Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
James Ward
Opinions, interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the United States Air Force. This work was sponsored by DARPA under Air Force Contract F19628-95-C-0002
Outline
Introduction STAP basics Partially adaptive STAP architectures STAP CFAR detection STAP parameter estimation Multidisciplinary STAP perspective Summary
Outline
Introduction
STAP basics Partially adaptive STAP architectures STAP CFAR detection STAP parameter estimation Multidisciplinary STAP perspective Summary
Ground Clutter
40 30 20 10 0 1 0 0 1 PRF/2
SNR (dB)
PRF/2
v
Surveillance Radar
STAP Tutorial-4 JW 1/12/2012
(STAP)
MIT Lincoln Laboratory
Doppler Filtering
STAP
CFAR Detection & Metrics Tracking & Display
RCVR A/Ds
Front-End filtering
Topics To Be Covered
properties Space-time covariance matrices Degrees of freedom Sample support / training data Pre-Doppler, post-Doppler algorithms SINR Loss MDV DPCA processing vs. STAP Principal components Cross spectral metric
canceller architecture Adaptive CFAR detection Maximum likelihood STAP Cramer-Rao bound on angle and Doppler accuracy Other application areas
Why Adaptive?
Outline
Introduction
STAP basics
Partially adaptive STAP architectures STAP CFAR detection STAP parameter estimation Multidisciplinary STAP perspective Summary
A/D
Pulse Compression
N
(Angle)
1 1 M 1
(Doppler Frequency)
80 70 Required SINR Improvement (dB) 60 50 40 30 20 10 0 50 100 150 Range (nmi) 200 250 Coherent SNR gain Additional rejection required from STAP
50
SINR (dB)
50
Range (nmi)
STAP Tutorial-12 JW 1/12/2012
v U R J
Clutter patch
v x v ! v y ! v uv v z
y
Array orientation (Linear array assumed):
d x d ! d y ! d ud d z
fc (J ,U ) !
2v T uv u (J ,U ) P d T ud u (J ,U ) P
[c (J ,U ) ! fcTr !
2vTr T uv u (J ,U ) P
] c (J ,U ) !
STAP Tutorial-13 JW 1/12/2012
Clutter Iso-Contours
Iso (Velocity) Iso (Range) Iso (Angle)
Clutter azimuth
-60 deg -30 deg 0 deg 30 deg 60 deg
F=1 9 km altitude
Clutter Ridges
Doppler unambiguous F=1 Doppler ambiguous F = 2.5
F!
2v df r
4v p P fr
! 2F
w11
w1M
wNM
7
Optimum weights
w=
R1v
Dimensionality can be very large: NM can be 102 to >104 Covariance matrix unknown a priori and must be
estimated from the radar data
STAP Tutorial-19 JW 1/12/2012
w ! QR v
1
Criterion Formulation Weight Normalization
Maximum SINR
max
w
w v
Q {0
w H Rw 1 Q ! H 1 1 / 2 (v R v ) 1 Q ! H 1 v R v
MIT Lincoln Laboratory
Maximum PD while maintaining CFAR PF Minimum output power subject to unit gain constraint in look direction
max PD (w ) PF ! L
w
min w H Rw
w
w Hv ! 1
Mainlobe clutter
N = 16 Elements M = 16 Pulses Sidelobe clutter Uniformly weighted transmit pattern CNR = 50 dB per element per pulse
F !0
F !1
F !2
F !3
2v F! df r
(F=1 is the DPCA condition)
rank ( Rc ) ! N ( M 1) F
Number of DOF occupied by the full (mainlobe plus sidelobe) clutter ridge
xnm ! e j ( n] m[ ) ! e
j 2T ( n mF ) dP1 sin J
Pulse #3 Time
~ d nm ! ( n mF )d
Rank(Rc) = N + (M-1)F = 4 + (3-1)1 = 6
Clutter rank is the number of distinct effective element positions, or the length of effective synthetic array aperture
kth Eigenbeam:
Pk (] , [ ) ! e v(] , [ )
H k
Eigenbeam #10
Eigenbeam #20
N ( M 1) F k !1
P (] , [ )
k
N ( M 1) F
P
k !1
Pk (] , [ )
x0 E_ 0 a! E v (] , [ ) x Cov_ 0 a! R x
SNR (dB)
Secondary snapshots (target-free range gates for covariance estimation) Noise Jamming Clutter
x1 , x 2 , - , x K E_ k a! 0 x Cov_ k a! R x
Assumptions
Target
Multivariate Gaussian Target only in primary snapshot Common interference covariance matrix
Pulses
Elements
T = pulse repetition interval z = A/D sampling period
Rangegates
z z z z
Outline
