Fundamentals of signal and

system analysis
Associate Prof. Dr. Ho Van Khuong
Office: Dept. of Telecom. Eng.
Email: khuong.hovan@yahoo.ca
 Outlines
• Signal models
• Signal classifications
• Generalized Fourier series
• Fourier series
• The Fourier transform
• Power spectral density and correlation
• Signal and linear system
• Sampling theory
• The Hilbert transform
• Discrete Fourier transform and fast Fourier transform
 Notes
• A signal is defined as the time history of some
quantity, usually a voltage or current.
• A system is a combination of devices and
networks (subsystems) chosen to perform a
desired function.
• Communication systems are complicated,
requiring a great deal of analysis and
experimentation with trial subsystems before
actual building of the desired system.
• Mathematical models for signals and systems.
 Signal models
• Deterministic and random signals
– Deterministic signals: completely specified
functions of time.
• Sinusoidal: x(t) = A cos(ω0t), -∞ < t < ∞
• Unit rectangular pulse:
1, t ≤ 0.5
Π (t ) =

0, otherwise
– Random signals: take on random values at any
given time instant.
 Signal models
• Periodic and aperiodic signals
– Periodic signal x(t): if and only if
x(t+T0) = x(t), -∞ < t < ∞ (*)
where the constant T0 is the period.
• Fundamental period is the smallest number satisfying
(*).
• Ex: x(t) = A cos(ω0t), -∞ < t < ∞
– Aperiodic signal: not satisfy (*).
 Signal models
• Phasor signals and spectra
– Consider a useful periodic signal with T0 = 2π/ω0
x ( t ) Ae
j (ω0t +θ )
Rotating 
= , −∞ < t < ∞
phasor
amplitude frequency (radians per phase in radians
second) or f0 = ω0/2π (Hz)

Phasor Ae jθ
– Relationship between the rotating phasor and the
real, sinusoidal signal
x ( t ) ) Re ( Ae (
+ θ ) Re ( =
x ( t ) A cos (ω0t= )
j ω t +θ )
= 0

1 1 1 j (ω0t +θ ) 1 − j (ω0t +θ )
A cos (ω0t + θ=
) x ( t ) + x * (=
t) Ae + Ae
2 2 2 2
 Signal models

• Amplitude and phase spectra for Acos(ω0t+θ)

Single sided spectra Double sided spectra

 Signal models
• Example: sketch the single-sided and double-
sided spectra of

 1 
=x ( t ) 2sin 10π t − π 
 6 

 1 
( t ) 2sin 10π t − π  + cos ( 20π t )
y=
 6 
 Signal models
• Singularity functions
– Unit impulse function (or delta function) δ(t)
∞ ∞

∫ x ( t ) δ ( t ) dt = x ( 0 )
−∞
∫ x ( t ) δ ( t − t ) dt =
−∞
0 x (t )
0

continuous at t = 0 continuous at t = t0

– Alternative definition of δ(t)

t2

∫ δ ( t − t ) dt=
0 1, t1 < t0 < t2 1, t1 < t < t2
t1 x (t ) = 
δ ( t − t0 )= 0, t ≠ t0 0, otherwise
 Signal models
• Singularity functions
– Properties of δ(t)
• δ(at) = (1/|a|)δ(t), a is a constant.
• δ(-t) = δ(t) t
 x ( t0 ) , t1 < t0 < t2
2

• Sifting property ∫ x ( t ) δ ( t − t0 ) dt =
 0, otherwise
t1 undefined, t = t or t
 0 1 2

