在高速傳輸時代中，隨著頻率的提升，保持信號完整性變得越來越重要。然而，從印刷電路板中獲取待測物響應是具有挑戰性的。因此，需要進行校正演算法來獲得純粹的待測物響應。許多校正演算法都要求夾具之間的匹配，但在印刷電路板中實現這一點非常困難。話雖如此，商用軟體中的2x-thru校正演算法已經實現了解決夾具不匹配的功能。 2x-thru是近幾年才開始盛行，且很少有研究從參考阻抗的角度分析2x-thru校正演算法，這些問題都是值得探討且不能被忽視的。例如，校正後的待測物參考組抗是什麼？如何決定？如何預估參考阻抗？本論文首先通過自行實現的Python校正演算法，驗證了校正兩種實際量測的PCB和模擬數據的精確度。其次，我們在驗證了自行實現的去嵌入演算法後，討論了兩個實際應用中常遇到的問題。第一個問題是先前提到的夾具不匹配修正，在PCB中，待測物與校正套件之間的不匹配特別常見，這是所有校正演算法面臨的主要挑戰。商用軟體都有相應功能，但並未公開演算法細節。因此，我們提出了解決方法，並且校正結果與商用軟體相當。第二個問題是金屬表面粗糙度，隨著頻率上升，電流更加集中在表面，金屬表面粗糙度也變得更加重要。但在實際應用中，獲取粗糙度得過程是非常繁瑣的。因此，我們將文獻中提出的估計粗糙度的流程應用於實際量測數據，並在過程中發現了一些值得進一步研究的問題。 總之，我們從參考阻抗的角度入手，剖析了幾個經常被忽視但非常重要的問題，特別是其對校正演算法的影響。在論文的結論部分，我們提及了工作的不足之處，並提出了未來研究和改進的潛在方向。 ;Various calibration algorithms exist for de-embedding, with the 2x-thru algorithm being widely used in PCB transmission line applications due to its simplicity. However, the issue of reference impedance alignment with this algorithm has not been extensively studied. This thesis focuses on this overlooked problem, validating our self-implemented de-embedding algorithms through real measurements and simulations. We also address two other important practical aspects: impedance correction for 2x-thru de-embedding and surface roughness of PCB transmission lines. Impedance correction involves aligning the reference impedance of the 2x-thru (Cal-Kit) with the fixture of the device under test (DUT). We propose a solution to address this challenge, achieving correction results comparable to commercial de-embedding tools. Regarding surface roughness, a critical factor in PCB channel characterization, we estimate its scale using S-parameters after 2x-thru calibration. Accurate calibration is essential for robust roughness characterization. During this estimation process, we identify and discuss issues related to the variability of roughness across different sides of the copper surface in a stripline, incorporating this variability into our estimation process for broader applicability. To summarize, we begin by examining the reference impedance and subsequently address several overlooked yet significant issues within the calibration algorithm. In the conclusion of this thesis, we acknowledge the limitations of our work and propose potential directions for future research and improvement.
