Open Access
Add to BibSonomy
Abstract
For a coaxial single-photon lidar system, amplified spontaneous emission (ASE) noise from the fiber amplifier is inevitable. The ASE backscattering from specular reflection annihilates the far-field weak signal, resulting in low signal-to-noise ratio, short measurement distance, and even misidentification. We propose a method for calibrating and mitigating ASE noise in all-fiber coaxial aerosol lidar and demonstrate the method for a lidar system with different single-photon detectors (SPDs). The accuracy of the coaxial aerosol lidar is comparable to that of the biaxial one. We conducted an experiment using three different detectors, namely, InGaAs/InP SPD, up-conversion SPD, and superconducting nanowire SPD in the same coaxial lidar system. Compared with the biaxial system, the three different detectors we used have achieved more than 90% ASE noise suppression, the measured visibility percent errors of InGaAs/InP SPD data, up-conversion SPD data, and superconducting nanowire SPD data all within 20%, and the percent error within 10% are 99.47%, 100%, and 95.12%, respectively. Moreover, time-sharing optical switching allowed to obtain background noise with high accuracy.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Lidar is an active optical remote sensing technology that provides high spatial and temporal resolution, real-time operation, and interference mitigation, thus being widely used for measurements of atmospheric environments and meteorology [1,2]. Atmospheric lidar consists of a laser source, optical transmitting system, optical receiving system, photoelectric detector, and data acquisition and processing system [3,4]. By using the interaction between the laser and environment, the atmospheric aerosol, wind field, cloud optical properties, and air pollutants can be detected and studied [3–5].
The development of fiber technology has led to the continuous improvement of all-fiber atmospheric lidar over the past two decades. The resulting system is more compact and robust than the space optical atmospheric lidar configuration [6]. However, limited by the core diameter of the fiber, backscatter in the telescope is difficult to fully couple into the fiber, hindering optical adjustment. In addition, the amplifier generates amplified spontaneous emission (ASE) when the gain medium in the fiber laser shows reverse excitation, which is a considerable part of amplifier noise and can deteriorate the signal-to-noise ratio (SNR) [7–10].
For coaxial atmospheric lidar measurements, the emitted pulse laser generates a strong specular signal on the mirror plane, the same is true of ASE noise. The specular ASE noise annihilates the signal of interest, especially for long-distance detection, for which the ASE noise submerges the backscattering signal, resulting in a low SNR and even misidentification. Most all-fiber coaxial atmospheric lidar technologies are based on coherent detection, which has a narrow band beat effect and high robustness against ASE noise. However, coherent atmospheric lidar is usually adopted in Doppler frequency measurement [11]. In single-photon detector (SPD) atmospheric lidar, the ASE and background noise is filtered out by the narrow band optical filter, but commercial fiber filters with 0.1 nm passband fail to eliminate specular interference, with the use of a biaxial optical layout being a common solution. Nevertheless, two optical axes introduce other problems, such as a larger blind area, more serious geometric overlap factor, lower stability, larger far-field signal jitter, and higher manufacturing costs compared with the coaxial system. Alternatively, ASE noise can be mitigated through dedicated methods in all-fiber coaxial single-photon lidar (CSPL). Yu et al. [12] used an algorithm that modifies the ASE noise along the vertical direction in a multimode fiber-coupled InGaAs/InP SPD 1.5 μm lidar system to obtain an accurate backscattering signal. Feng et al. [13] used an ASE filter to suppress unabsorbed pump light and ASE in the fiber after amplification. Liu et al. [14] used a narrow passband filter with 0.8 nm bandwidth to suppress forward propagating ASE. Qiao and Vella [15] calculated the ASE power and used it in an automatic signal-power control algorithm for EDFA Automatic Control.
Despite the variety of studies available, methods for CSPL have been largely unexplored. We propose a method for calibrating and mitigating ASE noise for all-fiber CSPL. Two SPD atmospheric lidar systems in the biaxial and coaxial optical layouts were built and set with the same detection path. Experiments using three different SPDs in the same coaxial lidar system were conducted. Compared with the biaxial system, the accuracy of retrieved ASE noise above 90%, measured visibility error within 20%, and proportion of data error within 10% were 99.47%, 100%, and 95.12%, respectively.
