Principles of Modern Radar - Volume 2 1891121537

10.5 STAP Fundamentals 4810−5−10−15SINR Loss ( φ, θ) = (0°, 0°)FIGURE 10-14SINR losscomparison foroptimal filter, STAP,and nonadaptiveprocessingapproaches.dB−20−25−30−35−40−45−50−800 −600 −400 −200 0 200 400 600 800Doppler Frequency (Hz)OptimalSTAP, 2*DoFNon-Adaptivecovariance matrix) SINR loss takes the formL s,1/opt ( f sp , ˜f d ) = sH s−t ( f sp, ˜f d )R −1k s s−t ( f sp, ˜f d )NM/σ 2 n(10.75)Figure 10-14 shows SINR loss corresponding to the data previously used to generate theresults given in Figure 10-10 and Figure 10-13. SINR loss curves are given for the optimalspace-time processor; STAP employing an IID training set of size 2*DoF = 2NM, or 1408training bins at 150 m range resolution; and a nonadaptive beamformer implementationwith a Hanning weight in Doppler, a 25 dB Taylor weighting in azimuth, and no weightingin elevation. We note from this figure the following key points: 0 dB represents noiselimitedperformance; the STAP nominally suffers a 3 dB loss with respect to optimal as aresult of the finite extent of the training data; and the performance of the STAP far exceedsthat of the nonadaptive space-time beamformer.10.5.2 Minimum Variance BeamformerThe MV space-time beamformer is another common formulation for the adaptive processorand employs a weight vector yielding minimal output power subject to a linear constrainton the desired target spatial and temporal response:minw kE [ y k yk∗ ] subject to wHk s s−t ( f sp , ˜f d ) = g (10.76)g is a complex scalar. The minimum variance weight vector is [2,9]w k =g ∗ R −1k s s−t( f sp , ˜f d )s H s−t( f sp , ˜f d )R −1k s s−t( f sp , ˜f d )(10.77)