You are on page 1of 43

Military Radar Summit

Smart Antennas
101
Frank Gross
2/25/15
5
5
0
4
3 -5
2 -10

1
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Relevant Books

5
5
0
4
3 -5
2 -10

2
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Examples of “Smart Antennas”
Precision Acquisition Vehicle Entry
Phased Array Warning System
450MHz: 1,792 Antennas, 145.6 kW

Detect
and
Track
ICBMs
PAVE PAWS
5
5
0
4
3 -5
2 -10

3
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Very Large Array (VLA) – New Mexico

74 MHz to 50 GHz

13 mi

82’

120o
5
5
0
4
3 -5
2 -10
1
-15

y (km)
0
-20

4
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
HF Antennas/ HAARP Facility
High Frequency Active
Auroral Research Program
Gakona, Alaska

HF: 2.8 – 10MHz / 180 Antennas 5


4
3
2
5

-5

-10

5
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Active Electronically Steered Array (AESA)

L-Band (1-2 GHz)


X-Band (8-12 GHz)

F-16

5
5
0
4
3 -5
2 -10

6
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Definition of “Smart Antennas”
 The term “Smart Antenna” refers to any antenna or
array which can adjust or adapt its beam pattern
according to a desired criteria.
w1
Desired signal
w1
w2
y
w2

y

wM
Suppressed
Interference
wM
+ d
Algorithm 
-

“Dumb” Antenna “Smart/Cognitive” Antenna


5
5
0
4
3 -5
2 -10

7
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Smart Antennas
Smart Antennas can encompass:
•Smart Switched Beam Arrays –
• Butler Matrix, Rotman Lens, Plasma Array
•Smart Reconfigurable Antennas –
• Change electrical properties to steer beam
•Smart Surfaces/Metamaterials –
• Substrate change causes beam change
•Smart Vector Antennas –
• Direction finding with single antenna using polarization
•Smart Adaptive Arrays –
• Steer the beam to any direction of interest while
simultaneously minimizing/nulling interfering signals 5
5
0
4
3 -5
2 -10

8
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Switched Beam Array – Rotman Lens

True Time Delay


Beam = Large
Bandwidth
Direction

Feed

Rotman Lens Feeding a Vivaldi Array REMCOM


32 Ports/32 Beams Software Rotman
Frontiers in Antennas, 2011 Design
5
5
0
4
3 -5
2 -10

9
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Applications for Smart Antennas

 Beamsteering
 Jammer Nulling

 Adaptive Tracking

 Multipath Mitigation

 Active Electronically Steered Arrays (AESA)

5
5
0
4
3 -5
2 -10

10
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Antenna Array Basics
z Antennas, d  /2

N-Element
Uniform Linear Antenna Array
Uniform Weights 

0 d 2d (N-1)d

N
Classic
AF  1  e j ( kd sin   )  e j 2 ( kd sin   )  ...  e j ( N 1)( kd sin   )   e j ( n 1)
Array Factor: n 1
 1 
 
 
 
 e j ( kd sin   )  T
Modern a ( )   1 e
j ( kd sin   )
... e j ( N 1)( kd sin   )
Steering Vector:   
 
 j ( N 1)( kd sin  )  2f
e  k
c
5
5
0
4
3 -5
2 -10

11
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Uniform Array Weights

30o

Equal Gains Across Array

Simple Array: Can’t manipulate


sidelobes or nulls
5
5
0
4
3 -5
2 -10

12
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Second Level of Control

 By applying varying scalar weights (gains) to the array elements, we


can suppress sidelobes
 Similar to windowing sampled data in the FFT world
 Window functions can be
 Bartlett
 Hanning
 Hamming
 Kaiser-Bessel
 Dolph-Chebyshev, etc…
 MATLAB has many of these window functions


wM w2 w1 w1 w2 wM
H
-(2M-1)d/2
  

-3d/2 -d/2 d/2 3d/2


  
(2M-1)d/2
AF ( )  w  a ( ) 5
4
5

3 -5
2 -10

13
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Scalar Weight Beamforming

 Kaiser-Bessel weights
 2
 k 
I o  1   ,   k = 0,1,... N/2,  > 1
  N /2 
w(k )   
I o   H
AF ( )  w  a ( )
0
Kaiser-Bessel
-5 Uniform Array

