Fiber artefact for performance evaluation of time domain distributed fiber sensor interrogators

Measurement Science and Technology PAPER • OPEN ACCESS Fiber artefact for performance evaluation of time domain distributed fiber sensor interrogators To cite this article: Xin Lu et al 2025 Meas. Sci. Technol. 36 115210 View the article online for updates and enhancements. You may also like Integration of BOTDA and -OTDR distributed fiber sensing for multiple parameters monitoring Zi Wang, Yang-En Zou, Chen-Yin Guo et al. - A wireless embedded passive sensor for monitoring the corrosion potential of reinforcing steel Sharmistha Bhadra, Douglas J Thomson and Greg E Bridges - Distributed Fiber Optic Sensing to Assess In-Live Temperature Imaging Inside Batteries: Rayleigh and FBGs Jiaqiang Huang, Laura Albero Blanquer, Charlotte Gervillié et al. - This content was downloaded from IP address 89.183.193.201 on 28/11/2025 at 10:06 Measurement Science and Technology Meas. Sci. Technol. 36 (2025) 115210 (10pp) https://doi.org/10.1088/1361-6501/ae214e Fiber artefact for performance evaluation of time domain distributed fiber sensor interrogators Xin Lu1,∗, Jesper B Christensen2, Peter J Thomas3, Guillaume Failleau4, René Eisermann5, Konstantin Hicke1and Stephan Krenek5 1Bundesanstalt für Materialforschung und -prüfung (BAM), Unter den Eichen 87, 12205 Berlin, Germany 2Dansk Fundamental Metrologi, Kogle Allé 5, 2970 Horsholm, Denmark 3Norwegian Research AS, Nygårdsgaten 112, 5008 Bergen, Norway 4Laboratoire National de métrologie et d’Essais (LNE), 29 avenue Roger Hennequin, 78197 Trappes, France 5Physikalisch-Technische Bundesanstalt (PTB), Abbestraβe 2–12, 10587 Berlin, Germany E-mail: [email protected] Received 30 July 2025, revised 4 November 2025 Accepted for publication 19 November 2025 Published 28 November 2025 Abstract Distributed fiber sensing (DFS) is a powerful tool for structural health monitoring (SHM), allowing continuous and seamless measurements of temperature and strain along the fiber. The spatial accuracy of a DFS interrogator, as a key parameter of the system, is vital for precisely locating structural perturbations or defects. Its evaluation and calibration methods however attract little attention. A fiber optic artefact based on a fiber loop is developed to evaluate distance accuracy and signal quality for both self-developed and commercial sensing systems based on Rayleigh, Raman, and Brillouin scattering effects, respectively. The measured distance is corrected to remove the influence of the pulse width. Additionally, the obtained SNRs are compared for different loop trips and pulse widths, assisting to assess signal quality for SHM applications. Keywords: distributed fiber sensing, distributed temperature sensing, fiber artefact, spatial correction, structural health monitoring 1. Introduction Distributed fiber sensing (DFS) utilizes an optical fiber as the sensing medium and relies on the backscattering effects inside the fiber to retrieve the environmental information ∗Author to whom any correspondence should be addressed. Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. continuously and seamlessly over long distances [1]. For structural health monitoring (SHM), the sensing fiber is embedded into or attached to the target structure. Based on the obtained signal, environmental parameters such as temperature and strain can be retrieved, allowing for the assessment of structural integrity. DFS systems can operate in timeor frequency-domain, with the former being particularly suitable for monitoring of large infrastructures, like pipelines [2], railway tracks [3], roads [4], bridges [5] and power cables [6], due to its long sensing distances up to tens of kilometers. Additionally, DFS has found numerous applications in harsh environments, thanks to the intrinsic properties of optical fibers, such as small size, chemical inertness and immunity to 1 © 2025 The Author(s). Published by IOP Publishing Ltd Meas. Sci. Technol. 36 (2025) 115210 X Lu et al electromagnetic interference [7]. Thus, the DFS can be applied in harsh environments to monitoring chemical process and structural response of ships [8,9]. Current DFS systems send light into the sensing fiber and collect the Raman, Brillouin and Rayleigh backscattered light. Raman based systems collect Stokes and anti-Stokes components of spontaneous Raman scattering, of which the antiStokes component is sensitive only to temperature, and the Stokes component is mainly used to compensate for fiber-loss effects [10]. Brillouin based DFS measures the Brillouin frequency shift (BFS) to quantify environmental variations like temperature and strain [11]. Rayleigh based DFS detects the interference pattern of the backscattered light and can essentially retrieve any perturbations that modify the local interference condition, including temperature [12,13]. All three types of DFS systems have been well developed