DSP
DSP
DSP
FIFTH SEMESTER
ELECTRICAL AND ELECTRONICS ENGINEERING
(REGULATION 2008)
UNIVERSITY QUESTION BAN !ITH ANS!ERS
UNIT"1
PART"A #2 MARS$
Q%1 !&'( ')* *+*, ',- .-- /01,'23(N.+ 4 D*5 2010)
Ans:
A continuous (or) discrete time signal x(t) is said to be an even
(symmetric) signal if it is satisfies the condition
) (t x
=
) ( t x
for all t
) (n x
=
) ( n x
for all n
A continuous (or) discrete time signal x(t) said to be an odd
signal (Asymmetric) signal it is satisfies the condition
) (t x
= -
) ( t x
for all t
) (n x
= -
n x (
for all n
E6'782*9
Q%2 D*:0,* A20'/0,1 *::*5(%(N.+ 4 D*5"2010)
Ans: Let us consider a band limited signal x(t) having no
frequency component for ||!
m
if e sample the signal x(t)
ith a sampling frequency f f
m
"#he periodic continuation of
x() results in spectral overlap" $n this case the spectrum x()
cannot be recovered using a lo pass filter" #his effect as
aliasing effect"
#o avoid aliasing:
%"x(t) is strictly band limited
&"
2
s m
f f >
Q%3 D*:0,* S'7820,1 (&*.)*7 (A8)024M'; 2011)
Ans: A band limited continuous time signal ith a higher
frequency
m
f
'( can be uniquely recovered from its samples
provided the sampling rate
2
s m
f f >
samples per seconds"
Q%4 !&'( 0/ 20,*') (07* 0,+')0',( /;/(*73 (A8)02 4 M';
2011)
A,/9 An L#$ system is one hich possesses to of the basic
properties using linearity and time invariance
L0,*')0(;: An L#$ system obeys super position principles hich
states that the output of the system to be a eighted sum of its
equal to the corresponding eighted sum of outputs to each of
the individuals inputs
[ ] [ ] [ ]
1 2 1 2
( ) ( ) ( ) ( ) T ax n bx n Ta x n Tb x n + +
T07* 0,+')0',( /;/(*79 if the input )output relation of a
system does not vary ith time" #he system said to be time
invariant"
Q%< D*:0,* N;1=0/( )'(*%(M'; 4 >=,* 2012)
A,/9 *yquist rate of sampling is the theoretical minimum
sampling rate at hich a signal can be sampled and still be
reconstructed from its samples ithout any distortion
*yquist rate
m N
f f 2
'(
#o avoid aliasing
m N
f f 2
Q%? S('(* ',- 8).+* (&* (07* )*+*)/'2 8).8*)(; .: :=))0*)
()',/:.)7 (M'; 4 >=,* 2012)
A,/9 $f
{ } ) ( ) ( w x n x F
{ } ) ( ) ( w x n x F
+roof:
{ }
n
n x n x F ) ( ) (
iwn
e
n
n x ) (
jwn
e
n
n x ) (
n w j
e
) (
{ } ) ( ) ( w x n x F
Q%@ G0+*, ' 5.,(0,=.=/ (07* /01,'2 A&'(
0/ (&* ,;B=0/( )'(* ',- :=,-'7*,('2 :)*B=*,5; .: (&*
/01,'23(M'; 4 >=,* 2013)
A,/9
500
m
C
500 f 2
m m
C
250
2
500
f
m
'(
*yquist rate
250 x 2 f 2 f
m N
500 '(
Q%8 D*(*)70,* A&*(&*) 6#,$D=#,$ 0/ 8.A*) /01,'2 (.)) ',
*,*)1; /01,'2 #M'; 4 >=,* 2013$
A,/9 ,nergy =
n
2
) n ( x
0 n
1
, =
, = - so given signal is poer signal
.ondition for energy signal is (/ 0 , 0 )-finite value of
energy
P')("B (1? M')E/)
Q%1 ') (0) !&'( -. ;.= 7*', F; N;1=0/( )'(*3 G0+* 0/
/01,0:05',5* (?) (N.+ 4 D*5 2010)
A,/9 *yquist rate of sampling is the theoretical minimum