The remainder of this paper is organized as follows. In Section 2, we introduce the ASE noise mitigation algorithm and experimental setup. In Section 3, we detail ASE noise acquisition and mitigation using the InGaAs/InP SPD [16], up-conversion SPD (UCSPD) [17], and superconducting nanowire SPD (SNSPD). We also report the calculation of the corresponding visibility and show the performance of ASE noise mitigation. In Section 4, conclusions and directions of future work are presented.
2. Principle and Instrument
2.1 ASE noise calibration and mitigation
In a CSPL system, ASE noise is generated by an erbium-doped fiber amplifier (EDFA) within the interval of the emitted pulse laser, and it shows a gradual increase between laser pulses [18]. The emitted pulse laser is reflected through the mirror plane of the telescope, and the intensity of the reflected noise depends on the specular reflectivity. Although the reflected ASE noise has a negligible effect on short-distance measurements, it is harmful in long-distance measurements, deteriorating the SNR of the backscattering signal or even completely masking the backscattering signal. We aim to eliminate ASE noise caused by specular reflection and enhance the CSPL performance. To this end, the Fernald method [19,20] is used to convert the measured range-corrected photon counts into an extinction coefficient, and the calculated visibility [21] is used to evaluate ASE noise mitigation.
Let us consider the elastic scattering lidar equation [22]. The power is replaced by photon counts in the equation to use near-infrared SPDs. As the experiment was conducted at nighttime with a laser wavelength of 1548.51 nm, the sky background radiation was negligible, and only the detector background noise was considered. Hence, the lidar equation can be written as
In addition to the detector background noise, lidar is also affected by the ASE noise reflected by the telescope mirror plane, whose distribution is stable and increases with increasing r. Therefore, the lidar Eq. (1) should add the ASE noise (${N_{ASE}}(r )$).
The ASE noise in CSPL suppresses the far-field backscattering signal, whereas biaxial single-photon lidar (BSPL) is not affected by the ASE noise from specular reflection. To obtain ${N_{ASE}}(r )$, the detector background noise, ${N_D}$, should be eliminated. The near-field blind area of BSPL can be used to obtain the detector background noise with more accuracy than by considering a fixed value. For CSPL, time-sharing optical switching can be used to obtain the background noise. After removing the effect of detector background noise, the number of photons, ${N_{B,D}}(r )$ and ${N_{C,D}}(r )$, received by the BSPL and the CSPL, respectively, are given by
2.2 Optical layout and instruments
The optical layouts of the CSPL and BSPL systems are shown in Fig. 1. A pulse fiber laser with a wavelength of 1548.51 nm generates laser pulses at 10 kHz. The laser pulses are amplified by the EDFA and transmitted to the atmosphere by the telescope through the fiber circulator. The receiving telescopes of the CSPL and BSPL systems obtain the backscattering signal simultaneously. The signal received by the CSPL enters the optical switch through the fiber circulator and is coupled into the fiber splitter with a 1:9 ratio by two filters with central wavelengths of 1550 and 1548.51 nm. Then, the signal is detected by the SPD. The optical switch is controlled by an arbitrary waveform generator with the control method described in the experiments (Section 3).
Fig. 1. Optical layout of CSPL and BSPL. AWG, arbitrary waveform generator; PC, personal computer; PFL, pulsed fiber laser; OS, optical switch; BT, beam trap
The signal beam acquired by the receiving telescope of the BSPL system enters the SPD directly through the optical fiber. The data detected by the SPD are collected by a multiscaler and transmitted to a computer for further analysis and processing. The sample rate of the multiscaler is 100 MHz, corresponding to a spatial resolution of 15 m.
In this study, we evaluated InGaAs/InP SPD, UCSPD, and SNSPD. Limited by the single-channel SNSPD, it was used in the CSPL system only, while the BSPL system used InGaAs/InP SPD as a reference. The InGaAs/InP SPD and UCSPD were used in the CSPL and BSPL systems.
The experimental setup shown in Fig. 1(b) is intended to obtain the ASE noise of the CSPL system for comparison. A beam trap (blue dashed box) is placed in front of the telescope of the CSPL system to absorb the laser pulse signal emitted by the telescope, while the other structures of the optical layout remain unchanged. The main parameters of the lidar system shown in Fig. 1 are listed in Table 1.
