Detection and Jamming Low Probability of Intercept (LPI) Radars
Detection and Jamming Low Probability of Intercept (LPI) Radars
Detection and Jamming Low Probability of Intercept (LPI) Radars
2006-09
Denk, Aytug.
Monterey California. Naval Postgraduate School
http://hdl.handle.net/10945/2541
NAVAL
POSTGRADUATE
SCHOOL
MONTEREY, CALIFORNIA
THESIS
by
Aytug Denk
September 2006
11. SUPPLEMENTARY NOTES The views expressed in this thesis are those of the author and do not reflect the official policy
or position of the Department of Defense or the U.S. Government.
12a. DISTRIBUTION / AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
Approved for public release; distribution is unlimited
13. ABSTRACT
An increasing number of LPI radars are integrated into integrated air defense systems (IADS) and modern
platforms and weapons, such as anti-ship missiles, and littoral weapon systems. These LPI radars create a
requirement for modern armed forces to develop new techniques, strategies, and equipment.
The primary objective of this thesis is to investigate methods and means to counter LPI radar threats
integrated into a modern platforms and weapons and focus on the related techniques, strategies, and technology. To
accomplish this objective both platform centric and network centric approaches will be examined thoroughly.
i
THIS PAGE INTENTIONALLY LEFT BLANK
ii
Approved for public release; distribution is unlimited
Aytug Denk
Captain, Turkish Air Force
B.S., Turkish Air Force Academy, 1997
from the
Orin Marvel
Second Reader
Dan Boger
Chairman, Department of Information Sciences
iii
THIS PAGE INTENTIONALLY LEFT BLANK
iv
ABSTRACT
An increasing number of LPI radars are integrated into integrated air defense
systems (IADS) and modern platforms and weapons, such as anti-ship missiles, and
littoral weapon systems. These LPI radars create a requirement for modern armed forces
to develop new techniques, strategies, and equipment.
The primary objective of this thesis is to investigate methods and means to
counter LPI radar threats integrated into a modern platforms and weapons and focus on
the related techniques, strategies, and technology. To accomplish this objective both
platform centric and network centric approaches will be examined thoroughly.
v
THIS PAGE INTENTIONALLY LEFT BLANK
vi
TABLE OF CONTENTS
I. INTRODUCTION........................................................................................................1
A. BACKGROUND ..............................................................................................1
B. SCOPE OF THE THESIS...............................................................................2
C. RESEARCH QUESTIONS .............................................................................2
D. METHODOLOGY ..........................................................................................2
E. BENEFITS OF THE STUDY .........................................................................2
F. THESIS OUTLINE..........................................................................................2
II. LOW PROBABILITY OF INTERCEPT (LPI) RADAR ........................................5
A. LPI RADAR PRINCIPLES ............................................................................5
B. CHARACTERISTICS OF LPI RADAR .......................................................6
1. Low Sidelobe Antennas .......................................................................6
2. Irregular Antenna Scan Patterns .......................................................6
3. High Duty Cycle/Wide Band Transmission.......................................8
4. Accurate Power Management.............................................................9
5. Carrier Frequency ...............................................................................9
6. Very High Sensitivity.........................................................................10
7. High Processing Gain ........................................................................11
8. Coherent Detection ............................................................................11
9. Monostatic/Bistatic Configuration ...................................................11
C. LPI RADAR WAVEFORMS........................................................................12
1. Frequency Modulation Continuous Wave (FMCW) Radar ..........12
2. Phase Shift Keying (PSK) Techniques .............................................15
a. Binary Phase Shift Keying (BPSK) ........................................16
b. Polyphase Codes......................................................................19
c. Polytime Codes ........................................................................24
3. Frequency Shift Keying (FSK) Techniques.....................................28
a Costas Code ............................................................................29
b. Hybrid FSK/PSK Technique (With Costas Code) .................31
c. Target Matched FSK/PSK Technique....................................33
D. EXAMPLES OF LPI RADAR......................................................................35
1. Airborne LPI Radars.........................................................................35
a. AN/APG-77 Multimode Radar: ..............................................36
b. AN/APG-79 AESA Radar .......................................................36
c. AN/APQ-181............................................................................37
d. AN/APS-147 Multimode Radar ..............................................37
e. AN/APG-78 Longbow Radar:.................................................38
f. Low-Altitude Navigation and Targeting Infra-Red for
Night (LANTIRN):..................................................................38
2. Maritime LPI Radars ........................................................................39
a. PILOT MK3 LPI Navigation and Detection Radar: .............39
b. SCOUT LPI Surveillance and Navigation Radar: ................40
c. SMART-L D-Band Radar:......................................................40
vii
d. RBS-15 MK3 ASCM: ..............................................................41
3. Land Based LPI Radars ....................................................................41
a. SQUIRE Ground Surveillance Radar:...................................42
b. Gerfaut (TRS 2620 and TRS 2630) Acquisition Radars: ......42
c. GB-SCOUT: ............................................................................43
d. MRSR Multi-Role Survivable Radar: ...................................43
e. MSTAR - Man-portable Surveillance and Target
Acquisition Radar: ..................................................................43
f. EL/M-2140 (Advanced Ground Surveillance Radar): ..........44
g. Improved HARD-3D Radar System: ......................................44
h. EAGLE Fire-Control Radar:..................................................45
i. POINTER Radar System:.......................................................45
j. CRM-100 Surveillance Radar: ...............................................46
k. JY-17A Surveillance Radar :..................................................46
l. CROTALE:..............................................................................47
III. DETECTION OF LPI RADARS..............................................................................49
A. ES RECEIVER CHALLENGES..................................................................50
1. Radar Processing Gain ......................................................................50
2. ES Receiver Sensitivity ......................................................................51
3. Coherent Integration .........................................................................53
B. ES RECEIVERS FOR LPI RADAR DETECTION...................................54
1. Channelized Receivers.......................................................................56
2. Superhet Receivers.............................................................................56
3. Matched Incoherent Receiver (MIR) ...............................................57
4. Acousto-Optic Receiver .....................................................................58
5. Digital Receivers.................................................................................58
C. ES RECEIVER EXAMPLES .......................................................................59
1. High Sensitivity Microwave Receiver (HSMR)...............................59
2. Vigile-300 ............................................................................................61
3. Sabre....................................................................................................62
4. NS-9003A-V2 ES System...................................................................62
D. SIGNAL PROCESSING ALGORITHMS ..................................................63
1. Adaptive Matched Filtering ..............................................................64
2. Parallel Filter Arrays and Higher Order Statistics ........................65
3. Wigner Ville Distribution (WVD) ....................................................67
4. Quadrature Mirror Filter Bank (QMFB)........................................70
5. Cyclostationary Processing (CP) ......................................................73
IV. CLASSIFICATION AND JAMMING OF LPI RADARS.....................................81
A. CLASSIFICATION OF LPI RADARS .......................................................81
B. NETWORK CENTRIC APPROACH .........................................................82
C. JAMMING OF LPI RADARS......................................................................86
1. Probability of Jamming.....................................................................86
2. Sensitivities Required for Jamming .................................................88
3. LPI Radar Jammer Design Requirements ......................................88
a. RF Bandwidth .........................................................................89
viii
b. Video Bandwidth .....................................................................90
4. Jamming FMCW Radars ..................................................................90
5. Jamming PSK Radars .......................................................................93
6. Jamming FSK Radars .......................................................................93
V. CONCLUSION ..........................................................................................................95
LIST OF REFERENCES ....................................................................................................101
INITIAL DISTRIBUTION LIST .......................................................................................105
ix
THIS PAGE INTENTIONALLY LEFT BLANK
x
LIST OF FIGURES
xii
LIST OF TABLES
xiii
THIS PAGE INTENTIONALLY LEFT BLANK
xiv
ACKNOWLEDGMENTS
Initially I would like to thank Mr. Edward Fisher and Professor Orin Marvel for
their support, guidance and patience during this thesis preparation. Without their help,
this work would not be possible.
I would also like to thank to my brothers in arms, Lt.J.G Aykut Kertmen Turkish
Navy, and 1st Lt. Ali Can Kucukozyigit, Turkish Army for their valuable contributions
and support to this work.
Finally, I would like to thank the Turkish Air Force for giving me the opportunity
to study at the Naval Postgraduate School.
xv
THIS PAGE INTENTIONALLY LEFT BLANK
xvi
I. INTRODUCTION
A. BACKGROUND
Most radars, such as surveillance and target tracking radars, have to contend with
very capable and advanced threats on today’s battlefields. These threats range from anti-
radiation missiles (ARMs), radar warning receivers (RWRs), electronic warfare support
(ES) interception capabilities, and electronic attack (EA) systems. All of these are
designed to contribute to the degradation of radar performance by jamming, evasion, or
destruction.
An increasing number of LPI radars are incorporated into integrated air defense
systems (IADS) and modern platforms and weapons, such as anti-ship missiles, and
littoral weapon systems. These LPI radars create a requirement for modern armed forces
to develop new techniques, strategies, and equipment to counter them.
1
B. SCOPE OF THE THESIS
The primary objective of this thesis is to investigate methods and means to
counter LPI radar threats integrated into a modern platforms and weapons and focus on
the related techniques, strategies, and technology. To accomplish this objective both
platform centric and network centric approaches will be examined thoroughly.
C. RESEARCH QUESTIONS
Primary Question
• Can LPI radars be jammed?
Subsidiary Questions
• How does LPI radar gain its advantage?
• What methods can be used to intercept LPI radars?
• What methods can be used to jam LPI radars?
D. METHODOLOGY
Articles, books, periodicals, thesis, IEEE, and DoD documents related to the
subject will be collected and thoroughly examined. The answers to questions stated in the
above section will be established in a reasonable fashion. In order to do this,
comprehensive knowledge of LPI radar systems, detection and jamming methods will be
studied and explained.
E. BENEFITS OF THE STUDY
The results from this thesis will be used to support ongoing efforts by the Turkish
Armed Forces. This thesis will enhance the perspective and knowledge of Electronic
Warfare officers, related project officers, and technical personnel. Furthermore, research
and results will assist the Turkish Armed Forces in evaluating future needs and
requirements of Electronic Warfare systems.
F. THESIS OUTLINE
Chapter II briefly describes the LPI radar techniques, characteristics and
waveforms used in this thesis work. Examples of airborne, maritime, and land-based LPI
radars are given.
Chapter III describes detection methods of LPI radars. For this purpose ES
receivers and signal processing algorithms are examined in detail. Examples of ES
receiver systems used in real operational environment are also given.
2
Chapter IV analyzes both platform and network centric classification and
jamming methods for LPI radars. LPI radar jammer requirements and jamming of LPI
radars are discussed.
3
THIS PAGE INTENTIONALLY LEFT BLANK
4
II. LOW PROBABILITY OF INTERCEPT (LPI) RADAR
Whether or not a radar is LPI depends on the purpose or mission of the radar, the
kind of receiver that is trying to detect it, and the applicable engagement geometry
(Adamy 2001, ).. These types of radars are also described as “quiet” radars.
In order to hide itself from the interception of ES systems and RWRs, the
detection range of radar RR should be longer than that of intercept receiver RI . From
RI
Figure 1, a range factor α can be defined as α = . If α > 1 , the radar will be detected
RR
by the intercept receiver. On the contrary, if α ≤ 1 the radar can detect the platform while
the intercept receiver platform can not detect the radar. In fact, so called LPI performance
is a probability event (GuoSui Liu et al. 2001, 120; 120-124; 124).
5
B. CHARACTERISTICS OF LPI RADAR
Many features distinguish LPI radar from conventional radar. These include:
• Low sidelobe antennas,
• Irregular antenna scan patterns,
• High duty cycle/wide band transmission,
• Accurate power management,
• Carrier frequency,
• Very high sensitivity,
• High processing gain,
• Coherent detection,
• Monostatic/bistatic configurations.
1. Low Sidelobe Antennas
The LPI radar antenna must have a transmit radiation pattern with very low
sidelobes. The low sidelobes in the transmit pattern reduce the possibility of an intercept
receiver detecting the radio frequency (RF) emissions from the sidelobe structures of the
antenna pattern. By applying a tapered illumination, the sidelobe level can be lowered
below -13 dB. For an LPI radar, ultra low sidelobes are required (-45 dB) (Pace 2004,
455)
The mainlobe can not be suppressed in the same manner, so the transmitting beam
should be wide with the radiated energy spread over a wide area. This increases the
difficulty to intercept the radar energy and determine direction of the signal. On the other
hand, the radar receiving antenna should use a narrow beam for high resolution and
detection. It is common to use adaptive arrays for leakage cancellation, multiple receiving
beams, and electronic scanning (GuoSui Liu et al. 2001, 120; 120-124; 124).
