You are on page 1of 4

10th International Conference on Information SCience, Signal Processing and their Applications (ISSPA 2010)

CFAR Detectors in Presence of Jammer Noise

Amir Zaimbashi, Abass sheikhi


Department of Electrical and Electronics Engineering, Shiraz University, Shiraz, Iran
Emails: zaimbashi@shirazu.ac.ir.sheikhi@shirazu.ac.ir

Abstract- Cell Averaging C ( A)-CFAR detector is an O.4 I-�--�-I-r========::::


- envelope pdf of jammer noise
:;-J
optimum detector under the assumption of Rayleigh 0.35
_. _. - envelope pdf of system noise
distributed homogenous noise background, but in presence of 0.3

clipped jamming noise, it does not have the best performance. 0.25

In this paper, we compareCA, GO, SO, as, OSGO and ossa


conventional CFAR detectors to obtain superior CFAR
detector in clipped jamming noise. Simulation results show
that the OS-CFAR detector not only provides low loss CFAR -.-

performance in exponential background but also performs 10 12 1


4
better than that ofCA in clipping jammer noise situations. On
other hand, by using as detector instead of CA detector in Fig.l Envelope of theremal system noise and clipping jammer noise
radar, it saddles about 3dB more loss to jammer noise of a radar receiver.
generator.

Index Terms-ECM, Jammer noise, system noise, CFAR


detector, CFAR loss
a DINA jammer should use a high power amplifier transmitter
whose compression point is 7dB to 10dB above the root mean
I. INTRODUCTION
square (rms) value of the noise voltage [2].
The objective of electronic countermeasures (ECM) In this paper, we consider the practical situation in which
systems is to reduce the military effectiveness of enemy the jammer noise is clipped. On the other hand, the In-phase
equipment involving electromagnetic radiation, such as radar, (I) and quadrature (Q) components of the jammer noise before
communication, guidance, or other radio-wave devices. To jammer transmitter are assumed independent and identically
accomplish this objective, both noise jammers and deception distributed (liD) Gaussian random variables with zero mean
jammers can be used. A noise jamming system is an ECM and unknown variance 0'2 . But, the envelope amplitude at the
device whose objective is to generate a disturbance in the output of the jammer transmitter is clipped in aO' level (See
radar receiver in such a way as to conceal the presence of a Fig. I).
target. For the jamming to be effective, it is necessary to Since the clutter or noise power is not known, a fixed
generate noise that emulates the thermal noise of the radar threshold detection scheme cannot be applied to the radar
receiver as much as possible. In this way, the radar operator returns in individual range cells if the false alarm rate is to be
cannot be sure that the radar is being jammed, especially if it controlled [3]. An attractive class of schemes that can be used
has a constant false alarm rate (CFAR) detector. One method to overcome this problem is the constant false alarm rate
of generating a noise of this type is called DIrect Noise (CFAR) detectors [4, 5, 6, 7, 9 and 10]. In this paper, we
Amplification (DINA) [1]. In this method the noise can be evaluate six conventional CFAR detectors, the CA [3], GO
generated at base band and then heterodyned to RF where it is [4], SO [5, 6], OS [7], OSGO and OSSO [8]-CFAR detectors,
power amplified and radiated. Therefore, if it could be to select superior detector in jamming noise in such a way that
possible to generate a Gaussian noise, the best noise jamming it not only provides low loss CFAR performance in
waveform, it can be statistically impossible to distinguish exponential background but also performs robustly in presence
between receiver noise generated in victim system and the of clipping jammer noise situations.
externally injected jammer waveform. However, there is a
significant practical problem in generating high-power
Gaussian noise, since most high-power microwave amplifiers
II. CFAR DETECTORS
are limited in the peak power they can handle. To preserve the
character of the Gaussian noise waveform, the designer of A CFAR detector employs an adaptive threshold in order
to remain CFAR irrespective of the interference power. A
block diagram of a CFAR detector is shown in Fig2. The

