You are on page 1of 4

1

SENSORS AND SIGNALS (DNM-MAE 203)


Matlab Practical
Eric Chaumette

I. R ADAR PRINCIPLE
A radar system entails a transmitter that synthesizes a base band signal, aka a waveform, e (t) (duration T0 and
bandwidth B ) which modulates a carrier signal with frequency fc :
n
o
j2fc t
mixer
(f
)
c
Re e (t) e
.
(1)
e (t)
The signal is transformed into an electromagnetic field (EMF)1 that propagates according to the radiation pattern
of the radars antenna. If the EMF encounters an obstacle, part of its energy is backscattered towards the radar and
received by the antenna. If the obstacle is non moving, the signal received at antenna output is:
o
n
(2)
ra (t) = Re a e (t 0 ) ej2fc t
where a and 0 denotes the complex attenuation2 and the round-trip delay of the target due to propagation. In
most modern radars, the first operation in the receiver is then to remove the carrier frequency in (at least) 3 stages.
The first one consists in decreasing the carrier frequency as follows:
a) to down-mix ra (t) from fc to fi :
o
n
ri1 (t) = Re a e (t 0 ) ej2fc t cos (2 (fi fc ) t)
 1

1
=
a e (t 0 ) ej2fc t + a e (t 0 ) ej2fc t
ej2(fi fc )t + ej2(fi fc )t
2
 2
1
j2fi t + e (t ) ej2fi t +

e
(t

)
e
0
a
0
a

= 14
j2(2fc fi )t + e (t ) ej2(2fc fi )t

e
(t

)
e
a
0
0
a
4
o 1
o
n
n
1
ri1 (t) =
Re a e (t 0 ) ej2fi t + Re a e (t 0 ) ej2(2fc fi )t ,
2
2
b) and with an analog pass-band filter centered on fi , to filter out the component with carrier frequency (2fc fi ):
o
n
a
ri1 (t) ri (t) = Re i e (t 0 ) ej2fi t , i =
.
(3)
2
The second stage consists in a analog to digital conversion (ADC), i.e. a sampling, which, provided that the sampling
frequency Fs satisfies Fs > 2 fi + B2 , yields3 :
X
{ri (nTs )}nZ rs (t) =
ri (nTs ) sin c (Fs (t nTs )) = ri (t) .
(4)
nZ

The last stage is a digital Hilbert filtering in order to get rid of the carrier frequency fi as follows:
a) to down-mix from fi to 0 (digital complex operation):

1
1
rhf
i e (t 0 ) ej2fi t + i e (t 0 ) ej2fi t ej2fi t
(t) = ri (t) ej2fi t =
2
1
1
1
rhf
(t) =
i e (t 0 ) + i e (t 0 ) ej2(2fi )t
2
2
1

The use of a carrier frequency fc is necessary to optimize the propagation of EMF in the air.
The complex attenuation includes the antenna gain, the propagation loss and the backscattering coefficient.
3
Notes on time-discrete signals.
2

b) and with a digital low-pass filter, to filter out the component with carrier frequency (2fi ):
a
i
1
=
.
(5)
rhf
(t) rhf (t) , x (t) = e (t 0 ) , =
2
4
Actually, at each stage of the receiver chain, the signal also incorporates some thermal noise, either received by
the antenna (sky noise) or generated by the receiver sub-assemblies (mixers, filters, ADC). Therefore at the output
of the Hilbert filter, in the presence of just one non moving point target, the baseband signal received can thus be
expressed as follows:
x (t) = e (t 0 ) + n (t) ,

n (t) : complex, centered, circular WSS random signal.

(6)

II. D ELAY ESTIMATION


One of the tasks of a RADAR (RAdio Detection And Ranging) is to estimate the range R0 of a point target via
the propagation delay 0 , since 2R0 = c 0 , where c is the speed of light in the air.
A. The noiseless case
n
o
If the noise n (t) can be neglected, i.e. E |n (t)|2  ||2 |e (t 0 )|2 , then (6) can be approximated by:
x (t) ' e (t 0 )

(7)

and x (t) is a deterministic signal.


Matlab: Run matlab practical, choice = noiseless case.
One way to assess 0 is to solve the following problem:
m ( ) = arg min {d (, )} ,

+
Z

d (, ) = kx (t) e (t )k ,

hx (t) |y (t)i =

x (t) y (t) dt,

(8a)

that is:

+
Z

m ( ) = arg min {d (, )} ,

d (, ) =

|x (t) e (t )|2 dt.

(8b)

Q1) Show that:


d ( ) = d (m , ) = kx (t)k2


2 
1 x,e ( ) ,

hx (t) |e (t )i
x,e ( ) =
,
kx (t)k ke (t )k

^
2

2 Re ( 0 )
x,e ( ) =
,
^
Re (0)

where Re (u) is the auto-correlation function of the finite energy signal e (t)4 .
Q2) Show that the minimum value d ( ) is obtained for , 0 .
Q2.1) What is the value of d ( 0 )?
Q2.2) Therefore what is the value of m ( 0 )?
Q3) Finally, which procedure could you use to estimate both (energy backscattered) and 0 ?
Matlab: Run matlab practical, choice = noiseless case (lambda).
In the following the function:
^
2


he (t) |e (t )i 2 Re ( )
2
2
=
Ae ( ) =
|m ( )| = || Ae ( )
^

ke (t)k ke (t )k
Re (0)
is called the ambiguity function of the waveform e (t), aka the delay/range ambiguity function.
4

Notes on correlation, energy/power spectrum.