• x(t)δ(t-t0) = x(t0)δ(t-t0), where x(t) is continuous at t = t0

• The superscript
t
n denotes the n th derivative
2

( ) ( n)
( ) ( ) x ( t0 ) , t1 < t0 < t2
n ( n)
∫ x
t1
t δ t − t 0 dt = −1

• If f(t) = g(t), where f(t) = a0δ(t) + a1δ(1)(t) + … + anδ(n)(t)

and g(t) = b0δ(t) + b1δ(1)(t) + … + bnδ(n)(t), then a0 = b0, a1
= b1, …, an = bn.
 Signal models
• Singularity functions
– Unit step
 0, t<0

t
=u ( t ) ∫= δ (λ ) dλ  1, t >0
−∞ undefined, t = 0

or
du ( t )
δ (t ) =
dt
 Signal classifications
• e(t) is the voltage across a resistance R
producing a current i(t).
• Instantaneous power per ohm is p(t) =
e(t)i(t)/R = i2(t).
• The total energy and the average power on a
per-ohm basis:
T T
1
E = lim ∫ ( t )dt P= ∫ ( t )dt
2 2
i lim i
T →∞ T →∞ 2T
−T −T
 Signal classifications
• For an arbitrary signal x(t), total (normalized)
energy and power:
T ∞ T
1
∫ x ( t ) dt ∫ x (t ) ∫ x (t )
2 2 2
E =
lim dt P= lim dt
T →∞ T →∞ 2T
−T −∞ −T

• x(t) is an energy signal if and only if 0<E<∞, so

that P=0.
• x(t) is a power signal if and only if 0<P<∞, thus
implying that E=∞.
 Signal classifications
• Determine the classification of the following
signals
x1 ( t ) Ae −α t u ( t ) , where α > 0 and A and α are positive constants.

x2 ( t ) = Au ( t )

x3 ( t ) = Ae
− j (ω0t +θ )

t −1/4 , t ≥ t0 ≥ 0
x4 ( t ) = 
 0, otherwise
 Signal classifications
• Periodic signal with the period T0
T0 + t0
1
∫ x p ( t ) dt
2
P=
T0 t0

t0 is an arbitrary starting time (chosen for

convenience)
• Ex:

x p ( t ) A cos (ω0t + θ )
=
 Generalized Fourier series
• Representation of signals as a series of
orthogonal functions.
• Consider the problem of representing a time
function x(t) on a T-s interval (t0, t0+T)
– A basis function set φn(t): φ1(t), φ2(t), …, φn(t) are
linearly independent (i.e., no one of them can be
expressed as a weighted sum of the other N-1).
– The approximation of x(t) Independent of time
N
xa ( t ) = ∑ X nφn ( t )
n =0
 Generalized Fourier series
– The choice of
1
X n = ∫ x ( t ) φn∗ ( t ) dt
cn T
is to minimize
∫ x (t ) − x (t )
2
= ε=
Error N a dt
T
where φn(t) is assumed to be orthogonal
t0 +T0
cn , n = m
∫ φm ( t ) φ (=
*
t ) dt c=
nδ mn  (all m and n)
 0, n ≠ m
n
t0
Kronecker delta function

– If cn = 1 for all n, φn(t) is an orthonormal basis set

 Generalized Fourier series
– Then N
(ε N )min ∫ x (t ) dt − ∑ cn X n
2
=
2

T n =0

– φn(t) is complete if
lim ( ε N )min = 0
N →∞

∫ x (t )
2
for any signal that is integrable square dt < ∞
T
– In the sense that the error is zero
∞
x ( t ) = ∑ X nφn ( t )
n =0
– Parseval’s theorem
∞

∫ x ( t ) dt = ∑ cn X n
2 2

T n =0
 Example
sin (π t ) , 0 ≤ t ≤ 2 xa(t)?
x (t ) = 
 0, otherwise Error?

orthonormal basis set?

 Fourier series
• Complex exponential Fourier series
– Given a signal x(t) defined over (t0, t0+T0) with
2π
ω0 2=
= π f0
T0
– Complex exponential Fourier series
∞
=x (t ) ∑
n = −∞
X n e jnω0t , t0 < t < t0 + T0 (*)