and widely used in various areas. In time domain DFS measurements, the light is pulsed before entering the sensing fiber and the backscattered light is recorded as a function of time. The time-of-flight of the light is converted into distance, and the spatial resolution is mainly dependent on the pulse length. Frequency domain DFS systems employ continuous and frequency scanned incident light, and retrieve the spatial information from the beating frequency of returning signal and a local oscillator. DFS systems working in time-domain are more suitable for large infrastructure monitoring because of their longer sensing distance, so they are the focus of this paper. The performance evaluation of DFS systems is based on four criteria: spatial resolution, measurand uncertainty, sensing distance and measurement time [14]. A tradeoff exists among these criteria, and the performance is eventually governed by the SNR of the obtained signal. For example, finer spatial resolution allows for more accurate location of the external perturbations, which is crucial for locating external perturbations, especially in large infrastructure monitoring. However, a finer spatial resolution usually results in a weaker signal, affecting the sensing distance and measurand uncertainty. Various SNR enhancement methods have been proposed over the last decades to address the tradeoff, significantly improving the sensing performance of the DFS [15]. However, the development of evaluation methods for spatial information has advanced more slowly. Only a few studies have focused on the metrological assessment of distributed temperature sensors based on Raman scattering, and several characterization facilities have been developed to evaluate spatial resolution, measurand uncertainty, etc [16–18]. To the best of our knowledge, research on the metrological assessment of interrogators for Rayleigh and Brillouin scattering based DFS in terms of spatial accuracy and signal quality is scarce [19]. Given the critical role of the metrological evaluation in standardization processes [20,21], increased attention and research in this area are highly required. Furthermore, with various DFS systems available in the market, a universal assessment method is essential for end-users to fairly compare the spatial accuracy of different systems. This paper improves upon the recent conceptualization of a fiber-optical measurement artefact as a standard for assessing performance of distributed DFS interrogators [19,22]. While initially designed for Brillouin optical time-domain reflectometry (BOTDR) [19], this work extends the applicability of the artefact to a broader range of the interrogators based on all scattering effects. In addition, it has been experimentally demonstrated that the artefact can be readily used for commercial DFS systems. Most importantly, this work extends the application of the artefact to assessing e.g. signal quality and spatial accuracy in distributed sensing systems. The spatial accuracy and the SNR of the obtained signal are compared respectively for different pulse widths and number of loop trips. The proposed scheme is therefore applicable to all time-domain DFS systems and is envisioned to improve industrial quality assurance and enable more consistent benchmarking of interrogator performance. 2. Working principle 2.1. DFS in time domain Most DFS systems working in time domain send optical pulses into the sensing fiber and record the spontaneously backscattered light as the pulse travels along the fiber, which is called optical time domain reflectometry (OTDR). Three types of scattering effects inside the fiber are exploited for distributed sensing: Rayleigh, Raman and Brillouin scattering. Rayleigh scattering is caused by inhomogeneities that induce refractive index fluctuations along the fiber. This is an elastic process, meaning the scattered light retains the same optical frequency as the incident light [1]. The other two mechanisms involve phonons in the fiber and are inelastic processes, resulting in a frequency shift compared with the incident light. For example, the frequency shifts for Brillouin and Raman scattering in a standard silica fiber are approximately 11 GHz and 13 THz, respectively, when the incident light is at 1550 nm [1]. Conventional OTDR employs a low coherence pulsed light source and detects the Rayleigh backscattered light. The obtained signal becomes weaker due to the losses in the fiber, so the system is mainly used for fiber attenuation characterization and fault localization. When highly coherent light sources are employed, the light backscattered at different inhomogeneities within half the pulse length maintains a given phase relationship and interferes with each other at the photodetector [23]. Thus, the detected signal is an interference pattern that varies randomly along the fiber, and any environmental perturbation that alters the local interference condition will change the signal at the corresponding position. Such systems are usually named as phase sensitive OTDR (φOTDR) and have been greatly improved in the past decades to accurately locate and quantify temperature and strain variations [12]. DFS systems based on Raman scattering, known as Raman OTDR, detect the anti-Stokes and Stokes components of the Raman backscattered light. The intensity of the anti-Stokes 2 Meas. Sci. Technol. 