Fundamentals of Radar Systems D.

Lawrence

Radar History
The term “radar” was originally an acronym for “radio detection and ranging” coined in 1940 by the
U.S. Navy for secrecy. Today the use of radar has become widespread including its familiar use in law
enforcement, weather sensing, and a wide array of military applications. The development of radar has
a rich and varied history with developments being made independently across the globe. It is generally
accepted that the history of radar began in the later part of the 19 th century with classical experiments
performed by Heinrich Hertz who helped validate predictions of electromagnetic theory published by
James Clerk Maxwell. Hertz demonstrated the basic reflection of electromagnetic waves from metallic
objects and refraction by dielectrics. In 1904, the German engineer Hulsmeyer demonstrated the use of
radio waves in a device to detect the presence of ships up to 3km, but the device lacked any ranging
information. The device was called a “telemobiloscope.” Despite its success, widespread development
of radar did not occur until decades later. A significant event that helped spread interest in radar was a
paper and speech presented to the Institute of Radio Engineers (now IEEE) by Marconi, considered the
father of the radio telegraph, in which he promoted the potential of the novel concept,

I also described tests carried out in transmitting a beam of reflected waves across country . . . and pointed
out the possibility of the utility of such a system if applied to lighthouses and lightships, so as to enable
vessels in foggy weather to locate dangerous points around the coasts...

It [now] seems to me that it should be possible to design [an] apparatus by means of which a ship could
radiate or project a divergent beam of these rays in any desired direction, which rays, if coming across a
metallic object, such as another steamer or ship, would be reflected back to a receiver screened from the
local transmitter on the sending ship, and thereby immediately reveal the presence and bearing of the other
ship in fog or thick weather.
Guglielmo Marconi, 1922

In the middle to late 1930s, the development of radar was accelerated with largely independent
developments in many countries around the world. The appearance of all-metal bomber aircraft
provided an urgent need for a warning device that could provide target detection and ranging
information. Early radars used relatively low frequencies in the HF and VHF bands from 10’s of MHz to
several 100’s of MHz. World War II began in 1939 and spurned the rapid development of many
supporting radar technologies. One such invention by the British was the high-power magnetron which
allowed radar operation at microwave frequencies. Interestingly, this invention was disclosed to MIT
Radiation Laboratory and Bell Telephone Laboratories in 1940 to speed the development of microwave
radar.

Figure 1. “Telemobiloscope” demonstrated by Hulsmeyer in 1904 to detect the presence of ships.

1
Fundamentals of Radar Systems D. Lawrence

Many early radar systems were not utilized to their full extent largely due to poor operator training
and lack of an adequate command and control system. A classic example is the 100 MHz SCR-270 radar
deployed on the Island of Hawaii on December 7, 1941 which detected the incoming Japanese raid on
Pearl Harbor almost an hour before the attack. The radar detections were tragically ignored because of
inexperience with the technology and poor communication to the command.

Following the end of WWII, radar continued to mature as new advanced technologies constantly
emerged. Even today, novel technologies and new applications of existing radar system concepts are
being developed. The military uses radar systems for surveillance, navigation, and weapons guidance for
many sea, air, and ground vehicles. While the military remains the largest developer of radar
technologies, many civilian applications have become more prevalent in areas such as weather radar, air
traffic control, police radar, and imaging applications.

Abbreviated Timeline:

1864 James Clerk Maxwell – EM Field Theory Published

1885 – 1888 Heinrich Hertz experimentally verified radio waves (~455 MHz) reflected from metallic
objects.

1922 Marconi – early pioneer of wireless radio, observed radio detection of targets during
experiments, presented famous speech to Institute of Radio Engineers (now IEEE).

1922 Taylor and Young – U.S. Naval Aircraft Radio Laboratory demonstrated detection of ship
in the Potomac River using bistatic transmitter and receiver arrangement.

1930s Independent development of radar by a number of countries.

1939 World War II begins. Rapid development of radar supporting technology.

1940 British invention of the high-power magnetron allowed microwave frequency radar
operation. Disclosed invention to U.S.

1940-present Post War Developments:

 High Power Amplifiers (Klystron, TWT, solid-state transmitters)

 Monopulse/Interferometer Angle Measurement

 Pulse Compression

 Synthetic Aperture Radar (SAR)

 Phased Array Antennas

 Target Recognition

 and many more…

2
Fundamentals of Radar Systems D. Lawrence

The Decibel
The decibel is a mathematical tool used to express power ratios. The advantage of the decibel is that it
is based upon a logarithmic function which allows a large range of power values to be expressed in a
compressed scale. Conversion of a linear power ratio to decibels is accomplished by

P 
Power Ratio in dB  10  log10  2 
 P1 

Beyond providing a compressed scale, decibels also allow multiplication and division to be performed by
simple addition and subtraction. The table below shows the power ratio for the integer decibel values.
The table also includes an approximate value for the decibels which can be useful for “back of the
envelope” calculations. Since it is a power ratio, values expressed in dB do not have units. However, the
dB scale is often used to express quantities of power, area, and antenna gain by referencing standard
units. For example, dBw is relative to 1 Watt, dBm is relative to 1 mWatt, dBsm is relative to 1 m 2, and
dBi is relative to an isotropic antenna.

Table 1. dB Tables

dB Power Ratio Power Ratio (Approx.)

0 1 1
1 1.2589 1.25
2 1.5849 1.5
3 1.9953 2
4 2.5119 2.5
5 3.1623 3 ()
6 3.9811 4
7 5.0119 5
8 6.3096 6
8.5 7.0795 7
9 7.9433 8
9.5 8.9125 9
10 10.000 10

Example calculations where the upper line is linear multiplication, and the bottom line is addition in dB:

4  7  28 100  5  500 5  1 / 4  1.25

1) 2) 3)
6  8.5  14.5 dB 20  7  27 dB 7  6  1 dB

4)
4    12.6
5)
4 3  1984 6)
8  0.01  0.08
6  5  11 dB 6  5 3  33 dB 9  20  11 dB

3
Fundamentals of Radar Systems D. Lawrence

Radar Frequencies
Radar frequencies vary from a few MHz to hundreds of GHz depending upon the application and
available components. The table below lists band designations for the commonly used frequencies. In
general, low frequencies have lower attenuation but require larger antennas for transmission. Higher
frequencies have higher attenuation but can be generated with smaller antennas. Components become
more expensive and harder to obtain at higher frequencies. Exceptions to this occur in bands where
commercial applications, such as the cell phone industry, have supplied the market with a large quantity
of components. It is also more difficult, and more expensive, to generate high powers at the higher
frequencies. During radar design, trade studies are often performed between available real estate, cost,
and performance when choosing the optimal radar frequency.

Table 2. Traditional Radar Band Designations

Band Frequency Wavelength

L 1 – 2 GHz 30 – 15 cm

S 2 – 4 GHz 15 – 7.5 cm

C 4 – 8 GHz 7.5 – 3.75 cm

X 8 – 12 GHz 3.75 – 2.5 cm

Ku 12 – 18 GHz 2.5 – 1.67 cm

K 18 – 27 GHz 1.67 – 1.11 cm

Ka 27 – 40 GHz 1.11 – 0.75 cm

V 40 – 75 GHz 7.5 – 4.0 mm

W 75 – 110 GHz 4.0 – 2.7 mm

The relationship between frequency, f, and wavelength, , is given by

c

f

where c is the speed of the wave. In free-space, c = 2.99792458x108 m/s and is usually approximated as
simply 3x108 unless very high accuracy is required. If the wave is traveling in a dielectric medium in
which the speed is slower, then the wavelength is reduced accordingly.

4
Fundamentals of Radar Systems D. Lawrence

Basic Radar Concepts

Radar uses electromagnetic phenomenon to detect, track, and/or image objects in the surrounding
environment. Electromagnetic energy is radiated from a source and reflected (or scattered) from target.
Reflected energy in the direction of the radar receiver can be processed to detect the presence of the
target and extract potentially useful information about the target’s position, speed, or identifying
geometric features. Radar’s primary advantage over other sensors such as optical and infrared devices is
the ability to operate in all weather, day or night.

Fundamental radar operation can be classified into three basic functions: 1) target detection,
2) tracking, and 3) imaging. At its most basic level, a radar system consists of a transmitter, receiver, and
antenna(s). A monostatic radar system is one in which the transmitter and receiver are in the same
physical location often sharing the same antenna for transmit and receive, while a bistatic radar has the
transmitter and receiver in different locations. Range is calculated from the round-trip time-delay to the
target. The range to the target is related to the speed of the wave and the time-delay, td1, and the range
from the target to the receiver is related to the time-delay, td2.

Transmitter-to-Target: R1  c t d1

Target-to-Receiver: R2  c t d 2

For the monostatic case, R1 = R2 = R. The target range is expressed in terms of the total time-delay,

R1  R2  c t d 1  t d 2 
2R  c td

c td
R (monostatic range to target)
2

*A useful rule to remember is c/2 = 150 m/sec.

Figure 2. Basic radar configuration.

5
Fundamentals of Radar Systems D. Lawrence

Pulsed Waveform

Perhaps the most common radar waveform consists of a series of rectangular pulses of short duration
modulating a sinusoidal carrier. A single transmit pulse having a carrier frequency, 0, is written

s(t )  P(t ) cos0t 

where
A , 0  t  
P (t )  
 0 otherwise

A series of N pulses separated by the pulse repetition interval, T, is written mathematically as

N 1
stx (t )   s t  nT 
n0
N 1
  Pt  nT cos 0 t  nT 
n0
N 1
  Pt  nT cos 0 t  n 0T 
n0

Fundamental pulsed waveform parameters:

 Carrier Frequency: 0  2f 0 (0 units are rads/sec, and f0 units are Hz)

 Carrier Wavelength: c
0 
f0
1 2
 Peak Power: Ppeak  A
2

 Pulse Width: 

 Pulse Repetition Interval: PRI = T

1
 Pulse Repetition Frequency: PRF 
T

 Duty Cycle/Factor: duty 
T

6
Fundamentals of Radar Systems D. Lawrence

Figure 3. Pulsed waveform example with peak amplitude = A, pulsewidth = , and PRI = T.

Average Power / Energy

Assuming that the pulsewidth is much larger than the period of the RF carrier signal,   1 f 0 , which is
almost always the case, the average power in the pulsed waveform can be calculated

T
1
Pavg   s 2 (t ) dt
T0
T
1
P 2 (t ) cos 2  0t  dt
T 0

Integrates to 0 when >> 1/f0

1 1 1 
  A 2   cos2 0t  dt
T 0 2 2 

1 1 2 1 
 
T 02
A dt  A 2 
2 T


Pavg  Ppeak   Ppeak  (Duty Cycle)
T

As will be shown later, the detection performance of a radar depends on the total energy in the
waveform rather than just the power. The energy in a pulsed waveform is calculated by multiplying the
power by the duration. The energy in a single pulse can be calculated using peak power and the
pulsewidth or average power and the PRI,

E  Ppeak    Pavg  T

7
Fundamentals of Radar Systems D. Lawrence

Unambiguous Range

For a pulsed radar waveform, there is a maximum target delay that allows the receiver to determine the
target range without ambiguity. A range ambiguity occurs when the target return from one pulse is
received after transmitting the subsequent pulse. Figure 4 shows the nominal case when the target
return arrives before the next pulse is transmitted. In this scenario the target range is easily determined
from the received pulse delay. Figure 5 shows the case where the target return appears after
transmitting the subsequent pulse. A range ambiguity results because there is uncertainty as to which
transmit pulse corresponds to the received pulse. The range corresponding to the maximum delay in
which the target range can be determined unambiguously is related to the PRI,

cT
Runambiguous 
2

Figure 4. Pulsed waveform illustrating sequential transmit pulses and receive pulses where the target
return from Pulse #1 arrives before transmitting Pulse #2. There is no ambiguity in calculating target
range.

