Open Access
Add to BibSonomy
Abstract
High three-dimensional (3D) resolution for a wide field-of-view (FoV) is difficult in LiDARs because of the restrictions concerning size, weight, and power consumption (SWaP). Using a static unitary detector (STUD) approach, we developed a photodetector and a laser module for a LiDAR. Utilizing the fabricated photodetector and laser module, a LaserEye2 LiDAR prototype for low-SWaP applications was built using the STUD approach, which efficiently enables short-pulse detection with the increased FoV or large photosensitive area. The obtained 3D images demonstrated a diagonal FoV of > 31°, a frame rate of up to 15 Hz, and a spatial resolution of 320 × 240 pixels within a detection range of > 55 m. This prototype can be applied to drones to rapidly detect small or thin hazardous objects such as power lines.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Light detection and ranging (LiDAR) has been widely used as one of the most important sensor technologies in various autonomous driving systems and other applications owing to its advantages in real-time, as well as accurate detection and high-resolution three-dimensional (3D) depth imaging [1–5]. Recently, LiDAR applications have expanded to battery-powered unmanned systems, such as autonomous vehicles, robots, and unmanned aerial vehicles (UAV) [6–12], thereby resulting in strict restrictions on the size, weight, and power (SWaP) of LiDARs with minimum deterioration in performance. Thus, to expand LiDAR usage in various fields, efforts to simultaneously increase the performance and minimize the SWaP condition are required.
Frequency-modulated continuous-wave (FMCW) LiDAR is a possible candidate for achieving the aforementioned characteristics because it can be implemented in a compact solid-state system using a photonic integrated circuit (PIC) including an optical phased array (OPA). The coherent detection in this LiDAR can provide noise immunity from solar background radiation (SBR) and unwanted signals from other LiDARs, resulting in a high signal-to-noise ratio (SNR) [13–16]. However, complex systems with advanced optical integration techniques are required for the inevitable linearization of the wavelength sweep in FMCW modulation. Additionally, the achievable detection range is limited by the linewidth of commercially available lasers and the significant optical loss in integrating the PIC module [17–19]. These limitations indicate that this technology is not yet mature and ready to use.
However, the traditional time-of-flight (ToF) LiDAR can be used in a relatively simple structure and is generally regarded as mature. Various high-performance commercial products using rotating [20–22], flash [23–25], and scanning [26–28] approaches have been attempted to expand the usage of this LiDAR to low-SWaP and high-performance conditions. A rotating LiDAR acquires 3D images using horizontally rotating modules containing multiple exquisitely aligned laser diode (LD) photodetector (PD) pairs. This LiDAR possesses excellent horizontal resolution owing to its horizontal scanning ability. This is also advantageous for implementing cost-effective systems because a rotating LiDAR generally contains inexpensive silicon avalanche PDs and 905 nm LDs. However, satisfying a high vertical resolution and low-SWaP conditions simultaneously is difficult. This is because the significantly increased module size owing to the increased number of LD-PD pairs renders its usage impractical for achieving high vertical resolution. In addition, since the 905 nm wavelength is easily affected by the SBR, unlike the 1.5 µm wavelength [26], increased peak power is required for deteriorated detection environments, such as long-range, the presence of weak reflection targets, and bad weather. However, the peak power must be sustained under the maximum permissible exposure (MPE) using eye-safe restrictions [29–31]. Therefore, the structure of a rotating LiDAR based on a 905 nm wavelength is substantially limited for achieving both low-SWaP and high spatial resolution simultaneously.
Flash LiDAR which can be implemented as a scan-less system using a receiver, is highly reliable and can achieve high resolution. The focal plane array (FPA) and readout integrated circuit (ROIC) are connected to each other using the flip-chip process. High resolution and reliability are achieved because no moving part is included, and the increased detector array format is possible. However, the laser source needs to illuminate the entire field-of-view (FoV) in this LiDAR and a small fraction of reflected signal power is collected to each pixel in an FPA. Additionally, a high peak power (∼1 MW) for the laser source at an eye-safe wavelength is essential. This is to ensure a sufficient SNR for receiving optical power at a pixel level [32]. Generally, a 1.5 µm wavelength is beneficial for bad weather or long-range detection because a much higher peak power than its counterparts (830–950 nm) is allowed even under eye-safe restrictions [33]. Furthermore, when using the diode-pumped solid-state lasers (DPSSL) with eye-safe, the volume and weight of the laser source can be decreased to less than 12.0 × 4.9 × 6.4 cm3 and 360 g, respectively from commercially available products such as [34]. However, even though the DPSSL is a good candidate in a viewpoint of the overall size of flash LiDAR, the frame rate is strictly limited by the available pulse repetition frequency (PRF). Since the conventional flash LiDAR conditions are a peak power greater than at least several hundred kilowatts and pulse width of less than 10 ns, the obtainable PRF is normally restricted to less than 10 Hz [35,36]. In addition, the ROIC design for high-speed parallel signal processing for every detector in an FPA cannot provide good performance. This is a result of the small circuit implementation area (e.g., 15 × 15 µm2 within pixel dimensions) for the signal processing blocks including a low noise amplifier (LNA) [37]. For these reasons, simultaneously satisfying low SWaP conditions and high performance is difficult for flash LiDAR.
Scanning LiDAR using a 2-axis mirror can structurally provide a high horizontal and vertical resolution owing to its arbitrary scanning ability in any desired direction for obtaining 3D points. Additionally, this LiDAR provides a high SNR. This is owing to the simple biaxial structure of a scanning LiDAR, in which the paths of the transmitter and receiver do not coincide. In particular, the high spatial resolution for the horizontal and vertical directions is especially important in certain applications, such as power lines in real-time UAV operations, bicycles and autonomous vehicles at a relatively long distance. These applications require accurate detection of small and thin objects that are placed in an arbitrary direction. Therefore, a biaxial structure with a 2-axis scan can be the most suitable choice for high-performance and low-SWaP applications only if a wide FoV is obtainable, which includes all the required scan areas [38,39]. Therefore, a large-area PD is an essential design parameter according to the receiving optics. However, a typical large-area PD introduces a high junction capacitance in the detector device and consequently, reduces the bandwidth of the LiDAR receiver, which makes the detection of short pulses difficult [40].
Additionally, a laser source must satisfy the low-SWaP and high-performance requirements in the LiDAR module. Simultaneously, the optical peak power under a given power consumption requirement should be sufficiently high and stable. Usually, a master oscillator power amplifier (MOPA) with a 1.5 µm wavelength is a good candidate for providing a suitable detection range and frame rate, and its transmission loss in the atmosphere is relatively small. This is because a MOPA simultaneously provides high beam quality in the optical output and superior PRF operation. The all-fiber structure without any bulk optics has another merit i.e., it can be used in harsh and contaminated conditions, such as UAV vibrations and operating environments [41]. This is owing to its ability to provide a stable optical output regardless of the environment. However, most conventional 1.5 µm all-fiber commercial MOPA lasers are used in high-power applications such as free-space optical communication and remote sensing. They are also generally large and have high power consumption due to complex amplification structures with more than three amplifying stages and multiple components [41–43]. Therefore, to be used in a low-SWaP LiDAR, an all-fiber MOPA laser must be further optimized to meet the SWaP conditions in terms of the required optical peak power, rather than the pulse energy or average optical power.
