BMS INSTITUTE OF TECHNOLOGY AND MANAGEMENT

YELAHANKA – BANGALORE - 64

DEPARTMENT OF ELECTRONICS & TELECOMMUNICATION

ENGINEERING

Course Name: Digital Signal Processing

Course Code: 18EC52

Semester : 5th semester

Prepared by : Prof. Thejaswini S
BMS INSTITUTE OF TECHNOLOGY AND MANAGEMENT
YELAHANKA – BANGALORE - 64

DEPARTMENT OF ELECTRONICS & TELECOMMUNICATION

ENGINEERING

CONTENTS
SL. No. Module Page No.
1 Module 1 1-69
2 Module 2 70-124

3 Module 3 125-156
4 Module 4 157-209
5 Module 5 210-243
 DIS CRETE FoURIE R 7AANSFOR ms DFT)
Fe
uen cy denat gavuplinq
2 Ju tontucio 0 uAH
wu ignall f T OS liniok t va.joynmali0n, 1
he la tiom Shup 0 t h othe 4 a s o ru 6H

Ln+reduut Fio vn
ie ed to deauihbe
uaunCHen
X ASinal tial Voliablo of phyiln
observed phy
an

PYDles diatlpHU)
MaHho naan (al a
obstract obeu Condi
hn
s ECOCeuCHYOran,
ECG(eu(tcaldroq entpha|oqkar )
6an okservaHOn Ghug 1

ranal
f
t
m a G e

pueh i qnal,
itfinie
itfitik
ne 0
af
ConkitK Can bR u i ein
uhte n
oles CLipHen 4 (an
Pogs
Pinulodg
0 dit

Sinln/ (t) + 9; (+)

A;(+)
St)

Plhat an
mplituo

a alo be olefine
Sye A
Afy8Rm
that perhomu
u an
tlhat perhoI
fical deu'
C
as p h y aCA Siq nal
on
Ceno
0Y Cn
ge nakd by 01
roiON
9pe olA
Cpech 3 na
Cokds
aY Hlo tho vo Cal CoYds cal tra
Vocal
tran
ConbiH d VoCal Covd
Vocal
yRM VoCal (avIA) 1
IM
 4iler wd to Joduo noit inprfelng
orpoY MU om onuaHonlt) On H

Phqnal tokie h ha the efeu of uduung (ileliny

not (ntreki na o the ouiltd (nfotawion
baling nal

S1qnalProMAing lal wiHh monipula hon 0r

iqnal Jo that i Jaul

In
in
wdiiatm 0
or InRls'e taba 5orm.
more deAikable

or more atulas
pa tul
On
CCtim echantine pludHUin
pAudenmim
T fo
(padamenH) 0a Siqal'attording
quimenu

phou et
auupli+u ou
P PuCh a polUs
protns
AuCh
undtg0gs
hui h
Th 9:qua (nput Signa
Kneuwn n a p
pnCMing Ming
thh u
f
perorm
Perhorms
9 huh
whuyh
Thu en HHy
en HH 0 p Pignal
P
u
tiqua
pwtMed

Th
A p l i i a ti en,
igna
Siq pnsing M OCU
ilRAtn,
Analog
+Yayul4o n FET, 0p-amp-
All thM meA
Can
Can
tiqnal pwaMinq
in
or
x A
platornm u
w9dt we can
t we Can duAg
diugy
Javo l09 t yfom, diuiqy
diuiqy k¢h niquu, R
Ahniq
fyth my uling prornano n
an analog an alup kabl
urormano
2
9m unHl A
Adosr
 Tu w On onothe r appYoath Called

Digital iqnal pwtsin telg he tnJat

tHiial m a t
AOLen td n
Jap

Anal
Ani Alia.in | Sauupo4 ADC
ilp hotd C Digial
analog On en
Oigiial
Piqnal
PyooMo

Ana leg Dipiia

RetenstrucHun D AC
iqnal

Ana log
b a l t en d

TW anHalrauiny lHr, an analoe Lps

band 1i'nai' p analo
anau g
th
wMd 4o
rqnal o Hhe Jugulcd eq Cop
pevenie Conup onen H dYond
Jang 4
as ali a s u
appeak inq
Hhi Jange rons u ilp a
In Hu tanup led spAHVUm

ilp iqnal
wR Hi wo
ana loq pitked uP
Tu anao ilp tig nal A
nivaA tin 4 nAe
a4 cliAUo se
value
lOM iq nA binaly
A PC Onap
the lp
lo app eos ing a
0 each Janup e
se+ 0 i n i e valu.
Chosen to a both wH tinf
diau1od 3
t a digilal lignal QuLplruAL
 Tht odtytal ignal to 6torLel PoasALd
b tnuploqino DP eohni
Olp u anether eqn 8 i n aly no'
oieh ConvK Rd fo analocqom tMinay
DAC
th unoiki
un dod
diui dod
Ke mo es
LpE at olp
The
noun 4 iLes 0ut h d
dAilsd

ueyh Hoq
Ana le Signo

Ow Anal19e Pignal
. Aun tLly es pigita

cliqi tal Pgumabt tys en allows o ibili

n Contiqu inq the otiqi+al ignal potmina
epas iuply by changinq tu pq

Rate nhigukagon 0 C anulo

Luonly tnupM a idesiqn 8
Hu Hluw follouXaol by RAHn fveliniCeuHo
40 se0 that t Ope rakA propetly
beRr Conth
diital fyetnn pwbiolus nnuch
a a ceulay JequtennnH
d Asd on
digial 91q nals ak eaily to

wionetuu uolo ( 4apt or dul

los s1qno
)Dip netheol allo allous hor the uuplaumuntaHo
mo e Sop hiatt CaA Hecl tp nal pw AM une alnl-
4
 Tht odtytal ignal to 6torLel PoasALd
b tnuploqino DP eohni
Olp u anether eqn 8 i n aly no'
oieh ConvK Rd fo analocqom tMinay
DAC
th unoiki
un dod
diui dod
Ke mo es
LpE at olp
The
noun 4 iLes 0ut h d
dAilsd

ueyh Hoq
Ana le Signo

Ow Anal19e Pignal
. Aun tLly es pigita

cliqi tal Pgumabt tys en allows o ibili

n Contiqu inq the otiqi+al ignal potmina
epas iuply by changinq tu pq

Rate nhigukagon 0 C anulo

Luonly tnupM a idesiqn 8
Hu Hluw follouXaol by RAHn fveliniCeuHo
40 se0 that t Ope rakA propetly
beRr Conth
diital fyetnn pwbiolus nnuch
a a ceulay JequtennnH
d Asd on
digial 91q nals ak eaily to

wionetuu uolo ( 4apt or dul

los s1qno
)Dip netheol allo allous hor the uuplaumuntaHo
mo e Sop hiatt CaA Hecl tp nal pw AM une alnl-
4
 The DisCYe teourier T'onszorns
I43 prop erthes 4* Applicati OnA
t i v gnab
u
analyfis 6 - dlifcete
Freq uenty
mos+ COnVenien tly pe rhornd bn q
usually genhalpupoR
woy be
a
ua PotMor [ié
dtoit olesgnud
oligital
or Spetially
CQpuRr
olr'otal

n aa odis uo
olis s
Hlw Analysi
Analysis s on
o

perhorm
vegunu Ha tinu
T0 TO Jwe
e
C
COon
neer 4

(nu e g u s va/e n t
u-odonARIn
+ivu Signal eguivalent
an HupresentaH On u
Sequn (e
to
u p r e s e nfaH
on
cAewu a Hh
Suth (w)
on. Transhorm
JupvesenaAi
Fourier
thu
qrven by
Sequunu (Cns
C o uutahsnatle
enatl,
Couputat
4 A L O