-10

-15
15dB
|AF|

-20

-25

-30

-35

-40 5

-90 -60 -30 0 30 60 90


5
0
4
3 -5
2 -10

 14
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Complex Weight Beamforming
 Weights can also include amplitude and phase to allow us to
steer the beam to a desired angle
 w1 
 
  jkd sin  0

 w2e
w ( )  

 w1 w2e  jkd sin  0
... wN e 
 j ( N 1) kd sin  0 T

  
0o 30o 60o
 
 w e  j ( N 1) kd sin 0 
 N 

d = spacing between antennas


k = wavenumber = 2/
Beamsteering + Kaiser-Bessel

H
AF ( )  w  a ( )
15
Propagation Over Water
 Low flying threats
 Low grazing angle means highly correlated multipath
 AOA accuracy is exacerbated by low-angle multipath
 The evaporation duct can also contribute to multipath
 Let’s use Maximum Likelihood detection
 Find the projection of measured data onto a 2-D space of Gram-Schmidt
orthonormalized steering vectors.
 Model glistening rough sea surface
 Various sea states (SS=0,…,5)
 Model direct path, specular path, & diffuse path, vertical and horizontal polarization, rms
wave height, and rms facet slope.
 Include divergence factor for curved earth, specular scattering factor, and diffuse scattering
factor for rough surfaces
 Use a vertical sparse array to find direct signal

5
5
0

16
4
3 -5
2 -10
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Diffuse and Specular vs. Roughness

5
5
0

17
4
3 -5
2 -10
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Glistening Surface
Like what you see when
looking out over the ocean at
sunset
Direct Specular Diffuse

h2
60

40
h2

Height (m)
20

h1
0
-15 5
-10 4
-5
3
Wid 0 m)
th ( 5 2 (k
m)
1 a nge
10 R
15 0

Glistening
Surface
5
5
0

18
4
3 -5
2 -10
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Total received field

 The signal received at the kth antenna element is composed of


the direct path, specular path, and diffuse path
2Rk
j ( t  )
sk  At e jt   k  s DAt e 
  k  d At  Fm e j (t m )
m
Direct Specular Scattering Diffuse Scattering

Where, k= Classic Fresnel reflection coefficient wrt antenna element


k
D = divergence factor
s= specular scattering factor
d =.5*(1- s) diffuse scattering factor
Rk = specular path length difference to element k
Fm = Amplitude function over range of glistening surface 5
4
5

3 -5
2 -10

19
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Multipath
 The direct path angle is positive relative to the
horizon
 direct
 +
Phased
Array 
 diffuse -

specular

 Need algorithm to search simultaneously for


positive angle direct paths and negative angle
multipath 5
5

0
4
3 -5
2 -10

20
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Maximum Likelihood Solution to Multipath

Helps tackle highly


correlated multipath

 s1( k ) 
N element array;  s ( k )
x( k )  a ( 1 ) a (  2 )  a (  D )  2   n( k )
D arriving angles   
kth time sample  s ( k )
 D 
a = array steering vector  A  s( k )  n( k )
5
5
0
4
3 -5
2 -10

21
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Maximum Likelihood Solution to Multipath

We estimate the D angles of arrival to be : ˆ1 ,ˆ2 ,...,ˆ D

Now form the NxD signal matrix.


C  a (ˆ1 ) a (ˆ2 )  a (ˆD ) 
We may now perform the Gram-Schmidt orthonormalization
procedure to find D basis vectors sd where d = 1,2,…,D.

The S matrix is given as:


S  s1 s2  sD 
Where,
si  s j  0; for i  j
 1; otherwise 5
4
3
5

-5
2 -10

22
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Maximum Likelihood Solution to Multipath

We define the likelihood function as:

L    x  S
Where,

L   is a 1  D  row vector

We can form the L2-norm of the likelihood function to be

L1    L2    ...  LD  
2 2 2
L ( ) 
2
5
5
0
4
3 -5
2 -10

23
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Example with 5 Element Array

N elements; 3 multipath
SS =terms; direct oangle
D = 12.4681
5; Known 12.470 o; d = -0.18043o
; r = =-12.5895
f = high; SS = 5; Estimated
Estimated direct angle:
direct angle = 12.50D  12.5
5