In this study, we designed and fabricated a new PD based on a static unit detector (STUD) approach that we developed [44–48]. Additionally, a compact low-power all-fiber MOPA laser with two amplifying stages was designed and fabricated to provide a sufficiently high optical peak power under low-SWaP conditions. By applying the resultant optical receiver, we finally developed a LaserEye2 LiDAR prototype based on the STUD approach, which simultaneously satisfied the required low-SWaP and high-performance conditions. The characteristics of the fabricated STUD-based PD and customized 1.5 µm MOPA laser were evaluated and optimized for the low-SWaP requirements. Subsequently, the LaserEye2 LiDAR prototype was implemented to provide a high-resolution 3D image. Section 2 describes the newly proposed InP/InGaAs PIN-PD design and fabrication for the LiDAR receiver, as well as the customized MOPA laser and developed LaserEye2 LiDAR prototype. Section 3 explains the optimization of the operating conditions and the results of implementing the LiDAR system.
2. Low-SWaP 3D LiDAR-based on a STUD receiver and customized MOPA
2.1 High-speed large photosensitive area PD based on the STUD approach
The large photosensitive area of the PD at a given receiving optics is an important factor for achieving a wide FoV in a conventional biaxial LiDAR; however, high-speed detection is difficult in a large PD because the bandwidth is limited by the high junction capacitance of the PD. Therefore, to achieve wide FoV and high-speed operation using the STUD approach [44,45], a newly designed PD structure with multiple small-area PDs was implemented in this study. More specifically, the dedicated PD based on the STUD approach was composed of several long and narrow small-area PDs. Each of these is known as a finger PD, which has a relatively small junction capacitance, thereby enabling the high-speed detection of short-pulses. The incident optical signals on each finger PD were grouped and converted into an electrical signal through a pad and subsequently transferred to the input of the corresponding low-noise transimpedance amplifier (TIA) (Maxim, Max3658). Thereafter, through an inverting post-amplifier (Analog Devices, ADA4939), all the TIA output signals were simultaneously combined and amplified to yield one unified single signal output for the LiDAR receiver. As a result, the operation of the STUD PD with multiple finger PDs was similar to that of one large PD without sacrificing the bandwidth for high-speed operation.
This approach provides a simple and effective way for a higher LiDAR performance than that obtained from the flash LiDAR approach, wherein an LNA and the related signal processing circuits should be implemented in every small pixel area. This is because the STUD approach uses a single output, and therefore, it does not impose any restrictions on the ROIC pixel design area. Additionally, parallel operation among multiple detectors is not required for achieving high resolution and performance. Moreover, the detection of short optical pulses and a wide bandwidth at a given FoV or photosensitive area can be easily achieved by dividing the total detection area into a small finger area for one pad. The number of TIAs can then be increased according to the increased number of pads to cover the same detection area. However, the rest of the design remains the same. Consequently, the LiDAR receiver structure using the STUD approach is advantageous because the photosensitive area and FoV can be increased without sacrificing the signal bandwidth or increasing the optical pulse width.
Based on the aforementioned principle, the implemented schematics of the newly designed STUD PD and its circuits are shown in Fig. 1(a). The designed STUD PD possessed a total of 32 fingers, eight of which were grouped into one pad and allowed optical short pulses to be detected. Finally, a total of four finger groups were contained in one STUD PD. The layout of the STUD PD with actual photosensitive dimensional information is shown in Fig. 1(b). A rectangular photosensitive shape is chosen as the PD design to reduce the junction capacitance as high as possible. This was achieved by reducing the photosensitive area by approximately 37%, compared with that of a general circular PD, to obtain the required diagonal FoV of a region of interest (ROI). [48].
The designed STUD PD, with a planar structure that minimizes dark current, was fabricated using a commercial InP/InGaAs PIN-PD process (Phosem Co., Ltd., South Korea). As shown in Fig. 2(a), for the cross-section of the fabricated device, the epitaxial layer consists of an InP substrate of 350 µm, an InP buffer layer of 0.4 µm (1 × 1017/cm3), an i-In0.53Ga0.47As absorption layer of 3.5 µm (1 × 1015/cm3), a cladding layer of 1 µm (1 × 1015/cm3), and an InGaAs cap layer of 0.1 µm (1 × 1019/cm3). After a p+ diffusion region was formed by the diffusion of Zn to define the photosensitive region, SiO2 was deposited and p-contact was made sequentially. Finally, the anti-reflective (AR) coating, p-pad, and n-contact were fabricated in order. An optical microscopy image of the fabricated STUD PD is shown in Fig. 2(b), in which 32 finger PD can be identified.
Fig. 2. (a) Cross-section of the InP/InGaAs PIN-PD for STUD PD, (b) optical microscopy image of the fabricated InP/InGaAs PIN-PD based on the STUD technique, and (c) total capacitance (red) and unit capacitance (black) for conventional PDs.
To confirm the unit capacitance of the PIN PDs based on the size of their photosensitive area, the total capacitances and unit capacitances for various diameters were measured using the same foundry process, as shown in Fig. 2(c). As expected, as the size increased, the unit capacitance decreased. By contrast, when the size decreased, the unit capacitance increased, which indicated that lowering the capacitance by simply reducing the area was difficult. For the STUD approach, this indicated that the finger size must be reduced further and the number of groups must be higher than what was initially prepared. Based on the measured results, the area of a single finger and finger group were carefully adjusted to 1,256 × 19 µm2 and 1,256 × 229 µm2, respectively. Here, the unit capacitance for a single finger and the total capacitance for eight fingers including the space between fingers were expected to be 1.1 pF and 13.3 pF, respectively. By including four finger groups in one STUD PD, the total photosensitive area was expected to be higher than 1,256 × 949 µm2, including the space between finger groups. However, if a conventional single detector was used for this photosensitive area, this PD would possess a high capacitance of approximately 55.3 pF, thereby making its implementation impractical.
To implement a LiDAR receiver board using the fabricated STUD PD, as shown in Fig. 3(a), each pad of the STUD PD (#1 in Fig. 3(a)) was connected to the input of the corresponding chip-on-board TIA (#2 in Fig. 3(a)) by wire bonding using the direct chip attachment technique [49]. To alleviate the noise at the power source and prevent local oscillations owing to the high TIA gain, a single-layer capacitor (SLC) (#3 in Fig. 3(a)) was placed for each TIA. Four additional SLCs (#4 in Fig. 3(a)) were placed beside the STUD PD to provide a sufficiently virtual short path to the ground for enhanced detector performance. All wire bonding was carefully conducted with a minimal wire-bonding length. To ensure a reliable and sufficient SNR, the sensor board was designed to have a transimpedance gain of > 80 dBΩ through the TIAs with low noise characteristics and an additional voltage gain of > 10 dB for the post-amplifier. Figure 3(b) shows the final LiDAR receiver board, including the STUD PD, four TIAs, eight SLCs, and a post-amplifier. To suppress the noise in the compact module size, the final outputs from the optical receiver were obtained from two micro-receptacle connectors in differential mode.
