AV-471

RADAR SYSTEMS
Lecture No 14
Constant False Alarm Rate (CFAR)
Detectors
Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A. Richards, J.
A. Scheer, W. A. Holm, (2010); Section 6.4

01 June 2020 – Academic Week No 07/18

Number of Slides: 14
1
In the Last Lecture
• Target fluctuating models
• Swerling models and their extensions

2
In This Lecture
• Adaptive thresholding or Constant False Alarm Rate (CFAR)
detector

3
Threshold Comparison on Range Cells
• In modern radars, threshold is detected on digital data

• Each digital sample corresponds to a range cell

• Threshold comparison for target detection is then done on
these range cells

4
CFAR Detectors
• A target is considered to be present if a threshold of
received power is crossed
• Threshold is to be set higher than the noise floor and
other interferences
• As the noise floor keeps on changing randomly, the
threshold should be adaptive to these changes
• We can do that!
• The adaptive changing of the threshold can be evaluated
based on keeping the as constant
• Therefore, the name constant false-alarm rate is coined to these
adaptive thresholding techniques
5
CFAR Thresholding

https://www.ll.mit.edu/sites/default/files/outreach/doc/2018-07/lecture%205.pdf 6
Basic CFAR Algorithmic Architecture

• CUT is Cell Under Test,

where the threshold is to
be calculated

• G are Guard cells that are

not used for interference
statistics because the
target (if present in CUT)
may also lie in guard cells

• Leading and Lagging cells

are used for calculating the
interference statistics

7
A 2D CFAR Window

8
Type of CFAR
• There are many types of CFAR techniques
• These include:
• Cell-Averaging (CA) CFAR – for homogenous environments
• Greatest-Of (GO) CFAR a.k.a. GOCA CFAR
• Smallest-Of (SO) CFAR a.k.a. SOCA CFAR
• Censored CFAR
• Order Statistic (OS) CFAR a.k.a. OSCA CFAR
and many others…

9
CA CFAR
• CA CFAR are used in “homogeneous” environment i.e.
• With a target in CUT, no additional targets are present
in the leading and lagging windows that bias the results

10
Effect of Rain on CA CFAR Thresholding

11
Effect of Rain on CA CFAR Thresholding

12
Greatest-Of (GO) CA CFAR

13
Summary
• Adaptive thresholding (or CFAR) adjust the threshold on-
the-go while detecting the targets
• CFAR algorithm architecture
• Two types of CFAR
• CA CFAR
• GOCA CFAR

14
 AV-471
RADAR SYSTEMS
Lecture No 17
Doppler Signal
(in Time and Frequency Domains)
Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A. Richards, J.
A. Scheer, W. A. Holm, (2010); Sections 8.7 – 8.9

15 June 2020 – Academic Week No 10/18

Number of Slides: 17
1
In the Last Lecture
• Notional datacube
• Basics of Doppler processing
• Total phase change

2
In This Lecture
• Doppler resolution
• Doppler Signal in time and frequency domain

3
1. Doppler Resolution (1/2)
• Earlier we noted that greater is the dwell time, better is
the Doppler processing and speed measurement
• Doppler resolution is the ability of a radar to distinguish
two objects moving at different speeds
• Greater is the dwell time, more better can the radar
distinguish two very closely differing speeds

4
1. Doppler Resolution (2/2)
• At detector stage, a square pulse train (time-domain) of
duration echoed from two moving targets at different
range bins gives two sinc functions (frequency domain)
• The Rayleigh bandwidth (peak-to-first-null length) is 1/
for each sinc function
• The two peaks can be reliably
resolved, regardless of initial
relative phase, only if
separated by approximately
the Rayleigh bandwidth or more
Doppler Resolution
5
Example
• Suppose a radar’s CPI has 20 pulses and it is operating at moderate
PRF of 5 kHz. What is its Doppler resolution? What is the “velocity
resolution” of the radar if the pulse frequency is
1. 1 GHz;
2. 10 GHz?

Solution
• PRI 1/5 KHz 0.0002 sec 20 0.0002 0.004 sec

Doppler resolution 250 Hz

.
#$% &'( &$ )
• We have Δ" ⇒ Δ+,
) #$%

⇒ Δ+,- 37.5 m/s and Δ+,0 3.75 m/s 6

Doppler Signal
(time-domain)
1. Measuring Doppler with Multiple Pulses
2. Coherent vs Noncoherent vs Coherent-on-Receive

7
1. Measuring Doppler with Multiple Pulses
• Suppose a target is moving towards the radar with velocity +
• The radar is transmitting 1 pulses with PRI
• Suppose the radar has sent a rectangular pulse with initial phase of
2 before modulation
• At the receiver detector stage, the echo from the target for pulse 3
4 3 will be

where 56 3 and 2′ is the phase change

• Change of phase in each successive pulse gives "
8
Phase Change in Video Pulses

Case of target moving away from radar

9
Moving Target Migrating Range Bins
• Moving target changes range bin during its motion
• However, in one CPI duration, this is rarely the scene
For example,
9:
• For CPI = 10 ms, pulse length of 1 8s, range bin
#
150 m.
• An aircraft is moving towards radar with Mach 2 (680 m/s).
• The aircraft in one CPI time will move only 680 10 ms =
6.8 m

10
2. Coherent vs Noncoherent vs Coherent-on-Receive

• Coherent pulsed radar has

every pulse generated
according to pre-defined
initial phase
• Non-coherent radar does not
have pre-defined pulses
generated according to pre-
defined initial phase
• Coherent-on-receive is non-
coherent on transmit side but
measure the initial phase for
use at the reception
11
Doppler Signal
(frequency domain – DTFT)
1. Notional Doppler Spectrum
1. Stationary Radar
2. Moving Radar

12
 Notional Doppler Spectrum, Stationary Radar
• “Principal period” of the spectrum of a radar
echo signal from one range bin
sampling interval = PRI
• Doppler processing is interesting when
L-1 SNR >> 1, but SCR << 1
sampling interval ≈ 1/β

• Therefore targets cannot be detected based on

fast time (range bin #)

amplitude alone

clutter not all features

Fourier moving are present in
Transform target each range bin

0 noise
0 M-1
slow time (pulse #) F

PRF clear clear PRF

− region
+
2 region skirt clutter skirt 2
region region region

Copyright Mark A. Richards, All Rights Reserved. 13

Notional Doppler Spectrum, Moving Radar
• Suppose the radar is on a moving aerial platform scanning
the Earth. Due to the echoes of side lobes
from clutter exactly normal to the
radar motion. => 0 for level flight

This spectrum
is usually re-
centered by
modulation
with carrier at
=> .

The targets at highest positive

Doppler wrapped around to
the negative Doppler axis.
14
Copyright Mark A. Richards, All Rights Reserved.
Notional Doppler Spectrum, Moving Radar

15
Summary
• Doppler signal gives more insights into stationary vs
moving targets and clutters

16
Thank You

17
 AV-471
RADAR SYSTEMS
Lecture No 18
Continuous Wave Radars

1. Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.

Richards, J. A. Scheer, W. A. Holm, (2010); Section 11.2
2. https://training.ti.com/sites/default/files/docs/mmwaveSensing-FMCW-
offlineviewing_0.pdf

15 June 2020 – Academic Week No 10/18

Number of Slides: 21
1
In the Last Lecture
• Doppler signal in time and frequency domain

2
In This Lecture
• Range-Doppler graphs
• Continuous Wave (CW) radars
• Just this lecture is about CW radars

3
Range-Doppler Maps
1. Range-Doppler Graphs of a Scenario
2. Range-Doppler Graph of Clutter

4
1. Range-Doppler Graph of Scenario

range
3°

20°

20 m

• Clutter around zero Doppler in

each range bin
• Clutter power fades with range
according to either R-3 (pulse
limited) or R-2 (beam limited)
• Moving targets appear in
appropriate range bins, with –PRF/2 0 PRF/2
appropriate Doppler shifts Doppler Frequency FD

• White noise floor throughout

Copyright Mark A. Richards, All Rights Reserved. 5
Simulated Range-Doppler Map

3D surface plot 2D contour plot

Fall 2010 Copyright Mark A. Richards, All Rights Reserved. 6

2. Range-Velocity Graph of Clutter
• Range-Velocity Distribution of a stationary clutter from a
stationary radar at an altitude of 3 km and boresight
intercept at 5.3 km:

7
Continuous Wave (CW)
Radar

8
Continuous Wave (CW) Radar
• CW radar transmits continuous wave (not pulsed) radio
energy
• Since it is not pulsed, there is no maximum or minimum
range
• Although, the broadcast power level impose limit on range
• It is mostly used in bistatic mode

9
 Continuous-
Wave Radars

Unmodulated Frequency-
CW Modulated CW

10
Unmodulated CW radar
• Easy to implement and
available at low cost
• Only detects moving target.
• Range cannot be calculated.
Application examples:
• Widely used in competition
sports i.e.
• racing
• golf
• tennis
• Used by police for finding
speed of roadway vehicles
11
Frequency Modulated CW (FMCW) radar
• Capable of measuring range of target along with velocity
• Application examples:
• Often used as radar altimeter to measure exact height during the
landing procedure of aircraft
• Used in semi-active radar homing
• FM may change frequency of transmitted signals in the
form of
• sinusoidal wave,
• sawtooth wave,
• triangular wave,
• square wave.
12
FMCW radar: sawtooth FM
Range measurement
- The vertical difference is
,
where is the slope / ,
is the frequency of the IF
signal, and is the round-trip
time.
- Then,
/

- Afterwards, the range .

