Chapter

2. Signals And Spectra

Chapter Objec7ves

Ø  Basic signal proper7es (DC, RMS, dBm and power)


Ø  Fourier transform and spectra

Ø  Linear systems and linear distor7on

Ø  Bandlimited signals and sampling

Ø  Discrete Fourier transform

Ø  Bandwidth of signals
1
 2.1 Proper7es of signals and noise
²  In communica+on systems, the received waveform is usually categorized
into two parts:
•  the desired part, called SIGNAL, contains the Informa7on
•  the undesired part, called NOISE

²  Proper+es of waveforms include:

•  DC value
•  Root-mean-square (RMS) value
•  Normalized power
•  Magnitude spectrum
•  Phase spectrum
•  Power spectral density
•  Bandwidth
 2.1 Proper7es of signals and noise
Physically realizable waveforms

Prac+cal waveform that that physically realizable waveforms (i.e. measurable

in a laboratory) sa+sfy several condi+ons:

²  The waveform has significant nonzero values over a composite +me interval
that is finite. --- produce finite amount of energy
²  The spectrum of the waveform has significant values over a composite
frequency interval that is finite. -- transmission medium has restricted bandwidth
²  The waveform is a con8nuous func8on of 8me. -- restricted bandwidth
²  The waveform has a finite peak value. -- physical devices protec8on
²  The waveform has only real values. That is, at any +me, it cannot have a
complex value a + jb, where b is nonzero. -- real waveform in real word

 2.1 Proper7es of signals and noise
Time Average Operator
1 T /2
DEFNITION. [•] = Tlim ∫ −T /2 [•]dt
−>∞ T

The +me average operator is a linear operator

α1w1 (t) + α 2 w 2 (t) = α1 w1 (t) + α 2 w2 (t)

DEFNITION. A waveform w(t) is periodic with period T0 if

w(t) = w(t + T0 ) for all t

THEOREM. If the waveform involved is periodic, the +me average operator

can be reduced to:
1 T0 /2
[•] =
T0
∫ −T0 /2
[•]dt
4
 2.1 Proper7es of signals and noise
DC Value

DEFNITION. The DC (Direct “Current”) value of a waveform w(t) is given by
its +me average, <w(t)>
1 T /2
Wdc = lim ∫ w(t)dt
−T /2
T −>∞ T

For any physical waveform, we are actually interested in evalua+ng the DC
value only over a finite interval of interest, say, from t1 to t2, so that the DC
Value is
1 t2

t2 − t1
∫ t1
w(t)dt

5
t1 t2
 2.1 Proper7es of signals and noise
Power

DEFNITION. Let v(t) denote the voltage across a set of circuit terminals,
and let i(t) denote the current into the terminal, as shown below.

the instantaneous power (incremental work divided by incremental
+me) associated with the circuit is given by

p(t) = v(t)i(t)

i(t)

+ v(t)

where the instantaneous power flows into the circuit when p(t) is posi+ve
and flow out of the circuit when p(t) is nega+ve
the average power

P = p(t) = v(t)i(t)
6
 2.1 Proper7es of signals and noise

Example 2-2. let the circuit of Fig. 2-2 (in textbook) contain a 120-V, 60-Hz
fluorescent lamp wired in a high-power-factor configura+on. Assume that the
voltage and current are both sinusoids and in phase (unity power factor), as
shown below. Find
1.) the DC value of this (periodic) voltage waveform?
2.) the instantaneous power?
3.) the average power?

7 7
 2.1 Proper7es of signals and noise
Voltage The DC values
Vdc = v(t) = V cosω 0 t
1 T0 /2
=
T0
∫ −T0 /2
V cosω 0 t ⋅ dt = 0

Where, ω 0 = 2π / T0 , and f0 = 1 / T0 = 60Hz

Similarly,
Current I dc = 0
The instantaneous power
1
p(t) = (V cosω 0 t)(I cosω 0 t) = VI(1+ cos2ω 0 t)
2
The average power
Instantaneous Power
1 VI T0 /2
P = VI(1+ cos2ω 0 t) =
2
∫ (1+ cos2ω 0t)
2T0 −T0 /2
VI
=
2
8
 2.1 Proper7es of signals and noise
RMS Value and Normalized Power

DEFNITION. The root-mean-square (RMS) value of w(t) is

Wrms = w 2 (t)


THEOREM. If a load is resis+ve (with unity power factor), the average power is
v 2 (t) V 2
P= = i 2 (t) R = rms = I rms
2
R = Vrms I rms
R R
DEFNITION. The average normalized power is
1 T /2
P = w 2 (t) = lim ∫ w 2 (t)dt
T −>∞ T −T /2

where w(t) represents a real voltage or current waveform

9
 2.1 Proper7es of signals and noise
Energy and Power Waveforms

DEFNITION. w(t) is a power waveform if and only if the normalize average

power P is finite and nonzero (i.e. 0 < P < ∞)

DEFNITION. The total normalized power is
T /2
E = lim ∫ w 2 (t)dt
T −>∞ −T /2

DEFNITION. w(t) is a energy waveform if and only if the total normalize

energy E is finite and nonzero (i.e. 0 < E < ∞)

10
 2.1 Proper7es of signals and noise
Energy and Power Waveforms

²  If a waveform is classified as either one of these types, it cannot be of the
other type

²  If w(t) has finite energy, the power averaged over infinite +me is zero


²  If the power ( averaged over infinite +me) is finite, the energy is infinite

²  However, mathema+cal func+ons can be found that have both infinite

energy and infinite power and, consequently, cannot be classified into
either of these two categories. (w(t) = e-t).

²  Physically realizable waveforms are of the energy type.
We can find a finite power for these!!