36 (2025) 115210 X Lu et al increases strongly with temperature, whereas the Stokes component intensity is only weakly temperature dependent [24]. Raman OTDR determines temperature from the intensity ratio between the Stokes and anti-Stokes components, but it can also measure temperature using only the intensity of the antiStokes component [25]. This system is an excellent candidate for distributed temperature sensing because the scattering process is sensitive only to temperature and immune to interference from other environmental parameters. Similar to Raman scattering, the strength of spontaneous Brillouin scattering is proportional to the average number of thermally activated phonons. Thus, temperature can be determined using a BOTDR system by detecting the intensity of the Stokes and anti-Stokes components of Brillouin scattering. Additionally, the BFS has a linear relationship with temperature, so most Brillouin-based DFS systems use this value for measurements due to its robustness against unexpected fiber loss in practice. Most systems rely on stimulated Brillouin scattering to obtain the BFS, with a popular system being the BOTDA. For the BOTDA measurement, the optical pulse (pump) and the continuous wave (probe) are launched into the sensing fiber at opposite fiber ends, and the power of the probe light is recorded along the fiber. By scanning the optical frequency difference around the BFS and plotting Brillouin gain spectrum based on the obtained power of the probe light, accurate BFS values can be determined as the peak of the gain spectrum and used to retrieve environmental information such as temperature and strain. 2.2. Fiber optic artefact The fiber optic artefact is essentially a fiber loop consisting of a combiner and a circulator, as shown in figure 1. The optical pulse first travels along a lead fiber and enters the fiber loop via the combiner. Then it passes through the circulator from port 2 to port 3, thus it makes several trips around the loop, with the intensity being progressively reduced due to the fiber attenuation and the insertion loss of the combiner. The optical pulse is continuously scattered as it propagates along the fiber. The light backscattered in the fiber section plotted as dotted line, consisting of the lead fiber (L0) and the right part of the loop (LA), travels back to the pulse entrance. Thus, this part can be considered as the fiber under test (FUT). The backscattered light from the left part (LB) of the loop is however blocked by the circulator in the loop, so it cannot be detected, and this section can be seen as the ‘dead zone’. To avoid any spatial ambiguity, the left fiber section should be longer than the right one, i.e. LB>LA[19]. Note that each optical pulse traverses the lead fiber only once, but circulates multiple times within the fiber loop. Consequently, backscattered light from the lead fiber is recorded a single time, whereas backscattered signals from the right part of the fiber loop are detected multiple times, and will exhibit a clear end for each loop trip. The end position of the first trip should be L0+LAwhere L0is the length of the lead fiber. Then, the signal is obtained at a spatial frequency of 2⁄(LA+LB). Therefore, the signal end of the Nth trip can be expressed as L0+LA+(N−1)∙(LA+LB)⁄2. In practice the Figure 1. Fiber optic artefact for metrological assessment of distributed fiber sensing. Optical pulses travel anticlockwise in the fiber loop and the backscattered light travels clockwise. lead fiber can be kept very short, so the right part of the fiber loop is treated as the FUT. While an optical isolator can effectively block backscattered light from the left part of the loop, the circulator is deliberately employed in this configuration to enable compatibility of the artefact with BOTDA systems. Specifically, the probe signal is injected into port 1 of the circulator, while the pump pulse is launched through the lead fiber. They propagate in opposite directions and interact within the right section of the loop. Following the interaction, the probe exits the loop via the optical combiner, where it is subsequently detected. The incident optical pulse, Rayleigh and Brillouin scattered light are located within the working wavelength of all the components in the loop. Therefore, the Rayleigh and Brillouin signals can be detected at the beginning of the lead fiber. A narrowband filter can be used before photodetection to select the signal of interest but