Abstract:
The use of wideband Multiple Input Multiple Output (MIMO) communication systems is currently subject to considerable interest. One reason for this is the latest development of 3rd Generation mobile communication systems, such as the wideband technology: Wideband Code Division Multiple Access (W-CDMA), which facilitates 5 MHz wide radio channels. This technology makes very efficient use of limited spectral resources and allows robust communications over time-varying radio channels. However, even though these new wideband systems offer a greater range of services such as higher bit rates, compared to existing systems they are still subject to many of the same fundamental limitations such as coverage and capacity.
For the design and simulation of these mobile radio systems for MIMO wireless propagation (e.g. like the wideband-CDMA), we need channel models that provide the required spatial and temporal information necessary for studying such systems, i.e., the basic modeling parameters in the space-time domains, e.g., the root mean square (rms) delay spread, (DS) is directly connected to the capacity of a specific communication system and gives a rough implication on the complexity of a receiver. Besides, these models must be easy to implement as well as to accomplish with practical important requirements such as simplicity and adaptivity.
The emphasis of the thesis lies in MIMO channel models based on geometry. However, analyses in angular and time domain for transmission scheme are shortly presented to get a better overview of the overall channel and communication systems. As mentioned before the scope of this thesis is the analysis of the spatio-temporal properties for MIMO systems using the clustering approach model proposed under stationary conditions, so, it does not take in to account the Doppler effect in the analysis presented. Assuming stationary conditions; however, the channel models consider the effect of local scatterers, which are grouped into clusters for the analysis. Besides, in Chapter 4, the far scatterers (again grouped into clusters), are analyzed assuming that the path loss will tend to limit their contribution to the overall channel. In addition, because of local scatterers introduce multipath differences that are small compared to the transmit-receive range, the focus is laid on microscopic (Rayleigh) fading only.
Geometric modeling with some stochastic features has becomes popular in describing the characteristics of spatio-temporal radio propagation channel [3–8] even if in some respects stochastic MIMO channels are easier to create from radio channel measurements [48]. One important merit of the geometric approach is that it enables accurate modeling of the correlation between different antenna branches, which is essential for multi-antenna applications such as diversity transmission or reception, beamforming, and the MIMO technique.
The spatio-temporal characteristics of urban environments have been studied using experimental results published in the literature for stationary conditions of the channel, and the environment to be slowing time-varying. Some important conclusions can be made that the comparisons with experimental results have good agreement with the clustering approach proposed. From the comparisons with experimental results, the following observations can be made. First, more clusters are observed at LOS environments or when the Tx–Rx separations are smaller because in such a case, MPCs impinging on the Rx have stronger power. If their path weights are beyond the channel sounder noise threshold, they will be observable, thereby leading to an increased number of effective clusters. Second, the cluster AS and DS are much smaller for LOS-propagation scenarios due to the smaller feature sizes of the pertinent structure. Third, the correlation between the system operating frequency and the number of clusters is weak.
The average number of clusters and MPCs distribution within a cluster is heavily dependent on the resolution of the parameter estimation algorithm. This also depends on the type of scenario, (indoor or outdoor); e.g., from experimental results for indoor scenarios, Chion et al., [41], found as most nine clusters. On the other hand from experimental results in outdoor scenarios, (including above, at, and below rooftop level of the base station locations), Toeltsch et al., [47] found as most four clusters. This investigation has shown that for each individual data set, four-six clusters can be identified with 7°-25° cluster angle spread (AS).
The statistical modeling approach can in turn be used to analyze different smart antenna configuration in UMTS, which fully exploit the characteristics of UMTS as well as the environment that the system operates in. The channel modeling based on the clustering approach described in Chapter 4 and analyzed in more details in Chapter 5 can be useful to simulate the cases of the appearance or disappearance of cluster due to the relative change in position due to small movements between the transmitter (Tx) or receiver (Rx); i.e., is possible to describe the change in the environments due to the different positions between the transmitter (Tx) and receiver (Rx). As also stated in Chapter 5, the derived PDF can be used to simulate a power delay angle profile (PDAP) and to quantify second order statistics, i.e., angle spread and delay spread for a given circular or elliptical shape of the cluster using the rab ratio parameter.