13
FMCW radar: sawtooth FM

Phase difference measurement

- For an object at a distance
from the radar, the IF signal
will be a sinusoid i.e.
sin 2

14
FMCW radar: sawtooth FM
Phase difference measurement
• If an object moves slightly
causing change in delay time
Δ , we have the IF signal as
sin 2

• The change in frequency

2Δ
Δ Δ

• The change in phase

Δ !"
4 Δ
Δ
15
FMCW radar: sawtooth FM
Velocity measurement
- Phase difference between
transmitted and received
chirp is used to find radial
velocity of the target.
4 !"
Δ

Δ
⇒ !
4 "

16
Example
Consider the chirp shown to the right. What
happens if an object in front of the radar changes
its position by 1mm (for 77GHz radar 1mm= λ/4)?
Solution:
Change in phase:
4 Δ
Δ

The IF signal frequency changes as

2Δ
Δ Δ 333.3 Hz

And the radial velocity of the target becomes

Δ
! 24.35 m/s
4 "
17
Ref: https://training.ti.com/sites/default/files/docs/mmwaveSensing-FMCW-offlineviewing_4.pdf
FMCW radar: sawtooth FM
• Maximum measurable velocity

- Unambiguous measurement of velocity ⇒ Δ * .

4 !"
⇒ *

⇒ ! *
4"
Greater is the required !, smaller should the chirps be separated.
18
Example
• How maximum should the two adjacent chirps be
separated for a maximum radial-velocity detection of 200
km/hr for a 77 GHz FMCW radar?

Solution:
" * 17.5 μs
4 !

19
Pulsed vs FMCW Radar (monostatic)
Pulsed radar FMCW radar
• Cannot detect target with • Can detect targets at vicinity.
* /2, where is pulse-
width. • No such requirement.
• Cannot resolve 2 targets
radially separated by /2. • More power consumption for
• Lower power consumption for a given range
a given range • Easily detected (jammed
• Not easily detected (not easily)
jammed easily) • Transmitter interfere with the
• Transmitter does not interfere reception
the reception • Initially designed for short-
• Initially designed for long- range applications
range applications

20
Assignment # 07 (Submit by Sunday)
1. Suppose a radar’s CPI has 17 pulses and it is operating at
moderate PRF of 4 kHz. What is its Doppler resolution?
2. What is the maximum radial velocity detectable by
FMCW radar if its chirps are separated by 20 .s and the
radar is operating at 78 GHz?

21
Summary
• Notional range-Doppler graphs gave insights into clutter
• CW radar:
• unmodulated only can measure the radial velocity of targets;
• modulated (FMCW) can measure both range and radial velocity
of targets.

22
Thank You

23
 AV-471
RADAR SYSTEMS
Lecture No 19
Matched Filter

Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.

Richards, J. A. Scheer, W. A. Holm, (2010); Sections 14.10, 20.1, 20.2

29 June 2020 – Academic Week No 11/18

Number of Slides: 14
1
In the Last Lecture
• Continuous wave radars
• FMCW

2
In This Lecture
• Correlation of signals
• Why use matched filter?
• What is matched filter?
• Implementation of matched filter

3
Matched Filter in Radar

4
Correlation of Signals
• The correlation coefficients of signal (a) with the
remaining signals are shown in red:

1 1 -1

0.961 0.628 0

5
Need of LNA and Detector Stages

Sent signal

Typical
received
signal

6
Detecting Echo
• It is about searching a known signal within an
unknown signal

• A radar sends a known signal

• The known signal reflects from a target along with
noise and other losses
• Radar receives this corrupted signal and treats it
as the unknown signal
• Now radar is to search where is that signal it
transmitted in the received signal
• How to do that?
7
Matched Filtering in Radar
• If we correlate the known signal with the unknown signal for the
whole span of the unknown signal, the point of maximum
correlation is where our transmitted (known) signal lies.
• This is the intuition behind a matched filter.
• Correlation of a complex signal with itself (auto-correlation)
is
∗
0

• Correlation of a signal with a delayed corrupted version of itself

(cross-correlation) is then
∗ ∗
, 1

where is the time delay.

even function
8
 ℎ ∗

Matched Filter as a System Matched

filter

• Matched filter is an LTI system with impulse

response being the complex-conjugate time-
reversal of the input signal s i.e.
∗
ℎ
⇒ℎ ∗

• Then, the output of the system (convolution)

is ℎ ∗

ℎ
∗
2

• Eqs. (1) and (2) are the same.

9
Example: Matched Filtering of a Rectangular Signal

• Suppose a radar sends a gated Continuous Wave

(CW) signal with PRF = 0.1 pps, pulse width of 0.4
sec. The radar collects data for one PRI, in which
there is an echo from a target as well.

• Matched filtering is going to show us clearly the

location of the target.

10
Example: Matched Filtering of a Rectangular Signal

11
Example: Matched Filtering of a Rectangular Signal
• fs = 1000; • subplot(3,1,1)
• ts = 1/fs; • plot(t, x)
• f = 10; % double the pulse width by • title('Signal sent')
making f = 5
• % axis([0 1.2])
• t = 1/fs:1/fs:4/f;
• x = 2*cos(2*pi*f*t); % signal sent • subplot(3,1,2)

• y = cos(2*pi*f*t)+randn(1,length(t)); % • plot(indy, y)
signal plus noise
• title('Signal received for one PRI = 10
• y = [randn(1,5000) y randn(1,4600)]; % sec')
signal received for one PRI = 10 sec.
• % axis([0 1.2])
• indy = (1:1:length(y))*ts; % abscissa
of y • subplot(3,1,3)
• z = xcorr(x,y); % correlation (matched • plot(indz, z)
filtering)
• % axis([0 1.2])
• indz = (1:1:length(z))*ts; % abscissa
of z • title('Correlation (output of matched
filter)')
• [mz,i] = max(z); % finding maximum
correlation point as sample number i
• i = i/fs; % finding maximum correlation
point as time in seconds

12
Summary
• Signals of same shape correlates high
• Matched filter brings out the sent signal in the noisy
received signal
• Matched filter is easily implemented using xcorr in
MATLAB

13
Thank You

14
 AV-471
RADAR SYSTEMS
Lecture No 20
Ambiguity Function and the Need of
Pulse Compression
Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.
Richards, J. A. Scheer, W. A. Holm, (2010); Sections 20.10

29 June 2020 – Academic Week No 11/18

Number of Slides: 17
1
In the Last Lecture
• Matched filter

2
In This Lecture
• Ambiguity function
• The need of pulse compression

3
The Waveform Ambiguity Function (AF)
• The waveform ambiguity function analytically
characterizes the behavior of a waveform using the
matched filter
• It helps in designing waveforms in terms of resolving
range and/or doppler resolution
• Consider a Doppler-shifted signal is input
to the matched filter at the receiver. The convolution
equation of matched filter is

; ∗ ,
4
The Waveform Ambiguity Function (AF)

• AF refers to the magnitude

, .

• AF is a function of two variables:

1. Time delay – relative to the matched filter’s peak;
2. Doppler shift at the receiver.

5
 What’s an Ideal AF?
• The “thumbtack” AF is often cited as the
ideal
– very low output from the filter unless the echo is
closely matched to the Doppler for which the filter
is designed
• implies good Doppler resolution
– a very narrow peak in range
also to get good range resolution
– low sidelobes in
both dimensions
– no additional peaks t
elsewhere in the (t,FD)
plane
FD
Fall 2010 Copyright Mark A. Richards, All Rights Reserved. Module #26 Slide 6
AF of a Simple Pulse (i.e. a Gated CW Pulse)

AF plot of a gated CW pulse

of pulse width and unit
energy.

sin | |
, ,

for " " .

7
AF of a Simple Pulse (i.e. a Gated CW Pulse)

(a) AF with zero doppler mismatch. (b) AF with zero delay.

8
The Need of Pulse Compression
• Suppose a simple gated CW pulse with longer
pulse width is received

9
The Need of Pulse Compression

Hence, to enhance the range resolution (# /2), the pulse

width must be reduced.
10
The Need of Pulse Compression
• The total energy of the simple pulse with pulse width
and instantaneous power $ is:

%
&

• The energy of the signal received after being attenuated

by a factor ' is ' . If the noise variance (power) is ( ,
then the SNR at the receiver is
SNR ∝ . Hence, for
' increasing SNR, the pulse
SNR
(
width must be increased.

11
The Need of Pulse Compression
• Summarizing:
1. Enhance the range resolution by decreasing pulse
width.
2. Enhance the SNR by increasing pulse width.

• Hence, enhancing the range resolution simultaneously

with enhancing the SNR and Doppler resolution are in
contradiction.

• How can one have a large enough pulse (better SNR)

as well as high range resolution?
• Pulse compression help out here.
12
The Objective of Pulse Compression

To make the range resolution and Doppler

resolution independent of one another so that
enhancing one resolution does NOT degrade the
other resolution.