Fig. 3. Images of (a) the fabricated STUD PD and related components in chip-on-board and (b) fabricated STUD PD-based sensor board.
2.2 High-performance all-fiber MOPA laser for a low-SWaP LiDAR
For an eye-safe LiDAR with high resolution and performance, an all-fiber 1.5 µm MOPA laser is a good candidate because it provides a narrow beam diameter for remote targets. This is owing to the high beam quality of the Gaussian profile, and reliable operation is possible because no bulk optics are included. In general, the multi-stage configuration in an all-fiber MOPA laser is designed for high optical power applications such as free-space optical communication and remote sensing [41–43], using multiple pumping LDs and a large amount of doped fiber. Therefore, the all-fiber MOPA structure must be optimized to meet the required low-SWaP conditions at a low cost. In this section, the design and implementation of an optical circuit and electrical control board for a custom-designed all-fiber MOPA laser that satisfies low-SWaP requirements, are described.
As shown in Fig. 4(a), the two amplification stages are adopted to satisfy the required low-SWaP conditions. A relatively inexpensive erbium-doped fiber (EDF) was used in the first amplification stage to efficiently amplify the low-power seed LD based on core pumping. Moreover, for the final amplification stage, which determined the final output, an erbium/ytterbium co-doped double-clad fiber (EYDCF) based on cladding-pumping was used instead of the EDF. This is because EDF is limited at the high-power amplifying stage owing to the excited-state absorption [50] and concentration quenching [51]. The fiber length of all amplifying stages was carefully optimized for high efficiency and low cost.
A distributed feedback (DFB) LD (EMCORE, 1752A-19-B-S-10) with a narrow linewidth (< 1 MHz) was used as the signal source for the overall MOPA operation with an average power of 10 dBm and a center wavelength of ITU-52ch (1535.82 nm). The feedback was amplified by the first amplification stage of the EDF amplifier (EDFA) with core pumping using a 976 nm pump LD (Coset, CM0001C) delivering an average power of 330 mW. The pump LD was connected to a 980/1550 wavelength division multiplexing coupler (WDM; Sungwoo engineering, SWE-S200-A), and the output port of the WDM was connected to the EDF. Based on the optimization experiment, the optimized 6 m single-mode (SM) EDF (OFS, GP980) of a 0.24 core NA was used for the first amplification stage. After the EDFA, the isolator/filter wavelength division multiplexer hybrid (IDWDM; AFR, SR15919B) was used to reduce the number of components and final volume. The IDWDM contained an isolator (50 dB isolation) to prevent the damage caused by the back return signal and a narrow band DWDM filter (0.36 nm pass-band width) to minimize unnecessary signals.
The SM EYDCF, which has a rare-earth-doped core and double-clad fiber structure for cladding-pumping, was used for the last amplification stage of the erbium/ytterbium-doped amplifier (EYDFA). Herein, the advantages of the SM EYDCF, such as high power and brightness, were utilized [52]. A (1 + 1) × 1 combiner (Huscent, MCS-01112SM) was used as a pump combiner to connect the pump laser to the EYDCF. A multimode 940 nm pump LD (BWT, K940EB2RN-10), which can provide up to 10 W, was connected to the input port of the combiner. Furthermore, the output port of the combiner was connected to an SM EYDCF (Nufern, SM-EYDF-10P/125-XP) with a core/clad NA of 0.11/0.46 for clad pumping. By optimizing the efficiency and implementation cost of the SM EYDCF, its length was carefully chosen to be 3 m. This is because the power efficiency does not increase beyond a certain length owing to the reabsorption process [43]. Finally, the end cap was manufactured by splicing a 100 µm-sized core-less fiber into a customized fiber (Huscent, G.657.B3) for high banding, and polished to 8° in the form of an FC connector, was connected to emit the final optical output. The residual pump power and back reflection can be alleviated simply, and the possibility of facet damage was minimized by the fabricated angled end-cap connector. These optical circuits were controlled by a separately designed electrical board and a program to adjust various driving conditions, such as the optical pulse width of MOPA, input current, and the temperature using a thermoelectric cooler (TEC). For example, an electrical pulse can be adjusted by a field-programmable gate array (FPGA) to generate an electrical short pulse with an arbitrary pulse width (1–10 ns) to the input of the DFB LD, which allowed the desired optical pulse width to be adjusted. The fabricated MOPA laser module with all the related components is shown in Fig. 4(b). Including the jig and the control module, as well as the optical circuit, the total dimensions and weight of the module were 11.5 × 8 × 2.7 cm3 and 188 g, respectively, which coincided with the low-SWaP requirements.
2.3 LaserEye2 LiDAR prototype based on the STUD approach
Figure 5 shows the overall system configuration for the implemented LaserEye2 LiDAR prototype using the STUD-based InP/InGaAs PIN-PD receiver and customized MOPA laser. In the optical transmission section, the 1.5 µm optical short-pulse generated by the fabricated MOPA laser was first collimated to a beam waist diameter of 1.6 mm and a full-angle divergence of 0.073° using a commercial collimation lens (Thorlabs, F240APC-1550). Under these conditions, the beam size at a distance of 50 m is calculated to be 6.56 cm. To cover the entire FoV range for a given ROI, the generated optical pulses were redirected using the 2-axis Galvano scanner (Cambridge Technology, 6210H) according to the generated X and Y scan coordinates. These coordinates were calculated by the FPGA on the main board using a snake scan. The backscattered signals of each point reflected on the target were collected using a customized lens, and the signals detected by the STUD PD device on the sensor board were transferred to the main signal process board with a differential output through subsequent amplification circuits. The customized receiver lens was optimally designed and manufactured using Zemax simulation based on a four-stage multi-element lens system to obtain the required horizontal FoV (26°) for the designed detection dimension (1,256 × 949 µm2). Finally, an aperture size of 28 mm, effective focal length of 2.76 mm, and entrance pupil diameter of 5.22 mm were achieved. In addition, a bandpass filter (Edmund optics, F85-903) with a full width at half-maximum (FWHM) of 50 nm and > 97% transmission at a wavelength of 1.5 µm was mounted on the front to prevent the increase in the thermal noise in the receiver from the SBR and other optical noise sources. From the simulation results, the customized receiving lens with the designed STUD PD with a diagonal length of approximately 1.57 mm was expected to have a diagonal FoV of 32.6° with good background optical noise suppression.
Fig. 5. Schematic of the LaserEye2 3D LiDAR prototype system configuration. HP: high power; HV: high voltage; LP: low power; LV: low voltage; MOPA: master oscillator power amplifier; PRF: pulse repetition frequency; SPI: serial peripheral interface; DAC: digital-to-analog converter; FPGA: field-programmable gate array; TIA: transimpedance amplifier; CPU: central processing unit; UDP: user datagram protocol; PC: personal computer.
