Download as pdf or txt
Download as pdf or txt
You are on page 1of 67

Discrete time -

signals & systems

chap 2
, Oppenheim -

Schafer -

Buck
Recall : LTI systems

n → →
y

* T is a linear operator

* T is time I shift -
invariant

* T is completely characterized by
h = TCS ) ,

a * h hats
y
- =
Properties of LT t
systems
Stability :

An LTI
system is stable if and
only if
+A

E Ih Ck 's ) Los .

K
summability
es
-
-
-

t absolute -
.

Causality :

An LTI
system is causal if and only if

hfk ] = 0
for all k co .
Stability .

(A) If h is absolutely -
summable :

§ lhlk ] / Eos .

K) is bounded
hCk3nCn
- n

{
.

ycn ]
,
=

N
F Ba ,
IMEDI E Ba for all .

f.
finite .
lyCn71 I § tick ] Mcn KT
)
-

( a -131
I )
sq
E tick ] Mcn KT
-

← last lbf

§ that
-1nCr ← E Bn .

⇐ Ba -

&lhI go .

E Ba -
S .
If h is absolutely summable
(B) not -
.

then T is not stable .

T I .

diverges
§ thick 31 → .

In ( n ] I
*
h C- n ] =L
Ncn ) =

-
.

Ihc NII -

y =
a * h .
For m -
-
o :

§
y lo ) ± Mlk] Mfk ]

& tick ] -

hk3
I hail

Sq 1h43 I
=


diverges .
* Finite -

Duration Impulse Response ( FIR ) system

. .
is an LTI system where h CK ] to for only

finitely many
values of k .

Observation :

Every FIR system is stable .

( ( hug is finite for all KS .

E ther )
{
cos
Incest
.

finitely
many K
* Infinite -
Duration Impulse Response ( HR ) system

. . .
is an LTI system that is not FIR .

* HR systems or not be stable


may may ,

too

depending on S th CK ] / .

k= -
as
Inverse System :

we that two LTI systems with


say

impulse responses h, and hz are

inverses of each other if

h, * bz = 8 .

Not LTI has inverse


*
every system an .
D TJ
T⇐I? ¥ .

at → .

h
,

n
a
.

he ¥42 = 8
Representation
Frequency Domain

eiw
#It -
.

Acosfwont →
.
Eigenfunctions of LTI
Systems

n →
t7→
-
an .

N is an eigenfunction of an LTI system T

if Tca ) is a scalar multiple of N .

* The effect of T on a is simply a

scalar multiplication .
Complex Exponentials are Eigenfunctions of LTI
Systems

suppose TC8Cn3 ) = hcn ]


'

' w
and ncn ] =
e ?

'

Teach ] ) = a * h

K)
+ as
jwcn
-

=
{ hck ] n Cn -
K ] → e
ke -
es
f f
Iwk
ejw ?
-

§ hck ]
-
e
Jwk ejwn
e-
§
Tca ) hck ] .

yay -
=

wk
]
=

enniwn .

[ { horse

t Iif it exists .

= man ] .

Hcejw )

Heem )
n →
I -

a .

Mn ] ejw ?
Frequency Response

The value Hcejw ) ,


as a function of w
,

is called the frequency response of the LTI

system .

'

( wt 2T ) n

* ejwn is indistinguishable from e


)

Hcejw ) is periodic in W
,
with period E 2T .

* Hcejw) is
generally specified for
-1T a WET
If tick ] is real for all k :

Hce Iw )
-

{ hCk3e+iw e
-

tjwsk
.

(tick ] e- jwkg*
{
=

=cE*¥ueiw #
,
=
H cejw )

fences'wD*
.
Response to sinusoids :

say ,
T =
LTI with freq .
response Hceiw )

n =
A cos ( won + of )

h Ck ] is real for all k .

Tca ) =
?
N =
Re ( Aejcwonto )
)
'

e- sole
Azeioeiwon
won
= +
Az

Iwm
Azeiotcejwon ) azeislotce )
-

Tcas =
+

=
Azejol .H .
ejwon
.es

wins?
.

+
ze.io
Tca > = Re ( Aejo . Hee's )
w
.
e
's won ]

= A .

I Hceiw ) l -

cos ( won -1ft Lince


's w
) )
-

Ta T
magnitude Phase shifted
multiplied by
by
I'i eiw .

T
t
magnitude phase response .

response .
via
Representing a
sequence

Fourier Transform


← d
*
Absolutely -
summable sequences
x

square summable
*
sequences qpec.kz/2aos
-

← .