where
t0 +T0
1
Xn = ∫ x ( t )e − jnω0t dt
T0 t0
 Fourier series
• Complex exponential Fourier series
– Note:
• The right-hand side of (*) is periodic with period T0,
since it is the sum of periodic rotating phasors with
harmonic frequencies.
• If x(t) is periodic with period T0, the Fourier series is an
accurate representation for x(t) for all t.
– Example: Find the complex exponential Fourier
series for the signal x(t) = cos(ω0t) + sin2(2ω0t)
where ω0 = 2π/T0.
 Fourier series
• Symmetry properties of the Fourier
coefficients • Magnitude of X is even
n
• Argument of Xn is odd
– x(t) is real: X n* = X − n
– x(t) is even/odd: Xn is purely real/imaginary and is
an even/odd function of n.
– Signals with odd half-wave symmetry, i.e.
 1 
− x (t )
x  t ± T0  =
 2 
where T0 is the period of x(t). Then
X n = 0, n = 0, ±2, ±4,...
 Fourier series
• Trigonometric form of the Fourier series
– x(t) is real ∞
) X 0 + ∑ 2 X n cos ( nω0t + ∠X n )
x ( t=
n =1
∞ ∞
X 0 + ∑ An cos ( nω0t ) + ∑ Bn sin ( nω0t )
x (t ) =
=n 1=n 1

t0 +T0 t0 +T0
2 2
An = ∫ x ( t ) cos ( nω0t ) dt Bn = ∫ x ( t ) sin ( nω t ) dt
0
T0 t0
T0 t0

– x0: average or DC component

– x1: the fundamental
– x2: the second harmonic
 Fourier series
• Parseval’s theorem
– Average power of a periodic signal
∞ ∞
1
∫ x ( t )= ∑ X0 + ∑ 2 Xn
2
= X=
2 2 2
P dt n
T0 T0 n = −∞ n =1

Sum of the powers in the

phasor components of the
Fourier series Sum of the powers in the
DC component and that in
the AC components
Fourier series
 Fourier series
• Line spectra

∞
=x (t ) ∑Xe
n = −∞
n
jnω0t
, t0 < t < t0 + T0

∞
) X 0 + ∑ 2 X n cos ( nω0t + ∠X n )
x ( t=
n =1
 Fourier series
T0
τ=
4
• Line spectra

Pulse train
∞
 t − nT0 − 0.5τ 
x (t ) = ∑ ∏ 
n = −∞
A
τ 


T0
τ= ; T0 same as in (a)
8

T0
τ= ;τ same as in (a)
4
Fourier transform of nonperiodic signals
• x(t) is nonperiodic but is an energy signal.
• The Fourier transform of x(t)
∞
X(f )=
ℑ  x ( t )  =
∫ x ( t )e − j 2π ft
dt
−∞

• The inverse Fourier transform of x(t)

∞
x (t ) =
ℑ−1  X ( f )  =
∫ X ( f )e j 2π ft
dt
−∞
Fourier transform of nonperiodic signals
• Amplitude and phase spectra
X ( f ) = X ( f ) e jθ ( f ) , θ ( f ) = ∠X ( f )

Amplitude spectrum Phase spectrum

– For real x(t)

X ( f ) =X ( − f ) θ(f )=−θ ( − f )
Fourier transform of nonperiodic signals
• Symmetry properties
– The Fourier transform of a real, even function is
real and even (i.e., Im(X(f)) = 0 and Re(X(f)) is an
even function of f).
– The Fourier transform of a real, odd function is
imaginary and odd (i.e., Re(X(f)) = 0 and Im(X(f)) is
an odd function of f).
∞
= =
R Re ( X ( f )) ∫ x ( t ) cos ( 2π ft ) dt
−∞
∞
I = Im ( X ( f ) ) = − ∫ x ( t ) sin ( 2π ft ) dt
−∞
Fourier transform of nonperiodic signals
• Example: plot the amplitude and phase
spectrum of
 t − t0 
x ( t ) = A∏  
 τ 
Fourier transform of nonperiodic signals
• Energy spectral density
G( f ) = X ( f )
2

