寬頻多入多出直接轉換傳收機之射頻劣化效應估算與補償技術

Abstract

在無線通訊傳收機射頻架構設計中，直接轉換架構(direct-conversion radio architecture)是一個具有低成本，低功率消耗和小體積的類比前端設計。然而，此架構卻會產生額外的射頻劣化效應，諸如I-Q失衡(I-Q Imbalance)與直流偏移(dc offset)等。這些劣化效應加上所有射頻架構都會產生的頻率偏移(frequency offset)，若無適當的補償，對通訊系統將造成嚴重的效能損失。此篇論文旨在針對多入多出(multiple input, multiple output, MIMO)直接轉換傳收機之射頻劣化效應估算與補償技術做深入研究。基本而言，射頻劣化效應估算與補償技術可分為兩類：第一類為估算補償技術，另一類則是自我校正技術。估算補償技術是在通訊傳輸中，在接收機端去除接收訊號的射頻劣化效應之技術;而自我校正技術則是在通訊傳輸前去除本身射頻劣化效應之技術。此兩類技術皆在本論文中做深入研究；所探討的射頻劣化效應包括了不隨頻率變動的I-Q失衡(frequency-independent I-Q imbalance)、隨頻率變動的I-Q失衡(frequency-dependent I-Q imbalance)、直流偏移與頻率偏移。

本篇論文在估算補償技術研究方面，探討在多入多出通訊系統下的傳送機與接收機之射頻劣化效應估算與消除方法。首先提出了一個二階段干擾消除架構：在第一階段消除接收端產生之I-Q失衡、頻率偏移與直流偏移，而在第二階段消除傳送端產生之I-Q失衡。此二階段消除架構能適用於各種多入多出之運作模式，包括了空間多工、時空區塊編碼與傳送端波束成形等技術，以及任意的傳送天線數與接收天線數。此外，此架構也概括了各種通訊應用下的消除架構，諸如應用於無線點對點對等(wireless peer-to-peer)通訊、上行鏈路(uplink)與下行鏈路(downlink)行動通訊。接著我們提出多種參數估算方法，第一種為在最小平方準則下最佳之聯合參數估算方法，分析指出此方法為不偏估計器(unbiased estimator)以及在有興趣範圍的訊號雜訊比(SNR)下其效能可達到Cramér-Rao下限(lower bounds)，但也是最高複雜度的。因此，我們設計其它多種低複雜度估測器，包含了特殊角度旋轉週期訓練設計、直流偏移與頻率偏移簡化估測器與藉由週期訓練協助之低複雜度遞迴估算法。電腦模擬顯示低複雜度設計所造成之效能損失幾乎可被忽略，與現今文獻所提之技術做比較，本論文所提之技術能擁有更低錯誤率以及較短之訓練符元長度需求。

在自我校正技術方面，本論文提出一個新的時域方法(time-domain method)，在迴路(feedback loop)不需要專門額外的類比硬體電路下，能夠同時自我校正傳送機與接收機之射頻劣化效應。此時域方法適用於各類的通訊系統並能同時校正不隨頻率變動與隨頻率變動的I-Q失衡和直流偏移。此外，我們亦提出最佳訓練數列化之設計方法，並由分析與模擬驗證了此方法之正確性與有效性。

Direct-conversion radio architecture is a low-cost, low-power, small-size design for the analog front-end of a wireless communication transceiver. Nevertheless, it induces extra radio impair-ments such as I-Q imbalance and dc offset that, along with frequency offset which is commonly encountered in all radio frequency (RF) architectures, incur severe degradation on communica-tion performance if not accurately compensated. This dissertation investigates radio impairments estimation and compensation techniques for wideband MIMO (multiple input, multiple output) direct-conversion transceivers. Basically, there are two types of techniques: one is estima-tion/compensation and the other is self-calibration. The estimation/compensation technique is to remove the impairments from the received signal during communication at the receiving side, while self-calibration is a technique to remove the transceiver’s own radio impairments before communication commences. Both types of techniques are studied in the dissertation with a com-plete set of radio impairments taken into consideration, including frequency-independent and dependent I-Q imbalances, dc offset and frequency offset.

For the estimation/compensation technique, this dissertation investigates the estimation and cancellation of the transmitter and receiver radio impairments in the MIMO communication sys-tems. Firstly, a two-stage cancellation architecture is proposed with the receiver fre-quency-independent and dependent I-Q imbalances, frequency offset, and dc offsets being can-celled in the first stage and the transmitter frequency-independent and dependent I-Q imbalances cancelled in the second stage. The architecture is general to accommodate different forms of MIMO operation including spatial multiplexing, STBC (space-time block coded) and transmit beam forming, with any number of transmit and receive antennas. In addition, it generalizes the cancellation architecture for various types of application configurations such as wireless peer-to-peer communication, downlink and uplink of mobile cellular communications. Secondly, several methods of estimation of radio parameters are proposed. One is the optimum joint esti-mation of all radio parameters based on least squares criterion. It is shown through analysis that the estimator is unbiased and can achieve the Cramér-Rao lower bound (CRLB) for the sig-nal-to-noise ratios (SNRs) of interest. The others are reduced-complexity methods, including the special phase-rotated periodic training design, simplified frequency and dc offset estimators and low-complexity iterative estimation aided by periodic training. Simulation results show that the proposed methods have negligible performance degradation when using the reduced-complexity designs and outperform the existing ones in error-rate performance and/or the number of training symbols required.

For the self-calibration technique, a new time-domain method is proposed to self-calibrate simultaneously the transmitter and receiver impairments without a dedicated analog circuit in the feedback loop. Thanks to the time-domain approach, the method is applicable to all types of sys-tems and is able to calibrate jointly the frequency-independent I-Q imbalance, fre-quency-dependent I-Q imbalance, and dc offset. In addition, training sequence design is investi-gated to optimize the performance of calibration, and analysis and simulations are conducted to confirm the effectiveness of the proposed method.