13
Pulse Compression
• The waveform exhibiting pulse compression has

• a long enough length so that the energy budget is correct

(for energy and detection);

• a design such that after matched filtering, the width of the

output signal is smaller in width (for range resolution). In
other words, the autocorrelation function is smaller in width.

14
Pulse Compression

Illustration of reducing the pulse width at the receiver to enhance

the range resolution 15
Summary
• Ambiguity function characterizes mathematically a
radar waveform
• Pulse compression is important for enhancing the
range as well as Doppler resolution at the same time

16
Thank You

17
 AV-471
RADAR SYSTEMS
Lecture No 21
Pulse Compression Waveforms

Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.

Richards, J. A. Scheer, W. A. Holm, (2010); Sections 20.5 – 20.7

06 July 2020 – Academic Week No 12/18

Number of Slides: 23 1
In the Last Lecture
• Ambiguity function
• The need of pulse compression

2
In This Lecture
• Pulse compression waveforms
• Characteristics and types
• LFM waveforms
• Barker-coded waveforms

3
Pulse Compression Waveforms
• The waveforms designed for pulse compression
are termed as the pulse compression waveforms.
Examples include
• Frequency-modulated pulse compression waveform,
• Phase-modulated pulse compression waveform.

• Two primary classes of pulse compression are

employed in radars:
• Analog pulse compression, when the matched filter
is before ADC in analog domain
• Digital pulse compression, where the matched filter
is after ADC in digital domain

4
Pulse Compression Waveforms
• Denote
• The transmitted pulse width as and
• The received compressed (processed) pulse width at
the output of matched filter as . Then,
Pulse Compression Ratio, /
• Characteristics of pulse compression waveforms
are that
1. The bandwidth ≫ 1/ . Hence, ≫ 1;
• Note: For a gated CW pulse, 1. For LFM, 1 (typically 20 to
30).
2. .
• is called the time-bandwidth product
5
Pulse Compression Waveforms
•There are many pulse compression
waveforms in the literature i.e.
• Frequency modulated waveforms
• Linear Frequency Modulated (LFM) waveform
• Stepped Frequency Modulated (SFM) waveform
• Non-linear Frequency Modulated (NLFM) waveform
• Phase modulated waveforms
• Barker-coded waveform
• Frank coded waveform
• Zadoff-Chu coded waveform
•Only the following two will be discussed:
• LFM waveform
• Barker-coded waveform
6
Linear Frequency Modulated
(LFM) Pulse Compression
Waveform

7
Frequency-Modulated Pulse Compression Waveform
• An LFM waveform
can be represented
as:

• The instantaneous
frequency is the
derivative of the
phase function i.e.

1
2

8
9
Ambiguity Function of LFM Waveform
sin " & '
, "
" &

For ' ( ( .

Let us use the ambiguity function to determine

how the range and doppler resolution are
enhanced together

10
LFM Ambiguity Function
• τ = 10 µs
B = 1 MHz
 Bτ = 10

11
Ambiguity Function of LFM Waveform for
Range Resolution
• The zero-doppler cut of the LFM ambiguity function is

sin 1 ' | |/
,0
Increasing the
product
reduces the
pulse width of
matched filter
output and
hence,
enhances the
range
resolution.

12
Concluding the Ambiguity Function of LFM
• At the output of the matched filter (i.e. the ambiguity
function) of LFM shows that
1. The pulse duration (that determines the range
resolution), ,
where is the transmitted LFM signal’s bandwidth.
2. The Doppler resolution Δ " ,
,

where is the transmitted LFM signal’s duration.

The range resolution and Doppler resolution are now

changed by two variables and independently. Hence,
one can enhance both resolutions simultaneously 13
 Phase-Modulated
Pulse Compression Waveform

14
Phase Coded Waveforms
• Phase coded waveforms consist of N contiguous subpulses
• subpulse length = τc τc
• total length = = Nτc

N −1
x(t ) =  xn ( t − nτ c ) =Nτc
n =0
exp ( jφn ) ; 0 ≤ t ≤ τ c
xn (t ) = 
0 ; elsewhere
- .
• Range resolution will be determined by subpulse length: Δ
2
• Each sub-pulse is a “chip” or “bit”
• τc called the “chip length”

15
Types of Phase Codes
N −1
x(t ) =  xn ( t − nτ c )
n =0

• Biphase • Polyphase
• Barker • Frank
• Combined Barker • P1, P2
• Minimum peak sidelobe • P3, P4
• Pseudo-random • P(n,k)

 φn = 0 or π ; exp ( jφn ) ; 0 ≤ t ≤ τ c
exp ( jφn ) , xn (t ) = 
xn (t ) =  0 ≤ t ≤ τc 0 ; elsewhere
0
 elsewhere
16
Barker Code Example

An example of biphase code (Barker code of length / 13) having only two
possible phases (commonly 0 and 180∘ )

It can be polyphase code having more than two possible phases

17
Barker Codes
1
Code Sequence 20 log
•Barker Codes - N
+/- Format
PSL (dB) /
“perfect” 2
2
+−
++
-6.0
-6.0
biphase codes 3 ++− -9.5
3 +−+ -9.5

• Lowest sidelobe 4
4
++−+
+++−
-12.0
-12.0
levels for the 5 +++−+ -14.0

values of N 7 +++−−+− -16.9

11 +++−−−+−−+− -20.8
listed in the 13 +++++−−++−+−+ -22.3
table

18
Ambiguity Surface of Barker Code of / 13

19
13-Bit Barker Matched Filter Output
14
•There is always a zero at
lag 1, so the main-lobe 12

width is τc 10

• therefore range resolution

magnitude
8

is cτc/2 = c/2Β 6

0
-15 -10 -5 0 5 10 15
range bin number

The matched filter output for zero Doppler shift

20
Range Resolution of the PC Waveforms

•Range resolution of LFM and Barker Coded

waveforms is
-
Δ
2

21
Summary
• Pulse compression waveforms enhance both range
and Doppler resolutions simultaneously
• Two famous PC waveforms are
• LFM waveform
• Barker-coded waveforms

22
Thank You

23
 AV-471
RADAR SYSTEMS
Lecture No 22
Doppler Processing: Moving Target
Indication
Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.
Richards, J. A. Scheer, W. A. Holm, (2010); Sections 17.4.1 – 17.4.2

06 July 2020 – Academic Week No 12/18

Number of Slides: 15
1
In the Last Lecture
• Pulse compression waveforms
• For improving range and Doppler resolutions
simultaneously

2
In This Lecture
• Removing clutter from target using Doppler
processing
• Moving Target Indication (MTI)
• Single Delay Line Canceller (DLC)

3
 Doppler Separating clutter
Processing in from targets
Pulse radar

Pulse Doppler
MTI Processing

• MTI radar usually works on low • Pulse Doppler radar usually works
PRF. on high PRF.
• Less informative but simpler to • More informative but more
implement complicated circuit 4
Moving Target Indication (MTI) Radar
• In MTI receiver,
returns from two
consecutive pulses are
subtracted from one
another. First sweep
(echo from first signal)
• Hence,
• echoes from a
stationary target are
cancelled out and
Second sweep
• echoes from a moving (echo from second signal)
target does not
perfectly cancels out
due to change in First sweep minus
frequency by Doppler. second sweep
5
MTI radar block diagram

Stable local oscillator (stalo)

signifies that the oscillator must
produce precise oscillations with
no phase change between
pulses.
Coherent oscillator (coho)
signifies that the phases of
transmitted and reference
signals are the same.

6
Frequency response of single DLC
1
y[m]
+ Σ z[m]
-
0 M-1
slow time (pulse #)

• Transfer function 1

• Frequency response:

1 2 sin
2
7
Frequency response of single DLC
• Periodic with period ω = 2π rad
• corresponds to Hz
• 3 periods shown
2

1.5

0.5

0-1.5 -1 -0.5 0 0.5 1 1.5

Doppler Frequency, (multiples of ) 8
Blind speeds
2

1.5

0.5

0-1.5 -1 -0.5 0 0.5 1 1.5

Doppler Frequency (multiples of )

• If speed of a target induces 0, , 2 , …, then

those targets will also be cancelled by MTI.
• Hence, targets moving at such speeds are invisible at the
output of MTI. These are called blind speeds and are
equated as
"# %#
!
2$ 2 9
Plot of first blind speed

10
Pros and cons of single DLC
• Pros: simpler to implement
• Cons:
1. Blind speeds – is small (low PRF) – ! placed
nearby
2. Clutter Doppler spectrum is not completely
cancelled

• How to avoid these disadvantages? 11

1. How to avoid blind speeds?
1. Operate at low frequencies (low &'
Disadvantages
• low $ has low resolution
• Spectrum of low $ (typically VHF) is already crowded for
communication
2. Operate at high PRF (high &( )
Disadvantage
• Unambiguous range is reduced
3. Operate at different PRFs from pulse-to-pulse
• Target with a blind speed at one PRF will not be blind at another
PRF
• Preferred choice
• This is also called PRF staggering.