The received LiDAR signals were first determined by a simple comparator and threshold voltage controlled by an internal digital-to-analog converter. The detected signals were then transferred to the time-to-digital converter (TDC) block, in which the ToFs were calculated. The TDC block was implemented in the FPGA by using two orthogonal digital signals with the frequency of 300 MHz, which finally provides 833.33 ps resolution for the ToF measurement. Thereafter, based on the calculated ToFs and their corresponding scan coordinates, 3D point cloud data were generated. Additionally, 3D images were displayed using the 3D point cloud in real-time by applying a specific false color coding according to the distance or height for each point through a separately developed 3D visualization program using the Rapidform library (3D Systems).
The final LaserEye2 LiDAR prototype was designed to have a resolution of 320 × 240 pixels with 76,800 points per frame and a frame rate of up to 15 Hz. All the components of the LaserEye2 LiDAR prototype were successfully integrated with a total size of 10 × 10 × 12 cm3 and a weight of < 1.08 kg, as shown in Fig. 6(a). A single external connector (TE Connectivity, M12) was used to provide an Ethernet interface for communication and a DC power supply to all electrical modules inside. The overall power consumption, including all integrated components, was designed to be under 28 W to comply with the low-SWaP requirements. In addition, a separately developed Python control program based on user datagram protocol (UDP) communication was developed, which allowed access to all subsystems and components from power on/off to the Galvano scanner range settings. This further optimized the operating conditions.
Fig. 6. (a) External and (b) internal structure (without the wire connection) of the LaserEye2 LiDAR prototype.
The detailed internal configuration of the LaserEye2 LiDAR prototype is illustrated in Fig. 6(b). As shown in the figure, to prevent any internal reflected signals from entering the receiver in the LiDAR module, the 2-axis Galvano scanner located at the bottom of the front side is covered by a black optical shield. The receiver module located at the top of the front side, including the fabricated lens and sensor board, was designed to completely encapsulate the LiDAR receiver using a jig for better optical isolation between the transmitting and receiving paths. The customized MOPA laser, main board with the Galvano scanner driver, power board for the power supply, fan for minimizing thermal effects, and M12 connector were located on the right, center, top, bottom, and left of the module, respectively.
3. Results and discussion
3.1 LiDAR receiver with the fabricated STUD PD
The block diagram for the manufactured LiDAR receiver characterization setup is shown in Fig. 7(a) to explain our approach to scrutinizing the receiver characteristics in the LiDAR prototype. The optical pulse was generated from the same 1.5 µm MOPA laser in the LaserEye2 LiDAR prototype and first attenuated by a variable optical attenuator (VOA; EXFO, FVA-600). Next, the resulting pulse was collimated toward the receiving lens by a triplet collimation lens (Thorlabs, TC25APC-1550), thereby resulting in a good Gaussian beam diameter of ∼5 mm [53]. The emitted optical pulse was collected by a receiving lens 1 m away and a small beam spot was made on the photosensitive area of the sensor board, which was mounted on the XYZ motorized linear stage (Newport, UTS 100 [3-axis]). The MOPA laser was controlled to produce similar optical pulse conditions such as a PRF of 120 kHz and a pulse width of 2 ns, by a waveform generator (Keysight, 33500 B), and an in-house control program was developed. The XYZ motorized linear stage adjusted the position of the sensor board precisely to the receiving lens at a micrometer scale using the developed in-house LABVIEW program and a dedicated controller (Newport, ESP301) through the IEEE 488 interface. This allowed the beam spot to illuminate in an arbitrary position on the photosensitive area. Finally, the electrical responses for all interesting positions on the photosensitive area were obtained using an oscilloscope (Keysight, MSOX4154A) and the developed LABVIEW program, thereby providing the detailed characteristics of the signal and noise for the test device in the LiDAR receiver.
Fig. 7. (a) Characterization setup of the LiDAR receiver for analyzing the STUD PD, (b)–(d) STUD PD characteristics: (b) signal output, (c) overall noise, and (d) derived SNR.
By replacing the required components in this setup, the signal output, overall noise, and resultant SNR of the sensor board are first extracted at a fixed optimal position in the sensor board, as shown in Figs. 7(b)–7(d). The results were obtained for the optical input peak power range of 0.983–983 µW at a step of 1 dB using the VOA, and the measurements were repeated 10,000 times to obtain sufficiently reliable outcomes. For the SNR calculation, the signal output for the LiDAR receiver was obtained from the average of the measurements. The overall noise consisting of thermal noise and shot noise generated in the LiDAR receiver was determined from the standard deviation. Generally, the overall noise was nearly the same as the read noise while the shot noise from the optical input was negligible when compared with the read noise. The overall noise began to proportionally increase according to the increased shot noise owing to the optical input peak power.
In the LiDAR using a comparator to obtain a ToF at a low cost, such as in the proposed LaserEye2 LiDAR prototype, the SNR obtained from the read noise was important because the received LiDAR signals must be greater than the read noise and not the overall noise. The overall noise included the shot noise from the optical source, as well as the thermal noise of the background light source. The read noise and SNR of the read noise can also increase in various situations, such as by increased stray light and/or an increased working temperature in a PD and LiDAR receiver board. Therefore, optimizing the detection threshold voltage, which determines whether the input is considered a signal or not, is important. In general, if the threshold is set near the read noise, spurious false alarms are generated owing to the noise output, thereby reducing the quality of the 3D point cloud. However, if the threshold is adjusted to be sufficiently high, it will no longer generate false alarms; although in this case, signal detection will be substantially decreased owing to the weak reflected signals from remote objects or low reflection on remote surfaces. This finally deteriorates the quality of the 3D point cloud.
As shown in Fig. 7(b), the signal output increases linearly with increasing input power after the optical power input is -25 dBm; however, the rate of increase slightly decreases after an amplitude of ∼1 V near the input power of -2 dBm, which indicates that the fabricated receiver is saturated at this input level. The read noise in the fabricated sensor board based on the STUD approach was observed to be ∼3.8 mVrms and was maintained at a level below the input power of -25 dBm as shown in Fig. 7(c). A rapid reduction in the overall noise above the saturation power was also observed, which can be explained by the reduced fluctuation caused by the saturation of the sensor output. Figure 7(d) shows the SNR calculated from the overall noise and read noise. The obtained SNR for the overall noise was saturated at approximately 30 dB at an input power of -10 dBm owing to the increased shot noise and gain saturation. In contrast, the SNR for the read noise linearly increased up to 49 dB before the output saturation at an input power of -2 dBm.
The main advantage of using this setup is that it allows the local SNR results to be measured for a specific point in the photosensitive area, which consequently generates a 2D SNR image similar to the laser beam-induced current (LBIC) approach [54–56]. To obtain the 2D SNR image, SNR measurements were successively conducted at every position by changing the sensor board moved in a snake scan. Another feature of this method was that the obtained SNR included all aspects of the optical path in the system, and consequently, allowed the characteristics very close to the actual receiving signal quality of a LiDAR module to be achieved. This was especially true when the receiving optics were nearly the same. Finally, the detector information regarding the photosensitive region can be obtained. This included the non-uniformity within the PD or defective region within the photosensitive area. By using such a setup, the detailed characteristics of the PD or the overall LiDAR receiver were easily acquired.