K
*
special sequences .
via
Representing a
sequence

Fourier Transform

*
Absolutely -
summable sequences

square summable
*
sequences
-

*
special sequences .
Fourier Transform of Absolutely -
summable Sequences .

Suppose Ncn ] is absolutely -


summable ,

+ as

i. e
S lacm ] I < as
-

M= -
as

The Discrete -
Time Fourier Transform CDTFT ) of R

is
-
Iwm
Xceiw ) =
E acne
n

for -

IT < w e- IT .
Note :
to

xcejw ) =
E ncnse-jwneim.it of
sequence
.

n= -
es

f a

-
try
wn
= him Encode .

N→ as N
n= -

un
t
xcejw.nl )

= lim xcejw.nl )
N → as
re Cn ] e
-

J Wh
peso .

-
"

i§il/
-
F-
N

th
. .

mk.mg/Encn3e-jwn.J-
t N Partial -

sum

X ceiw ,
N ) .

n= - N

Xceiw
,
i ) XCEIW 2)
,
,
Xcejw ,
3) .
. . .
.

sums )
( sequence of partial
.
The error between Xceiw ) and XCEJW ,
N )

Ixceiw ) -

xceiw.ms/=/Encn3e-iwn / n:

Inl > Ntl

S E lacrigeiwnl
n :

1h13 Ntl

) cinders of )
§
w
= Inch ] .

Int > Ntl

o as
N - os

.
In summary . if N is absolutely -
summable

① sup / xceiwj -
Xcejw .
N ) / → o

f as N → es .

over all WE C- T.lt ]


Supremum
.

we
say that
Xceiw .
N ) converges uniformly
-
-

N
to XCEJW ) as
→ as
.

② This implies that Xceiw ) is a continuous

function of W .
The inverse DTFT :

If N is absolutely -
summable ,
then

+IT

ncn ]
¥ J xceiwsejwndw
=
.

-
T
f
-

Xcejw ) exists for


Analysis all

Real
Analysis we C- As ]
.

it

Complex
✓ Tao z I
and is a Coutinho as

Terrence Analysis function of W

\
I
dmaostns since he is absolutely -

K summable .
proof :
Slim µ → os

w :

¥④ncm e- Jwmeiwndw
"

xceiw ) eiwndw
#
]
If =

=÷ .is#acmseiwcn-mdw ÷÷÷÷:]
+IT

J tzoejwcn Mdw
-

Em acm]
=

wa -
UT

1 if
Tf
hem

g
=

N Cn]
if
=
o NFM .
Theorem : ( DTFT for absolute -
summable sequences )

If a is absolutely -

summable then

xcejw ) = E ncnyejwn
n

exists for all w and is continuous in w .

The inverse DTET is

IT
xceiwjejwndw
UCM =

Igf .

-
I
via
Representing a
sequence

Fourier Transform

*
Absolutely -
summable sequences

square summable
*
sequences
-

*
special sequences .
Absolutely -
summable :

§ talk] ) cos .

:/
. .. . .. . .
t
set of
12
§ In a es
possible
.

all

sequences
as
.

quam
summable sequences .
Claim : Every absolutely seem masse seq .
is also

square summable .

Proof : Let M be absolutely summable .

{ lack 31 S s
= os .

→ flack ?k
'T
N
Consider Y =

E 5-1
. .

It is enough to show summability


square
-

for y .
"

{ (19¥31)
"

§ sail
' =

:#
'
=

§ I
'
I
-

E
& I I ⇐ ⇐

1)
.

lack ] I as as
k¥31 <

's in
M¥2
claim Not summable is
every
:
square seq .

absolutely summable
-
.

Yn

{
.
rise

① Nn ] ⇐

0 ,
M E O .

ht off Iese
£
.

diverges
§ In → .

,
, finite
value .
② If N is such that for all n #O :

)
Un) =
sinners
Tn

most absolutely smarm


- .

VI. square
-

summable .
DTFT
of square summable
sequences
-

* n is square
-

summable if

§ HEMI s os .

summability
Absolute summability square
.

*
-
-

* converse not true :

yn if n > I

Example : MCM =

{ o
if n SO
DTFT
of summable sequences
square
-

* If N is not absolutely -
summable

tht
Wn

µli→m• { rate may not exist for all w .

n
-

-
- N

we can
not define DTFT as

xcejw N )
xcejw ) = lim , .

N → as
Theorem : CDTFT of square
-

summable sequences )
The Riesz - Fischer theorem )

If U is square summable there exists


① ,
a

function fncw ) such that

IT

M¥8 S l faced -
xceiw.ms/2dw= 0 .