is unnecessary for the detection of Rayleigh scattered light because it is much stronger than the spontaneous Brillouin scattered light. A wavelength division multiplexer (WDM) is necessary in the loop after the combiner in order to detect the spontaneous Raman scattering because the combiner is unable to work in a broad wavelength range over 100 nm to cover the Raman frequency shift. The WDM has three reflection channels that work at the wavelength of the optical pulse, Stokes and anti-Stokes components, respectively. The common channel of the WDM is connected to the right part of the loop, and only the light backscattered between the WDM and the circulator is detected, and not for the lead fiber nor the length between the WDM and the combiner. As a result, the obtained Raman signal has a shorter fiber distance than signals of other scattering effects. 3. Experimental setup Figure 2shows the experimental setup based on a fiber optic artefact for the spatial accuracy and SNR investigation of the DFS. The pulse generation and amplification are almost the same as in many DFS systems [11,13,21]. The light source 3 Meas. Sci. Technol. 36 (2025) 115210 X Lu et al Figure 2. Experimental setup based on a fiber optic artefact for metrological investigation of distributed fiber sensing. The dash dot line represents fiber under text. Red parts denote the configuration for testing a commercial BOTDA system. is a semiconductor laser working at 1550.3 nm with a narrow linewidth of a 2.7 kHz, and the output light exhibits linear polarization oriented along the slow axis. The continuous light from the laser is gated into pulses by a semiconductor optical amplifier (SOA). The laser and the SOA are connected via polarization-maintaining fiber, ensuring that the light entering the SOA remains linearly polarized. This configuration guarantees a high extinction ratio of the generated pulses, which is critical for observing stochastic coherent Rayleigh signals and minimizing nonlocal effects in Rayleigh and Raman sensing. The pulse repetition rate is set as 10 kHz to enable signal acquisition from multiple loop trips, minimize inter-pulse interference, enhance peak power for Raman measurements, and ensure amplifier stability. The pulses are amplified by an erbium-doped fiber amplifier (EDFA) and the amplified spontaneous emission (ASE) from the EDFA is suppressed by a passband filter with a bandwidth of ±12.5 GHz. Then, the amplified pulse passes through a circulator and enters the artefact via the lead fiber as shown in figure 2. The lengths of the right and left fiber parts of the artefact (see figure 1) are LA=105.1 m and LB=173.8 m, respectively. The lead fiber (L0) is comparatively short, only ∼2 m long. The length is measured by a commercial OTDR with a spatial resolution of 30 cm. The whole setup is based on the standard single mode fiber (SMF), but the right part of the artefact (LAin figure 1) consists of three pieces of fibers from different manufacturers. The first piece after the WDM is about ∼51 m long and connected with the second (∼42 m long) by splicing; whereas the last one is comparatively short (∼10 m). This special fiber configuration is supposed to have negligible influence on the Rayleigh and Raman measurements, but the fibers have distinct BFSs, facilitating the test of Brillouin based DFS. The Rayleigh backscattered light from the artefact is redirected by the circulator in figure 2and passes through a preamplification scheme before it enters the photodetector. The scheme consists of another EDFA for pre-amplification and another filter with ±12.5 GHz bandwidth to suppress the ASE. Although Rayleigh scattered light is comparatively intense, the pre-amplification is necessary because the Rayleigh based DFS usually aims for dynamic measurement without signal averaging. Only the anti-Stokes component of the Raman scattering is detected and analyzed here because it is much more sensitive to temperature than the Stokes part [24]. The anti-Stokes component is selected by the WDM in the artefact and detected by an avalanche photodetector. The used WDM has a channel centered at 1465 nm with a bandwidth of ±25 nm. This broad spectral range enables efficient collection of Raman-scattered light, thereby enhancing the quality of subsequent analysis. The red parts in figure 2illustrate the configuration used to evaluate a commercial BOTDA system. The pump pulses from the system are directed to the input of the optical combiner, while the probe light is injected into the fiber loop via port 1 of the optical circulator. The pump and probe interact along the fiber segment indicated by the dotted line in figure 2. Then, the probe signal is collected by the BOTDA system to determine the BFS. Due to the small frequency separation between Brillouin and Rayleigh scattering, a narrowband filter is required to select the Brillouin component. This filtering is typically achieved using a fiber Bragg grating, as described in [11,14]. Various pulse widths are used to metrologically investigate the location accuracy and the signal quality. Due to the loss in the artefact, only the signal from three round loop trips can be detected and analyzed. The bandwidths of the used photodetectors for Rayleigh and Raman scattered light are 125 MHz and 100 MHz, respectively. The analog signal from the photodetector is digitized by a digitizer with a sampling rate of 250 MS s−1for digital data processing. Based on the time-offlight t, the distance can be calculated t∙c/2 n, with the speed of light in vacuum c≈2.99 ×108m s−1and the group index of the fiber n≈1.467. 