5
0
4
-5
3
2

-10
2

-15
1

-20
0 10 20 30 40 50
1
Maximum Likelihood 5

Surface
5
0
4
3 -5
2 -10

24
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Example 2: Highly Correlated Multipath
Three frequencies of 2GHz, 8GHz, and 14GHz. 5-element vertical
array. Assume direct, specular, and diffuse angles
N elements; 2 multipath terms; 1 = 10o; 2 = -5o N elements; 2 multipath terms; 1 = 10o; 2 = -5o
f = 2 GHz ; Estimated angles: 1  10o ; 2  -5.2499o f = 8 GHz ; Estimated angles: 1  10.0001o ; 2  -5.25o
10 1 10 1

5 5
0.8 0.8

0 0


0.6


0.6
-5 -5

2
2

0.4 0.4
-10 -10

-15 -15 0.2


0.2

-20 -20
-10 0 10 20 30 40 -10 0 10 20 30 40
1 1
N elements; 2 multipath terms; 1 = 10o; 2 = -5o
f = 14 GHz ; Estimated angles: 1  10.0001o ; 2  -5o
10 1

5
0.8
0


0.6
-5
2

0.4
-10

-15 0.2

-20
-10 0 10 20 30 40
5
5
0
4
3 -5
2 -10

1 25
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Adaptive Algorithms

 Adaptive algorithms attempt to minimize a “Cost


Function” J(w)
 The cost function is derived based upon a specified
criteria such as
 Maximum signal-to-interference ratio (SIR)
 Minimum variance distortionless response (MVDR)
 Minimum mean-square error (MSE)
 Constant modulus algorithm (CMA)
 Once the cost function is defined, various methods
can be used to find the minimum either in closed
form or by using a recursion relationship
5
5
0
4
3 -5
2 -10

26
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Maximum Signal-to-Interference Ratio (SIR)

 i1 (k ) 
0  i (k ) 
x (k )  a0 s (k )  a1 a2  a N   2   n (k )
s(k)
x1(k)
w1*   
x2(k)  
1
w2* i
N  ( k )
y(k)  xs (k )  xi (k )  n (k )
i1 (k)
 


N

iN (k)
wM* ?
xM(k)
y (k )  w H  x s (k )  xi (k )  n (k )  w H  x s (k )  u (k )
Let us define the signal-to-interference ratio
Power in desired received signal

 s2 w H  Rss  w
 s2  E  w H  x s   w H  Rss  w
2

 
SIR  2  H  J (w )
 u w  Ruu  w
Power in undesired received signal
Cost Function
 2
 u2  E  w H  u   w H  Ruu  w 5
4
3
5

-5

 
2 -10

27
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Maximum Signal-to-Interference Ratio (SIR)

 We can maximize the SIR by taking a derivative of the


cost function with respect to the needed weights
(amplitude and phase out of HPA)

 w J ( w )   0
 This leads to an eigenvector equation
1
R Rss  wopt  SIRmax  wopt
uu A emax  max I emax
 The largest eigenvalue gives the highest SIR
 The eigenvector solution yields the optimum weights

5
5
0
4
3 -5
2 -10

28
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Example of maximum SIR

 M=10 elements, 0 =300, 1 =-300, 2 =-400, 3 =500


300

0.9

0.8

0.7

0.6
|AF()|

0.5

0.4

0.3
-400
500
0.2
-300
0.1

0
-80 -60 -40 -20 0 20 40 60 80
 5
5
0
4
3 -5
2 -10

29
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Minimum Variance Distortionless Response
(MVDR)

 If the array output:

H
y sw u
Where,
s = undistorted desired output signal
u = undesired interferers
 We can define the variance as:

 2
MV    2
E y  E sw u H 2

H
 w Ruu w
5
5
0
4
3 -5
2 -10

30
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Minimum Variance Distortionless Response
(MVDR)

 Minimizing the cost function we derive the optimum weights


1
R a
wMV  H
uu 0
1
a R a
0 uu 0

Where

Ruu = covariance matrix of undesired signals +noise

a0 = steering vector for desired signal

5
5
0
4
3 -5
2 -10

31
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
3-D Example of MVDR

Airborne Cantor Ring


Nulled Interferers
Array
5
1 5
0
0.8

0.6 -5
0.4 -10
0.2
-15
y (km)
0 0
-20
-0.2

-0.4
-25

-0.6 -30
-0.8 -35
-1 -5
-40
-1 -0.5 0 0.5 1 -5 0 5
x (km)
5
5
0
4
3 -5
2 -10

32
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
High Altitude Constrained Beams

