Title: 非線性磁性記錄通道之第四類部份響應等化器研究
Class IV Partial-Response Equalization for Nonlinear Magnetic Recording Channels
Authors: 黃合淇
Ho-Chi Hwang
魏哲和
Che-Ho Wei
電子研究所
Keywords: 非線性磁性記錄通道; 部份響應等化; 適應性; 非線性濾波器;magnetic recording channels;partial-response equalization; daptive;nonlinear filters
Issue Date: 1992
Abstract: 應用第四類部份響應等化技術於線性磁性記錄通道﹐已經可以成功地提高 記錄密度了.然而﹐為了達到高記錄密度, 通道中的線性與非線性的符際 干擾(ISI) 變得相當嚴重﹐並且超過傳統等化器的能立範圍.因此在本論 文中﹐提出兩種類型的等化器來克服線性與非線性的符際干擾所造成的問 題.第一種型式是一個非適應性.類比的第四類部份響應 Minimax 等化 器;第二種型式是一個非線性數位適應性第四類部份響應等化器﹐包含 RAM-DFE 與CPWL-DFE.第四類部份響應 Minimax等化器是適應性等化器的 次佳替代者﹐如果考慮磁性記錄通道的參數變化情形﹐它在均方效益測量 是最佳地 robust.RAM-DFE 是用一個隨機存取記憶體(RAM) 取代傳統判 別式迴授等化器(DFE) 的迴授部份﹐它把通道上的脈衝反應塑造成想要的 部份響應信號.在此提出一些改進 RAM-DFE 收斂速度的方法. CPWL-DFE 使用一個分段式線性濾波器取代RAM-DFE中的隨機存取記憶體部份. 因此 有較快的收斂速度﹐但是在訓練過程中﹐ 它有收斂到局部極小值(local minimum)的缺點.另外一種具有不同架構且次佳結果的CPWL-DFE被提出討 論.透過電腦模擬結果﹐我們解釋與模擬出各種等化器之比較. Application of the class IV partial-response (PR_IV) equalization to linear magnetic recording channels has been shown successfully further increase the storage densities. However, the linear and nonlinear intersymbol interference (ISI) become severe and exceed the limits of the capabilities of traditional linear equalizers. Consequently, equalizers which are able to combat both the linear and nonlinear ISI nust be considered. In this thesis, two kinds of PR-IV equalizers are presented. The first type is a non-adaptive, analog minimax PR-IV equalizer and the second one is the nonlinear digital adaptive PR-IV equalizer including the RAM-DFE and CPWL-DFE. The minimax PR-IV equalizer is a "next-best" alternative for adaptive equalizer. It is optimally robust relative to a mean- square performance measure which takes account of recording channel parameters variations. The RAM-DFE which has a random access memory replacing the transversal feedback section of the DFE can shape the impulse response of the channel into the desired PR signaling. Some improvements for increasing convergence speed for the RAM-DFE are presented. The CPWL-DFE uses a canonical-piecewise linear(CPWL) filter to replace the RAM and hence has faster convergence speed. The drawback is that it may converge to a local minimum during the training mode. Another CPWL-DFE with different structure and suboptimum solution is proposed. Comparisons between various kinds of equalizers are simulated and interprted.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT810430019
http://hdl.handle.net/11536/56876
Appears in Collections:Thesis