sampling rate at hich a signal can be sampled and still be
reconstructed from its samples ithout any distortion
- $t is 1the theoretical minimum2 because hen the *yguist
rate of sampling is used
- 3nly ideal L+4 can be used to extract x() from x
s
()
- $t is alays equal to &f
m
here f
m
is minimum frequency
component present in the signal"
- A signal sampled at greater than nyquist rate is said to be
over sampled and a signal sampled at less than is nyquist
rate is said to under sampled
*yquist interval:-
*yquist interval is the time interval beteen anty to ad5acent
samples hen sampling rate is nyquist rate
*yquist rate N
f
=&f
m
'(
*yquist interval =
N
f
1
=
m
f 2
1
sec"
00)E682'0, (&* 52'//0:05'(0., .: -0/5)*(* (07* /01,'2%(10)
(N.+ 4 D*5 2010)
A,/9 .lassification of discrete time signal (%/)
a" +eriodic 6 Aperiodic signal"
b" ,nergy 6 +oer signal"
c" .ausal 6 *on-causal signal"
d" ,ven 6 odd signal"
a" +eriodic 6 Aperiodic signal
A discrete time signal x(n)is said to be +eriodic" $f
it is satisfies the condition
7(n)8x(n8*) for all integers n"
- #he smallest value of * hich satisfies the above
condition is 9non as fundamental period"
- $f the above condition is not satisfied even for n: then the
discrete time signal is Aperiodic
,nergy 6 +oer signal:
A periodic time signal x(n)",nergy signal , =
n
n x
2
) (
+oer signal:
+ =
Lt
N
1 2
1
+ N
N
-N
2
) (
n
n x
A signal is said to be an energy signal if only its total
enegy , is finite (/ 0 , 0 ) 6 average poer p=/
*on periodic signals are examples of poer signal
A discrete time signal is x(n)is said to be odd
(Asymmetric ) if it is satisfies the folloing condition
) ( n x
=-
) (n x
(or)
) (n x
= -
) ( n x
for all n
Q%2 (0) E682'0, 0, -*('02 (&* B=',(0G'(0., .: -010('2 /01,'2
(8) (N.+ 4 D**5 2010)
A,/9
Q=',(0G'(0.,9 the process of converting discrete time
continuous amplitude signal x(n) into a discrete time amplitude
signal x q (n) is 9non as quanti(ation hich is done by
rounding off each sample in x(n) to the nearest quanti(ation
level" #hen each sample in x q (n) is represented by a finite
number of digits using a coder
- $f a signal is represented by b8t % bit used ith amplitude
range ; than quanti(ation that can be represented is &
1 + b
#he difference beteen the quanti(ation step in terms of
the range of the signal is
q=
on quantizati of number
signal of range
=
1
2
+ b
R
- $t is necessary to scale the signal: if the range of the signal
exceeds 1 t
- <e can represent the sample values by a sequences
x(n)==/:/">&/:/"?@:/"?@:/"@A@:-/"/B:-/">&@:/"?@:-
/"?@:-/"@A@:/C
- Let us use a b8% bit AD. to represent the above sequence
ith b8% binary digit
1
2
+ b
quanti(ation levels can be
obtained and the input can be resolved to one part in
1
2
+ b
"
$f the input signal has a range of &v: then the quanti(ation
step si(e is equal to signal
q=
b
b
+
2