C a
not

i+ A
HowAY,

uq 4
uConVè nith + i6n
HOn
vepresen
i out Speltrun
t+o
Hu i
YpeLhrun
C8siolh
We fa
S a
rnn
ep lp
ngul.A
ts
dguLaLn vestnas
H
* er by
Jby On
n(n
Jequunte discre Fourier
T 7 rranshot »
ranshotn
a
Such
xt w COus Ltatanal
c o u p u f a t a n a l

leonoA to Hu
po
p ow
w h
p pu
u!

uwhich u a Heq Analyti a

LDET,
tool bor perjorming

au Prokais

6
 uny
0 F r e quenuy
D BUUOUln Sauupling The Distvete D0uher
7anshorny
Behove wt intodute DET, we Conkioler tu faupin
Hnu segh
an aperioclic
aper ditu -

ET thu anuplud
t u
thu lation hip beu'
Julas
Atoablish

Thus w
FT2 DFT
Reconst Yu(tiOn
Reconsivu(H iOn
Sauping
DevuoaLn
EVqueney
QisCvet 7HA inal
apenoodic
dit
oisuue
te
aperiodi c
nAiolr Cn
a n
us C with fouritr Transjorm
het fourier
T (n)
with

-H
t iu
n signa
Cn e
X(w
fignay
diftrek -H'me uh rom
nuDusun
C0nHnuvus
Xn)>
A COnti
Aiq + O to &r
w
w
oliftrete 2 1I SpeCuh,
paCinuk,
oliftrete
n iu huns
u
tha Xln) inuBuA
uam un
u l h oal i g '+ a l
Th
This
C o n t i n u o .

oigital
p
p oh
uM Or
uo r.

A on a
Can
Can
be
bt
X(
XCw eualuatd

g nna
a u
¥9
Cannt
be dis treR
tne
Bhly
Pinte +al pio tuaing
plouMing
problemT digi
odigi+al
e valuakd
thuA
h u problem unipommly (wri
unipormly (Pere
oertame

to SaupluAd
auupleod
t oo
X (w &wiadiarns ,
K penbol
pernboA
thu
a.
spettruh
e a spau wi th
wi th
periodi t
periodit
ktw u 0 to 7
SinSinte
rom
au
taKen
taken
wil JA
silI
fanuples JuttwMive J
np LA
p lu
Thu futtMive

the pating e foy conVenien1¢

2
t a k e N e9uidn
7
oswEåT with paeiuc fgupl rin HAR InkiVal
 ubgtituting w: k
N
in eq 0, we q
x() n: -03
N

K: D, , 3, N I

Ino 0 8auspu
* x (w) Ga is Calculoseo only at oliftve te Uluy
ta K2n 0uea
Hov
OV het N: 8. fanp uA L

will be ca lculand ad
perioc & 7 2 xCw)

w:0,, , , T 4
Odlohoseol os
7
0, 1, 2,
.

K: H
tuse valuus in eg.
fub! HFuNnq thown in
by OytainUd u
Spectrum
fanupud
ig belotwr

3 T7T S 3
4 4

K:0 2 3 A
laupling
d eun ain
tig veq
tlu fourier Transhormy

8
 uny
0 F r e quenuy
D BUUOUln Sauupling The Distvete D0uher
7anshorny
Behove wt intodute DET, we Conkioler tu faupin
Hnu segh
an aperioclic
aper ditu -

ET thu anuplud
t u
thu lation hip beu'
Julas
Atoablish

Thus w
FT2 DFT
Reconst Yu(tiOn
Reconsivu(H iOn
Sauping
DevuoaLn
EVqueney
QisCvet 7HA inal
apenoodic
dit
oisuue
te
aperiodi c
nAiolr Cn
a n
us C with fouritr Transjorm
het fourier
T (n)
with

-H
t iu
n signa
Cn e
X(w
fignay
diftrek -H'me uh rom
nuDusun
C0nHnuvus
Xn)>
A COnti
Aiq + O to &r
w
w
oliftrete 2 1I SpeCuh,
paCinuk,
oliftrete
n iu huns
u
tha Xln) inuBuA
uam un
u l h oal i g '+ a l
Th
This
C o n t i n u o .

oigital
p
p oh
uM Or
uo r.

A on a
Can
Can
be
bt
X(
XCw eualuatd

g nna
a u
¥9
Cannt
be dis treR
tne
Bhly
Pinte +al pio tuaing
plouMing
problemT digi
odigi+al
e valuakd
thuA
h u problem unipommly (wri
unipormly (Pere
oertame

to SaupluAd
auupleod
t oo
X (w &wiadiarns ,
K penbol
pernboA
thu
a.
spettruh
e a spau wi th
wi th
periodi t
periodit
ktw u 0 to 7
SinSinte
rom
au
taKen
taken
wil JA
silI
fanuples JuttwMive J
np LA
p lu
Thu futtMive

the pating e foy conVenien1¢

2
t a k e N e9uidn
7
oswEåT with paeiuc fgupl rin HAR InkiVal
 3) 6
X(n-An el kn
(7) n0 L: -
N

K 0-N- 1
V-1

p (n) e2Tkn

N-1
whe k: 0, ,

`xCn-Jn) >É4.I')
p n):
n +2n) 4 ( n t n ) + H(n)
+

Cn-N) +X (n-&n)
hlpeHtion
that tn) A a periooliC
This means Xp
N Sampu
oith the peviod &
tn) non-peHd dic
feuu arbiHary Shousv
Jhousen
Caaiodlur A
Let ws ClyntOarns Z
L
FanyPus
famplls
(n).
ignal 3)

ig(sa)An ovignal fignal

lengfh L

CAe- 1
AliaA in 9.tht ignal plhy Aplhy
N>L NO t i Signal
6
o aln)
ln)
NOW kof us
preparu
titr
vepe4itr en
orn
perigolicc Yepe
petiooli
OptainL
wWhich iA
ol by
by has
tha4L
Hhe
thu pericod N be gAa41 u h o w n tin
A the wh
Acd N:6 He sample
hehe Ct n - 4 45,
10
becawe tu igna upeag of h: 6, 19. +
hunaaN>L, Hue u no l i ovs Tnq
A ptn)

6 8lo I1
ig26 periodic upetitien o (n
>L
0ith peh od W
CcUACi), N<LAiasing

NOW Let ws
CMAiour W<L, bor 2 N:3
tu wi u thousn in g 3c)
Xp(n) over lap ping
duu to AloUtng

periodic repeiHon ln) wi

periocdi( N<L

drom tu w e observe that fina

NEL,
here
tu tampus ou Dverlappling at n:3, -..eC
Th 'uAlraling nu it unt posible to db+OUL
(n) drom p n)

TO auoid Aliasing H
tn - dgnnau h
thu ro bh Jomples m u Jpectu m u t
tivu-dlsuam
greaRr than 30 6Pamp lus ig
egnie N2L

11
 4
Re tontVut HOn
N
i'n) A erioduu wih pel ioel selies ay
wKT Xp foulie
dinuop
zpAKed b
Can R
Kn
N - 1

Xp(n) -
CkCk e ()
KD
N-T
wke n: 0, , etplmd
Can
Can e
be ypAihd
CkFouviey Co edhicie nt
Ck
as

Ck
C pn) e t
h:0
V-T
K:0,1,
wK7 nou eq X( e)
N -

Xpn ) e k n - G
x( 2)
4 ()
ue
q
e (
Coupekinq
D

XN ) (
CK)

Subh u Hing i n )
eky
X ( k)
N-

p(n) (8
K0
N

n 0,1
Ja cen thu cHe
tho
Th eqn(e po v)don
4om tha
puoli icnal pfn) XC)
m
SAec+Yu 12
a mples
 3) 6
X(n-An el kn
(7) n0 L: -
N

