Open Access
Add to BibSonomy
Abstract
A high resolution fiber-optic acoustic emission (AE) sensor using a random fiber laser (RFL) is proposed. The AE probe is undertaken by a random-gratings-based erbium-doped RFL. A narrow linewidth π-FBG is used as a wavelength locking and sensing element in the RFL. The random distributed feedback in RFL significantly extends the effective cavity length of the laser, thus reduces the thermal frequency noise of the laser and improves the resolution of AE signal. A narrow lasing operation with a 20 dB linewidth of ∼10.41 kHz and a frequency noise of ∼10 Hz/√Hz above 1 kHz is realized. The 3×3 coupler interrogation technique is used for signal demodulation. A high AE signal resolution of ∼280 fɛ/√Hz @ 1 kHz is obtained. To the best of our knowledge, this is the first time that RFL is used in the 3×3 coupler based AE demodulation scheme to improve the system resolution.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Acoustic emission (AE) detection, as an important nondestructive testing technology, has been widely used in the fields of structural health monitoring [1], mechanical engineering, geological exploration, etc. AE signals carries a lot of information related to structural damage, which usually has weak amplitude and a wide frequency range up to tens of kilohertz (kHz) to several megahertz (MHz). Thus, the AE sensors with a wide frequency bandwidth and high detection resolution are glaringly desired.
Compared with electromagnetic sensors, fiber-optic AE sensors have the advantages of anti-electromagnetic interference, small size, good durability, and feasibility in multiplexing. Fiber Bragg grating (FBG) is the most common type of fiber optic AE sensors because of its simple structure and embeddability [2,3]. However, the broadband spectrum of FBG and the performance of interrogator limit the measurement resolution, which is usually several hundreds of pico-strain (pɛ) [4]. Using a π-FBG-based sensing scheme, a dynamic resolution of ∼4 pɛ/√Hz is obtained [5]. Moreover, a slow light grating improves the resolution up to 130 femto-strain (fɛ)/√Hz [6], but it requires a very stable interrogation laser, and the slow light grating must be specially written in the deuterium loaded quartz fiber.
In addition, narrow linewidth fiber lasers, considered as a type of active gratings, can also be used in AE detection. In the scheme of intensity-modulated distributed feedback fiber laser (DFB), the minimum AE detection resolution of 2×10−6 m/s @ 5 kHz is obtained [7]. In the orthogonal polarization laser mode beat demodulation ultrasonic detection scheme of radio-frequency encoded fiber laser [8], the minimum signal detected is ∼74 µPa/√Hz @ 1 kHz. Based on phase generated carrier demodulation (PGC), the frequency response of DFB laser AE sensors is several kHz and the resolution is about ∼1 pɛ/√Hz, due to the limitation of external carrier frequency and high frequency noise of laser [9]. According to the previous research [10], the noise of optical path includes three parts: shot noise, electronic noise and laser frequency noise. The first two types of noises are very small if the frequency is much higher than the environmental noise frequency. Frequency noise is the main source of system noise, that is, reducing frequency noise becomes the key to obtain a high detection resolution.
Random fiber lasers (RFLs) with no definite oscillating cavity mechanism have demonstrated good output performance in narrow linewidth, low noise, etc. [11,12], and have been used in high-precision fiber sensing applications. In terms of the structure of fiber laser, the generation of random laser needs the random distribution feedback provided by the light localization effect. Considering that grating is a kind of effective scattering unit, the localization effect will be easy to produced when the light propagates in a single-mode fiber (one dimensional structure) with several randomly intervals identical weak reflectivity FBGs [13]. Localization is an excellent interference behavior of wave, leading most of the lasing photons to be confined in the random scattering medium, thus providing random coherent feedback [14], which is different from the determined cavity feedback of traditional fiber lasers. Various types of random gratings-based RFLs have been reported by writing a random grating structure on common passive fiber [15,16] or doped fiber [17]. Simultaneously, several filtering or mode-locking schemes are adopted to ensure a narrow lasing operation and high frequency stability [18,19]. The weak grating array with random intervals forms numerous disordered local cavities, which limits the multiple scattering of light, leading to one-dimensional light localization, and then realizes random distributed feedback [13,14]. Different from the traditional cavity fiber laser, the random distributed feedback of RFL can significantly extend the effective cavity length of the laser, which can suppress the thermally induced frequency noise. In the intensity demodulated RFL ultrasonic sensing scheme with random grating as probe [20], the estimated minimum detectable strain is ∼720 fɛ. In fact, the narrow and disordered spectral slope in grating array is easy to be affected by environmental disturbance, resulting in poor response consistency. The minimum detectable strain of 140 fɛ/√Hz @ 1 kHz is achieved in the scheme combining a Rayleigh backscatter-based RFL and the frequency-shift Pound-Drever-Hall (PDH) technology [21]. However, the tens of kilometers of optical fiber required for providing random feedback within the RFL and the complex frequency stabilization configuration cause inconvenience to realistic sensing.