2
2
1
- #he process of converting x(n) to a finite number of
introduces on error 9non as quanti(ation noise" $t is
defined as the difference beteen quantified value and
the actual value sample value
) ( ) ( n x n e
q
)
) (n x
(00) Eampling techniques:
$mpulse sampling
*atural sampling
4lat top sampling
$mpulse sampling
- #he above analog signal x(t) is multiplied ith a nit
impulse sequence ith uniform interval
- Eample an analog ave form ith a delta function
) (t
T
"
- Assume x(t) is band limited it is outside the interval (-
s
f
0
f
0
s
f
)
- Eample at x(t) at times t=n#s by periodic train of delta
function
) (t
T
) (t
T
n
s
nT t ) (
4ourier transform of
) (t
T
) ( f
T
s
T
1
n
s
nT t ) (
<here s
f
=
Ts
1
- Eampling can be achieved mathematically ass the product
of x(t) ith
) (t
T
) (t x
s
) (t x ) (t
T
"""""""" (%)
=
n
s
nT t t x ) ( ) (
=
n
s s
nT t nT x ) ( ) (
Epectrum of the sampled signal
) ( f X
s ta9e fourier
transform on both side of eqn (%)
) ( f X
s = ) ( f X
) (t
T
= ) ( f X
1
1
]
1
n
n
s
s
nT t
T
) (
1
n
n
s
s
nf f x
T
) (
1
*atural sampling
- *atural sampling also called sampling
- Fsing of a sequence pulses in the practical ay of
accomplishing sampling of a band limited signal
- #his is achieved by multiplying" #he signal ith a pulse
train
) (t p
T
- ,ach pulse of
) (t p
T
is short duration
and occurs at a
sampling period of # sec
- #he output of the sampler is same as the input during that
short duration
"
- 'ence it termed as natural sampling
- #he input analog signal multiplied (or) sampled ith
periodic step pulse input
) ( f X
s =
) (t x ) (t P
T
<here
) (t P
T
=
n
n
nT t p ) (
As
) (t P
T
is periodic pulse train" Let us rite its series
expansion
) (t P
T
=
n
n
nT t p ) (
=
n
n
t
s
f j
n
e c
2
""""""""(%)
n
c = ) (
1
2 /
2 /
t P
T
T
T
T
dt e
t
s
nf j 2
Eince
is the idth of
) (t P
single pulse in
) (t P
T
is very
much lesser than # and
n
c
=
) (
1
2 /
2 /
t P
T
T
T
T
t
s
nf j
e
2
=
) (
1
t P
T
n
c
=
) (
s s
nf P f
""""""""(&)
<here
) (
s
nf P
=
[ ]
s
nf t
t P F
) (
*o
) (t P
T
can be rite it as
) (t P
T
=
n
n
s s
nf f ) (
t
s
nf i
e
2
) ( f x
s
[ ] ) (t x F
s
= 1
]
1
n
n
t
s
nf j
s s
e nf P f F
2
) (
=
s
f ) ( ) ( ) (
s
n
n
s
nf f f X nf
4lat top sampling:
- #his is simplest 6 most popular sampling method that use
the sample and hold that circuit ith flat top sampling
Q%3 C&*5E A&*(&*) (&* :.22.A0,1 0/ 20,*')H (07*
0,+')0',(H 5'=/'2 I /('F2* (8) (A8)02 4 M'; 2011)
nx(n) + x(n) = y(n)
A,/9
nx(n) x(n) ! "(n)
[ ]
[ ] ) 1 ( ) (
) 1 ( ) (
+ +
+ +
n x n n x (n) x T (n) "
n x n n x (n) x T ! (n) "
# # # #
$ $ $ $
Eum of eighted outputs
) 1 ( ) 1 ( ) ( ) ( ) (
2
+ + + + + n bnx n x an n ax n " b n a"
$ $ # $
= [ ] ) 1 ( .......... ) 1 ( ) 1 ( ) ( ) (
2 1
+ + + + + n bx n ax n n x b n ax
# $
[ ] ) 2 .( .......... ) 1 ( ) 1 ( ) ( ) (
2 1
+ + + + + n bx n ax n n x b n x a
# $
,qn (%) 6 (&) are equal so
