有限脈衝數位濾波器之高階合成

Abstract

為了簡化有限脈衝濾波器設計過程及產生一高效能低複雜度之濾波運算，本論文提出一高階合成輔助器。其中包括了各種濾波器最佳化設計方式，IFIR，FRM，Multirate/Multistage。加上已熟知的快速濾波演算法的應用來進一步降低運算複雜度，並且加以延伸至較為一般性的形式。在有限精確度設計下，差分編碼被應用來抵抗量化誤差，我們並且提出彈性精確度設計，特別針對串接式濾波器架構，可以有效降低硬體複雜度及達成有限位元精確度下之高訊號及量化誤差比。最後，論文檢視了現階段實際應用的濾波器，應用於此高階合成軟體，並提出多項可行的設計分析以加速設計流程。

In this thesis, for the simplification of FIR filter design and complexity reduction of a filtering process, we propose a high-level FIR filter synthesizer. This synthesizer combines many efficient filter design techniques including IFIR, FRM, and Multirate/Multistage techniques. In addition, the existing fast convolution algorithms are extended to more general forms and adopted for further complexity reduction. The differential coding scheme is utilized to reduce the dynamic range of the coefficients against the quantization error. Moreover, we proposed an efficient scheme for fixed-point cascaded-form FIR filter implementation, which is suitable for low-precision applications. Finally, the designs of several modern communication applications are automatically generated and compared with the conventional design approaches. The results show effectiveness of the proposed designs.