12
Assignment # 08 (Submit by Monday)
1. Apply the concepts learned in pulse compression to
estimate the range and Doppler resolution in the
following scenarios:
i. Simple (gated CW) pulse with )* 1 +,
ii. LFM waveform with pulse width )* 1 +, and intra-pulse
bandwidth - 10 MHz

13
Summary
• Doppler processing to remove clutter
• MTI
• Single DLC
• Disadvantages of single DLC

14
Thank You

15
 AV-471
RADAR SYSTEMS
Lecture No 23
Moving Target Indication Filter and
Figures of Merit
Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.
Richards, J. A. Scheer, W. A. Holm, (2010); Sections 17.4

20 July 2020 – Academic Week No 14/18

Number of Slides: 14
1
In the Last Lecture
• Moving Target Indication (MTI)
• Disadvantages
• Blind Speed
• Incomplete clutter rejection

2
In This Lecture
• Incomplete clutter rejection: how to tackle with it?
• Higher order filters
• FIR (finite impulse response)
• IIR (infinite impulse response)
• MTI figures of merit
• Improvement factor
• Clutter attenuation

3
Pros and cons of single DLC
• Pros: simpler to implement
• Cons:
1. Blind speeds – is small (low PRF) – placed
nearby
2. Clutter Doppler spectrum is not completely
cancelled

• How to avoid these disadvantages? 4

2. How to avoid the insufficient clutter attenuation?
• If we use two single DLCs in cascade, their frequency
response becomes

H ( z ) = 1 − 2 z −1 + z −2
y[m]
+ Σ + Σ z[m]
- -
z-1 z-1

The nulls are

widened near
5
-pulse canceller

Coefficients of ∝
…
Single DLC 1 1
sin
2
Double DLC a.k.a. 1 2 1
three-pulse canceller sin
2
Four-pulse canceller 1 3 3 1
sin
2
-pulse canceller Coefficients of the binomial "
" sin
expansion of 1 ! 2

6
Transversal (non-recursive, FIR) filter
• In general, the -pulse canceller is shown by the
figure below:

where #$ are called weights i.e. coefficients of

1 "

7
Recursive (IIR) filter
• With a feedback network, the filter design also
employs the poles along with zeros. Such filter (IIR) is
then called a recursive filter.
• Such filter achieve desirable frequency response with
fewer delay lines than the FIR (non-recursive) filters.

Feedback
coefficients %$
8
MTI Figures of Merit: Improvement Factor
• Improvement Factor is the increase in S/C ratio
at the output of the clutter filter over that at the
input, averaged over all target radial velocities
of interest:

 ( S C )out 
I = E 
 ( S C )in vr

9
MTI Improvement Factor
Improvement factor is the product of the filter’s effect on the
clutter and on the target:
+
' ,-. +,-. '$
)* * * & ⋅ '(,
+ +$ ',-.
' $

where & is the signal power gain and ' ( is clutter

attenuation.

+12 +34 5
&* * * 5
+34 +34
10
Calculating the Gain
As Doppler shifts of targets is not apriori known,
&* 5 is averaged for all targets having
unambiguous Doppler shifts lying in the range
6! 6!
, i.e.
2
1.5
1
5 0.5
0
-1.5 -1 -0.5 0 0.5 1 1.5
π Doppler Frequency (multiples of )
 1 2
G= 

H (ω ) d ω (1)
 2π
11
−π
Calculating the Gain
For a 2-pulse canceller:
2 2
H (ω ) = 4sin (ω 2 )
Putting this in Equation (1), we get
&*2
2
1.5

5 7 1
0.5
0-1.5 -1 -0.5 0 0.5 1 1.5
Doppler Frequency (multiples of PRF) 12
Summary
• Stationary clutter can be attenuated with MTI filters
• Higher the order of MTI filter, more is the clutter
rejection
• MTI figures of merit
• Improvement factor
• Clutter attenuation

13
Thank You

14
 AV-471
RADAR SYSTEMS
Lecture No 24
Staggered PRF and
Pulse Doppler Processing
Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.
Richards, J. A. Scheer, W. A. Holm, (2010); Sections 17.5

20 July 2020 – Academic Week No 14/18

Number of Slides: 18
1
In the Last Lecture
• MTI disadvantages are
• Incomplete clutter rejection upon using a single DLC
• Use higher order transversal (FIR)/recursive (IIR) filters

2
In This Lecture
• MTI disadvantage
• Blind speeds
• Use staggered PRFs
• Pulse Doppler Processing
• Filter banks

3
Staggered PRF

4
Staggered PRF
• The use of different successive PRFs allow the
detection of moving targets that would otherwise be
eliminated with a constant PRF (if they were at blind
speed).

5
Staggered PRF: First blind speed
• We know that
1
2
• The new PRF becomes the LCM of all the individual
PRFs.
• Example: A set of two PRFs is used i.e. 750, 1000
Hz in an X-band (10 GHz) radar.
• What is the first blind speed if any of the existing is used?
Solution: 11.25 m/s for 750 Hz
15 m/s for 1000 Hz
• What is the first blind speed after the PRF staggering?
Solution: LCM 3000 Hz. Then 45 m/s
6
 750, 1000 Hz

750, 1000, 1250 Hz

7
Unambiguous range in staggered PRFs
• The unambiguous range is the range corresponding
to the highest PRF
• shortest of the individual unambiguous ranges

• In the last example,

• with 750, 1000 Hz, the unambiguous range is
150 km.
• with 750, 1000, 1250 Hz the unambiguous range is
reduced to 120 km.

8
Staggered PRF (pulse to pulse)

T1 = 1/PRF1 T2 = 1/PRF2 T1 = 1/PRF1

…
…
τ
• If we use a transversal filter, the frequency response is expressed
as
0 12 3 1 4 56789 3 16 4 5678 9 :9 3 ⋯ 3 1 4 5678 9 :⋯:9<

9
Considerations in designing staggered PRFs

• Highest PRF has to consider the minimum allowable

unambiguous range.
• PRF staggering should not “stress” the transmitter by
employing wide variations in the duty cycle.

10
 Doppler
Processing in
Pulse radar

Pulse Doppler
MTI Processing

• MTI radar usually works on low • Pulse Doppler radar usually works
PRF on high PRF
• More informative but more
• Less informative but simpler to complicated circuit
implement • Remove moving clutters as well
• Removes stationary clutter

11
The Concept of Pulse Doppler
sampling
interval = PRI frequency
samples
computed
L-1 by the FFT
1 complex
(I&Q) sample
fast time (range bin #)

in each cell
interval ≅ 1/B
sampling threshold

FD

• Each frequency sample

is subjected to a
Digital Filter threshold detector
0 or • threshold based on
0 M-1 Spectrum dominant interference at
slow time (pulse #) Analyzer that frequency
12
Doppler Filter Banks
• As the name is highlighting, it corresponds to a
set of attached filters for detecting targets of
different radial velocities.

13
Why to Use Doppler Filter Bank?
• MTI (delay line cancellers) removes stationary
clutters. What if the clutter is also moving?
• For example, detecting an aircraft while raining makes the
rain echoes as the moving clutter.
• Doppler filter banks place each moving target in a separate
output.
• More flexibility in tuning for a desired target
• e.g. if a bird and a target are in the same lobe, adjust the
threshold to reject the bird echoes easily.

14
Doppler filter bank

1=,> 4 5?67 =@ >/BC

where D 1,2, … , F represents

the F taps and G is an index
from 0 to F H 1 that
corresponds to a different set of
F weights, each for a different
filter

15
Frequency response (neglecting side lobes)

Here F I 8
16
Summary
• Staggered PRF helps in increasing the first blind
speed
• Pulse Doppler Processing involves filter banks, in
which each filter can be tuned separately
• Allows more flexibility

17
Thank You

18
 AV-471
RADAR SYSTEMS
Lecture No 25
Digital MTI and MTD

Textbook : Principles of Modern Radar Vol. I: Basic Principles by M. A.

Richards, J. A. Scheer, W. A. Holm, (2010); Sections 17.6

27 July 2020 – Academic Week No 15/18

Number of Slides: 15
1
In the Last Lecture
• MTI disadvantage
• Blind speeds
• Use staggered PRFs
• Pulse Doppler Processing
• Filter banks

2
In This Lecture
• Digital MTI
• Clutter Map
• MTD

3
Digital MTI Processing

4
Digital MTI

5
Blind phases;
Another At single DLC,
0. Hence,
reason to zero Doppler detected.
However, actually the
use I/Q target was moving.
channels This is called blind
phase.