Figure 5. Pulsed waveform where target return from Pulse #1 arrives after transmitting Pulse #2. In this
condition a range ambiguity results.

8
Fundamentals of Radar Systems D. Lawrence

Note that when the delay from the target return is within a pulsewidth of the PRI, there will be some
overlap with the next transmit pulse. In this condition, the target is said to be eclipsed. When the target
range exactly equals the unambiguous range, the target pulse is totally eclipsed. If there is only partial
overlap between the target return pulse and the next transmit pulse, the target pulse is partially
eclipsed. Many pulsed radar systems are unable to receive a target return while transmitting a pulse
(this fact depends upon the particular radar architecture and transmit/receive isolation), and the
eclipsed region is unusable. The target return will be partially or totally eclipsed by the subsequent pulse
when it falls within the following range extent

c c
Eclipsed Region: Runambiguous   R  Runambiguous 
2 2

Folded Clutter Return

In addition to targets, clutter return can also be ambiguous. Since ground clutter is usually distributed
over a long range, clutter return from past pulses can “fold” and appear at the same range as the target.
An airborne radar scenario is illustrated in Figure 6 where a downward looking radar detects the
presence of a target in the background of distributed clutter. The single transmit pulse produces both a
target return and a long trail of clutter return, including an increase due to the mountains in the
distance. Figure 7 demonstrates the effect of two transmit pulses. The second pulse is transmitted
before the target return is received, thus creating a range ambiguous target. Also, note that the clutter
return from the second pulse is received in addition to the delayed return from the previous pulse. In
this case, the original target return is much lower than the close-in clutter received from the second
pulse. When many pulses are transmitted, the cumulative effect of clutter folding into the current PRI is
shown in Figure 8. Here, the steady-state return from a given PRI shows the ambiguous target return
and clutter folded from many previous pulses. Doppler processing can often be used to separate a
moving target from stationary clutter, provided the residual clutter phase noise is not too large.

9
Fundamentals of Radar Systems D. Lawrence

Figure 6. Single pulse signal return from a target in the presence of distributed clutter.

Figure 7. Two pulses with ambiguous target return and “folded” clutter.

Figure 8. Steady-state return for a pulsed waveform, ambiguous target, and folded clutter.

10
Fundamentals of Radar Systems D. Lawrence

Range Resolution

In a pulsed radar system, the length of the pulsewidth determines the ability to separate two targets in
range. When the returns from two targets are separated by more than , the targets are capable of
being resolved in range as illustrated in Figure 9. When two target returns are separated less than , the
two targets cannot be resolved, see Figure 10. The range resolution of a pulsed radar is defined to be

c
Rresolution 
2

Note that the range resolution is not the same as the range accuracy, although they are often related.

Figure 9. Two targets separated by more than a pulsewidth are easily resolved in range.

Figure 10. Two targets separated by less than a pulsewidth cannot be resolved.

11
Fundamentals of Radar Systems D. Lawrence

Figure 11. Small transmit power over long pulse duration has poor range resolution.

Figure 12. Large transmit power over short pulse duration has good range resolution.

Radar detection performance is proportional to the energy in the waveform. Waveforms having the
same energy can in theory produce the same detection performance. For a simple pulsed waveform, the
same energy can be achieved by a small transmit power over a long pulse or a large transmit power over
a short pulse.

Energy  Psmall   long  Pbig   short

The short pulse provides better range resolution but requires a larger peak transmit power which may
be unrealizable. The long pulse is beneficial since the required peak transmit power is low, but it suffers
from poor range resolution. The short pulse has a much wider bandwidth than the long pulse and leads
to the basic relationship.

More Bandwidth = Smaller Range Resolution

It will be shown later that the long pulse can be modulated to give wider bandwidth, and thus smaller
range resolution, while keeping the peak transmit power low. The technique is called pulse
compression.

12
Fundamentals of Radar Systems D. Lawrence

Radar Range Equation

The radar range equation provides a convenient method for calculating radar performance based upon
the radar’s fundamental parameters. The most basic form of the range equation can be written

Pt Gt Gr 2
Pr 
(4 ) 3 R 4 L
where
Pt = Transmit Power
Pr = Received Power
Gt = Transmit Antenna Gain
Gr = Receive Antenna Gain
 = Wavelength
 = Target’s Radar Cross Section
R = Range
L = System Losses

The minimum detectable signal can be calculated from the ubiquitous thermal noise density and basic
radar system parameters,

S min  kT  B  Fn  SNRmin

kT = Thermal Noise Density (174 dBm/Hz)

B = System Bandwidth
Fn = Receiver Noise Figure
SNRmin = Minimum SNR

Smin
a

SNRmin

Figure 13. Example received power calculation for a short range Ku-band radar.

13
Fundamentals of Radar Systems D. Lawrence

14
Fundamentals of Radar Systems D. Lawrence

Atmospheric Attenuation in the Radar Range Equation

The radar range equation including atmospheric attenuation is shown below where the term (R) is
added to show the range-dependent atmospheric loss.

Pt Gt G r  2   R 
Pr 
( 4 ) 3 R 4 L

The attenuation is most commonly quoted as a one-way attenuation with units of dB/km. To calculate
the 2-way attenuation in dB using the 1-way attenuation, , in dB/km and the range, R, in meters the
following expression is used.

R
 dB R   2 
1000

In the linear (non-dB) form of the range equation shown above the attenuation is written

  dB  R  
 
 R   10  10 

Range Equation in dB
It is useful in practice to work with the radar range equation in its dB form,

Pr _ dBm  Pt _ dBm  G t _ dB  G r _ dB  20 log     dBsm   dB R   30 log 4   40 log R   LdB

Along with the minimum detectable signal is written,

S min_ dBm  174  10 log B   NFdB  SNR min_ dB

Table III. Example Attenuation Coefficients

Condition One-Way Attenuation,  (dB/km)

Clear air (10GHz) 0.01
Clear air (35 GHz) 0.05
4mm/hr Rain (10 GHz) 0.05
10mm/hr Rain (10GHz) 0.15
4mm/hr Rain (35 GHz) 1.0
10mm/hr Rain (35GHz) 2.5

15
Fundamentals of Radar Systems D. Lawrence

Average Atmospheric Absorption

16
Fundamentals of Radar Systems D. Lawrence

Doppler Frequency
The radar return from a moving target can have a frequency that is shifted from the transmit frequency.
This is called the Doppler frequency and is very useful in radar systems for measuring target velocity and
separating targets from stationary clutter. Consider a transmit pulsed waveform,

stx (t )  P(t ) cos0 t 

If the received waveform is delayed in time, td, and scaled in amplitude as predicted by the radar range
equation,

srx (t )  Ar stx t  t d   Ar P(t  t d ) cos0 t  t d 

Ar is the amplitude of the received pulse and P(ttd) is the time-delayed pulse used to extract target
range information. The phase term the for the received signal can be expressed with the time delay
written explicitly as a function of time since the target is moving,

 (t )  0 t  t d (t )  0 t  0 t d (t )

First, consider a constant velocity target to relate range and time delay. The range is expressed

R(t )  R0  vt

The time delay is proportional to the two-way range to the target. The range of interest, however, is the
range that occurred when the signal was at the target location. This occurs at half of the time-delay prior
to the time the signal is received, ttd(t)/2. The time delay is written

 t (t ) 
2 R t  d 
 2 
t d (t ) 
c
  t (t ) 
2  R0  v   t  d 
 2 
 
c
2 R0  2vt  vt d (t )
 
c c

The time delay appears on both sides of the equation. Solving for td,

2 R 0  2v
t d (t )   t
cv cv

17
Fundamentals of Radar Systems D. Lawrence

Substituting this time delay into the function for the delayed sinusoidal phase results in 3 terms: the
original carrier, a constant phase term, and a Doppler component.

 2 R0   2v 
 (t )   0 t  t d (t )   0 t   0    0  t
cv cv

The Doppler frequency in Hz is simply expressed,

2v 2v
fd  f0  f0
cv c

The approximation holds when v << c which is the case for many practical situations. The easily
remembered Doppler expression is

2v
fd 
0

In general, the velocity used in the Doppler calculation is only the velocity component in the direction of
the radar. If Pt is the vector location of the target and Pr is the vector location of the radar, then the
Doppler calculation is modified to include the dot product of the closing velocity vector, v , and the line-
of-sight range vector.

fd  
2
v
P  P 
t r

0 Pt  Pr

Expressed in terms of the range vector,

2 R
fd   v
0 R

where R  Pt  Pr and R  R .

For a general target that is moving but with non-constant velocity, the range vector can be expanded in
a Taylor series to include velocity, acceleration, jerk components, and higher order terms if necessary

1  2 1  2
R (t )  R0  R t  R t  Rt  
2! 3!

18
Fundamentals of Radar Systems D. Lawrence

When calculating the time-delay associated with this range, we will neglect the fact that the range
should be evaluated at time td/2 earlier since the approximation is valid in most all cases. Substituting
into the expression for the received signal phase and expanding,

2 R (t )
 (t )   0 t  t d (t )   0 t   0
c
2 R (t )
 0t  0 R (t ) 
c R (t )
2 R (t ) 2 R (t ) 2R R (t )
 0t  0 R0    0 R  t  0  t2 
c R (t ) c R (t ) c 2 R (t )

The first term is the original carrier, the second term is a constant phase term that is not usually of
interest, the third term is the Doppler frequency term calculated earlier, and the fourth is a quadratic
phase term resulting from target acceleration. The term R (t ) R(t ) is the unit vector in the line-of-sight
direction from the radar to the target and is generally not a strong function of time.

19
Fundamentals of Radar Systems D. Lawrence

Radar Coordinates
Relating a target’s position coordinates to the radar measurement coordinates is frequently
encountered in radar engineering analysis. Consider a radar located at the origin in an inertial
coordinate frame with the boresight direction of the radar along the z-axis. The target location is (xt,yt,zt)
in the inertial frame shown in Figure 14. Conversion from spherical to Cartesian coordinates is written

x t  R sin  cos 
y t  R sin  sin 
z t  R cos 

Radar coordinates are usually given in range (R), azimuth angle (az), and elevation angle (el). Direction
cosines are commonly used to express the target angle location in radar coordinates.  and  represent
the angles as measured from the x and y reference axes, respectively. These are depicted graphically in
Figure 15 and Figure 16. The direction cosines are related to the Cartesian target coordinates by

xt
u  cos  
R
y
v  cos   t
R

These are also related to the common spherical coordinate system,

u  cos   sin  cos 

v  cos   sin  sin 

Often the target angles are given in azimuth and elevation rather than direction cosines since these
angles are referenced from the radar boresight direction. The azimuth and elevation angles are defined


 az  
2

 el  
2

Thus, the relationship to u,v angle coordinates becomes

u  sin  az
v  sin  el

Note that (az,el) = (0,0), and thus (u,v) = (0,0) corresponds to the boresight direction along the z-axis.
To be in real angle space, the target must be located within a unit circle in the u,v plane. Figure 17 shows
an example target. The outer circle represents  = 90 deg in spherical coordinates.