Eystem is linear system
nx(n) x(n) ! "(n)
[ ] $) nx(n x(n) x(n) T ! "(n) + +
#he output due to input delayed by 9
[ ] %) & x(n T %) "(n'
%) & x(n x(n)
"(n)(
=
( ) % & $ n nx %) & x(n + +
""""""
($)
#he output delayed by 9 inputs
) 1 ( ) ( ) ( % n x % n % n x "(n)( %) & "(n
% & n n
+ +
"""""""
(#)
eqn (%) 6 (&) not equal
%) & "(n %) "(n'
system is time invariant
nx(n) x(n) ! "(n)
future nx()) x(#) "(#) # n
future x($) x(+) "(+) + n
x(&$) x(&#) "(&#) &# ! n
+
+
+
Eo system is non- causal
(00) C&*5E A&*(&*) (&* :.22.A0,1 ')* *,*)1; (.)) 8.A*)
/01,'2(8) (A8)02 4 M'; 2011)
) (
2
n u
$
x(n) ($)
n
,
_
jwn
,e x(n) (#)
) (
2
n u
$
x(n) ($)
n
,
_
N n
N n
n
n x Lt
2
) (
=
1
1
]
1
,
_
N n
N n
n
n
n u Lt
2
) (
2
1
'ere
'
<
0 0
0 1
) (
n for
n for
t u
0 1 ) ( n for t u
that means limit change as (/ to *) from (-*
to*)
, =
1
1
]
1
,
_
N n
n
n
N
Lt
0
2
2
1
Apply limit N eqn becomes
,=
1
1
]
1
,
_
n
n
n
0
2
2
1
n
n
,
_
0
4
1
a =
4
1
, =
3
4
5oules
,
1
1
0
jwn
,e x(n) (#)
,nergy =
N n
N n
N
n x Lt
2
) (
=
2
N n
N n
jwn
N
,e Lt
=
2
2
N n
N n
jwn
N
e Lt ,
= [ ]
2
2
1
N n
N n
N
Lt ,
=
[ ] 1 2
2
+
N Lt ,
N
= ) (
2
,
=
,nergy is infinite : then the given signal is poer signal
Q%4 (0) C&*5E (&* A&*(&*) * :.22.A0,1 ')* 8*)0.-05 (?)
(A8)02 4 M'; 2011)
(%)
) 3 cos( ) ( n n x
(&)
) 3 sin( ) ( n n x
A,/9
(%)
) 3 cos( ) ( n n x
3
3 2 f
3
2
T
) (or
2
T =
3
2
=
3
2
# is the ratio of to integers then
) (n x
periodic
(&)
) 3 sin( ) ( n n x
3
T
2
=
3
2
#
<here
) (t
t
n
n
t
nT t ) (
#a9e the fouier series on both sides
)] ( [ t X F
s
] ) ( ) ( [
n
n
s
nT t t x F
) (
s
X =
1
]
1
n
n
s
n X ) ( ) (
2
1
0
) (
s
X
=
n
n
X
T
) (
1
) (
s
n
) (
s
X
=
[ ] .......) 2 )(cos ( 2 ) cos( ) ( 2 ) ( + + + t t x t t x t x F
s s
) (
s
X
=
n
n
s
n X
T
) (
1
#o receive original signal
) ( X
(i) 4ilter ith a gate function
) (
2
m
w
-
of ith &
m
(ii) Ecale it by #
) ( ) ( ) (
2
m
s
- TX X
Aliasing: (
m s
2 <
)
- Aliasing is a phenomenon here the high frequency
components of the sampled signal interfere ith each
other because of inadequate sampling
m s
2 <
#o avoid alias
(i)
) (t x
strictly band limited
(ii)
s
f
is !than &
m
f
3ver sampling :
$n practice signal are oversampled here
s
f
significantly higher
than *yquist rate to avoid aliasing
Hunti(ation
- #his is conversion of discrete time continuous time valued
signal into a discrete time discrete valued
- #he difference beteen the unquanti(ed sample
) (n x
and
qunati(ed output
) (n x
q is called quani(ation error"
Q%< (0) D0/5=// A&*(&*) (&* :.22.A0,1 ')* *,*)1; (.))
8.A*) /01,'2
(1) ) (
2
3
) ( x n u n
n
,
_
(2)
jwn
Ae n ) ( x
A,/9
(%) ) (
2
3
) ( x n u n
n
,
_
N n
N n
n
n x Lt
2
) (
, =
2
) (
2
3
1
1
]
1
,
_
N n
N n
n
n
n u Lt
'
<
0 0
0 1
) (
n for
n for
t u
1 ) ( n u
for n / that means limit changed as / to * from )* to
*
a=
4
3
,=
4
3
1
1
= I 5oules
,
1
1
0
a
a
n
n
n
1
1
0
a
a
n
n
n
1
1
0