4. Analysis and discussion of experimental results 4.1. Rayleigh signal The Rayleigh signals are obtained with 10 ns, 20 ns, 50 ns and 80 ns pulses, respectively. To remove noise, the obtained signal is averaged by 20 000 times. The averaged signal for the 20 ns pulse is shown in figure 3. Due to the random nature of the Rayleigh scattering, the obtained signal exhibits a stochastic profile along the fiber [26]. Nonetheless, the signal profiles for different trips are very similar because they are obtained under the same condition. However, it is obvious that the signal intensity obtained in the 3rd loop trip is much lower than that of the 1st trip due to the loss in the artefact. Owing to the Fresnel reflection at the combiner, high peaks are visible at the beginning of the right part of the loop. According to the figure, the peak interval between the 1st and the 2nd trips is ∼139 m, very close to half of the loop length, that is (LA+LB)/2 =139.45 m. The signal is obtained at the right part of the loop and the intervals between the traces are due to the left part of the loop. It is very important to set LBlarger than LA, otherwise, the signals from different trips will overlap in time domain and cannot be discriminated from each other. The gray regions indicate the signal used for quality analysis, which discards the beginning of the trip that is subject to the strong Fresnel reflection. 4 Meas. Sci. Technol. 36 (2025) 115210 X Lu et al Figure 3. Averaged Rayleigh signal with 20 ns incident pulse. 4.1.1. Location accuracy analysis for Rayleigh scattered light. The obtained signal is at first analyzed to evaluate location accuracy. As shown in figure 3, the signals at different loop trips exhibit distinct power levels due to the optical loss in the artefact. More importantly, the signal level is expected to vary with pulse width, as longer pulses introduce more optical power into the fiber. To eliminate these level differences and facilitate analysis, the averaged signals for each pulse width are normalized to its mean value within the gray region for each trip. The normalized signals are compared in figure 4for different pulse widths and different trips, respectively, with a focus on the end of each trip. Despite the random variation of the obtained signal, it is straightforward to identify the last sampling point of the signal, that is taken as the measured distances determined under different conditions, as summarized in table 1. The directly obtained end position varies with the pulse width and the fiber seems to become longer for longer pulses. This width-dependent offset is due to the additional time the longer pulse needs to travel through the loop and it can be corrected by removing the impact of the pulse width, e.g. reducing ∆t∙c/2nfrom the directly measured result, where ∆trepresents the pulse width. The corrected results are also listed in table 1and it can be seen that the largest distance difference between the results of 10 ns and 80 ns is 0.21 m, smaller than the sampling resolution of ∼0.41 m. Despite the distance correction, the obtained result differs inevitably from the actual value owing to the quantization of the analog-to-digital conversion and detection noise, etc. The relationship between the obtained and actual values is assumed to be linear according to the international standard of the International Electrotechnical Commission [27]. Since the lengths of each section in the fiber loop L0,LAand LBare accurately determined, the actual distance at each signal end can be determined more precisely than the distance directly obtained from the measurement. A linear least-squares regression is used to find out the relationship between the measured and actual values. The results of different pulse widths are plotted in figure 5. The red and blue colors represent the measured and corrected results, respectively. The difference between them is very small for Figure 4. Rayleigh signal obtained with different pulse widths at the end of the right loop section for (a) 1st loop trip, (b) 2nd loop trip and (c) 3rd loop trip. short pulses, and it becomes obvious for long pulses because of the width dependent offset. However, the pulse width seems to have no influence on the regression