20
Fundamentals of Radar Systems D. Lawrence

Figure 14. Target location in inertial coordinate frame showing relationship of cartesian to spherical
coordinates.

Figure 15. Azimuth direction cosine angle measurement relative to x-axis.

21
Fundamentals of Radar Systems D. Lawrence

Figure 16. Elevation direction cosine angle measurement relative to y-axis.

Figure 17. u,v angle coordinates for a radar target.

22
Fundamentals of Radar Systems D. Lawrence

23
Fundamentals of Radar Systems D. Lawrence

Antenna Principles
An antenna provides a transition from a guided wave on a transmission line to an unbounded “free-
space” wave and vice versa. An example antenna is shown in Figure 18. There are many different
antenna types used in radar systems depending upon the application. Basic antenna types with
examples of each are listed below:

 Wire Antennas
– Dipoles
– Loops
– Yagi-Uda
 Aperture Type Antennas
– Pyramidal Horns
– Parabolic Reflector (dish)
– Slots
 Microstrip Antennas
– Patches
– Printed Dipole
– Vivaldi
 Antenna Arrays
– 1D Linear
– 2D Planar
– Phased Arrays
– Adaptive Arrays
 Broadband
– Helix
– Spiral
– Multi-resonant

No matter what antenna type is chosen, all antennas are characterized by certain fundamental
quantities such as antenna pattern, gain, polarization, and input impedance. The antenna radiation
pattern is a mathematical and/or graphical depiction of the radiation properties of an antenna often
shown as a function of the spherical coordinates  and . Two basic types of antenna patterns:

– Field Pattern: Voltage pattern proportional to |E(,)| on a linear scale.

– Power Pattern: Power pattern proportional to |E(,)|2 on a linear or decibel (dB) scale.

An example three-dimensional field pattern is shown in Figure 19 illustrating important pattern features.
A normalized pattern is obtained by dividing the field pattern by the maximum field pattern value,

E  ,  
F  ,   
E max

24
Fundamentals of Radar Systems D. Lawrence

Important pattern parameters are the half-power beamwidth (HPBW), the beamwidth between first
nulls (FNBW), and the peak sidelobe level (SLL). A voltage pattern cut is illustrated in Figure 20 as a polar
plot showing the HPBW and FNBW. Figure 21 is a plot of the power pattern on a rectangular plot
illustrating the sidelobe level measurement relative to the pattern peak. A power pattern plotted in dB
scale is the preferred method for viewing pattern details.

Figure 18. Example antenna functioning as transition from bounded to un-bounded wave.

Figure 19. Three dimensional antenna pattern example.

25
Fundamentals of Radar Systems D. Lawrence

Figure 20. Beamwidth definitions of voltage antenna pattern cut.

Figure 21. Antenna power pattern example illustrating beamwidth, peak sidelobe level, and backlobe.

26
Fundamentals of Radar Systems D. Lawrence

The antenna pattern is independent of range when taken far enough away from the antenna such that
the radiating phase-front becomes mostly planar rather than spherical. The pattern becomes valid in
what is known as the far-field region, and the pattern is called the far-field pattern. The standard far-
field conditions are

2D 2
R

In addition to
R  D
R  

D is the largest dimension of the antenna. The defined near-field and far-field regions of an antenna are
shown in Figure 22. The reactive near-field involves direct coupling through parasitic capacitance and/or
inductance to the antenna elements. The radiating near-field is beyond the reactive near-field, but the
wavefront is still largely spherical rather than planar. The phase front of the radiating signal becomes
planar in the far-field region.

The reciprocity theorem in electromagnetics implies that the antenna pattern is the same whether
transmitting or receiving. This is a very useful property since antenna pattern measurements can be
performed with a transmitting or receiving antenna under test, whichever is more convenient. Analytical
calculations are usually performed assuming a transmitting antenna , but due to reciprocity the resulting
pattern is valid for receive as well.

Figure 22. Near-field and far-field regions of an antenna. Antenna patterns are valid in the far-field
region.

In addition to the radiation pattern, the gain of an antenna is of fundamental interest. The radar range
equation requires calculation of the antenna gain, and it is useful to directly relate the gain to the

27
Fundamentals of Radar Systems D. Lawrence

radiation pattern. To derive the relationship between the two, consider a quantity defined to be the
radiation intensity of an antenna

1

U  ,    Re E  H *  r 2 rˆ
2
 Watts/steradian (sr)

The radiation intensity is a power density that has units of Watts/steradian. The steradian is simply a
“square radian”. Integrating the radiation intensity over a spherical angular space must equal the total
radiated power.

Pt  U  ,   d
2 
   U  , sin  d d
0 0
2 

  F  , 
2
 U max sin  d d
0 0

Define the beam solid angle in terms of the antenna pattern,

2 

  F  , 
2
A  sin  d d
0 0

Thus, the total power can be written as the maximum radiation intensity multiplied by the beam solid
angle,

Pt  U max  A

The average radiation intensity can be calculated by dividing the total power by the total beam solid
angle over a sphere. The beam solid angle of a sphere is calculated

2 
 sphere    sin  d d
0 0

 4

This is the beam solid angle for an antenna that radiates isotropically. The average radiation intensity is
equivalent to the transmitted power divided by the beam solid angle of an isotropic radiator,

Pt 1
U avg   U max  A
4 4

28
Fundamentals of Radar Systems D. Lawrence

Now the peak gain of an antenna is defined to be the ratio of maximum radiation intensity to average
radiation intensity1,

U max
G
U avg

Substituting from above, the peak gain is expressed

4
G
A

This important result indicates that the antenna gain can be calculated from knowledge of the radiation
pattern. Also, note that the beam solid angle for an antenna is inversely proportional to the gain. A
narrow beam solid angle results in high gain and vice versa. The largest possible beam solid angle is 4 ,
equivalent to an isotropic radiator, and corresponds to a gain of unity. Often the precise radiation
pattern is not known, but the half-power beamwidth is known. A useful approximation when the
pattern detail is not known has been proposed by Stutzman for high-gain antennas 2,

26,000
G
HPEo HPHo

where the half-power beamwidths in the E-plane (HPEo)and H-plane (HPHo) are specified in degrees. The
peak gain for an antenna is the gain at the peak of the radiation pattern. The general expression for gain
as a function of angle is simply the peak gain multiplied by the normalized power pattern,

G ,    G  F  ,  
2

An example antenna gain pattern normalized to the peak gain is shown in Figure 23 where the peak of
the antenna pattern corresponds to the 20 dB antenna gain. The gain at angles other than boresight is
determined by the peak gain reduced by the relative pattern in the direction of interest.

Recall gain is defined to be the ratio of peak to average radiation intensity. Radiation intensity is
converted to radiated power density by dividing by R2. Thus, the radiated power density from an
isotropic antenna is calculated by

1
In most antenna literature the antenna directivity, D, is introduced as the ratio of peak to average radiation
intensity. Gain, G, is proportional to directivity through an efficiency factor accounting for losses in the antenna.

2
W. L. Stutzman, G. A. Thiele, Antenna Theory and Design, 2 nd ed., pp. 297-298, Wiley, 1998.

29
Fundamentals of Radar Systems D. Lawrence

Pt
Sisotropic 
4R 2

This result can also be derived by dividing the radiated power by the surface area of a sphere. The
radiated power density from an antenna with gain, G, is by definition calculated as the increase in
power density relative to that of an isotropic radiator. Hence, the radiated power density becomes

Pt G
Sisotropic 
4R 2

Figure 23. Antenna pattern normalized to the peak gain.

Figure 24. Calculating radiated power density using antenna gain.

30
Fundamentals of Radar Systems D. Lawrence

Antenna gain is also related to the effective aperture size of the antenna. The effective aperture, Ae, is
not necessarily the same as the physical aperture, Ap, but is related by an efficiency factor, ap, between
0 and 1.

Ae   ap A p

The gain is directly calculated from the effective aperture and wavelength,

4 Ae
G
2

It is important to note that the gain is proportional to the electrical size of the antenna. The electrical
size is defined by the physical dimension normalized by the wavelength. In this case, the area is
normalized by the wavelength squared. Thus, the size of an antenna is only relevant once the
wavelength is specified. A 1m x 1m antenna aperture is electrically small at 1 MHz, but is huge at 10
GHz. The concept of effective area is also useful when considering the amount of power received by an
antenna from an incident power density. The incident power density (Watts/m 2)multiplied by the area
(m2) gives the total received power (Watts). This is illustrated in Figure 25.

Figure 25. Incident power density collected by antenna effective aperture area, Ae.

31
Fundamentals of Radar Systems D. Lawrence

Polarization
The orientation of the electric field as the wave propagates defines its polarization. Polarization is
defined to be linear, circular, or elliptical. Elliptical is the most general, but most antenna designs
attempt either pure linear or circular radiation. Linear polarization occurs when the electric field
remains oriented in the same linear direction. It oscillates back and forth at the carrier frequency, but
the pointing direction is always in the same line. Electric field components for a linear polarized wave
can be written,

E x  E1 cost  z 
E y  E 2 cost   z 

Note that the x and y components do not necessarily have the same amplitude, but they are in-phase.
The relative amplitudes of E1 and E2 determine the tilt angle of the linear polarized wave. Examples of a
linear polarized E-field are shown in Figure 26.

Figure 26. Linear polarization examples.

Circular polarization is characterized by a rotating electric field as the wave propagates. Just as in the
linear case, the E-field vector can be decomposed into orthogonal x and y components,

E x  E cost   z 
E y  E cos t   z  2 

The orthogonal components have the same amplitude but are out of phase by 90 degrees. The sign of
the 90 degree phase term determines the sense of rotation. The + sign indicates clockwise rotation and
is called left hand circular polarization (LHCP) and the – sign indicates counter-clockwise rotation and is

32
Fundamentals of Radar Systems D. Lawrence

called right hand circular polarization (RHCP). Examples of the rotating circular polarized E-field is shown
in Figure 27. A RHCP wave propagating in space is illustrate in Figure 28.

Figure 27. Right and left-hand circular polarization vector examples.

Figure 28. RHCP wave propagation in space.

33
Fundamentals of Radar Systems D. Lawrence

The most general case is elliptical polarization in which the E-field traces out an ellipse as it propagates.
The x and y components can be written in the general form

E x  E1 cost   z 
E y  E 2 cost   z   

The amplitudes E1 and E2 , along with the phase  , can be independently specified to vary the axial ratio
and tilt angle of the ellipse. The axial ratio is simply the ratio of the major axis to minor axis of the ellipse
and is usually specified in dB.