IT

[ convergence
in the mean -

square sense ]
7

IT

fab ) ejwndw
② Moreover
,
acn ] =

Iz f .

IT
-
We call this
function facws the DTFT
of se

and denote it as

xcejw) =
facw ) .

need not exist for all w

*
mlim→ ftp.anfiaalmy
,

* we can the limit in the mean


-

square

sense .

d
* xcejws .
Example :

suppose we want A such that

1 WE two

{
-

we E

xceiw )
.

O
,
else .

Such an a is not
absolutely
-

summable ,

since X is not continuous .


n÷÷÷¥i÷÷n .

N is
NIT absolutely summable
-

.
to obtain through inverse DTFT
Try a :

IT

ejwndw-fzfxeeiwjeiwndwtfn-efyn.to
we
Ncn ) =

If f
-1T
we
-

IF [ eY ] !!
We

f-
do if neo
toss
=
-

iiie-iw.cn
.
.

- we

-
÷ .

sin ifn.to
If
{
= -

Yf ifn -

- o
* Inch ] I decays like Yn

This can be used to show :

① n is not absolutely -
summable .

② a is square -
summable .
via
Representing a
sequence

Fourier Transform

*
Absolutely -
summable sequences

square summable
*
sequences
-

*
special sequences .
Special sequences

There that would like to use


* are
sequences we

that not even


square summable
-

are .

N =
complex exponential

or

unit step .

* Their DTFT
requires the theory of
generalized functions
'
'

.
① Complex Exponential
+ as

's won 2t8( w wot2TD


Ncn ] = e → xcejw ) =
{ -

re -
as

ii.tw#-..
'

② unit step
+ as

xeeiws

few
acre vents → IT ECwt2tr )

toilet
- -

-
-

If at
Fourier
Symmetry Properties of Transform
-
I

By default we

mean

DT FT

( in this counsel .
Claim : If the DTFT
of Ncis is Xcejw ) then

the DTFT
of NERD is xttceisw)

to *

Heiny ]
-

[
Claim : If Ncn ] is real for all n then

Jw
xce
-

) =
X
*
cejw ) for all w .
Fourier Transform Theorems
Linearity .

If U ←s X and es Y
y

then

xat → 2X Y
By t
p
Time
shifting

Nn -

no ] → e- jwnoxcejw )
Frequency shifting

eiwon xcejcw )
Wo )
-

seen ] →
Time -

Reversal

-
Jw
a C- n] a- ice ,
Differentiation in Frequency

cejws
na Cn ] a- j
dX-dw
-
's
Parser al Theorem :


nI§acn3y*cn3 =

ftp.#xceiwsx*ceiwsdw
T inter product
X. Y
product between
.

inner
a y .

between ,


§ lack =

If ! ! lxceiwsldw

'
* I Xcejwsl is called the energy density

spectrum of a .
Convolution Theorem

If y = a * h

then

YCEIW ) = Xcejw ) -

Hcejw ) for all w .


modulation Theorem

If ycn ] = a [ n] -

w Cn ] for all n

then
IT

Jxcejojwcejcw
o'

ycesnwj )
-

do
If
=

@= -

IT
The Z -

Transform

chap 3 Oppenheim Schafer Buck


of ,
&
Drawback DTFT
of

* Xceiw ) is well -

defined for all w


only if

is
R
absolutely summable .

Acn ] ?

Difficult to capture =3 Uh ] .

"
'

* workaround : dampen a before DTFT


as
+
Jun
xcreiw ) =
E acnzrm .
e-
as
MI
-

=
§ seen ] ( re-sign .
* If V is such that

§ tacos ni s
-

-
r as

then Xcrejw ) is well -


defined for all w .

* xcrejw ) can be used instead of Xceiw )


not
( which may
exist )

* rejw is denoted as 3 ,
i -
e .

8=131 .

,
The Z -

Transform
The Z -

Transform

Definition : To
every sequence n associate a

function X as follows

Msas
§ncn35 ? §1acn31l3
-

if

×C3 > =

{ undefined ,
eye .
Region of convergence ( Roc )

* Roc is the set of values


of z for which

Xcz ) is well -
defined .

* The function Xcz ) is holomorphic / analytic


in its Roc .


xcz > is continuous and infinitely
differentiable in the Roc .
Note :

* The Z transform of x is a function X


-

X is determined by

is
( is its Roc ( where the function
defined )

AND

dig the value


of xcz ) when ze ROC -

You might also like