analysis as the obtained slope is 1.0082 for all cases. Owing to attenuation within the artefact, signals from only three round trips can be detected with an acceptable SNR, providing three data points for regression analysis. This is considered sufficient, as the relationship is expected to be linear according to [27], and the R2value is close to 1, indicating an excellent fit as shown in figure 5. The location accuracy is also studied when the sensing fiber is subject to external vibrations. Numerous studies have revealed that the signal detected in a Rayleigh based DFS system is actually the interference result of the light backscattered within half of the pulse length [1,12,13,28]. Any perturbation that modifies the interference condition can cause a signal change at the corresponding position. The signal should change at the same frequency as the perturbation in the ideal case. For the dynamic measurement, ∼13 m fiber in the right loop section of the artefact is wound around a piezoelectric tube (PZT) that is driven by a sinusoidal voltage oscillating at 500 Hz. The measurement is repeated using different pulse widths and the raw data is analyzed in the frequency domain for a better visualization of the vibration. The frequency signals obtained by standard fast Fourier transform are shown in figure 6with a focus on the perturbed fiber section. The vibration signal at 500 Hz is clearly visible for all the used pulse widths and the perturbed region seems to have the same start position. The result of the longer pulse exhibits an ‘extended’ vibration region because the end position shifts to greater distances owing to the width dependent offset. This phenomenon has also been observed for the Raman based DTS system [16]. Such an offset can also be removed as described above, so the actual perturbed region can be determined, and its length is very similar to the fiber wound around the PZT with a difference less than the sampling interval. 4.1.2. Signal quality analysis for Rayleigh scattered light. As mentioned in the introduction, the four performance criteria of the DFS are interdependent and ultimately governed by the SNR of the obtained signal. The artefact can be used to assess the quality of the obtained signal and the result can 5 Meas. Sci. Technol. 36 (2025) 115210 X Lu et al Table 1. Sensing distance measured with different pulse widths and loop trips. 10 ns 20 ns 50 ns 80 ns Measured Corrected Measured Corrected Measured Corrected Measured Corrected 1st trip 110.30 m 109.28 m 111.52 m 109.48 m 114.38 m 109.27 m 117.65 m 109.48 m 2nd trip 250.90 m 249.88 m 252.13 m 250.09 m 254.99 m 249.88 m 258.26 m 250.09 m 3rd trip 391.50 m 390.48 m 392.73 m 390.69 m 390.59 m 390.48 m 398.86 m 390.69 m Figure 5. Linear least-squares regression of the measured and actual distances of different loop trips for the pulse widths of (a) 10 ns, (b) 20 ns, (c) 50 ns and (d) 80 ns. be exploited to estimate the SNR for long distance sensing. Noise is inevitable in the measurement and causes a slight and random change of the signal obtained at the same position over time, deteriorating the SNR. Averaging is a powerful tool for noise reduction, and the obtained signal is averaged over 20 000 times for the quality analysis. Since pre-amplification is used, the dominant noise in this case is the beat between the signal and the ASE from the pre-amplifier, referred to as sigASE noise [29]. This noise is proportional to the strength of the local signal; thus, it also varies randomly along the fiber due to the random nature of the obtained Rayleigh signal. At a given position, the standard deviation Istd of the signal obtained over the 20 000 measurements is taken as the noise. Therefore, the SNR is defined as SNR(z)=10 log 10( I(z)/Istd(z)), where I represents the averaged signal at a given position, respectively. As mentioned above, both Iand Istd change randomly along the fiber, the SNR over the entire fiber can thus only be analyzed in a statistical manner. Probability density function (PDF) is an effective tool for such a task. The PDFs of the SNR obtained at every sampling point within the shaded fiber sections as shown in figure 3are calculated and plotted in figure 7for different incident pulse widths. The SNR calculated under each condition produces a PDF peak and the comparison of the peak positions provides meaningful information. As the loop trip number increases, the PDF Figure 6. Vibration induced frequency signals for the pulse widths of (a) 10 ns, (b) 20 ns, (c) 50 ns and (d) 80 ns. peak shifts to the left, indicating a degradation in the overall SNR, primarily due to the loss of the artefact in the setup. The total loss within the artefact can be precisely quantified. Based on the attenuation coefficient of the standard SMF, the spatial information from multiple loop trips can be accurately converted into real distance measurements. This enables reliable estimation of