用於功率放大器之非線性特性之特殊測試訊號設計

Abstract

射頻電路設計隨著無線通信的日益增多，成為一個明顯重要的研究課題。射頻元件的特性，例如在這篇論文裡研究的功率放大器，在確認系統性能中，能起重要作用。因此，評估功率放大器的特性，無論是透過測量還是模擬，都能提供對設計過程有價值的訊息，而這篇論文，報告我的對功率放大器的模型化和模擬方法的研究。首先，我集中於非線性和線性記憶的模型化。適當的模型化，能為我們提供功率放大器的非理想特性的洞察力。實際上，模型參數可以被萃取，那靠一些測試訊號信號，例如單音調，雙音調和多重音調的信號。在這篇論文的第二個部分，我透過模擬來取代實際電路的量測，集中研究正交分頻多工系統。在這樣的系統裡，非理想的效應透過錯誤向量振幅的測量。如果用隨機的正交分頻多工信號作模擬，要錯誤向量振幅的結果將非常費時。為加速模擬時間，我研究一種特別設計正交分頻多工信號。實驗顯示，使用被設計的信號能大量降低模擬時間，並且與使用隨機正交分頻多工信號相比，可產生粗略的錯誤向量振幅值。

RF circuit design has become a tremendously important research area along with the growing popularity and importance of wireless communications. The linearity of RF devices, such as the power amplifier studied in this thesis, plays a significant role in determining the overall system performance. Therefore, evaluating power amplifiers’ behavior, whether by measuring or simulating with the design, can provide valuable information for the design process. This thesis reports my research on the modeling and simulation of power amplifiers. First, I studied the modeling issues with the focus on nonlinearity and linear memory. A proper modeling can provide us the insights of the causes and effects of the non-ideal behaviors of power amplifiers. In practice, the modeling parameters can be extracted by measuring the outputs with selected excitation signals, such as single-tone, two-tone and multi-tone signals. In the second part of this thesis, I try to evaluate the non-ideal effects by simulations instead of physically measuring a circuit. I focus my effort for Orthogonal Frequency-Division Multiplexing (OFDM) systems. In such systems, non-ideal effects are summarized by measuring the Error Vector Magnitude (EVM). It is very time-consuming if the simulation is conducted by inputting pseudo-random OFDM signals and measuring the EVM output. To speed up the simulation time, I developed a specially designed OFDM signal which satisfies certain spectral and statistic properties.. Experiments show that using the designed signal can drastically reduce simulation time and generate comparable EVM values as compared to using pseudo-random OFDM signals.