Figure 29. Elliptical polarization vector example.

The polarization of a wave or antenna can be specified conveniently using a complex polarization vector.
Consider a plane wave traveling in the +z-direction with E-field components in the x-y plane. The
complex vector representation of the polarized electric field is written using phasor notation,

E  E1 xˆ  E2 e j yˆ (1)

where E1 and E2 are the magnitudes of the x and y components, respectively, and the phase  is the
relative phase between y and x components. Note that the phase of the x-component is taken to be zero
without loss of generality.

E E1 E
eˆ   xˆ  2 e j yˆ
E E E (2)

34
Fundamentals of Radar Systems D. Lawrence

The polarization phase vector, ê , is a complex unit vector that carries all the polarization information
for a plane wave. The intensity of the wave is not included in the polarization phase vector since it is a
normalized quantity.

eˆ  cos  xˆ  sin  e j yˆ

The parameters (,) are sufficient to completely describe the polarization ellipse and useful for
calculations involving the polarization phase vector. A general polarization ellipse is shown in Figure 30.
An equivalent description is given by the parameters ( where  is an angle derived from the axial
ratio and  is the tilt angle of the ellipse. Well known conversions between these representations are

cos 2  cos 2 cos 2

tan 2
tan  
sin 2

The axial ratio is related to the angle  by,

AR   cot  , with 1  AR   , and  45    45

Figure 30. General polarization ellipse geometry.

35
Fundamentals of Radar Systems D. Lawrence

A useful quantity for calculating the interaction between an incoming wave and an antenna is called the
normalized complex voltage. The normalized complex voltage is defined to be the dot product of the
wave polarization vector and the conjugate of the antenna polarization vector,

veˆw , eˆa   eˆw  eˆa*

 cos  w cos  a  sin  w sin  a e j  w  a 

The polarization efficiency, p, sometimes referred to as the polarization mismatch factor, is equal to the
squared magnitude of the normalized complex voltage

p  veˆw , eˆa 
2

The polarization efficiency is bounded between 0 and 1, and quantifies the power coupling between an
incoming wave and receiving antenna due to polarization mismatch. This loss factor is most often given
in dB. A polarization insertion phase, p, can be defined to be the argument of the normalized complex
voltage

 p  argveˆw , eˆa 

The polarization insertion phase preserves the residual phase information from the interaction of an
incoming wave with the receiving antenna. The polarization efficiency alone does not capture this phase
since it is a power quantity. The insertion phase is useful when comparing the polarization dependent
effects on interferometer/monopulse angle measurements.

36
Fundamentals of Radar Systems D. Lawrence

Example calculations of polarization loss factor, p or PLF:

37
Fundamentals of Radar Systems D. Lawrence

38
Fundamentals of Radar Systems D. Lawrence

Antenna Input Impedance

From the source/transmitter perspective, the antenna can be viewed as a complex impedance fed by a
transmission line with characteristic impedance, Z0, which is often 50 . An example of this equivalence
is shown in Figure 31. The antenna input impedance can be separated into three terms: radiation
resistance, loss resistance, and reactance. Figure 32 illustrates the various impedance terms. The
radiation resistance and loss resistance are both real, while the reactance is complex. The radiation
resistance is the real load placed upon the antenna by free space. Power absorbed in this resistance is
equal to the radiated power from the antenna. The loss resistance accounts for the ohmic losses in the
antenna. Antenna efficiency becomes degraded when the radiation resistance is on the same order (or
lower) than the loss resistance. The reactance accounts for the residual capacitive/inductive impedance
presented by the antenna.

Figure 31. Complex antenna impedance.

Figure 32. Antenna impedance can be decomposed into radiation, loss, and reactance terms.

39
Fundamentals of Radar Systems D. Lawrence

Mismatch between the antenna impedance and transmission line impedance causes a reflection at the
antenna terminal and results in a loss of power radiated from the antenna. The square of the reflection
2
coefficient, in , is the ratio of power reflected to power incident at the antenna terminals. The
reflection coefficient is defined,

Z A  Z0
in 
Z A  Z0

A useful quantity for calculating the loss in transmitted power is the insertion loss,

2
Insertion Loss  1  in

Insertion loss is the reduction in transmitted power relative to the incident power at the antenna. It is
usually given in dB. For the same antenna/transmission-line configuration, the loss would be
experienced on both transmit and receive. Oftentimes, this loss is included in the antenna gain
measurement when the same feeding transmission line impedance is used to measure gain.

40
Fundamentals of Radar Systems D. Lawrence

Noise
The ultimate limit in radar performance is determined by the noise power in the receiver. Noise power is
the result of random processes arising from many potential sources both external and internal to the
radar. The most basic noise source is the thermal noise generated by any component operating above
absolute zero. For practical radar frequencies, the RMS voltage of thermal noise generated in a resistor,
R, is given by the expression

v n  4kT0 BR

where k is Boltzmann's constant, 1.38 x 10-23 J/oK, and T0 is the temperature in Kelvin. Noise power
dissipated in a resistor matched to the source represents the maximum available noise power,

2
v 
PN   n   R
 2R 
 kT0 B

Notice that the maximum available noise power is independent of the resistance and is only a function
of the temperature and bandwidth. Under typical radar operating conditions, the temperature is taken
to be standard temperature, T0 = 290 oK. Small variations in temperature do not significantly affect the
resulting noise power. The factor kT0 represents the thermal noise density in Watts/Hz and is a useful
quantity to remember,

kT0  4  10 21 W
Hz

 204 dBw
Hz

 174 dBm
Hz

Figure 33. Circuit model for thermal noise generated in a resistor.

41
Fundamentals of Radar Systems D. Lawrence

vn
I 
2R

2
v 
PN   n   R
 2R 
 kT 0 B

Figure 34. Maximum available noise power delivered to a matched load.

Noise Figure
In addition to thermal noise, electronic components in the radar receiver also contribute to the noise
floor. The output signal-to-noise will always be less than the input signal-to-noise due to the added
noise. Noise Figure is defined to be the ratio of input to output signal-to-noise, SNR.

S in
N in S in N out
Fn    1
S out S out N in
N out

This is always a number greater than or equal to 1 since the input SNR is always greater than or equal to
the output SNR. An illustration of the noise figure applied to a noisy device with gain, G, is shown in
Figure 35. Adjusting the thermal noise power by the noise figure accounts for the additional noise power
added by the receiver,

PN  kT0 BFn

Here the noise power is expressed without the gain, G, since it is common to both signal and noise
terms at the receiver output. Essentially the noise is referenced to the front end of the receiver in order
to compare to the received signal power calculated from the range equation.

Figure 35. Calculating output noise power using noise figure.

42
Fundamentals of Radar Systems D. Lawrence

Noise figure is often expressed in dB,

NF  10  logFn 

Using dB notation, the thermal noise power adjusted by the noise figure becomes

PN  174 dBm Hz  10  logB   NF

The noise power calculated from this equation is in units of dBm. It can easily be converted to dBw by
subtracting 30 dB.

Effective Noise Bandwidth

The frequency spectrum of thermal noise expressed differently depending upon whether the noise is
taken to be one-sided or two-sided. One-sided noise only considers positive frequencies. For a one-sided
spectrum, the power spectral density of thermal noise is a constant

S n ( )  kT0  N 0

The bandwidth used for noise calculations is derived from the filter bandwidth used in the detection
process. This can be the IF matched filter bandwidth or the digital FFT bandwidth depending upon the
specific radar processing architecture. For one-sided noise, the effective noise bandwidth is formally
calculated from the filter transfer function, H(f),

 H( f )
2
df
B 0
2
(one-sided noise bandwidth)
H ( f0 )

2
H( f ) 
B  H( f0 )   H ( f ) df
2 2

0
2
H ( f0 )

B
Figure 36. One-sided equivalent noise bandwidth.

43
Fundamentals of Radar Systems D. Lawrence

For a two-sided spectrum, the power spectral density of thermal noise is also constant but extends over
both positive and negative frequencies. The two-sided density is half of the one-sided value.

kT0 N 0
S n ( )  
2 2

Similarly, the effective noise bandwidth for the two-sided case is calculated

 H( f )
2
df
B 
2
(two-sided noise bandwidth)
H ( f0 )

Noise Temperature
A useful concept in thermal noise calculations is the effective noise temperature of a device. The noise
temperature of a device is defined to be the effective temperature, T e, that represents the noise power
added by the device. This noise is referred to the input. The output noise power can be expressed as the
sum of the input noise and equivalent device noise both multiplied by the gain of the device.

N out  k T0  Te BG  kT0 BGFn

By equating the output noise expressed in terms of equivalent noise temperature with the output noise
expressed in terms of the noise figure, the following relationship is observed

Te
Fn  1 
T0

Rearranging terms, the equivalent noise temperature can be written

Te  Fn  1T0

Figure 37. Equivalent noise temperature of a noisy device.

44
Fundamentals of Radar Systems D. Lawrence

Figure 38. Cascaded devices with noise represented by noise temperatures.

Figure 39. Equivalent model for cascaded devices.

The noise temperature is perhaps most useful when calculating the thermal noise of cascaded devices in
a receiver.

N out  k T0  T1 BG1G2 G3  kT2 BG2 G3  kBT3G3

 kTe BGT

Dividing both sides by the total gain, GT = G1G2G3 ,and solving for the equivalent temperature

𝑇 𝑇
𝑇 =𝑇 + +
𝐺 𝐺 𝐺

Using the connection between noise figure and noise temperature, it can be shown that the equivalent
noise figure of the cascaded devices is written

F2  1 F3  1
Fn  F1  
G1 G1G2

For a low overall noise figure, it is important for the first component to have low noise figure and high
gain. Under this condition, the noise figure of the cascaded devices is dominated by the noise figure of
the first component since the gain divides the noise figure of subsequent stages. Conversely, if the first

45
Fundamentals of Radar Systems D. Lawrence

stage has low gain, the subsequent stages may increase the overall noise figure. An important case to
consider is the noise figure of a passive lossy component, such as an attenuator or lossy transmission
line. It can be shown that a lossy line, G = 1/L, at a physical temperature T has an equivalent noise
temperature given by

Te  L  1T

and the noise figure is

T
Fn  1  L  1
T0

For many cases, the lossy line is at a physical temperature T0, and the noise figure is simply equal to the
loss, Fn = L.

For the preceding discussion, it has been assumed that the input thermal noise always corresponds to
the standard temperature T0. In fact, it is standard practice for components to specify noise figure using
this convention so that a common reference is used. However, a radar system in operation may not
always have an input temperature of T0. For example, when an antenna is pointed toward the sky the
noise temperature can be well below 290 oK. When the sun is in the main beam, the equivalent input
temperature can far exceed 290 oK. To account for this difference, another noise figure is introduced
called the system noise figure, Fs. Sometimes this is referred to as the operating noise figure. When the
input noise temperature is Ts, the output noise power is calculated

N out  k Ts  Te BG  kTs BGFs

And the system noise figure becomes

Te
Fs  1 
Ts

The system (operating) noise figure can be calculated directly from the standard noise figure,

T0
Fs  1  Fn  1
Ts

46
Fundamentals of Radar Systems D. Lawrence

Detection of Signals in Noise

Target detection in the presence of noise is the primary role of a surveillance radar. Tracking radars also
require target acquisition before a track can be initiated. The detection of signals in the presence of
noise must be described in terms of statistical quantities since the noise, and sometimes the target
amplitude, is random in nature. For a particular radar dwell there are two possible decisions:

1) Target Present 2) No Target Present

The probabilities of interest in radar detection are defined to be

Probability of Detection, Pd: The probability that the radar declares a target is present given that a
target is indeed present.