SNR in long-distance sensing applications that spans tens of kilometers using only a few hundred meters of fiber. Regardless of the pulse width, the PDF peaks of the 1st loop trip tend to cluster in a similar SNR region, as shown in figure 7. Although a longer pulse width results in higher optical power launched into the fiber and consequently a stronger detected signal, the sig-ASE noise increases proportionally with the signal, leading to an almost unchanged SNR. Therefore, the SNRs obtained in the 1st loop are very similar. In contrast, the PDF peaks of the 3rd loop trip are located in different SNR regions depending on the pulse width. The SNR can drop below 0 dB for short pulses (10 ns and 20 ns) but remains above 0 dB for longer pulses (50 ns and 80 ns). This behavior can be attributed to a shift in the dominant noise due to the weak signal. As shown in figure 3, the backscattered light in the 3rd loop trip is extremely weak, even after 6 Meas. Sci. Technol. 36 (2025) 115210 X Lu et al Figure 7. The probability density function of the obtained SNR over different numbers of trips in the fiber loop with the pulse widths of (a) 10 ns, (b) 20 ns, (c) 50 ns and (d) 80 ns. pre-amplification, due to losses in the setup. Consequently, the sig-ASE noise is reduced, and the impact of other noise sources, such as thermal noise and the ASE from the preamplifier, becomes comparatively more significant. As a result, the dominant noise is no longer proportional to the signal, and the weak signal obtained with short pulses leads to a lower SNR. 4.2. Raman signal The WDM shown figure 1can separate light between 1550 nm and 1465 nm. This wavelength is deliberately selected so that the anti-Stokes component of the spontaneous Raman scattering can be detected, which is much more sensitive to temperature than the Stokes component. The Raman scattering is intrinsically 30 dB weaker than the Rayleigh scattering [30], and the whole setup shown in figure 2is not optimized for Raman signal detection, i.e. using a standard SMF instead of the multimode fiber, due to the limitation of the hardware. Consequently, a long averaging time is required to enhance the obtained signal and the SNR. The Raman signal is detected with pulse widths of 20 ns, 50 ns and 80 ns and averaged over 6 ×106times, and a short section of the fiber is subject to temperature changes. The averaged signals with 50 ns width are shown in figure 8(a) for the three loop trips. Compared with figure 3, the obtained signal is obviously lower and noisier, and it becomes even lower for the 2nd and 3rd trips due to the optical loss in the artefact. At about 53 m, a signal drop is visible for all traces due to a splicing loss. The fiber section in the gray region is placed in a water bath and experiences a temperature change from 14.4 ◦C to 75 ◦C. The local signal increases at a higher temperature and can be used for temperature sensing. However, it is difficult to identify the temperature induced signal increase for the 2nd and 3rd loop trips due to the weak and noisy signal. Figure 8. Averaged Raman signals from the artefact (a) with 50 ns incident pulse under different temperatures, and with different pulse widths at the end of the right loop section for (b) 1st loop trip, (c) 2nd loop trip and (d) 3rd loop trip. Note the change of scale. The Raman signals from the fiber end are compared at various trips and pulse widths, as shown in the lower row of figure 8. In figure 8(b), it is obvious that the end position of the first trip shifts to greater distances with longer pulses. Note that the sensing fiber becomes shorter according to the Raman signal because the fiber section between the combiner and the WDM is not measured here. As the loop trip increases, the signal becomes weak and noisy, making it challenging to identify the fiber end. It is thereby impossible to assess the spatial accuracy of the Raman signal. Nonetheless, a similar study has been documented in [16]. In practice, the sensing fiber used for the Raman DFS is usually multimode in order to collect more scattered light, enhancing the SNR of the measurement. In this case, the spatial correction is more important for the Raman measurement owing to the intermodal dispersion. This work will be performed in the future. The quality analysis for the Raman signal differs from that in section 4.1.2. Compared with the random Rayleigh signal, the Raman signal is supposed to be much smoother, being subject to the loss along the sensing fiber. However, quasi-periodic spurious spikes are visible in figure 8, which are probably due to the impedance mismatch between the photodetector and the digitizer. Thermal noise is expected to dominate the photodetection, being constant over the fiber, and its value is dependent on the performance of the used photodetector. As a result, the SNR of the Raman signal should exhibit a lower level of variability, which is confirmed by the