The peak power and pulse width used for the SNR image, in this case, were approximately 200 µW and 2 ns, respectively. A 2D SNR image with a sufficiently large beam diameter (> 400 µm) was first obtained, as shown in Figs. 8(a)–(c). The 2D scan results in Figs. 8(a) and (b) were obtained in steps of 50 µm for the 2 × 2 mm2 scan area to include all photosensitive areas of the fabricated STUD PD. The photosensitive area measured 1,290 × 975 µm2 for the FWHM of the vertical and horizontal regions, which was 2.5% larger than the designed dimension (1,256 × 949 µm2). Figure 8(c) shows the intensity vs. y-axis cross-section graph of the 2D SNR image that clearly reveals the uniformity among all the photosensitive areas. In addition, as shown in Figs. 8(d)–(f), a 2D SNR image with a narrow beam spot (< 7 µm) using a commercial receiving lens (NAVITAR, SWIR-8) was obtained to observe the difference in the images when using a wider beam. The 2D scan was executed in steps of 4 µm for an area of 1.6 × 1 mm2, and the remaining conditions were maintained. Owing to the narrow beam, the exact shapes of 32 fingers with a pitch of 30 µm are observed, as shown in Figs. 8(d)–(f). Additionally, the final photosensitive area was measured to be 1,266 × 956 µm2, which almost coincided with the designed dimensions of 1,256 × 949 µm2. Therefore, from this experiment, the test beam size did not substantially change the photosensitive area; however, the resolution of the 2D SNR image could be determined. To provide exact information, a y-axis cross-sectional graph is shown in Fig. 8(f). This graph reveals the exact information on the photosensitive area, including the overlapping area of each finger and areas between the fingers. Additionally, the repeated change in the SNR around each finger was identified and can be explained according to the layout and vertical structure of the PD; however, this is beyond the scope of this study.
Fig. 8. Results of measuring the sensitivity of the STUD PD: (a)–(c) result obtained from the customized receiving lens, (d)–(f): result obtained from the commercial receiving lens, (a), (d) 2D SNR image, (b), (e) tilted 2D SNR image and (c), (f) cross-section of the y-axis with respect to intensity.
3.2 Low-SWaP MOPA laser
To evaluate the fabricated MOPA laser module, the peak wavelength of the optical signal for a PRF of 292.88 kHz and pumping current of 3.35 A was measured to be 1535.82 nm using an optical spectrum analyzer (OSA; Agilent, 86140B), as shown in Fig. 9(a). Additionally, by applying various input currents of the pump LD (Pump LD2 of Fig. 4(a)) in the last amplifying stage at the same PRF, the average power was measured using a power meter (Gentec-eo, UP19K-30H-W5-D0). The pump LD could be driven up to 10 A. However, the operating current range was 2–4 A to meet the required low-SWaP conditions, which corresponded to an average optical output in the range of 40–420 mW based on the measured results, as shown in Fig. 9(b).
Fig. 9. Customized MOPA laser measurement result: (a) center wavelength, (b) optical average power (PRF = 292.88 kHz), (c) optical peak power (PRF = 292.88 kHz), and (d) peak power with variable PRF (input current = 3.35 A).
Since the exact optical peak power of the customized MOPA laser is important in a conventional pulse-mode ToF LiDAR module, this peak power was measured separately using a digital communication analyzer (DCA; HP, 83480A). Because the 83480A DCA can measure a peak power up to 2 mW, the peak power was measured by applying an intentional attenuation of 63 dB. After compensating for the attenuation, the final peak power was derived to be ∼1.2 kW at a pumping current of 4 A, as shown in Fig. 9(c). However, because the power consumption of the LaserEye2 LiDAR prototype is restricted, the final pumping current was chosen to be 3.35 A, such that the overall power consumption of the customized MOPA laser could be sustained at 12 W and an output peak power of ∼880 W could be achieved. In general, because the peak power of the MOPA laser module is dependent on the applied PRF even at the same pumping current, increasing the peak power is possible if the PRF is decreased to widen the detection range. This is also possible in the current LiDAR prototype by decreasing the frame rate. The peak power dependency on the PRF was measured as shown in Fig. 9(d), which shows the peak power of ∼2.16 kW at 73.2 kHz and ∼265 W at 1.17 MHz for a pumping current of 3.35 A.
The stability of the wavelength and the output power are also important factors for applying the fabricated MOPA in various LiDAR environments. For this purpose, the 1 h aging test for wavelength stability was conducted in steps of 1 min the same OSA, as shown in Fig. 10(a). The result was measured to be an average of 1535.817 nm, a standard deviation of 0.0024 nm, and an RMS stability of 0.000156%, thereby indicating that the TEC controls were sufficient for various environmental changes, such as heat and electrical noise generated in the module. The average power stability was verified at a step of 1 s for 1 h PRF of 292.8 kHz, and pumping current of 3.35 A. The measured result was an average of 328.2 mW and a standard deviation of 2.2 mW, as shown in Fig. 10(b), which indicated that the RMS stability was approximately 0.69%.
Fig. 10. Results of the 1 h aging test of the customized MOPA laser: (a) wavelength stability and (b) average power stability.
3.3 High-resolution and wide FoV LiDAR prototype integration
As mentioned in Section 3.1, the characterization of the overall noise, as well as the read noise, on the sensor board is important for determining the threshold voltage for optimal detection. This is especially true when all components are integrated into a LiDAR module. When the LaserEye2 LiDAR prototype was assembled within a small size of 10 × 10 × 12 cm3 including all necessary functional blocks, newly introduced noise could be easily induced, thereby resulting in much higher overall noise than that obtained in the separate sensor board evaluation in Section 3.1. Moreover, the optimum threshold voltage can easily change over time owing to the changes in circumstances, such as background light and operating temperature. Therefore, to sustain the best LiDAR operating conditions, monitoring and adjusting the optimum threshold is important. However, the optimal value is difficult to determine in such an integrated environment. If a high-speed analog-to-digital converter is available in a LiDAR sensor, this problem can be easily alleviated; however, this is not the case for a low-cost LiDAR. Therefore, in the LaserEye2 LiDAR prototype, a simple method was developed and adopted by using the same TDC for detecting the LiDAR signal without any additional components. Additionally, the optimum threshold was determined for the best performance of the final assembled LiDAR according to the continuously changing environments.
To characterize the noise distribution of the integrated sensor boards for the optimum threshold, a false detection probability plot was used, which was obtained from the measurement of the detection rate at a given threshold voltage without any incoming LiDAR signal. The noise distribution generally decreases from 100% to 0% on increasing the threshold level because the chance of signal detection decreases at an increased threshold. To prove its effectiveness and show the difference, we included a standard sensor with a bandpass filter, which was installed in the receiver module, as well as a degraded sensor and a standard sensor without a bandpass filter. The measured false detection probability plots for the three sensor boards in the integrated LaserEye2 LiDAR prototype are shown in Fig. 11(a).