3. Experiments and results
Two experiments were designed and conducted in this work. The first experiment was based on BSPL to eliminate the ASE noise in CSPL. The second experiment used CSPL to obtain the specular reflection of the ASE noise, obtaining the pure ASE noise data in the CSPL system.
Table 1. Main parameters of the lidar system
3.1 CSPL ASE noise mitigation
The original backscattering signals were measured using the CSPL and BSPL systems installed in Hefei, China in October 2021. Three SPDs were used for comparisons, as shown in Fig. 1. Continuous observations were collected from 17:50 to 19:35 on October 25, 2021 for the UCSPD, from 22:21 on October 25 to 01:04 on October 26, 2021 for the InGaAs/InP SPD, and from 19:50 on October 26 to 03:17 on October 27, 2021 for the SNSPD.
As a filter with a center wavelength of 1548.51 nm was fused in the output end of the EDFA, to accurately obtain the background noise of the SPD in the CSPL system, we used time-sharing optical switching. The optical switch, laser pulses, and transmittance of the backscattering signal were controlled by the timing sequence of the arbitrary waveform generator. The time sequence is shown in Fig. 2. Figure 2(a) shows the periodic TTL (transistor–transistor logic) signal generated by the arbitrary waveform generator. In one period, the TTL signal is a 10 kHz pulse (pulse width of 4 μs) lasting 3 min followed by a high-level signal lasting 15 s. When the TTL is set to the high level, the optical switch selects the 1550 nm filter, and the detector does not receive any signal except for background noise and a partial high-power pulsed laser signal caused by specular reflection. When the TTL is set to the low level, the optical switch selects the 1548.51 nm filter, and the detector receives a normal backscattering signal, as shown in Fig. 2(c). When the laser pulse shown in Fig. 2(b) is emitted, the detector receives both a normal backscattering signal and a weakened strong specular reflection signal in the first 3 min of a period. Then, it receives background noise and the weakened strong specular reflection signal in the last 15 s.
Fig. 2. Timing sequence of signals received by detector. (a) Optical switch (OS) control signals, (b) pulsed laser signals, and (c) backscattering signal.
After using the three SPDs to obtain the atmospheric backscattering signals and remove the background noise, the data acquired by the BSPL required proportional transformation. In the measured atmospheric path, the near-field backscattering signal of the biaxial system is greatly affected by the blind area and geometric overlap factor, while the far-field backscattering signal of the coaxial system is greatly affected by the specular reflection of ASE noise. Although the near-field backscattering signal of the coaxial system should also consider the influence of the geometric overlap factor, its effect is lower than on the biaxial system, and the influence distance is shorter. To correctly determine the proportional transformation, a distance range that is less affected by the geometric overlap factor of the biaxial system and the specular reflection of ASE noise in the coaxial system should be determined. This distance range can be determined as follows.
First, the distance gate with the strongest received SPD signal in the BSPL is identified, and the following calculation is made from that distance gate:
When $G(r )\mathrm{\geqslant }0$, according to the analysis of Eqs. (3) and (4), BSPL and CSPL can be regarded as insensitive to the geometric overlap factor, that is, ${Y_B}(r )= 1$ and ${Y_C}(r )= 1$. Meanwhile, the influence of ASE noise is very small. Take the first distance gate as the starting point of the selected distance range when $G(r )\mathrm{\geqslant }0$. Use Eq. (10) and ${K_{SD}}({{r_0}} )$ as the initial estimate value of the proportional transformation: Thereby,Once the distance range of the proportional transformation is determined, value k of the proportional transformation can be determined. As the ASE noise is small over this distance range and the variation is small relative to the backscattering signal, the ASE noise can be considered as almost constant. Thus,
The advantage of the method above is that when the ASE noise is unknown and the change relative to the signal is small, the signal itself can change greatly, thus reducing the impact of the ASE noise and allowing to determine k of the proportional transformation more accurately. After obtaining k, the proportional transformation can be performed using Eq. (5), and the ASE noise can be estimated.