11
 2.1 Proper7es of signals and noise
Decibel

The decibel is a base 10 logarithmic measure of power ra+os. For example,
the ra+o of the power level at the output of circuit compared with that at

the input is oeen specified by the decibel gain instead of the actual ra+o

DEFNITION 1. The decibel gain of a circuit is.

average power out Pout
dB = 10 log( ) = 10 log( )
average power in Pin

12
 2.1 Proper7es of signals and noise
Decibel

DEFNITION 2. The decibel signal-to-noise ra7o .
P signal
! s (t) 2 $
(S / N ) = 10 log(
dB ) = 10 log ## &
&
P
2
noise " n (t) %
2

2
s (t) V
Since Signal power (S) = =
rms signal

R R
n 2 (t) n 2rms
noise power (N) = = noise

R R
This defini+on is equivalent to
V rms signal
(S / N )dB = 20 log( )
V rms noise
13
 2.1 Proper7es of signals and noise
Decibel

DEFNITION 3. The decibel power level with respect to 1 mW is

actual power level (watts)
dBm = 10 log( −3
)
10

=30 + 10log(Actual Power Level (wags))

Here the “m” in the dBm denotes a mlliwag reference. When a 1-W reference
level is used, the decibel level is denoted dBW; when a 1-kW reference level is
used, the Decibel level is denoted dBK.

14
 2.1 Proper7es of signals and noise
Phasors

DEFNITION. A complex number c is said to be a “phasor” if it is used to
represent a sinusoidal waveform. That is,

w(t) = c cos [ω 0 t + ∠c ] = Re {ce jω0t }

where the phasor
c = c e j∠c and Re{.} denotes the real part of the
complex quan+ty {.}.

The phasor (complex number) can also be wrigen in either Cartesian form

c = x + jy
or polar form
c = c e jϕ

15
 2.2 Fourier transform and spectra
DEFNITION. The Fourier Transform (FT) of a waveform w(t) is
∞
W ( f ) = f[w(t)] = lim ∫ [w(t)]e− j 2 π ft dt
T −>∞ −∞

Where f[*] denotes the Fourier transform of [*], and f is the frequency
parameter with units of hertz (i.e., 1/s). This defines the term frequency.
It is the parameter f in the Fourier transform.

W(f) is also called a two-sided spectrum of w(t), because both posiFve and
NegaFve frequency components are obtained from previous equaFon.

What is Fourier and Fourier Transform???

1
2.2 Fourier transform and spectra

What is Fourier and Fourier Transform ??

Fourier is a man, a genius

Name: Jean BapFste Joseph Fourier

Year: 1768-1830
Na@onality: French
Fields: MathemaFcian, physicist, historian

2
 2.2 Fourier transform and spectra
Fourier Series and Fourier Transformer
A weighted summaFon of Sines and Cosines of different frequencies can be used to
represent periodic (Fourier Series), or non-periodic (Fourier Transform) funcFons.
Is this true?
People didn’t believe that, including Lagrange, Laplace, Poisson, and other big wigs.

But, yes, this is true?

Possibly the greatest tool used in Engineering, one of the the fundaments of
modern communicaFon, control, signal processing, and etc.
3
 2.2 Fourier transform and spectra
Fourier Series
Approxima@ng a periodic signal with trigonometric func@ons

x! (t)
For a periodic signal which is periodic with period T 0 has the property
x! (t + T ) = x! (t)

T0/2

-A
T0

Periodic square-wave signal

4
 2.2 Fourier transform and spectra
Fourier Series
Approxima@ng a periodic signal with trigonometric func@ons

x! (t)
The best approximaFon to using only one trigonometric funcFon is
4A
x! (1) (t) = sin(ω 0 t)
π
!
x(t) x! (t)
(1)
ε!1 (t) = x(t)
! − x! (1) (t)

A A

T0 -A
-A
 2.2 Fourier transform and spectra
Fourier Series
Approxima@ng a periodic signal with trigonometric func@ons
x! (t)
Let’s try a three-frequency approximaFon to and see if the approximate
error can be reduced.
x! (3) (t) = b1 sin(ω 0 t) + b2 sin(2ω 0 t) + b3 sin(3ω 0 t)
ε!3 (t) = x! (t) − x! (3) (t) = x! (t) − b1 sin(ω 0 t) − b2 sin(2ω 0 t) − b3 sin(3ω 0 t)
!
x(t)
x! (3) (t) ε!3 (t) = x(t)
! − x! (3) (t)
A A

-A T0 -A
 2.2 Fourier transform and spectra
Fourier Series
Approxima@ng a periodic signal with trigonometric func@ons
x! (t)
Let’s try a 15-frequency approximaFon to and see if the approximate
error can be reduced.
x! (15) (t) = b1 sin(ω 0 t) + b2 sin(2ω 0 t) + . . . + b15 sin(15ω 0 t)
x! (t)
x! (15) (t) ε!15 (t) = x! (t) − x! (15) (t)
A A

T0 -A
-A
 2.2 Fourier transform and spectra
Fourier Series

Trigonometric Fourier Series (TFS)

∞ ∞

x! (t) = a0 + ∑ ak cos(kω 0 t) + ∑ bk sin(kω 0 t)

k=1 k=1

e jθ = cos(θ ) + j sin(θ )

Exponen@al Fourier Series (EFS)

∞

x! (t) = ∑ ck e jkw0t
k=−∞
2.2 Fourier transform and spectra
Fourier Transform
 2.2 Fourier transform and spectra
Fourier Transform for con@nuous-@me signals

Fourier Transform (Forward Transform)

