Introduction to Radar Systems

Detection of Targets in Noise

and
Pulse Compression Techniques

Radar Course_1.ppt
MIT Lincoln Laboratory
ODonnell 6-18-02
 Disclaimer of Endorsement and Liability

• The video courseware and accompanying viewgraphs presented on this

server were prepared as an account of work sponsored by an agency of the
United States Government. Neither the United States Government nor any
agency thereof, nor any of their employees, nor the Massachusetts Institute
of Technology and its Lincoln Laboratory, nor any of their contractors,
subcontractors, or their employees, makes any warranty, express or
implied, or assumes any legal liability or responsibility for the accuracy,
completeness, or usefulness of any information, apparatus, products, or
process disclosed, or represents that its use would not infringe privately
owned rights. Reference herein to any specific commercial product,
process, or service by trade name, trademark, manufacturer, or otherwise
does not necessarily constitute or imply its endorsement, recommendation,
or favoring by the United States Government, any agency thereof, or any of
their contractors or subcontractors or the Massachusetts Institute of
Technology and its Lincoln Laboratory.

• The views and opinions expressed herein do not necessarily state or reflect
those of the United States Government or any agency thereof or any of
their contractors or subcontractors
Radar Course_2.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Detection and Pulse Compression

Propagation Waveform
Transmitter
Medium Generator

Signal Processor
Target
Cross
Section Pulse Doppler
Antenna Receiver A/D
Compression Processing

Main Computer
Console /
Tracking & Display
Detection Parameter
Estimation
Recording

Radar Course_3.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Outline

• Detection of Target Echoes in Noise

– Basic Concepts
– Integration of Pulses
– Fluctuating Targets Issues
– Adaptive Thresholding Techniques
• Pulse Compression

Radar Course_4.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Target Detection in the
Presence of Noise
15 Detectable Noise
Targets
Marginal
Relative Power (dB)

Threshold
10
Undetectable
5

-5

-10

-15

-20
0 10 20 30 40 50 60 70 80 90 100
Range Gate
• The radar return is sampled at regular intervals with A/D (Analog to
Digital) converters
• The sampled returns may include the target of interest and noise
• A threshold is used to reject noise
Radar Course_5.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
321-00395
DPC 9/8/2008
 The Detection Problem

• The area under the noise

0.6
probability curve, from
the detection threshold
Probability Density

Noise
0.5 Detection
Threshold
to infinity (way, way out
to the right) is the
0.4
probability of false alarm.
0.3 • The entire area under the
noise density curve is 1.
0.2

0.1

0
0 1 2 3 4 5 6 7 8
Probability
of False Noise Voltage
Alarms

Radar Course_6.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 The Detection Problem

0.6

Detection
Probability Density

Noise
0.5 Threshold PD = Detection Probability

0.4

0.3 Signal + Noise

SNR = 15 dB
0.2

0.1

0
0 1 2 3 4 5 6 7 8
Probability
of False Voltage
Alarms

Radar Course_7.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Detection Examples with Different SNR

Signal-to-Noise Ratio = 15 dB Signal-to-Noise Ratio = 20 dB

0.6 Noise 0.6 Noise

Probability Density

0.5 0.5 Detection

Detection
Threshold Threshold
0.4 0.4
Signal + Noise
0.3 Signal + Noise 0.3
Higher PD
0.2 0.2 (Detection
PD Probability)
0.1 (Detection 0.1
Probability)
0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Voltage Voltage
Probability
of False For a fixed threshold, a higher SNR (or S/N) will result in
Alarm
a higher of probability of detecting the target

Radar Course_8.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Probability of Detection vs. SNR

Numbers to Remember

Figure by MIT OCW.

Radar Course_9.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Outline

• Detection of Target Echoes in Noise

– Basic Concepts
– Integration of Pulses
– Fluctuating Targets Issues
– Adaptive Thresholding Techniques
• Pulse Compression

Radar Course_10.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Integration of Radar Pulses

• Improve ability of radar to detect targets by combining the

returns from multiple pulses
• Coherent Integration
– No information lost (amplitude or phase)
• Non-coherent integration techniques
– Some information lost (phase)
– Non-coherent (video) Integration
– Binary Integration
– Cumulative detection
– For most cases, coherent integration is more efficient than non-
coherent integration

Radar Course_11.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Coherent Integration

• Real and Imaginary (In-phase and Quadrature) parts of the

complex radar return are added, and the magnitude of the
voltage is calculated
– V=(I2 + Q2 )1/2
• This quantity is then thresholded
• The coherent integration gain is equal to the number of pulses
coherently integrated
– 2 pulses 3 dB
– 10 pulses 10 dB
– 20 pulses 13 dB
• For this gain to be realized, the noise samples, from pulse to
pulse must be independent
– The background noise is white Gaussian noise

Radar Course_12.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Noncoherent Integration
Steady Target
Single Pulse
5
4
3
2
Normalized Power

