Eigenvector matrix partition and radio direction finding performance

The problem of high frequency radio direction finding (HFDF) is solved, using eigenstructure analysis of the spatial correlation matrix. The general approach is to compute a matrix of eigenvectors which are ordered according to decreasing eigenvalues. The eigenvector matrix is partitioned so that one part is associated with signals and the other is associated with noise. The performance of a direction finder is investigated under the conditions of an incorrectly partitioned eigenvector matrix. In particular, observations are presented for various conditions of 1) signal coherence, 2) bearing separation, 3) unequal signal strength and 4) additive noise. It is noted that underestimation of signal order produces biasing and spurious spectral peaks. In the high signal-to-noise ratio (SNR) case it appears that overestimation of signal order is a desirable strategy. In the low SNR case the estimation of signal order is extremely critical. The effect of the noise is to merge the eigenvalues so that separation of signal and noise components, based on size of the eigenvalues, is difficult to do reliably.