For ease of analysis, we unified the ASE noise data acquisition from the 200th to the 950th distance gates, that is, for a distance range of 2.625–13.875 km. Then, we fit the ASE noise data. Figure 3 shows the fitting results after processing the data acquired by the UCSPD to estimate the ASE noise.
After fitting function ${N_{ASE,fit}}(r )$ was obtained, we removed the ASE noise from the backscattering signal data received by the CSPL to obtain ${N_{C,cor}}(r )$. For ${N_{C,cor}}(r )$ and ${N_{B,prop}}(r )$, we applied 10-point Gaussian kernel smoothing and 5-distance-gate averaging and then inverted them to obtain the extinction coefficient using the Fernald method. Finally, we calculated the visibility. To confirm the effect of the ASE denoising, we performed visibility retrieval of the data detected by the CSPL without ASE noise mitigation. The visibility retrieval of the uncorrected and corrected ASE noise detected using UCSPD data are shown in Fig. 4. The visibility retrieval error distribution obtained from UCSPD, SNSPD and InGaAs/InP SPD data for uncorrected and corrected ASE noise is shown in Fig. 5.
Fig. 4. Visibility retrieval for CSPL SPD data (a) without and (b) with ASE noise mitigation. (c) Visibility retrieval for BSPL SPD data. The average visibility was 9.97 km from October 25 to October 28, 2021 in Hefei, China. The CSPL SPD data retrieved visibility of 2.025–5.025 km, and the BSPL SPD data retrieved visibility of 2.625–5.025 km.
Fig. 5. (a) Uncorrected and corrected ASE noise visibility error distribution of UCSPD data. (b) Uncorrected and corrected ASE noise visibility error distribution of InGaAs/InP SPD data. (c) Uncorrected and corrected ASE noise visibility error distribution of SNSPD data. The percent error is calculated by ${{\delta \textrm{ = (}{\textrm{V}_C}\textrm{ - }{\textrm{V}_B})} / {{V_B}}}$, where ${\textrm{V}_C}$ and ${\textrm{V}_B}$ are the visibility inverted by CSPL and BSPL data, respectively
Figure 4 intuitively show that the visibility of CSPL without correcting ASE noise deviates greatly from that of BSPL. When CSPL data is corrected, the visibility of both is almost the same. We calculated the visibility percent error ($\delta $) between CSPL and BSPL, the results show in Fig. 5. Figures 5(a), 5(b) and 5(c) show that when the UCSPD, SNSPD and InGaAs/InP SPD data in CSPL were not correct the ASE noise, the $\delta $between CSPL and BSPL of them can exceed 40%. After ASE noise mitigation, the $\delta $ obviously reduced. We compare the probability of uncorrected and corrected ASE noise of three detectors in CSPL, as shown in Table 2, which shows that without corrected ASE noise, the proportion of $\delta $ below 20% for UCSPD data, SNSPD data, InGaAs/InP SPD data are only 17.58%, 36.08% and 29.67%. After ASE noise mitigation, the $\delta $ of the UCSPD, SNSPD and InGaAs/InP SPD are all within 20%. Moreover, 100%, 95.12% and 99.47% of the UCSPD, SNSPD and InGaAs/InP SPD data percent errors are within 10%, respectively. Hence, after CSPL SPD data mitigated ASE noise, the visibility retrieval of CSPL is comparable to that of BSPL.
Table 2. The $\delta $ probability comparison of CSPL uncorrected and corrected ASE noise
In data processing, considering that the data acquired by the SNSPD and UCSPD are less affected by the dead time and after-pulse, no correction of these parameters is necessary. On the other hand, the data acquired by the InGaAs/InP SPD are greatly affected by the dead time and after-pulse, rendering their correction necessary.
3.2 ASE noise measurement
We measured the ASE noise data (${N_{ASE,D}}(r )$) of CSPL using the optical circuit diagram of Fig. 1(b) and the optical switching control consistent with CSPL ASE noise mitigation. We measured 3 h of data, and after correction for dead time and after-pulse, ${N_{ASE,D}}(r )$ was obtained by averaging. The fitting function (${N_{ASE,D,fit}}(r )$) of ${N_{ASE,D}}(r )$ was normalized together with the ASE noise fitting function obtained from UCSPD data, and the comparison is shown in Fig. 6.