Probability of False Alarm, Pfa: The probability that the radar declares a target is present given that a
target is not present.

Probability of Miss, PM: The probability that the radar declares a target is not present given
that a target is indeed present. The probability of miss is related to
the probability of detection, PM = 1 - Pd.

We begin with a discussion of noise alone without a target in order to calculate Pfa. Following this, a
constant target amplitude is introduced in the presence of random noise to calculate Pd for different
signal-to-noise ratios (SNR). Noise in a radar receiver is often modeled as a bandlimited random process.
Written in complex notation, the in-phase and quadrature components take the form

n(t )  x(t )  jy (t )

where x(t) and y(t) are independent, zero-mean Gaussian distributed random processes each having
variance 2. The variance of the complex noise process n(t) is directly proportional to the noise power in
the radar processing bandwidth, B.

 
E n 2 (t )  2 2  kT0 B

47
Fundamentals of Radar Systems D. Lawrence

To make a target detection, a radar receiver often employs an envelope detector to remove the
underlying carrier. Receiver noise most often contains uncorrelated Gaussian distributions in the in-
phase and quadrature components of the signal, assuming no target signal is present. The output of an
envelope detector when noise alone is present has Rayleigh probability density function (pdf) statistics.
The Rayleigh pdf is given by

z  z2 
f Z ( z)  exp    , for z > 0
2  2
2


where the random variable, z, is the magnitude of the complex noise process, z(t) = |x(t)2 + y(t)2|. An
amplitude threshold, Vt, is defined in order to decide whether a target is present or not. If the received
amplitude exceeds the threshold, then a target is declared present. Otherwise, a target is not declared
present. A false alarm occurs when noise alone exceeds the threshold. An example of the amplitude of a
bandlimited noise process is shown as a function of time on the left in Figure 40. The corresponding
Rayleigh pdf of the amplitude of the noise process is shown on the right in Figure 40. The location of a
voltage threshold is shown on both plots to illustrate the connection between the noise magnitude
crossing the threshold in time and the probability of false alarm calculated by integrating the Rayleigh
pdf. Formally, the probability of false alarm is calculated by integrating the Rayleigh pdf from Vt to
infinity. The integral can be easily evaluated and results in a convenient closed-form relationship
between Pfa and the detection threshold. Defining the SNR threshold to be SNRthreshold = Vt2/22,


Pfa   f Z ( z ) dz
Vt

 V  2
Pfa  exp  t 2   exp SNRthreshold 
 2 

Figure 40. Amplitude of bandlimited noise process (left). Rayleigh pdf illustrating Pfa calculation (right).

48
Fundamentals of Radar Systems D. Lawrence

Figure 41. Relationship between Pfa and the SNR threshold.

Often in system requirements the average time between false alarms is a more useful quantity than the
probability of false alarm. The average time between false alarms can be calculated using the probability
of false alarm and the receiver processing bandwidth,

Tdwell 1
T fa  
Pfa B  Pfa

Probability of Detection
The probability of detecting a target is determined by calculating the probability that the signal-plus-
noise exceeds the threshold given that a target is present. When a target is present, the radar return can
be written simply as the target amplitude, A, added to the bandlimited noise process. Using the complex
notation as before,

z (t )  A  n(t )

A constant RCS target produces the same amplitude for all dwells and is not considered a random
variable. Later, fluctuating target amplitude will be considered in statistical terms. The probability
density function describing the magnitude of a constant signal plus complex, Gaussian random noise is
known as the Rice (or Rician) pdf,

z   z 2  A2   zA 
f Z ( z )  2 exp   I  2 
2  0 
   2     

where I0 is the modified Bessel function of order 0.

49
Fundamentals of Radar Systems D. Lawrence

It is convenient to re-write the Rician pdf in terms of SNR,

z   z2   z 
f Z ( z)  exp   
 2 2  SNR  
 I  2 SNR 
2
0
    

where SNR = A2/22. The Rician pdf for several values of SNR are shown in Figure 42. When the target
amplitude is 0 (SNR = 0), the Rician pdf is identical to the Rayleigh pdf. As the SNR gets large, the Rician
pdf approaches Gaussian with mean equal to the target amplitude A. Calculating the probability of
detection is achieved by integrating the Rician pdf above the detection threshold, Vt.


Pd   f Z ( z )dz
Vt


z   z2   z 
 exp   2  SNR  I 0  2 SNR  dz
Vt
 2
  2    

Unfortunately, the integral is not as easily evaluated as the Rayleigh case. The expression can be written
in terms of Marcum's Q function,

A V 
Pd  Q , t 
  

 Q 2SNR , 2SNRthreshold 
where Q(a,b) is Marcum's Q function defined to be


   2  a 2 
Qa, b     exp   I 0 a  d
b   2 

This is a built-in function, marcum(a,b), in Matlab's Signal Processing Toolbox. An infinite series solution
has also been derived and can be conveniently written in terms of the target SNR and the threshold SNR,

 
SNRthreshold 
m
SNR k k
Pd  e SNRthreshold  SNR      m! 
k 0  k! m0 

50
Fundamentals of Radar Systems D. Lawrence

Figure 42. Rician pdf examples for normalized noise,  = 1.

The SNR = 0 (-  dB) case is identical to the Rayleigh pdf.

Figure 43. Pd obtained by integrating the Rician pdf.

Another option is to evaluate the Pd integral numerically. This can produce unstable results when the
argument of the Bessel function becomes large. A useful approximation for numerical implementation
utilizes the large argument approximation for the Bessel function. The Rician pdf can be written
approximately

51
Fundamentals of Radar Systems D. Lawrence

 z   z 2  A 2   zA  zA
 2 exp   I 0  2  for 0  2  80
   2     
2

f Z ( z)  
 1   z  A  2
zA
exp  for  80
 2  2  2
2
 

where the factor of zA/ = 80 is used as the threshold for the Bessel function approximation. This
threshold could be adjusted depending upon the specific behavior of the numerical approach being used
and the required accuracy. Furthermore, better results can be achieved by avoiding the infinite limit
when calculating the Pd integral numerically. Instead, the integral can first be evaluated from 0 to Vt
giving the probability of miss. Subtracting from 1 then produces Pd. Specifically,

Vt

Pd  1   f Z ( z ) dz
0

Another convenient form for the Rician integral can be used for numerical integration. Defining and SNR
variable u = z2/22, the following pdf is written exclusively in terms of the SNR,




exp u  SNR I 0 2 u  SNR  for 0  2 u  SNR  80
fU (u )   1

 4  SNR exp  u  SNR  2 u  SNR  for 2 u  SNR  80


The Pd integral then becomes

SNRthresh

Pd  1  f 0
U (u ) du

There are conflicting design goals for Pd and Pfa. Figure 44 illustrates the relationship of Pd and Pfa to the
threshold setting Vt. As the threshold is lowered, Pd increases as desired but Pfa also increases. Raising
the threshold reduces Pfa but also reduces Pd. An ideal but unrealistic design is to have Pd = 1 and Pfa = 0.
This can only occur when the noise-alone pdf and the signal-plus-noise pdf have no overlapping regions
and the threshold is placed in the gap between them. This situation is only approached when the SNR
becomes large. In most radar systems, the threshold is set by defining a tolerable Pfa and the remaining
system parameters are designed to achieve the desired Pd.

52
Fundamentals of Radar Systems D. Lawrence

Figure 44. Pd and Pfa dependence upon threshold Vt.

Receiver Operating Characteristic
A useful tool in evaluating receiver detection performance is known as receiver operating characteristic
(ROC) curves. ROC curves depict the Pd versus Pfa for a given target SNR as the threshold varies. Figure
45 illustrates ROC curves for a single radar dwell. The curves are generated by sweeping the threshold
from 0 to infinity and calculating the corresponding Pd and Pfa for each threshold. The calculated values
are then plotted versus each other without explicitly showing the threshold setting. Since the detection
probability is dependent upon the SNR, separate curves are generated for different SNR values.

Figure 45. ROC curve examples for a single radar dwell. More information can be obtained from the plot
if the Pfa axis is plotted on a logarithmic scale (right).

53
Fundamentals of Radar Systems D. Lawrence

False Alarm Probability for Multiple Range/Doppler Cells

Radar systems often employ processing that allow multiple, independent opportunities for a false alarm
for a single target detection opportunity. An example of such a case is shown in Figure 46 where Doppler
processing of each radar dwell produces multiple range bins and Doppler bins. The increase in false
alarm probability can be accounted for using classical probability theory. If there are N independent
opportunities for a false alarm, the false alarm for the whole radar dwell can be calculated in terms of
the single cell false alarm probability,

Pfa _ dwell  1  1  Pfa _ cell 

N

When the false alarm probability is much less than 1, the following approximation can be used,

Pfa _ dwell  N  Pfa _ cell

Figure 46. Each cell in a range/Doppler matrix is a potential false alarm opportunity.

54
Fundamentals of Radar Systems D. Lawrence

Signal Integration
Target detection performance can be improved by integrating multiple pulses before making a target
present/ not present decision. The term "integration" can be misleading since the processing is actually
performed on discrete outputs and is probably better described by summation. However, due to well
established convention, the term integration will be used here. Integration can be done in three ways:

1) Coherent Integration
Applied after coherent demodulation and the matched filter but before envelope detection.
This approach requires that phase information be preserved from pulse to pulse.
2) Non-Coherent Integration
Applied after envelope detection. Only magnitude information is needed.
3) Binary Integration
Applied after threshold detection to the target present/not present decision.

Figure 47. Typical radar signal processing operations for detection.

Binary Integration
Rather than making a target present decision on a single opportunity, better performance can often be
achieved by utilizing multiple dwells. Binary integration refers to the decision rule requiring detections
on at least M out of N opportunities before declaring a target present. An example of at least 3 of 5
detection is shown in Figure 48. Detection opportunities can be consecutive dwells/CPIs or separate
antenna channels. Noise in sequential dwells or channels must be independent to qualify as separate
detection opportunities. Using Bernoulli trials from probability theory, the cumulative false alarm
probability for an M out of N decision rule is calculated

N
  k  P 1  P 
N
N k
PCfa  k
fa fa
k M  

When Pfa << 1, the following approximation is valid

N
PCfa    PfaM
M 

Similarly, the cumulative probability of detection is calculated

55
Fundamentals of Radar Systems D. Lawrence

N
N
  k  P 1  P 
N k
PCd  d
k
d
k M  

A common binary decision rule utilizes 2 of 3 detection logic. This is illustrated in Figure 49. In this case
the cumulative probabilities can be simplified,

PCd  3Pd 1  Pd   Pd
3
2

2 3
 3Pd  2 Pd

2 3 2
PCfa  3Pfa  2 Pfa  3Pfa

Figure 48. Binary Integration example showing 3 of 5 detection logic. Out of 5 opportunities, detections
are made on # 1,3, and 5.

