You are on page 1of 45

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

RADAR signal processing


Radar basics Dr. Ir. Xavier Neyt Associate Professor
Communication, Information, Systems and Sensors Departement

Royal Military Academy

April, 2012

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Denition
Radar Transmits an electromagnetic wave The wave is reected by objects Reection is received Compares transmitted and reected wave
in amplitude and in phase

Requirement The radar need to be coherent there is a coherency between the transmitted and the received wave. Conguration monostatic bistatic moving non-moving

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Measured quantities

Measurements Time-delay Distance (Range) Frequency shift Radial velocity Amplitude Radar cross section (Detection) Other quantities Change in polarization Target identicaiton

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Elements of a radar system


Transmitter stage Waveform Generator Local oscillator

Signal Processor Receiver stage

Detection

Tracking

Display

We will concentrate on the Signal processor block Central question: How to process the data to maximize system performance?

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Down-conversion: typical analog

IF RF BPF LPF

LPF
900

Q
D

LPF

I
D

LO1

LO2

Issue: preserve the phase


IF: I: Q: I+jQ: sIF (t ) sIF (t ) cos IF sIF (t ) sin IF = A cos( t + ) =A 2 [cos(( IF )t + ) + cos(( + IF )t + )] =A 2 [sin(( IF )t + ) + sin(( + IF )t + )]
A j j ( IF )t 2 e [e

+ e j (+IF )t ]

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Down-conversion: typical analog

IF RF BPF LPF

LPF
900

Q
D

LPF

I
D

LO1

LO2

Issue: preserve the phase

Advantage Low sampling frequency fs > B

Inconvenience Imbalance DC-oset of amplier

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Down-conversion: Digital Down Conversion (DDC)

RF

BPF

LPF

IF

LPF

I+jQ

LO1

e j 2 nT LO2

Issue: preserve the phase


IF: I+jQ: sIF (t ) = A cos( t + )
A j j ( IF )t 2 e [e

+ e j (+IF )t ]

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Down-conversion: Digital Down Conversion (DDC)

RF

BPF

LPF

IF

LPF

I+jQ

LO1

e j 2 nT LO2

Issue: preserve the phase

Advantage No I/Q imbalance Perfect I/Q orthogonality

Inconvenience Higher sampling frequency fs > 2 B

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Range-resolution

Denition Range resolution: Smallest distance that must exist between two targets to permit the discrimination of the two targets. Pulsed radar Pulse of length p Resolution : r = p Bandwidth: p
1 B

Conclusion The range resolution is proportional to the bandwidth of the transmitted signal.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Doppler frequency resolution


range of a moving target: r (t ) = r (0) + vr t vr = radial velocity r time delay: (t ) = (0) + 2 v c t transmitted signal: s (t ) = A(t )e j (t ) e j t received signal: sr (t ) = s (t (t )) = A(t (t ))e j (t (t )) e j (t (t )) with (0) = 0: vr j (t (12 vcr t )) j t (12 vcr t ) r sr (t ) = s (t (12 v e c t )) = A(t (12 c t ))e neglect envelope A() and phase () change during observation vr sr (t ) = A(t )e j (t ) e j t e j 2 c t = A(t )e j (t ) e j t e 2j D t Doppler frequency fD =
vr 1 2 2 c
r = 2v

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Doppler frequency resolution


Denition Doppler freq. resolution: Smallest Doppler frequency separation that must exist between two targets to permit the discrimination of the two targets. Pulsed radar Pulse of length p Resolution :
1 p

Conclusion The Doppler frequency resolution is proportional to the length of the transmitted signal.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Pulse-Doppler radar

Situation High range resolution: small p High frequency resolution: large p Contradictory requirements Solution Coherent pulse train

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Pulse-Doppler radar
Re (sPD )

p TR

Pulses are repeated at the Pulse Repetition Frequency (PRF). 1 PRF = T R Mathematical expression
N 1

PD

(t ) =
k =0

rect

t kTR p

e j t

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Doppler-frequency resolution
1

|Scw|

0.5

1000

500

0 f (Hz)

500

1000

|Spd|

0.5

1000

500

0 f (Hz)

500

1000

Resolution
1 = 2 T 1 = 2 NT R

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Range ambiguities

Ambiguous range Targets separated by Ramb will produce the same echo. Ramb =
TR c 2

Solutions Antenna radiation pattern only applicable for down-looking radar Increase PRI large PRI for surveillance radars Pulse decorrelation PRI staggering