Fig. 11. Integrated noise characteristics for the sensors in the assembled LaserEye2 LiDAR prototype under various conditions: (a) false detection probability and (b) probability density.
To demonstrate the difference intuitively, the derivatives of Fig. 11(a) are calculated and are plotted in Fig. 11(b). As shown in the figure, the peak position and width are 63 and 20 DN for the standard sensor board, respectively, whereas the other two boards show considerably high peak positions and/or wide peak widths. In particular, the peak position and width of the degraded board increased to 91 DN and 38 DN, respectively. Only the peak position was increased to 155 DN for the standard sensor without the filter; however, the peak width remained the same. From the measured results, the background light can be concluded to increase the threshold level but not the peak width. This indicated that background light did not increase the read noise component substantially and only increased the signal offset in the developed LiDAR receiver. Additionally, the peak position and width for the degraded sensor were increased, thereby indicating that the threshold level should increase further than the increased noise offset. Moreover, this was regarded as a worse situation than when only the peak position was increased. Thus, the proposed approach confirmed the internal state of the LiDAR receiver and sensor conditions (e.g., sensor degradation) in real-time.
Using the explained approach, the optimum operational threshold of the LaserEye2 prototype was chosen to be 88 DN with the false alarm rate of < 2% considering the worst outdoor daylight condition. Based on the optimally chosen threshold voltage, the 3D point cloud data were finally generated based on the calculated coordinates and subsequently transferred to the outside of the LiDAR module as UDP packets. To display 3D scenes from the transferred UDP packets, we developed a 3D visualization program with various false color options. Finally, the LaserEye2 LiDAR prototype was optimized, and the required performance was achieved under low-SWaP conditions such as a total size of 10 × 10 × 12 cm3, weight of < 1.08 kg, and power consumption of < 28 W. Additionally, a detection range of > 55 m, diagonal FoV of > 31°, and frame rate of up to 15 Hz can be achieved, simultaneously, while the horizontal and vertical resolutions of the obtained 3D images can be up to 320 × 240 pixels with a high angular resolution of approximately 0.08° in the horizontal and vertical directions. Several representative 3D scenes were obtained using the LaserEye2 LiDAR prototype for various environments, as shown in Fig. 12. In particular, the developed 3D image visualization program was executed to display the false color coding for each point according to the distance or height of the point such that the point could be clearly recognized. Figures 12(a)–(c) are based on the distance, and Fig. 12(d) shows the height.
Fig. 12. Real-time 3D images obtained from the LaserEye2 LiDAR prototype in a single frame (320 × 240 pixels): (a) indoor laboratory with various products (< 5 m), (b) a person in an indoor corridor (< 30 m), (c) outdoor rooftop with a wire (daylight, > 55 m), and (d) various targets in an outdoor environment (daylight, < 50 m).
As shown in Fig. 12, because the images were obtained in a single frame without any accumulation of point clouds from multiple frames, the high-resolution image can be used in real-time operation under indoor short-range and outdoor long-range conditions. Based on the achieved point cloud data, thin and small objects, such as a 3.3 mm-thick wire at a distance of 15 m in Fig. 12(c) and (a) rod of a pergola at a distance of 28 m in Fig. 12(d), were clearly recognized in real-time in a single frame. This indicated that sufficient resolution in the horizontal and vertical directions was successfully achieved in this low-SWaP LiDAR module. Such high-resolution detection in a single frame with a sufficiently wide FoV can be actively applied in power-restricted unmanned applications such as UAVs, which require prompt detection and response to any potentially hazardous objects through real-time recognition at a sufficiently long distance.
In our use case of the fabricated LaserEye2 LiDAR prototype, the real-time high-resolution 3D images were mainly required only under 50 m in the mission; however, the detection distance can be easily increased in the module by decreasing the frame rate at a given resolution and/or decreasing resolution at a given frame rate without the increase in power consumption of the customized MOPA laser. In conclusion, even though the operating conditions are restricted by the low-SWaP requirements, the properties including the high-resolution 3D image and wide FoV were simultaneously achieved and confirmed because of the STUD approach. Additionally, the eye-safe MOPA laser was successfully characterized under low-SWaP conditions to provide good beam quality, high PRF, and sufficient optical peak power.
4. Conclusion
A large photosensitive area STUD PD for wide FoV and high-speed detection in a LiDAR receiver and a customized MOPA laser for high-performance and low-SWaP operation was designed and fabricated. The LaserEye2, which is a 3D LiDAR prototype based on the STUD approach, was assembled, and verified for the required low-SWaP conditions and high LiDAR performance.
More specifically, the fabricated STUD-based InP/InGaAs PIN-PD was designed to have a large photosensitive area of 1,256 × 949 µm2 consisting of 32 finger PDs and was placed in the LiDAR receiver along with a TIA. The TIA possessed a transimpedance gain of 80 dBΩ, and a post-amplifier was also included for a voltage gain of > 10 dB. The final photosensitive area was measured to be 1,290 × 975 µm2, and the characteristics of the overall LiDAR receiver showed a signal output of up to 1 V, read noise of 3.8 mVrms, and SNR of up to 32 dB from the overall noise. In addition, the normalized sensitivity and uniformity of the photosensitive area were measured and analyzed for the broad and narrow beams using the proposed 2D scan measurements. The dimensions of the fabricated 1.5 µm MOPA laser for the low-SWaP requirement are 11.5 × 8 × 2.7 cm3, weight is 188 g, and individual power consumption is < 12 W. A peak power of 2.16 kW, an average power of 328.2 mW, and a PRF of 0.732–1.17 MHz were achieved. To evaluate the performance reliability of the fabricated MOPA laser module, the RMS stability of the wavelength and average power were measured to be 0.000156% and 0.69%, respectively.
The dimensions of the fabricated LaserEye2 LiDAR prototype, including the developed receiver board, customized MOPA laser, and all other related submodules, are 10 × 10 × 12 cm3, weight is 1.08 kg, and power consumption is < 28 W, which complies with the required low-SWaP conditions. Under these restrictions, small objects, such as a wire of 3.3 mm, were clearly and promptly detected in a single frame within a maximum range of 55 m using the fabricated LiDAR module. These optimum performances were achieved using the explained optimization techniques without introducing any additional modules. The final diagonal FoV was measured to be > 31° with the fabricated STUD PD and designed receiving lens. Within the achieved diagonal FoV, the final obtained 3D images possessed a high resolution of 320 × 240 pixels with angular, horizontal, and vertical resolutions of approximately 0.08° each at a frame rate of up to 15 Hz.
In conclusion, judging from the various results obtained from the fabricated LaserEye2 LiDAR prototype, the STUD approach is beneficial for easily sustaining high LiDAR performance, such as high resolution in the horizontal and vertical directions for a wide FoV even under a given SWaP requirement, which can be adopted for various power-restricted LiDAR applications such as UAVs and robots.