K 0-N- 1
V-1

p (n) e2Tkn

N-1
whe k: 0, ,

`xCn-Jn) >É4.I')
p n):
n +2n) 4 ( n t n ) + H(n)
+

Cn-N) +X (n-&n)
hlpeHtion
that tn) A a periooliC
This means Xp
N Sampu
oith the peviod &
tn) non-peHd dic
feuu arbiHary Shousv
Jhousen
Caaiodlur A
Let ws ClyntOarns Z
L
FanyPus
famplls
(n).
ignal 3)

ig(sa)An ovignal fignal

lengfh L

CAe- 1
AliaA in 9.tht ignal plhy Aplhy
N>L NO t i Signal
6
o aln)
ln)
NOW kof us
preparu
titr
vepe4itr en
orn
perigolicc Yepe
petiooli
OptainL
wWhich iA
ol by
by has
tha4L
Hhe
thu pericod N be gAa41 u h o w n tin
A the wh
Acd N:6 He sample
hehe Ct n - 4 45,
10
T h wConctude that tar SnOCtVuAm 0 an
apeitodit dciR -me eiqnal wt V
Co ya Y.
Yom îtfmp{y at oA
JeCe LAAd

) Xp (n) 0 n N-/|

XCw) Ca d Copueol ( )
& hinally
XCw)
-CO

14
Pn X Cu
 Discrete Fourrer 7 fanshornm fDFT]
n ) has a hinite dulh Jength L<m
hun p (n) A peyiodic repetiti oh th n)
p (n): n ) 0 n e L-
0 L h M-1
dehinol hov o n

period & N
onsiolu eq i -

pln)e2)kn
x() N
n0
N

thown 1ha4 if n t
have t hu u
rom big (1.2 6)
we thuN
N thun
thin
Fhan N
ou
los than h u a b o e

n (n) lculan

ranp Ca
Iwe e Can wriR
no Aliain Hu
hor X(n),
e

:0 Hnu donsa
AllOAIng Ja
bA
auoid tCan
an
N>L to uum mma h
on
o n
wkT th
S m aH

limit
hent HL upp e r
made ab N-
(9 betoms

e N-I

0
DFT
K: 0,1 . .N-1
A
R
writen a x Ck).
Y ( ae) u
NOR adebuWed by k 6ny
au
Ualus X(3K)
N Fourier Tranljorky 15
ea u (alud a Distve k
 N OW (osiolur eq (
N-

Apln)-

K:0 N-/
n:0, 1,
a bove eqh or
.

evaluae He
iwe
Cn )
Hun Mp (n):

IDET Xn) L 3 xCkJ e Jn Kn

N N
K D

h 0, I,

ThL above eqn gi es the origr nal se xtn) o

Catd InveHe
DF T: Hence" it u as
t
Discrek
Fourher Transhorms DET
i n a o r 7VanáhommaltHon
DFT Os a
WN e (2
N
kut us desine
TWidle hattoy.
wnhn
a4
be
Can
Can
ZDFT e qn
DFT 2
hente
-
K: 0,1, . N-| ( s )
S ztn) wN
XCK)
Kn .N-l-

( K )X(K) WN
-

n:0,,
):
K:0 batlr
botlr cal4
cally
olenoe ol
olLn0te tya
1ynt
Can be
Jo eg(H A
O hollows

K DET Y (K) X(nJy

16
 HeLweea t cloRAv p y +hat ue
Can toLy XCw) oy 7(n) DDm thu a,

uol tu Cona
0 a (COnLplakh thiu tul

ue helaton ehtp Jt4 p )

4 7n)
0 7n
ino XpCn) U neliodt exknin

pCn)
Cn - Jw)
i.

n) Can be Je totae
than
Ul in clocY no aliou in
aliodinm
tala u
non p(n) i
tioleuaiy as
illtra t i
n

Ohon
initdulaHom too
we contider a InRAUal
non TeMO in thi
tn) which ia
oEneL-I
u
u unn N2
hat
we obsele 0nN-|
2(n) p(n)
n an dse ske0OLuled
SO tha4
o p(n) wi+houi ambrq uwy
i+ n04 poib to Atovy
1 NL

n) rom it paiooi ex enaion du

e cdowaun Alios inq
13
 Kn Kn wN
K:0 Wn K:0,n:0 WN wWN K:0, n: I

Kn WN
k:|W WN WNK:1, n:
Kn
K:2 Wn K 2, n o
W W K.1, P: I
N
Kn
wNK:N-i, ns 0
WN WN
KN-IE NK:N-I,n:
n:
Kn Kn
WN WNk:0, nen
WNK0,n: N-
Kn kn
WN WN K,n:M
2(N-)
Kn Kn
WN K:2, n:N-
K:2,JK:2, n:
N-)C-)
(w-l N
Kn
Wn NK:N-ly h:M
kn
w K:N-l, D: NXN

0
wN
WN WN N
w .wN
2(NH)
WN
wN WN
w-y (n-1
N-
(n-l)
ahL
elutnti
al
hu Hu
individu
KYOwS nColumy
K withK

18
 DET 7 E 137 Can dbe upkinkd
Then N-polnt
Cy
18
XN
eq 1 Can ba uprensad
TDFT H
c
th wmatHix homm
in

N
N
Ww
u Coypley lonjugat

w N - 2 m a t Y n a r p H a n

a YmmL
1Ymmuic

that
t ha wNu ver fe P
obsera haA
we umu
eM G
iwt Can
slwriARn

maLtr Hune

L u aning
w
WN
WN WN

(O Wny NT

I an NY 1olLn H'+yma

19
 20

hlt Ser tha valus w bor tu otl valuu

i) N:8
wkT lW : e
N
with N: 8
8 4
ThLR
W e
Wg e a8 t
e u k: 0, 1,1. ...
K:0 w e 1
K:1
Co -Sin
)K: w e Co n, - JSinnh : 0-J
: e JnL
GH) K: 3
Cos -jtin34 -
K:4 w :e to87-}4in 7: 1
er
k: 5
wg-ug
J3
VK: 6 W e e:Cos3m+-jiin3 2
Wg-Wg
) K: 7 tuf : e : 0f - i n
 N OW (osiolur eq (
N-

Apln)-

K:0 N-/
n:0, 1,
a bove eqh or
.

evaluae He
iwe
Cn )
Hun Mp (n):

IDET Xn) L 3 xCkJ e Jn Kn

N N
K D

h 0, I,

ThL above eqn gi es the origr nal se xtn) o

Catd InveHe
DF T: Hence" it u as
t
Discrek
Fourher Transhorms DET
i n a o r 7VanáhommaltHon
DFT Os a
WN e (2
N
kut us desine
TWidle hattoy.
wnhn
a4
be
Can
Can
ZDFT e qn
DFT 2
hente
-
K: 0,1, . N-| ( s )
S ztn) wN
XCK)
Kn .N-l-

( K )X(K) WN
-

n:0,,
):
K:0 batlr
botlr cal4
cally
olenoe ol
olLn0te tya
1ynt
Can be
Jo eg(H A
O hollows

K DET Y (K) X(nJy

16
 22

DET oP gona HAndaro diyna

(1)
Conpute the N-point DFT ho H oll ignaly
o tn) d(n) [unit fanple)
The Unit fanuple Sn) in gien as
(n) 1 ,n:o
0, nt0
N-poin1 DT
DET A& gien by
-
kn 2 (n) 6n)
YCK) S An)e
:0
7 tn):6 n -no)
Subs HHuting o Zn)
XCK X Co) e
x tn) o (n-no)
XK) xn) e t
S S n no) e k n n-no) I;n2no
0, th
S(no) e kno
kn0
e N
 EM M 2MT
M
|29
z

(O

23
 ()
X(n) l Ko
e , 0 n<n-l
X(K) A n)e N
n0

e J3TT on- N

o3(-Ko
-

(enN Ckko) ]
-pd -ko)

- eJ (k-k)
wN k# ko

I - eltk.e
hun K Ko
N

e (k *))

N
n0

0 Km N . dlk-m
X(K) N k:m
24
 DET 7 E 137 Can dbe upkinkd
Then N-polnt
Cy
18
XN
eq 1 Can ba uprensad
TDFT H
c
th wmatHix homm
in

