[Back]


Doctor's Theses (authored and supervised):

M. Herdin:
"Non-Stationary Indoor MIMO Radio Channels";
Supervisor, Reviewer: E. Bonek, A. Burr; Institut für Nachrichtentechnik und Hochfrequenztechnik, 2004.



English abstract:
Stationarity is an often used assumption for the mobile radio channel. For a single-input single-output (SISO) channel this means that the channel statistics does not change with time or frequency. Unfortunately, real channels never fulfil this exactly. They can only be assumed to be quasi-stationary, i.e. the channel statistics stays approximately constant within a specific timefrequency stationarity region. If this stationarity region is large enough, a transmission scheme can take advantage of the channel statistics by estimating it and adapting the transmission accordingly. For such systems, the performance depends essentially on the extent of this stationarity region. In case of multiple-input multiple-output (MIMO) systems, the spatial structure has to be taken into account, additionally. Because of the great importance of the spatial domain for MIMO systems, especially the time and frequency intervals within which the spatial statistics stay approximately constant, become essential.

The main goal of this thesis is to investigate to what degree indoor MIMO channels can become non-stationary. As a first step, I consider different metrics to measure the non-stationarity of the MIMO channel. It turns out that a space-only stochastic description of the MIMO channel is a good basis for a stationarity definition for MIMO channels. On the one hand, it covers the main aspect of MIMO channels, i.e. the spatial structure, on the other hand, it can easily be applied to measured channels. Based on this approach, I introduce the correlation matrix distance (CMD) to measure to what extent the channel statistics changes over time or frequency. Although, it is a simplified viewpoint, I treat correlation, i.e. the spatial structure of the channel, at transmit and receive side separately. This allows for comparison to MIMO transmission schemes that make use of the channel statistics, since they often adapt the transmission to the spatial structure at transmit side only. Before investigating measurements, I analyse the CMD using synthetic data. It turns out that in case of a CMD below 0.2, the spatial structure of the channel did not change largely, whereas a value of 0.4 or more means that there were significant changes.

In the main part of the thesis, measured channels are investigated. I analyse measurements that were taken in an indoor office environment, and on the Vienna International Airport, at 2.45GHz and 5.2GHz centre frequency. I analyse the non-stationarity of the MIMO channel regarding different aspects including mutual information, variation of the spatial structure, and the influence of the non-stationarity on the performance of a MIMO transmission scheme. I will show that the indoor MIMO channel cannot be assumed to be stationary, a priori.

Movements of the mobile tend to signicantly inuence the spatial structure of the channel, which shows up as changed mutual information and also as performance degradation of MIMO transmission schemes. It will turn out that it becomes necessary to distinguish between downlink and uplink for indoor environments where a fixed base station and a moving mobile is considered. The spatial structure at the base station, i.e. the fixed transmitter in the measurements, is found to be relatively constant if movements of the mobile (the receiver) within one office room are considered. As soon as the mobile moves to another office room, the spatial structure at the base station side changes significantly. The spatial structure at the mobile, however, varies strongly, also for small movements within one office room. The consequence for MIMO transmission schemes is that knowledge of the spatial structure of the channel may be assumed at a fixed base station, but at a moving receiver, this assumption is problematic. People that are moving around, act mainly as blockers, hence they do not signifficantly contribute to the spatial structure of the channel. They only shadow multipath components dependent on their movements.


Electronic version of the publication:
http://publik.tuwien.ac.at/files/PubDat_108035.pdf


Created from the Publication Database of the Vienna University of Technology.