Funding
Ministry of Science and ICT, South Korea (19PR1230); Ministry of Trade, Industry and Energy (21FB6310); Ministry of SMEs and Startups (21FB2510);
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. Y. Gao, S. Zhang, X. Wang, R. Shi, Y. Zhao, and Z. Li, “Reconfigurable optical time delay array for 3D lidar scene,” Opt. Express 28(26), 39688–39699 (2020). [CrossRef]
2. J. Li, H. Bao, X. Han, F. Pan, W. Pan, F. Zhang, and D. Wang, “Real-time self-driving car navigation and obstacle avoidance using mobile 3D laser scanner and GNSS,” Multimed Tools Appl 76(21), 23017–23039 (2017). [CrossRef]
3. L. Chen, Y. He, J. Chen, Q. Li, and Q. Zou, “Transforming a 3-D LiDAR Point Cloud Into a 2-D Dense Depth Map Through a Parameter Self-Adaptive Framework,” IEEE Trans. Intell. Transport. Syst. 18(1), 165–176 (2017). [CrossRef]
4. J. Hecht, “LiDAR for Self-Driving Cars,” Opt. Photonics News 29(1), 26–33 (2018). [CrossRef]
5. U. Lee, J. Jung, S. Jung, and D. H. Shim, “Development of a self-driving car that can handle the adverse weather,” Int.J Automot. Technol. 19(1), 191–197 (2018). [CrossRef]
6. M. Li, H. Zhu, S. You, L. Wang, and C. Tang, “Efficient Laser-Based 3D SLAM for Coal Mine Rescue Robots,” IEEE Access 7, 14124–14138 (2019). [CrossRef]
7. R. Moss, P. Yuan, X. Bai, E. Quesada, R. Sudharsanan, R. L. Stann, J. F. Dammann, M. M. Giza, and W. B. Lawler, “Low-cost compact MEMS scanning ladar system for robotic applications,” Proc. SPIE 8379, 837903 (2012). [CrossRef]
8. B. L. Stann, J. F. Dammann, M. M. Giza, R. R. Gregory, P. S. Jian, and W. B. Lawler, “Low-cost compact ladar sensor for ground robots,” Proc. SPIE 7323, 73230X (2009). [CrossRef]
9. L. Wallace, A. Lucieer, C. Watson, and D. Turner, “Development of a UAV-LiDAR System with Application to Forest Inventory,” Remote Sens. 4(6), 1519–1543 (2012). [CrossRef]
10. B. Brede, A. Lau, H. M. Bartholomeus, and L. Kooistra, “Comparing RIEGL RiCORTER UAV LiDAR Derived Canopy Height and DBH with Terrestrial LiDAR,” Sensors 17(10), 2371 (2017). [CrossRef]
11. J. Zhou and K. Qian, “Low-voltage wide-field-of-view lidar scanning system based on a MEMS mirror,” Appl. Opt. 58(5), A283–A290 (2019). [CrossRef]
12. L. Ye, G. Gu, W. He, J. Lin, and Q. Chen, “Adaptive Target Profile Acquiring Method for Photon Counting 3-D Imaging Lidar,” IEEE Photonics J. 8(6), 1–10 (2016). [CrossRef]
13. P. Bhargava, T. Kim, C. V. Poulton, and J. Notaros, “Fully Integrated Coherent LiDAR in 3D-Integrated Silicon Photonics/65 nm CMOS,” 2019 Symposium on VLSI Circuits C262-C263 (2019).
14. Z. Li, Z. Zang, Y. Han, L. Wu, and H. Y. Fu, “Solid-state FMCW LiDAR with two-dimensional spectral scanning using a virtually imaged phased array,” Opt. Express 29(11), 16547–16562 (2021). [CrossRef]
15. G. Kang, D. Lee, H. Yoon, and H. Park, “High-speed and power-efficient beam-steering using 1 × 16 silicon optical phased array with electro-optic phase shifters,” Proc. SPIE 11286, 21 (2020). [CrossRef]
16. Y. Li, B. Chen, Q. Na, and Q. Xie, “Wide-steering-angle high-resolution optical phase array,” Photonics Res. 9(12), 2511–2518 (2021). [CrossRef]
17. M. Okano and C. Chong, “Swept Source Lidar: simultaneous FMCW ranging and nonmechanical beam steering with a wideband swept source,” Opt. Express 28(16), 23898–23915 (2020). [CrossRef]
18. J. D. Chen, H. L. Ho, H. L. Tsay, Y. L. Lee, C. A. Yang, K. W. Wu, J. L. Sun, D. J. Tsai, and F. Y. Lin, “3D chaos lidar system with a pulsed master oscillator power amplifier scheme,” Opt. Express 29(17), 27871–27881 (2021). [CrossRef]
19. J. He, T. Dong, and Y. Xu, “Review of Photonic Integrated Optical Phased Arrays for Space Optical Communication,” IEEE Access 8, 188284–188298 (2020). [CrossRef]
20. H. M. Tulldahl and H. Larsson, “Lidar on small UAV for 3D mapping,” Proc. SPIE 9250, 925009 (2014). [CrossRef]
21. Schwarz, “Mapping the world in 3D,” Nat. Photonics 4(7), 429–430 (2010). [CrossRef]
22. C. Glennie and D. D. Lichti, “Static Clibration and Analysis of the Velodyne HDL-64E S2 for High Accuracy Mobile Scanning,” Remote Sens. 2(6), 1610–1624 (2010). [CrossRef]
23. K. I. Schultz, M. W. Kelly, J. J. Baker, and M. H. Blackwell, “Digital-Pixel Focal Plane Array Technology,” Lincoln Laboratory Journal20(2), (2015).
24. C. Zhang, S. Lindner, I. M. Antolović, J. Mata Pavia, M. Wolf, and E. Charbon, “A 30-frames/s, 252×144 SPAD Flash LiDAR With 1728 Dual-Clock 48.8-ps TDCs, and Pixel-Wise Integrated Histogramming,” IEEE J. Solid-State Circuits 54(4), 1137–1151 (2019). [CrossRef]
25. R. Lange, “3D Time-of-flight distance measurement with custom solid-state image sensors in CMOS/CCD-technology,” PhD thesis, University of Siegen (2000).
26. A. Hassebo, B. Salas, and Y. Y. Hassebo, “Monostatic and bistatic lidar systems: simulation to improve SNR and attainable range in daytime operations,” Proc. SPIE 10094, 1009421 (2017). [CrossRef]
27. Y. Y. Markushin, G. S. Pati, and R. Tripathi, “Multi-pulse detection technique to improve the timing/range resolution in a scanning LADAR system,” Proc. SPIE 9342, 93420L (2015). [CrossRef]
28. E. Garcia and H. Lamela, “Low-cost three-dimensional vision system based on a low-power semiconductor laser rangefinder and a single scanning mirror,” Opt. Eng. 40(1), 61–66 (2001). [CrossRef]
29. C. Zhang, X. Sun, R. Zhang, S. Zhao, W. Lu, and Z. Fan, “Impact of solar background radiation on the accuracy of wind observations of spaceborne Doppler wind lidars based on their orbits and optical parameters,” Opt. Express 27(12), A936–A952 (2019). [CrossRef]
30. F. Seeber, “Light sources and laser safety,” Fundamentals of Photonics 1(2), 3 (2007). [CrossRef]
31. M. Pfennigbauer and A. Ullrich, “Multi-Wavelength Airborne Laser Scanning,” Proc. of the International Lidar Mapping Forum (ILMF) (2011).
