The effect of the covariance matrix of ionospherically propagated signals on the choice of direction finding algorithm

A large number of different superresolution algorithms are available, but not all are suitable for every application. Some methods, for instance,are restricted to particular array geometries, such as linear arrays, we only consider algorithms which are applicable to arbitrary array geometries. Some techniques also rely on the incident signals having particular properties, such as being uncorrelated, rendering them unsuitable for multipath signals. Thus the algorithms to be implemented for HF superresolution DF need to be matched to the HF multipath environment. Ionospherically propagated signals may consist of several modes, non-stationary in bearing and highly correlated with each other. Furthermore, due to the signal being refracted from an inhomogeneous region, each mode can be considered to consist of a specular ray surrounded by a cone of diffracted rays. The resulting wavefront for each mode may therefore be far from planar. However, for radio direction finding algorithms, the wavefronts are generally modelled as being planar. Any algorithm chosen must be robust, and impervious to a degree of error in the model. The detection of the number of signals present from the information contained in the covariance matrix is a problem, and particularly difficult in the environment described. A set of criteria for signal detection are also proposed. Algorithm implementation is also discussed, since an execution which is both robust and computationally efficient is required