| Summary: | 碩士 === 國立清華大學 === 電機工程學系 === 98 === The throughput requirement of advanced MIMO-OFDM systems increases tremendously recently. The complexity of tone-by-tone QR decomposition becomes the bottleneck of implementation. The interpolation-based QR decomposition(IQRD) algorithm has been proven to reduce the complexity. In the thesis, a one step function is proposed to reduce more complexity of IQRD, and the scalability for different-size channel matrices is attractive to implementation. Moreover, a group-ordering interpolation-based QR decomposition (GO-IQRD) is presented by introducing group-ordering concept into IQRD to increase detection performance. To our best knowledge, the proposed QR decomposition has the lowest complexity among QR decomposition algorithms. A configurable architecture of IQRD without group-order is proposed. Besides, QR-based successive interference cancellation (SIC)
with multiple-candidate selection (MCS-QR-SIC) which has scalable complexity and flexible performance is chosen to cooperate with IQRD. These flexible algorithms are integrated together to be a MIMO detector which could be realized in many MIMO-OFDM systems. The iterative architecture of MCS-QR-SIC is presented. The shorten critical path enables it to work on higher frequency, and it shows the proposed hardware could achieve ML detection performance by controller in low-order modulation. Hence, a QR-based MIMO architecture which supports 4 x 4, 4 x 2, and 2 x 2 channel matrices, with QPSK, 16-QAM, and 64-QAM modulations in the specified bandwidth in the 3GPP-LTE standard is presented. The synthesis result of proposed interpolation-based QR decomposition with UMC 90nm 1P9M CMOS technology shows the corresponding throughput is up to 31.25 MQRD/s, so it could meet the requirement of the highest downlink data rate in hardware categories for 3GPP-LTE Rel. 8.
|
|---|
