| Summary: | 碩士 === 國立清華大學 === 通訊工程研究所 === 103 === Due to the increasing demand for high data-rate indoor wireless communications, the large amount of unlicensed bandwidth around 60 GHz has attracted much attention. The IEEE 802.15.3c Task Group established a channel model and developed a physical layer framework to support the wireless personal area networks operating at the 60 GHz band. In this thesis, the multiple-input-multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) scheme is used over the IEEE 802.15.3c channel model. We consider the scenario in which multiple transmitter-receiver pairs communicate simultaneously such that the transmitter-receiver pairs interfere each other. Interference alignment (IA) is applied to mitigate the effects of the inter-user interference by aligning and suppressing the inter-user interference with the precoding matrices at the transmitters and the combining matrices at the receivers.
The performances of the closed-form IA (CIA), the minimum projection error (MPE) algorithm, the iterative interference alignment (IIA) algorithm, and the maximum signal-to-interference-plus-noise ratio (MSINR) algorithm are compared. The difference of the performance phenomena between IEEE 802.15.3c line-of-sight CM1.1 and non-line-of-sight CM2.1 channel models is investigated. In CM2.1, the MSINR algorithm has better performance than CIA, MPE, and IIA since the power of additive white Gaussian noise is considered in MSINR. In CM1.1, the MSINR algorithm outperforms other algorithms at low signal-to-noise ratios while the four IA algorithms have similar performance at medium to high signal-to-noise ratios. It is also found that different closed-form solutions may result in different performances. For iterative algorithms, initialization of the precoding matrices also affects the performance. Random initialization may result in the convergence to the local optimum, instead of the global optimum. In CM1.1, using the results of the iterative procedure in the previous channel realization as the initialization of the precoding matrices can improve the performance; yet, however, no improvement can be obtained by using this method in CM2.1.
|
|---|