2. Irregular Antenna Scan Patterns
Intercept receivers can use scan type and scan rate information to search for,
detect, and identify radars. With confusing radar scan techniques, such as changing the
scan parameters randomly, LPI radar will have a greater chance to avoid interception.
Phased array Electronically Scanned Antennas (ESAs) can be used to produce irregular
6
scan patterns by creating multiple beams to search different scan volumes at different
frequencies. Electronic scanning with software control also helps the LPI radar limit its
illumination time.
Omnidirectional LPI (OLPI) radars use another antenna technique related to the
scan pattern. They use a non-scanning, wide beam transmitting antenna and multiple
receiving beams as shown in Figure 3. This technique increases target dwell time and
reduces radar vulnerability to ES receivers.
7
The German Floodlight radar is an example of OLPI radar. For transmitting, eight
dipoles in a column are combined by a micro strip feed network resulting in a horizontal
fan beam pattern about 20° in elevation and 120° in azimuth. The receiving antenna
consists of an array of 64 columns, each column containing 8 dipoles combined by a
micro strip network similar to the transmitting antenna (GuoSui Liu et al. 2001, 120; 120-
124; 124).
3. High Duty Cycle/Wide Band Transmission
LPI radars escape detection by spreading the radiated energy over a wide
spectrum of frequencies. The ES receiver must search a large bandwidth to find the LPI
radar. The LPI radar is thus able to exploit the time bandwidth product by reducing its
peak transmitted power to bury itself in the environmental noise. Due to the mismatch in
waveforms for which the ES receiver is tuned, the LPI radar is effectively invisible to the
ES receiver (Ong and Teng 2001, ).Since the high peak power transmitted by the pulsed
radar can easily be detected by ES receivers; continuous wave (CW) radars can transmit
very low power while maintaining the same energy profile (Taboada 2002, 271). A
comparison is shown in Figure 4.
8
4. Accurate Power Management
Power management is a radar technique that is becoming more practical with
improvements in digital signal processing. Power management encompasses a host of
techniques including:
• Antenna sidelobe control/suppression
• Pseudo-random illumination of a target
• Dynamic control of transmitter power to maintain a minimal SNR
The French CROTALE system makes effective use of power management.
Shortly after lock on, the tracking radar reduces its transmitter power such that the SNR
of the received level is kept to a minimal value. This process is continued during the
course of engagement reducing the range at which the radar can be detected (McRitchie
and McDonald 1999, ). This LPI technique causes some ES receivers to calculate the
range of the threat incorrectly and categorize the threat as a low priority
5. Carrier Frequency
An LPI radar can use frequencies of 22, 60, 118, 183, and 320 GHz at which peak
absorption occurs. This will serve to maximize attenuation in order to mask the transmit
signal and limit reception by hostile receivers (atmospheric attenuation shielding).
Because of the high absorption of the emitter’s energy, this technique is always limited to
short range systems.
5 (Klein 1997, )
9
Using a radar frequency that is outside of the current ES receivers’ working band
(generally between 0.5 GHz and 20 GHz) is another option for LPI radar carrier
frequencies.
KTB=-114dBm+10log(B) (2.1)
where KTB is the thermal noise in dBm and B is the bandwidth on Hz (Adamy 2001, ).
6 (Adamy 2001, )
10
It is clear that reduction of the radar noise temperature and losses will improve
LPI radar performance.
7. High Processing Gain
Processing gain has the effect of narrowing the effective bandwidth of the radar
receiver by taking advantage of the signal modulation. Thus, the radar receiver achieves a
processing gain while the hostile receiver cannot. An LPI radar achieves bandwidth
advantage over an intercept receiver because the radar knows its own signal. In contrast,
the intercept receiver must accept a wide range of signals and must typically make
detailed parametric measurements to identify the type of signal it is receiving (Taboada
2002, 271).
8. Coherent Detection
Coherent detection is another technique used by LPI radars to avoid interception.
An Electronic Warfare Support (ES) receiver cannot achieve coherent detection of a radar
signal unless it knows the parametric details of the signal. When the signal modulation is
random, this property becomes even more effective. Using true noise to modulate a radar
signal is a good illustration of these characteristics. Radars using true noise modulation
are called random signal radars (RSR). This kind of radar correlates the returning signal
with a delayed sample of the transmitted signal. The amount of delay necessary to peak
the correlation determines the range of a target. Since the transmitted signal is completely
random, the intercepting receiver has no reference for correlating the received signal
(Adamy 2001, ).
9. Monostatic/Bistatic Configuration
Monostatic and bistatic configurations may both be used in LPI radar designs. For
monostatic radar, the leakage of the CW signal from the transmitter must be isolated from
the receiver. For bistatic radars, the transmitting antenna and receiving antenna(s) are
separated by distance. Bistatic radar designs face technological challenges preventing
widespread operational use, such as the synchronization of time and direction, etc. From
all considerations, the bistatic spread spectral CW radar is the most ideal form of LPI
radar. In addition, bistatic radar can minimize the attack of ARMs and increase the
detection of stealth targets (GuoSui Liu et al. 2001, 120; 120-124; 124).
11
C. LPI RADAR WAVEFORMS
There are several LPI radar techniques available to the modern radar designer that
may be used singly or in various combinations, depending on the application. Reducing
the radar’s peak effective radiated power (ERP) by using some form of pulse
compression technique is the most common LPI radar technique. The objective is to
spread the radar’s signal over a wide bandwidth and a period of time. This is typically
done with frequency modulation, phase shift keying and frequency shift keying
techniques (McRitchie and McDonald 1999, ).
1. Frequency Modulation Continuous Wave (FMCW) Radar
Most of the LPI radars use FMCW which is a frequency modulation, pulse
compression technique. This is the simplest and easiest technique to implement with
simple solid-state transmitters. Another advantage of FMCW radars are their extremely
high time bandwidth product which makes them very resistant to interception by ES
systems. Large modulation bandwidth provides very good range resolution. The
deterministic nature of this waveform provides practical advantages over other modulated
CW waveforms because the form of the return signal can be predicted. FMCW technique
provides:
• Resistance to jamming since any signals not matching are suppressed,
• It is simpler to find range information with FFT from IF signals,
• Implementation of sensitivity time control (STC) to control dynamic range
and prevent saturation in the receiver will be easier in the frequency
domain.
The most popular linear frequency modulation is triangular modulation. This
consists of two linear frequency modulation sections with positive and negative slopes.
With this configuration, and by using a continuous 100% duty-cycle waveform, target
range and Doppler information can be measured unambiguously by taking the sum and
difference of the two beat frequencies (Pace 2004, 455). These characteristics are shown
in Figure 7.
12
Figure 7. Linear Frequency Modulated Triangular Waveform and the Doppler Shifted
Return Signal7
∆F ∆F
f1 = f c − + t (2.2)
2 tm
for 0 < t < tm and zero elsewhere. Here f c is the RF carrier, ∆F is the transmitted
modulation bandwidth, td is the round-trip delay time and tm is the modulation period.
t
φ1 = 2π ∫ f1 ( x)dx (2.3)
0
13
Assuming that φ0 = 0 at t=0,
⎡⎛ ∆F ⎞ ∆F 2 ⎤
φ1 (t ) = 2π ⎢⎜ f c − ⎟t + t ⎥ (2.4)
⎣⎝ 2 ⎠ 2tm ⎦
⎡⎛ ∆F ⎞ ∆ F 2 ⎤
s1 (t ) = a0 sin 2π ⎢⎜ f c − ⎟t + t ⎥ (2.5)
⎣⎝ 2 ⎠ 2tm ⎦
The frequency of the transmitted waveform for the second section is similarly
⎡⎛ ∆F ⎞ ∆F 2 ⎤
s2 (t ) = a0 sin 2π ⎢⎜ f c + ⎟t − t ⎥ (2.6)
⎣ ⎝ 2 ⎠ 2t m ⎦
Under normal operating environments, the FMCW radar will generally receive
many signals from targets at different ranges simultaneously. These signals will combine
to form a complex waveform at the output of the receiver mixer. The complex waveform
at the output, after A/D conversion, is resolved into its frequency components using a
Fast Fourier Transform (FFT). The width of each frequency bin of the FFT represents a
range increment and the amplitude of that bin is the echo strength of the target at that
range. The output of the FFT is normally further processed and converted into a ‘regular’
analog video signal which is suitable for PPI display or used for tracking purposes (Ong
and Teng 2001, ).
For any radar waveform, the ideal range resolution, ∆R , is linearly proportional
to time resolution, ∆T , and inversely proportional to the bandwidth of the transmitted
waveform, ∆F , as given below:
14
c∆T c
∆R = = (2.7)
2 2∆F
Figure 8(a) illustrates the triangular modulation signal for a FMCW signal with a
modulation bandwidth of 250 Hz, modulation period of 50 ms and carrier frequency of
1000 Hz. Figure 8(b) shows the power spectral density (PSD) of the triangular FMCW
signal described.
(a) (b)
Figure 8. Triangular Modulation and PSD of the FMCW signal8
Binary phase shifting codes are popular while the most useful codes are the
polyphase codes. Polyphase codes allow the phase shift value within the sub code to take
15
on many values and the code length to be made extremely long. These codes have better
sidelobe performance and Doppler tolerance than binary phase codes.
The PSK techniques can result in a high range resolution waveform, while also
providing a large SNR processing gain for the radar. The average power of the CW
transmission is responsible for extending the maximum detection range while improving
the probability of target detection. PSK techniques are also compatible with new digital
signal processing hardware and solid state transmitters (Pace 2004, 455).
a. Binary Phase Shift Keying (BPSK)
Binary Phase Shift Keying (BPSK) is a modulation technique that has
proven to be extremely effective in communication and radar systems. Even though
BPSK is not a technique presently employed in LPI radar modulation, the technique is
useful as a test signal in evaluating the performance of the signal processing (Taboada
2002, 271).
With BPSK, two output phases are generated for a single carrier
frequency. One output phase represents logic 1 and the other logic 0. As the input digital
signal changes state, phase of the output carrier shifts between two angles that are 180°
out of phase. BPSK is a form of suppressed carrier, square wave modulation of a
continuous wave signal (Jarpa 2002, 154).
Binary phase coded signals exhibit the same range sidelobes seen in FM
chirp signals and mathematicians have spent years developing codes to minimize them.
Three codes are commonly seen in use today are Barker codes, compound Barker codes
and pseudo-random codes (McRitchie and McDonald 1999, ).
16
Table 1. Barker Codes9
Range Sidelobe Processing Gain
Code Length Code
Level (dB) (dB)
2 + - OR + + -6.0 3.0
3 ++- -9.5 4.8
4 + + - + OR + + + - -12.0 6.0
5 +++-+ -14.0 7.0
7 + + +- - + - -16.9 8.5
11 +++---+--+- -20.8 10.4
13 +++++--++-+-+ -22.3 11.1
18
• The feedback register makes them programmable
• Although the spectral content of a PN coded signal is noise-like, the fact
that the phase transitions are binary allows a hostile receiver to make use
of the frequency doubling to remove the coded structure and thereby
detect the signal (McRitchie and McDonald 1999, ).
b. Polyphase Codes
Polyphase codes have many useful features, such as low range–time
sidelobes, ease of implementation, compatibility with digital implementation, and low
cross–correlation between codes. Polyphase codes also have compatibility with bandpass
limited receivers and code lengths of any size are possible.
The major disadvantage of this kind of code is that as the phase increment
becomes smaller, the equipment needed to generate them becomes more complex and
therefore more costly. In addition, the resulting processing is more sensitive to Doppler
shifts. This property will restrict the number of the phase levels employed.
(1) Frank Code: The Frank code is one of the modulation codes
that have been successfully implemented in LPI radars. A Frank waveform consists of a
constant amplitude signal that is phase modulated by the phases of the Frank code
(Persson 2003, 127).
19
2π
φi , j = (i − 1)( j − 1), i=1,2,...,M j=1,2,...,M (2.8)
M
where φi , j describes the phase of the i-th sample of the j-th frequency.
(a) (b)
Figure 10. Frank Code Phase Values for M=8 ( N c =64)11
20
−π
φi , j = [ N − (2 j − 1)][( j − 1) N + (i − 1)] (2.9)
N
where i = 1,2,…,N and j = 1,2,…,N code. Figure 11(a) shows the phase values that
result for the P1 code for M=8, ( N c =64) and Figure 11 (b) shows the signal phase
modulo 2π.
(a) (b)
Figure 11. P1 Code Phase Values for M=8 ( N c =64)12
−π
φi , j = [2 j − 1 − N ][2i − 1 − N ] (2.10)
2N
21
where i=1, 2, …, N and j= 1, 2, …, N. This code has the frequency symmetry of the P1
code while also containing the property of being a palindromic code since the phases are
symmetric in the center of the code (Lewis 1986, ). The P2 polyphase code has more of a
symmetrical frequency spectrum than a Frank coded signal due to its symmetry in the
carrier. Figure 12 (a) shows the phase values that result for the P2 code for M=8,
( N c =64) and Figure 12 (b) shows the signal phase modulo 2π.