• Rayleigh’s energy theorem or Parseval’s

theorem for the Fourier transform
∞ ∞
x ( t ) dt ∫ X ( f )
∫=
2 2
=E df
−∞ −∞

• Ex.: Find the energy in

 f 
x (t ) =
40sin c ( 20t ) ↔ X ( f ) =
2∏ 
 20 
Fourier transform of nonperiodic signals
• Convolution
∞ ∞
x ( t ) = x1 ( t ) * x2 ( t ) = ∫ x ( λ )x ( t − λ ) d λ = ∫ x ( t − λ )x ( λ ) d λ
1 2 1 2
−∞ −∞

– Time reversal to obtain x2(-λ)

– Time shifting to obtain x2(t-λ)
– Multiplication of x1(λ) and x2(t-λ)
• Ex.: Find the convolution of
x1 ( t ) e −α t u ( t )=
= , x2 ( t ) e − β t u ( t ) , α > β > 0
Fourier transform of nonperiodic signals
- Transform theorems
• Superposition a x ( t ) + a x ( t ) ↔ a X ( f ) + a X ( f )
1 1 2 2 1 1 2 2

• Time-delay x ( t − t ) ↔ X ( f ) e
0
− j 2π ft0

• Scale-change x ( at ) ↔ 1a X  af 

• Duality X ( t ) ↔ x ( − f )
• Frequency translation x ( t ) e ↔ X ( f − f ) j 2π f 0 t
0

• Modulation x ( t ) cos ( 2π f t ) ↔ 12 X ( f − f ) + 12 X ( f + f )
0 0 0

Differentiation d x ( t ) ↔ ( j 2π f ) X ( f )
n
• n
n

dt
• Integration ∫ x ( λ ) d λ ↔ ( j 2π f ) X ( f ) + 1 X ( 0 ) δ ( f )
t
−1

−∞
2
∞ ∞

• Convolution ∫ x ( λ )x (t − λ ) d λ = ∫ x (t − λ )x ( λ ) d λ ↔ X ( f ) X ( f )
1 2 1 2 1 2
−∞ −∞ ∞
• Multiplication x ( t ) x ( t ) ↔ X ( f ) * X ( f ) =∫ X ( λ )X ( f − λ ) d λ
1 2 1 2 1 2
−∞
Fourier transform of nonperiodic signals -
Transform theorems

• Ex: Use the duality theorem to show that

 f 
2 AW sin c ( 2Wt ) ↔ AΠ  
 2W 

• Ex: obtain the following Fourier transform

pairs
1. Aδ ( t ) ↔ A
2. Aδ ( t − t0 ) ↔ Ae − j 2π ft
0

3. A ↔ Aδ ( f )
4. Ae j 2π f t ↔ Aδ ( f − f 0 )
0
Fourier transform of nonperiodic signals -
Transform theorems

• Ex: Use the differentiation/convolution

theorem to obtain the Fourier transform of
the triangular signal
t  1 − t / τ , t <τ
Λ =
τ   0, otherwise

• Ex: obtain the Fourier transform of

∞
ys ( t )
= ∑ δ ( t − mT )
m = −∞
s
Fourier transform of nonperiodic signals -
Transform theorems

• Convolution of x(t) with an impulse occurring

at time t0 simply shifts x(t) to t0
∞
δ ( t − t0 ) * x ( t ) =∫ δ ( λ − t0 )x ( t − λ ) d λ =x ( t − t0 )
−∞

• Ex: obtain the Fourier transform of

t
AΠ   cos (ω0t ) , ω0 =
x (t ) = 2π f 0
τ 
 Fourier transform of periodic signals
• x(t) is a periodic power signal
 ∞  ∞
)  ∑ δ ( t − mTs ) ∗ p ( t=)
x ( t= ∑ p ( t − mT ) s
 m = −∞  m = −∞

Pulse-type signal
∞
=X(f) ∑ f s P ( nf s ) δ ( f − nf s ) P ( f ) = ℑ  p ( t ) 
n = −∞

