Linear equalizer is known to effectively suppress intra-cell interference caused by the multi path propagation in high speed downlink packet access (HSDPA) system. Equalizer algorithms are commonly based on linear minimum mean square error (LMMSE) estimators, which can also take into account the colored interference from other cells. However, several receive antennas are required to efficiently cancel inter-cell interference. On the other hand, non-linear methods are attractive if only one receive antenna is present. However, these algorithms tend to require several auxliary parameters to be known. In this paper, we study a two stage single antenna receiver, combining non-linear and linear processing in HSDPA system. The first stage uses blind interference cancellation (BIC) algorithm to cancel the dominant part of the other cell interference and the second stage uses the LMMSE equalizer to suppress intra-cell interference. The performance of the two stage receiver is compared against conventional LMMSE equalizer. The results show that the studied non-linear algorithm achieves improved performance especially on the cell edge.

The exact bit error probability of a matched filter receiver in a known discrete Rayleigh fading multipath channel is determined for a binary antipodal system. The matched filter receiver or the RAKE is one form of a diversity receiver. Traditionally, both the interpath interference (IPI) and the intersymbol interference (ISI) caused by the non-ideal autocorrelation function of the bit waveform are neglected in the analysis of the RAKE. Previously, the exact matched filter bounds; which include IPI, were derived. This bound gives the performance of the matched filter if only one symbol is transmitted. In our analysis, a symbol sequence is transmitted instead of only one symbol. Thus, our analysis includes both IPI and ISI as well as the correlation between the paths, The analysis permits direct comparisons between different kinds of bit waveforms.

The bit-error probability is determined for a binary antipodal diversity system with an adaptive receiver using decision-directed channel estimation. The diversity channels are assumed to be nonselective Rayleigh-fading channels with additive white Gaussian noise (AWGN). The analysis is valid for all linear estimators which use a finite-impulse-response (FIR)-type estimation filter. Numerical results are presented for a minimum mean-square error (MMSE) filter and a moving average (MA) filter.

It is already known that the performance of narrowband interference adaptive filtering in DS spread spectrum systems is additionally degraded in the presence of a frequency hopped jamming signal compared to the case with CW interference. Even so, the adaptive filter can be rather efficient because during the hopping interval, while the jamming signal parameters are constant, the filter adjusts its weights; and if the hopping rate is not too high, there are time periods when the interfering signal is suppressed. In order to have a better insight into the performance of these systems, we analyze the DS spread spectrum receiver behavior, when the jamming signal frequency is continuously changing (chirp signal). We find that this is a reasonable step for a jammer that is trying to be more efficient. In such a case, a two-sided filter structure shows worse overall performance than a prediction error structure. As the jamming frequency sweeping rate or bandwidth is increased, the system performance is degraded. The analysis provides an insight into how much the system performance will be degraded as a function of the interfering (frequency sweeping) signal parameters.>

The exact matched filter bound for an L-path Rayleigh fading channel is determined. The result is used to study the effect of interpath interference. The numerical analysis shows that the loss in the signal-to-noise ratio due to the interpath interference depends on the selected waveform.>

The effect of channel estimation errors upon the performance of an adaptive rake matched filter in a DS spread-spectrum system is studied in a two-path channel. A moving average filter with a memory of M bit intervals is used in the channel estimation. The estimation errors are mainly caused by channel noise and channel impulse response variations during the observation interval. The performance of the adaptive rake depends strongly on the selection of M. The optimal value of M depends on the signal-to-noise ratio and on the variation rate of the channel.>