978-1-4244-7167-6/101$26.00 ©2010 IEEE 426


i = min(U,V) (4)

C. OS-CFAR Detector
In the OS-CFAR [7] the threshold is obtained from one of
the ordered samples of the reference window. The range
samples are first ordered according to their magnitude, as
x(1) $ ... $ X(kos) $ ... $ X(N)

In this case, the statistic i is taken to be the kth largest


Fig.2 structure of a conventional CFAR detector sample X(kos)
,
(5)

range samples are processed through a square law detector and


then are sent serially into a tapped delay line of length N+I. D. OSGO\SO- CFAR Detector
These N+1 samples include N samples in reference (R)
In the OSGO (OSSO)-CFAR detector, the reference
window (Xi i=l, 2, . . . ,N/2, N/2+1,. . . ,N) surrounding the test
samples of the leading (Ie) and lagging (la) windows are
cell Xo and test cell Xo . The samples in the reference window
sorted separately, and the mth sorted sample
are independent of each other and of the sample in the test
cell. When a target is present in the test cell, the use of guard
( xla (m) , Xl, (m) , 0 $ m $ N 12 ) of each window is selected. Then
cells(G) between the test cell and the reference window these samples are compared and the greater (smaller) one is
prevents target energy from corrupting the reference window. selected as the estimation of the statistic i. Therefore, for
The statistic Z which denotes the estimate of the total noise OSGO detector, we have
power is computed by processing the reference cells i = max(Xla(m),XI'(m)) (6)
surrounding the test cell. A target is declared to be present if and for the OSSO-CFAR detector, we have
the cell under test, Xo, exceeds the adaptive threshold pi. The i = min(Xla(m), XI'(m)) (7)
scaling factor fJ is selected so that the probability of false
alarm does not exeed a predefined limit in homogenous
background. In the following subsections, six conventional III. PERFORMANCE STUDY
CFAR detectors such as CA, GO, SO, OS, OSGO and OSSO­ In the simulations of this section, as many similar works,
CFAR detectors, which are presented in [3-8], are introduced. we consider the Rayleigh model for fluctuating targets,
representing Swerling I (SWI) and Swerling II (SWII) (single
A. CA-CFAR Detector pulse assumption) and we examine and compare the
performance of the CA, GO, SO, OS, OSGO and OSSo­
In the CA-CFAR, the statistic i is the sum of N range CFAR detectors in presence and absent of jammer noise. We
cells in the refrence window. This is a complete, sufficient consider N=24, kos=20 and m=IO [9]. In the presence of
statistic for the noise power under the assumption of jammer noise, it is assumed that the power of jammer noise is
exponentially distributed homogenous noise background [3]. larger than that of system noise, for range larger than bum
For this detector, we have through range, and clipping level varied from 0. 2 to 3,
, N
Z = LX, (1) O.2$a$3. In the simulations of this paper, it is assumed that
i=! the threshold is set based on system noise, unless otherwise
Excessive numbers of false alarms in the CA-CFAR at clutter mentioned. Actually, it is assumed that the system donot have
edges and degradation of detection in multiple target situations any intelligence about presence of jammer noise.
are prime motivation for exploring other CFAR processors to In the CFAR detectors, it is desired to have PCa less than a
cure the above problems. Some of these detectors are brought desired value. Therefore, in Fig. 3, the PCa regulation versus
in the following. clipping level, a, of the CA and OS detectors is compared
together. It is seen that the PCa of both detectors are less than
B. GO\SO-CFAR design value, Pca=1O·3, and OS detector has lower masking
target effect.
In the GO (SO)-CFAR detector, the reference samples of In Fig.4, we present the probability of detection (PD)
the leading window, X],X2," " XN/2 , and lagging window, against the signal-to-noise ratio (SNR) for CA, GO, SO, OS,
XN/2+],X N2/ +2, , XN, are sumed separately, as
• • •
OSGO and OSSO-CFAR detectors in presence of noise
N/2 tV
U= LX" V= LX, (2) system only. As it is expected, the CA detector performs better
i=l i=N/2+1 than the other four detectors. Figs. 5 and 6 present the
In these detectors, the statIstIc i is selected the larger performance of the aforementioned detectors in the presence
(smaller) of two seprate sums U and V. For the GO-CFAR of noise jamming for various clipping level of jamming
detector, we have transmitter. From these figures, it can be seen that the
Z = max(U,v) (3) performance of the mentioned CFAR detectors improve as a
and for the SO-CFAR detector, we have increases. Fig. 7 shows that the performance of the mentioned