∞
W ( f ) = ℑ[w(t)] = ∫ −∞ [w(t)]e− j2 π ft dt

Inverse Fourier Transform (Inverse Transform)

∞
−1
w(t) = ℑ [w(t)] = ∫ −∞ [W ( f )]e j2 π ft dt
 2.2 Fourier transform and spectra
Alterna@ve Evalua@on Techniques for FT Integral

²  Direct integraFon.
²  Tables of Fourier transforms or Laplace transforms.
²  FT theorems.
²  SuperposiFon to break the problem into two or more
simple problems.
²  DifferenFaFon or integraFon of w(t).
²  Numerical integraFon of the FT integral on the PC via
MATLAB or MathCAD integraFon funcFons.
²  Fast Fourier transform (FFT) on the PC via MATLAB or
MathCAD FFT funcFons.
11
 2.2 Fourier transform and spectra
DEFNITION. The Fourier Transform (FT) of a waveform w(t) is
∞
W ( f ) = ℑ[w(t)] = lim ∫ [w(t)]e− j2 π ft dt
T −>∞ −∞

W(f) is a complex funcFon of frequency, and can therefore be represented in as

Quadrature / Cartesian Magnitude-Phase / Polar

W ( f ) = X ( f ) + jY ( f ) W ( f ) = W ( f ) e jθ ( f )

W ( f ) = X 2( f ) +Y 2( f ) ⎛ Y( f ) ⎞
−1
θ ( f ) = tan ⎜ ⎟
⎝ X ( f ) ⎠
 2.2 Fourier transform and spectra
DEFNITION. The Inverse Fourier Transform (FT) of a waveform w(t) is
∞
w(t) = ∫ −∞ W ( f )e j 2 π ft
df

The funcFons w(t) and W(f) consFtute a Fourier transform pair

Fourier transform
w(t) W(f)
Inverse Fourier transform
Time domain Frequency domain
 2.2 Fourier Transform and Spectra

The waveform w(t) is Fourier transformable if it saFsfies both

Dirichlet condi@ons:

² Over any Fme interval of finite length, the funcFon w(t) is single
valued with a finite number of maxima and minima, and the
number of disconFnuiFes (if any) is finite.

∞
²  w(t) is absolutely integrable. That is, ∫ −∞ | w(t) | dt < ∞

Above condiFons are sufficient, but not necessary
 2.2 Fourier Transform and Spectra
A weaker sufficient condiFon for the existence of the Fourier
transform is:

∞ 2
E = ∫ w(t ) dt < ∞ Finite Energy
−∞

Where E is the normalized energy.

This is the finite-energy condi@on that is saFsfied by all physically
realizable forms.

Conclusion: All physical waveforms encountered in engineering
pracFce are Fourier transformable.

 2.2 Fourier Transform and Spectra
Example 2-3. Spectrum of an exponen@al pulse

Let w(t) be a decaying exponenFal pulse that is switched on at t = 0. That is

" −t
$ e if t > 0
w(t) = # find its spectrum?
$ 0 if t < 0
%

real magnitude

phase
image
 2.2 Fourier transform and spectra
Proper@es of Fourier Transforms
THEOREM. Spectral symmetry of real signals. If w(t) is real, then
W (− f ) = W * ( f )

The superscript asterisk denotes the conjugate operaFon.

x(t) = a + bj x * (t) = a − bj

Proper@es of the Fourier transform:

Ø  f, called frequency and having units of hertz, specifies the specific frequency
in the waveform w(t).
Ø  The FT looks for the frequency f in the w(t) over all Fme. That is, over

−∞ < t < ∞

Ø  W(f) can be complex, even though w(t) is real
Ø  If w(t) is real, then W(-f)=W*(f)
17
 2.2 Fourier transform and spectra
Example 2-4. Spectrum of a damped sinusoid
"
$ e−t/T sin ω 0 t if t > 0,T > 0
Let damped sinusoid be given by w(t) = #
$ 0 if t<0
%
find its spectrum?

18
 2.2 Fourier transform and spectra
Proper@es of Fourier Transforms

Parseval’s Theorem:
∞ ∞

∫ 1
w *
(t)w
2 (t)dt = ∫ 1 2 ( f )df
W ( f )W *

−∞ −∞

If w1(t)=w2(t)=w(t), then the theorem reduces to
Rayleight’s Energy Theorem:
∞
2
∞
2
∫ w (t)
1 dt = ∫ W (f)
1 df
−∞ −∞

The energy calculated from the @me domain is equal to

the energy calculated from the frequency domain
19
 2.2 Fourier transform and spectra
Parseval’s Theorem and Energy Spectral Density

DEFNITION. The Energy Spectral Density (ESD) is defined for energy waveforms by

We can see that the total normalized energy is given by the

area under ESD func@on
2.2 Fourier transform and spectra
Some Fourier Transform Theorems
 2.2 Fourier transform and spectra
Dirac Delta Func@on
DEFINATION. The Dirac delta funcCon is defined by
δ (x)
∞ δ(x)
∫ w(x)δ (x)dt = w(0)
−∞

t
where w(x) is any funcFon that is conFnuous at x = 0.
An alternaFve definiFon of δ(x) is:

#
% 0 if x≠0 ∞

δ (x) = $ and ∫ δ (x)dx = 1

% ∞ if x=0 −∞

&
 2.2 Fourier transform and spectra
Dirac Delta Func@on
The SiDing Property of is
δ (x)
∞

∫ w(x)δ (x − x )dt = w(x )

0 0
−∞
 2.2 Fourier transform and spectra
Unit Step Func@on

! u(t)
# 1 if t>0
u(t) = " 1
# 0 if t<0
$
t

Time shiD of the unit-step funcFon

" u(t)
$ 1 if t > t1
u(t − t1 ) = # 1
$ 0 if t < t1
%
t1 t
 2.2 Fourier transform and spectra
The rela@onship between unit-step and Delta funcFons
∞
du
u(t) = ∫ δ (λ )d λ δ (t) =
−∞
dt
 2.2 Fourier transform and spectra
Example 2-5. Spectrum of a sinusoid
Find the spectrum of a sinusoidal voltage waveform that has a frequency f0 and
A peak value of A volts. That is where
v(t) = Asin ω 0 t ω 0 = 2π f0 find its spectrum?

