| 摘要: | 近年來，光達（Light Detection and Ranging, LiDAR）技術在自駕車、智慧製造與機器人等領域中扮演關鍵角色。其中，傳統的飛行時間式（Time-of-Flight, ToF）LiDAR 主要透過計算雷射脈衝從發射至接收的時間差以進行距離量測。雖具有實用性，惟其對環境光極為敏感，且無法直接擷取目標物之速度資訊，限制其於高精度動態場景中的應用。
相較之下，調頻連續波（Frequency-Modulated Continuous-Wave, FMCW）LiDAR 系統則透過發射與回波之間的頻率差，同步量測目標的距離與速度。此技術將距離資訊編碼於拍頻訊號之頻率成分中，透過頻域分析可準確解析出目標深度，提升空間解析度並有效抑制雜訊干擾。同時，FMCW LiDAR 可藉由都卜勒效應（Doppler Effect）擷取目標速度資訊，實現即時 4D 成像（即三維空間資訊加上速度維度），已成為次世代光達系統之主流技術。
然而，FMCW LiDAR 系統仍面臨兩大挑戰：雷射線寬過寬與波長掃描線性不足。雷射線寬決定其頻率穩定性與同調性，進而影響拍頻訊號之頻譜解析與速度偵測靈敏度。窄線寬雷射能產生穩定、尖銳之拍頻，有助於提升都卜勒頻移解析能力與整體訊號雜訊比（Signal-to-Noise Ratio, SNR）；反之，過寬的線寬將導致頻譜模糊與量測誤差，降低對低速目標的偵測能力。
為解決此問題，本研究首先以一款商用 DFB 雷射（PL-DFB-C-A81-W1550-PA，原始線寬約 2.2 MHz）進行自注入鎖定實驗，成功將雷射線寬壓縮至 38.9 kHz，達到約 56.6 倍的壓縮效果，並實現高達 40.5 dB 的邊模抑制比（SMSR）。儘管此方法可行，但由於其對操作條件敏感、穩定性不足，不利於系統整合與長時間運行。
因此，本研究提出一套基於延伸腔分布式布拉格反射（Extended-Cavity Distributed Bragg Reflector, EC-DBR）雷射的 4D FMCW LiDAR 系統。該雷射採用 InP generic foundry 製程平台，實現單晶圓整合放大器、相位調變器、DBR 光柵與波導等元件，並具備約 25 kHz 的窄線寬，有效提升拍頻訊號穩定性與速度解析度。系統利用配置於外腔之相位調變器進行反向偏壓驅動，於不需預失真補償的情況下實現高線性掃頻，避免因功率與折射率同時變化所引發之線寬增強效應（linewidth enhancement effect）與相位雜訊，並可維持穩定的雷射輸出。
實驗系統採用同頻偵測（homodyne detection）架構，將拍頻訊號維持於低頻基頻範圍，有效抑制 flicker noise。透過提升調變頻率與頻率濾波處理，系統可穩定解析速度低至 10 mm/s 的慢速目標。同時，結合 MEMS 掃描鏡進行高速二維掃描，展現出優異的影像解析度與動態量測靈敏度。
綜合上述成果，本研究展示了一套不依賴外部高頻調變器與複雜偏移架構之 FMCW LiDAR 系統，僅以高線性掃頻與窄線寬雷射源，即可實現 4D 成像功能。該設計具備低功耗、低成本與高度整合潛力，未來可廣泛應用於車載雷達、無人載具導航、精密工業量測等高階應用場域。
;In recent years, Light Detection and Ranging (LiDAR) technology has played a pivotal role in autonomous vehicles, smart manufacturing, and robotics. Traditional Time-of-Flight (ToF) LiDAR estimates target distance by measuring the time delay between transmitted and reflected laser pulses. While practical, ToF systems are highly susceptible to ambient light interference and cannot directly acquire target velocity, limiting their application in high-precision dynamic environments.
In contrast, Frequency-Modulated Continuous-Wave (FMCW) LiDAR systems measure both distance and velocity simultaneously by analyzing the frequency difference between the emitted and received signals. This technique encodes distance information into the frequency domain of the beat note signal, allowing depth retrieval via spectral analysis. FMCW LiDAR also enables velocity extraction through the Doppler effect, supporting real-time 4D imaging—three-dimensional spatial mapping with an additional velocity dimension—making it a promising next-generation LiDAR solution.
However, FMCW LiDAR systems face two critical challenges: broad laser linewidths and nonlinearity in wavelength tuning. Laser linewidth significantly affects frequency stability and coherence, thereby influencing beat note sharpness, Doppler resolution, and signal-to-noise ratio (SNR). A narrow linewidth results in stable and sharp beat signals, improving velocity detection sensitivity. Conversely, broad linewidths lead to spectral blurring and reduced measurement accuracy, especially for slow-moving targets.
To address these issues, this study first demonstrates linewidth compression of a commercial DFB laser (PL-DFB-C-A81-W1550-PA, ~2.2 MHz) via self-injection locking, achieving a linewidth as narrow as 38.9 kHz—a 56.6× improvement—with a side-mode suppression ratio (SMSR) of 40.5 dB. However, this method is highly sensitive to operating conditions, posing challenges to system integration and long-term stability.
This work therefore proposes a 4D FMCW LiDAR system based on an extended-cavity distributed Bragg reflector (EC-DBR) laser, fabricated using an InP generic foundry platform. The monolithically integrated chip includes semiconductor optical amplifiers, phase modulators, DBR gratings, and waveguides, achieving a linewidth of ~25 kHz. By applying a reverse-bias voltage to the external phase modulator, a highly linear frequency sweep is achieved without pre-distortion, while avoiding linewidth enhancement effects and maintaining output stability.
The system employs a homodyne detection architecture to ensure beat notes remain in the low-frequency baseband, effectively suppressing flicker noise. With optimized modulation frequency and spectral filtering, it can resolve slow target velocities down to 10 mm/s. High-resolution 2D scanning is achieved using a MEMS mirror, enabling both spatial precision and velocity sensitivity.
In summary, this study presents a 4D FMCW LiDAR system that eliminates the need for external high-frequency modulators or frequency shifters. Through high-linearity chirped modulation and narrow-linewidth laser sources, the system achieves robust 4D imaging with low power consumption, compact form factor, and strong integration potential. It holds significant promise for applications in automotive radar, autonomous navigation, and precision industrial metrology. |