1
0
8 Pulses Noncoherently Averaged
5
4
3
2 SNR Unchanged
1
0
0 20 40 60 80 100
Range Gates
Noise Variance Reduced after Integration (Allows Lower Threshold)

Radar Course_13.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Different Types of Non-Coherent
Integration
• Non Coherent Integration – General (aka video integration)
– Generate magnitude for each of N pulses
– Add magnitudes and then threshold

• Binary Integration
– Generate magnitude for each of N pulses and then threshold
– Require at least M detections in N scans

• Cumulative Detection
– Generate magnitude for each of N pulses and then threshold
– Require at least 1 detection in N scans

Radar Course_14.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Outline

• Detection of Target Echoes in Noise

– Basic Concepts
– Integration of Pulses
– Fluctuating Targets Issues
– Adaptive Thresholding Techniques
• Pulse Compression

Radar Course_15.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Target Fluctuations
Swerling Models

Fluctuation Interval
scan-to-scan pulse-to-pulse
(multiple pulses/scan)

similar amplitudes

Swerling I Swerling II
−σ
p(σ ) =
1 σ av
e
σ av
Nature of
Scatterers one amplitude much
larger than others

Swerling III Swerling IV

4σ −2σ σav
p(σ ) = 2 e
σav

Radar Course_16.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 RCS Variability for Different
Target Models
20
Non-fluctuating Target
15

10

Measured RCS (dBsm)

0
20

Swerling I/II 15

10

0
20

Swerling III/IV 15

10

0
0 20 40 60 80 100
Sample #
Radar Course_17.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
321-00400
DPC 9/8/2008
 Detection Statistics for Fluctuating Targets
Single Pulse Detection

Figure by MIT OCW.

Fluctuating Targets Require More SNR than Non-fluctuating

Targets to Maintain a High Probability of Detection
Radar Course_18.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Outline

• Detection of Target Echoes in Noise

– Basic Concepts
– Integration of Pulses
– Fluctuating Targets Issues
– Adaptive Thresholding Techniques
• Pulse Compression

Radar Course_19.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Constant False Alarm Rate
(CFAR) Thresholding

• Problem: Must know (or 40

estimate) noise floor to Absolute
Signal
set threshold False alarm
threshold
30
• Solution: Estimate noise

Power (dB)
floor using noise-only Noise floor
samples 20

– Adaptive thresholding
10
• CFAR thresholding:

test cell
> threshold 0
noise floor estimate
0 20 40 60 80 100
Time (µs)

Radar Course_20.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 The Mean Level CFAR

• Use mean value of surrounding range cells to determine

threshold for cell under test

Window Slides Through Data

Cell Under Test

“Guard” Cells
Data Cells for Mean Level Computation

• Nearby targets can raise threshold and suppress detection

Radar Course_21.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Effect of Rain on CFAR Thresholding
Radar Backscatter (Linear Units)

Range Cells C Band

5500 MHz
9 dB
Rain Cloud
2.2 dB

Receiver Noise
Receiver Noise

2.6 Slant Range, nmi 4.5

Window Slides Through Data

Cell Under Test

“Guard” Cells
Data Cells for Mean Level Computation

Radar Course_22.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Effect of Rain on CFAR Thresholding
Amplitude (Linear Units)

Range Cells
C Band
5500 MHz
9 dB
Rain Cloud
2.2 dB

Receiver Noise
Receiver Noise

2.6 Slant Range, nmi 4.5

Window Slides Through Data

Cell Under Test

Sharp Clutter or Interference Boundaries
Can Lead to Excessive False Alarms “Guard” Cells
Data Cells for Mean Level Computation

Radar Course_23.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Greatest-of Mean Level CFAR

• Find mean value of N/2 cells before and after test cell
separately
• Use larger noise estimate to determine threshold

Window Slides Through Data

Cell Under Test

Data Cells for Mean Level 1 Data Cells for Mean Level 2
“Guard” Cells

Use Larger Value

• Helps reduce false alarms near sharp clutter or interference

boundaries
• Nearby targets still raise threshold and suppress detection
Radar Course_24.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Outline

• Detection of Target Echoes in Noise

• Pulse Compression
– Introduction
– Phase Coded Waveforms
– Linear Frequency Modulation Waveforms

Radar Course_25.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Pulsed CW Radar Fundamentals
Range Resolution
1 μsec pulse Frequency spectrum of pulse
3 20

Bandwidth
Pulsewidth

Power (dB)
Amplitude

2 10
1
T T
1 0

0 -20
1 2 3 4 0 1 2 3 4 5
Time (μsec) Frequency (MHz)
• Range Resolution ( Δ r )
– Proportional to pulse width (T) Δr =cT
2
– Inversely proportional to bandwidth (B = 1/T)
Δr= c
1 MHz Bandwidth => 150 m of range resolution 2B

