Experimental results of high-frequency direction finding with modern methods

The authors present experimental results concerning the application of modern parametric direction-finding (DF) methods to high-frequency (HF) signals. HF is challenging since the long wavelength typically implies a small electrical aperture (for reasonably sized arrays), which reduces the resolution of conventional techniques. On the other hand, the complex signal and noise environment is difficult to characterize, which can be challenging to model-based parametric techniques. In these experiments, field data was collected from a uniformly spaced linear HF array and analyzed with four DF methods: the conventional beamformer, Capon´s method, Schmidt´s multiple signal classification (MUSIC) method, and the deterministic-signals maximum-likelihood (ML) method. The field data consists of several long-range, high signal-to-noise ratio (SNR) intercepts of international broadcasts with frequencies between 15 and 25 MHz, as well as several noise-only intercepts. The field data is validated, examined for indications of multipath propagation, and tested against the discrete wavefront model assumed by modern DF methods. The DF methods are then exercised in a variety of ways. Overall, the authors find that improved direction finding is possible at useful SNRs by using modern methods such as MUSIC and ML to model multiple modes. However, it is important that the correct number of modes can be identified, and automatic mode enumeration methods such as the minimum description length (MDL) test do not perform well at HF due to temporal nonstationarity and departure from the discrete-wavefront model