26
 2.2 Fourier transform and spectra
Rectangular Pulses
DEFINATION. The single rectangular pulse is denoted as ∏(•)
"
t Δ$ 1, | t |< T / 2
∏( ) = #
T $ 0, | t |> T / 2
%

sin x
DEFINATION. Denoted the funcFon
Sa(•) Sa(x) =
x
 2.2 Fourier transform and spectra
Example 2-6. Spectrum of a rectangular pulse

Find the spectrum of a rectangular pulse w(t) = ∏ (t / T )

28
 2.2 Fourier transform and spectra
Spectrum of a Rectangular Pulse
⎛t ⎞
w(t ) = Π ⎜ ⎟ ⇔ W ( f ) = T ⋅ Sa ( fT )
⎝T ⎠
•  Rectangular pulse is a time window.
•  FT is a sinc function, infinite frequency content.
•  Shrinking time axis causes stretching of frequency axis.
•  Signals cannot be both time-limited and bandwidth-limited.

Note the inverse relaFonship between the pulse width T and the zero crossing 1/T
 2.2 Fourier transform and spectra
Triangular Pulses
DEFINATION. The single triangular funcCon is denoted as Λ(•)
$
& 1− | t | , | t |≤ T
t Δ& T
Λ( ) = %
T &
&' 0, | t |> T
 2.2 Fourier transform and spectra
Example 2-7. Spectrum of a triangular pulse
Find the spectrum of a triangular pulse w(t) = Λ(t / T )

⎛t ⎞ 2
w(t ) = Λ ⎜ ⎟ ⇔ W ( f ) = T ⋅ Sa (π fT )
⎝T ⎠
31
 2.2 Fourier transform and spectra
Convolu@on
DEFNITION. The convoluCon of a waveform w1(t) with a wave w2(t) to produce
a third waveform w3(t) is ∞
ω 3 (t) = ω1 (t)∗ ω 2 (t) = ∫ ω1 (λ )ω 2 (t − λ )d λ
−∞
∞

= ∫ ω1 (λ )ω 2 (−(λ − t))d λ
−∞

Where is a shorthand notaFon for this integraFon operaFon and * is read

ω1 (t)∗ ω 2 (t)
“convolved with.”

The convoluCon can be obtained through three steps:

1.  Time reversal of to obtain .
ω 2 (t) ω 2 (−λ )
ω2
2.  Time shiding of by t seconds to obtain ω 2 (−(λ − t))
ω1
3.  Mul@plying this result by to form the integrand
ω1 (λ )ω 2 (−(λ − t))

32
 2.2 Fourier transform and spectra
Convolu@on of a rectangle with and exponen@al
ω1 (t) = e−t u(t) and ω 2 (t) = ∏ (t −1)

33
 2.2 Fourier transform and spectra
Example 2-8. Convolu@on of a rectangle with an exponen@al
# 1 &
%t − T (
let w1 (t) = Π % 2 (
% T (
$ '
and

w2 (t) = e−t/T u(t)

34
 2.3 Power spectral density and autocorrela@on
func@on
Power spectral density
DEFNITION. The power spectral density (PSD) for a determinisFc
power waveform is
# W (f)2 &
pw ( f ) = lim %% r (
(
T −>∞
$ T '
Where and has units of wa6s per hertz.
wT (t) ↔ WT ( f ) pw ( f )
Note:
1.) The PSD represents the normalized power of a waveform in its
frequency domain
2.) The PSD is always a real nonnegaCve funcFon of frequency.
3.) The PSD is not sensiCve to the phase spectrum of w(t).
35
2.3 Power spectral density and autocorrela@on
func@on

The Normalized Average Power

∞
P =< w 2 (t) >= ∫ P (f)
w
−∞

This means the area under the PSD func@on is the normalized
average power

36
2.3 Power spectral density and autocorrela@on func@on
Autocorrela@on Func@on
DEFNITION. The AutocorrelaCon of a real (physical) waveform is
1 T /2
Rw (τ ) = ω (t)ω (t + τ ) = lim ∫ ω (t)ω (t + τ )dt
T −>∞ T
−T /2

Wiener-Khintchine Theorem: The PSD and the autocorrelaCon

funcFon are Fourier transform pairs:
Rw (τ ) ↔ Pw ( f )

The PSD can be evaluated by either of the following two methods:

²  Direct method: by using the definiFon.
²  Indirect method: by first evaluaFng the autocorrelaFon funcFon
and then taking the FT.
Pw ( f ) = ℑ[Rw (τ )]
37
2.3 Power spectral density and autocorrela@on func@on

