ON THE INFLUENCE OF INCOMPLETE DATA MODELS ON ESTIMATED ANGULAR DISTRIBUTIONS IN CHANNEL CHARACTERISATION M. Landmann*, W. Kotterman, R.S. Thomä *Technische Universität Ilmenau, Ilmenau, Germany Email: markus.landmann@tu-ilmenau.de Keywords: channel characterisation, parameter estimation, antenna data model, polarimetry, artefacts. Abstract Despite the popularity of the use of high-resolution parameter estimation for channel characterisation and modelling purposes, the influence of the antenna data model on robustness and accuracy of the estimation has been underexposed. The point to be addressed in this paper is that the use of an incomplete data model inherently will result in biased and artificially spread angular estimates. For some antenna array types, estimation bias is even unavoidable, irrespective of the used data model. It is not inconceivable that the popular approach of clustering multi-paths components for channel modelling is spurred by the artefacts resulting from data model choices as described in this paper. Practical examples are given for two popular types of arrays, the uniform linear and uniform circular array. 1 Introduction The impetus for the present work stems from the following observations: 1. Parameter estimation is able to achieve high resolution by incorporating a priori knowledge. 2. Part of that a priori knowledge are the data models that describe antenna responses to particular incident wave-fields, for all possible angles of incidence and all polarisation states of fields. Obviously, the use of inadequate data models precludes high-quality estimates. 3. The fact that many arrays are fitted with antenna elements that only offer one electrical input/output port is easily mistaken for insensitivity to one of the two fundamental polarisations, or for identical behaviour for both polarisations. 4. Full polarimetric calibration of antennas over 4π space angle to make up these data models is time- consuming and cumbersome, as is storage of the calibration data and handling (interpolation of) these data during estimation. 5. Therefore, often, short-cuts are taken, resulting in using a cut through the azimuthal plane and/or calibrating for one polarisation only or even not calibrating at all but relying on theoretical patterns or only partly calibrating. Table 1 gives an overview from recent literature 1 , as opposed to the very few publications that use complete data models [4,5,7]. 6. Linear arrays have an inherent angular response ambiguity that must lead to estimation bias. 7. Up to now, non decent treatise has been published which consequences particular choices for array types and data models have for estimation accuracy, as far as the authors are aware of. 8. As using incomplete data models during estimation may give rise to the occurrence of clusters of multi- path components where there are none, it was felt our findings will be of interest for channel modelling. In view of the aforementioned points, we will discuss the two major effects of incomplete data models: 1. Inadequate treatment of the change of phase distributions over non-linear arrays with elevation; 2. Disregarding the fact that virtually every antenna does receive signal from both polarisation directions, by considering only array patterns for a single polarisation. Ignoring Type of experiments Polarisation Elevation Polarisation + elevation Simulations 6 1 7 Measurements 3 1 4 Table 1 Amount of publications ignoring certain aspects of antenna radiation characteristics in their data model. 1 To give an impression of the standing of these papers, they appeared in Transactions and Proceedings of the IEEE societies APS, CS, SPS, and VTS.