In this paper, a phase modulated high resolution AE sensor with a random grating-based RFL is proposed. In the RFL, a random fiber Bragg grating array provides random feedback, an erbium fiber provides gain, and a π-FBG is included as a filter and AE sensor probe to realize a narrow linewidth laser output and the strain sensing function of the laser. A phase demodulation scheme based on the 3×3 coupler with a sampling rate of 1 MHz is used to interrogate the captured AE signals in the wide response frequency band. The narrow laser linewidth and the Lorentz envelope imposed by the random distribution feedback on the original laser frequency noise jointly suppress the laser frequency noise (especially in the high frequency), resulting in a low system bottom noise level of ∼3.4×10−7 pm/√Hz within 500 kHz, corresponding to a high detection resolution of ∼280 fɛ/√Hz, which is almost an order of magnitude better than that of the previous scheme [22].
2. Experimental setup
The experimental setup of the RFL-based acoustic emission sensor is shown in Fig. 1. The proposed RFL of random grating base has a half cavity structure, shown in the yellow dashed box in Fig. 1.
Fig. 1. The experimental setup of RFL-AE sensor. RGA, random grating array; CIR, circulator; EDF, erbium-doped fiber; PC, polarization controller; FG, function generator; WDM, wavelength division multiplexer; OC, optical coupler; PD, photo-detector; DAQ, data acquisition equipment; FRM, faraday rotator mirror, OSA, optical spectrum analyzer.
A 4 m erbium-doped fiber provides sufficient linear gain. Random distributed feedback of light is provided by a random grating array (RGA), consisting of 30 equivalent weakly reflective FBGs on a single-mode fiber (Corning-28). Each grating was etched by 248 nm ultraviolet exposure and phase mask method, and the length, center wavelength and reflectivity of each FBG are controlled to 3 mm, 1537.70 nm, and ∼6%. The phase mask is fixed on a precision electric displacement stage with a moving accuracy of 1 µm. The random intervals of RGA are randomly selected and recorded by the upper computer between 0-3 mm, and the total length of RGA is 0.149 m. The polarization controller is used to adjust the polarization state of laser. The tail isolator prevents any disturbance of Fresnel reflection. At the left end of RFL, a circular mirror composed of an optical circulator (OC) provides total reflection and induces the unidirectional transmission of light, thus avoiding the spatial hole burning effect. A half-open cavity structure formed by the ring mirror and RGA increases the efficiency of optical oscillation. When the maximum gain of the sub cavity modes overcomes the total loss, the lasing behavior can occur through the mode competition. A 35 mm long π-FBG is embedded in the RFL to lock the laser within the narrow transmission peak (linewidth ∼16 MHz), and it is also used as the sensing probe of the laser.
A simulated AE test is carried out on an aluminum plate. The π-FBG, as the AE probe of RFL, its two ends are glued to the surface of the aluminum plate with dimensions of 450 mm×450 mm×6 mm. A PZT driven by the periodic sinusoidal voltage output from the function generator (FG) acts as an AE source (called, PZT-S). By disturbing the probe, the stress wave forces the laser to produce a laser frequency drift.
The configuration of the 3×3 coupler interrogation system is shown in the green dotted box in Fig. 1. The output laser goes into the unbalanced Michelson interferometer (MI) through a circulator (CIR) and a 3×3 coupler. Two Faraday rotating mirrors (FRMs) are used to eliminate the polarization fluctuation in the interferometer. The arm length difference of MI is 5 m. Then, the interference light is divided into three channels with a phase difference of 120° through a 3×3 coupler and enters the multi-channel high sensitivity photo-detector with a response bandwidth of 1 MHz. An NI FlexRIO (NI 5751 and PXIe-7972R) is used to obtain the original signal. The sampling rate is set to 1 MHz. The symmetrical demodulation algorithm is written into the field programmable gate array (FPGA) to complete the complex data calculation under high-speed sampling. A high-pass FIR filter is designed by Xilinx IP core to suppress low frequency interference in the original signal.