The average power can be obtained by any of the

four techniques

P =< w 2 (t) >= Wrms

2
= ∫ P ( f )df = R
w W (0)
−∞

38
2.3 Power spectral density and autocorrela@on func@on

Example 2-9. PSD of a sinusoid

let w(t) = sin ω 0 t

39
2.4 Orthogonal Series Representa@on of Signal and Noise
Orthogonal Func@on
DEFNITION. FuncFons φn(t) and φm(t) are said to be orthogonal
with respect to each other over the interval a < t < b
if they saFsfy the condiFon
" &
b
$ 0 n≠m $
∫ ϕ n (t)ϕ m (t)dt = #
*
' = K nδnm
a $ Kn n = m $
% (
# '
% 0 n≠m %
δnm ≡ $ (
% 1 n=m %
& )
²  δnm is called the Kronecker delta funcCon
²  If the constants Kn are all equal to 1 then the φn(t) are said to be
orthonormal funcCons.
40
2.4 Orthogonal Series Representa@on of Signal and Noise
Orthogonal Series
Assume that w(t) represents some pracFcal waveform (signal, noise,
or signal-noise combinaFon) that we wish to represent over the
interval a < t < b. Then we can obtain an equivalent orthogonal series
representaFon by using the following theorem.
THEOREM. w(t) can be represented over the interval (a,b) by the series
w(t) = ∑ anϕ n (t)
n

where the orthogonal coefficients are given by

1 b
an =
Kn
∫a w(t)ϕ *
n (t)dt

And the range of n is over the integer values that correspond to the
subscripts that were used to denote the orthogonal funcFon in the
complete orthogonal set 41
 2.4 Orthogonal Series Representa@on of Signal and Noise
Applica@on of Orthogonal Series
Ø  It is also possible to generate w(t) from the ϕj(t) funcFons and the coefficients aj.

Ø  In this case, w(t) is approximated by using a reasonable number of the ϕj(t) funcFons.

w(t) is realized by
adding weighted
versions of
orthogonal
func@ons

42
 2.5 Fourier Series
Complex Fourier Series
The complex Fourier series uses the orthogonal exponenFal funcFon

THEOREM. A physical waveform (i.e. finite energy) may be represented

over the interval a < t < a+T0 by the complex exponenFal Fourier series
∞
w(t) = ∑c e n
jnω 0t

n=−∞

where the complex (phasor) Fourier coefficient are

1 a+T0
cn = ∫ w(t)e− jnw0t dt
T0 a

2π
and where ω 0 = 2π f0 =
T0

43
 2.5 Fourier Series
Complex Fourier Series
∞
1 a+T0
w(t) = ∑c e n
jnω 0t cn = ∫ w(t)e− jnw0t dt
n=−∞
T0 a

²  Cn is the Fourier Series. In general, it is a complex number. The Fourier coefficient

C0 is equivalent to the DC value of the waveform w(t).

²  If the waveform w(t) is periodic with period T0, this Fourier series representaFon
Is valid over all Fme.

²  For this case of periodic waveforms, the choice of a is arbitrary and is usually
taken to be a = 0 or a = -T0/2 for mathemaFcal convenience.

²  The frequency f0 = 1/T0 is said to be the fundamental frequency and the
frequency nf0 is said to be the nth harmonic frequency, when n>1.
44
 2.5 Fourier Series
Some Proper@es of the Complex Fourier Series

45
 2.5 Fourier Series
Some Proper@es of the Complex Fourier Series

Note that these proper@es for the complex Fourier series coefficients are similar
to those of Fourier transform as given Sec. 2-2
46
 2.5 Fourier Series
Quadrature Fourier Series
The Quadrature Form of the Fourier series represenFng any physical waveform
w(t) over the interval a < t < a+T0 is,
∞ ∞
w(t) = ∑ an cos(nω 0 t) +∑ bn sin(nω 0 t)
n=0 n=0
Where the orthogonal func@ons are cos(nw0t) and sin(nw0t). we find that
these Fourier coefficients are given by

47
 2.5 Fourier Series
Polar Fourier Series
The Polar Form of the Fourier series represenFng any physical waveform is,
∞
w(t) = D0 + ∑ Dn cos(nω 0 t + ϕ n )
n=1
Where w(t) is real and
"
$ D0 , n = 0
an = # bn = −Dn sin ϕ n n ≥1
$ Dn cos ϕ n , n ≥ 1
%

These two equaFons may be inverted, we got

" "
$$ a0 , n = 0 c0 , n = 0 " bn %
$ −1
Dn = # =# ϕ n = − tan $ ' = ∠cn , n ≥ 1
$ an2 + bn2 , n ≥ 1 $ 2 | cn |, n ≥ 1 # an &
$% %

48
2.5 Fourier Series

What is the best form to use?

 2.5 Fourier Series
Line Spectra for Periodic Waveforms
THEOREM. If w(t) is periodic with period T0 and is represented by
∞ ∞
w(t) = ∑ h(t − nT ) = ∑ c e
0 n
jnw0t

n=−∞ n=−∞
!
where # T0
# w(t), t<
h(t) = " 2
#
#$ 0, elsewhere

Then the Fourier coefficients are given by

cn = f0 H (nf0 )

and where
H ( f ) = ℑ[h(t)] and f0 = 1/T0
50
 2.5 Fourier Series

THEOREM. For a periodic waveform w(t), the normalized power is

∞
2
2
Pw = w (t) = ∑c n
n=−∞

Where the {cn} are the complex Fourier coefficients for the waveform

51
 2.5 Fourier Series

THEOREM. For a periodic waveform w(t), the power spectral density

(PSD) is
∞
2
P( f ) = ∑c n δ ( f − nf0 )
n=−∞

Where T0 = 1/f0 is the period of the waveform, and the {cn} are the
complex Fourier coefficients for the waveform

52
 2.6 REVIEW OF LINEAR SYSTEMS
Linear Time-Invariant System

An electronic filter or system is Linear when Superposi,on holds, that is

when,
y (t ) = L[a1 x1 (t ) + a2 x2 (t )] = a1L[ x1 (t )] + a2 L[ x2 (t )]

•  Where y(t) is the output and x(t) = a1x1(t)+a2x2(t) is the input.

