時-空碼相關議題之研究

Abstract

在無線通訊環境中，由於建築物的遮蔽或人車的移動，往往會造成傳輸訊號在振福上的衰減，以及相位上的改變。而此種不良的效應，會造成接收端在估測接收訊號上很大的困難性。因此，接收端通常需要藉著「分集」 (diversity) 的技術，來增加接收訊號的正確估計。由於無線通訊系統逐漸邁入數據通訊的時代，不在只是聲音上的傳輸，因此更需要高頻寬效率、低功率消耗的設計，過去各種的分集技術在實現上，有其困難性。 在1993年，Wittneben提出一種延遲分集 (delay diversity)的架構；而在1998年，Alamouti和Tarokh將之加以改良後，分別提出一種時間—空間方塊 (space-time block) 及籬柵 (trellis) 架構的編碼調變方式。而這兩種不同形式的編碼調變方式，都具有「完全的分集增益」 (full diversity gain)。 本篇論文，主要是延續Tarokh的時間—空間籬柵編碼調變 (space-time trellis coded modulation) 的架構來加以探討。在第一部份，我們利用串接碼連接交錯排列器 (interleaver) 的形式，結合時間—空間編碼調變及反覆解碼 (iterative decoding) 方式，來增加系統的編碼增益 (coding gain)。另外，為了有效提高交錯排列增益 (interleaving gain) ，我們提出另一種遞迴 (recursive) 形式的時間—空間籬柵編碼調變方式來替代原來無遞迴 (nonrecursive) 形式，作為串接碼系統中的核心碼 (inner code) 。 第二部分，主要討論時間—空間編碼如何有效結合連續相位調變 (continuous phase modulation) 系統。在此章節，我們將提出設計上的原則，並且討論在不同調變參數 (modulation index) 、記憶長度 (memory length)、及頻率脈波整型濾波器 (frequency pulse shaping filter)下對系統錯誤率表現的影響。 第三部份，我們考慮在緩慢及非頻率選擇性的複衰減通道 (slow and frequency nonselective complex fading channel)，如何結合解碼器 (decoder) 和通道等化器 (channel equalizer) 來補償通道造成的intersymbol interference.(交互位元干擾)。而在通道估計 (channel estimation) 方面，我們利用兩種方式來達成: 第一，使用偽亂碼序列 (PN sequence) 作為引導位元 (pilot symbol) 加以估計通道資訊；第二，使用標準化最小平均平方誤差法 (NLMS algorithm) 來估計通道。由於，此兩種方式需要引導位元或是訓練序列 (training sequences)，會降低頻寬效率及增加功率消耗，因此，最後我們將嘗試利用決策性迴授 (decision feedback) 的方式，結合上維特比 (Viterbi) 或是最大後驗機率 (MAP) 解碼器作為系統的解碼方式。而最後我們將利用數值模擬，加以驗證及分析討論我們提出的各種架構。

Space-time (ST) codes introduce inter-dependence of transmitted coded symbols in both time and spatial domains to attain high coding gain while taking full advantage of the transmit diversity. Tarokh's space-time trellis coded (STTC) structure, assuming a flat fading environment and perfect channel state information on the part of the receiver, enjoys many advantages like the high bandwidth efficiency, full diversity gain, and easy hardware implementation, their coding advantages are not especially impressive. This thesis proposes two new ST coding schemes with improved coding gains and suggests a receiver structure that takes the selective multipath fading into account. We first examine a class of serial-concatenated ST coded schemes. The proposed serially-concatenated ST code consists of a nonrecursive convolutional code as the outer code and a recursive space-time trellis codes (RSTTC) as the inner codes with an interleaver placed in between them. Such an arrangement is based on the fact that a serially-concatenated turbo code should a nonrecursive outer code and a recursive inner code. The second ST coding scheme differs from earlier ST schemes in that continuous phase modulation (CPM) is used in conjunction with ST codes, resulting a continuous phase space-time coded modulation (CPSTC). The constant envelop property of the CPM waveforms can not only ease the linearity requirement of the transmitter power amplifier but also provide extra coding gain. The associated performance demonstrate that 3-4 dB gain is achievable. If we add an interleaver between the delay diversity encoder and the CPM encoder then an extra 4 dB decoding gain can be obtained by applying iterated decoding. Finally, we consider the effect of channel frequency selectivity on the receiver design. Although it has been shown that the same design criteria are still valid if the multipath fading channel can be modeled as an FIR filter, the optimal receiver structure is different from that for flat fading channels. As has been shown before, channel memory can be advantageous for a coded system. This is because one can model the channel effect as a nonbinary convolutional encoder so that the receiver can perform iterated decoding on such a concatenated coded system. Therefore, we place an interleaver and then a rate-1 recursive encoder between the output of a ST encoder and the modulator so that the resulting effect on the received waveform is equivalent to a concatenated coding system with a recursive inner code. We evaluate the associated system performance under three different conditions: (i) perfect channel state information is available, (ii) pilot-assisted channel estimate is used, (iii) only blind decision-feedback channel estimate is available. The appendix examines two typical pilot-assisted channel estimation methods, namely, one that uses a PN sequence and those that employ an adaptive method. When a PN pilot sequence is used, correlation seems to be the most effective non-adaptive approach for channel sounding while if adaptive approach is desirable the normalized least mean squared (NLMS) algorithm is an attractive alternative.