(a) (b)
Figure 12. P2 Code Phase Values for M=8 ( N c =64)13
π
φi = (i − 1) 2 (2.11)
Nc
the quadratic discrete phase values that result for the P3 code for N c =64 and Figure 13
(b) shows the signal phase modulo 2π.
(a) (b)
Figure 13. P3 Code Phase Values for N c =6415
π (i − 1) 2
φi = − π (i − 1) (2.12)
Nc
where i=1,2,…, N c , and N c is the compression ratio. Figure 14 (a) shows the discrete
phase values that result for the P4 code for N c =64 and Figure 14 (b) shows the signal
phase modulo 2π.
23
(a) (b)
Figure 14. P4 Code Phase Values for N c =6416
c. Polytime Codes
The polytime waveforms are developed by letting the phase change
approximate a stepped frequency or linear frequency modulation waveform. With
polytime waveforms the subcode period is not uniform in size. That is the size of the
phase step varies as needed to approximate the underlying waveform while the time spent
at any given phase state is a constant. To generate polytime waveform the approximation
of a stepped frequency or linear frequency modulation waveform is generated by
quantization of the underlying waveform into a user selected number of phase states.
Four types of polytime waveforms exist. The first two variants of polytime
coded waveforms, denoted T1(n) and T2(n) where n is the number of phase states, can be
generated using the stepped frequency model. The T3(n) and T4(n) polytime waveforms
are approximations of a linear frequency modulation model. Increasing the number of
phase states increases the quality of the polytime approximation to the underlying
waveform; but it also reduces the time spent at any given phase state, complicating the
generation of the waveform (Fielding 1999, 716-721).
24
(1) T1(n) Code: The T1(n) sequence waveform is generated using
the stepped–frequency waveform where the first code segment is at “zero” frequency.
The equation for the wrapped phase, φ(t) versus time for the T1(n) polytime sequence is
⎧ 2π ⎡ jn ⎤ ⎫
ϕ (t ) = MOD ⎨ INT ⎢ (kt − jT ) ⎥ , 2π ⎬ (2.13)
⎩ n ⎣ T ⎦ ⎭
where j=0,1,2,…,k-1 is the segment number in the stepped RF waveform, k is the number
of segments in the T1 code sequence, t is time, T is the overall code duration, and n is the
number of phase states in the code sequence (Pace 2004, 455).
25
the waveform has an odd number of segments, the zero–beat frequency is the frequency
of the center segment. If an even number of segments are used, the zero frequency is the
frequency halfway between the two center most segments. The expression for the
wrapped phase versus time for the T2(n) polytime sequence is
⎧ 2π ⎡ ⎛ 2 j − k +1⎞ n ⎤ ⎫
ϕ (t ) = MOD ⎨ INT ⎢(kt − jt ) ⎜ ⎟ ⎥ , 2π ⎬ (2.14)
⎩ n ⎣ ⎝ T ⎠ 2⎦ ⎭
where the variables are the same as defined under T1(n) (Fielding 1999, 716-721). An
example of converting a stepped RF waveform into a T2(2) polytime waveform with k=4
segments and n=2 phase steps is shown in Figure 16.
⎧⎪ 2π ⎡ n∆Ft 2 ⎤ ⎫⎪
ϕ (t ) = MOD ⎨ INT ⎢ ⎥ , 2π ⎬ (2.15)
⎩⎪ n ⎣ 2T ⎦ ⎭⎪
18 (Pace 2004, 455)
26
where t is the time, T is the overall pulse duration, ∆F is the modulation bandwidth and is
the number of phase states in the code sequence (Fielding 1999, 716-721). An example of
converting a stepped RF waveform and its conversion into a T3(2) polytime waveform
with k=4 segments and n=2 phase steps is shown in Figure 17.
⎧⎪ 2π ⎡ n∆Ft 2 n∆Ft ⎤ ⎫⎪
ϕ (t ) = MOD ⎨ INT ⎢ − ⎥ , 2π ⎬ (2.16)
⎪⎩ n ⎣ 2T 2 ⎦ ⎪⎭
The variables are the same as defined under T3(n) (Fielding 1999,
716-721). An example of converting a stepped RF waveform into a T4(2) polytime
waveform with k=4 segments and n=2 phase steps is shown in Figure 18.
27
Figure 18. Polytime Waveform T4(2) Derived from Linear FM Waveform20
{t , t ,..., t } . The frequencies are placed in the various time slots corresponding to a
1 2 NF
binary time-frequency matrix. Each frequency is used once within the code period, with
one frequency per time slot and time slot per frequency. The expression for the complex
envelope of the transmitted CW FSK signal is given by
j 2π f j t
s (t ) = Ae (2.17)
28
The transmitted waveform has N f contiguous frequencies with a band B, with
In contrast to the FMCW and PSK techniques, the FSK technique of rapidly
changing the transmitter frequency does not lower the power spectrum density (PSD) of
the emission, but instead moves the PSD about according to the FH sequence (Pace 2004,
455). Besides the advantages of FH radars mentioned above, other important advantages
are:
• Large bandwidths can easily be generated,
• Range resolution depends on the hopping rate and not on bandwidth,
• The use of secret hopping codes,
• The capability to be built with very simple architecture and circuits
(Burgos-Garcia et al. 2000, 23-28).
a Costas Code
In a frequency hopping system, the signal consists of one or more
frequencies being chosen from a set { f1 , f 2 ,..., f m } of available frequencies, for
transmission at each of a set {t1 , t2 ,..., tn } of consecutive time intervals. For modeling
purposes, it is reasonable to consider the situation in which m=n, and a different one of n
equally spaced frequencies { f1 , f 2 ,..., f n } is transmitted during each of the equal duration
time intervals {t1 , t2 ,..., tn } . Such a signal is represented by an nxn permutation matrix A,
where the n rows correspond to the n frequencies, the n columns correspond to the n
intervals, and the entry aij equals 1 means transmission and 0 means no transmission
29
Figure 19. Binary Matrix Representation of (a) Quantized Linear FM and (b) Costas
Signal21
30
Figure 20. The Coding Matrix, Different Matrix and Ambiguity Sidelobes Matrix of a
Costas Signal22
31
As illustrated in Figure 21, with NF frequency hops and NP as the number
of phase slots of duration TP in each sub-period TF, the total number of phase slots in the
FSK/PSK waveform is given by N= NFxNP (Donohoe and Ingels 1990, 268-273).
Figure 21. General FSK/PSK Signal Containing NF Frequency Hops with NP Phase Slots
per Frequency23
32
(a) (b)
Figure 22. ((a) PSD for a Costas Coded Signal (b) PSD of a FSK/PSK Costas Coded
Signal24
33
The matched FSK/PSK radar will then use a correlation receiver with a
phase mismatched reference signal instead of a perfectly phase matched reference. This
allows the radar to generate signals that can match a target’s spectral response in both
magnitude and phase.
Figure 23. Block Diagram of the Implementation of the FSK/PSK Target Matched
Waveform25
34
Figure 24. Frequency Probability Distribution AND Components Histogram26
26(a) FSK/PSK Target 64 Frequency Components and Frequency Probability Distribution (b)
FSK/PSK Target 64 Frequency Components Histogram with Number of Ocurrences per Frequency for 256
Frequency Hops (Taboada 2002, 271)
35
a. AN/APG-77 Multimode Radar: F/A-22 Raptor tactical
fighter’s AN/APG-77 (Northrop Grumman with Raytheon)
multimode radar incorporates a low-observable, Active
Electronically Scanned Array (AESA - incorporating
approximately 2,000 transceiver modules) and is described as
offering long-range, multi target, all-weather, stealth vehicle
detection, electronic intelligence gathering and multiple missile
engagement capabilities. The active array provides frequency
agility, low radar cross section, agile beam steering, and a wide
bandwidth capability typical of LPI radar.
Figure 25. The AESA Antenna Used in the AN/APG-77 Radar27
36
The radar's receiver/exciter features four channels with programmable
waveform generation and is billed as offering a wide bandwidth/fast frequency
agility/low noise/spurious signals. The sensor has a range exceeding 100nm (180km) -
almost twice that of some of today's radars - and can track more than 20 targets
simultaneously (Jane's Radar and Electronic Warfare Systems 2004a, ).
29 (Raytheon )
30Aboard the MH-60R, the Scanner for the APS-147 Multimode Radar is Mounted Below the Helicopter's
Cockpit (Sikorsky) (Jane's Radar and Electronic Warfare Systems 2005, )
37
The AN/APS-147 uses flexibility through programmability, providing a
product optimized for the maritime surveillance mission. Advanced processing allows the
APS-147 to use a collection of waveforms to perform its mission at an output power
substantially lower than traditional counterparts in maritime surveillance radars. This
results in a radar with an extremely Low Probability of Intercept (LPI). Using a low peak
power waveform with frequency agility, the radar can detect medium- to long-range
targets without the threat of electronic warfare support system interception (Jane's Radar
and Electronic Warfare Systems 2005, ).
38
2. Maritime LPI Radars
Just as LPI techniques are useful for covert navigation and targeting for air
applications, they are equally useful for covert maritime applications. LPI is well suited
for this environment as the relatively slow speed of the ship allows for long integration
times and extremely large radar cross sections (RCSs).
In the maritime environment the most significant threat to navies are anti-ship
cruise missiles (ASCMs) with LPI seekers. These ASCM seekers will have power
managed operation in the 8-20GHz range as well as 35-96GHz ranges, by incorporating a
number of advanced electronic technologies. These technologies will enable the missile
to generate a broad collection wideband programmable waveforms with bandwidths
reaching 500MHz to 1GHz. Using a variety of wideband techniques and coherent range
Doppler processing, these seekers will effectively target low RCS ships while
simultaneously allowing the seeker to escape detection and reject decoys such as chaff
(Pace 2004, 455). The following are examples of maritime LPI radars.
The low power causes the ES system to have a very short detection range
while the PILOT has the same navigation radar detection range as a conventional pulsed
radar with peak power levels of several kW. The new Mk3 has frequency agility, which
makes the probability of detection by ES much harder (SPG Media ).
32 (SPG Media )
39
b. SCOUT LPI Surveillance and Navigation
Radar: Scout (Signaal) is an I-band (8 to 10 GHz)
radar with max of 20nm range using FMCW
techniques with low transmitter power (operator
selectable 10 mW, 100 mW or 1 W) for LPI purposes.
It was modified and improved from the FMCW Pilot
radar concept by Thales (Netherlands). SCOUT is used
on especially low cross section corvettes and fast patrol
crafts. Scout Mk 2(S) variants are being offered for use
in mobile or fixed-site coastal surveillance applications
(Jane's Radar and Electronic Warfare Systems 2006b).
Figure 31. SCOUT LPI Radar33
40
d. RBS-15 MK3 ASCM: Saabs RBS-
15 medium range, radar guided, air to
surface missile is one of a family of long
ranged ASCMs produced in Sweden that
can be launched from the air, land, or sea.
Figure 33. RBS-15 MK3 ASCM35
The missile makes use of low RCS materials to reduce the likelihood of
early detection by enemy radar and also has a low infrared signature to reduce the
probability of detection by infrared search and track systems. The seeker uses FMCW
technology and has output power in the milliwatt range that is progressively reduced as
the missile approaches the target.
Saab is developing a future land attack version of the RBS-15 MK3 and is
working on several new seeker technologies that maybe applicable and that may be
retrofitted to existing variants. These include synthetic aperture radar, which would boost
seeker resolution by more than 100% and substantially increase the seeker’s target
discrimination capability as well as the terminal aimpoint accuracy.
Another option is an LPI radar seeker that would use long, coded pulses
that are difficult to detect and difficult to jam. Prototypes for both the synthetic aperture
and LPI seekers are currently under test. The improved MK3 version uses a global
positioning system (GPS) data link, and the range is 400km (Jane's Air-Launched
Weapons 2002, ).
3. Land Based LPI Radars
There are many examples of land based LPI radars generally performing ground
surveillance and short range air surveillance. In the case of ground surveillance role,
these radars can be used to covertly detect ground targets because long integration times
are possible. In the air surveillance role, the high speed of ingressing aircraft does not
permit for extended integration times but typically these radars are used to cue short
range SAM systems. LPI can also be used effectively in the detection of hovering
helicopters. Since there is a little motion, long integration times can be used and this
41
helps the radar to detect the target even though it is embedded within the surrounding
clutter (McRitchie and McDonald 1999, ). The following are examples of land based LPI
radars.