3. Theoretical analysis
The random feedback mechanism of RFL is guided by the light localization effect in RGA, which requires that the light localization length is shorter than that of the random medium. The light localization length can be estimated according to Eq. (1) [14]:
where T is the average transmittance of RGA, L is the length of RGA (random medium), and ξ is the light localization length.An ASE light source and an optical spectrum analyzer (APEX, AP2061A) with a resolution of 0.04 pm were used to test the spectra of RGA and π-FBG. The results are shown in Fig. 2. In order to effectively perform spectral filtering, the position of the transmission peak of π-FBG should be within the reflection band of RGA. Multitudinous disordered narrow-band chirp peaks can be found in the reflection band of the RGA (the blue curve in Fig. 2), which are caused by multiple interference and reflection of random Fabry-Perot (F-P) cavities formed in any grating pair. Therefore, to a certain extent, the grating array can be regarded as the envelope of narrow-band filter superimposed by many F-P interferometers [23], which is beneficial to prove the generation of random local modes.
The parameter T of the average transmission coefficient of the RGA can be calculated by Eq. (2):
where, n is the relative value of the transmission power characterized by the transmission spectrum of the RGA, that is, the relative transmission depth of RGA, and the unit is dB. n is about ∼ 6.2 dB from the transmission spectrum of grating array (the red curve in Fig. 2.), and T(L) is estimated to be ∼ 0.239. According to Eq. (1), the light localization length ξ can be calculated as 52.05 mm, which is shorter than that of the array. The light localization can exist in this random laser. Note that the random feedback of this RFL is wavelength dependent, that is, light localization only occurs in the reflection band of RGA.Perturbation of the temperature and strain on the π-FBG will cause the wavelength of its transmission peak to change, forcing RFL to generate a corresponding frequency shift. The AE signal is essentially an elastic wave that propagates in the form of surface waves (Rayleigh waves) on a thin aluminum plate. When the two ends of π-FBG are bonded to the aluminum plate, the AE wave modulates the refractive index and geometric length of π-FBG with dynamic longitudinal strain. In this state, the frequency response of RFL to elastic waves can be expressed by Eq. (3) [24]:
The conversion factor of the laser wavelength variation and the corresponding frequency drift can be expressed by Eq. (4):
where Δf is the frequency drift of the laser, Hz; c is light velocity, 3×108 m/s. Combining Eqs. (3) and (4) can estimate the conversion factor is ∼6.57×10−15 ɛ / Hz at 1537 nm. Generally, the amplitude of stress wave generated by PZT-S is in the nano-strain (nɛ) level. Considering the strain response of RFL, the frequency drift of laser is estimated at the level of femto-strain.4. Results and discussion
Figure 3 shows the lasing spectrum of RFL in free running state and wavelength-locked state. When the π-FBG is removed, multimode lasing occurs, as shown in Fig. 3(a). The flat gain profile of EDF and the approximate quality factor of sub-cavity modes lead to fierce modal competition. Multiple sub-cavity modes are fully oscillated and amplified synchronously, resulting in the randomness of the number and wavelength of laser. The unstable multimode lasing and mode hopping make the laser not have the consistency and robustness of high temperature and strain, so it is difficult to meet the requirements of high-resolution sensing.
In Fig. 3(b), a single lasing spectrum with a side-mode suppression ratio (SMSR) of 40 dB occurs stably near 1537.7 nm. The insertion of the π-FBG can adjust the loss distribution of the resonators and suppress the effective distribution amplification of numerous sub-cavity modes. The energy is concentrated on a few modes in the Bragg transmission peak to achieve the wavelength-locked effect.