Figure 49. Common 2 out of 3 detection logic example with cumulative probability expressions.

56
Fundamentals of Radar Systems D. Lawrence

Figure 50. ROC curves for different binary integration decision rules.

57
Fundamentals of Radar Systems D. Lawrence

Coherent Integration
Coherent integration is possible when the phase information from the target return is known and
preserved from pulse-to-pulse. The target dynamics must be benign enough to stay coherent over the
coherent processing interval (CPI). Target dynamics often limit the time available for integration. This
occurs when the target's radial velocity is large enough such that the target moves through a range bin
during a dwell or the lateral velocity moves out of the antenna beam. Coherent integration consists of
multiple pulses added together before the envelope detector. When a target is present with constant
amplitude A for each pulse, the coherent addition of N pulses can be written

s (t )  z1 (t )  z 2 (t )    z n (t )
 A  n1 (t )    A  n 2 (t )     A  n N (t )
N
 NA   n k (t )
k 1

The signal power after integration is calculated by recognizing the first term as the signal output

PS   NA   N 2  A 2
2

which is N2 times the power of a single pulse. The noise power is calculated from the power in the
output noise term and uses the fact that the noise is uncorrelated from one pulse to the next.

 N  
2


E   nk (t )    E n12 (t )  n22 (t )    n N2 (t ) 
 k 1  
 N  E n 2 (t ) 
 N  2 2

Since uncorrelated noise adds in power rather than voltage, the output noise power is increased by N
rather than N2. As a result, the signal-to-noise ratio after coherently integrating N pulses is

N 2  A2
SNR N   N  SNR1
N  2 2

Note that the SNR is increased by a factor of N above that of a single pulse. This result can be
incorporated into the range equation to quantify the improvement in detection performance. Consider
the form of the range equation written in terms of the minimum required signal, Smin,

Pt Gt Gr 2
4
Rmax 
(4 ) 3 L S min

58
Fundamentals of Radar Systems D. Lawrence

The minimum required signal can be expanded in terms of the noise density, including the system noise
figure, and the minimum SNR. Explicitly, the minimum required signal is written

S min  kTBFn  SNRmin

Substituting into the range equation yields,

Pt Gt Gr 2
4
Rmax 
(4 ) 3 L  kTBFn  SNRmin

Note that SNRmin is the minimum SNR (A2/22) needed to satisfy the probability of detection and
probability of false alarm statistics. When coherent integration is used, the minimum SNR is reduced by
a factor of N compared to the single pulse SNR. For a single pulse SNR denoted SNR1, the range equation
for coherent integration of N pulses becomes

Pt Gt G r  2  N
4
R max 
(4 ) 3 L  kTBFn  SNR1

Non-Coherent Integration
Non-coherent integration takes place after the magnitude detector and operates only on the amplitude
of the received signal rather than the phase. The summation of N magnitudes is written

z (t )  z1 (t )  z 2 (t )    z N (t )

Calculating the improvement in signal-to-noise requires the evaluation of the pdf of the summed signal.
The pdf of the sum can be calculated by convolving the pdfs of the individual terms.

f Z ( z )  f Z 1 ( z )  f Z 2 ( z )    f ZN ( z )

Assuming that each pulse is an independent, identically distributed random variable, the pdf of the sum
can be written conveniently using the Fourier transform of the pdf,

f Z ( z )  f Z1 ( z)  f Z1 ( z)   f Z1 ( z)

  1  f Z 1 ( z )
N

where we make use of the fact that convolution becomes multiplication in the Fourier transform
domain. The individual pulse pdf fZ1(z) is Rayleigh for the noise alone case and is Rician for the signal-
plus-noise case. As the number of pulses increases, the resulting pdf tends toward Gaussian as predicted

59
Fundamentals of Radar Systems D. Lawrence

by the Central Limit Theorem from probability theory. Even with the Fourier transform simplification,
this equation is difficult to evaluate analytically. As a result, a useful approximate expression has been
derived for the relationship between Pfa , Pd, and the required single pulse SNR when integrating N
pulses non-coherently. The expression is known as Albersheim's equation,

  4.54 
SNR N  5log10 N  6.2     log10  A  0.12 AB  1.7 B 
  N  0.44 

The result is in dB and the parameters A and B are defined in terms of the detection statistics,

 0.62 
A  ln  
 P 
 fa 
 P 
B  ln  d 
 1  Pd 

Albersheim's equation is accurate within 0.2 dB for 10-7 < Pfa <10-3 and 0.1 < Pd < 0.9. This signal to noise
can be used in the range equation to evaluate performance given the specified detection statistics.

Pt Gt Gr 2
R 4

(4 ) 3 L  kTBFn  SNR N
max

It can be shown that the equivalent improvement in SNR that is gained by non-coherent integration is
approximately N0.7 to N0.8. This is below the optimal gain of N provided by coherent integration but is
still useful since it does not require phase coherency between pulses. In general, the range equation can
be written in terms of the SNR with no integration, SNR1, and an integration improvement factor, I(N),

Pt Gt Gr 2  I ( N )
4
Rmax 
(4 ) 3 L  kTBFn  SNR1

where I(N) is N for coherent integration and approximately N0.7 to N0.8 for non-coherent integration. If
more accuracy is needed for the non-coherent case, I(N) can be calculated using Albersheim's equation,
computed analytically, or evaluated from a table.

60
Fundamentals of Radar Systems D. Lawrence

Figure 51. Required single pulse SNR is reduced by pulse integration. The detection statistics for this
case are Pfa = 10-5, Pd = 0.9, with the required SNR without integration being about 12.5 dB.

Figure 52. Improvement factor for coherent and non-coherent integration. The detection statistics for
the non-coherent case are Pfa = 10-5, Pd = 0.9. Note that the non-coherent improvement factor is
generally between N0.7 and N0.8.

61
Fundamentals of Radar Systems D. Lawrence

Radar Cross Section

The radar cross section (RCS) for a target is defined to be the equivalent area that when multiplied by
the incident power density and re-radiated isotropically produces the same power density at the
receiver as the target. It is related to the physical cross section area but is generally not the same. The
RCS can vary significantly with aspect angle and a single RCS value is usually not sufficient to completely
describe a target. Furthermore, a bistatic RCS can be defined that varies with both incident angle and
scattered angle. It is often convenient to decompose the RCS into three basic components,

  Geometric Area  Reflectivity  Gain

Calculating RCS in this way is approximate but provides accurate results in the high frequency limit. The
geometric area is simply the physical area presented to the incident wave. The reflectivity is the ratio of
re-radiated power to incident power and is bounded between 0 and 1. The reflectivity is 1 for a perfectly
conducting target, and the reflectivity is 0 for a target that completely absorbs the incident power.
Finally, the gain is determined by the geometry of the target and is treated as an antenna having the
same geometry and current distribution. Formally, the RCS of a target can be calculated from the
scattered and incident electric fields,

E s  ,  
2

  ,   lim 4R 2 2
R 
Ei

Computational electromagnetics is often employed in order to calculate the RCS of complex scatterers
using the above formalism.

Calibration Targets
1. Sphere
A number of basic geometric shapes are used for calibration targets since their RCS is well known. The
sphere is a useful calibration target since the monostatic RCS is independent of angle and does not
require precise positioning. For a sphere, the normalized RCS is plotted below showing several distinct
scattering regions. In the low frequency limit, the RCS follows the Rayleigh scattering law

  9a 2 ka 4
ka  0

The Rayleigh approximation is valid for a < 0.1 . Note that the RCS varies as  in the low frequency
region. In the high frequency limit, the RCS becomes independent of frequency and equal to the
geometric area,

  a 2
ka  

62
Fundamentals of Radar Systems D. Lawrence

This is known as the optical region and is satisfied for a > 2  or so. The oscillating region in between the
low and high frequency limits is known as the resonance region. The oscillation in RCS is due to the
constructive and destructive interference of surface waves travelling around the sphere with the direct
reflection. These surface waves are sometimes called creeping waves.

Figure 53. Sphere RCS illustrating the low frequency Rayleigh region, resonance region, and high
frequency optical region depending upon the sphere radius, a.

63
Fundamentals of Radar Systems D. Lawrence

2. Flat Plate
Another fundamental target is a metallic flat plate. Consider a square flat plat with dimension a x a. In
the high frequency limit the RCS can be estimated by using the heuristic approach described earlier. The
geometric area, reflectivity, and gain are each determined separately and combined to form the RCS:

Geometric Area: A  a
2

Reflectivity: 1 (perfect conductor)

4A
Gain: G 
2

The total RCS is then estimated

4A 2

2

This value applies when the incident wave is normal to the plate and represents the maximum value. If
the plate is tilted with respect to the incident wave, then the RCS will be less. Consider a flat plate tilted
in azimuth with angle  relative to the incident plane wave. The RCS can be calculated using the heuristic
approach by taking into account the projected area and phase taper across the plate,

4A 2  sin ka sin   

2

   cos 2 
  ka sin  
2

The main disadvantage of the flat plate as a calibration target is the dependence on orientation. Slight
misalignments produce vastly different RCS values. The development is useful, however, since many
targets can be approximated using a combination of flat plate surfaces.

3. Dihedral Corner Reflector

A dihedral corner reflector is constructed by arranging two flat plates with a common edge at a 90 deg
angle. The advantage of the corner reflector is that it keeps the high RCS value over a much larger range
of incidence angles compared to an isolated flat plate. The principle of operation is shown in the figure
below where the incident wave arrives at an angle  with respect to one of the plates. Using ray tracing,
it is easily shown that the reflected ray will be in the same direction as the incident ray. The RCS of the
dihedral is approximately

4Aeff
2
8a 2 b 2
 
2 2

where the effective area, Aeff  2ab , is the effective area of the reflector when  = 45 deg.

64
Fundamentals of Radar Systems D. Lawrence

Figure 54. Dihedral corner reflector composed of two flat plates.

Figure 55. Corner reflectors reflect waves back in the direction of the incident waves.

65
Fundamentals of Radar Systems D. Lawrence

4. Trihedral Corner Reflector

One of the most commonly used calibration targets for large RCS is the trihedral. It uses the same
principle as the dihedral but adds a third plate oriented 90 deg to the other two. Common
configurations are the square and triangular trihedral. The RCS is slowly varying over a wide range of
incidence angles with the peak RCS,

12a 4 4a 4
 square _ trihedral  and  trianglular _ trihedral 
2 32

The half-power response is about 23 deg for the square trihedral and 40 deg for the triangular trihedral.
So the square trihedral presents a larger RCS over a smaller angular extent compared to the triangular
trihedral which presents less RCS but extends over a larger angle.

Figure 56. Square trihedral (left) and triangular trihedral (right).

Figure 57. L-band triangular-trihedral corner reflector in the field.