Radar Course_26.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Pulse Width, Bandwidth and Resolution
for a Square Pulse
Resolution: Pulse Length is Larger than Target Length
Cannot Resolve Features Along the Target

Δr =cT
2
Δr= c
2B
Pulse Length is Smaller than Target Length
Can Resolve Features Along the Target
0
RCS (dB)

High Bandwidth
Relative

Example : Δr = .1 x Δ r
-20
BW = 10 x BW
-40 Low Bandwidth

Relative Range (m)

Shorter Pulses have Higher Bandwidth and Better Resolution

Radar Course_27.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Motivation for Pulse Compression

• Hard to get “good” average power and resolution at the

same time using a pulsed CW system
– Higher average power is proportional to pulse width
– Better resolution is inversely proportional to pulse width
• A long pulse can have the same bandwidth (resolution) as
a short pulse if the long pulse is modulated in frequency
or phase
• These pulse compression techniques allow a radar to
simultaneously achieve the energy of a long pulse and the
resolution of a short pulse

Radar Course_28.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Matched Filter Concept

E = Pulse Energy (Power × Time)

2E
Matched N0
Filter

Fourier Transform

Pulse Spectrum Matched Filter Noise Spectrum

Amplitude

Amplitude
Amplitude
Phase

Phase
N0

Frequency Frequency Frequency

• Matched Filter maximizes the peak-signal to mean noise ratio

– For rectangular pulse, matched filter is a simple pass band filter

Radar Course_29.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Frequency and Phase Modulation of Pulses

• Resolution of a short pulse can be achieved by

modulating a long pulse, increasing the time-bandwidth
product
• Signal must be processed on return to “pulse
compress”
Binary Phase Linear Frequency
Square Pulse Coded Waveform Modulated Waveform
Pulse Width, T Pulse Width, T Pulse Width, T

TCHIP
Frequency F1 Frequency F2
Bandwidth = 1/T Bandwidth = 1/TCHIP Bandwidth = ΔF = F2-F1
Time × Bandwidth = 1 Time × Bandwidth = T/TCHIP Time × Bandwidth = TΔF

Radar Course_30.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Binary Phase Coded Waveforms

Binary Phase
Coded Waveform • Changes in phase can be used to
increase the signal bandwidth of a
Pulse Width, T
long pulse
• A pulse of duration T is divided into
N sub-pulses of duration TCHIP
• The phase of each sub-pulse is
changed or not changed, according
to a binary phase code
• Phase changes 0 or π radians (+ or -)
• Pulse compression filter output will
TCHIP be a compressed pulse of width TCHIP
and a peak N times that of the
Bandwidth = 1/ TCHIP uncompressed pulse

Pulse Compression Ratio = T/ TCHIP

Radar Course_31.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
No overlap – Output 0 Time

Radar Course_32.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
No overlap – Output 0 Time

Radar Course_33.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
One sample overlaps 1x1 =1
Time

Radar Course_34.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
Two samples overlap (1x1) + (1x1) = 2
Time

Radar Course_35.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
Three samples overlap (1x1) + (1x1) + (1x1) = 3
Time

Radar Course_36.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
Two samples overlap (1x1) + (1x1) = 2
Time

Radar Course_37.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
One sample overlaps 1x1 =1
Time

Radar Course_38.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Implementation of Matched Filter

• Matched filter is implemented by “convolving” the

reflected echo with the “time reversed” transmit pulse
1

Reflected echo Time reversed pulse

• Convolution process:
– Move digitized pulses by each other, in steps
– When data overlaps, multiply samples and sum them up
3

Matched Filter
Output of
2

0
Use of Matched Filter Maximizes S/N Time

Radar Course_39.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Pulse Compression
Binary Phase Modulation Example

Figure by MIT OCW.

Radar Course_40.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Linear FM Pulse Compression

Because range is measured by a shift in

Doppler frequency, there is a coupling
of the range and Doppler velocity
measurement

Figure by MIT OCW.

Radar Course_41.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 Summary

• Detection of Targets in Noise

– Both target properties and radar design features affect the ability to
detect signals in noise
– Coherent and non-coherent integration pulse integration can improve
target detection
– Adaptive thresholding (CFAR) techniques are needed in realistic
environments
• Pulse compression offers a means to simultaneous have high
average power and good resolution
– A long pulse can have the same bandwidth (resolution) as a short
pulse, if it is modulated in frequency or phase
– Phase-encoded pulse compression divides long pulses into binary
encoded sub-pulses
– With frequency-encoded pulse compression, the radar frequency is
increased linearly as the pulse is transmitted

Radar Course_42.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01
 References

• Skolnik, M., Introduction to Radar Systems, New York,

McGraw-Hill, 3rd Edition, 2001
• Toomay, J. C., Radar Principles for the Non-Specialist,
New York, Van Nostrand Reinhold, 1989

Radar Course_43.ppt
MIT Lincoln Laboratory
ODonnell 10-26-01