Several experimental results are available to which is possible to compare the theory. In the indoor case, Chong et al. [41] have characterized the indoor wideband channel model to the angular domain through experimental results obtained by a wideband vector channel sounder together with an eight-element uniform linear array (ULA) receiver (Rx). MPCs parameters were estimated using a super-resolution frequency domain algorithm Space Alternating Generalised Expectation (FD-SAGE) and clusters were identified in the spatio-temporal domain by a nonparametric density estimation procedure. The clustering effect also gives rise to two classes of channel power density spectra PDS intercluster and intracluster PDS, which are shown to exhibit exponential and Laplacian functions in the delay and angular domain respectively.
In the outdoor case, Toeltsch et al. [47] used a wideband channel sounder together with a planar antenna array to determine the parameters of the incident waves. A super-resolution algorithm Unitary Estimation of Signal Parameters via Rotational Invariance Tecniques (U-ESPRIT) allows resolving individual MPCs in such clusters and hence enables a detailed statistical analysis of the propagation properties.
Comparisons to experimental results published in [47] are summarized in Tables 5.1, 5.3 and 5.5 respectively. These Tables presents the parameters of the clusters extracted from measurement campaign, i.e., the excess delay, delay spread (DS), (both in terms of distance), DOA, and angle spread (AS), (both in degrees). Then from these experimental results it uses the solution of the system of equations (4.18) from the channel modeling based on the clustering approach proposed in the previous chapter, in order to get the parameters “a” and “rab”, and from there obtain the position of each cluster and the separation distance between each cluster (Sc) and the receiver (Rx). These results are obtained numerically in order to compare the results obtained using the analytical tractable solution. The theoretical parameters obtained from Chapter 4 are also verified, the boundaries of the delay spread (DS) and angle spread (AS) plotted in Figures 4.6 and 4.7 for time domain and Figures 4.8 and 4.9 for angular domain respectively. The results obtained from Tables 5.1, 5.3, and 5.5 are summarized in Tables 5.2, 5.4, and 5.6 respectively.
Figures 5.12, 5.14, and 5.16 respectively, show the position of each cluster based on the measurements of the power delay angle profiles (PDAPs) published in [47]. Tables 5.2, 5.4, and 5.6 respectively, summarize the parameters using analytical tractable solution as derived in details in [40, 87, and 88] for the angular and time domain parameters respectively.
Comparison of theoretical with experimental results is done in two cases. In Indoor, MPC parameters were estimated using an algorithm and DOA is calculated. There gets the two classes of power density spectra (PDS) through clustering effect: Inter-cluster and Intracluster PDS which exhibit the Laplacian function in angular domain like power angular spectrum.
In outdoor, the incident wave parameters are determined by wide band channel with planar array. The individual MPC’s in the clusters are given by an algorithm by which the analysis of detailed statistics of the propagation properties are enabled.
Advantages:
• Simple model in which the number of clusters and their descriptive parameters (TOA, DOA, DS, AS), is limited.
• Energy is clustered into isolated intervals in delay and angle at the BS and MS.
• Clustering in the angular domain influences the MIMO techniques (beamforming, spatial multiplexing, and diversity), while in the delay domain influences the design of receivers.
• The maximum DOA of the signals to the receiver can be calculated with lower time delay.
Disadvantages:
• This cannot explain about the signals received out of the ellipse which are of huge time delay. This cannot explain the signal statistics in presence of the Doppler effects.
• The average number of cluster and MPCs within a cluster is heavily dependent of the parameter estimation algorithm.
Regarding to the average number of clusters and MPCs distribution within a cluster is heavily dependent on the resolution of the parameter estimation algorithm. This also depends on the type of scenario, (indoor or outdoor); e.g., from experimental results for indoor scenarios, Chion et al., [41], found as most nine clusters. On the other hand from experimental results in outdoor scenarios, (including above, at, and below rooftop level of the base station locations), Toeltsch et al., [47] found as most four clusters. Furthermore, as stated in [41], the number of clusters and MPCs detected are also dependent on several others factors such as the Tx-Rx separation, and location, the physical layout of the environment, as well as the dynamic range of the channel sounder.