427
detectors in jamming noise with clipping level equal to 3 are
the same as that in exponetional clutter or noise only. Fig.8
shows the jammer loss of the CA and OS detectors versus the 0.9
value ofa for Pd=O.9, Pfa=1O·3• The jammer loss is defined by
. SNR(Pja, Pd,a) 0.8

Jammer loss = -----""--''---- (8) "-u


= 00
SNRcA (Pja,P",a ) c·
0.7
0

Where SNRcA is the required signal-to-noise ratio for CA­ UQ) 0.6
;;
0
CFAR detector. As it is observed from Fig.8, the more '0
0.5

clipping of noise at output of jamming transmitter, the more >-

:0 0.4
.,
loss is forced to jammer. For example, if a radar use OS­
e
D --CA
"- 0.3
CFAR detector, and transimitter of jammer clip noise at �GO
- - -so
level 0.5cr, then the jammer loss is approximately equal to - ·- os
0.2

�OSGO
11. 6dB at Po=O.9. This loss is decreased to about 8.85 dB for 0.1
--ossa
the CA-CFAR detector. So, a jammer designer should O ����_L_�_L�_L__'_-L__'_L_��
-5 -2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30
consider the above noticeable loss in computing jammer Signal-to- Noise Ratio. SNR(dB)
power and a radar desiner should use the CFAR detector such
as OS to force more loss to jammer. From the Fig. 8, it is Fig.4 Pd comparison of CA, GO, SO, OS, OSGO and OSSO·CFAR in
presence of system noise only.
observed that if there is an intelligent algorithm to detect the
presence of jammer and then set CFAR threshold accordingly,
the above losses can be increased up to 18dB for OS and CA
3
detectors. P
FA
=10- (Threshold Desiged for Exponential Distribution) ,a=0.5

0.9

IV. CONCLUSION 0.8

u
From the results obtained so far, we may make the "- 0.7
C
following general statements about the performance of OS­ .2
13 0.6
CFAR detector relative to CA-CFAR. Though the OS-CFAR *
0
'0
0.5
detector exhibits some loss of detection power in system noise

:0 CA
.,
0.4
only compared with the CA-CFAR detector, its performance --

�os
e
D

in non-homogenous and presence of clipped jamming noise is "- 0.3


-- OSSO
clearly superior. On other words, a jammer designer should 0.2
--OSGO
�GO
consider the 2. 75dB more loss in computing jammer power - - -so
0.1
when a radar system use OS-CFAR detector instead of CA­
CFAR detector. This loss also can be increased by using an O ���-�...L-�-'-----'-L--'---'�-'-�-��
-10 -7.5 -5 -2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25
Signal-to- Noise Ratio, SNR(dB)
intelegent algorithm to detect presence of jamming noise.
Fig.S Pd comparison of CA, GO, SO, OS, OSGO and OSSO·CFAR in the
presence of noise jamming when clipping level of jamming transmitter is
3 equal to a = 0.5 .
Threshold designed for exponential distribution, P =10·
'a