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: construction

Start from one pulse Replicate (convolve) in time with a comb-function


Is a multiplication in the spectral domain

Consider a nite comb function


To have a pulse train of nite length.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: 1 pulse

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 t (s)

0.02

0.025

0 0.03 1

0.5

0 f (Hz)

0.5 x 10

1
4

The spectrum of a (single) pulse is sin x /x -shaped.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: comb function

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 t (s)

0.02

0.025

0 0.03 1

0.5

0 f (Hz)

0.5 x 10

1
4

The spectrum of a comb-function is also a comb function. Temporal spacing: TR 1 Spectral spacing: T R

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: Innite pulse train

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 f (Hz)

0.02

0.025

0 0.03 1

0.5

0 f (Hz)

0.5 x 10

1
4

The spectrum of a pulse train is composed of peaks.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: Innite pulse train

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 f (Hz)

0.02

0.025

0 0.03 1000

500

0 f (Hz)

500

1000

The spectrum of a pulse train is composed of peaks. causes ambiguities

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Doppler ambiguities
Target with velocity v and v + frequency.
2TR

will have same Doppler

Velocity of target with velocity > vmax will be erroneously estimated. Ambiguous velocity Maximum unambiguous velocity: vmax =

1 4 TR

Solutions Decrease the PRI Large PRF for Doppler radars Use multiple PRI sequentially

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: N-pulses

Multiply innite pulse train with a square window of duration N TR


1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 t (s)

0.02

0.025

0 0.03 1

0.5

0 f (Hz)

0.5 x 10

1
4

The spectrum of the window is sin x /x -shaped (but very narrow).

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: N-pulses

Multiply innite pulse train with a square window


1
1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 t (s)

0.02

0.025

0 0.03 1000

500

0 f (Hz)

500

1000

The spectrum of the window is sin x /x -shaped (but very narrow).

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: N-pulses


The spectrum of the innite pulse train is convolved with the spectrum of the square window.
1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 t (s)

0.02

0.025

0 0.03 1

0.5

0 f (Hz)

0.5 x 10

1
4

The spectrum is non-zero between the peaks.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: N-pulses


The spectrum of the innite pulse train is convolved with the spectrum of the square window.
1
1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.005

0.01

0.015 t (s)

0.02

0.025

0 0.03 1000

500

0 f (Hz)

500

1000

The spectrum is non-zero between the peaks. will lead to leakage of a target onto another will limit detection of weak targets

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Spectrum of a pulse train: Comparison

0.8

0.8

0.8

0.6

0.6

0.6

0.4

0.4

0.4

0.2

0.2

0.2

0 1000

500

0 f (Hz)

500

0 1000 1000

500

0 f (Hz)

500

0 1000 1000

500

0 f (Hz)

500

1000

Summary: Pulse train High range resolution


There are range ambiguities

High Doppler resolution


There are Doppler (velocity) ambiguities

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Radar equation

Transmitted power Pt Radar cross section = Received power


Ae Pr = Prt 4 R2

Power density at range R


G Pd = Pt 4 R2

Prt Pd

Antenna eective area Ae =


G 2 4

Radar equation
G G Pr = Pt 4 G = Pt (4 R 2 (4 )2 R 2 )3 R 4
2 2 2

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Radar equation

Radar equation
G Pr = Pt (4 )3 R 4
2 2

Links radar parameters to predict performance


Targets with large RCS are easier to detect (obvious) A doubling of the range imply multiplying the power by 16!

Received power must be larger than (thermal) noise power


High transmitted power Pt Antenna with high gain (high directivity) G Receiver with low noise Preferably long wavelengths

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Signal model

Sampled baseband transmitted signal (delayed): s (k ) = A(k )e j (k )


A(k ) = amplitude of the transmitted signal (k ) = phase of the transmitted signal

Received (baseband) echo (in the presence of a target): y (k ) = s (k ) + n(k )


= target (complex) reectivity n(k ) = noise

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Signal model: vector notation

Consider a sample sequence of length N


s = [s (0), s (1), . . . , s (N 1)]T y = [y (0), y (1), . . . , y (N 1)]T n = [n(0), n(1), . . . , n(N 1)]T

Received signal y = s + n

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Signal model: noise


Gaussian noise Noise is characterized by covariance matrix R = E {nn } Noise probability density function (PDF) p (n) = 1 N |R| e n
R1 n

Can take any noise into account, including non-white noise White noise:
R = 2 I p (n) =
1 2N e N n
n 2 2 n