N
N
Ww
u Coypley lonjugat

w N - 2 m a t Y n a r p H a n

a YmmL
1Ymmuic

that
t ha wNu ver fe P
obsera haA
we umu
eM G
iwt Can
slwriARn

maLtr Hune

L u aning
w
WN
WN WN

(O Wny NT

I an NY 1olLn H'+yma

19
 X(4) + wg+ u +w I -I +1 -

/
:]0
Xls) t wg +luug +lug
t t/
6

XCk) I-j3 AI, 0,1joHI, 0,

1+J0.ul, 0,

bnd Ha 4-poin DET tha eqh2en

Je low xCn) , 0, 1, 0
above ind it nersg ws tne
uAing
using XCk) ound
ha dehinlng eq"

N: 4 , WNA e
WA e -J/
w:I w -i, w 1, wJ
DFTtn} *(K)4 n) w
Kn
04kej7
(D)w+n (1Dw
1+ w4 2
X(o I+|= 2
2
D : 1+ -I: 0
Y[2 1 w1t uw tl: 2
9
0

X(K) , o, 3, o} 26
 - Kn
NOR. Ww w
W) 1

ww7* J
xt): IOFT{ T ( ) 4 S x ( -Kn
N
K-0
-KN
X() S_X(K) 4
KD n: 0,',2,3

+ 0 W
- n t 3rw +0t
n -3n
-2n7
J 14 W-an

xn):1, 0, 1, 0

27
 (11)
upuH e - pon D7 D he eQ1 At
matic uuth0d n) 0,1, 2, 3
het N:4 evaluate u Hul
K 0, 1, 2,? 3
w 1 -1 W J
N
2 N7
3

N N4 w Wa WA WA0

3
w W4
w w w w
w w
:
-1

-1 1

-1

NOW wKT

DET

A
1 1 110
1 -1 J | |
-1 1 1
1i-1 i) |3 28
 22

DET oP gona HAndaro diyna

(1)
Conpute the N-point DFT ho H oll ignaly
o tn) d(n) [unit fanple)
The Unit fanuple Sn) in gien as
(n) 1 ,n:o
0, nt0
N-poin1 DT
DET A& gien by
-
kn 2 (n) 6n)
YCK) S An)e
:0
7 tn):6 n -no)
Subs HHuting o Zn)
XCK X Co) e
x tn) o (n-no)
XK) xn) e t
S S n no) e k n n-no) I;n2no
0, th
S(no) e kno
kn0
e N
 Relaron ship DET 10_0thst 7ransjorma,
Relatiot &hip fo 700 hOuriep 7lantypim,

o Cnon-pe Yi0clie Sequent halin

peritdic seg
n habin
[n) habtn
The FT 3 a non

Jength 'N'is qiven by

N-
X(eJw S (n) own
0

he x(eJu) a (on Hn usus fu" b5

FT 2(n) u given by
holis CAL R
e kn
XCK) 7n)
:0 1,I . .

N-
K: 0,
tha
hind
COuypahing e 0 B
version
verdion
ln is Panpled
DFT ven y
Fr d H seqh
the
XCE)xCeJu)
N

2 N-
K: 0,1,

the - t a nsjo
fo
Rolatton Ship n)
Xh) Hte
DHhe
Linie
nin
seq n
(Bnsider WanAurm.
At
At Z
WIth
cAnN

n ) zh
X(z):2
n-O
m t l u o l t s
thu
Hu until tire
until (ure(

tha qualy
eg ualiy
wih Ro c
anpled
af N
z) i Cirdk
Cirrla
i Bn HL
Unit -

pated poins
30
 (n) eJ kn
00 (2)
s Idni Cal to the f r X(elu)
ca
evalua Hd a+ N equaty p9 Cod .
Wk k, 0eEs M-

0ih -+Yons otm

tn)z"
wL ha

N
K:0
0nsN-/

Cutuin9 eq i n eg
X(z) x Ck)eJT kn
K-0
N

N N
K:0

2 01-0n:0
- a 4|

N
K0 31
 - N
X(a) XCk) eleN.
N
N
K D
elnk-
N

XCk
K: 0

N-
x(
-2
N K0
Circh a

i u
evaluad on a unit

e: e'ak.
poins |it,
equally tpaud OkN-I
N zCn)
Cn)
h gnik dul" Seg
ET
we ge+ He DFT
DfT
1 S xCk)
trmns
n
-e ei(w-2
x(e ): Kz0

to the FouYiey
See
Rela tion
priodi C Seq tent iod nw
CoeiCieN-) enep
D fJu
undauuntalju
ndamuntal u hod
n) with
wiHh
c seq" p as
A periodi in a F8
Nepresenkd - Cn 0
Can be
p (n): Ckek

Giren
A l G'rn
aA by the
tke
Coeh{icients
Coey{iclens

Fourier Serey
usua Hu
Ck
epn p n e/27n

n0
ol th
(Ouypaking eg 2 }
XCk:xn) e t n n):
0
32
XCk) N Ck
Conpute8- 8-point DFT for th all feqh:.
xn):1, 1,L1, 0, 0, o, o?
O01h'

WKT WNe
Sinu N:8. K- 0,, 2,7
3
ug . ug,
u
wg-I ug J : i4

K 0,1,2 .N-
XK)S(n) w 1K

XCK) 3k
3
+W +Wg
Xto) +| +| 4 4j
x(D + 3
Wp tw
+L -jL -J -L -jL

xt) I-jj-J
I
j 9.A1
-

aa): 1+ wt wtWg w
1-J -1+j 0
X(3)1+ Wg +Wg tu
25
In.L
(8 Find the
N- polnt DET 0 the Se2"

z(n):;h even 04 n < N-|

0,h 0dd
? N:0dd
N-

XCK) S 7n) p"J3Kn

n:0

J9T K, :
+X(8) e 4 ) e -J9TK. 4
+ tn-1) el2K[N-I)

:0

N
wKt a aN O Ng +

n-N

Ni 0N*e
N

N
N, 0 o t | -N-I+
N-H2

N+I
y) 1- ( a ) N+

34
 2

a-b(atb) (a-b) -1

(1-e2) (14e4k)
DET eh
i n g th seg un
matix mthsd

hu 4-point
Find the
2tnl: £i, 0, 0,1J tàkiny
by
v e i y tht andwe the
he
vv
J Auu
ultt
1D£T 8
-polnt

XN:WN ZN
(a
N: 4
wa w wwu
wy Lw4 W4
XC1)
xto) by 6
xCI) w4 w4'
w 0
y (2)
Y1

-J
-I-

35
 N

to)
C)
72)
-j -

1+

|1-J

36
 (11)
upuH e - pon D7 D he eQ1 At
matic uuth0d n) 0,1, 2, 3
het N:4 evaluate u Hul
K 0, 1, 2,? 3
w 1 -1 W J
N
2 N7
3

N N4 w Wa WA WA0

3
w W4
w w w w
w w
:
-1

-1 1

-1

NOW wKT

DET

A
1 1 110
1 -1 J | |
-1 1 1
1i-1 i) |3 28
38
39
40
 (n) eJ kn
00 (2)
s Idni Cal to the f r X(elu)
ca
evalua Hd a+ N equaty p9 Cod .
Wk k, 0eEs M-

0ih -+Yons otm

tn)z"
wL ha

N
K:0
0nsN-/

Cutuin9 eq i n eg
X(z) x Ck)eJT kn
K-0
N

N N
K:0

2 01-0n:0
- a 4|

N
K0 31
42
43
44
(8 Find the
N- polnt DET 0 the Se2"

z(n):;h even 04 n < N-|

0,h 0dd
? N:0dd
N-

XCK) S 7n) p"J3Kn

n:0

J9T K, :
+X(8) e 4 ) e -J9TK. 4
+ tn-1) el2K[N-I)