•  L[.] denotes the linear system operator ac>ng on [.].

y (t ) = x(t ) ∗ h(t )
Y ( f ) = X ( f )H ( f )

1
 2.6 REVIEW OF LINEAR SYSTEMS
Linear Time-Invariant System

Example. Are the following system is linear system or not?

1.) y(t) = 5x(t)

2.) y(t) = 5x(t) + 3

2
 2.6 REVIEW OF LINEAR SYSTEMS
Linear Time-Invariant System
CondiAons for Ame-invariance
Sys{x(t)} = y(t) implies that Sys{x(t-τ)} = y(t-τ)

•  If the system is >me invariant for any delayed input x(t – t0), the
output is delayed by just the same amount y(t – t0).
•  That is, the shape of the response is the same no maLer when the
input is applied to the system.

3
 2.6 REVIEW OF LINEAR SYSTEMS
Linear Time-Invariant System

Example. Are the following system is Fme-invariant system or not?

1.) y(t) = 5x(t)

2.) y(t) = 5x(t) + 3

4
 2.6 REVIEW OF LINEAR SYSTEMS
Impulse Response
The impulse response is the solu>on to the differen>al equa>on when
the forcing func>on is a Dirac delta func>on. That is
y(t) = h(t) when x(t) = δ(t).

A general waveform at the input may be approximated by
∞
x(t) = ∑ x(nΔt) [δ (t − nΔt)]Δt
n=0
The output may be approximated by
∞
y(t) = ∑ x(nΔt) [ h(t − nΔt)]Δt
n=0
This expression becomes the exact result as Δt becomes zero,
leSng nΔt = λ, we obtain
∞
ConvoluAon y(t) = ∫ −∞
x(λ )h(t − λ )d λ = x(t)∗ h(t)
5
 2.6 REVIEW OF LINEAR SYSTEMS
Transfer FuncAon
²  The output waveform for a >me-invariant network can be obtained
by convolving the input waveform with the impulse response of
the system.
²  The spectrum of the output signal is obtained by taking the Fourier
transform of both sides. Using the convolu>on theorem, we get

Y ( f ) = X( f )H ( f )

or H( f ) =
Y( f )
X( f )
or is said to be the transfer func,on or frequency
H ( f ) = ℑ [ h(t)]
response of the network.

6
 2.6 REVIEW OF LINEAR SYSTEMS
Transfer FuncAon
The impulse response and frequency response are a Fourier transform pair

h(t) H(f)

Of course, the transfer func>on H(f) is, in general, a complex quan>ty and
can be wriLen in polar form as

H ( f ) = H ( f ) e j∠H ( f )

where is the amplitude (or magnitude) response and

| H( f ) |
# Im{H ( f )} &
−1
θ ( f ) = ∠H ( f ) = tan % (
$ Re{H ( f )} '

is the phase response of the network.

7
 2.6 REVIEW OF LINEAR SYSTEMS
Transfer FuncAon

Since h(t) is a real func>on of >me (for real networks), it follows

²  |H(f)| is an even func>on of frequency ( |H(-f)| = |H(f)| )
²  θ(f) is an odd func>on of frequency. ( θ(-f) = -θ(f) )

Example Output
Input

x(t) = A cos w0 t y(t) = A H ( f0 ) cos[w0 t + ∠H ( f0 )]

j∠H ( f0 )
H ( f0 ) = H ( f0 ) e
∞ ∞

X( f ) = ∑ c δ ( f − nf )
n 0
Y( f )= ∑ c H (nf )δ ( f − nf )
n 0 0
n=−∞ n=−∞

8
 2.6 REVIEW OF LINEAR SYSTEMS
Power Transfer FuncAon

We also can obtain the rela>onship between the power spectral density
(PSD) at the input, and the output for a linear >me-invariant network as:
1 2
Py ( f ) = lim YT ( f )
T →∞ T

using Y ( f ) = X( f )H ( f )
2 1 2
we get Py ( f ) = H ( f ) lim XT ( f )
T →∞ T

or
2
Py ( f ) = H ( f ) Px ( f )

Py ( f ) 2
Power transfer funcAon Gh ( f ) = = H( f )
Px ( f )

9
 2.6 REVIEW OF LINEAR SYSTEMS
DistorAonless Transmission
In communica>on systems, a distor>onless channel is oben desired.
This implies that the channel output is just propor>onal to a delayed
version of the input
y(t) = A(x − Td )
Where A is the gain (which may be less than unity) and Td is the delay

In the frequency domain

Y ( f ) = AX( f )e− j 2 π fTd
The transfer funcAon of the channel is
Y( f )
H( f ) = = Ae− j 2 π fTd
X( f )
10
 2.6 REVIEW OF LINEAR SYSTEMS
DistorAonless Transmission
For a linear >me-invariant system, two requirements are needed as
Distor>onless system.
²  The amplitude response if flat. That is.
|H(f)| = constant = A
No Amplitude DistorAon
²  The phase response is a linear func>on of frequency. That is
θ ( f ) = ∠H ( f ) = −2π fTd
No Phase DistorAon
We define the ,me delay of the system as
1 1
Td ( f ) = − θ( f ) = − ∠H ( f )
2π f 2π f
For distor>onless system, Td(f) must be constant.
2.6 REVIEW OF LINEAR SYSTEMS
Example 2-18 RC Low-Pass Filter

12
2.6 REVIEW OF LINEAR SYSTEMS
Example 2-18 DistorFon caused by a filter
 2.6 REVIEW OF LINEAR SYSTEMS
Effects of DistorAon

