IEEE 802.3an標準要求10GBase-T系統的傳輸速度為每秒10十億位元(Gbps)且位元錯誤率需低於10^(-12)。在此要求下，決策點信噪比至少要23.4dB。本論文針對10GBase-T系統中的回音消除濾波器之抽頭(tap)數之選擇進行探討，我們提出了一個演算法，讓系統在傳送訓練資料的階段，可以依據實際等效回音通道長度適性調整其抽頭數量，以避免欠模擬(under-modeling)或過模擬(over-modeling)的情況。
目前已存許多之可適性最小均方演算法用來進行未知系統長度進行估測，但其多為用通道與濾波器差值作為回授，調整濾波器長度調適演算法中的洩漏因子(leaky factor)、步階值等參數，然而這些既有的方式無法直接應用在10GBase-T系統中的回音消除濾波器之長度預測。本論文在既存的可適性最小均方演算法上用信噪比做二次回授的方法進行改良。此二次回授之方法，適用於對輸出信噪比有較嚴苛要求而通道單方向後半段較小之環境。在此篇論文裡，10GBase-T之回音通道(長度約450)模擬，原先既存在各方法只能感應到通道係數值較大(長度100)之部分，改良後二次回授則可以自動調整到接近實際長度值(400~ 450)。

IEEE 802.3an requires the data transmission rate is 10 Gbps and the bit error rate is less than 〖10〗^(-12). This necessitates the signal-to-noise ratio (SNR) at the decision point should be higher than 23.4 dB. This thesis studies the selection of the tap length for the Echo canceller in the 10Gbase-T systems. We propose an adaptation algorithm to estimate a suitable tap length for Echo cancellers to avoid the under-modeling or over-modeling.
Some related works adopted the least mean squared (LMS) based algorithms to estimate the tap length of the unknown systems. Most of these approaches use error signals as a feedback signal to adapt the leaky factor or step-size for the tap-length recursions. However, they cannot directly apply to the tap-length estimation of Echo cancellers in the 10GBase-T systems. We modify the LMS-based algorithm and adopt the value of SNR as a second feedback signal such that a near-optimal tap length for Echo cancellers can be estimated, which would be useful especially in a monodirectional decrescent channel.
We use computer simulation to validate the effectiveness of our proposed method. In our simulation, the effective tap length for the Echo channels is approximately 450. The conventional approaches report the optimal tap length of Echo cancellers is about 100, which is severely under-modeling. On the other hand, our proposed method is able to correctly estimate the required tap length for Echo cancellers should be arranged from 400 to 450.

第一章、 緒論…………………………………………… 1
1.1前言……………………………………… 1
1.2動機與目標……………………………… 1
1.3論文架構………………………………… 2
第二章、 基頻通道模型與10GBase-T系統概觀……… 3
2.1通道之非理想效應……………………… 3
2.2通道模型………………………………… 7
2.3 10GBase-T系統目標與要求…………… 11
2.4 10GBase-T系統傳送端與接收端……… 11
2.4.1 傳送端……………………………… 11
2.4.2 接收端……………………………… 14
第三章、 既有可變濾波器長度最小均方演算法與提出改良方法……………………………………………… 16
3.1 最小均方演算法………………………… 16
3.2 梯度下降可變濾波器長度最小均方演算法17
3.3 非整數可變濾波器長度最小均方演算法及
延伸…………………………………………… 20
3.3.1非整數可變濾波器長度最小均方演算法20
3.3.2可適性參數之可變長度濾波器…… 23
3.3.3可適性步階值之可變長度濾波器… 26
3.3.4簡化複雜度變數長度濾波器……… 30
3.3.5二次回授可變長度濾波器………… 32
第四章、 10GBase-T系統處理流程與回音消除演算法置入模擬結果………………………………………… 34
4.1 10GBase-T收發器啟始流程…………… 35
4.1.1 起始流程…………………………… 35
4.1.2等化及串擾消除調整……………… 36
4.2模擬流程………………………………… 39
4.3 可變濾波器長度LMS模擬……………… 43
第五章、 結論與未來展望……………………………… 67

參考文獻………………………………………………… 68

[1]李展華,”經銅線傳輸之百億位元乙太網路系統經片架構設計,”國立台灣大學電信工程學研究所碩士論文,July ,2006.

[2]S. Powell, B. Shen, and G. Ungerboeck,“10GBASE-T modulation &co-ding, set of fixed precoders, and start-up,”IEEE P802.3an Task Force,Nov. 2004.

[3]S. Powell and B. Shen,“Specification and performance of proposedLDPC (2048,1723) code,”IEEE P802.3an Task Force, Jan. 2005.

[4]G. Ungerboeck,“10GBASE-T coding and modulation : 128-DSQ + LDPC,” IEEE P802.3an Task Force, Pp.9-10 2004.

[5]Z. Pritzker and A. Feuer,”Variable Length Stochastic GradientAlgorithm,”IEEE Trans. Signal Processing, Vol.39, Pp.997-1001,Apr. 1991.

[6]Yuantao Gu, Kun Tang, and Huijuan Cui,”LMS Algorithm With GradientDescent Filter Length,”IEEE Signal Processing Letters, Vol.11, No.3,March 2004.

[7] Yu Gong and Colin F. N. Cowan,”An LMS Style Variable Tap-LengthAlgorithm for Structure Adaption,”IEEE Transactions on Signal Proce-ssing, Vol.53, No.7, July 2005.

[8] Yu Gong and Colin F. N. Cowan,”A Novel Variable Tap-LengthAlgorithm for linear adaptive filters,” in Proc. IEEE Int. Conf. Acoust, Speech Signal Process, Montreal, QC, Canada, May 2004.

[9] ---,”Structure Adaption of Linear MMSE Adaptive Filter,”Proc.Inst. Elect. Eng.-Vision, Image, Signal Process., Vol.151, No.4,Pp.271-277, Aug. 2004.

[10]Dingjie Xu, Bo Yin, Wei Wang,”Variable Tap-Length LMS AlgorithmBased on Adaptive Parameters for TDL Structure Adaption,”IEEE SignalProcessing Letters, Vol.21, No.7, July 2014.

[11]Ying Wei, Zhibin Yan,”Variable Tap-Length LMS Algorithm withVariable Tap-Length Adaption Step Size,”Chinese Control and DecisionConference 28th, Aug 2016.

[12] Ying Wei, Zhibin Yan,”Simple Variable Tap-Length Algorithm forHigh-Noise Environment,”Electronics Letters Vol.53, No.5, Pp.320-322, 2nd March 2017.

[13] K.Seki,“Startup protocol for 10GBASE-T,”IEEE P802.3an TaskForce, Nov 2004.

[14] G.Ungerboeck,“10GBase-T TxFE solutions,dpSNR vs length ofprecoding response, and PMA training,”IEEE P802.3an Task Force, May 2005.