Digital Signal Processing Notes
Digital Signal Processing Notes
Lian Zhao
Professor
Department of Electrical and Computer Engineering
Ryerson University
Source
encoder
Channel
encoder
Pulse
modulate
Bandpass
modulate
Channel
2
1.5
1
0.5
0.5
0
0.5
0.5
1
1
1.5
2
1.5
2.5
2
2
Time [sec]
x 10
2
Time [sec]
4
3
x 10
Knowing the format of the transmitted signal (but not the order of the pulses), we can still
correctly identify the most likely (probability theory) pulse train which might have been
transmitted. Hence
y(t) 0 1
1 5
With amplitude shift keying, as long as the noise does not change the amplitude from
one level to another level, the original bits can be recovered.
With QAM: as long as the received signal is more close to the original symbol than
its neighboring symbols.
Each repeater can regenerate the analog modulated signals from demodulated bits
(noise do not accumulate).
Source coding: can apply digital compression techniques to reduce data rate subject to
distortion criterion.
Error correction coding: can insert parity bits before sending data to allow
detection/correction of errors.
Different signals can be multiplexed more easily (Internet packets can contain any types
of signals, cell phones can send different types of data).
6
(bits/s)
g (t)dt =
|g(t)|2 dt
For example, -30 dBm represents signal power of 106 W in linear scale.
10
Deterministic signal: there is no uncertainty with its value at any time. For example:
x(t) = 5 cos(10t)
Random signal: some uncertainty. For example, noise process. We then analyze its
statistical properties.
Periodic and nonperiodic signals: a signal x(t) is called periodic in time if there exits a
constant T0 > 0 such that
x(t) = x(t + T0 )
The smallest value of T0 that satisfies this condition is called the period of x(t).
Continuous time and discrete time signals
A continuous time signal x(t) is a continuous function of time and t can take any real
value; For example: x(t) = 5 cos(10t), audio and vedio recordings.
A discrete signal x(kT ) is one that exists only at discrete times. Sometimes, we
denote x(kT ) as x(k) where k { , 3, 2 1, 0, 1, 2, 3 }.
11
A signal whose amplitude can have any value in a continuous range is an analog
signal.
A digital signal is one whose amplitude can take on only a finite number of values.
The term continuous time and discrete time qualify the nature of signal along the
time (horizontal) axis.
The term analog and digital describes the nature of the signal amplitude (vertical)
axis.
12
g(t)
g((t t0 ))
13
t %= 0
(t)(t T ) = (T )(t T )
!
(t) dt = 1
x(t)(t t0 ) dt = x(t0 )
1 t0
u(t) =
0 o.w.
n=
(t n)
14
0
|t| 1
15
sinc(x) =
sin(x)
x
x0
x0
sin(x)
cos(x)
= lim
=1
x0
x
1
4. sinc(x) is the product of an oscillating signal sin x (of period 2) and a monotonically
decreasing function 1/x. Therefore, sinc(x) exhibits sinusoidal oscillations of period 2,
with amplitude decreasing continuously as 1/x.
5. In summary, sinc(x) is an even oscillating function with decreasing amplitude. It has a
unit peak at x = 0 and zero crossing at integer multiples of .
16
!
1
1
Dn =
xp (t) ej2nf0 t dt and f0 =
(2)
T0 T0
T0
The FS coefficients Dn , n = , 2, 1, 0, 1 are usually complex valued. We can use
the set of FS coefficients {Dn } to sketch the two-sided line spectrum.
Example
'
: pulse train xp (t) =
(t nT0 ). Then xp is periodic, and its FS as
! T0 /2
1
1
1
Dn =
xp (t)ej2nf0 t dt =
(f0 =
)
T0 T0 /2
T0
T0
1 % j2nf0 t
e
T0 n=
(3)
17
1 %
n
F(xp (t)) =
f
T0 n=
T0
To
fo
(4)
2fo
18
1
Dn =
T0
we can have
Dn =
xp (t)ejn0 t dt
T0
1
1
X (n0 ) =
Xf (nf0 )
T0
T0
( )
t
A rect
A sinc (f )
19
n
T0
where
sin()
Review: sinc function property: (1) sinc(0)=1; (2) an even function; (3) equal zeroes at
= , 2,
sinc() =
t
103
A sinc(f )
* n +
1
A
X(nf0 ) =
sinc(nf0 ) = 1.5sinc
T0
T0
2
20
(5)
t
= 1.5 cos(1000t) rect
1 103
[( 1000) + ( + 1000)]
(
)
103
3
10 sinc
2
=
=
=
1
X()| = n0
T
,
( 3
)
( 3
)1.5 103
10
10
sinc
(n
1000)
+
sinc
(n
+
1000)
0
0
2 103
2
2
.
*
+
*
+/
0.75 sinc
(n 1) + sinc
(n + 1)
2
2
21
and Dn ej2nf0 t
with Dn = Dn . Let
Dn = |Dn |ejn
Dn = Dn = |Dn |ejn
then,
Dn e
j2nf0 t
j2nf0 t
+ Dn e
Hence, negative frequencies are essential, they are the product of the mathematical
22
Dn ej2nf0 t = c0 +
n=
where
cn [cos(n0 t + n )]
n=1
|D |
0
cn =
2|Dn |
Hence,
n=0
n = 1, 2,
c0 = |D0 |
n=0
c
n
=
Rn =
2|Dn |
n = 1, 2,
2
n = n = Dn n = 0, 1, 2
|Rn | is the RMS value of the cosine component corresponding to the frequency
= n0 .
23
Parsevals Theorem
Rayleighs Energy Theorem : given x(t) X(), then the energy of x(t) is
!
!
1
2
Ex =
x (t) dt =
|X()|2 d
2
Given the periodic function g(t) with the FS expansion {Dn }, then the average power is
g(t) = C0 +
Cn cos(n0 t + n )
Pg = C02 +
Pg =
n=1
g(t) = D0 +
Dn ej2nf0 t
n=(&=0)
n=
1% 2
C
2 n=1 n
|Dn |2
Pg = D02 + 2
n=1
|Dn |2
24
(6)