In a typical orthogonal frequency division multiplexing (OFDM) broadband wireless communication system, a guard interval using cyclic prefix is inserted to avoid the inter-symbol interference and the inter-carrier interference. This guard interval is required to be at least equal to, or longer than the maximum channel delay spread. This method is very simple, but it reduces the transmission efficiency. This efficiency is very low in the communication systems, which inhibit a long channel delay spread with a small number of sub-carriers such as the IEEE 802.11a wireless LAN (WLAN). To increase the transmission efficiency, it is usual that a time domain equalizer (TEQ) is included in an OFDM system to shorten the effective channel impulse response within the guard interval. There are many TEQ algorithms developed for the low rate OFDM applications such as asymmetrical digital subscriber line (ADSL). The drawback of these algorithms is a high computational load. Most of the popular TEQ algorithms are not suitable for the IEEE 802.11a system, a high data rate wireless LAN based on the OFDM technique. In this thesis, a TEQ algorithm based on the minimum mean square error criterion is investigated for the high rate IEEE 802.11a system. This algorithm has a comparatively reduced computational complexity for practical use in the high data rate OFDM systems. In forming the model to design the TEQ, a reduced convolution matrix is exploited to lower the computational complexity. Mathematical analysis and simulation results are provided to show the validity and the advantages of the algorithm. In particular, it is shown that a high performance gain at a data rate of 54Mbps can be obtained with a moderate order of TEQ finite impulse response (FIR) filter. The algorithm is implemented in a field programmable gate array (FPGA). The characteristics and regularities between the elements in matrices are further exploited to reduce the hardware complexity in the matrix multiplication implementation. The optimum TEQ coefficients can be found in less than 4µs for the 7th order of the TEQ FIR filter. This time is the interval of an OFDM symbol in the IEEE 802.11a system. To compensate for the effective channel impulse response, a function block of 64-point radix-4 pipeline fast Fourier transform is implemented in FPGA to perform zero forcing equalization in frequency domain. The offsets between the hardware implementations and the mathematical calculations are provided and analyzed. The system performance loss introduced by the hardware implementation is also tested. Hardware implementation output and simulation results verify that the chips function properly and satisfy the requirements of the system running at a data rate of 54 Mbps.

In a typical orthogonal frequency division multiplexing (OFDM) broadband wireless communication system, a guard interval using cyclic prefix is inserted to avoid the inter-symbol interference and the inter-carrier interference. This guard interval is required to be at least equal to, or longer than the maximum channel delay spread. This method is very simple, but it reduces the transmission efficiency. This efficiency is very low in the communication systems, which inhibit a long channel delay spread with a small number of sub-carriers such as the IEEE 802.11a wireless LAN (WLAN). To increase the transmission efficiency, it is usual that a time domain equalizer (TEQ) is included in an OFDM system to shorten the effective channel impulse response within the guard interval. There are many TEQ algorithms developed for the low rate OFDM applications such as asymmetrical digital subscriber line (ADSL). The drawback of these algorithms is a high computational load. Most of the popular TEQ algorithms are not suitable for the IEEE 802.11a system, a high data rate wireless LAN based on the OFDM technique. In this thesis, a TEQ algorithm based on the minimum mean square error criterion is investigated for the high rate IEEE 802.11a system. This algorithm has a comparatively reduced computational complexity for practical use in the high data rate OFDM systems. In forming the model to design the TEQ, a reduced convolution matrix is exploited to lower the computational complexity. Mathematical analysis and simulation results are provided to show the validity and the advantages of the algorithm. In particular, it is shown that a high performance gain at a data rate of 54Mbps can be obtained with a moderate order of TEQ finite impulse response (FIR) filter. The algorithm is implemented in a field programmable gate array (FPGA). The characteristics and regularities between the elements in matrices are further exploited to reduce the hardware complexity in the matrix multiplication implementation. The optimum TEQ coefficients can be found in less than 4µs for the 7th order of the TEQ FIR filter. This time is the interval of an OFDM symbol in the IEEE 802.11a system. To compensate for the effective channel impulse response, a function block of 64-point radix-4 pipeline fast Fourier transform is implemented in FPGA to perform zero forcing equalization in frequency domain. The offsets between the hardware implementations and the mathematical calculations are provided and analyzed. The system performance loss introduced by the hardware implementation is also tested. Hardware implementation output and simulation results verify that the chips function properly and satisfy the requirements of the system running at a data rate of 54 Mbps.

In this thesis, a new design methodology for pipeline FFT processor has been proposed. The pipeline FFT processor can achieve high throughput rate, and is very suitable for those systems where the continuous data sequences that call for the FFT processing enter systems sample by sample sequentially. However, the traditional pipeline FFT design based on the common single delay feed-back approach suffers low hardware utilization for the butterfly unit. In addition, the resulted transformed sequence is in the form of bit-reverse order which is not suitable for some FFT applications such as OFDM (Orthogonal Frequency Division Multiplexing). Therefore, this thesis proposes a novel pipelined FFT design by first splitting the input sequence into two data streams, which can then be applied to the FFT data-path based on the feed-forward dual-delay path data commutator. The resulted FFT architecture can achieve full butterfly utilization such that the required number of adders can be reduced by almost a half. One potential drawback of the proposed approach is that some additional large storage buffer is required at the last stage. However, the additional storage buffer can be re-organized and merged with the output reordering buffer together such that the normal-order transformed output sequence can be generated. The proposed approach has been applied to the design of 8-K point FFT in this thesis. The 8-K FFT architecture proposed in this thesis is designed based on the radix- 2^4 algorithm such that the required number of general complex number multipliers can be minimized to three. The multiplication of is realized by the dedicated constant multiplier architecture. By proper data partition and allocation, the large buffer required for many data commutator and the output reordering buffer can both be efficiently realized by multi-bank single-port memory modules. The other salient features of the 8-K FFT also include the table reduction for twiddle factors as well as the optimized variable internal data representation. The proposed FFT processor has been implemented by the TSMC 0.18um 1P6M CMOS process technology with core area of 8.74 which is the smallest design reported in the literature for normal sequential input/output order FFT.