6
Clutter Map
• It is a technique for detection of moving targets with
zero or low Doppler shift
• It is typically used by ground-based scanning radars
such as airport surveillance radars
• The purpose of clutter maps is to detect targets on
crossing paths – that is, passing orthogonal to the
radar so that the radial velocity is zero
• It is effective if the RCS of target is greater than that
of clutter
• For example, to reduce ground clutter, radar antenna is
pointed upward
7
Moving Target
Detector

8
Moving Target Detector
• It is an example of MTI processing system
• Originally developed by MIT Lincoln Lab for local air traffic
control

9
 Moving Target Detector
For removing Weighting to reduce
For removing side lobe level of
moving clutter
stationary Doppler filters
clutter /
magnitude =

CPI = 10
pulses

Adaptively
changing
thresholds for
• For detecting aircraft on a crossing target detection
trajectory with zero radial velocities
• Clutter map then detect targets
based on thresholds 10
MTD: Unmasking aircraft while raining
• Moving aircraft may be masked (suppressed) by rain
echoes.
• Usually (but not always) the target is moving at a
higher speed than the first blind speed of one PRF.
• Sampling rate for Doppler processing (of slow-time
pulses) is equal to PRF.
Aircraft echo alias Aircraft
folded over PRF/2 echo

-PRF/2 PRF/2 11
MTD: Unmasking aircraft while raining
• Staggered PRF helps MTD in separating moving target from moving clutter
even if both occurs in the same filter.
• Suppose rain echo and true aircraft velocities are as shown in the figure
below.

• prf-2 alias of aircraft is masked by rain. Then

• prf-1 alias of aircraft is separated from rain.
12
Limitations to MTI Performance
MTI performance is degraded by factors that include
1) Antenna scanning modulation
• When clutter’s spectrum is spread because of a time-limited
signal.
• Reduce it by having longer look time on the clutter.
2) Internal modulation of clutter
• Echoes from clutters like sea, trees, etc are not perfectly
stationary.
3) Equipment instabilities
• Changes in amplitude, phase, or frequency of stalo or coho.
4) Filter straddling loss
• When signal spectral lines are not centred on the filter.

13
Summary
• Digital MTI
• Clutter Map
• MTD

14
Thank You

15
 AV-471
RADAR SYSTEMS
Lecture No 26
Windowing

Text Book : Principles of Modern Radar Vol. I: Basic Principles by M. A.

Richards, J. A. Scheer, W. A. Holm, (2010); Section 17.5

27 July 2020 – Academic Week No 15/18

Number of Slides: 21
1
In the Last Lecture
• Digital MTI
• MTD
• Clutter Map

2
In This Lecture
• Windowing
• A signal processing technique

3
The DTFT of a Moving Target - 1
• The coherent receiver baseband output of an ideal,
constant-radial-velocity moving target for an M-pulse
dwell is a pure complex exponential:

, 0, … , 1

4
The DTFT of a Moving Target - 1
20
• The DTFT that results is a 18
“digital sinc” or “aliased sinc” 16 M = 20
function 14 FD = PRF/4
12  fD = 0.25
• peak at Doppler frequency

|Y(F)|
10
• −13.2 dB sidelobes 8
• Rayleigh width = 6

• 1/M cycles/sample 4
2
• 1/MT Hz
0
• − 3 dB mainlobe width = -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Doppler Frequency (multiples of PRF)
• 0.89/M cycles/sample
• 0.89/MT Hz

sin
sin

5
Windowing
1
0.6
0.2 1
0
-0.2 0.8
0.6
-0.6
0.4
-1-2 -1.5-1 -0.50 0.5 1 1.5 2 0.2
time (microseconds)
0
-0.2
1 -0.4
Hamming
0.8 window -0.6
-0.8
0.6
-1
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
0.4 time (microseconds)
0.2
0
-2 -1.5-1 -0.50 0.5 1 1.5 2
time (microseconds)
6
Effect of Windows on the Moving Target DTFT
• Response shape changes from digital sinc function to
the Fourier transform of the window:

#
!"
12

20
18 10
16
14
8
12
BEFORE AFTER
|Y(F)|

|Y(F)|
10
8 6
6
4
2
4
0
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Doppler Frequency (multiples of PRF) 2
sin
sin 0
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Doppler Frequency (multiples of PRF) 7
Effect of Window on the Fourier Transform
0
• Good no window
• reduction of -10 Hamming window

sidelobes -20

-30
• Bad
power (dB)
-40
• reduction
in peak -50
• widening of -60
mainlobe
-70
• reduction in SNR -50 -40 -30 -20 -10 0 10 20 30 40 50
frequency (MHz)

8
 Why Do We Care About Sidelobes?
• Sidelobes from a strong target can mask the return from a
nearby (same range, different Doppler) weaker target
• Solution: windowing for sidelobe suppression
0 0

-10
-10

-20
-20

-30

-30
-40

-40
-50

-50 -60

-60 -70
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5

without window with Hamming window 9

Window Metrics
• We can window the data in slow time to reduce
Doppler sidelobes in the usual fashion
• Windows also widen mainlobe  decreased Doppler
resolution
• but this also reduces straddle loss
• Windows also reduce the peak of the DTFT by the
loss in peak gain, or LPG
• Windows also cause a modest signal to noise ratio
loss, called the processing loss, PL

10
Loss in Peak Gain - 1
• LPG describes the reduction in coherently integrated
power when a window is used on the data
• Consider a slow time signal of form
, 0, … , 1

• Peak power in the Doppler spectrum without a

window is just

• With a window it is
"

!" !"
11
Loss in Peak Gain - 2
• The ratio is the LPG:

1
$%&
!"

• In this equation, M = total window length

• It doesn’t matter if some w[m] = 0
• 0 dB for rectangular window, about −5.7 to
−5.4 dB for Hamming
12
Processing Loss
• PL is the reduction in SNR due to
use of a window on noisy data
• It is the ratio of SNR with windowing
to the SNR without windowing

'()*
'()
M −1 2
• Function of the window shape
and length only  w[ m ]
• 0 dB for rectangular, m =0
PL = M −1
−1.7 to −1.34 dB for Hamming 2
M  w[ m] 13
m=0
Sampling the DTFT: The DFT
• The DFT is a sampled version of the DTFT
• sampled at frequencies k(PRF/K) Hz, k = 0,...,K-1
• to get denser set of samples, increase K (zero
padding)
• sample points are fixed on the frequency axis
M −1
Y [ k ] =  y [ m] e − j 2π mk K
, k = 0,K, K − 1
m =0

= Y (ω ) ω = 2π k K = Y ( F ) F = k PRF K

14
DFT Sampling of the DTFT - 1
25

An ideal result y[m] = exp(j2πfDm), 0≤m≤19

for a pure 20 fD = 0.25

|Y(f)| and |Y[k]|

K = 20
complex
exponential 15

input!
10

0
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 f
10 12 14 16 18 0 2 4 6 8 k

15
DFT Sampling of the DTFT - 2
25

Much different
y[m] = exp(j2πfDm), 0≤m≤19
result from 20
fD = 0.275

|Y(f)| and |Y[k]|

very similar K = 20
waveform! 15

10

0
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 f
10 12 14 16 18 0 2 4 6 8 k
16
Straddle Loss
• Straddle loss is apparent loss of gain when sinusoid
frequency does not correspond to a DFT sample
frequency
• DFT samples miss the peak of the DTFT sinc
• Maximum straddle loss occurs when sinusoid is 1/2
bin off center
• With no data window, maximum loss is 3.9 dB and
average loss is 1.5 dB.
• Hamming window reduces maximum loss to 1.7 dB,
average to 0.65 dB
• varies somewhat with window length
17
Windows Reduce Straddle Loss
4
• With no data
window, maximum
3
loss is 3.9 dB and

Straddle loss (dB)

average loss is 1.5 no window
dB. 2

• Hamming window
reduces maximum 1
loss to 1.7 dB,
Hamming window
average to 0.65 dB
0
0 0.1 0.2 0.3 0.4 0.5
Normalized deviation from bin center (bins)

18
Metrics for Some Common Windows
3 dB
Peak Gain
Mainlobe
(dB relative Peak Signal-to- Maximum
Width
Window to Sidelobe Noise Ratio Straddle
(relative to
rectangular (dB) Loss (dB) Loss (dB)
rectangular
window)
window)
Rectangular 1.0 0.0 -13.2 0 3.92
Hann 1.68 -6.3 -31.5 -1.90 1.33
Hamming 1.50 -5.6 -41.7 -1.44 1.68
Kaiser, β= 6.0 1.63 -6.3 -44.1 -1.80 1.42
Kaiser, β = 8.0 1.85 -7.5 -57.9 -2.35 1.11
Taylor, 35 dB, n =5 1.34 -4.4 -35.2 -0.93 2.11
Taylor, 50 dB, n =5 1.52 -5.7 -46.9 -1.49 1.64
Taylor, 50 dB, n =11 1.54 -5.8 -49.8 -1.55 1.60
Dolph-Chebyshev
1.54 -5.6 -50.0 -1.54 1.61
(50 dB equiripple)
Dolph-Chebyshev
1.78 -7.2 -70.0 -2.21 1.19
(70 dB equiripple)
19
Summary
• Windowing is used to
• Reduce the peak sidelobe level
• Reduce the Straddle loss
• But in the process, windowing
• Results in SNR loss
• Results in reduction of main-lobe’s peak

20
Thank You

21
 AV-471
RADAR SYSTEMS
Lecture No 27
Pulse Pair Processing

Textbook : Principles of Modern Radar Vol. I: Basic Principles by M. A.

