In this work, we investigate the performance of High-Speed Downlink Packet Access (HSDPA) for WCDMA when used in combination with Double Space-Time Transmit Diversity (DSTTD) over temporally and spatially correlated Multi-Input Multi-Output (MIMO) channels with narrow Laplacian angle spread.
We first investigate various fading channel models, their statistical properties and generation methods before introducing the ray-based spatial channel model (SCM) as formalized by the 3GPP/3GPP2 SCM Ad-Hoc Group (AHG). We investigate two generation methods to emulate power azimuth spectra with a Laplacian distribution and show that the equal sub-ray power method suggested by the SCM-AHG is less effective in accurately reproducing the theoretical spatial correlation properties of a channel with Laplacian angle spread than the more straight-forward sampled Laplacian approach.
In order to assess the performance of the combined HSDPA-DSTTD system over the SCM channel we first determine the effect of temporal and spatial correlations on each of their constituent parts. DSTTD is shown to effectively combine receive diversity, transinit diversity and spatial multiplexing and we analyze the behavior of each of these subsystems individually. Receive diversity can be obtained with a maximum ratio combiner(MRC) and we show flow MRC can still extract significant diversity gain in channels with high levels of spatial correlation. Transmit diversity can by provided by means of space-time block codes (STBCs) and we show the very detrimental effect of large Doppler spreads on the performance of STBCs. Very high spectral throughputs can be realized by the Vertical Bell Layered Space-Time (V-BLAST) spatial multiplexing scheme for which we investigate and compare various detection schemes such as zero-forcing (ZF) and minimum mean squared error (MMSE) equalization techniques with optimum successive interference cancellation (OSIC). We investigate efficient algorithms to reduce the decoding complexity, such as the square root MMSE algorithm (SQRA) and sorted QR decomposition(SQRD). Since all these methods are known to be sub-optimal, we discuss the sphere projection algorithm (SPA) which is an efficient add-on to such sub-optimal detection schemes. SPA is specifically designed to improve the decoder decisions in badly conditioned channels and shown to achieve near-maximum-likelihood (ML) performance in uncorrelated channels. We also present ML performance results using the sphere decoder (SD) and show the negative impact of spatial correlation on V-BLAST detection. More in particular, we show that MMSE is more robust to spatial correlation than ZF and that SPA is still notably superior to optimal MMSE detection although it no longer achieves near-ML performance.
Based on our findings on MRC, STBC and V-BLAST, we investigate the performance of DSTTD and determine that for a given spectral throughput DSTTD performs significantly better than spatial multiplexing alone. The performance results for DSTTD over the spatial channel model are shown to be consistent with the previous results for MRC, STBC and V-BLAST. In addition, we show that DSTTD detection using the sphere detector offers better performance than joint ML detection /MMSE interference cancellation.
The main component of HSDPA that determines its performance is the WCDMA turbo codec. We show that the decoding performance of the turbo codec degrades at lower fading rates, due to less effective interleaver operation. Decoding of turbo encoded V-BLAST streams is shown to improve if the MIMO detector can provide the turbo decoder with soft information based on the maximum a posteriori (MAP) criterion. We explain how this soft information can be obtained using the list sphere decoder (LSD) and flow error rates can further be decreased if the MIMO MAP detector and turbo decoder operate in an iterative fashion. At a given level of spatial correlation, the performance of turbo encoded DSTTD is shown to have an optimum in function of the maximum Doppler shift.
We conclude this study by evaluating the performance of the complete HSDPA coding chain over SISO AWGN and Rayleigh channels as well as over the spatial channel model. We present performance results for HSDPA combined with DSTTD using the spatial channel model and the results are shown to be consistent with all our previous findings. With DSTTD, the theoretical maximum throughput of HSDPA is doubled to about 25.6 Mbps/s, when 15 spreading codes are used simultaneously with 16QAM.