66
Fundamentals of Radar Systems D. Lawrence

Composite RCS for Complex Targets

Targets can often be characterized by multiple scatterers located on the target body. Consider the case
of a target consisting of N scatterers with the nth scatterer located at the coordinates (xn,yn,zn) having
RCS n and scattering phase n. The contribution from each scatterer is added in voltage rather than
power to account for constructive/destructive interference between components. The scattering phase
between scatters depends upon the individual scattering phases and the two-way path length
differences between scattering centers. Using the origin as the phase reference, the composite RCS from
N scatterers is calculated

N 2

  ,     ne j n
exp2 jk  x n sin  cos   y n sin  sin   z n cos  
n 1

where k = 2 is the propagation constant, and the composite RCS is a function of incident angles  and
. It should be noted that, in general, the RCS and scattering phase from each component could also
depend on the incident angles. For a two-dimensional scattering scenario with all scatterers in the x-y
plane, the composite RCS can be written in the simpler form,

N 2

     ne j n
exp2 jk  x n cos   y n sin  
n 1

A two-dimensional scattering example is shown in the figure below illustrating the path length
calculation for the nth scatterer.

Figure 58. Composite RCS calculated from individual scatterers.

67
Fundamentals of Radar Systems D. Lawrence

Fluctuating RCS – Swerling Models

The RCS for a complex target cannot usually be described by a single value. Rather, the range of
potential RCS values must be described in statistical terms. For large targets consisting of multiple
scatterers spaced multiple wavelengths apart, the RCS will vary significantly with aspect angle. Swerling
models describe the statistical fluctuation of targets depending on 1) the target scatterer distribution
and 2) the variation during or between dwells. Table 3 lists four distinct target fluctuation models
introduced by Swerling for fluctuating targets. The non-fluctuating case is labeled 0 or 5 and is usually
added to the Swerling list for comparison. Cases 1 and 2 share the same amplitude pdf but differ in the
rate of fluctuation. Case 1 represents slow fluctuations that vary from dwell-to-dwell but remain
constant across sequential pulses used for integration. Case 2 represents fast fluctuations that vary from
pulse-to-pulse. Cases 3 and 4 both have the same amplitude pdf, but Case 3 has slow fluctuations and
Case 4 has fast fluctuations. In what follows, detailed descriptions of each case are given.

Table 3. Swerling Target Model Descriptions

Swerling 1 and 2
For Swerling 1 or 2, the pdf of the target return amplitude is Rayleigh distributed. This case best
describes a target that consists of many scatterers of similar amplitude. The Rayleigh pdf for the
amplitude, A, is written

A  A2 
f ( A)  exp  
2 
A0  2 A0 

The term A02 represents the average value of the received signal power. It is straightforward to show
that this is equal to the expected value of the normalized signal power,

1 
A02  E  A 2 
2 

68
Fundamentals of Radar Systems D. Lawrence

It is also instructive to evaluate the pdf of the RCS. Since the RCS is directly proportional to the square of
the amplitude, the pdf of the RCS is exponential for the Swerling 1 or 2 case. (Recall a Rayleigh
amplitude pdf corresponds to exponential power pdf.)

1  
f ( )  exp  
  

where the mean RCS is denoted  . The range equation provides the connection between A and .
Specifically, the expected value of the received signal power is written

1 2   Pt Gt Gr 2  Pt Gt Gr 2
E  A   A0  E 
2

 ( 4 ) R L  (4 ) R L
3 4 3 4
2 

The signal-plus-noise pdf for a single pulse must take into account the random nature of the signal
amplitude. The familiar Rician pdf becomes a conditional pdf since A is a random variable. This is
formally written

z   z 2  A2   zA 
f Z ( z | A)  exp    I 0  
2     
2
  2

The single pulse pdf is then obtained by integrating the conditional pdf over the possible values of A
weighted by the pdf,


f Z 1 ( z )   f Z ( z | A) f A ( A)dA
0

z   z 2  A2   zA 
  2 exp    I 0    f A ( A) dA
0    2    
2

  
z z2  1 
exp 


 2 1 A02
2
  2 2



2
1  A02



This represents the signal-plus-noise pdf for a single pulse for both Swerling 1 and 2 targets.

69
Fundamentals of Radar Systems D. Lawrence

Swerling 1 and 2 Pdf

0.7

0.6

0.5

0.4
f(A)
0.3

0.2

0.1

0
0 1 2 3 4 5
RCS Amplitude / A0

Figure 59. Swerling 1 and 2 RCS models use a Rayleigh amplitude pdf.

Swerling 1
The difference between Swerling 1 and 2 becomes relevant when integrating multiple pulses. Since
Swerling 1 assumes slow fluctuations, the amplitude can be assumed constant during the coherent
integration time. The amplitude varies, however, from dwell-to-dwell. This is illustrated in the figure
below. Coherent integration of N pulses for Swerling 1 is straightforward since the same amplitude is
assumed for each pulse. The amplitude is still a random variable and must be integrated to obtain the
pdf for coherent integration of N pulses. The conditional pdf is altered by multiplying the amplitude A by
N to yield a factor of N improvement in the SNR. The pdf for coherent integration of N pulses
becomes,


f ZN ( z )   f Z ( z | N A) f A ( A)dA
0

z  z 2  NA 2   z N A 
  2 exp   I 0  
    f A ( A) dA
0
   2  2
  
  
z z2  1 
exp


 2 1  NA
 2
2
0
  2  2  1  NA202
  



The probability of detection is calculated by integrating the resulting pdf,

70
Fundamentals of Radar Systems D. Lawrence


Pd   f ZN ( z )dz
Vt

   
z  z2  1 

Vt 
 2 1  NA
 2
2
0
 exp 
 2  2  1  NA202
  
 dz


Fortunately, this integral can be simplified to give the following probability of detection for Swerling 1
with coherent integration of N pulses,

 SNRthreshold 
Pd  exp  
2  (Swerling 1 Coherent Integration)
 1  NA0 /  
2

Non-coherent integration of Swerling 1 follows a similar development. In this case, the conditional pdf
for non-coherent integration is obtained by convolution of N single-pulse conditional pdfs. Since
Swerling 1 is slowly fluctuating, the random amplitude is considered constant for all the pulses being
non-coherently integrated. The signal signal-plus-noise pdf is written

f Z _ NonCoherent ( z | A)  f Z ( z | A)  f Z ( z | A)    f Z ( z | A)

f Z _ NonCoherent ( z )   f Z _ NonCoherent ( z | A) f A ( A) dA
0

The noise-alone pdf is the same as the non-fluctuating case from before. Analysis with these pdfs
becomes involved and will not be presented here. However, a close approximation is given for Pd when
N pulses are integrated non-coherently,

N 1
 1   SNRthreshold 
Pd  1  
2 
exp  
2  (Swerling 1 Non-Coherent Integration)
 NA0
2
/    1  NA0
2
/  

Swerling 2
Swerling 2 target fluctuations have the same pdf as Swerling 1, but the RCS exhibits fast fluctuations
varying from pulse-to-pulse. Coherent integration of Swerling 2 targets is usually not possible since the
phase may also change from pulse-to-pulse. Non-coherent integration is analyzed by convolving the
single pulse pdf N times. In this case, the integral over the random amplitude occurs before convolution.


f Z 1 ( z )   f Z ( z | A) f A ( A) dA
0

71
Fundamentals of Radar Systems D. Lawrence

f Z _ NonCoherent ( z )  f Z 1 ( z )  f Z 1 ( z )    f Z 1 ( z )

Figure 60. Swerling 1 and 3 targets exhibit slow fluctuation, changing only from dwell-to-dwell.

Figure 61. Swerling 2 and 4 targets exhibit fast fluctuation, changing from pulse-to-pulse.

Swerling 3 and 4
For Swerling 3 or 4, the pdf of the target return power is chi-square distributed with 2 degrees of
freedom. This case best describes a target that consists of a dominant scatterer plus many small
scatterers of similar magnitude. The pdf for the amplitude, A, is written

9 A3  3A2 
f ( A)  exp   
2 A04 2
 2 A0 

In this case, the term A02 is proportional to the average value of the received signal power. It is
straightforward to show that this is proportional to the expected value of the normalized signal power,

2 2 1 
A0  E  A2 
3 2 

72
Fundamentals of Radar Systems D. Lawrence

The pdf for the RCS is chi-square,

4  2 
f ( )  exp  
   
2

where the mean RCS is denoted  . The range equation provides the connection between A and .
Specifically, the expected value of the received signal power is written

1  2  P G G 2  P G G 2
E  A 2   A02  E  t t 3 r 4   t t 3 r 4
2  3  ( 4 ) R L  ( 4 ) R L

As before, the signal-plus-noise pdf for a single pulse must take into account the random nature of the
signal amplitude. The single pulse pdf is then obtained by integrating the conditional rician pdf over the
possible values of A weighted by the pdf,


f Z 1 ( z )   f Z ( z | A) f A ( A)dA
0

z   z 2  A 2   zA 
 exp    I 0    f A ( A) dA
2   2     
2
0

z  z2   z2  1 
  1  exp   
2 
 2 1  
X 2
2  2  1 
2
  2 2  1  X2
X 



1 
E  A2  2 2
A
where X  2 
 2  3 0
 2 is the average SNR. This represents the signal-plus-noise pdf for a single pulse
for both Swerling 3 and 4 targets.

73
Fundamentals of Radar Systems D. Lawrence

Swerling 3 and 4 Pdf

1.4

1.2

0.8
f(A)

0.6

0.4

0.2

0
0 1 2 3 4 5
RCS Amplitude / A0
Figure 62. Swerling 3 and 4 RCS models use a chi-square RCS pdf.
Coherent integration of Swerling 3 targets results in an SNR improvement of N. The probability of
detection for coherent integration N pulses is expressed in closed form,

 SNRthreshold   SNRthreshold 
Pd  1  2 
exp   (Swerling 3 Coherent Integration)
 1  2 1  NX   1  NX / 2 
NX

Non-coherent integration for Swerling 3 follows the same approach as Swerling 1 since it exhibits slow
fluctuation, but uses the Swerling 3 amplitude pdf. Swerling 4 exhibits fast fluctuation and coherent
integration is generally not possible. Non-coherent integration for Swerling 4 follows the approach
described for Swerling 2 but with the chi-square amplitude pdf.

In this section, results have been given for a linear detector. If a square-law device is used for detection,
the statistics are slightly different but the Pd, Pfa results are very similar. In some cases, analysis of
square-law devices results in closed-form expressions for non-coherent integration. These expressions
are generally very complicated and will not be pursued here.