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Signal model: target reectivity


Model of the complex amplitude of the reected signal Deterministic constant Steady target (Marcum model) Not very realistic Leads to easy analytical results Stochastic quantity Fluctuating target (Swerling models) PDF of : p () =
|2 1 |2 e 2

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Signal model: probability density function


Two hypotheses: H0 and H1 H0 : No target is present sr = n 1 1 p (sr |H0 ) = N e sr R sr |R| H1 : A target is present sr = s + n 1 1 p (sr |H1 ) = N e (sr s) R (sr s) |R|

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Decision: likelihood
Likelihood L(H0 ) = p (y|H0 ) L(H1 ) = p (y|H1 ) Likelihood ratio (y) = Likelihood ratio test (y)
H0 H1

L(H1 ) L(H0 )

Maximize the probability of detection for a given probability of false alarm.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Test statistics: PDF


Likelihood ratio (y) = e
s R1 y

e y

R1 s

e ||

2 s R1 s

With s R1 y = (y R1 s) (y) = e 2Re( and taking the logarithm (y) = 2Re( s R1 y)||2 s R1 s. Test statistic T (y) = |w y|2 with w = k R 1 s and k=
1 s R1 s
s R1 y)

e ||

2 s R1 s

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Test statistics: Signal to noise ratio


Consider the result of the lter w: = w y Noise only noise power y=n Power = E {| |2 } = 1 (due to the normalization with k ) Signal only signal power y = s Power = E {| |2 } = ||2 s R1 s y = s Signal to noise ratio SNR = ||2 s R1 s

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Test statistics: PDF


Consider z = w y z |H0 CN (0, 1) z |H1 CN ( , 1) k thus Re(z ) N (, 1) & Im(z ) N (, 1)

T (y) = |z |2 = Re(z )2 + Im(z )2 T (y) is chi-squared distributed Test statistic PDF T (y )


1 2 2 2 1 2 2 1 2 2 (2|| s R s)

H0 H1

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Test statistics: PDF


Marcum target (constant target amplitude)
0.07 0.06 0.05 PDF(T(y)) 0.04 0.03 0.02 0.01 0 0 H0 H1

10

20

T(y)

30

40

50

The test statistics distribution is dierent for H0 and H1


makes the discrimination between H0 and H1 possible

The separation depends on the signal to noise ratio SNR.

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Test statistics: PDF (sidenote)

H0 1 T ( y ) 2 2 2 A chi-square with 2 degrees of freedom = exponential distribution H1 1 T (y) 22 (2||2 s R1 s) 2 A non-central chi-square with 2 degrees of freedom = Rayleigh distribution

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Test statistics: Probability of false alarm


Probability of false alarm
+

PFA = Q ( ) =

p (T |H0 )dT

= Q 1 (PFA )
0.07 0.06 0.05 PDF(T(y)) 0.04 0.03 0.02 0.01 0 0 PFA

10

20

T(y)

30

40

50

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Test statistics: Probability of detection


Probability of detection
+

PD =

0.07 0.06 0.05 PDF(T(y)) 0.04 0.03 0.02 0.01 0 0

p (T |H1 )dT
PD

10

20

T(y)

30

40

50

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Performance
1 0.9 0.8 0.7 0.6 PD 0.5 0.4 0.3 0.2 0.1 0 0 5 10 SINR (dB) 15 20 PFA = 10 PFA = 10
4 5

PFA = 106 PFA = 107

Performance Depends critically on the SNR


even a small decrease in SNR will reduce PD

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Performance
Comparison of PD for steady and uctuating targets
1 0.9 0.8 0.7 0.6 PD 0.5 0.4 0.3 0.2 0.1 0 0 5 10 SINR (dB) 15 20 Marcum model SwerlingI model

large SNR: steady targets are easier to detect low SNR: uctuating targets are easier to detect
due to uctuation, target signal may exceed the treshold

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Performances: multiple uctuating pulses

Probability of detection as a function of the number of pulses


1 0.9 0.8 0.7 0.6 PD 0.5 0.4 0.3 0.2 0.1 0 0 1 radar 2 radars 3 radars 4 radars

Increase of performance due to Incoherent integration gain Diversity gain

10 SINR (dB)

15

20

Introduction

Radar system

Resolution

Pulse-Doppler radar

Radar equation

Target detection theory

Performance

What are the performance of a particular lter w? Performance is measured in terms of SNRloss SNRloss = SNRoutput SNRinput

direct link between SNRloss and detection performance

You might also like