narrower PDF peaks in figure 9compared to those in figure 7. For the 1st trip, the PDF peaks at a comparatively higher SNR value and peak width is caused by the fiber attenuation, splicing loss and the spurious spikes. It is noteworthy that the PDF peaks for the first trip are observed at −14 dB, −11.8 dB, and −11 dB for pulse widths of 20 ns, 50 ns, and 80 ns, respectively. This indicates that the SNR enhancement scales unproportionally with increasing pulse width. In particular, the SNR shows negligible improvement between the 50 ns and 80 ns pulses. This behavior can be attributed to the variation in peak power across different pulse widths at a fixed repetition rate; specifically, longer pulses exhibit reduced peak power. Consequently, the 7 Meas. Sci. Technol. 36 (2025) 115210 X Lu et al Figure 9. The probability density function of the obtained SNR for Raman signal in the region of interest with the pulse widths of (a) 20 ns, (b) 50 ns and (c) 80 ns. SNR improvement is limited and fails to increase linearly with pulse duration. Similar to the Rayleigh signal measurement, the SNR degrades as the trip increases due to optical loss in the setup, causing the PDF peak to shift leftward, as shown in figure 9. Meanwhile, the PDF peaks become broader owing to the influence of the noise on the weak signal, as shown in figure 8(a). By converting the measured distance into actual spatial information, the signal quality of the Raman backscattered light can be estimated for practical applications. 4.3. Test with commercial systems In previous sections, the artefact is utilized to test the selfdeveloped DFS systems based on Rayleigh and Raman scattering effects. It is now applied to two commercial systems in order to showcase the capability of the artefact to evaluate all types of DFS in the time domain, and its readiness and suitability for testing the systems on the market. 4.3.1. Application to conventional OTDR system. Conventional OTDR systems use Rayleigh scattering to troubleshoot optical communication networks by identifying losses and locating faults. The artefact is directly connected to the OTDR system that is used to determine LAand LBin order to demonstrate its readiness for testing commercial systems. Figure 10(a) shows OTDR traces with different pulse widths. A broader incident pulse results in a stronger signal because more optical power is launched into the fiber. However, due to the loss in the fiber loop, the signal weakens and becomes noisier on the 2nd and 3rd trips, especially with narrower incident pulses. The OTDR measurement indicates a signal drop caused by the splicing around 55 m, similar to the Raman signal shown in figure 8(a). The incident pulse is reflected at the end of the right part of the loop, resulting in reflection peaks at the end of the OTDR traces. Thus, the width of the reflection peaks is dependent on the pulse width, longer pulses result in wider peaks, as shown in the lower row of figure 10. It is interesting to note that all the peaks start at almost the same position for the same loop trip and the distance is close to the actual length of the fiber. The same analysis as in section 4.1.1 can be used here to correct the measured distance of the OTDR system. Figure 10. Standard OTDR traces measured from the artefact (a) with various incident pulses, and zoom-in views of the trace end for (b) 1st loop trip, (c) 2nd loop trip and (d) 3rd loop trip. Figure 11. Brillouin frequency shift obtained from the artefact with 5 ns and 10 ns pulses. The gray region represents the fiber section in a water bath at 50 ◦C. 4.3.2. Application to a commercial BOTDA system. The artefact has been applied to calibrate BOTDR systems [19], it is used in this paper to test a commercial BOTDA system in order to demonstrate its versatility. The BOTDA measurement relies on stimulated Brillouin scattering, requiring optical pulses to act as pump and continuous lightwave as probe. The pump is launched into the artefact via the combiner and the probe is launched via the port 1 of the circulator in the artefact, as shown in figure 2. The pulses with 5 ns and 10 ns widths are used for the measurement and the obtained signal is averaged by 214 times for a reliable determination of the BFS. Like the Raman test, a short fiber section is placed in a water bath to cause a local BFS change. The measured BFSs are plotted in figure 11 and exhibit very similar profiles in the three trips. The three fiber sections used to construct the right part of the loop exhibit very different BFSs with the boundary at 55 m and 100 m, and the BFS is even larger in the gray region which is induced by the hot water at 50 ◦C. The sudden change of the BFS between different fiber sections offers another possibility for distance correction and even calibration. In the 1st trip, the BFS profiles with 5 ns and 10 ns pulses almost overlap, only a slight difference can be observed in the 8