Richards, J. A. Scheer, W. A. Holm, (2010); Section 17.7

03 August 2020 – Academic Week No 16/18

Lecture slides taken from Professor Mark A. Richards
Number of Slides: 22 1
In the Last Lecture
• Windowing to reduce the side lobe levels

2
In This Lecture
• Application of radar in weather using
• Pulse pair processing

3
 Pulse Pair Processing
• Technique used widely in
meteorological radar for estimating
echo
– power
– mean velocity
– spectral width
• These measures used in turn in
algorithms for detecting severe
weather, accumulated rainfall, etc.
• Often referred to as “PPP”
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 4
 PPP Spectral Model
• PPP assumes the Doppler spectrum of the
data for a given range
• consists of a single moving target
component, plus a noise floor
– no clutter
– no multiple targets
– Usually (always?) assume Gaussian shape
Sy(F)
area =
power spectrum
width

Sn(F) Sw(F) F
white noise floor

F0 mean velocity
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 5
 Autocorrelation Function and
Power Spectrum
• Given a finite sequence of complex data samples
from a given range, y[m], m ∈ (0,…,M-1), the
(deterministic) autocorrelation function is defined
as M − k −1
sy [k ] ≡  y [ m] y∗ [ m + k ]
m =0

• Its Fourier transform is

2
S y (ω ) = F{s y [ k ]} = Y (ω )
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 6
 Estimating Power
• In the time domain:
M −1 M −1
 y [ m] y [ m] =  y [ m]∗ 2
Pˆx = s y [ 0] =
m =0 m =0

– this is really the energy in the finite-

length signal
– dividing energy by time duration
would convert it to power

Copyright Mark A. Richards . All Rights

Fall 2010 Reserved. Module #52, Slide 7
 Mean Frequency, Time Domain
• Suppose the signal is an ideal complex
sinusoid, and consider sy[1]:
y [ m ] = Ae j 2π F0Tm
M −2
s y [1] =  y [ m ] y∗ [ m + 1]
m =0
M −2 M −2
− j 2π F0T ( m +1)
 
2
= Ae+ j 2π F0Tm A∗e = A e − j 2π F0T
m =0 m =0
M −2

2 − j 2π F0T 2 − j 2π F0T
= A e (1) = A ( M − 1) e
m =0
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 8
 Mean Frequency, continued
M −2

2 − j 2π F0T 2 − j 2π F0T
s y [1] = A e = A ( M − 1) e
m =0
amount of phase
rotation per sample
• By inspection, we can obtain
−1
ˆ
F0 =
2π T
arg s y [1] { }

Copyright Mark A. Richards . All Rights

Fall 2010 Reserved. Module #52, Slide 9
 PPP Mean Estimator
−1
Fˆ0 =
2π T
arg s y [1] { }
• This is the pulse pair processor for estimating
mean frequency
– works for more general signals than the pure
sinusoid, provided there is a single dominant
frequency component with adequate SNR
– assumes no aliasing
• of gives you aliased frequency
• Multiply by λ/2 to convert to velocity units
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 10
 Gaussian Spectrum
• Work with continuous-time case
• Assume that Doppler mean is estimated and
removed from data to give new zero-Doppler
data sequence y'(t) (sampled version y'[m])
• Assume spectral energy is Gaussian shaped:
2
A − F 2 2σ F2
S y′ ( F ) = e
2πσ F
• Then so is the continuous-time autocorrelation
function:
2 −2π 2σ F2 z 2
s y′ ( z ) = A e
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 11
 Sampled Version

• Sampled autocorrelation function:

2 −2π 2σ F2 k 2T 2
s y ′ [ k ] = s y′ ( z ) z = kT = A e

• Corresponding DTFT, still in units of Hz:

2
A − F 2 2σ F2
S y′ ( F ) = e
2πσ F T
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 12
 PPP Spectral Width Estimator
• Assuming a Gaussian spectrum, we can
estimate its variance from the
autocorrelation lags:
2 −2π 2σ F2 T 2 −2π 2σ F2 T 2
s y′ [1] = A e = s y′ [ 0 ] e 

2 1  s y′ [1] 
σˆ F = − 2 2 ln  
2π T  s y′ [ 0] 

Copyright Mark A. Richards . All Rights

Fall 2010 Reserved. Module #52, Slide 13
 Simplified Width Estimator
• Use a series expansion of ln{ } to get
rid of the logarithm calculation:
2 3
x −1 1  x −1 1  x −1
ln x = +   +   +L
x 2  x  3 x 
x −1 1
≈ = 1−
x x
• Apply to width 2 1 
 s y ′ [ 0 ] 

estimator: σˆ F = − 2 2 1 − 
2π T  s y′ [1] 
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 14
 Frequency-Domain PPP

• There are equivalent estimators for

power, mean frequency, and spectrum
width that operate on the power
spectrum instead of the autocorrelation
function
• See text for details

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #52, Slide 15
 NEXRAD Processing Algorithms
• The WSR-88D
(“NEXRAD”) radar of the
National Weather Service
uses pulse pair processing
for power and spectral
moments:
– power: average between 24
and 512 samples, 1 dB
accuracy
– velocity: 40 to 200 samples,
σv ~= 4 m/s
– spectral width: 40 to 200
samples, σv ~= 4 m/s
Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 16
 Nice NEXRAD Display

Reflectivity (power) Velocity

Copyright Mark A. Richards . All Rights
Fall 2010 Reserved. Module #52, Slide 17
 Tornado Signatures

Copyright Mark A. Richards . All Rights

Fall 2010 Reserved. Module #52, Slide 18
 Hurricane Katrina reflectivity 8/29/05 08:49:26 CDT
NEXRAD data, KLIX radar (New Orleans)

Copyright Mark A. Richards . All Rights

Fall 2010 Reserved. Module #52, Slide 19
Hurricane Katrina velocity 8/29/05 08:49:26 CDT
NEXRAD data, KLIX radar (New Orleans)

Copyright Mark A. Richards . All Rights

Fall 2010 Reserved. Module #52, Slide 20
Summary
• Pulse pair processing refers to estimating the
• power
• velocity
• spectral width
of weather phenomena for meteorological applications

21
Thank You

22
 AV-471
RADAR SYSTEMS
Lecture No 28
Radar Tracking

Reference : Introduction to Radar System by M. I. Skolnik, 3rd Edition,

(2001); Section 4.1 – 4.3

03 August 2020 – Academic Week No 16/18

Number of Slides: 14
1
In the Last Lecture
• Pulse pair processing for application of radar in
meteorology

2
In This Lecture

3
Tracking with Radar
• After a target is initially detected, the radar must continue
to detect the target, hence radar can track the target.

4
Tracking Tasks

5
Tracking Tasks

6
Tracking types
• Tracking can be of
A. Angle
B. Range

7
A. Angle Tracking
• Two beams are
used to adjust the
boresight direction
in a single spherical
angle (azimuth or
elevation)

8
Angle Tracking
• Beams illuminated at different angles can be
1. Time shared (not simultaneous)
Examples: conical scan, sequential lobbing

2. Simultaneous
Example: monopulse

9
Sequential Lobbing Tracking

• Time sequence of beams directed around track location (two shown above)
• Reuses single receiver hardware for multiple beams
• Control loop redirects track location to equalize the beam response

10
Conical Scan Tracking – World War II

Advantage over lobe switching was that it gave a 2-D view

11
Conical Scan Tracking

target

that you are off the axis

12
Summary
• Radar tracking
• After detection
• Predicting the next location of a moving target
• Beam displacement
• Angle scanning
• Time-shared
• Sequential lobbing
• Conical scan
• Simultaneous
• Monopulse (next lecture)

13
Thank You

14
 AV-471
RADAR SYSTEMS
Lecture No 29
Radar Tracking II

Reference : Introduction to Radar System by M. I. Skolnik, 3rd Edition,

(2001); Sections 4.1 – 4.3, 4.6, 4.9

10 August 2020 – Academic Week No 17/18

Number of Slides: 21
1
In the Last Lecture
• Tracking was introduced
• Flow chart
• Time-shared angle tracking

2
In This Lecture
• Continuing the tracking topic
• Simultaneous tracking (monopulse)
• Amplitude based
• Phase based
• Range tracking
• Predictions in tracking

3
Monopulse Tracking
• Monopulse angle estimation
compares two or more
simultaneous receive beams
four feeds
• Monopulse improves example
performance over conical
scan and sequential lobing
whose performance degrade
with time varying radar
returns
• Monopulse measurements
can be made via two methods
1. Amplitude-comparison
2. Phase-comparison
4
1. Amplitude Comparison Monopulse

• Amplitudes of signal from the two feeds are added

and subtracted to get error signal
• Error signal re-steer the antenna boresight on to the
target

5
1. Amplitude Comparison Monopulse

4-port
microwave
device
2-inputs
and 2-
outputs

wider main lobe 6

Sum and Difference by Hybrid Junction

7
Two Dimensional Monopulse

8
Block diagram of Two Dimensional
Monopulse

9
2. Phase Comparison Monopulse

10
Range and Velocity
Gate Tracking
Section 4.6

11
B. Range Tracking
• Early days witnessed the use of an operator to
manually watch and track a target by looking at the A-
scopes.
• It was soon replaced by automatic tracking system
called split-gate tracker.

