異地多輸入多輸出（Multi-site MIMO）雷達可透過空間分集與多視角觀測提升偵測與定位潛力，但其系統能否發揮效益，關鍵取決於分散站點間是否能建立高精度且可長時間維持的時間與頻率同步。為解決 GPS 1PPS 具備絕對時間基準但存在短期抖動、以及各站點本地振盪器長時間累積漂移等問題，本論文提出並實作一套「GPS 1PPS + 銣原子鐘」的混合式同步架構，並以有限狀態機（FSM）導入容差式校準機制，使系統在多數時間維持本地穩定時脈域運行，僅在相位偏差超過門檻時進行校正，以兼顧短期穩定性與長期準確性。本系統以 Xilinx ZCU111 RFSoC 平台為核心，整合基頻即時處理與收發控制，並將 GPS 1PPS 經由邊緣化與同步化電路轉換為 FPGA 內部觸發事件，進一步生成對齊 PRI 的關鍵控制訊號以驅動發射閘控與接收擷取窗。在戶外 2T2R 多站點實證中，所提同步機制透過有限狀態機設定 28 ns 為漂移之容許門檻 (Maximum Drift Threshold)，確保系統在長時間運作下，站間時序偏移被嚴格限制在此邊界之內。此機制有效抑制了發散風險，並結合後端校正算法，成功支撐異地同調收發與後續資料融合流程，並完成對遠距離無人機目標之偵測與定位驗證。;Multi-site Multiple-Input Multiple-Output (MIMO) radar can significantly enhance detection and localization by exploiting spatial diversity and multi-aspect observations. However, practical deployment is fundamentally limited by the ability to achieve and sustain high-precision time and frequency synchronization across geographically separated stations. This thesis presents and implements a hybrid synchronization architecture that combines a GPS one-pulse-per-second (1PPS) signal as an absolute time reference with a rubidium atomic clock as a highly stable frequency reference. To balance GPS short-term jitter and long-term oscillator drift, an FSM-based, tolerance-driven calibration strategy is developed: the radar primarily operates in a locally stable clock domain and performs phase realignment only when the monitored offset exceeds a predefined threshold. The prototype is realized on a Xilinx ZCU111 RFSoC platform, integrating real-time baseband processing and transceiver control. The external 1PPS input is conditioned and converted into an internal trigger event, which is then used to generate PRI-aligned control signals for precise transmit gating and repeatable receive window capture. Outdoor 2T2R field experiments validate that the proposed scheme maintains inter-site synchronization within 28 ns, enabling coherent multisite operation and supporting subsequent data fusion. The system further demonstrates successful detection and localization of a long-range UAV target, confirming the feasibility of the proposed synchronization-enabled distributed sensing framework.
