正交分頻多工(Orthogonal Frequency Division Multiplexing,OFDM)因為高效率的頻寬效益以及多路徑通道的穩定傳輸，使得其成為現代無線通訊中不可或缺的技術。然而，此技術本身所擁有的高峰值對均值功率比(Peak-to-Average Power Ratio,PAPR)問題，造成功率放大器之非線性失真，導致調變訊號會有頻譜再生(Spectral regrowth)的現象而干擾鄰近通道的傳輸訊號。考慮到寬頻系統中的記憶性問題，本篇論文研究了功率放大器的新預失真技術。在過去的文獻中，已有許多預失真器模型能夠同時對功率放大器進行記憶非線性化的補償。在本論文的架構中，首先將真實功率放大器的模型表示成一個維納(Weiner)模型通式，預失真器的部分，則是在傳統的漢默斯坦(Hammerstein)模型中加入了基頻的單型規範分段線性(Simplicial Canonical Piecewise Linear, SCPWL)函數，用以模擬系統的非線性。與過去的預失真方法比較，所提出的預失真器能簡易的透過調整SCPWL函數分段區間的數量來達到所需求的補償效能。接著，於間接學習結構中再利用遞迴預測誤差法(Recursive Prediction Error Method,RPEM)來對所提出的預失真器參數做適應性的估測，實現理想的系統性能以及加快收斂的速度，就能達成高度的線性化補償。;It is well known that Orthogonal Frequency Division Multiplexing (OFDM) has become indispensable in modern wireless communications because of high frequency efficiency and high transmission stability in multi-path channel environments. However, OFDM has an inherent characteristic of high Peak-to-Average Power Ratio (PAPR) subject to nonlinear distortion of a power amplifier, leading to the phenomenon of spectral regrowth for modulated signals such that the adjacent communication channels are interfered. Taking into account the memory problem in wideband systems, this thesis studied a new pre-distortion scheme for the power amplifier. From previous researches in the literature, there have been many predistorter’s models considering to compensate for the nonlinearity effect of memory in a power amplifier. In our framework, we first establish the Weiner model for characterizing the power amplifier and then use the Hammerstein model for the predistorter along with the baseband Simplicial Canonical Piecewise Linear (SCPWL) function for nonlinear modeling. Compared with previous predistorting methods, the proposed predistorter is easier to reach the required performance by adjusting the number of segments of the SCPWL function. For the indirect predistortion architecture, we use the Recursive Prediction Error Method (RPEM) for adaptive estimation of the parameters of the proposed predistorter, which can achieve a satisfying system performance and fast convergence speed for better linearization.