3
P =10· (Threshold Desiged for Exponential Distribution),a=1
FA

0.9

0.8

"-u
C
0.7
0

UQ) 0.6
;;
0
'0
0.5


:0 0.4
.,
e
D --CA
"- 0.3
--GO
10·7 L-_-'-_-'-_----'__.l..-_-'-_-'-_--'__L-_..J - - -so
-as
0.5 1.5 2.5 3 3.5 4.5 0.2
Cliping Level (a) -OSGO
--ossa
0.1

Fig.3 PFA regulation versus clipping level, a , of CA and OS·CFAR in the O��__��...L_�-'-----'_L--'---'�-'-� _��
presence of noise jamming. -10 -7.5 -5 -2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25
Signal-to- Noise Ratio, SNR(dB)

Fig.6 Pd comparison of CA, GO, SO, OS, OSGO and OSSO·CFAR in the
presence of noise jamming when clipping level of jamming transmitter is
equal to a = 1 .

428
3
REFERENCES
P =10- (Threshold Desiged for Exponential Distribution) ,0.=3
FA

[ 1 ] J, Boyd, Electronic Countermeasures, Los Altos, CA: Peninsula


0.9 Publishing, 1 978,
0.8
[2] D.C. Scheleher, Electronic Warfare in the information Age,
Artech House, 1 999.
Q."C 0.7
[3] H.M. Finn, and RS. Johnson, Adaptive detection mode with
.�
� 0.6 threshold control as a function of spatially sampled clutter-level
c3
'0 0.5
estimate, RCA Review, no.29, ppAI4-467, Sept. 1 968,
[4] V,G, Hansen and J,H. Sawyers, "Detectability loss due to greatest

:0 0.4 -- CA
of selection in a cell averaging CFAR", IEEE Trans, Aerospace
'"
.0
-GO
£. 0.3 - - -SO
and Electron, Syst. 1 6(,) ( 1 980) 1 55- 1 69,
-OS [5] G,V, Trunk, B.V, Cantrell and F,D, Queen, "Modified
0.2 -OSGO
generalized sign test processor for 2-D radar", IEEE Trans.
--0550
0.1 Aerospace and Electron. Syst. 1 0 (.) ( 1 974) 574-582.
[6] M. Wesiss, "Analysis of some modified cell-averaging CFAR
O ���� __L-�-L�__L-�-L�__ ���
-5 -2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 22.5 25 27.5 30 processors in multiple target situation", IEEE Trans. Aerospace
Signal-to- Noise Ratio, SNR(dB)
and Electron, Syst. 1 8 (,) ( 1 982) 1 02- 1 1 3,
Fig.7 Pd comparison of CA, GO, SO, OS, OSGO and OSSO-CFAR in the [7] H. Rohling, Radar CFAR thresholding in clutter and multiple
presence of noise jamming when clipping level of jamming transmitter is target situations, IEEE Trans, Aerospace and Electron, Syst. 1 9
equal to a = 3. (4) ( 1 983) 608-62 1 .
[8] A. R. Elias-Fuste, M. G. Mercado and E.R. Davo," Analysis of
some modified ordered statistic CFAR: OSGO and OSSO CFAR"
IEEE Trans. Aerospace and Electron. Syst. 26 ( 1 ) ( 1 990) 1 97-
202.
[9] A. Zaimbashi, M,R Taban, , M,M Nayebi and y, Norouzi ,
Weighted Order Statistics and Fuzzy Rules-CFAR Detector for
Weibull Clutter , EURASIP Journal on signal processing,
33(3)(2008) 558-570.
[ 1 0] A. Zaimbashi and y, Norouzi, Automatic Dual Censoring Cell
Averaging CFAR Detector In Nonhomogenous Environments ,
:z EURASIP Journal on signal processing, 88( 1 1 )(2008) 26 1 1 -
g 10
a; 262 1 .
E
E 8
.'1

O ������-L��������
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
a

Fig.8jammer loss of the CA and OS detectors versus a value for Pd=O.9 and
Pfa<IO·3.

429

You might also like