:0

N
wKt a aN O Ng +

n-N

Ni 0N*e
N

N
N, 0 o t | -N-I+
N-H2

N+I
y) 1- ( a ) N+

34
46
47
48
37
50
51
52
40
54
55
56
43
58
59
60
46
62
63
64
49
66
67
68
52
 EFFICIFNT (Oupuialnor b
FE
.
LotYo dut HOn, aorihmn
A wiolly us ed Dsp
WKT, DET
mnny
in uany
1portau nle
DET also play an
mcludnay iuso
uol
Dspmusing
applitaHons 8 t , A uayir
SpettunM analykiu Hh eniS +ena
rlHehing . thi eyis+ena
nsorHan
Yeas on
Yeosovn hov
h0Y Ht i C9puta1ion
Copuati
on b DF
bDF7
alonth or
ethiciuut efhreps
Couuputaionally

hor tha
h a
The dipjelsn
The Cumed w or
olis
0u
Algovithmd DFT
evaluaing DFT vonluctn
e
approarhus
fo»
ebhitlens m
maan
nit

a 2 effitiens
Thuuha
Thus onally
Conputat
in a
DFT

) p i v i d 4 (onguel _approah
Hhe N-/point DET
DAFr
Hu N-poins
approuch, Snua llen
Sma lon
in
this C o u p u t a k 1 6 n
o
thl Cauputed
cluud t
lalger
DF7 u
tu
DFrS ohich
o m wr
W- COuLpOS/H h0)
talliol
algorithnu
C o m p u t a H b n a l
F F T ) algotithms
alqohithma
TM FFT)
Hu Siz
Trankhorms
Fast-FoUrier
wul
DFT
Hu KnswnOy
knsuon.oy
Couputtn
tor Tus
2
powy
N racli-2 FfT algonthmy

approath '.
Alintok filelivng DFT a
i1) Hu hormulafin eh
en
i id a ded H data
linuay hilkAiuq
0pelh on
a
Hu algotithuus ilike
/ thus approati Handjotay70
Gte+ (A0ertzef 4 Chiip 7
6EHeL
 bhrCievlt ComputolHm-6 th DF7
.2)E
DFT
Diveet CenupulaHon o
have
oFT, W
Fronm tu dezn a
N Kn
X(K Xtn) >()

0kSN-
Xn) way e Jual Covupun

wN hcutter shith in
7widdk
COLpU n
tncluolu multiplitaions
Sumnuation
T abo
Sunuuing Conplex o
(gusle
adloli
H s

Cw-)

Couuplo unulip coH OK

o any valuss okwe ha
ons O
Coupun nuultiplhcati
Nn0
4N-1) |
aoldiHi0hy
For evaluaton A[E) h1om
O to I
qui y
MO BConupux nLulHphtaHCus : NKN >N

o Coupuv acdldi Hons (N- aN: NC-1)

L 71
Ako eq 0 Can be wnten as
tn) R n) +J X (n)

(un) +/ Jm[u,)

X(K)
n 0

R tn) Re (uSn) +j a n) Tnlus)

h: 0
t ( n ) Ae (ws") - I m luon)

a nd Re lu,") x,(n)
-

Inlush)
:0

4 ual xios

Rn) xRe(u")e2, (nXT m(yn)

- Jual oddiHo
ako (sun ed as 9
NOR 4 Subtrattion A
ineit Jiguly
addit)on in Dsp
CaluoF Sanu Hnuu ad addihe

a jb A ef enlluHo

usta way B Jeprejen Hing

Cpupur o 72
55
 0 to N
YK Valiej pron

hunte
Dov Coplln DFT
CsULplun a d d i t i o r s Xn-1XN

a['.I
O Convedin.
N-9N

Jual additi@uA
aA. fww eg G
70tal
7Ota Lal addiFTOM 1n
DEI 2 aN-9N +2
Couuputa tlen
4N 3N
: N[AN-
Thus for ivect Conyputa tim & N-point
M Holl avithueH
DFT uquus
velue /
uulHplitaHons jor each
N Conpux all Valus of k
mulhplitationA Hor
N Complex
addiHisns hor each vCtuug o k

(N-1) Conypux r all

all valuA
values o kK
e}
addihons jo
NN-) Couupux

oY) or
jor each value 8k k
multipitaions
AN dsal all valuu k
por
naulHphtatews

4N2Jeal each vaul k

4N additions hor
4N-9) Mal
(2N-2 +2N AIN-2) values Gk
all
addiw'su or
Aual
N(4N-3)
74
 earithuuai olarha oliba
In
0TheThe ng
Coupuration 6 DET large dthus i u
ti ConAunlnq. AA vlyÅapidly
0fulations
10tal ao
Th
dunlop
i n R t to
N
pracilal
lourpuny
Huna i4u algonthns
Dor
e6icien
has+
more

tha DFT

wn
Property 05
Peyiacliciry

K +N

0WN jaT
WN: e N
PY wN e N

uplau N by NL|
, KAN-2T|k+n

WwWN e T
N 2

e 22
C

Sumty
K
W
- 1 aloy
- ea
K+M [k+ML?
75
 (A
RocllM -. FF7 lqoiithm
4 (onqu approah,
B1 employin diuid
employin to evaluae
OGonputationally ephicient algorithm
DFT an duelopud.
deC0wp05/Fim
toumposItion
approach dapends on thu de
Th n to
+o Su tuive ly
N-poin+
DFT in
N-p
Salew Si3e DFT

N : Y T3 . .Yv
Yy
V, : % : Y3 Yy Y

N: y FFT algorithy
Catud Yodix
Y
YOLoli'
a explalnd
Lun Y: ,
alqorithm
FfT
2 FF7 algorkk
F£7 algortuy
raolix-
drao lix-2
Tuul o tyos
t y al FfT algohtho
FfT algorith
Decimati9n
in Time
:
a)Podix DIT-FF7} algorieh
tu FfT alg
n vequn ty
Ávequun ty FF aly.
Detiniahon
DIF -FFL] algonher
RaOM-2,

76
58
 Cosiour that H
given Jengh -N Sogk i

.(r)
x(N-D,
I-Sta9
DecimaH thus Seq)X(n) into two Sqh e
en -

Inolened
guL comsioliol
tngth N aM odd-Inoe yio
0dd-Inde yiol
n)2 h
Val 6
Ualuus e ntn)

segh t o , Xa), xa) . . lNa)

X(N-1)
Even Ind ne od
D , x1). x(*D
odd-inoltud caqh
DFT,
wk7 by olefh
N Kn
XCK
n:0
int
u
eunn ddo
0dol
ain) in eq O
olLtimaHng
N
N-2 Kn ,kn
(n) w tS
XCK)
n
n 0
n:0dd
h: e n
Summati on
5ummatiOy
n: r in Se
ivS+
nol SuMmakien uegu
SubtituHing Summaih

in 5elonol
n:ar+

kr Kar+L)
XCK) Cav). w+ X(ar n) W
Y: 0 Y:0
KY
NOH W W
2kr
KY
S h(r) w u
78
 . KY
N KY K h(r) u , ( )
gl) wy t
Y: 0
2
Y: 0
-poin DE7 i/
Point DF7 0
0 dd-Inck u o l

etn Indi t o Seq

N.point DF7f 7
n (K)4 H(K) indugd t
een 2 odd

0EK
4they aM alko eriod c with a foiod

w ha
n Ck: On(K-)1->6
4 H[k): H(k- MI)
aA
eg Can b w ritn
u HCK)(6
Y(KJ GiCK) +
oKS - )

ing e O. e EsE
Can be wnen a
e
CK)

gN-
CK)+ w HCK) ,0ksN-)

on()+H(KR). skerd
2 79
Coiger N: N : 8 Thun 04K2
rom e9 ee qt

XCK Gn(K) + HCK) K:D, NI

GnkK) +w RK+), Rr N-1

X(K) Ck)+ w HCK) K: 0, 1,3,3
L
3
(KFA) twK H(KF4), K: 4, s, 6, 7

X(o) (0) + Wg Hto)

XCu)