5
5
0
4
3 -5
2 -10

33
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
AESA

 Modern Active Electronically Steered Arrays


(AESA) are filled arrays of wideband elements
 Vivaldi elements
 Fragmented patch elements
 The arrays have GaAs T/R modules for every
antenna element
 The cost per element can be $40 to $500
depending on the power demands

5
5
0
4
3 -5
2 -10

34
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
AESA

 Vivaldi Array
 Fragmented Patch Array GTRI

Applied Radar, Inc. & UMASS

Fragmented
Aperture
Vivaldi Array 33:1
Bandwidth
12:1
Bandwidth 5
5
0
4
3 -5
2 -10

35
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
AESA

Wideband
12-Channel Dual-Pol T/R Module Vivaldi
Array

MMIC
T/R Modules

High-Performance
Composite Radome

5
5
0
4
3 -5
2 -10

36
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Wideband Dual-Pol AESA Antenna

3”

Vivaldi Array T/R Module


Backing Plate Housings

5
5
0
4
3 -5
2 -10
1
-15

y (km)
0
-20
-1

37
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Low-Cost Phased Arrays

Applied Radar
UMASS Design
Vivaldi
Elements

Slant Left/Right
Polarizations

24

192
8 elements
5
5
0
4
3 -5
2 -10
1
-15

38

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Low-Cost Phased Arrays

 AESA arrays are filled arrays with large numbers of


elements
 Wideband array elements cannot be removed
because the inter-element coupling is crucial
 A template can be devised to thin the array and
preserve performance requirements
 Thinned elements are terminated in dummy loads
reducing T/R modules, power demands, heat, $$$

5
5
0
4
3 -5

39
2 -10
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Low-Cost Phased Arrays
 We can thin the number of active AESA elements?
 Resolution
 Uses fewer active elements
– Side Lobe Levels -15 to -20 dB
– Robust with 5 or 4-bit phase shifters
– Use same element phase centers
Thinned AESA Array Configuration; No. Elements eliminated =82

Distinct “thinned”
10
43% thinning pattern

5
z(cm)

-5

110 elements
-10
not 192
5
5
0
4
3 -5
2 -10
1
-15

-10 -5 0 5 10 40

y (km)
0
-20
-1
-2 -25

-3 -30

y(cm)
-4
-35
-5
-40
-5 0 5
x (km)
Low-Cost Phased Arrays

o
AESA Elevation Pattern, Gain = 0 dBi,  o =0 , 6 Bit phase quantization
0
Thinned Array
Robust with 4 or 5-bit -5 Filled Array

phase shifters (11o to 16o) Elevation; 9 GHz; 5-bit phase


-10 -13dB

Gain(dBi)
43% of elements -15
-22dB
eliminated -20

43% fewer elements -25

means lower cost -30


-90 -75 -60 -45 -30 -15 0 15 30 45 60 75 90

Requires ~ 2x HPA power 

for same ERP AESA Azimuth Pattern, Gain = 0 dBi, o =90o, 6 Bit phase quantization
0
Thinned Array
-5 Filled Array

Azimuth; 9 GHz; 5-bit phase


-10
Gain(dBi)

-15 -13dB
-20
-16dB

-25

-30 5
5

-90 -75 -60 -45 -30 -15 0 15 30 45 60 75 90


0
4
3 -5
2 -10


1
-15

y (km)
41
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
AESA Summary

 Filled Vivaldi Array Challenges


 NxM elements, cost, temperature gradients, cooling, weight
 Thinning is a viable option
 Array Thinning
 A novel new approach
 Uses 43% fewer elements for a 192 element array
 Reduces total # elements, weight, heat, etc…
 Performance
 Produces nearly the same beamwidths with lower sidelobes
 Thinned array can be weighted/tapered to suppress sidelobes
 Performs well with 5 bit phase and amplitude quantization

5
5
0
4
3 -5
2 -10
1
-15

y (km)
42
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)
Questions?

5
5
0
4
3 -5
2 -10

43
1
-15

y (km)
0
-20
-1
-2 -25

-3 -30
-4
-35
-5
-40
-5 0 5
x (km)

You might also like