The integration of ultra-wideband, cognitive radio, and multiple-input multiple-output (MIMO) radio technologies represents a powerful tool to improve the spectral efficiency of different wireless communication systems. In that way, new strategies for MIMO channel modelling and characterisation are necessary in order to investigate how the central frequency and the bandwidth affect the system performance. Previous investigations have focused less attention how these parameters affect the MIMO channel characteristics. In this PhD Dissertation, a frequency dependent MIMO channel characterisation is presented from both experimental and theoretical view points. The problems addressed in this Dissertation treat five main areas: measurements, data post-processing, channel synthetic generation, multivariate statistics of MIMO data, and MIMO channel modelling.
A measurement setup based on a vector network analyzer (VNA) has been designed and validated, and wideband measurements between 2 and 12 GHz in static indoor scenarios in both line-of-sight (LOS) and non-line-of-sight (NLOS) conditions have been carried out. Besides, a procedure for post-processing, channel synthetic generation and experimental MIMO channel analysis based on frequency domain measurements has been proposed and validated. The experimental procedure is based on normalized non-frequency selective channel transfer matrices, estimation of complex covariance matrices (CCM), and the Cholesky factorisation of the CCM to obtain finally the coloring matrix of the MIMO system. Moreover, a validated correction procedure (CP) for synthetic channel generation is presented for MIMO cases with large-scale arrays when the CCM is indefinite. This CP permits the Cholesky factorisation of such CCM. On the other hand, the multivariate characteristics of the MIMO data have been investigated and tested using multivariate complex normal distribution (MCND) analysis. This MCND analysis has indicated both large-scale array and wideband dependency on the normality of the MIMO data.
On the other hand, different channel metrics have been selected for future performance analysis of beamforming, space time block coding (STBC) and spatial multiplexing (SM) in LOS and NLOS for indoor environments. Based on statistical results, all studied metrics showed frequency dependency between 2 and 12 GHz under a high signal to noise ration (SNR), equal electric separation between array elements and isolation of the path loss. Remarkable changes on the magnitude and phase distributions of the complex spatial correlation coefficients (CSCC), distribution of MIMO system eigenvalues, MIMO capacity, and multipath richness were observed for different central frequencies.
Finally, a new channel model useful for wideband MIMO system simulations, which describes frequency-dependent (FD) wideband MIMO channels, has been formulated imposing wide-sense-stationary-uncorrelated-scattering (WSSUS) conditions. The model is a FD-deterministic-Gaussian-uncorrelated-scattering (FD-DGUS) MIMO channel model, which offers useful statistical properties for experimental characterisation. This model considers the effect of both the central frequency and the bandwidth on the space-time-frequency variation of the channel. Besides, environment FD effects are considered within the model by means of homogeneous plane waves (HPW) and inhomogeneous plane waves (IPW), being the first model in the literature using IPWs and applying perfect MIMO channel modelling. The FD-DGUS-MIMO channel model characteristics are analysed, and its 2D-space-time-frequency correlation function (2D-STFCF), power spectral density (PSD) functions, and quantities are closed formulated. Moreover, the perfect MIMO channel modelling strategy is formulated for parameters estimations using measurements, and the simulation model with fixed parameters is presented.