H(3)
X(3) n (3)+ wR
X(4) C(o) + wi H0)
(5) n CID + HCI)
X(b) (2) tW H(9)
X(T) C1 3)+ w H(3).
egh ue Bbtau ha jlots
thse SeK
wAing dleCimaHion ef
dlecimatHon o
1Sf Stage
graph ah
8-pOint DFT

80
61
Tno.siag eathN polnt Segh a hurther detimaed

into Seqna length N

4
we hae
NA KY
nCK)
even 00dd.
dd.
n: 0
into
int 0 ven
g(r e
olecinuaed
Indud Seq"
KY
KY
CK)
o Y
Y0
sumnuoutim
Y:3A h I m we 9
Sulsitutinq
Y:2+
k (i1+1)
aK
2gta) w
: 0

2K

9 (a) wN, t2 921+1) W W

K
S 9 Cal +1) Wg
gta) w w
:
ACk)
K
CnCK)A(K)+ W2 8(K) (2)
indsNed
meuzud Seg" J gto)
Jeg" H ir
v en

-point
DF7 H
AK)
0do
82
BK)
 H(R) CCK) + U, D(k) (13)

sinu A(K), B(K), CCK), & Drk) au perioodic

oith a p eriod 4 we Can wuh

CK) ACK)+uy B() K 0, 1, -I

ACK)+ WN B(K+), K:

H(K): cCk)+wy DCK) K: , .

, -
c(k) +okog),
M:- 5
o N:8
nCK) A (K+ wBCk), K:0, |
A
(kF3)+ uB(Ka), K: 3, 3
H(k)C(K) +W4 BCk) , k:0,
(kTa)+ wD(KA), K:3,3
10) Alo) +40 B to)
G A) +w B1)
Ut(L) A (o)
+w B(o)
(n(3) A C)
tw B(i)
ACO): C(o) +w DiD
A)CCD + w Di)
H/2)C(o) +w4 D(o)
83
 A (D)
7

PointA C) X(1)
1 DFT

Gn2)
B Co)
K(2)
poin B) G (7)
DFT

CCo)
X(u)
/WA
poin
ACTCC)
ACU)
DET

DLD) x(8)
Poit
D D4) )
W R3)
Hhe gn Srage clotôpositioy
lo -haph affpA Algorthim
N8
fer N
DIT-FCr
in catens
Cenuplen mulhpli
In qeneral
the no gibe
delina Hm Mion iquu.
aRY n taqe nox

Jelou
B: 2+ )+

xiopis
C o n ple ne a
DO Couuple
Juquilcl for xio nA AuquAnd
vuputaHon B
diveet tO
7 0 - N poit Dy o mul#p'y
te sacto LW,
4( 2/)+N
84
B N N
64
xto) A o)
C poin+
A C) Cn Awr Hen y
D
in g
K
ACK) n wN
n: 0

AK: Kto)+al4) W
AO): lo)+ w xC4) G
AO) nt0) + w nl4)1

A o)
Aw

XA) AO)
a/
flow-qiaph
Shousn in519 1 n
the utaHom of
Inserhing -9apn hor coup
ComY lLe zlow
7he in7
A aA Shown
&-point DET

86
to) ACb) Go)
Awg
xl4) ACI) u
X{2)
(9) Bt)

(6) 13)

BC w
l4)
CCo

W4
X)
2/s wg

H)
W4

DP
-NPtin4

DI7-FfT

CouLpluk
fleuw ghaph N:8
gThi
thm or
1lgori

Shoun abon
hlow haph muullH
Hppiltia
lHa io
onng
g
m
the lex
rom Conpex
8 Conyp
t h l l aik
sraqe
o r each lun adoiH
ons
4 COup DF
4 8 N-point

genhal for CeuyputaHm au N Conyp lek

In thull
DiT- Ff7 A19, adolih
addiwo on
ding (onplu
Mulph(aHin
N
OY each Jtana
fin telu a Vlog agu total
N log,
87
tu C0uPuk Hu D0in DfT udney DIT-FfT d(1
 utOui0n -
In-place (onup
we have the holl points
rom bi9
bit ve vers eol order
The inp ut data app tar in
(a bi't veeHa

aA illustYOLHd belons
o 0 oI 2
basiC CouuputaHonal
EachEach
iutalud
lock in t u oiagra m I6>s

buterhly' e caue ai o13

a
o icug rcam norial
appea's
in u
X[k) ave
The N-point D T ordk
(C) dlomaun olp)
Co) Puq
t u Stag9e
t m ypvesent
n th
In tha SHhgy
S+ngy
posinon no
P2 n Rd
repelenRd
repvel
M
ay
Can be
e
Ca n
each butterhly
n big
belonw
Shouwn in i g

me(P)

Xm Xm+1 ()
H
The value Y is Variable d depends upon
position Hu buttergly.
t h bu Hey
bukAy
CP 4 m (a)
The
Th e olps Xm+ erm
Caltu la ed in
at Staqe (m +I) au1 asM Hu
Valun,
shich
Xn(P) 4 Xml4) hi
th no
tha h 1tage
orn 18
(P) + w m(4)
m+1 (P)- Km
mlq)19
Xm+ (4) Xm(P +- as in-plau 88
pLt%
Thid krlhd conLputa Hn1 iy Knoon
67
 (IL-

X(o)

X()
Aw
W4

20)- X(4)

wg

W4

X ()
1
-1
S+a9e-

SFage-
DI7-FFT flow raph
8-point
wg1
NOFe,
wgW Wg wA

ug0-707-j0 707
-0 707-/0707
90
 8-point
DET f h e Seq n qtn)
Conupute
uAin
In
a t m : L L 0 , 0, 0, o
AQoi thm
OTT-FET
olptd
blpod
nn) olpiSnge-4 S0
1

-iH
1-J D : I-j2414
w
0 X0

al6).0 J
Xt4O
wg: 4-2
-J
Is):0 XW:1+J-HIL
wg:0:707-0/70 -)-1
Y(90o
a)-1
w:-
XA
up-0.70 I+2415
j0-9
0 404 (1-j)+ (0707-/0.76) C1-)
1-jt0,07 -jo72 -0.93 -0-f6,
1 - J9.414
T 4 1 4

x(R):H, -/341H, 0, 1/ 0-Hl4, 0, 1+0-ul

O, 1tj3-414
91
 Hho
(1.
find the 8-point DE7 t Seqr
X(n): ah 0h7, use DIT-FTAl

Xtn):0, 3, 4, 8,16, 33, 64, 13

olp T-51a9, 7,-is, b, -6o, 34,-30,136,-1
op T-5ag: 8s, - IS tjbo, -5 i, - IS] 60, /70

-30tJlao, -lo 2, -30 Fjl2o

XCk): a55, h8 -63+/J66-05 -5 t/ I02,

-78 63-/46-05
-78-63 +/Ab-05, -85,
48-83- 168-05
-51-j1o DFT
yiven xtn) : fo, 1, 2, 3} ind i
Aing DIT-FFT AIg N:4
+H): X(R)

2(:1- H
w +44)
Xy3 9
w - 2+(-)

XCK) -2+j3, -2, -2-]2

92
 EFFICIFNT (Oupuialnor b
FE
.
LotYo dut HOn, aorihmn
A wiolly us ed Dsp
WKT, DET
mnny
in uany
1portau nle
DET also play an
mcludnay iuso
uol
Dspmusing
applitaHons 8 t , A uayir
SpettunM analykiu Hh eniS +ena
rlHehing . thi eyis+ena
nsorHan
Yeas on
Yeosovn hov
h0Y Ht i C9puta1ion
Copuati
on b DF
bDF7
alonth or
ethiciuut efhreps
Couuputaionally

hor tha
h a
The dipjelsn
The Cumed w or
olis
0u
Algovithmd DFT
evaluaing DFT vonluctn
e
approarhus
fo»
ebhitlens m
maan
nit

a 2 effitiens
Thuuha
Thus onally
Conputat
in a
DFT

) p i v i d 4 (onguel _approah
Hhe N-/point DET
DAFr
Hu N-poins
approuch, Snua llen
Sma lon
in
this C o u p u t a k 1 6 n
o
thl Cauputed
cluud t
lalger
DF7 u
tu
DFrS ohich
o m wr
W- COuLpOS/H h0)
talliol
algorithnu
C o m p u t a H b n a l
F F T ) algotithms
alqohithma
TM FFT)
Hu Siz
Trankhorms
Fast-FoUrier
wul
DFT
Hu KnswnOy
knsuon.oy
Couputtn
tor Tus
2
powy
N racli-2 FfT algonthmy

approath '.
Alintok filelivng DFT a
i1) Hu hormulafin eh
en
i id a ded H data
linuay hilkAiuq
0pelh on
a
Hu algotithuus ilike
/ thus approati Handjotay70
Gte+ (A0ertzef 4 Chiip 7
6EHeL
 wKT kNL-J e -J2T
WN
Yedu ty to
eq
Kn Kn
X (K) = n) W +
0