74
Fundamentals of Radar Systems D. Lawrence

Surveillance Radar Range Equation

For surveillance radar applications, the standard radar range equation can be re-written in a form that
highlights parameters most relevant to a radar search mode. A basic form of the range equation is
written
Pt GAe  I ( N )
4
Rmax 
(4 ) 2 L  kTBFn  SNR1

The range equation will be rearranged using some basic relationships between parameters. First, note
the antenna gain can be written in terms of the beam solid angle,

4
G
 beam

Also, the total scan time can be written in terms of the ratio of total scan beam solid angle, scan, to the
solid angle of a single antenna beam multiplied by the dwell time.

 scan
t scan   t dwell
 beam

The total dwell time is the number of pulses multiplied by the PRI, or equivalently the inverse of the PRF,

N
t dwell 
PRF

And the average power is written in terms of the total power, pulse width, and PRF,

Pavg  Pt   PRF

Using these relationships, the peak transmit power / antenna gain product can be re-written in terms of
the scan time and scan beam solid angle,

Pavg 4
Pt G 
  PRF  beam
Pavg 4
 t
  PRF  scan t dwell scan
Pavg 4  PRF
 t scan
  PRF  scan N
Pavg 4 t scan

 N  scan

75
Fundamentals of Radar Systems D. Lawrence

Substituting into the standard range equation gives the useful surveillance range equation,

Pavg Ae  I ( N ) t scan

4
Rmax  
(4 )  L  kT ( B ) Fn  N  SNR1  scan

The bandwidth is usually matched to the pulse, B  1

 . Taking the product B to be unity and assuming
coherent integration, I(N) = N, the surveillance range equation can be simplified,

Pavg Ae t scan

4
Rmax  
(4 )  L  kTFn  SNR1  scan

The surveillance equation demonstrates that for a specified scan region and scan time, the parameter
under the control of the radar designer to improve detection performance is the power-aperture
product, PavgAe. The equation also demonstrates that under these conditions there is no explicit
frequency dependence.

Figure 63. Scanning a search volume with narrow beam antenna patterns. For discrete beam steps
(rather than continuous scanning) the radar dwell is tdwell for each fixed beam. The search pattern is
repeated every scan time, tscan.

76
Fundamentals of Radar Systems D. Lawrence

Pulse Waveform Spectrum

The frequency spectrum of a single pulse of width  is a sinc function with firs null equal to 1/. This is
easily derived by taking the Fourier transform of the pulsed waveform. For a single pulse of amplitude A,

t
s(t )  A  rect 
 

The Fourier transform is calculated,

 s(t )e
 jt
Fs ( )  dt


  Ae  jt dt
0

 e  j  1 
 A 
  j 
 j 2  e j 2  e  j 2

 A e  
 2 j  2 
2 sin   2
 A e  j
 2

The normalized amplitude plot of the pulse spectrum is shown below in both linear and power (dB)
units. The most salient feature of the pulse spectrum is the inverse relationship between pulse width
and bandwidth. Narrow pulses produce wide bandwidth and vice versa.

77
Fundamentals of Radar Systems D. Lawrence

Frequency Spectrum of a Pulse

1

0.8

Amplitude Spectrum
0.6

0.4

0.2

0
-4 -2 0 2 4
frequency (1/)
Frequency Spectrum of a Pulse
0

-5
Power Spectrum (dB)

-10

-15

-20

-25

-30
-4 -2 0 2 4
frequency (1/)
Figure 64. Normalized amplitude (left) and power (right) frequency spectra for a single pulse of width .

Pulse Train Spectrum

The frequency spectrum of an infinite train of pulses produces discrete frequency components at the
PRF. The weights of the discrete frequencies are calculated using a standard Fourier series. The Fourier
series coefficients are calculated

 sin n0  2
Fn  A e  jn0 2

T n0  2

where  = 2/T is the fundamental frequency. Note the fundamental frequency is simply the PRF.
When the pulse train is made up of a finite number of pulses, the frequency response has spectral lines
at the PRF spread inversely proportional to the total pulse train length.

78
Fundamentals of Radar Systems D. Lawrence

Figure 65. Pulse train with M pulses.

Figure 66. Pulse train frequency spectrum for M pulses. The envelope follows the single pulse sinc
spectrum with spectral lines occurring at the PRF. PRF lines are spread by the inverse of the total pulse
train duration 1/MT.
Doppler Ambiguities and Blind Speeds
Similar to the range ambiguities introduced by periodic pulses in time, the periodic PRF lines in
frequency introduce Doppler ambiguities. The unambiguous Doppler is simply equal to the PRF. In terms
of velocity, the unambiguous velocity (sometimes called blind speed) depends on both the PRF and
carrier frequency,

c  PRF
vun 
2 f0

It is interesting to note that the PRF and unambiguous range are coupled and often require design
tradeoffs to meet system performance objectives.

c
Run  PRF 
2

79
Fundamentals of Radar Systems D. Lawrence

Figure 67. Periodic waveform consists of discrete PRF lines which cause Doppler ambiguity.

Figure 68. Design tradeoff between first blind speed and unambiguous range.
Matched Filter
The purpose of the matched filter is to maximize the SNR at the output of the filter for optimum
detection performance. The matched filter does not necessarily preserve the shape of the input
waveform, but instead is specifically designed to maximize the SNR. Consider the linear filter H() which
has signal and noise at the input.

Using frequency-domain notation, the filter output can be analyzed by separating the signal and noise
components.

Y ( )  H ( )   X ( )  N ( )
 H ( ) X ( )  H ( ) N ( )
 
Output Signal Output Noise

80
Fundamentals of Radar Systems D. Lawrence

The output signal component in the time-domain is obtained by taking the inverse Fourier transform,


1
 H ()X ()e
jt
so (t )  d
2 

and the signal power is written,

 2
1
 H ( )X ( )e
2
so (t )  j t
d
2 

Since noise is a random quantity, the output noise power must be expressed in terms of power spectral
density. For two-sided noise density, the input white noise power spectral density is written


E N ( ) 
2
 N0
2

The output noise power is calculated by integrating the output noise density over frequency,


1 N0
 H ( )
2
PN  d
2 2 

The output SNR can be written

 2
1
 H ( )X ( )e
jt
2 d
s (t ) 2 
SNRout  o  
PN 1 N0
 H ( )
2
d
2 2 

The Schwartz inequality states that

 2  

 H ( )X ( )e d    X ( )
2 2
jt
H ( ) d  d
  

Applying the inequality to the output SNR expression,

81
Fundamentals of Radar Systems D. Lawrence

2  
 1 
 H ( )  X ( )
2 2
  d  d
 2 
SNRout  



1 N0
 H ( )
2
d
2 2 

1
 X ( )
2
d
2
 
N0
2
2E

N0

Where Parseval's theorem has been used to simplify the integral over the square of the Fourier
transform to be simply the signal energy, E. Recall,

 
1
  x(t )
2 2
E X ( ) d  dt
2  

In order to make the output SNR inequality a true equality, then the optimum filter response must take
the form

H    X   e  jt0

where t0 is the time corresponding to the output peak. This filter is called the matched filter since it is
matched to the input waveform having a conjugate frequency-domain response. The phase term
exp(jt0) allows the filter to be causal. The output signal from the matched filter can be calculated by
substituting the matched filter frequency response and performing the inverse Fourier transform,


1
 X ( ) e j t t0 d
2
so (t ) 
2 

At the peak sampling time, t = t0, the output signal becomes


1
 X ( )
2
so (t 0 )  d  E
2 

In the time-domain, the impulse response of the matched filter becomes

82
Fundamentals of Radar Systems D. Lawrence

 
1 1
h (t )   H ( )e d 
j t
X
*
( )e j t t0  d
2 
2 

 x (t 0  t )
*

For real signals, the conjugation of x(t) has no effect. Using the filter impulse response, the output signal
signal can also be calculated using the convolution integral,

so (t )  x(t )  h(t )

  x( )h(t   )d


  x( )x (  t  t )d
*
0


At the peak sampling time, t = t0, the output signal becomes identical to that calculated using the
frequency-domain approach,


so (t0 )   x( )x ( )d
*




 x( )
2
 d  E


Matched Filter Example: Rectangular Pulse

Consider a pulse waveform shown below having amplitude A and pulse-width . The matched filter
impulse response for this waveform is also shown. Note that t 0   in order for h(t) to be a causal filter.
Convolution of the input pulse with the impulse response illustrates the generation of the output
waveform with a peak at t0. For a pulse input, the matched filter output is a triangle. This result could
also be obtained in the frequency domain where the input waveform has a sinc() shaped spectrum.
Since the matched filter transfer function will also have a sinc() shape, the output of the matched filter
has a sinc2() spectrum. Performing the inverse Fourier transform of a sinc2() results in a triangular
waveform in time.

83
Fundamentals of Radar Systems D. Lawrence

Figure 69. Input rectangular pulse and matched filter impulse response.

Figure 70. Illustration of convolution process to generate matched filter output.

Matched Filter Example: Triangular Pulse

Consider a triangular pulse waveform shown below having peak amplitude A and pulsewidth . The
matched filter impulse response for this waveform is also shown. Convolution of the input pulse with
the impulse response illustrates the generation of the output waveform with a peak at t0. Note that
when convolving with the impulse response, the integral essentially becomes an auto-correlation. Since
the impulse response is a flipped version of the input signal, the convolution process flips the impulse
response again such that the input signal and flipped impulse response produce a correlation. For the
triangular pulse, the matched filter output becomes a quadratic triangular shape.

84
Fundamentals of Radar Systems D. Lawrence

x(t)

Figure 71. Input triangular pulse and matched filter impulse response.

Figure 72. Convolution with matched filter impulse response becomes autocorrelation of input signal.

Ambiguity Function
The ambiguity function is a useful tool in the performance analysis of different radar waveforms. Ideally,
the matched filter implementation in a radar system is perfectly matched to the received waveform
from the target. In reality, however, the received waveform is often shifted in frequency due to the
target motion. The ambiguity function is a measure of the matched filter output when the input is
Doppler shifted relative to the matched filter design. The ambiguity function also provides a measure of
the time sidelobes of the matched filter output for both time and Doppler offsets. When x(t) is the
complex modulation envelope of the signal, the ambiguity function is calculated

85
Fundamentals of Radar Systems D. Lawrence

 x( ) x (  t )e
j 2f d 
 (t , f d )  *
d


The ambiguity function is a function of both time and Doppler frequency and is visualized using a surface
or contour plot. Note that some authors define the ambiguity function to be  2 (t , f d ) , but the un-
squared version will be used here since it is a direct measure of the output of the matched filter. Some
interesting properties of the matched filter include:

1)  (t , f d )   (0,0)  E
2)   (t, f d ) 2 dt df d  E 2

Figure 73. Ideal "thumbtack" ambiguity function has good range/Doppler resolution, no range/Doppler
ambiguities, and constant sidelobes.

The ambiguity function for a single pulse is calculated

sinf d   t 
 (t , f d )  for   t  
f d

86
Fundamentals of Radar Systems D. Lawrence

Figure 74. Ambiguity diagram for pulse waveform.

10

5
Doppler (f d)

-5

-10
-1 -0.50 0.5 1
Delay/
Figure 75. Ambiguity diagram contour for pulse waveform.

87
Fundamentals of Radar Systems D. Lawrence

Figure 76. Ambiguity function plot for fd = 0.

Figure 77. Ambiguity function plot for  = 0.

88
Fundamentals of Radar Systems D. Lawrence

Figure 78. Matched filter output with varying Doppler mismatch.

89
Fundamentals of Radar Systems D. Lawrence

90
Fundamentals of Radar Systems D. Lawrence

Radar Processing

Range, Angle, and Velocity Measurement

Radar Waveforms

Targets in Clutter

Target Glint

Antenna Arrays

SAR Processing

ISAR Processing

91