Fig. 6. Comparison of ASE noise fitting function between ${N_{ASE,D}}(r )$ and UCSPD data. The error (blue dashed line) of normalized ASE noise fitting function is obtained from UCSPD data and normalized ${N_{ASE,D,fit}}(r )$.
Figure 6 shows that the two fitting functions overlap after normalization, and the fitting error is within 10% and gradually decreases with increasing distance. Hence, using the data detected by BSPL as reference to obtain the CSPL ASE noise is correct and feasible, and it enables ASE denoising in CSPL. Moreover, considering the temporal stability of ASE noise when building a CSPL system, the ASE noise reflected by the telescope mirror can be premeasured and calibrated to increase the detection range.
4. Conclusion
We propose a method to mitigate ASE noise in CSPL and improve the SNR and measurement range. After the backscattering signal of CSPL mitigated ASE noise, we compared the visibility retrieval of CSPL with BSPL. The results show that the visibility retrieval error of the InGaAs/InP SPD, UCSPD, and SNSPD can be controlled within 20%, and 95% of the visibility retrieval error remains around 10%. A direct measurement experiment of ASE noise to verify the accuracy of CSPL ASE noise mitigation also be designed and built. The error is within 10% between the ASE noise fitting functions of measurements and noise mitigation results using UCSPD data, indicating that the proposed method can increase the accuracy of CSPL to the level of BSPL. In addition, time-sharing optical switching was implemented to obtain accurate detector background noise.
As CSPL uses a monocular telescope, it provides a small size, light weight, simple structure, low cost, and structure stability compared with BSPL as well as smaller and more stable near-field blind area and geometric overlap factor. Thus, CSPL is widely used in atmospheric environments and meteorology, and CSPL improved by the proposed ASE noise mitigation method can achieve higher SNR, larger measurement distance, and more accurate signal acquisition, being suitable for airborne and spaceborne platforms and long-term observations. In future work, we will investigate long-term observations to evaluate aspects such as the generated ASE noise changing owing to laser power attenuation during long-term operation.
Funding
State Key Laboratory of Pulsed Power Laser Technology (SKL2020KF08); National Natural Science Foundation of China (42125402, 42188101); Key-Area Research and Development Program of Guangdong Province (2020B0303020001); Shanghai Municipal Science and Technology Major Project (2019SHZDZX01); Joint Open Fund of Mengcheng National Geophysical Observatory (MENGO-202106); Fundamental Research Funds for the Central Universities.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper can be obtained from the authors.
References
1. X. Ma, C. Wang, G. Han, Y. Ma, S. Li, W. Gong, and J. Chen, “Regional atmospheric aerosol pollution detection based on LiDAR remote sensing,” Remote Sens. 11(20), 2339 (2019). [CrossRef]
2. A. Pantazis, A. Papayannis, and G. Georgousis, “Lidar algorithms for atmospheric slant range visibility, meteorological conditions detection, and atmospheric layering measurements,” Appl. Opt. 56(23), 6440–6449 (2017). [CrossRef]