The linewidth of RFL is measured by the beat frequency of two lasers [11]. The test result of the beat signal is shown in Fig. 4. Two beams of light from two RFLs are beat in a 50×50 coupler. The signal is then recorded in an electronic spectrum analyzer (ESA) built by An NI FlexRIO (NI 5772, and NI PXIe 7966R) via a photo-detector (response bandwidth 1 GHz). The sampling rate and frequency resolution are set to 800 MHz and 200 Hz, respectively. In the two RFLs, the wavelength difference of π-FBGs is less than 3 pm, which ensures that the beat frequency peak of the two lasers is located in the detectable band of ESA. A 20 dB linewidth of the beat signal is measured to be ∼10.2 kHz, corresponding to the RFL Lorentzian 3dB linewidth of ∼510 Hz.
To a certain extent, narrow laser linewidth means that the lower laser frequency noise can be achieved in a high frequency part, which is beneficial to obtain a high sensing resolution. However, the laser frequency noise is not determined by the lasing linewidth. In conventional cavity erbium-doped fiber lasers, the equilibrium thermal noise and non-equilibrium thermal noise, which is inversely proportional to the length of the passive part and the gain medium, respectively, are both limit the frequency stability of the output light. Although the length of the proposed RFL is relatively short, about tens of meters, the random distributed feedback greatly increases the effective cavity length of the laser by the large number of random cavities formed within the RGA and between the RGA and the ring mirror, which can significantly suppress thermally induced frequency noise.
The jitter of the beat peak between two RFLs was recorded with a time step of 0.2 s. to characterize the laser frequency stability (see the upper inset of Fig. 5). A frequency jitter of ∼± 60 kHz is maintained within 6 min, and the Allen standard deviation is calculated to be ∼3.975×10−11 within 100 s. The test result of RFL frequency noise is recorded in the red curve in Fig. 5. Although there is a small modulation peak near 30 kHz, the frequency noise of less than 10 Hz/√Hz above 1 kHz. Such a low noise level is due to the effective suppression of the original frequency noise by the Lorentz envelope line imposed by the random distribution feedback characteristic, which cannot be realized in standard cavity erbium-doped fiber laser. Simultaneously, the higher noise in the low frequency part (below hundreds of Hz) is caused by the pump noise and the millisecond intrinsic relaxation time of the EDF [25]. Considering the conversion factor of 6.57×10−15 ɛ/Hz, the ultimate resolution of RFL AE sensor can be estimated as 98.55 fɛ/√Hz above 1 kHz. Thus, the proposed RFL is expected to be used for a high-resolution AE detection.
For optical fiber AE sensor, the system noise level is the key index to measure its signal detection resolution. Generally, when the signal response level and the noise floor are equal, the minimum AE detection resolution of the system can be obtained. In order to verify the AE response reliability of the RFL AE sensor, and simultaneously to obtain the system bottom noise, a simulation experiment was carried out according to the configuration in Fig. 1. A function generator drives PZT-S to generate a 5 kHz sinusoidal stress wave to simulate AE source, and a high pass filter with a 1 kHz threshold is used to eliminate the low-frequency interference contained in the original AE signal.
Figure 6 shows the time-domain sequence and power spectral density of the response of the RFL AE sensor. In Fig. 6(a), a periodic sinusoidal signal with good waveform is presented. Accordingly, a peak signal with a signal-to-noise ratio of ∼60 dB and a frequency of 5 kHz is also obtained in Fig. 6(b). The above results verify that the proposed RFL AE sensor has the ability to accurately detect AE signals. It can be seen from Fig. 6(b) that although a low amplitude hopping peak recorded near 30 kHz is caused by the intrinsic relaxation oscillation of the erbium fiber laser, the system maintains a relatively flat and low bottom noise, 3.3×10−7 pm/√Hz within 500 kHz, corresponding to a minimum AE resolution of 280 fɛ/√Hz, considering the conversion factor of 6.57 ×10−15 ɛ/Hz.
Fig. 6. The reliability results of AE signal detection of rfl-ae sensor. (a) the response of RFL-AE sensor to continuous AE signal @ 5 kHz; (b) the power spectral density of the measured AE signal.