Audio
Human ear more sensi>ve to amplitude distor>on, but less so to
phase distor>on

Analog Video
Human visual system more sensi>ve to >me delay errors, which
result in smearing of edges, but less so to intensity varia>on

Digital Signals
Pulses smearing into other >me slots – “Inter-Symbol
Interference” (ISI)
 2.7 Bandlimited Signals and Noise
Bandlimited Waveform

DEFINIATION. A waveform w(t) is said to be (absolutely) bandlimited to

B hertz, if
W(f) = [w(t)] = 0 for |f| ≥ B

DEFINIATION. A waveform w(t) is said to be (absolutely) @me limited if

w(t) = 0, for |t| ≥ T

 2.7 Bandlimited Signals and Noise
Bandlimited Waveform
THEOREM. A absolutely bandlimited waveform cannot be absolutely @me
limited, and vice versa.
A physical waveform that is @me limited, may not be absolutely bandlimited, but it may be
bandlimited for all prac@cal purposes in the sense that the amplitude spectrum has a
negligible level above a certain frequency.
 2.7 Bandlimited Signals and Noise
Bandlimited Waveform
THEOREM. A absolutely bandlimited waveform cannot be absolutely @me
limited, and vice versa.
A physical waveform that is @me limited, may not be absolutely bandlimited, but it may be
bandlimited for all prac@cal purposes in the sense that the amplitude spectrum has a
negligible level above a certain frequency.
 2.7 Bandlimited Signals and Noise
Sampling Theorem
Sampling Theorem. Any physical waveform may be represented over the
interval -∞ < t < ∞ by
∞
sin {π fs [t − (n / fs )]}
w(t) = ∑a n
π fs [t − (n / fs )]
n=−∞

∞ sin {π fs [t − (n / fs )]}
where an = fs ∫ w(t) dt
−∞
π fs [t − (n / fs )]
And fs is a parameter that is assigned some convenient value greater
than zero. Furthermore, if w(t) is bandlimited to B hertz and fs >= 2B,
then previous equa@on becomes the sampling func@on representa@on,
where
an = w(n/fs)
That is, for fs >= 2B, the orthogonal series coefficients are simply the
values of the waveform that are obtained when the waveform is
sampled very 1/fs seconds.
 2.7 Bandlimited Signals and Noise
Sampling Theorem
²  The MINIMUM SAMPLING RATE allowed for reconstruc@on without error is
called the NYQUIST FREQUENCY or the Nyquist Rate. ( fs )Min = 2B

²  Suppose we are interested in reproducing the waveform over a T0-sec
interval, the minimum number of samples that are needed to reconstruct the
waveform is:
T0
N= = fsT0 ≥ 2BT0
1 / fs
•  There are N orthogonal func@ons in the reconstruc@on algorithm. We can
say that N is the Number of Dimensions needed to reconstruct the T0-second
approxima@on of the waveform.

•  The sample values may be saved, for example in the memory of a digital
computer, so that the waveform may be reconstructed later, or the values
may be transmibed over a communica@on system for waveform
reconstruc@on at the receiving end.
 2.7 Bandlimited Signals and Noise
Sampling Theorem
Example 2-19. Sampling theorem for a rectangular pulse
 2.7 Bandlimited Signals and Noise
Sampling Theorem
Example 2-19. Sampling theorem for a rectangular pulse
 2.7 Bandlimited Signals and Noise
Sampling Theorem
Example 2-19. Sampling theorem for a rectangular pulse
 2.7 Bandlimited Signals and Noise
Sampling Theorem
Example 2-19. Sampling theorem for a rectangular pulse
 2.7 Bandlimited Signals and Noise
Impulse Sampling and Digital Signal Processing
The impulse-sampled series is another orthogonal series. It is obtained when the
(sin x) / x orthogonal func@ons of the sampling theorem are replaced by an
orthogonal set of delta (impulse) funcBons. The impulse-sampled series is
iden@cal to the impulse-sampled waveform ws(t): both can be obtained by
mul@plying the unsampled waveform by a unit-weight impulse train, yielding
∞ ∞
ws (t) = w(t) ∑ δ (t − nTs ) = ∑ w(nT )δ (t − nT )
s s
n=−∞ n=−∞

Waveform Impulse sampled waveform

 2.7 Bandlimited Signals and Noise
Impulse Sampling and Digital Signal Processing

∞ ∞
ws (t) = w(t) ∑ δ (t − nTs ) = ∑ w(nT )δ (t − nT )
s s
n=−∞ n=−∞

Take the Fourier transform on both sides of this equa@on:

1 % ∞ jnw t ( 1 ∞
Ws ( f ) = W ( f )* ℑ' ∑ e s * = W ( f )* ∑ ℑ[e jnwst ]
Ts &n=−∞ ) Ts n=−∞
∞
1
= W ( f )* ∑ δ ( f − nfs )
Ts n=−∞

1 ∞
Or = ∑ W ( f − nfs )
Ts n=−∞
 2.7 Bandlimited Signals and Noise

Ø  The spectrum of the impulse sampled signal is the spectrum of the unsampled signal
that is repeated every fs Hz, where fs is the sampling frequency (samples/sec).

Ø  This is quite significant for digital signal processing (DSP).

Ø  This technique of impulse sampling maybe be used to translate the spectrum of a

signal to another frequency band that is centered on some harmonic of the sampling
frequency.
 2.7 Bandlimited Signals and Noise
Dimensionality Theorem

THEOREM: When BT0 is large, a real waveform may be completely specified by

N=2BT0
independent pieces of informa@on that will describe the waveform over a T0 interval. N
is said to be the number of dimensions required to specify the waveform, and B is the
absolute bandwidth of the waveform.