32. J. P. Siepmann, “Fusion of current technologies with real-time 3D MEMS ladar for novel security and defense applications,” Proc. SPIE 6214, 621409 (2006). [CrossRef]
33. T. Plank, M. Czaputa, E. Leitgeb, S. Sheikh, N. Djaja, B. Hillbrand, P. Mandl, and M. Schonhuber, “Wavelength selection on FSO-links,” Proceedings of the 5th European Conference on Antennas and Propagation (EUCAP), 2508–2512 (2011).
34. FindLight, OT-03: Er: Glass Laser Transmitters with Diode Pumping, https://www.findlight.net/lasers/solid-state-lasers/dpss/ot-03-er-glass-laser-transmitters-with-diode-pumping, access date: 01.03.2022.
35. M. V. Bogdanovich, V. V. Kabanov, G. I. Ryabtsev, A. G. Ryabtsev, and Y. V. Lebiadoc, “High-performance LD-pumped solid-state lasers for range finding and spectroscopy,” Proc. SPIE 8677, 86770X (2013). [CrossRef]
36. N. S. Sankar, A. Ramalingam, V. L. N. Sridhar, and A. S. Laxmiprasad, “Theory and Engineering Design Aspects of a Mini Diode-Pumped Q-Switch Er: Glass Laser for Space Applications,” AIP Conf. Proc. 1391, 279–281 (2011). [CrossRef]
37. J. X. Zhang, W. Wang, Z. B. Li, H. F. Ye, R. Y. Huang, Z. P. Hou, H. Zeng, H. X. Zhe, C. Liu, X. Y. Yang, and Y. L. Shi, “Development of a High Performance 1280×1024 InGaAs SWIR FPA Detector at Room Temperature,” Front. Phys. 9, 678192 (2021). [CrossRef]
38. X. Zhu, A. DesLauriers, C. Bell, L. Gagnon, and M. Guibert, “A wide angle bistatic scanning lidar for navigation,” Proc. SPIE 8379, 83790V (2012). [CrossRef]
39. B. L. Stann, J. F. Dammann, M. M. Giza, P. S. Jian, W. B. Lawler, H. M. Nguyen, and L. C. Sadler, “MEMS-scanned ladar sensor for small ground robots,” Proc. SPIE 7684, 76841E (2010). [CrossRef]
40. H. Tsuji, M. Imaki, N. Kotake, A. Hirai, M. Nakaji, and S. Kameyama, “Range imaging pulsed laser sensor with two-dimensional scanning of transmitted beam and scanless receiver using high-aspect avalanche photodiode array for eye-safe wavelength,” Opt. Eng. 56(3), 031216 (2016). [CrossRef]
41. P. Kaczmarek, D. Stachowiak, and K. M. Abramski, “40W All All-Fiber Er/Yb MOPA System Using Self-Fabricated High-Power Passive Fiber Components,” Appl. Sci. 8(6), 869 (2018). [CrossRef]
42. E. Bayaki, D. S. Michalopoulos, and R. Schober, “EDFA-Based All-Optical Relaying in Free-Space Optical Systems,” IEEE Transactions on Communications 60(12), 3797–3807 (2012). [CrossRef]
43. M. Michalska and J. Swiderski, “Highly efficient, kW peak power, 1.55 um all-fiber MOPA system with a dffraction-limited laser output beam,” Appl. Phys. B 117(3), 841–846 (2014). [CrossRef]
44. B. Mheen, J. Shim, M. Oh, J. song, M. Song, G. Choi, H. Seo, and Y. Kwon, “High-resolution three-dimensional laser radar with static unitary detector,” Electron. Lett. 50(4), 313–315 (2014). [CrossRef]
45. B. Mheen, M. Han, G. Choi, H. Seo, and Y. Baek, “Reconfigurable Low-SWaP Eye-safe LiDAR based on Static Unitary Detector (STUD) Technology,” Conference on Lasers and Electro-Optics, JTu2G.4 (2020).
46. M. Han, G. Choi, H. Seo, J. Choi, and B. Mheen, “Noise Characteristic Analysis Technique for TDC-based LIDAR using High-resolution Eye-safe Low-SWaP LIDAR,” Laser Congress 2019 (ASSl, LAC, LS&C), JM5A.46 (2019).
47. M. Han, G. Choi, M. Song, H. Seo, and B. Mheen, “Analysis and Enhancement of 3D Shape Accuracy in a Single-shot LIDAR Sensor,” Proc. SPIE 10089, 100890R (2017). [CrossRef]
48. G. Choi, M. Han, H. Seo, and B. Mheen, “Analysis of the SNR and sensing ability of different sensor types in a LIDAR system,” Proc. SPIE 10427, 72 (2017). [CrossRef]
49. J. H. L. Pang and D. Y. R. Chong, “Flip Chip on Board Solder Joint Reliability Analysis Using 2-D and 3-D FEA Models,” IEEE Trans. Adv. Packag. 24(2), 499–506 (2001). [CrossRef]
50. A. D. Guzman-Chavez, Y. O. Barmenkov, and A. V. Kir’yanov, “Spectral dependence of the excited-state absorption of erbium in silica fiber within the 1.48 - 1.59 um range,” Appl. Phys. Lett. 92(19), 191111 (2008). [CrossRef]
51. J. Jilsson, P. Scheer, and B. Jaskorzynska, “Modeling and Optimization of Short Yb3+-Sensitized Er3+-Doped Fiber Amplifiers,” IEEE Photonics Technol. Lett. 6(3), 383–385 (1994). [CrossRef]
52. D. J. Richardson, J. Nilsson, and W. A. Clarkson, “High power fiber lasers: current status and future perspectives,” J. Opt. Soc. Am. B 27(11), B63–B92 (2010). [CrossRef]
53. Thorlabs, Inc., TC25APC-1500, https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=5124&pn=TC25APC-1550, access date: 01.03.2022.
54. C. H. Seager, “The determination of grain-boundary recombination rates by scanned spot excitation methods,” J. Appl. Phys. 53(8), 5968–5971 (1982). [CrossRef]
55. L. A. Bitzer, M. Meseth, N. Benson, and R. Schmechel, “A new adaptive light beam focusing principle for scanning light stimulation systems,” Rev. Sci. Instrum. 84(2), 023707 (2013). [CrossRef]
56. J. S. Wu, D. Y. Lin, Y. G. Li, G. P. Hsu, M. C. Kao, and H. Z. Chen, “Optical Characterization and photovoltaic Performance Evaluation of GaAs p-i-n Solar Cells with Various Grid Spacings,” Crystals 9(3), 170 (2019). [CrossRef]