• Ex: Find the Fourier transform of a periodic

cosinusoidal train
 ∞  t 
y ( t )  ∑ δ ( t − mTs )  ∗ Π 
=  cos ( 2π f 0t ) , f 0 >> 1/ τ
 m = −∞  τ 
 Fourier transform
• Poisson sum formula
 ∞
 ∞
ℑ−1  ∑ f s P ( nf s ) δ ( f − nf s )  =∑ f s P ( nf s ) e j 2π nf s t

 n = −∞  n = −∞

∞ ∞

∑ p ( t − mT ) ↔ ∑
m = −∞
s
n = −∞
f s P ( nf s ) δ ( f − nf s )

∞ ∞

∑
m = −∞
f s ∑ P ( nf s ) e j 2π nf s t
p ( t − mTs ) =
n = −∞
Power spectral density and correlation
• For power signals, power spectral density S(f)
∞ T
1
=P ∫ S (=
f ) df x ( t ) lim
=2
∫ x 2 ( t ) dt
T →∞ 2T
−∞ −T

• Ex: Consider x(t) = A cos(2πf0t + θ). S(f)?

Power spectral density and correlation
• Time-average autocorrelation function
– For energy signals
∞ T

(τ )
φ= ∫ x (λ ) x (λ +τ =
) dλ lim
T →∞ ∫ x (λ ) x (λ +τ ) dλ
−∞ −T

• φ(τ) gives a measure of the similarity, or coherence,

between a signal and a delayed version of the signal.
• E = φ(0)
• φ(τ) ↔ G(f)
Power spectral density and correlation
• Time-average autocorrelation function
– For power signals
T
1
(τ )
R= x ( t ) x ( t +=
τ ) lim
T →∞ 2T ∫ x ( t ) x ( t + τ ) dt
−T
– x(t) is periodic with period T0
1
=R (τ ) ∫ x ( t ) x ( t + τ ) dt
T0 T0
• R(τ) gives a measure of the similarity between a power
signal at time t and at time t+τ.
• P = R(0)
• R(τ) ↔ S(f)
Power spectral density and correlation
• Properties of R(τ):
– R(0) = 〈x2(t)〉 ≥ R(τ), ∀τ
– R(-τ) = 〈x(t)x(t-τ)〉 = R(τ): R(τ) is even
– limτ→∞R(τ) = 〈x2(t)〉 if x(t) does not contain
periodic components.
– If x(t) is periodic in t with period T0, then R(τ) is
periodic in τ with period T0.
– The Fourier transform of R(τ) is nonnegative.
Power spectral density and correlation
• Example: Find the autocorrelation function
and power spectral density of the signal
x (t ) = ( ) 2 + 3cos (10π t ) − 4sin (10π t )
Re 2 + 3e j10π t + 4 je j10π t =
 Signals and linear systems
• Operator representation of a linear system
x(t) H y(t)

y(t) = H[x(t)]
• Linear time-invariant (LTI) system
– Linear: y(t) = H[α1x1(t)+α2x2(t)]
= α1H[x1(t)]+α2H[x2(t)]
= α1y1(t)+α2y2(t)
– Time-invariant: y(t-t0) = H[x(t-t0)]
 Signals and linear systems
• Impulse response and superposition integral
– The impulse response h(t) of an LTI system is the
response of the system to an impulse applied at
t=0: h(t) = H[δ(t)]
– LTI system:
N N
x (t )
= ∑ α δ ( t − t ) 
n
n 1=
n linearity
→ y (t ) ∑α h (t − t )
=
time invariance

n 1
n n

– Superposition integral
∞ ∞
y (t ) = x (t ) ∗ h (t ) = ∫ x ( λ ) h ( t − λ )d λ = ∫ x ( t − λ ) h ( λ )d λ
−∞ −∞
 Signals and linear systems
• Stability
– A LTI system is bounded-input, bounded-output
(BIBO) stable if every bounded input results in a
bounded output.
– A system is BIBO stable if and only if
∞

∫ h ( t ) dt < ∞
−∞
 Signals and linear systems
• Causality
– A system is causal if it does not anticipate the
input.
– For a time-invariant causal system, the impulse
response
h(t) = 0, t < 0
 Signals and linear systems
• Transfer (frequency-response) function
– Applying the convolution theorem of Fourier
transform
∞ ∞
y (t ) = x (t ) ∗ h (t ) = ∫ x ( λ ) h ( t − λ )d λ = ∫ x ( t − λ ) h ( λ )d λ
−∞ −∞