12
Split-Gate Tracker

Courtesy: Electronics World

13
Prediction in Tracking
Section 4.9

14
Prediction in Tracking
• On a series of past target detections, the tracker
makes a “smoothed” (filtered) estimate of the target’s
present position or velocity.
• The tracker does the following:
• estimate the future position/velocity of the target;
• correct its estimates at each iteration (e.g. minimise the
mean square error)
• Some smoothed estimators include
• - tracker
• Kalman filter
• Multiple hypothesis tracker (MHT)
15
 - tracker
• - tracker was an earlier and simpler estimator (predictor) for
estimating the target’s future position.

where ⋅ and ⋅ are the position and velocity by filter,

⋅ and ⋅ are the measured and predicted position,
and are the position and velocity smoothing
parameters.
• This tracker does not inherently handle a manoeuvring target.
16
Example
• Maroon = 1800
actual 1600 α=0.5, β=0.3
trajectory
1400

1200
• ◊=
Position (m)

1000
measured
positions 800

600

• Green = 400
tracker 200
estimate
0
( x̂ )
-200
0 10 20 30 40 50 60
Time Step 17
Kalman Filter
• More sophisticated tracker than the - tracker
• Can predict trajectory of manoeuvring targets as well
• Most modern radar trackers use this recursive
Kalman filter
• Kalman filter assumes probabilistic models for
• the measurement error;
• target’s trajectory;
• disturbances in the target’s trajectory e.g.
• target manoeuvres,
• atmospheric turbulence,
• neglecting higher order derivatives in the model,
• radar system’s limitations i.e. calibration, beam width, etc.
18
Kalman Filter
• When the Kalman Filter is modelled with
• Straight line for the target’s trajectory;
• White Gaussian with zero mean for the measurement noise
and the trajectory disturbance,
the Kalman Filter equations reduce to the - tracker
equations.

19
Summary
• Radar tracking
• Basic flow chart
• Angle tracking
• Simultaneous
• Non-simultaneous
• Phase tracking
• Predictors
• - tracker
• Kalman filter

20
Thank You

21
 AV-471
RADAR SYSTEMS
Lecture No 30
Radar Antennas
References :
[1] Introduction to Radar System by M. I. Skolnik, 3rd Edition, (2001);
Sections 9.7, 9.8
[2] First Course in Radar Systems by Dr. Robert O'Donnell, IEEE
Aerospace and Electronic Systems Society, Online Videos and
Lectures, (2013)

10 August 2020 – Academic Week No 17/18

Number of Slides: 21
1
In the Last Lecture
• Radar tracking

2
In This Lecture

Focus on phased array antennas

3
Radar Antenna
• Antenna is “mean for radiating or receiving radio waves”
(IEEE STD 145-1983)
• Transitional structure between guiding device and free
space.

4
Antenna Gain

• Radiation intensity is power density over sphere (watt/steradian)

• Gain is radiation intensity over that of an isotropic source
• The narrower is the beam, the larger is the radar coverage given
same transmitted power 5
Antenna Pattern

6
Antenna Pattern Characteristics

Aperture diameter, D = 5 m Gain, G = 24 dBi

Frequency, 300 MHz Sidelobe level = -18 dB
Wavelength, 1m HPBW = 120

7
Effect of Aperture Size on Gain

8
 ALTAIR MMW
Reflector comparison
example:
Kwajalein Missile
Range
https://www.ll.mit.edu/about/facilities/reagan-test- 9
site
 Kwajalein Atoll

ALTAIR MMW
Reflector comparison
example:
Kwajalein Missile
Range
https://www.ll.mit.edu/about/facilities/reagan-test- 10
site
Phased Array Antennas

• Radiation pattern is determined by the amplitude and phase of the

11
current at each element.
Phased Array Antennas offer…
Advantages
• Agile, rapid beam-steering
• Flexible
• Multiple-target tracking

Disadvantages
• Expensive
• Complex

12
Phased Array Antenna Types
• Geometrical configurations:
• Collinear(a.k.a. linear), planar, circular, triangular

• Arrays can be
• Broadside
• End-fire

Collinear array elements

13
Array Controls
• Phased array has many factors through which the beam
of antenna can be controlled.

• Relative time delays

between the elements 14
Increasing Array Size: by adding elements

15
Increasing Array Size: by adding space

16
Planar Arrays

17
Mutual Coupling

18
Phased Array Antenna Generations
• Passive Electronically
Scanned Array (PESA)
• All antenna elements are
connected to a single
transmitter and/or receiver
• Example: Microwave Landing
Microwave Landing
System System
• Active Electronically
Scanned Array (AESA)
• Also referred to as the second
generation PESA
• Each antenna element has its
own transmit/receive module
• Example: PAVE PAWS,
CAPTOR CAPTOR PAVE
19
PAWS
Summary
• Radar antenna
• Parabolic reflectors
• Phased array

20
Thank You

21
 AV-471
RADAR SYSTEMS
Lecture No 31
Introduction to Electronic Warfare
References : EW 102: A Second Course in Electronic Warfare by D. L.
Adamy, (2004)

17 August 2020 – Academic Week No 18/18

Number of Slides: 24
1
Electronic Warfare
Electronic battlefield to gain superiority over the
enemy

“Any action involving the use of the electromagnetic

spectrum (EM spectrum) or directed energy to control
the spectrum, attack an enemy, or impede enemy
assaults.”

“The next war will be won by the side that best exploits
the electromagnetic Spectrum”
[Soviet Admiral Sergei Gorshkov, 1973]
2
Purpose of EW
• “To deny the opponent the advantage of, and ensure
friendly unhindered access to, the EM spectrum.”

• EW can be applied by manned and unmanned systems

from
1. air,
2. sea,
3. ground,
4. space,
5. cyberspace.

• EW can target humans, communication, radar, or other

assets (military and civilian).

3
 EW

Electronic Electronic Electronic

Support Counter Counter
Measure Measure Counter
(ESM) (ECM) Measure
(ECCM)
Defined by

Electronic Electronic Electronic

warfare Attack (EA) Protection
NATO

Support (EP)
(ES)
4
Electronic Attack (EA)
ECM signal will refer to the “attacking signal”.

5
(A) Electronic Attack (EA)
• Previously known as Electronic Countermeasures
(ECM)
• EA is the offensive use of electromagnetic (EM)
spectrum or anti-radiation weapon for the purpose of
• degrading
• neutralizing and/or
• destroying
enemy combat capability.
• Types
• EM: Attacking enemy in EM domain by jamming or
deception.
• Anti-radiation weapon: EA refers to missiles/bombs that
are guided by signals to follow a path to destroy target.
6
EA type: EM
• The offensive use of EA type:
EM energy EM

Active
Passive
(Jamming)

Chemical Mechanical

Noise Deceptive
Jamming Jamming Absorbent • Chaff
paints • Decoys
• Metallic
• Spot • Range shaping
• Barrage • Angle
• Sweep • Velocity
7
Passive EM: Mechanical
• (Or mechanical jamming) reflects radar signals to produce
false target returns.
Examples
Chaff thrown Decoys thrown by aircraft are
by aircraft to manoeuvring objects to deceive radars.
swamp the radar • Corner reflectors are usually used
screen with because of higher RCS.
echoes

8
Noise jamming (concealment)
• Broadcasts white noise at higher jamming
(ECM signal) to signal (victim radar’s echo
signal) ratio (J/S) to overpower enemy’s radar
• Attack enemy radar receiver with a directed
beam
• Considers enemy radar’s bandwidth
• Spot jamming occurs when a jammer focuses all of its
power on a single frequency.
• Sweep jamming is when a jammer's full power is shifted
from one frequency to another.
• Barrage jamming is the jamming of multiple frequencies at
once by a single jammer.
9
Deceptive jamming
• A.k.a. self-screening jamming a.k.a. Deceptive ECM
• Does not intend to overpower the enemy radar
• Rather it creates false information about the real target
• Hence, deceives the enemy radar into false target information
Deceptive jamming might be
• Range gate pull-off break the
enemy’s range lock-on
• learns enemy radar signal
• sends false echoes
• progressively changes the range
gate
• Velocity gate pull-off (Doppler
Remember split-gate tracking?
radars) break the enemy radar’s
velocity tracking
• Angle deception jamming (obsolete) break the enemy radar’s
10
angle tracking
EA type: Anti-radiation weapon
example: Radar homing
• Homing (of a weapon) means that the weapon has
an electronic system that enables it to find and hit a
target automatically.
• Radar homing weapon is a weapon having a radar
system for automatic guidance.
• Radar homing may be
• Active
• Semi-active

11
Active Radar Homing
• A missile contains a radar transceiver to find and track
its target independently
• Advantages
• Higher kill probability
• Fire-and-forget capability
• Disadvantages
• More expensive
• Lower range because of
small antenna size and
battery run PL-12 SD-10A on JF-17
thunder
Mass: 180 kg
Range: 70 – 100 km 12
Semi-Active Radar Homing (SARH)
• In SARH, a missile only
possess a radar receiver
and no transmitter.
• Uses CW radars
• Advantages:
• Simpler in design (low
weight)
• Can have higher range
• Disadvantage:
• Less accurate than active
radar homing

13
 EA type: Anti-radiation weapon example:
IR homing

IRIS-T infrared homing air-to-air FIM-92 Stinger infrared homing surface to air
missile missile
Christopher O'Quin, U.S. Marine Corps -
http://www.marines.mil

