Linearized SIMO Models and Associated Multichannel Receivers for GMSK Signals

• Main Authors: Yi-Ming Tseng, 曾一鳴
• Other Authors: Jeng-Kuang Hwang
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/39766107001807371980

碩士 === 元智大學 === 電機與資訊工程研究所 === 87 === As is well known, GMSK is a spectral and power efficient modulation scheme that is adopted by the GSM system. However, the non-linearity of GMSK often brings about difficulties in design and analysis. In view of this, we first base on the linear approximation GMSK (Gaussian-filtered Minimum Shift Keying) signals model proposed by Laurent [1], and utilize it to derive a linearly approximated GMSK signals model for replacing the traditional GMSK model. More specificcally, a linearized SI4O (single input and four outputs) model and associated receivers for single-channel GMSK reception are proposed. The model is called DF-SIMO model since it is resulted from the usage of De-rotation and T/2 Fractionally-spaced sampling of the received signal. Next, We study the resulting multi-channel estimation problem by applying three least squares estimators, namely, (1) batch LS estimator based on training sequence only, (2) structured LS estimator using the known pulse-shaping function, and (3) a semi-blind channel estimator. Finally, With our proposed DF-SIMO model and channel estimators, four optimum multichannel receivers according to MMSE and ML criteria are developed, including linear MMSE (Minimum Mean Squared Error) receiver, non-linear DFE (Decision Feedback Equalizer) receiver, non-linear EDFE (Erasures Decision Feedback Equalizer) receiver, and non-linear MLSE (Maximum Likelihood Sequence Estimation) receiver. Using extensive and realistic computer simulations, the signal-channel model, channel estimators, and T-spaced or T/2-spaced receivers are thoroughly investigated, by hypothesizing various GSM and non-time-variant channel models. It is demonstrated that the DF-SIMO scheme is superior to the traditional T-spaced D-SI20 scheme. And we concluded that among all the receiver structures, the EDFE receiver is recommended since its complexity is not only much less than the MLSE, but also its performance can be very close the MLSE under various time-variant channels. Moreover, in some important case such as typical urban (TU) channels, it is found that the bit error rate (BER) performance of the EDFE receiver is even better than that of the MLSE.