42
c. GB-SCOUT:
The GB-Scout, developed in the Netherlands by Signaal, differs from its
contemporaries in using an FMCW waveform with a very low output power (min 10mW,
max 1W). This reduces the likelihood of it being intercepted while still giving a 90%
detection probability against a truck with 50 square meter RCS at 25km (Jane's
International Defense Review 1994, ).
d. MRSR Multi-Role Survivable Radar:
MRSR (Raytheon), tactical target acquisition and tracking radar, is a US
Army Missile Command program to meet future tactical air defense requirements in the
High-to-Medium Air Defense (HIMAD) and Forward Area Air Defense (FAAD) mission
areas. The radar is a 3D track-while-scan, phased array in elevation radar designed to
acquire and track multiple airborne targets over a 360º azimuth at extended ranges and at
tactical altitudes. Targets include tactical aircraft, unmanned aerial vehicles, and hovering
and slowly moving helicopters. The radar incorporates solid-state, low noise transmitter
technology, and operates over a wide bandwidth with frequency agility.
43
MSTAR features selectable power outputs (1 and 10W) while maintaining
low probability of intercept (Jane's Radar and Electronic Warfare Systems 2006a, ).
f. EL/M-2140 (Advanced Ground Surveillance Radar):
Improved HARD features a LPI capability which is due to its very low
electromagnetic signature, low peak output power (240W, 30W average), broadband
frequency agility, low sidelobes and narrow antenna beam. The radar is difficult to detect
with warning receivers and virtually impossible to attack with anti-radiation missiles,
according to Ericsson Microwave Systems (Jane's Land Based Air Defense 2004, ).
44
h. EAGLE Fire-Control Radar: The Ericsson
Eagle fire-control radar is a silent millimetric system
intended for use in mobile ground and naval-based air
defense systems. The equipment operates in the K-
band (20 to 40 GHz) enabling tracking of low-flying
targets. The Eagle system has been designed with an
extremely low radar signature which has been
achieved by pulse compression, high antenna gain and
almost no sidelobes, in combination with low peak
output power (Jane's Radar and Electronic Warfare
Systems 2004e, ).
Figure 39. EAGLE Fire-Control Radar41
41 BAMSE Missile Control and Launch Vehicle With the Mast-Mounted Eagle Radar/TV/IFF and
the Giraffe 3-D Radar in the Background (Jane's Radar and Electronic Warfare Systems 2004e, )
42 (Jane's Land Based Air Defense 1999, )
45
j. CRM-100 Surveillance Radar: CRM-100 (Poland),
I-band (9.3 to 9.5 GHz sub-band) surface surveillance radar, is
described as being a quiet, solid-state, frequency modulated
continuous wave radar that is designed to detect surface targets,
determine their co-ordinates and automatically hand-off
tracking data (target number, range (from own position),
bearing (from own position), course and speed) to a command
system.
Figure 41. CRM-100 Surveillance Radar43
46
10 km, a light vehicle at 15 km, a helicopter at 20 km, and a ship at 30 km (Jane's Radar
and Electronic Warfare Systems 2004d, ).
l. CROTALE:
Power managed systems such as
French CROTALE have been placed in land
based LPI radar category. Experience has shown
that CROTALE is capable of quickly acquiring
a target and decreasing its transmitting power to
maintain a minimal SNR. This makes it very
difficult for a hostile receiver to detect unless
special techniques are employed (McRitchie and
McDonald 1999, ).
Figure 43. The Monopulse-Doppler Radar Fitted to the Crotale Firing Unit45
48
III. DETECTION OF LPI RADARS
Menahem Oren
Electronic Warfare Support (ES) receivers must perform the tasks of detection,
parameter identification, and classification in a complex environment of high noise
interference and multiple signals in order to exploit LPI radar signals.
Detection of LPI radar signals requires a large processing gain because of the
wideband nature of the LPI radar. The basic idea behind the use of wideband signals is to
spread the radiated power over a large bandwidth in order to produce a Power Spectral
Density (PSD) below the noise at receiver input. Under these conditions, detection is only
possible if the signal is integrated over a long observation time. During that time, a
special integration procedure must be used to ensure that the noise is not being added in
the same amount (Burgos-Garcia et al. 2000, 23-28).
LPI radars are assumed to be low power, high duty cycle signals with phase or
frequency coding. As the coding is unknown and can be complex, and assuming the
frequency is also unknown, then coherent detection is not possible and non-coherent
detection must be performed first. To achieve the maximum sensitivity the RF and video
bandwidth must be matched to the signal modulation allowing detection of the total
signal energy (Rayit and Mardia 1994, 359; 359-362; 362).
49
The detection process is followed by the task of classification. Classification
requires sorting the signal into groups having similar parameters. Parameters such as:
• LPI radar type
• Carrier frequency
• Modulation bandwidth
• Modulation period
• Code period
• Time and angle of arrival.
These are the parameters that distinguish one LPI radar signal from another and
they are required for effective exploitation (jamming). Correlation with existing signals in
a database (identification) can then aid in signal tracking and response management.
To identify the emitter parameters, Fourier analysis techniques have been used as
the basic tool. From this basic tool, more complex signal processing techniques have
evolved, such as the short-time Fourier transform (STFT), so as to track signal
parameters over time. More sophisticated techniques have also been developed, called
time-frequency and bi-frequency distributions, to identify the different modulation
schemes used by the LPI radar. These techniques include the Wigner Ville Distribution
(WVD), Quadrature Mirror Filter Bank (QMFB), and Cyclostationary Processing (CP)
(Pace 2004, 455).
A. ES RECEIVER CHALLENGES
To detect LPI radar signals, ES receivers have to overcome three main
difficulties. These are:
• Processing gain of the LPI radar
• High sensitivity requirement
• LPI radar’s coherent integration
1. Radar Processing Gain
RI
In the second chapter of this thesis, range factor α was defined as α = where
RR
RI was the detection range of the interceptor and RR was the detection range of the
50
radar. If α > 1 , the radar will be detected by the intercept receiver. On the contrary, if
α ≤ 1 the radar can detect the platform while the intercept receiver platform cannot detect
the radar.
When the same radar antenna is used to both transmit and receive ( GTr ≈ GRr ) and
an omnidirectional intercept antenna ( Gi = 1 ) is used for the interceptor, then for a certain
energy or average power transmitted range factor α can be expressed directly in terms of
the radar waveform, antenna pattern and radar cross section as;
RI 1 1
α= = K ⎡⎣( GTi / GTr )(1/(τ Bi ) ) ⎤⎦ 2 (1/ σ ) 4 (3.1)
RR
where K is the constant parameter of the equation, GTi is the antenna gain in the direction
of the interceptor, Bi is the equivalent noise bandwidth of the intercept receiver and τ is
the integration time of the LPI radar. From (3.1), the α is directly proportional to square
root of the antenna gain of the radar antenna in the direction of the interceptor, fourth-
root of the radar cross section and inversely proportional to the time-bandwidth factor
(τ Bi ) which is also the processing gain of the radar receiver over the intercept receiver
(Lee 1991, 55).
LPI radars are effective against some ES receivers when a low-peak power and
long duration signal is used with a large time-bandwidth product. Large bandwidth
signals greater than 10MHz which give 15 meters range resolution may not be needed
unless very high range resolution is required. This implies that signals of relatively
narrow bandwidths and high duty cycles are effective for LPI applications. An effective
time-bandwidth product (processing gain) of around 1000 or 30dB with 10MHz
modulation bandwidth and 1ms integration time is practicable and can be achieved with
some LPI radars. (Lee 1991, 55).
2. ES Receiver Sensitivity
Some ES receivers do not have sufficient sensitivity for the detection of LPI radar
signals. Mr. Jim P. Lee states that a system sensitivity requirement of about -100dBmi
51
will be adequate even for over-the-horizon operation. Sensitivity of ES receiver ( δ I ) can
be calculated as;
where ni is the receiver noise power density, SNRi is the threshold signal-to-noise ratio,
post detection bandwidth, and γ is a parameter 0 < γ < 0.5. (Lee 1991, 55).
intercept receiver over which it can detect signals. The post-detection bandwidth B2
defines the maximum modulation rate that the intercept receiver can measure.
The parameter γ determines the effective bandwidth of the receiver and varies
from a value of 0.5 when B1 > B2 , characteristic of a wide-open, high probability of
intercept receiver, to 0 when the two bandwidths are comparable, characteristic of a high
sensitivity search receiver.
The ES receiver has three basic means for increasing its sensitivity: increasing the
antenna gain, reducing the pre-detection bandwidth and reducing the post-detection
bandwidth. In order to improve sensitivity further, both the noise figure and transmission
loss of the ES receiver should be minimized.
The first two means involve a probability of intercept (POI) loss by reducing
either the angular or frequency instantaneous coverage. The third merely represents a
reduction in the measurement bandwidth of the intercept receiver. Therefore, for
operation against high duty cycle LPI waveforms, there is scope within conventional ES
receivers for increasing sensitivity at negligible cost by reducing the post-detection
bandwidth without compromising the POI (Ruffe and Stott 1992, 200; 200-202; 202).
Table 2 below shows the results from calculations of free space detection ranges
for the PILOT radar, one of the most common FMCW tactical navigation LPI radar,
52
compared with a conventional pulsed radar at 10kW peak power. The detection ranges
are calculated assuming that frequencies and antenna beams all coincide in time.
It can be seen from Table 2 that the PILOT radar with 1W output power can
detect its 100m² RCS target at 28km, whereas its transmissions can only be intercepted at
0.25km with -40dBmi sensitivity. It can also be seen that ES receiver interception range
is coming closer to radar’s maximum detection range with -80dBmi sensitivity. ES
receiver interception range can be calculated as 250km, too much above radar’s
maximum detection range, if the sensitivity of ES receiver were -100dBmi.
Table 2 also shows that the effectiveness of LPI radar performance is strongly
influenced by the radar cross-section of the target to be detected. If the PILOT radar were
required to detect a smaller target, for example an aircraft, with an RCS of 1m²,
transmitted power of 1W would give 8.8km radar detection range and the ES receiver
with -80dBmi sensitivity would intercept PILOT radar much before it detects aircraft.
3. Coherent Integration
LPI radars can integrate their reflected signals coherently over the whole of the
integration time, thus narrowing the receiver noise bandwidth and increasing sensitivity.
On the other hand ES receivers cannot coherently detect the radar’s signals and hence
they cannot narrow their bandwidths in the same manner (Fuller 1990, 1-10).
47 (SPG Media )
53
B. ES RECEIVERS FOR LPI RADAR DETECTION
Some wide-open ES receivers such as the Instantaneous Frequency Measurement
(IFM) and Crystal Video Receivers (CVR) work well in a low density signal environment
where the pulses are short in duration. However, they are susceptible to interference in a
dense signal environment where radar pulses overlap in time. This problem has become
more severe with the introduction of pulse compression waveforms and pulse Doppler
radars with their higher duty cycles. The problem associated with signal overlap may
become worse with LPI signals which are expected to maintain even higher duty cycles.
On the other hand, LPI signals are expected to be of much lower in peak power,
and thus those LPI radars which are far away will not affect the performance of the ES
receiver. However, there are likely to be “friendly” LPI radars on the same platform or
nearby which will cause interference.
As a result, with the proliferation of pulse compression and LPI signals, current
wide-open IFM and crystal video receivers will be more susceptible to interference and
thus are poor candidates for future ES receiver systems. In addition, they do not have the
sensitivity for the detection of current and projected LPI signals (Lee 1991, 55).
With a scenario involving an FMCW LPI radar and an IFM receiver, the effects
of processing gain and sensitivity on detection ranges can be seen. In the scenario the
range at which 100% probability of intercept can be achieved against the main beam of
the radar will be taken as the baseline measure of performance (MOP). Parameters of
both FMCW LPI radar and IFM receiver are based on a reported calculation described in
(Stove, Hume, and Baker 2004, 249-260). These parameters are given in the Table 3
below.
54
Table 3. Parameters of the FMCW Radar and IFM Receiver System48
Radar Type FMCW ES Receiver Type IFM
Mean Transmitter Power 1W IF Bandwidth 2GHz
Antenna Gain 30dB ES Receiver Antenna Gain 0dB
Antenna Sidelobe Level -35dB Video Bandwidth 10MHz
Effective Radiated Power (ERP) 60dBmi Effective Bandwidth 200MHz
Frequency 9GHz Processing Losses 3dB
Integration Time 1ms Minimum SNR for Detection 17dB
Bandwidth 1KHz Net Sensitivity -60dBmi
Incident Power Density from -19
Received Power at 20km Range -125dBm
60dBmi at 2.5km dBm/m²
Target RCS 100m² Received Power at 2.5km -60dBm
Noise Figure 4dB Noise Figure 10dB
Noise Floor -144dBm Noise Floor -80dBmi
Incoherent Integration Gain 4dB Effective Aperture -41dBm²
SNR at 20km Range 15dB
Agile Bandwidth 100MHz
It can be calculated from the parameters in Table 3 that the FMCW radar can
detect its target at 20km range, while its transmissions can only be intercepted at 2.5km
by the IFM receiver. If the FMCW radar is replaced by a pulsed radar with 0.1% duty
cycle, the peak power will be increased by a factor of 1000 and the free space intercept
range increased by about a factor of 30. In other words, the IFM receiver will easily
detect the radar emissions before the radar system detects its target. As a result, it can be
seen that although an IFM receiver can be suitable for low duty cycle pulsed radars, it is
not a suitable ES receiver for LPI radar detection.