X ( f ) = ℑ{ x ( t )}
H ( f ) = ℑ{h ( t )}
Y ( f ) = ℑ{ y ( t )}

Y( f )= H( f )X ( f )
Transfer (frequency-
response) function of
the system
 Signals and linear systems
• Symmetry properties of H(f)
H ( f ) = H ( f ) e j∠H ( f )

Amplitude-response function Phase-response function

– Real-time function h(t)

H ( f=
) H (− f )
∠H ( f ) = −∠H ( − f )
Ex: Find the frequency-response function of the
lowpass RC filter
Differential equation

Fourier transform

Fourier transform pair

Impulse response of the filter
 Signals and linear systems
• Input-output relationships for spectral
densities
LTI
x(t) y(t)
H(f)

– For energy signals x(t) and y(t), energy spectral

densities: Gx(f)=X(f)2, Gy(f)=Y(f)2
G y ( f ) = H ( f ) Gx ( f )
2

– For power signals,

S y ( f ) = H ( f ) Sx ( f )
2
 Signals and linear systems
• Response to periodic inputs
– The steady-state response of a LTI system to
Aexp(j2πf0t)
yss ( t ) = H ( f 0 ) Ae j 2π f0t
– An arbitrary periodic input
∞
y (t ) = ∑
n = −∞
X n H ( nf 0 ) e j 2π nf0 t
∞
y (t ) ∑ H ( nf )
n = −∞
0 (
X n exp j  2π nf 0t + ∠X n + ∠H ( nf 0 )  )
 Signals and linear systems
• Response to periodic inputs
– Ex: Given the response of a filter
 f  − jπ f /10
H ( f ) = 2Π   e
 42 
and its input being a unit-amplitude triangular
signal with period of 0.1s.
Find the output of the filter.
 Signals and linear systems
• Distortionless transmission
– A system is distortionless if it introduces the same
attenuation and time delay to all spectral components of
the input, for then the output looks like the input.
y ( t=
) H 0 x ( t − t0 ) → H ( f=) H 0e− j 2π ft ; H 0 and t0 are constants
0

– Three main types of distortion:

• If the system is linear but the amplitude response is not
constant with f, the system introduces amplitude
distortion.
• If the system is linear but the phase shift is not a linear
function of f, the system introduces phase, or delay
distortion.
• If the system is not linear, we have nonlinear distortion
 Signals and linear systems
• Group and phase delay
– The group delay of the LTI system Phase response

1 dθ ( f )
Tg ( f ) = −
2π df
– For distortionless system,
θ(f )= −2π ft0 → Tg ( f ) = t0
– Group delay is the delay that a group of two or more
frequency components undergo in passing through a
linear system.
– The delay of a single-frequency component as the
input is defined as the phase delay θ(f)
Tp ( f ) = −
2π f
 Signals and linear systems
• Nonlinear distortion
– Consider a nonlinear system with the input-output
characteristic:
y ( t ) a1 x ( t ) + a2 x 2 ( t ) ; a1 and a2 are constants
=

=x ( t ) A1 cos (ω1t ) + A2 cos (ω2t )

– The output
= is y ( t ) a1  A1 cos (ω1t ) + A2 cos (ω2t ) 
 
desired output

1
2
( 1
)
+ a2 A12 + A22 + a2  A12 cos ( 2ω1t ) + A22 cos ( 2ω2t ) 
2
  
harmonic distortion

{ }
+ a2 A1 A2 cos (ω1 + ω2 ) t  + cos (ω1 − ω2 ) t 
 
intermodulation distortion
 Signals and linear systems
• Nonlinear distortion
 Signals and linear systems
• Ideal filter
– B is the single-sided bandwidth of the filter
– Lowpass
H LP ( f=
) 0 (
H Π f / 2 B ) e − j 2π ft0