Kn

bbtained
(3
now
in
The olueimation
The
& odd Rr X(e)
ven
9efting
SubjtitL 9 r in eq )

grn
n ) +(-Dz(n+M)|
x(ar)

Yn
W

YN
xtn)+ (n + ) l

Y
Nb oY - |
(aY)

94
95
96
 one Conyplex nuultipli(ah øn
*one >4 Aual ro
t
A Cenver Hed in

K 2)
hov each valus
N ual rios
ee couypUx mulHip Catin7
li 4-N

C u Hd in to 3N Mal it
A

+0 N-/
p10na 0
K VoA
o Conplk DET >4N4Nn lual xie
in
C a u p u x naulriplitationg
N keal t
-
a C&nled in
additiong
additiony
tsuupa
SL
S u hw
Many (6upa
n e u w nuany
1l4 WA
NOw

(a +j (Ctjd)

(atjb)+(C+)d) a1) +j btd)

J e a l7
aeloliHon

o d d iH a

o n e tonupun addiHon
ual additio
1t

A Convel.eod
Hul ak (v-L)
- )
valuus K,
hor each
ConLpuX acddiHns

hun (N-) (oupln an-1) Jual

&N-D Jaf
ddityms ia centelHd! adoi'i

a 73
98
99
100
 (A
RocllM -. FF7 lqoiithm
4 (onqu approah,
B1 employin diuid
employin to evaluae
OGonputationally ephicient algorithm
DFT an duelopud.
deC0wp05/Fim
toumposItion
approach dapends on thu de
Th n to
+o Su tuive ly
N-poin+
DFT in
N-p
Salew Si3e DFT

N : Y T3 . .Yv
Yy
V, : % : Y3 Yy Y

N: y FFT algorithy
Catud Yodix
Y
YOLoli'
a explalnd
Lun Y: ,
alqorithm
FfT
2 FF7 algorkk
F£7 algortuy
raolix-
drao lix-2
Tuul o tyos
t y al FfT algohtho
FfT algorith
Decimati9n
in Time
:
a)Podix DIT-FF7} algorieh
tu FfT alg
n vequn ty
Ávequun ty FF aly.
Detiniahon
DIF -FFL] algonher
RaOM-2,

76
102
103
104
 . KY
N KY K h(r) u , ( )
gl) wy t
Y: 0
2
Y: 0
-poin DE7 i/
Point DF7 0
0 dd-Inck u o l

etn Indi t o Seq

N.point DF7f 7
n (K)4 H(K) indugd t
een 2 odd

0EK
4they aM alko eriod c with a foiod

w ha
n Ck: On(K-)1->6
4 H[k): H(k- MI)
aA
eg Can b w ritn
u HCK)(6
Y(KJ GiCK) +
oKS - )

ing e O. e EsE
Can be wnen a
e
CK)

gN-
CK)+ w HCK) ,0ksN-)

on()+H(KR). skerd
2 79
106
107
108
Tno.siag eathN polnt Segh a hurther detimaed

into Seqna length N

4
we hae
NA KY
nCK)
even 00dd.
dd.
n: 0
into
int 0 ven
g(r e
olecinuaed
Indud Seq"
KY
KY
CK)
o Y
Y0
sumnuoutim
Y:3A h I m we 9
Sulsitutinq
Y:2+
k (i1+1)
aK
2gta) w
: 0

2K

9 (a) wN, t2 921+1) W W

K
S 9 Cal +1) Wg
gta) w w
:
ACk)
K
CnCK)A(K)+ W2 8(K) (2)
indsNed
meuzud Seg" J gto)
Jeg" H ir
v en

-point
DF7 H
AK)
0do
82
BK)
110
111
112
Continuinq this Pro(tls b dotimaion, we
DE7 a
Can upisent each N-poin4
a Combinahim 82wo p o i n t DET
8
So. 0n
N apouer b 2i, N:9
unil
thu pnum u continwd
V loq Stagy

In the above PMaunp u N:8

8-Poin DFT has bun Juduud to
the, C0nuputatien
DFT
Conp utahm 3-point
ar th secon d stage ole timatleh.

t Us Consiur thu9-Poin DET

Xlo) Xl4) we hak

kn
A CK)
0KJ
n:0
4

K
n) w
n:0

A(K) () +)w
A tto)i_ Zl4) Wn[ 85
114
115
116
 utOui0n -
In-place (onup
we have the holl points
rom bi9
bit ve vers eol order
The inp ut data app tar in
(a bi't veeHa

aA illustYOLHd belons
o 0 oI 2
basiC CouuputaHonal
EachEach
iutalud
lock in t u oiagra m I6>s

buterhly' e caue ai o13

a
o icug rcam norial
appea's
in u
X[k) ave
The N-point D T ordk
(C) dlomaun olp)
Co) Puq
t u Stag9e
t m ypvesent
n th
In tha SHhgy
S+ngy
posinon no
P2 n Rd
repelenRd
repvel
M
ay
Can be
e
Ca n
each butterhly
n big
belonw
Shouwn in i g

me(P)

Xm Xm+1 ()
H
The value Y is Variable d depends upon
position Hu buttergly.
t h bu Hey
bukAy
CP 4 m (a)
The
Th e olps Xm+ erm
Caltu la ed in
at Staqe (m +I) au1 asM Hu
Valun,
shich
Xn(P) 4 Xml4) hi
th no
tha h 1tage
orn 18
(P) + w m(4)
m+1 (P)- Km
mlq)19
Xm+ (4) Xm(P +- as in-plau 88
pLt%
Thid krlhd conLputa Hn1 iy Knoon
118
119
120
 8-point
DET f h e Seq n qtn)
Conupute
uAin
In
a t m : L L 0 , 0, 0, o
AQoi thm
OTT-FET
olptd
blpod
nn) olpiSnge-4 S0
1

-iH
1-J D : I-j2414
w
0 X0

al6).0 J
Xt4O
wg: 4-2
-J
Is):0 XW:1+J-HIL
wg:0:707-0/70 -)-1
Y(90o
a)-1
w:-
XA
up-0.70 I+2415
j0-9
0 404 (1-j)+ (0707-/0.76) C1-)
1-jt0,07 -jo72 -0.93 -0-f6,
1 - J9.414
T 4 1 4

x(R):H, -/341H, 0, 1/ 0-Hl4, 0, 1+0-ul

O, 1tj3-414
91
122
123
124
 wKT kNL-J e -J2T
WN
Yedu ty to
eq
Kn Kn
X (K) = n) W +
0

Kn

bbtained
(3
now
in
The olueimation
The
& odd Rr X(e)
ven
9efting
SubjtitL 9 r in eq )

grn
n ) +(-Dz(n+M)|
x(ar)

Yn
W

YN
xtn)+ (n + ) l

Y
Nb oY - |
(aY)

94
126
127
128
97
130
131
132
100
134
135
136
103
138
139
140
106
142
143
144
109
146
147
148
112
150
151
152
115
154
155
156
118
 Angly attordan0 ith
ClaksiiCaHm Filk
th q
euCHLe Challh
ouw pouM iler. and pahs
b'lt
High paM ile band Stop