55
1. Channelized Receivers
This is a system of many narrowly spaced receiving channels used to measure RF.
This aims to give the best of both worlds, having a large probability of intercept with a
high degree of sensitivity. Each channel is a complete radio receiver tuned to a particular
filter characteristic and the assembly of many channels constitutes a fully parallel
receiver with inherently high data rate capabilities (Fuller 1990, 1-10).
Channelized receiver techniques offer greater sensitivity than the IFM receiver
described in the scenario, by dividing the IF bandwidth (of 2 GHz in the scenario) into a
large number of narrow channels. For example, a sensitivity improvement of about 20 dB
is possible using a channel bandwidth of typically 10 MHz with a lower noise figure and
losses than the IFM based system. The detection range against the FMCW radar in the
scenario with 1W will then be increased to 25km, i.e. it will be approximately equal to
the FMCW radar’s detection range.
A potential counter to this is the random noise (RN) radar. This can have a very
instantaneous bandwidth and thus the intercept range will be reduced if the transmission
bandwidth is greater than the channel bandwidth. This is due to signal in any one channel
potentially being below the detection threshold, even if the total power (which is spread
over several channels) exceeds it.
The linear FMCW waveform does not have RN radar’s advantage because the
signal is not instantaneously wideband and in any practical scenario the received signal
will ‘dwell’ in a channel for a period longer than the reciprocal of the channel’s
bandwidth, and so will be detected (Stove, Hume, and Baker 2004, 249-260).
2. Superhet Receivers
A lower-cost alternative to the channelized receiver is to use a superheterodyne
receiver which uses filtering and mixing to translate the signal to a lower intermediate
frequency (IF). This has the advantage of enabling a narrowband channel with higher
sensitivity to be tuned over a desired operating range. Superheterodyne receivers are also
able to analyze one signal at a time without interference from signals close in frequency,
and hence are suitable for emitter identification. This form of receiver can be especially
useful if a search is to be made for a specific radar type.
56
Table 4. Sensitivity of the Superheterodyne Receiver49
IFM Receiver Sensitivity (from Table 3) -60dBmi
Lower Losses -3dB
Lower Noise Figure -4dB
Narrower Bandwidth -22dB
Net Sensitivity -89dBmi
Table 4 shows the sensitivity of the superheterodyne receiver with the IFM
receiver system sensitivity. Even in the ‘non-tuned’ case the receiver outlined in Table 4
would still detect the main beam of the FMCW radar, in free space, at 70km range, i.e.
considerably greater range than that at which the radar can detect its target (Stove, Hume,
and Baker 2004, 249-260).
3. Matched Incoherent Receiver (MIR)
The matched incoherent receiver overcomes the mismatch currently found
between the bandwidths of radars and intercept receivers (Stove, Hume, and Baker 2004,
249-260). Growth in computing power makes it feasible for a parallel processor to carry
out matched filtering in a number of channels to combat a number of potential threats
simultaneously. The MIR would be matched to the RF information and information
bandwidths of the radar, but not to its actual transmitted waveforms. This is because it
still does not match to the phase of the signal as does a coherent matched receiver.
Moreover, the radar no longer has the advantage of a mismatch between its bandwidth
and that of the intercept receiver, only the advantage of knowing its own waveform and
which part of its agile bandwidth it is actually using at any given time.
For the scenario above, the MIR would have an effective bandwidth of 200KHz,
making it 30dB more sensitive than an IFM receiver. If MIR has 7dB improvement over
the IFM receiver due to lower losses and noise figure which was assumed for the
channelized receiver and the superhet receiver, the MIR will have a sensitivity of -
97dBmi, giving it a free-space detection range of 177km against FMCW radar in the
scenario (Stove, Hume, and Baker 2004, 249-260).
57
4. Acousto-Optic Receiver
The receiver requirement for a relatively a large number of narrow channels with
a narrow video bandwidth for the detection of LPI radars can be easily met by the use of
a time-integrating acousto-optic receiver. The narrow video bandwidth and the relatively
large number of channels can be implemented relatively easy by using a time-integrating
photodetector array.
Considerable progress has been made on the development of both 1-D and 2-D
acousto-optic receivers. In a 1-D configuration, the acousto-optic receiver performs
spectrum analysis on the received signals while in the 2-D configuration both spectrum
analysis and direction-finding are carried out. An Acousto-optic receiver which is
suitable for the detection of LPI signals can be implemented easily using “off-the-shelf”
photodetector arrays with variable integration times.
The effective integration time (video bandwidth) of the acousto-optic receiver can
be adjusted to match the duration of the signal intercepted for maximum sensitivity. This
can be accomplished by either changing the integration period on the photo detector array
or changing the number of samples integrated digitally (Lee 1991, 55).
5. Digital Receivers
Most recent receivers deployed for LPI radar detection are digital, using mainly
Fast Fourier Transform (FFT) as a signal processing technique. With these digital
processing techniques such as FFT, the processing gain of the LPI radar is overcome.
Digital receivers, often called software radios, place a high performance burden
on the ADC, but allow a good deal of flexibility in post detection signal processing. ES
receiver parameters of interest include sensitivity, dynamic range, resolution,
simultaneous signal capability, complexity, and cost. Figure 45 shows a block diagram of
wideband digital ES receiver (Pace 2004, 455).
58
Figure 45. Wideband Digital ES Receiver50
After the signal is down-converted, digitized and sorted, the parameter encoder
forms a pulse descriptor word (PDW). For LPI CW emitters, the PDW contains the center
frequency, the signal coding details such as the modulation period and bandwidth
(FMCW), code period and subcode period details (PSK), and frequency-hopping
frequencies (and order), as well as the signal’s angle of arrival. In all cases, the signal is
down-converted to a baseband frequency that depends on the analog-to-digital converter
(ADC) technology that is available.
The trend in ES digital receivers is to push the ADC as far towards the antenna as
possible, and eliminate the down-conversion stage. This is due to bothersome spurious
signals, nonlinearities, and image frequencies that the mixing and filtering operations
cause. Although the development of standard components, such as ADCs, that are
essential for such a concept have made considerable advancements recently, more
wideband solutions are required using electro-optics (extremely wideband) and
superconductivity (high sensitivity) (Pace 2004, 455).
C. ES RECEIVER EXAMPLES
The following are some of ES receivers that have the capability of LPI radar
detection, identification and classification.
1. High Sensitivity Microwave Receiver (HSMR)
Tenix Defense Electronic System Division’s High Sensitivity Microwave
Receiver (HSMR) is an ES receiver capable of operating in aircraft, ships, submarines
and ground vehicles, and can stand-alone or be integrated with existing ES systems.
59
The signal detection of the HSMR system is designed specifically to detect LPI
radars and, combined with its very high sensitivity, provides the early detection of LPI
emitters. The HSMR has an instantaneous bandwidth of 0.5, 1.0 or 2.0 GHz and is fed
from a wide-band receiver operating nominally between 0.5 to 18.0 GHz. For LPI radars
with very low effective radiated power, the HSMR provides significant warning times.
The level of integration with existing ES equipment can range from none at all to a fully
automated operation. As a stand-alone system, the HSMR will display the presence of
LPI signals. This allows the operator to manually task the ES system to search for and
identify the LPI signal. As a fully integrated system, HSMR detections can be integrated
to include automatic library identification and then pass parametric information to the ES
system on predetermined, or operator selected, emitters for further analysis (Tenix
Defense 2005, 2).
The HSMR Acousto-Optic Module (AOM) is a custom, three slot VXI module
containing the Radio Frequency (RF) amplifiers, 0.5, 1.0 or 2.0 GHz Bragg Cell, a 1024
element photo detector array and a digital signal processor. A wide-band analog signal
(3.0 GHz intermediate frequency) is fed into the HSMR AOM and the resultant spectral
power information is sent via an Ethernet connection to the HSMR executive processor.
The VXI backplane is only used for power and ground, with no use of the data paths. The
specifications for the 2.0 GHz system are shown in the Table 5. HSMR specifications at
the Table 5 show that the receiver system covers the most common radar band of 0.5-
18GHz with a very high sensitivity of >-90dBm which is a requirement for LPI radar
detection.
Specifications of HSMR51
Table 5.
Wide-band Receiver
Input Frequency Range 0.5-18.0 GHz
Output Frequency Range 2.0-4.0 GHz
HSMR Core Module
Instantaneous Frequency Range 2.0-4.0GHz
Instantaneous Dynamic Range >45dB
Detector Array 1024 Photo-diode elements
Pixel Resolution 2.0MHz
Detector Integration Time 3ms
51 (Tenix Defense 2005, 2)
60
Frame Update Time 48ms
Overall System Sensitivity >-90dBm
Interfaces
Control, Access to Spectral and Processed Data Commercial Standard VXI
High Speed Access to Spectral Data Ethernet, RS232, RS422
2. Vigile-300
Vigile-300 is a digital scanning receiver giving an enhanced sensitivity over a
narrow band, allowing detection of LPI radars and fine-grain analysis ('fingerprinting').
Thales representatives state that "There are lots of systems that do fingerprinting, but the
challenge is to be able to do it automatically". Vigile 400 features high-accuracy DF with
a wideband interferometer array that exploits a dual-polarized anti-roll antenna design,
61
minimizing ambiguity and giving sub-degree accuracy on E-J bands, suitable for passive
targeting (JANE'S NAVY INTERNATIONAL 2004, ).
3. Sabre
Thales Airborne Systems has also developed the earlier Sabre EW as shown in
Figure 47 for the Royal Netherlands Navy's four De Zeven Provincien class air defense
and command frigates. The first three are at sea, and in tests Thales states that Sabre has
shown a very high level of accuracy and sensitivity, including an ability to detect LPI
emitters such as the company's own Scout FMCW radar.
62
Table 6. Specifications of NS-9003A-V2 ES System
Frequency Coverage 2-18 GHz (0.5-40 GHz option)
Frequency Accuracy 2 MHz
Azimuth Coverage 360°
DF Accuracy 2° RMS
POI 100% (claimed)
Sensitivity -65 to -75 dBm
Dynamic Range 60dB
63
use of the Fast Fourier Transform (FFT) (Taboada 2002, 271). Below are some signal
processing algorithms that are used for LPI radar detection.
1. Adaptive Matched Filtering
To detect LPI radar signals at operationally useful ranges, the intercept receiver
must overcome the processing gain advantage of the radar. One method to regain the
advantage is to form a matched filter to the LPI radar waveform. Achieving similar
processing gain, the interceptor’s signal detection capability will be identical to that of
the LPI radar; that is, dependent only on the energy contained within the signal. If the
adaptive filter is matched with the LPI waveform, mismatch disadvantage will be
eliminated.
To construct a matched filter, the transmitted frequency, the slope of the FM and
the repetition period have to be known. However, these features of interest are not
normally known to the LPI radar interceptor. As such, the matched filter has to be
adaptively formed. The LPI radar signals have to be estimated and incorporated into the
matched filter and adaptively changed as part of the detection process. In the construction
of the adaptive matched filter, any inaccuracy in the feature estimates (mismatched filter)
will lead to a loss in processing gain.
An adaptive matched filter for the PILOT radar waveform was developed by Mr.
Peng Ghee Ong using a technique employed by pulse compression radar, called
deramping. The deramping process mixes the input signal with a locally generated linear
FM signal to produce an output signal of reduced FM slope in comparison with the input
signal (Ong and Teng 2001, ).
From the analysis of the deramping process, the frequency range of the output can
be predicted when the features of the matched filter are closely tuned to that of the target
LPI radar waveform. The output of the deramped signal can then be easily processed
using a FFT filter bank that covers the expected frequency range. Figure 48 shows an
example of a LPI radar detector using analog deramping (Ong and Teng 2001, ).
64
Figure 48. LPI Radar Detector Block Diagram54
The use of Higher Order Statistics (HOS) and parallel filter arrays along with the
extraction of the most important features provide an accurate analysis and interpretation
of unknown signals in real time (Taboada 2002, 271).
The use of parallel filter arrays and HOS is an effective technical approach for
detecting and classifying LPI radar signals where the waveform of the signal is unknown.
