A new method of partially adaptive constant false-alarm rate (CFAR) detecti
on is introduced. The processor implements a novel sequence of orthogonal s
ubspace projections to decompose the Wiener solution in terms of the cross-
correlation observed at each stage. The performance is evaluated using the
general framework of space-time adaptive processing (STAP) for the cases of
both known and unknown covariance. It is demonstrated that this new approa
ch to partially adaptive STAP outperforms the more complex eigen-analysis a
pproaches using both simulated DARPA Mountain Top data and true pulse-Doppl
er radar data collected by the MCARM radar.