Ø  The informa@on which can be conveyed by a bandlimited waveform or a
bandlimited communica@on system is propor@onal to the product of the bandwidth of
that system and the @me allowed for transmission of the informa@on.

Ø  The dimensionality theorem has profound implica@ons in the design and

performance of all types of communica@on systems.
 2.8 Discrete Fourier Transform

DEFINIATION: The discrete Fourier transform (DFT) is defined by

k=N−1
X(n) = ∑ x(k)e− j(2 π /N )nk
k=0
Where n = 0, 1, 2, …, N-1, and the inverse discrete Fourier transform (IDFT)
Is defined by
k=N−1
1
x(k) = ∑ X(n)e j(2 π /N )nk
N k=0

Where k = 0, 1, 2, …, N-1,

²  The defini@on could be different according to different authors, which

Only effect the “scale factor” and “frequency factor”.

²  The fast Fourier transform (FFT) is a fast algorithm for evalua@ng the DFT
 2.8 Discrete Fourier Transform
Using the DFT to Compute the ConKnuous Fourier Transform
In digital signal processing, we use DFT (approxima@on)to represent
CFT (truth) through three steps.

²  Step 1: the @me waveform is first windowed (truncated) over the interval
(0, T) so that only a finite number of samples N are needed
"
$ w(t), 0 ≤ t ≤ T ( t − (T / 2) +
ww (t) = # = w(t)∏ * -
$ 0, otherwise ) T ,
%

²  Step 2: do the Fourier transform on the windowed waveform

∞ T
Ww ( f ) = ∫ −∞ ww (t)e− j2 π ft dt = ∫ 0 w(t)e− j2π ft dt
 2.8 Discrete Fourier Transform
Using the DFT to Compute the ConKnuous Fourier Transform
In digital signal processing, we use DFT (approxima@on)to represent
CFT (truth) through three steps.
²  Step 3: approximate the CFT by using a finite series to represent the
integral
N−1
Ww ( f ) f =n/T ≈ ∑ w(kΔt)e− j(2π /N )nk Δt
k=0
Where t = kΔt, f = n / T, dt = Δt, and Δt = T / N

The rela@on between CFT and DFT is

Ww ( f ) f =n/T ≈ ΔtX(n)
k=N−1
Where X(n) = ∑ x(k)e− j(2 π /N )nk
k=0
2.8 Discrete Fourier Transform

The DFT may give significant

errors when it is used to
approximate the CFT. The
errors are due to a number
of factors that may be
categorized into three basic
effects: leakage, aliasing,
And the picket-fence effect.
 2.8 Discrete Fourier Transform
Using the DFT to Compute the Fourier Series
The DFT may be also used to evaluate the coefficients for the complex
Fourier series.
1 T
From cn = ∫ w(t)e− j2 π nf0t dt
T 0
We approximate this integral by using a finite series
N−1
1
cn ≈ ∑ w(kΔt)e− j(2 π /N )nk Δt
T k=0
where t = kΔt, f = n / T, dt = Δt, and Δt = T / N
The Fourier series coefficient is related to the DFT by
1
cn ≈ X(n)
N
 2.8 Discrete Fourier Transform
Using the DFT to Compute the Fourier Series

The Fourier series coefficient is related to the DFT by

1
cn ≈ X(n)
N
For posi@ve n, we use
1
cn = X(n) 0≤n< N /2
N

For nega@ve n, we use

1
cn = X(N + n) −N / 2 < n < 0
N
 2.9 Bandwidth of Signals
In engineering definiKons, the bandwidth is taken to be the width of
posiKve frequency band.
We will give six engineering defini@ons and one legal defini@on of
Bandwidth that are ojen used.
²  Absolute bandwidth is f2 – f1: where the spectrum is zero outside the
Interval f1 < f < f2 along the posi@ve frequency axis.


 2.9 Bandwidth of Signals
²  3-dB bandwidth (or half-power bandwidth) is f2 – f1: where for frequency
inside the band f1 < f < f2 , the magnitude spectra, say, |H(f)|, fall no lower
than 1/√2 @mes the maximum value of |H(f)|, and the maximum value
occurs at a frequency inside the band.

3dB
 2.9 Bandwidth of Signals
²  Equivalent noise bandwidth: the width of a fic@@ous rectangular spectrum
such that the power in that rectangular band is equal to the power
associated with the actual spectrum over posi@ve frequencies.

1 ∞ 2
Beq = 2
| H ( f0 ) |
∫0 H ( f ) df
 2.9 Bandwidth of Signals
²  Null-to-null bandwidth (or zero-crossing bandwidth) is f2 – f1: where f2 is
the first null in the envelope of the magnitude spectrum above f0 and, for
bandpass system, f1 is the first null in the envelope below f0, where f0 is
the frequency where the magnitude spectrum is maximum. For baseband
systems, f1 is usually zero.

|X(f)|
Null-to-null Bandwidth Bn

0
2Bn
 2.9 Bandwidth of Signals
²  Bounded spectrum bandwidth is f2 – f1 such that outside the band
f1 < f < f2 , the PSD, which is propor@onal to |H(f)|2, must be down by
at least a certain amount, say 50 dB, below the maximum value of the

power spectral density.

²  Power bandwidth is f2 – f1 where f1 < f < f2 defines the frequency band
in which 99% of the total power resides. This is similar to the FCC defini@on
of occupied bandwidth, which states that the power above the the upper
band edge f2 is 0.5% and the power below the lower band edge is 0.5%,
leaving 99% of the total power within the occupied band.

²  FCC bandwidth is an authorized bandwidth parameter assigned by the
FCC to specifiy the spectrum allowed in communica@on systems.