54 (Ong and Teng 2001, )
65
The HOS processing is one time-frequency approach to the detection of LPI signals as
shown in Figure 49. The objective of parallel filter arrays is to separate the input signal
into small frequency bands, providing a complete time-frequency description of the
unknown signal. Then, each sub-band signal is treated individually by a third-order
estimator in order to suppress the noise and preserve the phase of the signal during the
correlation process. Finally, the resulting matrix is entered into a feature extraction
module whose resulting characteristics from the signal are used to determine what type of
modulation was detected.
The detection can be performed without knowing any of the characteristics of the
input signal. The parallel array of filters can approximate the behavior of a matched filter.
The purpose of this filter bank is to separate the observed signal into frequency bands. An
increment in the number of filters in bank also increases the resolution of the system. The
implementation of HOS, particularly third-order cumulant estimators, shows the potential
of the method to suppress white Gaussian noise. The detection method also indicates that
the third-order cumulant of a signal grows out the third-order cumulant of the noise with
increasing SNR (Taboada 2002, 271).
66
The efficiency of the proposed method varies with the modulation used in the LPI
radar signal. High efficiency was gained for detecting and identifying all the parameters
of BPSK and FMCW signals. The detection and identification of polyphase-coded signal
is satisfactory, even though some of the parameters cannot completely be distinguished.
This method also exhibits a good discrimination among different polyphase-coded signals
as Frank, P1, P2, P3 and P4 (Taboada 2002, 271).
3. Wigner Ville Distribution (WVD)
The WVD is a two-dimension function describing the frequency content of a
signal as a function of time. The WVD has been noted as one of the more useful time-
frequency analysis techniques for signal processing. Using the WVD, frequency and time
changes in most of the LPI radar signals can be identified (Gau 2002, 147).
⎛ τ⎞ ⎛ τ⎞
W (t , ω ) = ∫ x ⎜ t + ⎟x* ⎜ t − ⎟ e − jwt dτ (3.3)
⎝ 2⎠ ⎝ 2⎠
where t is the time variable and ω is the frequency variable. The WVD is a two
dimension function describing the frequency content of a signal as a function of time.
This continuous time and frequency representation can be modified for the
discrete sequence x(l), where l is a discrete time index, l=…,-1,0,1,…. The discrete WVD
is defined as
∞
W (l , ω ) = 2∑ x(l + n) x* (l − n)e− j 2 wn . (3.4)
−∞
N
W (l , ω ) = 2 ∑ fl (n)e− j 2 wn , (3.5)
n =− N
67
where
πk
ω= , k = 0,1, 2,..., 2 N -1. (3.7)
2N
j 2π k
⎛ πk ⎞ N −
W ⎜ l, ⎟ = 2 ∑ f l ( n)e . (3.8)
2N
⎝ 2N ⎠ n =− N
Adjusting the limits of n in order to use the standard FFT algorithms, equation (3.8)
becomes
j 2π k
⎛ πk ⎞ 2 N −1 −
W ⎜ l, ⎟
⎝ 2N ⎠
= 2 ∑
n=0
f l
'
( n ) e 2N
. (3.9)
In (3.9) the kernel function has been adjusted to fl ' (n), where
⎧ f l (n), 0 ≤ n ≤ N -1
⎪
fl (n) = ⎨0,
'
n=N (3.10)
⎪ f (n − 2 N ), N + 1 ≤ n ≤ 2 N -1.
⎩ l
68
The resulting WVD is, therefore
2 N −1 jπ kn
−
W ( l , k ) = 2 ∑ f l ' ( n)e N
. (3.11)
n=0
Equation (3.11) is the final WVD equation used to calculate the WVD of the
detected signals (Gau 2002, 147). Figure 50 shows the 2D frequency-time output of an
FMCW signal with 1KHz carrier frequency at the IF band, 250Hz modulation bandwidth
and 20ms modulation period processed after WVD signal processing algorithm.
(a) (b)
Figure 50. Frequency-Time Output of FMCW signal (a) Signal Only (b) SNR=-6dB56
The WVD is a good time-frequency signal processing algorithm for observing the
modulation characteristics of non-stationary signals. It is, however, extremely costly with
respect to computation time. Improvements in ES receiver hardware are helpful but the
analysis computations to extract the detailed modulation parameters are still expensive.
Additionally, the signal must be known for somewhat large periods of time to derive
useful results. For real-time signal processing, an efficient coding algorithm and a very
69
fast digital processor of field programmable gate array (FPGA) are required (Milne and
Pace 2002, IV3944; 3944-7 vol.4; 7o.4).
4. Quadrature Mirror Filter Bank (QMFB)
Complex sinusoids are used by the Fourier transform to perform the analysis of
signals using appropriate basis functions. This approach is difficult since local
information, such as an abrupt change in the signal, is spread out over all frequencies
based on the infinite extension of the Fourier transform. This problem has been addressed
by introducing windowed complex sinusoids as basis functions. This leads to the doubly
indexed windowed Fourier transform:
∞
X WF (ω ,τ ) = ∫e
− jωt
w(t − τ ) x(t )dt , (3.12)
−∞
By varying the window used, resolution in time can be traded for resolution in
frequency. To isolate discontinuities in signals, it is possible to use some basis functions,
which are very short, while longer ones are required to obtain a fine frequency analysis.
The wavelet transform achieves this by obtaining the basis functions from a single
70
1 ⎛ t −b ⎞
hab (t ) = h⎜ ⎟, (3.13)
a ⎝ a ⎠
where is a positive real number and b is a real number. For large a, the basis function
becomes a stretched version of the prototype wavelet (low frequency function) while for
small a, the basis function becomes a contracted wavelet (short high frequency function).
The wavelet transform (WT) is defined as
∞
1 ⎛ t −b ⎞
X W ( a, b) = ∫h
*
⎜ ⎟x(t )dt. (3.14)
a −∞ ⎝ a ⎠
Figure 51 shows the basic two–channel QMFB. Here, the input signal x[n] is first
passed through a two–band analysis filter bank containing the filters, H 0 ( z ) and H1 ( z ),
which typically have lowpass and highpass frequency responses, respectively, determined
by a cutoff frequency π/2.
71
Figure 51. The Two-Channel Quadrature Mirror Filter Bank57
The sub-band signals {vk [n]} are then down-sampled by a factor of 2 in the
“signal analysis section” to be transmitted to the “signal synthesis section.” Here the
signals will be up-sampled by a factor of 2 and passed through a two-band synthesis filter
bank composed of the filters G0 ( z ) and G1 ( z ), whose outputs are then added yielding
y[n]. The analysis and the synthesis filters in the QMFB are chosen to ensure that the
reconstructed output is a reasonable replica of the input x[n].
Some filters, such as the Haar filter, meet the wavelet requirements that perfectly
tile the input energy in time but, unfortunately, does not tile well in frequency. The
opposite of the Haar filter, in this respect, would be the sinc filter. The correct filter is the
“modified sinc filter,” which will return a good tile in time and frequency (Jarpa 2002,
154).
72
(a) (b)
Figure 52. Frequency-Time Output of an FMCW signal (a) Signal Only (b) SNR=-
10dB58
The QMFB tree receiver does a good job of distinguishing the signals in the time
domain. It is very good at picking out when frequency shifts occur and what those
frequencies are. However, the image frequencies that appear can be strong enough to
totally mask the actual signal frequency, making interpretation difficult. While the
receiver’s ability to pick out signals at low SNR appears to be poor, further analysis on
the output matrices can add significant signal detail (Copeland and Pace 2002, IV-3960;
IV-3963 vol.4).
5. Cyclostationary Processing (CP)
The cyclostationary attribute in the periodicities of the second order moments of
the signal can be interpreted in terms of the generation of spectral lines from the signal by
putting the signal through a quadratic non-linear transformation. This property explains
the link between the spectral-line generation property and the statistical property called
“spectral correlation”, corresponding to the correlation that exists between the random
fluctuations of components of the signal residing in distinct spectral bands. The
correlation integral is very important in theoretical and practical applications and may be
defined as
58 (Pace 2004, 455)
73
∞
h( x ) = ∫
−∞
f (u ) g ( x + u )du (3.15)
If f(x) and g(x) are the same function, the integral above is normally called the
autocorrelation function, and called cross-correlation if they differ. The autocorrelation
function is a quadratic transformation of a signal and may be interpreted as a measure of
the predictability of the signal at time t + τ based on knowledge of the signal at time t
(Lima 2002, 162).
T
1 2
⎛ τ ⎞ *⎛ τ⎞
Rx (τ ) lim
T →∞ T ∫ x ⎜ t + 2 ⎟⎠x
−T ⎝
⎜ t − 2 ⎟ dt
⎝ ⎠
(3.17)
2
⎛ T
⎞
⎛ τ ⎞ * ⎛ τ ⎞ − j 2πα t ⎟
2
⎜ 1
Rx (τ ) ⎜ lim ∫ x ⎜ t + ⎟x ⎜ t − ⎟ e
α
dt ⎟ ≠ 0 (3.18)
T →∞ T
⎜ −T ⎝ 2⎠ ⎝ 2⎠ ⎟
⎝ 2 ⎠
where α is the cycle frequency. Rxα (τ ) is the cyclic auto-correlation function, also known
as the “time-frequency limit autocorrelation function”. Since (3.18) is a generalization of
(3.17), when α = 0, the DC component of (3.18) yields the time average autocorrelation
74
function of (3.17). Therefore, the process defined by (3.18) is able to extract more
information from the signal than the process defined by (3.17).
Due to the fact that the power spectrum magnitude may be obtained from the
Fourier transform of the autocorrelation function, the spectral-correlation density (SCD)
or the cyclic-spectral density may also be obtained from the Fourier transform of the
cyclic autocorrelation function (3.18)
∞
1 ⎛ α⎞ ⎛ α⎞
S xα ( f ) ∫R
α
x (τ )e − j 2π T dτ = lim X T ⎜ f + ⎟ X T* ⎜ f − ⎟ (3.19)
−∞
T →∞ T ⎝ 2⎠ ⎝ 2⎠
T
2
∫ x ( u )e
− j 2π fu
XT ( f ) du (3.20)
−T
2
which is the Fourier transform of the time domain signal x(u). The additional variable α
leads to a two-dimensional representation S xα ( f ) which is the bifrequency plane or (f,α)
plane (Lima 2002, 162).
t +∆t
2
1
∆t t −∆∫t TW
α α
S x ( f ) ≈ S xT (t , f ) = S xα (t , u )du, (3.21)
W
where
75
1 ⎛ α⎞ * ⎛ α⎞
S xαT (u , f ) = XT ⎜ u, f + ⎟ X TW ⎜ u, f − ⎟ , (3.22)
W
TW W ⎝ 2⎠ ⎝ 2⎠
and ∆t is the total observation time of the signal, TW is the short-time FFT window
length, and
t +TW
2
X TW (u, f ) = ∫ x (u )e
− j 2π fu
du (3.23)
t −TW
2
is the sliding short-time Fourier transform. Figure 53 shows that, for any signal x(t), the
frequency components are evaluated over a small time window TW along the entire
observation time interval ∆t. The spectral components generated by each short-time
Fourier Transform have a resolution, ∆f=1/ TW . The variable L is the overlapping factor
between each short-time FFT. In order to avoid aliasing and cycle leakage on the
estimates, the value of L is defined as L ≤ TW / 4 (Lima 2002, 162).
76
Figure 54 shows that the spectral components of each short-time FFT are
multiplied for the cyclic-spectrum estimates. Note that the dummy variable u has been
replaced by the time instances t1 ,..., t p . At each window ( TW ), two components centered
on some frequency f 0 and separated by some α 0 are multiplied together and the
resulting sequence of products is then integrated over the total time (∆t).
desired for any given t and, for all f, by making ∆t large. Finally, an illustration of the
relationship between the frequency plane and the bi-frequency plane is shown in Figure
55 (Lima 2002, 162).
77
Figure 55. Bi-frequency and Frequency Plane61
(a) (b)
Figure 56. Bi-Frequency Output FMCW signal (a) Signal Only (b) SNR=-6dB62
61 (Lima 2002, 162)
62 (Pace 2004, 455)
78
Table 7 below summarizes requirements to detect LPI radars. These include
receiver challenges to overcome, ES receiver candidates and their operational examples,
and some of the signal processing algorithms used in ES receivers.
79
THIS PAGE INTENTIONALLY LEFT BLANK
80
IV. CLASSIFICATION AND JAMMING OF LPI RADARS
This task is normally called specific emitter identification (SEI). SEI is a method
of recognizing individual electronic emitters through the precise measurement of selected
signal and characteristics. The problem that arises is that in order to be identified by SEI
techniques, the emitter must have parameters that are stable and unique, within the
measurement capabilities of the ES receiver. For LPI signals, this is typically not the
case, since the signal is on for only a few code periods (Pace 2004, 455).