Q4) The range resolution of a radar is its ability to distinguish targets closely spaced in range. It is mostly determined
by the shape and width of the mainlobe of the ambiguity function Ae ( ). A common metric used to define the
range resolution is the Rayleigh criterion, i.e., the separation between the peak and the first null of the mainlobe
response.
Matlab: Run matlab practical, choice = ambiguity function.
What is the range resolution of the radar under consideration?
Q5) Prove that (8b) is actually equivalent to:
Z
+T0

m ( ) = arg min {d (, )} ,

d (, ) =

|x (t) e (t )|2 dt.

(9)

B. The noisy case


n
o
If the noise n (t) can not be neglected, i.e. E |n (t)|2  ||2 |e (t 0 )|2 , then (6) must be taken into account:
x (t) = e (t 0 ) + n (t) ,

n (t) : complex, centered, circular WSS random signal,

n (t) = nr (t) + jnj (t) , E {n (t)} = 0, nr (t) and nj (t) are mutually WSS,

Rnr ( ) = Rnj ( )
E {n (t) n (t )} = 0
Rnr ,nj (0) = 0
Rnr ,nj ( ) = Rnr ,nj ( )

(10a)
(10b)
(10c)

and x (t) is a random signal.


Matlab: Run matlab practical, choice = noisy case.
Q6) What kind of noise is n (t)? What is the power of n (t)5 ?
In that case, it is logical to replace (8b)(9) by:
+T
+T
Z 0 n
Z 0

o
2
m ( ) = arg min {d (, )} , d (, ) = E
|x (t) e (t )| dt =
E |x (t) e (t )|2 dt ,

(11)

that is:
m ( ) = arg min {d (, )} ,

d (, ) = kx (t) e (t )k2 ,

hx (t) |y (t)i = E

Q7) Is x (t) a WSS random signal ?


Q8) Show that7 :

+T
2
Z 0

E
x (t) e (t ) dt




x,e ( ) 2 =
+T

+T0
Z
Z 0

E
|x (t)|2 dt E
|e (t )|2 dt

Notes on wide-sense stationary random signals (process)


Notes on wide-sense stationary random signals (process)
7
Notes on correlation, energy/power spectrum
6

+T
Z 0

x (t) yx (t) dt

(12)

Q9) Using the properties of WSS random signals8 , prove that:


^
2

2
||
R
(

)
e
0

2
||2
x,e ( ) = 
=

^
^
n (0)
||2 + R
T0
^
||2 Re (0) + Rn (0) T0 Re (0)
R (0)
e

^

R ( ) 2
e
0
^

R (0)
e

(13)

where Rn ( ) is the autocorrelation function of n (t).



2
Q10) Show that the minimum value d ( ) is obtained for , 0 and give the expression of x,e ( 0 ) as a
function of the signal to noise ratio (SNR) defined as:
+T

Z 0

E
|e (t 0 )|2 dt
^

||2 Re (0)

(14)
=
SN R =
+T

T0 Rn (0)
Z 0

E
|n (t)|2 dt

Matlab: Run matlab practical, choice = noisy case (lambda).


Note that
n
o
(
)
2
2
n
o
E
|hx
(t)
|e
(t

)i|
|hx (t) |e (t )i|
E |m ( )|2
= E
=
ke (t )k4
ke (t )k4


hx (t) |e (t )i
hE {x (t)} |e (t )i
he (t 0 ) |e (t )i
E {hx (t) |e (t )i}
E {m ( )} = E
=
=
=
2
2
2
ke (t )k
ke (t )k
ke (t )k
ke (t )k2
E {m ( 0 )} =


Q11) Let us assume that n (t) has a constant PSD over F2s , F2s , Fs > B (bandwidth of e (t)), i.e. Sn (f ) = Sn
and that |e (t)|2 = 1, 0 t T0 . We want to achieve the lowest possible value of d (min , 0 ). It can be done
2
by maximizing x,e ( 0 ) (13), or equivalently, by maximizing the SN R (14) with the filtering of x (t) with an
ideal low-pass filter with bandwidth Bf and transfer function H (f ).
Q11.1) Show that, if Bf B , then the SNR at the output of the ideal low-pass filter has become:
SN RH =

||2
Sn Bf


2
Q11.2) What is the optimal choice for Bf for which the maximum value of x,e ( 0 ) is obtained?
Matlab: Run matlab practical, choice = noisy case filtered (lambda).
Q12.1) What is the bandwidth at 3dB of the filter ? Is the filter physically implementable?
Q12.2) What is the noise power after filtering?
Q12.3) Does the filtering improve the behaviour of m ( ) =

Notes on wide-sense stationary random signals (process)

hx(t)|e(t )i
?
ke(t )k2

Why?

You might also like