3. T. Foken, Springer Handbook of Atmospheric Measurements (Springer, 2021).
4. A. Comerón, C. Muñoz-Porcar, F. Rocadenbosch, A. Rodríguez-Gómez, and M. Sicard, “Current Research in Lidar Technology Used for the Remote Sensing of Atmospheric Aerosols,” Sensors 17(6), 1450 (2017). [CrossRef]
5. R. Ceolato and M. J. Berg, “Aerosol light extinction and backscattering: A review with a lidar perspective,” J. Quant. Spectrosc. Radiat. Transf. 262, 107492 (2021). [CrossRef]
6. S. Kameyama, T. Ando, K. Asaka, Y. Hirano, and S. Wadaka, “Compact all-fiber pulsed coherent Doppler lidar system for wind sensing,” Appl. Opt. 46(11), 1953–1962 (2007). [CrossRef]
7. T. I. Li and M. C. Teich, “Bit-error rate for a lightwave communication system incorporating an erbium-doped fibre amplifier,” Electron. Lett. 27(7), 598–600 (1991). [CrossRef]
8. P. Diament and M. C. Teich, “Evolution of the statistical properties of photons passed through a traveling-wave laser amplifier,” IEEE J. Quantum Electron. 28(5), 1325–1334 (1992). [CrossRef]
9. T. Li and M. C. Teich, “Photon point process for traveling-wave laser amplifiers,” IEEE J. Quantum Electron. 29(9), 2568–2578 (1993). [CrossRef]
10. A. Mecozzi, “Quantum and semiclassical theory of noise in optical transmission lines employing in-line erbium amplifiers,” JOSA B 17(4), 607–617 (2000). [CrossRef]
11. R. T. Menzies and R. M. Hardesty, “Coherent Doppler lidar for measurements of wind fields,” Proc. IEEE 77(3), 449–462 (1989). [CrossRef]
12. C. Yu, J. Qiu, H. Xia, X. Dou, J. Zhang, and J.-W. Pan, “Compact and lightweight 1.5 μ m lidar with a multi-mode fiber coupling free-running InGaAs/InP single-photon detector,” Rev. Sci. Instrum. 89(10), 103106 (2018). [CrossRef]
13. Y. Feng, J. Zhou, S. Du, and W. Chen, “All-fiber pulsed laser transmitter for space-based 3D imaging lidar,” in Conference on Lasers and Electro-Optics/Pacific Rim (Optica Publishing Group),(2009), paper TUP4_12.
14. Y. Liu, J. Liu, and W. Chen, “Eye-safe, single-frequency pulsed all-fiber laser for Doppler wind lidar,” Chin. Opt. Lett. 9(9), 090604 (2011). [CrossRef]
15. L. Qiao and P. J. Vella, “ASE analysis and correction for EDFA automatic control,” J. Light. Technol. 25(3), 771–778 (2007). [CrossRef]
16. C. Yu, M. Shangguan, H. Xia, J. Zhang, X. Dou, and J.-W. Pan, “Fully integrated free-running InGaAs/InP single-photon detector for accurate lidar applications,” Opt. Express 25(13), 14611–14620 (2017). [CrossRef]
17. H. Xia, G. Shentu, M. Shangguan, X. Xia, X. Jia, C. Wang, J. Zhang, J. S. Pelc, M. M. Fejer, Q. Zhang, and X. Dou, “Long-range micro-pulse aerosol lidar at 1.5 μm with an upconversion single-photon detector,” Opt. Lett. 40(7), 1579–1582 (2015). [CrossRef]
18. J. Xiaodong, X. Haiyun, S. Mingjia, and D. Xiankang, “Measurement of amplified spontaneous emission noise in high power pulsed fiber laser,” High Power Lasers Part. Beams 27(5), 51006 (2015). [CrossRef]
19. F. G. Fernald, “Analysis of atmospheric lidar observations: some comments,” Appl. Opt. 23(5), 652–653 (1984). [CrossRef]
20. C. Li, Z. Pan, F. Mao, W. Gong, S. Chen, and Q. Min, “De-noising and retrieving algorithm of Mie lidar data based on the particle filter and the Fernald method,” Opt. Express 23(20), 26509–26520 (2015). [CrossRef]
21. Z. Yang, D. S. Sun, and Q. Li, “Retrieval of atmosphere extinction coefficient for imaging laser radar,” INFRARED LASER Eng. 33(3), 239–242 (2004).
22. V. E. Privalov and V. G. Shemanin, “The lidar equation solution depending on the laser radiation line width studies,” Opt. Mem. Neural Netw. 22(4), 244–249 (2013). [CrossRef]
23. D. Kumar and F. Rocadenbosch, “Determination of the overlap factor and its enhancement for medium-size tropospheric lidar systems: a ray-tracing approach,” J. Appl. Remote Sens. 7(1), 073591 (2013). [CrossRef]
24. W. Gong, F. Mao, and J. Li, “OFLID: Simple method of overlap factor calculation with laser intensity distribution for biaxial lidar,” Opt. Commun. 284(12), 2966–2971 (2011). [CrossRef]
25. S. L. Wang, K. F. Cao, S. H. U. and H, and L. Wei, “Analysis and det-ermination of lidar geometrical factor,” Laser Technol. 32(2), 147–150 (2008).