First a good amount of preprocessing must be done before the NN processed the
image. The NN is trained with different LPI radar signals to recognize the numerous
modulations that might be used by the LPI radar. Once the modulation type is identified,
it is used to select the proper parameter measurement algorithm to process the time-
frequency or bi-frequency output image. After the parameters of the signal are measured,
the results are weighted to select the highest probable signal parameters, and then sorted
into emitter classes by a clustering routine. It is only by directly digitizing the signal at
the antenna, and taking advantage of high-speed parallel processing to run the
sophisticated algorithms, that autonomous classification of LPI emitters can take place
(Pace 2004, 455).
B. NETWORK CENTRIC APPROACH
There are limitations to the use of intercept receivers in a platform-centric
configuration. Geometrical limitations include extended stand-off ranges and alignment
problems, which make it especially difficult to detect and jam LPI emitters. Also, the
intercept receiver is limited by “look through”. The look through process allows the
jammer to observe its effectiveness on the LPI emitter by stopping the jamming
assignment to listen periodically. This results in inefficient jammer management and
limited coordination during a mission.
During the jamming process, a certain amount of look through is required. For
example, with an EA-6B reactively jamming a frequency-hopping radar, the jamming
must stop in order to sense the radar’s transmit frequency. Duty cycle of the intercept
receiver look through process must be less than the time necessary for the radar to sense
it is being jammed, and switch radar parameters such as frequencies. Any amount of look
through is not desired, since this allows the threat radar a window in which to detect the
strike aircraft (Pace 2004, 455).
82
Figure 58. Tip-and-Tune: Solution for the Look Through Problem64
If, however, the EA-6B integrates threat parameters from an electronic order-of-
battle database, a reconnaissance aircraft with near-real time on scene intelligence
collection, analysis, and dissemination capabilities (e.g., Rivet Joint), and frequency data
from an off-board stand-in sensor (e.g., UAV) to cue the on-board intercept receiver (tip
and tune) as shown in Figure 58, a fast reactive electronic attack can be performed that
eliminates the need for look through. For the reactive jamming assignments to be
effective, however, the data link used to provide the cuing data must not induce a delay
time of any significance to the reactive assignment. That is, if the frequency hopping
radar can switch frequencies faster than the cueing data can arrive from the off-board
intercept receiver, then effectiveness is significantly degraded (Pace 2004, 455).
64 (Pace 2006, )
83
Eliminating the limitations inherent in a platform-centric configuration comes
from a distributed system of systems. A distributed system of systems provides
significant geometric flexibility, and can reduce or eliminate the need for look through.
Figure 59 above demonstrates the detection and jamming of an LPI emitter using
a network-centric architecture. The LPI emitter is detected using a number of sensors that
relay the information to both a C2 point, and the proper shooter. C2 allows the shooter to
apply the appropriate electronic attack to disable the LPI emitter. The shooter also relays
65 (Pace 2004, 455)
84
its information concerning the jammed emitter back to the both sensors and C2. That is,
instead of each platform making decisions on information received by only its own
intercept receiver (the platform-centric approach), modern ES receivers integrate
information from many sensors and databases for targeting (Pace 2004, 455).
85
Figure 60. Swarm Intelligence Approach66
1. Probability of Jamming
So far, this thesis has considered the detection and the classification of LPI radar
signals, and not the extraction of any information that would enable the jammers to
exploit the transmissions. If we consider the information that can be extracted from the
waveform, then we can examine the potential to further exploit the LPI radar signal
(Stove, Hume, and Baker 2004, 249-260).
where C is the channel capacity in bits/second, W is its bandwidth, and SNR is the signal
to noise ratio. In this case, if W is set to 1 then C becomes the capacity in bits per
66 (Pace 2006, )
86
interception. For the detection threshold level of 17dB SNR, each interception may be
assumed to provide the opportunity for extracting 5.7 bits of information concerning the
LPI radar. It can be assumed that in order to exploit transmissions, the following
information is needed:
• Scan timing, i.e. where the radar is pointing at any time,
• Carrier frequency,
• Modulation bandwidth,
• Modulation or code period,
• Synchronization, i.e. when the modulation pattern starts.
If each of these parameters must be known to 4 bits precision, i.e. slightly better
than 10% accuracy, 20 bits of information is needed to characterize the radar, not
including trying to replicate its waveform in any detail. The scan timing and carrier
frequency can readily be discovered from the way in which intercept is made.
Information on the LPI radar would effectively be obtained from multiple looks at the
receiver output. After this process 12 bits giving 36dB SNR of information has to be
recovered from the signal. In some conventional ES receivers this sensitivity is achieved
through integrating multiple looks, using a receiver with a wider bandwidth than the
signal’s information bandwidth (Stove, Hume, and Baker 2004, 249-260).
The problem with attempting to match the jamming receiver for a single look
interception and exploitation is that it is more efficient in energy terms to obtain
information from separate looks at lower SNR. For example, to send 8 bits of information
in one go requires 24dB SNR in the channel’s information bandwidth. To send it in two
4-bit messages requires twice as much time but only 12dB SNR, a saving of 9dB in the
amount of energy used.
It should be noted in passing that the true matched filter removes all the
modulation information from the signal, leaving only the information about the energy
spectrum. It can be a fundamental fact that optimizing the detection sensitivity involves
removing as much as possible of the information bearing capacity of the original
waveform, by whatever means the filtering is achieved (Stove, Hume, and Baker 2004,
249-260).
87
2. Sensitivities Required for Jamming
The process of coding the information onto the radar waveform and decoding it in
the receiver is 6dB less efficient than a typical communication channel, which may itself
be assumed to be 6dB less efficient than the Shannon limit. Therefore, 48dB SNR is
needed to recover the required information from the signal. This is approximately
equivalent to removing the 30dB increase in sensitivity obtained by going from an IFM
type receiver to a matched ES receiver. It can be hypothesized that the channelized
receiver achieves an intermediate exploitation performance by being less lossy than the
IFM type receiver in recovering the information, but that it will require additional SNR to
stitch together the outputs of the different channels to recover all the information required
(Stove, Hume, and Baker 2004, 249-260).
The results derived above show that it is possible to exploit the radar’s
transmissions. The simple radiometric detector is able to cope with any waveform, but at
the cost of destroying most of the information contained within it. This makes it
unsuitable for use in a busy environment, but it can be useful during normal ‘radar
silence’ when very few emitters will be present. In fact, in busy environments, it can be
argued that the best way of transmitting covertly is to make the transmissions look like a
commercial radar, such as a conventional marine radar or aircraft weather radar. Thus,
they may not be noticed. However, if an LPI waveform is detected, it is clearly not
coming from an innocent source (Stove, Hume, and Baker 2004, 249-260).
3. LPI Radar Jammer Design Requirements
Two parameters are of critical importance when considering the design of an LPI
receiver front-end for EA applications. First one must consider the RF bandwidth; too
wide a bandwidth allows too much signal to enter the detector and unnecessarily
degrades the receiver noise figure while too narrow a bandwidth eliminates too much of
the signal lowering the average power to the detector.
The second factor to be considered is the detector video bandwidth. A wide video
bandwidth provides for fast rise and fall times necessary for processing narrow pulses,
but this is done at the expense of allowing more noise to the detector as well. A carefully
designed jammer will address both of these design areas (McRitchie and McDonald
1999, ).
88
a. RF Bandwidth
In order to define the RF bandwidth requirements for a jammer, the
dynamics of the radar/target engagement must be considered. A limiting factor in this
discussion is the time-bandwidth product of the radar.
T1 given by:
∆R = T1.ν r . (4.2)
For coherent integration, the target must not move out of the radar range cell during the
integration time, so that:
c.τ c
∆R ≤ ≅ (4.3)
2 2.B
89
1msec < T1 < 10msec
20MHz < B < 500MHz
90
significant suppression of many interfering waveforms that are uncorrelated. (such as
narrowband interference and pulsed radar emissions)
There are some basic approaches for jamming FMCW radars. One approach is to
predict the frequency-versus-time characteristics of the signal and use a jammer that will
input energy to the receiver at the same frequency as the FM signal it is attempting to
receive. This will allow the maximum jammer-to-signal ratio (JSR) to be achieved for
any given jammer power and jamming geometry (Adamy 2001, ).
Another approach is to cover all or part of the modulation range with a broadband
jamming signal that is received by the LPI radar receiver with adequate power to create
adequate JSR in the “de-chirped” output. As shown in Figure 61, the FM modulated
signal has an anti-jam advantage equal to the ratio between its information bandwidth and
the frequency range over which it is modulated.
91
Figure 61. Anti-jam Advantage of FMCW Radar Signal67
67 (Adamy 2001, ) Unless a jammer can be swept in synchronization with the signal modulation rate,
a jammer’s power must be spread across the whole swept range. The ratio of the modulation sweep range to
the information bandwidth is the antijam advantage.
92
signal will affect the integration efficiency, in order to be effective, the VGPO must
create at least 180° phase shift over the signal (McRitchie and McDonald 1999, ).
Narrow band Doppler noise may also be quite effective since the SNR in the LPI
receiver is already at quite a low value. A Digital RF Memory (DRFM) can be used to
focus the available power of the jammer and inject Doppler noise that is only a few KHz
wide, matching to the instantaneous bandwidth of the FMCW radar (McRitchie and
McDonald 1999, ).
5. Jamming PSK Radars
These phase coded signals can be affected by VGPO type techniques. The
introduction of an additional Doppler shift, which will be interpreted by the radar as an
additional phase shift, will cause a spreading of the received signal and therefore decrease
the effective processing gain. If enough Doppler shift can be added then there will be a
corresponding loss of integration gain within the radar processor (McRitchie and
McDonald 1999, ).
Range bin masking techniques should also be quite effective. If a section of the
radar waveform recorded by DRFM or repeater is used by the jammer as it’s transmit
waveform, the truncation will cause an increase in the sidelobe levels of the processed
return. The merging of the sidelobes can create a threshold problem for the radar leading
to total signal loss; but at the very least will degrade the SNR of the true target return
causing a loss in processing gain (McRitchie and McDonald 1999, ).
6. Jamming FSK Radars
FSK radars are said to have an anti-jam advantage as seen in Figure 62. This
advantage is based on the assumption that the jammer knows only the full hopping range
and must spread its jamming power over that full frequency range. Assume an FSK radar
that has a 2000 frequency hopping sequence which is random or unknown to the ES
receiver. The FSK radar can be said to have a jamming advantage of 2000, which
converts to 33dB. This means that it takes 33dB more jammer power to achieve a given
JSR against this frequency hopper than would be required if it were a fixed-frequency
conventional radar.
93
Figure 62. Anti-jam Advantage of FSK Radar Signal68
One jamming approach that can be used for this kind of radar is to perform
“follower jamming”. A follower jammer detects the frequency of each hop and then jams
on that frequency. This solution requires an extremely fast frequency measurement
technique in order to deny the enemy the transmitted information in each hop (Adamy
2001, ). Furthermore, the frequency hopping sequence of FSK radar is unknown and
appears random to the ES receiver. Unless this frequency sequence is solved, the
possibility of a jammer following the changes in frequency is very remote.
68 (Adamy 2001, ) In order to jam a FSK radar signal with a conventional jammer, the jammer’s
power must be spread across the whole hopping range. The ratio of the hopping range to the information
bandwidth is the antijam advantage.
94
V. CONCLUSION
The signal environment is changing at a rapid pace with new Low Probability of
Intercept (LPI) radars coming into service worldwide. These radars exhibit lower power
and higher duty cycle than previous radar technology. By 2010, approximately 30% of all
radars will emit LPI signals and will be employed in all classes of radar including
battlefield, navigation, surveillance, target acquisition and missile seekers on airborne,
maritime, and land-based platforms (Tenix Defense 2005, 2).
Airborne LPI radars are used for search, tracking, location, identification,
acquisition, designation, target imaging, periscope detection, and weapon delivery. These
LPI radars also have modes for covert navigation, weather detection, terrain following,
and terrain avoidance. Just as LPI techniques are useful for covert navigation and
targeting in airborne applications, they are equally useful for maritime applications. LPI
is well suited for this environment as the relatively slow speed of a ship and extremely
large radar cross sections (RCSs) allows for long integration times. Besides maritime
applications, there are many examples of land-based LPI radar generally performing
ground surveillance and short range air surveillance. In the case of the ground
surveillance role, these radars can be used to covertly detect ground targets due to long
integration times.
