IRE 37.1 (Jan. 1949), pp. 10–21. issn: 0096-8390.

[10] ADS54J66 Quad-Channel, 14-Bit, 500-MSPS ADC with Integrated DDC.

Texas Instruments, Inc. 12500 TI Boulevard, Dallas, TX 75243, USA, Dec. 2015. url: http://www.ti.com/lit/ds/symlink/ads54j66.pdf.

(accessed: 24.04.2018).

[11] TSW14J56 JESD204B High-Speed Data Capture and Pattern Generator Card User’s Guide. Texas Instruments, Inc. 12500 TI Boulevard, Dallas, TX 75243, USA, Jan. 2016. url: http://www.ti.com/lit/ug/slwu086c/

slwu086c.pdf. (accessed: 24.04.2018).

[12] Fast Fourier Transformation FFT. NTi Audio, Inc. url: https://www.

nti - audio . com / en / support / know - how / fast - fourier - transform - fft. (accessed: 20.04.2018).

[13] Mini-Shaker Type 4810. Brüel & Kjær Sound & Vibration Measurement A/S. Skodsborgvej 307, DK-2850 Nærum, Denmark. url: https : / / www.bksv.com/en/products/shakers- and- exciters/measurement- exciters/mini-shaker-type-4810. (accessed: 16.04.2018).

(41)

[14] Ando AQ4321A/AQ4321D Tunable Laser Source Data Sheet. Ando Elec- tric Co., Ltd. 3-484, Tsukagoshi, Saiwai-ku, Kawasaki, Kanagawa, 212- 8519, Japan.

[15] GLD200 Series DFB Laser Datasheet. Shenzhen Greatway Technology Co., Ltd. 5F West Building 2, Lihe Industrial Park, Baimang, Shenzhen, 518055, China.

[16] BOA1004PXS Specification Sheet. Thorlabs, Inc. 56 Sparta Avenue, New- ton, NJ 07860, USA, Mar. 2011.

[17] An Extended Series of Fiber Amplifiers. GN Nettest A/S. Kirkebjerg Allé 90, DK-2605, Brøndby, Denmark, Dec. 2000. url: http://sup.xenya.

si/sup/info/nettest- photonetics- yanista/GNNE_BT17_BT20.pdf.

(accessed: 11.04.2018).

[18] 25 dB Gain Pre-Amp EDFA Module, C-band. Optilab LLC. 600 E. Camel- back Rd., Phoenix, AZ 85012, USA, Jan. 2010. url: http : / / www . optilab . com / images / datasheets / EDFA - PA - M _ 05252016 _ v1 . pdf.

(accessed: 13.04.2018).

[19] Manually Tunable Bandpass Filter. DiCon Fiberoptics, Inc. 1689 Regatta Blvd., Richmond, CA 94804, USA, 2014. url: https://www.diconfiberoptics.

com/products/scd0124/SCD0124B.pdf. (accessed: 11.04.2018).

[20] APD430x Operation Manual. Version 1.3. Thorlabs, Inc. 56 Sparta Av- enue, Newton, NJ 07860, USA, Sept. 2017. url: https://www.thorlabs.

com/thorproduct.cfm?partnumber=APD430C. (accessed: 25.05.2018).

[21] AFG3000 Series Programmer Manual. Tektronix, Inc. 14200 SW Karl Braun Drive, Beaverton, OR 97077, USA.

[22] K. O. Hill et al. “Bragg gratings fabricated in monomode photosensitive optical fiber by UV exposure through a phase mask”. eng. In: Applied Physics Letters 62.10 (Mar. 1993), pp. 1035–1037. issn: 0003-6951.

[23] Rüdiger Paschotta. Fiber Bragg Gratings. url: https://www.rp-photonics.

com/fiber_bragg_gratings.html. (accessed: 18.04.2018).

[24] P. G. Kazansky, P. S. J. Russell, and H. Takebe. “Glass fiber poling and applications”. In: Journal of Lightwave Technology 15.8 (Aug. 1997), pp. 1484–1493. issn: 0733-8724. doi: 10.1109/50.618381.

[25] B. Ortega et al. “Electrooptic poled fibre switch/modulator”. In: Pro- ceedings of 2005 IEEE/LEOS Workshop on Fibres and Optical Passive Components, 2005. June 2005, pp. 416–421. doi: 10.1109/WFOPC.2005.

1462165.

[26] Rui M. Almeida. Non-Linear Optical Glasses I - Fundamentals. url:

https://www.lehigh.edu/imi/teched/OPG/lecture37.pdf. (accessed:

17.05.2018).

[27] A. Yariv and P. Yeh. “Third-Order Nonlinear Optical Processes”. In: Pho- tonics. Optical Electronics in Modern Communications. 6th ed. Oxford University Press, Inc., 2007. Chap. 8.6.

(42)

[28] Rüdiger Paschotta. Pockels Effect. url: https://www.rp- photonics.

com/pockels_effect.html. (accessed: 17.05.2018).

[29] L. M. Ward and P. E Greenwood. “1/f noise”. In: Scholarpedia 2.12 (2007).

revision #90924, p. 1537. doi: 10.4249/scholarpedia.1537.

[30] PZ2 High-efficiency Fiber Stretcher Data Sheet. Optiphase, Inc. 7652 Haskell Avenue, Van Nuys, CA 91406, USA, Nov. 2015. url: http : / / www . optiphase.com/data_sheets/PZ2_Data_Sheet_Rev_G.pdf. (accessed:

22.05.2018).

Optimization and Miniaturization of a Fiber-Optic φ-OTDR