On interference suppression techniques for communication systems

Doctoral thesis, 2008

The work presented in this thesis concerns interference suppression techniques in two different communication systems. In the first part we study fiber-optic systems with practical non-ideal transmitters and receivers, which give rise to intersymbol interference (ISI). A unique property of this ISI is that it depends on the transition pattern of the transmitted symbols, which changes for different transmitter configurations. The resulting asymmetric signal constellations of the detected symbols lead to unequal bit error rates (BER) in quadrature phase shift keying (QPSK) systems and unequal symbol error rates (SER) in differential quadrature phase shift keying (DQPSK) systems. By deriving the theoretical BER and SER for QPSK and DQPSK systems, respectively, we explain these phenomena, which have not been described previously. Based on the analytic results, system modifications are proposed to mitigate the performance degradations. These modifications require no hardware add-on but merely the adjustment of transmitter coefficients or receiver decision boundaries. With our solutions, the overall BER performance can be improved by up to 2 dB and 3 dB for QPSK and DQPSK systems, respectively. The analytic results can also be used to optimize the bandwidth of the receiver low-pass filter for QPSK systems, or to improve channel mutual information for DQPSK systems. The second part of the thesis presents a fixed-point successive interference canceller (SIC) from an implementation point of view. A SIC is applied to suppress multiple-access interference (MAI) in direct-sequence code-division multiple-access (CDMA) systems. The minimum precision of the fixed-point SIC for a system to achieve satisfactory performance is determined and is used to implement an SIC in a rapid single flux quantum (RSFQ) digital signal processor. A model of the fixed-point SIC is developed to predict the system performance at a given precision, which greatly simplifies the design procedure. We show that the theoretical results agree well with simulations for a moderately loaded system, especially with a high precision, which is of most interest.