Electronic Warfare Support (ES) receivers currently in service are not optimized
for the detection of LPI radars as they lack the sensitivity to detect the signals at
sufficient range to provide military crews with an operational range advantage. LPI
Radars use advanced radar and signal processing techniques “to see and not be seen”69
by ES receivers. To survive Electronic Attack (EA) and Anti Radiation Missile (ARM)
threats and mask their presence, LPI radars use:
• Low sidelobe antennas,
• Irregular antenna scan patterns,
• High duty cycle/wide band transmission,
• Accurate power management,
95
• Carrier frequencies at peak atmospheric absorption,
• Very high sensitivity,
• High processing gain,
• Coherent detection,
• And monostatic/bistatic configurations.
There are several LPI radar techniques available to the modern radar designer that
can be used singly or in various combinations, depending on the application. Reducing
the radar’s peak effective radiated power (ERP) by using some form of pulse
compression technique is the most common LPI radar technique. The objective is to
spread the radar’s signal over a wide bandwidth and a period of time. This is typically
done with frequency modulation (FM), which is the most common technique, phase shift
keying (PSK), and frequency shift keying (FSK) techniques.
In order to jam LPI radar signals the ES receivers and associated EA systems
must accomplish specific steps. First and most important is LPI radar detection. Without
detection no countermeasures are possible. To detect LPI radar signals, ES receivers have
to overcome three main difficulties. These are:
• LPI radar’s coherent integration,
• High sensitivity requirement,
• And processing gain of the LPI radar.
LPI radars have low power, high duty cycle signals with phase or frequency
coding. As the coding is unknown and can be complex, and assuming the frequency is
also unknown, then coherent detection is not possible and non-coherent detection must be
performed first. To achieve the maximum sensitivity the RF and video bandwidth must
be matched to the signal modulation, allowing detection of the total signal energy (Rayit
and Mardia 1994, 359; 359-362; 362). Detection of LPI radar signals also requires a large
processing gain because of the wideband nature of the LPI radar. Detection is possible if
the signal is integrated over a long observation time.
Detection of LPI radar signals also requires sophisticated receivers that use time
frequency signal processing, correlation techniques and algorithms to overcome the
processing gain of the LPI radar. Fourier analysis techniques have been used as the basic
tool. From this basic tool, more complex signal processing techniques have evolved, such
96
as the short-time Fourier transform (STFT), so as to track signal parameters over time.
More sophisticated time-frequency and bi-frequency distribution techniques have also
been developed to identify the different modulation schemes used by LPI radars. These
techniques include the Wigner Ville Distribution (WVD), Quadrature Mirror Filter Bank
(QMFB), and Cyclostationary Processing (CP) (Pace 2004, 455).
These signal processing algorithms require large amounts of computing speed and
memory. Managing processing speed is not a problem with current digital capabilities,
but carrying enormous amounts of data is still difficult. Increasing the sensitivity of the
receiver allows for detecting sidelobes of the emitter, but at the same time obligates the
receiver to process a significantly large number of signals.
The trend in ES receivers for LPI radar detection is toward digital receivers and
incorporates the concept of digital antennas in which the analog-to-digital converter
(ADC) is at the antenna. The future digital receiver will incorporate optical technologies
for speed and bandwidth, and will also incorporate high-temperature superconductors for
required sensitivities (Pace 2006, ).
Once the detection hurdle has been overcome, the ES receiver must next perform
classification. Classification requires sorting the signal into groups having similar
parameters. These parameters are;
• LPI radar type,
• Carrier frequency,
• Modulation bandwidth,
• Modulation or code period,
97
• Scan timing, i.e. where the radar is pointing at any time,
• And synchronization, i.e., when the modulation pattern starts.
These parameters distinguish one LPI radar signal from another and they are
required for effective jamming. Correlation with existing signals in a database, which is
called identification, can then aid in signal tracking and response management.
The NN is trained with different LPI radar signals to recognize the numerous
modulations that might be used by the LPI radar. Once the modulation type is identified,
it is then used to select the proper parameter measurement algorithm to process the time-
frequency or bi-frequency output image (Pace 2004, 455). Adaptive Resonance Theory
(ART) is a very strong candidate for NN and has pattern recognition algorithms that the
time-frequency or bi-frequency output images of LPI radar signals can be processed and
classified effectively.
98
For FMCW LPI radars, noise techniques such as white Gaussian noise and
continuous wideband noise are not effective, and FMCW signals can be recovered in
such moderate noise conditions (Fu and Ke 1996, 605). On the other hand; if the
modulation period tm and modulation bandwidth ∆F can be determined, then coherent
deception jamming is feasible and very effective, since the jammer waveform looks like
the radar waveform (Pace 2004, 455). False range targets, velocity gate pull-off, and
narrow band Doppler noise created by using digital RF memory (DRFM) are the coherent
EA techniques which can be used effectively against FMCW LPI radars.
Against phase coded LPI radars, noise type EA techniques are also not effective.
Phase coded LPI radars can be affected by VGPO type techniques. The introduction of an
additional Doppler shift, which will be interpreted by the radar as an additional phase
shift, will cause a spreading of the received signal and therefore decrease the effective
processing gain. If enough Doppler shift can be added then there will be a corresponding
loss of integration gain within the radar processor (McRitchie and McDonald 1999, ).
Range bin masking techniques should also be quite effective against phase coded radars.
If a section of the radar waveform recorded by either DRFM or a repeater is used by the
jammer as it’s transmit waveform, the truncation will cause an increase in the sidelobe
levels of the processed return. The merging of the sidelobes can create a threshold
problem for the radar leading to a blank display, and at the very least will degrade the
SNR of the true target return, causing a loss in processing gain (McRitchie and
McDonald 1999, ).
For FSK radars ”follower jamming” can be used. A follower jammer detects the
frequency of each hop and then jams on that frequency. But this solution requires an
extremely fast frequency measurement technique in order to deny the enemy the
transmitted information in each hop (Adamy 2001, ). Furthermore, the frequency hopping
sequence of the FSK radar is unknown and appears random to the ES receiver. Unless
this frequency sequence is solved, the possibility of the jammer following the changes in
frequency is very remote.
99
alignment problems, which make it especially difficult to detect and jam LPI radars.
Also, the ES receiver is limited by “jamming look through”. Eliminating the limitations
inherent in a platform-centric configuration is accomplished by a distributed system of
systems. A distributed system of systems provides significant geometric flexibility, and
can reduce or eliminate the need for look through.
While modern LPI radar systems and waveforms present formidable challenges to
older and presently deployed ES receivers, there are techniques and technologies
available on the near-horizon equipped to meet this challenge. Digital receivers with high
sensitivities, time-frequency and bi-frequency signal processing techniques with high
processing gains working in the neural networks will overcome the LPI radar’s signal
masking and hiding techniques. Furthermore, using miniaturized receivers at UAVs with
network centric and swarm intercept strategies will carry LPI radar and ES receiver battle
into a different dimension.
100
LIST OF REFERENCES
Adamy, David L. 2001. EW 101: A first course in electronic warfare. Norwood, MA:
Artech House.
Bonabeau, E., Dorigo M, and G. Theraulaz. 1999. Swarm intelligence from natural to
artificial systems. New York: Oxford University Press.
Burrus, S., A. Gopinath, and H. Guo. 1998. Introduction to wavelets and wavelets
transforms. First Edition ed.Prentice Hall.
Copeland, D. B., and P. E. Pace. 2002. Detection and analysis of FMCW and P-4
polyphase LPI waveforms using quadrature mirror filter trees.
Donohoe, J. P., and F. M. Ingels. 1990. The ambiguity properties of FSK/PSK signals.
Fu, J. S., and Y. Ke. 1996. Anti-jamming aspects of linear FM and phase coded pulse
compression by simulation. Paper presented at CIE International Conference of
Radar, .
Fuller, K. L. 1990. To see and not be seen [radar]. Radar and Signal Processing, IEEE
Proceedings F 137, (1): 1-10.
Gau, Jen-Yu. 2002. Analysis of low probability of intercept (LPI) radar signals using the
wigner distribution. M.S. in Systems Engineering., Naval Postgraduate School.
GuoSui Liu, Hong Gu, WeiMin Su, and HongBo Sun. 2001. The analysis and design of
modern low probability of intercept radar. 2001 CIE International Conference on
Radar Proceedings (Cat no.01TH8559) (2001): 120; 120-124; 124.
Hou Jiangang, Tao Ran, Shan Tao, and Qi Lin. 2004. A novel LPI radar signal based on
hyperbolic frequency hopping combined with barker phase code. 2004 7th
International Conference on Signal Processing Proceedings (IEEE
Cat.no.04TH8739) (2004): 2070; 2070-3 vol.3; 3o.3.
Jane's Air Defense Radar – Land and Sea. 1997. Section 9 – product descriptions (united
states). (February 1997).
101
Jane's Air-Lauched Weapons. 2002. Air to surface missiles, RBS-15F. (September 12,
2002).
Jane's C4I Systems. 2001. Intelligence systems – surveillance and reconnaissance. (April
06, 2001).
Jane's International Defense Review. 1994. Radar eyes on the battlefield. (October 01,
1994).
Jane's Land Based Air Defense. 2004. Self-propelled anti-aircraft guns (netherlands).
(August 17, 2004).
JANE'S NAVY INTERNATIONAL. 2004. Radar ESM faces up to the littoral challenge
(SEPTEMBER 01, 2004).
Jane's Police and Security Equipment. 2005. Surveillance equipment. (December 29,
2005).
Jane's Radar and Electronic Warfare Systems. 2006a. Battlefield, missile control and
ground surveillance radar systems. (February 22, 2006).
———. 2006b. Naval/Coastal surveillance and navigation radars. (February 22, 2006).
———. 2005. Airborne surveillance, maritime patrol and navigation. (August 11, 2005).
———. 2004a. Airborne fire-control radars, united states. (March 10, 2004).
———. 2004b. Airborne fire-control radars, united states. (October 22, 2004).
———. 2004c. Airborne fire-control radars, united states. (December 03, 2004).
———. 2004d. Battlefield, missile control and ground surveillance radar systems.
(October 22, 2004).
———. 2004e. Battlefield, missile control and ground surveillance radar systems.
(October 26, 2004).
———. 2003. Battlefield, missile control and ground surveillance radar systems. (March
25, 2003).
Jarpa, Pedro F. 2002. Quantifying the differences in low probability of intercept radar
waveforms using quadrature mirror filtering. M.S.: Electrical Engineering., Naval
Postgraduate School.
102
Klein, L. A. 1997. Milimeter-wave and infrared multisensor design and signal
processing. Norwood, MA: Artech House, INC.
Lewis, Bernard. 1986. Aspects of radar signal processing. Norwood, MA: Artech House,
INC.
Lima, Antonio F. 2002. Analysis of low probability of intercept (LPI) radar signals using
cyclostationary processing. M.S. in Systems Engineering., Naval Postgraduate
School.
McRitchie, Dr W. K., and S. E. McDonald. 1999. Detection and jamming of LPI radars.
Canada: MC Countermeasures INC, W7714-7-0133/003/SV.
Milne, P. R., and P. E. Pace. 2002. Wigner distribution detection and analysis of FMCW
and P-4 polyphase LPI waveforms. 2002 IEEE International Conference on
Acoustics, Speech, and Signal Processing.Proceedings (Cat.no.02CH37334) (2002):
IV3944; 3944-7 vol.4; 7o.4.
Ong, Peng G., and Teng, Haw K. Digital LPI radar detector.2001.
Pace, Philip E. 2006. Introduction to LPI radar, EC4690 class notes. Class Notes ed.
Monterey: .
Pace, Phillip E. 2004. Detecting and classifying low probability of intercept radar.
Artech house radar library. Boston: Artech House.
Rayit, R. K., and H. K. Mardia. 1994. Detection of LPI radars using an adaptive front-
end. Microwaves 94.the Applications of RF.Microwave and Millimetre Ware
Technologies.Conference Proceedings (1994): 359; 359-362; 362.
Ruffe, L. I., and G. F. Stott. 1992. LPI considerations for surveillance radars.
International Conference Radar 92 (Conf.Publ.no.365) (1992): 200; 200-202; 202.
Schleher, D. Curtis. 1986. Introduction to electronic warfare. Artech house radar library.
SPG Media. SAAB BOFORS DYNAMICS. [cited April 16, 2006]. Available from
http://www.naval-
technology.com/contractors/weapon_control/celsiustech/celsiustech2.html.
103
Stove, A. G., A. L. Hume, and C. J. Baker. 2004. Low probability of intercept radar
strategies. Iee Proceedings-Radar Sonar and Navigation 151, (5) (OCT): 249-260.
Tenix Defense. High sensitivity microwave receiver. in Tenix Defense Pty Ltd [database
online]. Australia, 2005 [cited November 6, 2006]. Available from
http://www.tenix.com/Main.asp?ID=849.
104
INITIAL DISTRIBUTION LIST
3. Chairman
Information Sciences Department
Monterey, California
105