Furthermore, we compare the response of RFL AE sensor and a PZT sensor to the same AE signals. The schematic diagram of the test scheme is shown in Fig. 7. A reference PZT sensor (called, PZT-R) with a maximum detection frequency of 1MHz, the RFL-AE sensor and PZT-S are glued on the surface of the aluminum plate in an equilateral triangle with a side length of 200 mm, and the sensing grating of RFL is axially pointing to the AE source. The sampling rate is set to 1 MHz. Under ideal conditions, the stress waves transmitted to the PZT-R and RFL AE sensors through the aluminum plate are consistent. Figure 7(b) shows the power spectral density of two sensors for a continuous AE signal with a frequency of 50 kHz. The power spectral density of the RFL-AE sensor shows a power peak with a signal-to-noise ratio (SNR) of ∼58 dB, which is 25 dB higher than that of PZT-R sensor. As can be seen from Fig. 7(c), below 500 kHz, the overall SNRs of the RFL AE sensor to the AE signals with different frequencies remains above 50 dB. The statistical SNRs are smoothed to make it easier to observe their evolution. It is considered that the overall decreasing trend of two curves with frequency is caused by the more easily attenuated amplitude of the high frequency wave. The coupling inadequacy and the non-uniformity of the aluminum plate may limit the uniform modulation of the AE wave in the aluminum plate to the sensing probe, resulting in the curve floating.
Fig. 7. The response of RFL AE sensor and PZT to AE signals. (a) the schematic diagram of AE signals response comparison; (b) the power spectral density of the two sensors response to an AE signal @ 50 kHz; (c) the trend of SNRs with frequency between RFL AE sensor and PZT.
5. Summary
We demonstrate a high-resolution fiber-optic AE sensor based on a RFL and the 3×3 coupler interrogation scheme. A π-FBG in RFL is used as a filter and the AE probe to ensure a narrow laser with a 20 dB linewidth of ∼10.14 kHz and the AE sensing function of RFL. Due to narrow lasing linewidth and random distribution feedback, a low frequency noise of ∼10 Hz/√Hz above 1 kHz is obtained. The reliability of RFL AE sensor for AE detection is verified, and a low bottom noise of ∼3.4×10−7 pm/√Hz is obtained, corresponding to a high AE detection resolution of ∼280 fɛ/√Hz. To the best of our knowledge, this is the first time that a random grating-based RFL is used in the phase demodulated system of 3×3 coupler to obtain a high AE detection resolution, providing a new choice for weak AE signal detection in structural health monitoring and biomedical research.
Funding
National Key Research and Development Program of China (2018YFB2101003); National Natural Science Foundation of China (U1939207); The Scientific Instrument Developing Project of the Chinese Academy of Sciences(Broadband fiber optic seismometer acquisition instrument and system); Youth Innovation Promotion Association of the Chinese Academy of Sciences (2016106); Innovation Project of Shijiazhuang Tiedao University (YC2019049).
Acknowledgments
The authors thank Mr. Yuanhui Liu for the help in the experiment.
Disclosures
The authors declare no conflicts of interest.
References
1. A. Nair, C. S. Cai, and K. Xuan, “Using Acoustic Emission to Monitor Failure Modes in CFRP-Strengthened Concrete Structures,” J. Aerosp. Eng. 33(1), 04019110 (2020). [CrossRef]
2. X. Bai, M. Hu, T. Gang, and Q. Tian, “An ultrasonic sensor composed of a fiber Bragg grating with an air bubble for underwater object detection,” Opt. Laser Technol. 112, 467–472 (2019). [CrossRef]
3. H. Tsuda, “Ultrasound and damage detection in CFRP using fiber Bragg grating sensors,” Compos. Sci. Technol. 66(5), 676–683 (2006). [CrossRef]
4. L. Bin, Y. Ruan, Y. Yu, J. Xi, Q. Guo, J. Tong, and G. Rajan, “Laser Self-Mixing Fiber Bragg Grating Sensor for Acoustic Emission Measurement,” Sensors 18(6), 1956 (2018). [CrossRef]
