Iterative Multi-User Receivers for CDMA Systems

Mobile communication networks of the third and future generations are designed to offer high-data rate services like video-telephony and data-transfer. The current Rake receiver architecture will create a shortage in available bandwidth offered to the users. This is not due to a shortage in spectrum but results from inefficient receiver architectures. Spectral efficiency can be increased considerably through multi-user detection techniques in the receiver algorithms. The present thesis investigates iterative re- ceivers for encoded CDMA transmission in the uplink. The iterative receiver is a suboptimal receiver algorithm with manageable complexity. It consists of an inter- ference mitigating multi-user detector, a bank of single-user decoders, and a channel estimator. Instead of deciding on the transmitted symbols right after the first decod- ing, the receiver feeds back tentative decision symbols to mitigate multiple-access interference in the next iteration. Similarly, soft decision symbols are used to sup- port channel estimation in every iteration. This thesis analyses the possibilities and limits of an iterative multi-user receiver in realistic mobile radio channels with multi-path propagation. First, we ask the question under which assumption fading channels show block- fading characteristic. It turns out that, for the duration of a UMTS-slot, channels are quasi-block static if the velocity of the mobile does not exceed 30 km/h. We investigate the influence of inter-symbol interference on bit error rate. In the frame- work of UMTS, inter-symbol interference can only be neglected for spreading factors larger than 32. With short spreading factors, the bit error rates can only be min- imised if inter-symbol interference is eliminated in the detector. Further, we analyse whether extrinsic or a-posteriori decoder information shall be used for interference mitigation. In this context we define load as the quotient between number of users to spreading factor. In general, soft decision symbols that are derived from a-posteriori information allow for faster convergence over a large region of the load. However, the maximal load is only achieved when soft decision symbols are used that are derived from extrinsic information. We develop a low-complexity linear MMSE interference suppression filter with which the load can be increased significantly over a single-user matched ilter. Parallel interference cancellation is compared to successive interference cancellation. The achievable loads of the two schemes are practically the same, although successive cancellation requires fewer iterations. Weextend the multi-user receiver to multiple-antenna reception and show that the load can be increased considerably through the stabilisation of average receive power. In particular, we extend the iterative multi-user detector with a channel estima- tor. Additionally to pilot symbols, also soft decision symbols are used for channel estimation. We design channel estimators that use different a-priori knowledge on the second order statistics of the noise, the channel coefficients, and the soft deci- sion symbols. After six iterations the least squares estimator, that does not require any knowledge on the statistics, attains nearly the same bit error rates as the more advanced estimators based on linear MMSE estimation. We further observed, that a simple random QPSK pilot sequence, spread with the same signature sequence as the data, shows a loss of up to 1,75 dB, over a perfect correlation pilot sequence, depending on the particular multi-path channel. Perfect correlation pilot sequences are particularly advantageous when the channel is estimated only once with the pilot symbols. The gained knowledge on multi-user detection and channel estimation is used in the design of a UMTS compliant iterative receiver with low-complexity components. With symbol synchronous transmission the receiver decreases the bit error rates significantly to practically interesting levels. Limits in practical deployment will be discussed. Finally, we analyse the convergence behaviour of iterative receivers with inte- grated channel estimation by means of density evolution for a channel with constant amplitude and random phase. The results for the derivation of the properties of the channel estimator are gained by using analysis tools from statistical physics. The performance of the iterative receiver is assessed by its multi-user efficiency which is parameterised by the transmission block length, the number of pilot symbols, the load, and the signal to noise ratio. Keywords: CDMA, MAP decoding, multi-user detection, multi-path channels, interference cancellation, inter-symbol interference, channel estimation, multiple re- ceive antennas, pilot sequences, UMTS, density evolution, large system analysis.