用於正交分頻多工數位視訊廣播系統之接收多樣與相位雜訊補償技術

Abstract

正交分頻多工為一種具高頻譜效益，並能有效地克服多路徑衰落效應的調變技術。使用正交分頻多工技巧之數位視訊廣播(OFDM-based DVB)系統，具有對抗碼際干擾之強韌性及高頻譜效益，這些利基驅使研發單位與產業界競相投入該種系統之設計與實現。在本論文中，吾人提出一種具有接收多樣與相位雜訊補償能力之OFDM-based DVB接收機，能有效對抗多路徑衰落及相位擾動現象。一般而言，行動通訊系統受到多路徑干擾之時變性衰落通道的影響，易造成系統效能大幅降低，吾人將針對此問題對接收機之各個層級提出接收多樣方法，藉以減緩時變性衰落效應。另一方面，為了能夠充分地改善空時頻無線通道之相位雜訊問題，吾人針對相位雜訊補償，提出「空時頻回授決策法」，利用相鄰載波通道間高相關之特性，選擇最佳資料取樣大小，並將資料以最小平均平方誤差法則計算相對應之載波加權值，以取代原本的通道估測。此種批次處理法可在載波通道變化及相位雜訊補償上取得一平衡點。最後，吾人藉由電腦模擬驗證上述架構在市區無線環境中，具有優異的傳輸表現。

Orthogonal frequency division multiplexing (OFDM) is a high spectral efficiency modulation technique that can successfully alleviate multipath fading effects. Digital video broadcasting (DVB) systems incorporating the OFDM technique, which can promise robustness against intersymbol interference (ISI) and efficiency in spectrum utilization, are attracting researchers and industry attention to its design and implementation. In this thesis, an OFDM-based DVB receiver with diversity reception and phase noise compensation schemes is proposed, which provides diversity gain and phase noise suppression simultaneously. Performance of a mobile communication system is typically significantly degraded due to time-varying fading channels. To mitigate this problem, diversity reception schemes at different stages of the receiver are designed to alleviate the fading effects. On the other hand, to alleviate the phase noise effect over space-time-frequency wireless channels, we propose a space-time-frequency decision feedback algorithm for phase noise compensation, in which the optimum data sample size is selected based on the high correlation between adjacent data tones. Batches of the data samples are then processed with the minimum mean square error (MMSE) criterion to obtain the optimum weights at the corresponding data tones. The batch processing can achieve a good balance between channel variation and phase noise compensation. Finally, we evaluate the performance of the proposed system and confirm that it works well in a typical urban environment.