– Highpass
( f ) H 0 1 − Π ( f / 2 B ) e− j 2π ft0
H HP =
– Bandpass
H BP ( f=
)  H1 ( f − f0 ) + H1 ( f + f0 ) e− j 2π ft0
where H1 ( f=) H 0Π ( f / B )
 Signals and linear systems
• Ideal filter
 Signals and linear systems
• Ideal filter
=hLP ( t ) 2 BH 0 sin c  2 B ( t − t0 ) 

hBP ( t ) = 2 H 0 B sin c  B ( t − t0 )  cos  2π f 0 ( t − t0 ) 

 Signals and linear systems
• Approximation of ideal lowpass filter by
realizable filters
– Ideal filters are noncausal, thus unrealizable
devices.
– nth-order Butterworth filter
ω3n
H BW ( s ) =
( s − s1 )( s − s2 ) ... ( s − sn )
where the poles s1, s2, …, sn are symmetrical w.r.t.
the real axis and equally spaced about a semicircle
of radius ω3 in the left half s plane and f3 = ω3/2π
is the cut-off frequency.
 Signals and linear systems
• Approximation of ideal lowpass filter by
realizable filters
– The Chebyshev lowpass filter
1
HC ( f ) =
1 + ε 2Cn2 ( f )
• ε is specified by the minimum allowable attenuation in
the passband
 f 
Cn ( =
f ) 2   Cn −1 ( f ) − Cn − 2 ( f ) =
, n 2,3,...
 fc 
f
C1 ( f ) =
fc
C0 ( f ) = 1
 Signals and linear systems
• Approximation of ideal lowpass filter by
realizable filters
the nominal delay
– The Bessel filter of the filter
• Cut-off frequency
ωc
f C (=
2π t0 )
−1
=
2π
• The nth-order Bessel filter
Kn
H BE ( f ) =
Bn ( f )
– Kn is a constant chosen to yield H(0)=1.
2
B0 ( f ) = 1
 f 
Bn ( f ) =
( 2n − 1) Bn−1 ( f ) −   Bn−2 ( f )  f 
 fc  B1 ( f ) = 1 + j  
 fc 
 Sampling theory
• Represent a signal x(t) as an ideal
instantaneous sampled waveform
Sampling interval
∞
=xδ ( t ) ∑ x ( nT ) δ ( t − nT )
n = −∞
S S

• Questions:
– What are the restrictions on x(t) and TS to allow
perfect recovery of x(t) from xδ(t)?
– How is x(t) recovered from xδ(t)?
 Sampling theory
• Theorem: if a signal x(t) contains no frequency
components for frequencies above f = W Hz,
then it is completely described by
instantaneous sample values uniformly spaced
in time with period TS < 1/2W. The signal can
be exactly reconstructed from the sampled
waveform by passing it through an ideal
lowpass filter with bandwidth B, where W < B
< fS-W with fS=1/TS. The frequency 2W is
referred to as the Nyquist frequency.
 Sampling theory
• Spectrum of the sampled signal
∞
Xδ ( f ) fs
= ∑ X ( f − nf )
n = −∞
s

• Sampling simply results in a periodic repetition of

X(f) in the frequency domain with a spacing fS.
• fS<2W: the separate terms in Xδ(f) overlap → no
way to recover x(t) from xδ(t) without distortion.
• fS>2W: the term for n=0 is easily separated from
the rest by ideal lowpass filtering.
 Sampling theory
• LPF

 f  − j 2π ft0
H ( f ) = H 0Π  e , W ≤ B ≤ fS − W
 2B 
• The output spectrum with xδ(t) at the input, is
Y ( f ) = f S H 0 X ( f ) e − j 2π ft0

Distortionless
y ( t ) f S H 0 x ( t − t0 )
= recovery of
x(t) from xδ(t)
 Sampling theory
• If the conditions of the sampling
theorem are not satisfied (either
because x(t) is not bandlimited
or because fS<2W), the
distortion (referred to as
aliasing in (a)) at the output of
the reconstruction filter is
inevitable.
• Error due to the nonideal
frequency response
characteristics of practical filters
(see (b)).
 Sampling theory
• Theorem: if a signal has a spectrum of W Hz
and upper frequency limit fu, then a rate fs at
which the signal can be sampled is 2fu/m,
where m is the largest integer not exceeding
fu/W. All higher sampling rates are not
necessarily usable unless they exceed 2fu.
• Ex.: fs = 2(3)/2 = 3 samples per second
∞
X δ ( f ) 3 ∑ X ( f − 3n )
=
n = −∞