• Playing video file: AV-471_14 - EW Videa IR EA

14
Electronic Protection
(EP)
ECCM signal will refer to the signal for protecting against
the enemy.

15
(B) Electronic Protection (EP)
• Previously known as
• Electronic Protective Measure (EPM) or
• Electronic Counter Counter Measure (ECCM)
• EP is the ability to defeat EA.
• EP are the actions taken to ensure friendly (defensive)
use of EM spectrum under hostile environment.
• Electronic protection can be
• Anti-active (protection against active EA)
• Anti-passive (protection against passive EA)
• Concepts
• Overcome jamming
• Pattern recognition to distinguish the deception from the real
target 16
Overcome Jamming
Cannot jam the radar
1. More Powerful ECCM
• Use a more powerful
transmitter than the
enemy jammer to ‘burn-
through’ (override) the Jams the radar
jammer.
• Radar burn-through
range is the range at
which the strength of the
radar echo is dominant
over the jamming signal.
• It is different from
crossover range, which
is the range when
jamming signal power
equals victim’s echo
signal power i.e. J = S 17
Reference: http://www.tscm.com/burnthru.pdf
Overcome Jamming
2. Frequency agility a.k.a. frequency hopping
• Continuously change the frequency of the transmitted radar
signal to foil the jammer
• Use spread spectrum technique to deliberately increase the
bandwidth of the transmitted signal
• It is also useful against barrage jamming
3. Polarization agility a.k.a. polarization hopping
• The ECM signal in one polarization would be reduced and the
echo signal would dominate.
4. Anti-radiation missiles (ARM)
• Throw missiles to detect and target the enemy EM source
5. Artificial intelligence – pattern recognition
• Distinguish (classify) actual targets from false targets
18
(C) Electronic Warfare Support (ES)
• Previously known as Electronic Support Measure (ESM)
• ES refers to EW actions taken under direct control of an operational commander
• ES provides the necessary information required for decisions involving EA and EP
• ES gathers intelligence through passive “listening” to EM
• Three keywords to discuss are:
i. Electronic Warfare Support (ES) – immediate action
• Collects enemy signals to immediately take actions
• Collects real-time less amount of data
• Less extensive processing with high throughput
ii. Signals Intelligence (SIGINT) – strategic action
• Collects enemy signals for long period of time to understand their systems
• It has two types COMINT and ELINT.
Communication Intelligence (COMINT)
• Collects communication signals from enemy to extract information
• Collects data for long time for profound understanding of the enemy communication signals
Electronic Intelligence (ELINT)
• Collects non-communication signals from enemy to extract information about enemy’s systems
• Collects data for long time for profound understanding of the enemy EM systems

19
Electronic Warfare Evolution
• Electronic warfare is a very wide field of study, which
is hard to keep up with because
• it is very rapidly evolving;
• most of its technology is classified.
• Example:
• Unmanned Aerial Systems (UAS) or drones are
autonomous EA system that currently pose a growing
threat.
• EP against UAS is expensive. For instance, low-flying
small-size drones are hard to track and home-in and may
cause collateral damage.

20
Summary
• Electronic Warfare is a very important field of study
for military applications in a country
• Three branches
• EA
• EP
• ES

21
 AV-471
RADAR SYSTEMS
Lecture No 32
Advanced Topics in Radar
References : Principles of Modern Radar Vol. I: Basic Principles by M.
A. Richards, J. A. Scheer, W. A. Holm, (2010), Chapter 20

17 August 2020 – Academic Week No 18/18

Some slides are taken from Professor Mark A.
Richards with his permission
Number of Slides: 22
1
In the Last Lecture
• Electronic Warfare and its applications

2
In This Lecture
• Advanced and modern topics in radar systems:
• Synthetic Aperture Radar
• Artificial Intelligence in radars

3
 Albuquerque Airport
• 3 meter resolution, Ku band (15 GHz)
– http://www.sandia.gov/RADAR/sar_sub/images/

3 m SAR Optical
Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 4
 What the Albuquerque Airport Might Look
Like on a Cloudy Night

3 m SAR Optical
Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 5
 Advantages of Radar Imaging

• Images can be acquired day or night

• Images can be acquired through cloud cover
• And in some cases
– Images can be acquired through foliage
– Images can be acquired from subterranean
features

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 6
 Washington, DC at Night in a
Snowstorm

Image courtesy of Sandia National Laboratories

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 7
 NYC & Long Island, Optical & Radar
• Collected by
SIR-C, April
1994

• Optical @ 3:00
AM April 20

• SAR @ 3:00
AM April 16
http://www.geo.hunter.cuny.edu/terrain/radar1.html

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 8
 Radar Image Applications
• Reconnaissance • 2D and 3D
– Target Planning Cartography
– Battle Damage • Earth Resources
Assessment – Terrain Classification
– Obstacle Detection – Oceanography
for Friendly Troop
– Land-Use
Movements
Classification
– Change Detection
• RCS Analysis
• Semi-Fixed Target
Detection • and more …

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 9
SAR is About Cross-Range Resolution
• A good 2D image requires good resolution in both dimensions
• Range (a.k.a cross-track) resolution is obtained with appropriate high-
bandwidth waveforms and matched filtering (pulse compression)

• SAR is a technique for obtaining good resolution in the other dimension,

namely cross-range (a.k.a. along-track)

cross-range (x)
range (y)

Image courtesy of Sandia National Laboratories

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 10
 SAR is About Imaging Stationary “Clutter”
• Basic SAR is about imaging terrain
– often considered the “clutter” in many other radar modes
• The “target” is therefore stationary
– moving objects in scene (vehicles) require special treatment
– space-based SAR must deal with earth rotation
• Clutter-to-noise more relevant than signal-to-noise

Image courtesy of Sandia National Laboratories

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 11
 Cross-Range Resolution in
Conventional Radar
λ
cross-range

θaz θ az =
range
Daz
R0

∆CR ≈ Rθ az
cross-range

θaz Rλ
=
range Daz
R0

• At a fixed range, two scatterers are considered to be just at

the point of being resolved in cross-range when they are
separated by the width of the (3 dB) main beam
Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 12
 “Real-Beam” Cross-Range Resolution
is No Good for Mapping
• For realistic parameters, cross-range resolution
is large compared to range resolution
– At X band, 10 km, Daz = 1 m, we get ∆CR = 300 m
– At C band (5 GHz), low earth orbit (770 km), Daz =
10 m  ∆CR = 4.6 km
• Cross-range resolution degrades as range
increases
• But resolution improves with larger apertures
• (Range resolution is determined by waveform
bandwidth; not affected by SAR processing)
Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 13
 What Do We Have to Do to Get Good
Cross-Range Resolution?
∆CR = Rλ D az
• Example: airborne multimode radars typically operate
at X band (10 GHz, λ = 3 cm) with antennas on the
order of Daz = 1 m across
• To achieve ∆CR = 3 m at a nominal range of R0 = 10
km we have to have
– λ = 0.3 mm (1,000 GHz = 1 teraHz), or
– Daz = 100 m (one football field)

antenna (to scale!)

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 14
 The Synthetic Aperture Concept:
Synthesizing a Virtual Large Array
future transmit positions
X X X X

• The physical antenna is one “element” of the synthetic

array
• Data is collected at each position sequentially, and then
processed together
• Effective aperture size is determined by the distance
traveled while collecting a data set

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 15
Physical vs. Synthetic Array Data Collection
• Physical array radiates from all elements at
once, collects at all element locations

• Synthetic array collects from same locations, but

over time
– and doesn’t radiate from all locations at once

Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 16
 future transmit positions
X X X X

Forming

low cross-range

range bins
resolution
One Line
of High
Cross- Signal
Processor

Range
cross-range

range bins
resolution

Resolution
high

Samples
Fall 2010 Copyright Mark A. Richards . All Rights Reserved. Module #67, Slide 17
Artificial Intelligence in
Radar: Examples
- Multi-objective optimization algorithms in phased-
array radars
- Classification using radar returns

18
Phased Array Radars as Multi-
Objective Optimization Problems
• Phased array antenna has a lot of parameters that
can control its beams
• For example, number of elements, its configuration, current
input to each antenna element, etc.
• Multi-objective optimization algorithms like genetic
algorithm are suitable to find efficient solutions to
• Gain required beamforming with lower number of active
antenna elements, hence reducing the power consumption
• Achieve side-lobe level reductions with optimum main-lobe
beam-width

19
Classification using Radar Returns
• Classification is a task in machine learning, which is a
sub-field of artificial intelligence
• Classification involves training an algorithm on part of a
dataset and then launching that algorithm’s model to
distinguish objects automatically by computer itself
• In radars, the datasets are the received echoes or images
in the case of SAR
• Different objects/phenomena have different radar
signatures
• Hence, classification techniques are applied on received radar
data
• For example,
• SAR data can be used to classify land-covers
• Radar echoes are used to classify different objects
20
Summary
• Radar imagery has advantages over optical
imageries
• SAR is the technique to provide practically-viable
high-resolution radar images
• Artificial intelligence techniques in radars are
increasingly adopted

21
Thank You

22