5. D. Srivastava and B. Das, “Pico-strain-level dynamic perturbation measurement using π-FBG sensor,” (2017).
6. G. Skolianos, A. Arora, M. Bernier, and M. J. F. Digonnet, “Photonics sensing at the thermodynamic limit,” Opt. Lett. 42(10), 2018–2021 (2017). [CrossRef]
7. T. Yang, Y. Song, W. Zhang, and F. Li, “Acoustic emission detection using intensity-modulated DFB fiber laser sensor,” Chin. Opt. Lett. 14(12), 120602 (2016). [CrossRef]
8. B. Guan, L. Jin, L. Cheng, and Y. Liang, “Acoustic and Ultrasonic Detection With Radio-Frequency Encoded Fiber Laser Sensors,” IEEE J. Sel. Top. Quantum Electron. 23(2), 302–313 (2017). [CrossRef]
9. W. Huang, W. Zhang, and F. Li, “Acoustic Emission Source Location Using a Distributed Feedback Fiber Laser Rosette,” Sensors 13(10), 14041–14054 (2013). [CrossRef]
10. J. Chow, I. Littler, G. de Vine, D. McClelland, and M. Gray, “Phase-sensitive interrogation of fiber Bragg grating resonators for sensing applications,” J. Lightwave Technol. 23(5), 1881–1889 (2005). [CrossRef]
11. M. Pang, X. Bao, L. Chen, Z. Qin, Y. Lu, and P. Lu, “Frequency stabilized coherent Brillouin random fiber laser: theory and experiments,” Opt. Express 21(22), 27155–27168 (2013). [CrossRef]
12. Y. Xu, S. Gao, P. Lu, S. Mihailov, L. Chen, and X. Bao, “Low-noise Brillouin random fiber laser with a random grating-based resonator,” Opt. Lett. 41(14), 3197–3200 (2016). [CrossRef]
13. O. Shapira and B. Fischer, “Localization of light in a random-grating array in a single-mode fiber,” J. Opt. Soc. Am. B 22(12), 2542–2552 (2005). [CrossRef]
14. V. Milner and A. Genack, “Photon Localization Laser: Low-Threshold Lasing in a Random Amplifying Layered Medium via Wave Localization,” Phys. Rev. Lett. 94(7), 073901 (2005). [CrossRef]
15. A. Zhang and L. Hao, “Random phase-shift Bragg grating-based random fiber laser with a half-open cavity,” Appl. Opt. 57(34), 10017–10021 (2018). [CrossRef]
16. J. Deng, M. Han, Z. Xu, Y. Du, and X. Shu, “Stable and low-threshold random fiber laser via Anderson localization,” Opt. Express 27(9), 12987–12997 (2019). [CrossRef]
17. Z. Guo, J. Song, Y. Liu, Z. Liu, S. Ping, and X. Dong, “Randomly spaced chirped grating-based random fiber laser,” Appl. Phys. B 124(3), 48 (2018). [CrossRef]
18. B. Hu, W. Zhang, R. Ma, J. Guo, A. Ludwig, and Y. Rao, “Wavelength locking of Er-doped random fiber laser,” Laser Phys. Lett. 16(5), 055102 (2019). [CrossRef]
19. Y. Li, P. Lu, F. Baset, Z. Ou, J. Song, A. Alshehri, V. R. Bhardwaj, and X. Bao, “Narrow linewidth low frequency noise Er-doped fiber ring laser based on femtosecond laser induced random feedback,” Appl. Phys. Lett. 105(10), 101105 (2014). [CrossRef]
20. Y. Xu, L. Zhang, S. Gao, P. Lu, S. Mihailov, and X. Bao, “Highly sensitive fiber random-grating-based random laser sensor for ultrasound detection,” Opt. Lett. 42(7), 1353–1356 (2017). [CrossRef]
21. P. Liu, W. Huang, W. Zhang, and F. Li, “Ultrahigh resolution optic fiber strain sensor with a frequency-locked random distributed feedback fiber laser,” Opt. Lett. 43(11), 2499–2502 (2018). [CrossRef]
22. X. Mao, J. B. Huang, and H. C. Gu, “Demodulation Technology of Distributed Feedback Fiber Laser Sensor Based on 3×3 Coupler,” Chin. J. Lumin. 38(3), 395–401 (2017). [CrossRef]
23. B. Hu, W. Zhang, R. Ma, J. Guo, A. Ludwig, and Y. Rao, “Quasi mode-locking of coherent feedback random fiber laser,” Sci. Rep. 6(1), 39703 (2016). [CrossRef]
24. A. Minardo, A. Cusano, R. Bernini, L. Zeni, and M. Giordano, “Response of fiber Bragg gratings to longitudinal ultrasonic waves,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 52(2), 304–312 (2005). [CrossRef]
25. R. Hui and M. O’Sullivan, Fiber Optic Measurement Techniques, Fiber Optic Measurement Techniques, (Academic, 2009).