BPF to recover x(t)

 Hilbert transform
• Definition
x (λ )
∞ ∞
x (t −η )
=xˆ ( t ) ∫= dλ ∫ dη
−∞
π (t − λ ) −∞
πη
 1, f > 0
x(t) H(f) = -jsgnf x̂ ( t ) 
=
sgn f =0, f 0
−1 f < 0


• Ex: Find the Hilbert transform of

x(t)=cos(2πf0t)
 Hilbert transform
• Properties
– The energy (or power) in a signal and its Hilbert
transform are equal.
– A signal and its Hilbert transform are orthogonal.
– If c(t) and m(t) are signals with nonoverlapping
spectra, where c(t) is lowpass and m(t) is highpass,
then
 Hilbert transform
• Analytic signals
– An analytic signal xp(t), corresponding to the real
signal x(t)
( t ) x ( t ) + jxˆ ( t )
x p=
– The envelope of a signal is defined mathematically
as the magnitude of the analytic signal xp(t).
– Spectrum of xp(t)
X p ( f ) X ( f ) [1 + sgn f ]
=
 Hilbert transform
• Complex envelope representation of bandpass
signals
– x(t) is a bandpass signal
– x ( t ) is a complex-valued lowpass signal (complex
envelope)
– f0 is a reference frequency chosen for
convenience.
x p ( t ) = x ( t ) e j 2π f0t
x ( t ) = x p ( t ) e − j 2π f0t
 Hilbert transform
• Complex envelope representation of bandpass
signals

( t ) x ( t ) e = x ( t ) + jx ( t )
x p=  j 2π f t 0

{
x ( t ) = Re x ( t ) e j 2π f0t }
= Re { x ( t )} cos ( 2π f 0t ) − Im { x ( t )} sin ( 2π f 0t )
 
inphase component quadrature component

• Ex: find the complex envelope corresponding

to x(t)=cos(22πt)
 Hilbert transform
• Complex envelope representation of bandpass
systems
h ( t ) ∗ x ( t ) =
y ( t ) = ℑ−1  H ( f ) X ( f )  Bandpass input

{
h ( t ) = Re h ( t ) e j 2π f0t }
x ( t ) = Re { x ( t ) e j 2π f 0 t
}
y (t ) =
1
2
{
Re y ( t ) e j 2π f0t }
• Ex: find the complex envelope of the filter
output t
x ( t ) = Π   cos ( 2π f 0t )
−ατt 
h ( t ) = α e u ( t ) cos ( 2π f 0t )
 Discrete Fourier transform (DFT)
• DFT N −1
=Xk ∑ n =
x e
n =0
− j 2π nk / N
, k 0,1,..., N − 1

– Xk is an approximation of the Fourier spectrum of

a signal at frequencies k/(NTS)
– x0, x1,…, xN-1 are N sample values of the signal
taken at TS-s intervals.
• Inverse DFT
N −1
=xn ∑ k =
X e
k =0
j 2π nk / N
, n 0,1,..., N − 1
Discrete Fourier transform (DFT) and
fast Fourier transform (FFT)
 Assignments
• 2.1, 2.4, 2.6, 2.7, 2.8
• 2.9, 2.10, 2.11, 2.12, 2.13
• 2.14
• 2.15, 2.16, 2.18
• 2.24, 2.25, 2.27, 2.28, 2.29, 2.30, 2.31, 2.32, 2.34,
2.36, 2.37
• 2.40, 2.41, 2.42
• 2.46, 2.47, 2.52, 2.53, 2.54, 2.58, 2.62, 2.63
• 2.66, 2.67, 2.68
• 2.72, 2.74