Introduction STAP basics Partially adaptive STAP architectures STAP CFAR detection STAP parameter estimation Multidisciplinary STAP perspective Summary
Preprocessor
Estimate Interference
Element-Space Pre-Doppler
Element-Space Post-Doppler
Spatial filtering
Beam
Beam-Space Pre-Doppler
Doppler filtering
Beam-Space Post-Doppler
Pulse
T=FG
Reduced Dimension
Ds Spatial DOF Dt Temporal DOF
Output
STAP
STAP Tutorial-30 JW 1/12/2012
Separable processors easily implemented with cascade of beamformer and Doppler filter Toeplitz structures represent equivalent subaperture filtering in space and/or time Judicious preprocessor design can lessen required DOF after preprocessor MIT Lincoln Laboratory
Duality
Beamspace Pre-Doppler Element-space Post-Doppler
Element #1 Pulse #1 Space Pulse #2 Pulse #2 Element #4
Element #1 Pulse #1
Element #4
d
Space
Pulse #3 Time
Pulse #3 Time
Pulse #3 Time
Rank(Rc) = Ds + (Dt-1)F = 4
Pulse Data
monopulse radars Need more than two spatial DOF (beams/subapertures) for
Jammer nulling Simultaneous clutter cancelling and angle estimation
Doppler FFT(s)
Doppler FFT(s)
important
Step 2
B Beams M Pulses
Beamspace
STAP
Clutter Nulling
Jammer Training
Clutter Training
Lessens total DOF required for STAP Requires training data free of mainlobe clutter for Step 1
Beyond the horizon range gates in low PRF Doppler filter away from mainlobe clutter
80 70 Required SINR Improvement (dB) 60 50 40 30 20 10 0 50 100 150 Range (nmi) 200 250 Coherent SNR gain Additional rejection required from STAP
50
SINR (dB)
50
Range (nmi)
STAP Tutorial-35 JW 1/12/2012
Distance
Element / subaperture phase center Transmit phase center Receive phase center Equivalent monostatic phase center
DPCA: Subtract clutter signals from same equivalent monostatic phase center for perfect clutter cancellation
Distance
Clutter signal on nth element, mth pulse has phases commensurate with effective position dnm = (n+mF)d = (n+m)d
Pulse #2
Pulse #3
Time
Tr
+ -
Tr
+
2Tr
Time
7
+
Doppler FFT
0 1 w! 1 0
2v fr ! d
d vTr ! 2
Subtract signals from same effective phase center for perfect clutter cancellation
F (] , [ ) ! w H v ! e j 2T] e j 2T[
DPCA filter produces a null along the F! clutter ridge ]![
0 0.4 -5 -10 Relative Power (dB) -15 -20 -25 -30 -35 -40 -0.4 -0.2 0 0.2 Spatial Frequency 0.4
Temporal Frequency
0.2
Deep null along F! clutter ridge Doppler cuts are two-pulse canceller responses Performance depends only on true target velocity
-0.2
-0.4
DPCA Issues
Advantages Simple, straightforward processing chain Can be implemented with subaperture beams, sum and difference beams Disadvantages Requires PRF matched to platform velocity Requires precise element pattern matching to get cancellation over whole ridge Velocity and array axis misalignment degrades performance No inherent provision to suppress clutter and jamming simultaneously No inherent provision to adapt to intrinsic clutter motion
Solution: STAP
STAP Tutorial-40 JW 1/12/2012
T
+ -
Doppler FFT
7j(
7j(
Left subaperture
Right subaperture
Outline
STAP CFAR detection STAP parameter estimation Multidisciplinary STAP perspective Summary
15
1Bin
50
Outline
DLOHSERHT NOITCETED
EVITPADANON
EVITPADA
)mk( EGNAR
561 061 551 051 541 01 01 02 03 0
ADAPTIVE DETECTOR
OUTPUT
H1
THRESHOLD
H0
N DIMENSIONAL (Complex)
1 DIMENSIONAL (Real)
Outline
Introduction STAP basics Partially adaptive STAP architectures STAP CFAR detection STAP parameter estimation
Uncertainty indicated by error ellipses (95% confidence) Angle-Doppler estimation a joint problem with STAP
STAP Tutorial-48 JW 1/12/2012
Outline
Introduction STAP basics Partially adaptive STAP architectures STAP CFAR detection STAP parameter estimation Multidisciplinary STAP perspective
Summary