正交分頻多工系統中數位射頻處理之直流偏差和IQ不平衡的聯合盲式估測與補償/等化

Abstract

隨著寬頻無線通訊的發展蓬勃，更高的傳輸速率與更寬的通道頻寬的通訊標準被訂定來支援更好的服務；也越來越多的無線通訊應用標準被訂定來支援更多的服務。然而，這些標準的嚴苛需求以及不同標準的支援需求，加劇了收發機硬體的設計的困難，需要更多的元件更複雜的電路，此舉不僅增加了電路面積、消耗功率，也增加了成本。因此，單晶片設計的概念被提出來解決複雜電路設計以及不同製程間相容性的問題，進一步數位射頻的概念被提出。 數位射頻實現射頻功能於數位邏輯閘電路，徹底減少了類比電路設計的非理想性和噪音干擾，然而，實現射頻於數位電路須要先進的數位互補式金屬-氧化層-半導體(CMOS)製程來支援高頻的設計，除此之外，數位射頻降頻所需的高取樣率高功率類比數位轉換器，仍然難以實現在目前的製程和技術中。因此，數位中頻降頻的概念被提出來實現數位射頻的概念，像是德州儀器(TI)提出的數位射頻處理收發機(DRP)，然而，他所造成的代價包括高取樣率、高功率的類比數位轉換器以及額外多出複雜的混合式信號校正電路。所以，在傳統架構下，使用系統觀點來運用數位訊號處理(DSP)演算法來處理射頻損害以及設計的問題，不僅可以減少類比電路設計的複雜度；也可減少額外類比元件電路的使用。 因此，在本論文中，我們考慮一個直接轉換收發機為基礎的數位射頻架構，描述套討在此架構中對於正交分頻多工系統影響較嚴重的直流偏差和IQ不平衡兩個問題，以系統的觀點來推導此兩損害一起造成的模型，並推導出基頻對等的不連續信號模型。然後提出一個以數位訊號處理為基礎的盲式估計和補償/等化演算法，並分析証明盲式估計所利用的正交分頻多工系統時域信號的二階統計特性。模擬的結果驗證此演算法的有效性，並顯現了在不同調變和不同通道環境的強健性，以及在系統符元錯誤率有好的表現，證明了此一系統的可行性。最後總結並點了參數估計可能遇到的問題以及其他架構問題也套討了數位射頻架構實現或演算實現所遇到的問題。

Radio-frequency system-on-chip (RF-SoCs), which incorporate the RF, analog and digital circuit into the same chip, are believed to be the next generation of RF system design technology in order to realize low cost, low power dissipation and small form factor transceiver solution. Since with the broadband wireless communication progressing to support ambitious data rates, this imposes serious challenges on the underlying hardware of the transceiver and inevitably increases the circuit components and area, in particular to the radio frequency (RF) components. Furthermore, the RF-SoCs not only facilitate the software defined radio (SDR) and cognitive radio (CR) fulfillment, but provide a promising way to multi-standard or multi-mode capabilities of the transceiver. To realize the RF-SoCs, several Digital RF paradigms are being proposed and we classify them into two categories: (i) Digital-RF technology, which utilize the advantage of the process scaling to implement the RF circuit functionalities in digital domain, (ii) Digital-centric CMOS technology, which use the digital signal processing (DSP) technique to calibrate or compensate the RF components. However, the first digital-RF method taking advantage of the process scaling also add the derivative circuit design issues, even further the cost of the advanced fabrication process exponentially increase with the process scaling. Thus, the third digital-centric method is more attractive and more promising from system design perspective by shifting some analog functionality to be addressed in the digital domain instead of changing the process technology and adding extra analog form. Hence, in this thesis, the digital-centric scenario are discussed and addressed, and we focus on the more challenging receiver side of the transceiver. The objective of this thesis is to identify the effects introduced due to RF/analog components imperfection and introduce a system that reduce or remove those effects. Thus, a novel frame work for the RF components impairments during multicarrier communications are identified and summarized and then the accurate problem formulations are derived. And the performance degradation resulting from those issues are simulated. From the performance degradation owning to those effects, the most serious problems: DC offsets and I/Q mismatch are addressed. In order to provide short term digital front end signal processing solution for this system, a novel complete approach for the digital estimation and compensation of the DC offset and I/Q imbalance are proposed. In our view, the blind, i.e. nonreference-based, character is one of the key benefits of the novel algorithm. By merely observing the statistics of the received, corrupted samples, the algorithm is capable to estimate the unknown parameters of the DC offset and I/Q imbalance. Thus, the algorithm is independent of any specialized training sequence or calibration signals, such as standard-specific pilots or receiver generated test signals. Implied by the deliberate abdication on any reference symbols, an inherent robustness to large scale DC offsets and I/Q imbalance are observed.