Cormpalieion be Analog 2 019ital

i
Analeg D'gital
ConHnULU Ilo'Y adu dis LR H
H eiqnay rgna
alu tmplwnr
Impmun ta tHm df Thy e
hilkA Colled O a diqital Comput,
Hue Wlng DP
Out ustny paMie r HlD
Conyd oNnta elumunt fulh a
M CLddlar, y de loy

Ana leg 4ilky Hhask, band m

sotod m lina
diifun Hal eqy lim ditfsunu eqn
Laplata
trans jo rms 2 2-Hran horms a w
4 or analys oY Cnglytid in 2-plar
a l ed
in pla

clad- Digitellk
highev noik seraiivin, Juquius adldHona
non liLaL it ADC DAC Comer
lacle 8 xiblia,
Palaneitn L Yang e tn
fnvironmental,
UM CHol to hal
t fa mpling Jak
4nor
mperoman (6) e gli gle effuu
envio nmental palamr158
 this chap er
thu
couig 1IR
ilt
In
Jealizt ablt 4 P Hable
hat a detai|.
in
dis Cud
hCn)
hín) a
Fo r q
Aeapon
Impulhe
The
Jualizable er
h Cn) 0
nso (1)
+ mu+ hy
PaHIy
m w t faHI
thu
foy srabiliy
Co n d i h

htn) e0 (

haue
7rans
7tansfer un 8
fey hun
drgial i leA
IR

hCn) z-) bez-

H(2)
K
/taz-t

K
ThA doAiqn aan IIR il u qiun
etiica Hes A hindune ilte
CoerCient aKA bks
types_0f Analug tlky ban td
un ban
un ed
which yects
one
APlter i qlous
allous A
he ilp ignal 2
u4 Prom thu luquilud
Auquiled
to
obtain
doui d Hse na
hane a olp Piq
159
 Hq's S19
8 F19 nal tho
Pa band The sang e theilky
Pamed hMo'
that
al bloCka
e 4
Sop ban d tho
Ang lo drhfuent tyy bad
thely uagnituds upom
LP,HP, IP
- LP, HP, p

theiy (ut 8 u
Phau t,
har haiy awpitucle Jespom
hau 0% Hhai y phae JpoM
nd PauMi'y
hatu 6h oluura Achu

allows Low uqA

Leuw LOw pak tilte to pom thto' it
shi aHe nua ky
hig Hi
Idaa
pYaCAco
P
SLC
Design diqital il ftom Analog He
Th mo94 (ommon techniq wmol foY dlesignunq IR
1R di
ilres is knewn as indiLct mothod - 9hich vovel
at dsiqning an ana ls
pYofo Hype Pley & thun Hrasjovmuny
pYo to Hype to a d ) +od i 1
fo ven speijicaHors of dliqttallter, the coluiva Hion
digial ilR Tras eru Mqui Jus 3 Pen
(D Map h dlasiud diqital b'l RY Specijica Hons into
equilont Analog ilter
the ana TE Hor the
() dolive
Hw TF 6 Ana
analoq pwtoty
TYansh0rrM
tut a equrva un C'q'Eal hilRr ptotyfu 160
tansdey tm
121
 TIR-Puey
HA Analoq Fil trsuvfh

-p
ranaitHnband

t Pas
wand
Tray

ds 4

p S
k p.bands

Snetiiraion LPF
des 19n
in Hu
An Importan HD he
deini Hom 6h
A th
analoq
JNLuifCaHo
ns the
tha
Heg AAs ponM he
Sheuld bA t e satie fied by
by
eY spon
lse hous H
hau Hhe
SleustcoHous
descai
Thun
in t SHondy -Sran
ilRY a tts
to SinuAeidal y

aboLe Phous t y pical

J u pork
iagnitudo taq
Lp F
> ThaAange Hi4 h Siq nal hatall
POMJand
pakd hdo' thu dlk,

tbpband ho uaA that a dloc ked162

 p aM band edgo Hea
s S4op band edge ha

Th Aange
Catud ransilHon band
A pobidnd
ho Sneeitcai8n

hatchuo ala tndr'cah

thu
yTha in
nuaghl+udt valuy
forbidoden

andy
tra nsisHon and,
feA
M0n otonilal4
uayni'tudt
h n thu band

mathemaHCal
desaipHon h e
JsponM

ran (e e4 magniAuoe Ausp onle

dp t0le
in pom band
dAiLd Maqhl +ud
n paMband u
OP Pasbanel+ttelan
163
 na anituol AeAponte
Os 1o levane
n Stopband.
AponAe

olaike o niaonif ud
A
n the Stopbano

dp-Pa band a H e
Xwe define
o log (1-dp)
OAP
Pasband ppu in d

AP 2o loq (I -6P)
A doinao os pos band gaiy

Ss- Stop banod atenua t on.

3AS:-a0loq 6S A deinnol
OM +Op band ua i on
a4Hen

or Jpple in do

KA As aolog dS
JA
A doinud CA
top beLnd qan
ad De.

&P
Sp DolA band 7olakante .ipple
NO d topband
pom band aHe nuatia
f o p bamd 164
124
 N
N- 10
AC

ninuolt Jespondy
79p)cal nLaq
buetwotth filt
ale uads )
The hetf obserua Heus
t auboe i9.

Hio)| 1 orall N
a c foy
a u finite N

Tuun OA tha o fogHC7A¢S: -3.01de

ChalaeKHU u Sand
Sald tu
b TuL wlaghitud
be mainLatly lat

0 foY n:1,,
d /D0
S)H | u a dataakinq
monotonitally deakasing

by any
i8,14C) )|<HCi4,)
Values J 2 Juch hat 166
0S
The waaniude- Pquaki
Pu uponse (6
Ha (6
BuHerworth filer u
nomalizac :1) LP
2
H(J
+ N
Al: A xA*

HNCA) H(-la
1+ N 8
Suplacing jn by SS hena A:. in
we qet

HN(33 H-S)
N
(
no finik
has
H / ) Hl[ )
ThL +ranaher u

pmduct HN) Hw/-8)

+The poles of h
eq uaHny Hu
H
a deel m ined by
dnonuinatoY to e o

IN

167
 S
9N-1
1

,
Jn(8k +u)/ ks0,

eT-1'= e

y
aie qiren
The poles
jT(IK
k) inlz
SK N

( )
Sk e

a Ha
HN-e)
HN(4J
T poes that that lie
ChoateSH Ceq
Yo0ts 6Hu uni+ ladiu al
On he Cirde

p lased
g144n at angluy

N N

168
127
N -N-I i v brcloy Pilr

SK

IN-I
K 0,
foY N:I, K: 0, 1

JT
So I 1Go - | |7 Ie Co8 M+|Sin7
1+J0
So

Cas&n t jtinan
S113 :
1e
170
 BnC) T (f-SK LHP 1) RRP 8

-So

B,(& S-(-1) H)
(S ) unl+ o
B,C) (s +1) S+I Cro Check

H,):
N29n
orden filiey )
9
fY N: K 0, 1, 3,3

o2

K: 0,1, 8,3
Ox +
So: 110p
3T
2

o 3l4 Cos 13Stjsin 13s

i) K:1

Cas St /Sin s

--i 171
 3IS

33tar 9T40
4 4

S3 |9 Co89+j sin
14

H8)
-e) Cr-s)
Ba(t): t:) (t-)
S

h-3) (? )

H CA)
172
130
174
175
176
133
178
179
180
136
182
183
184
139
186
187
188
142
190
191
192
145
194
195
196
148
198
199
200
151
202
203
204
154
206
207
208
 IR File Ydosign
ed analoq DilRYj
M
chasaCHns H'u e Commonly to análoq
to a náloq
BuHeluwor+) Cheby g hee il
PYS. Analo
Analg
aHon
eq HanjYm

n+rodutH B4n unuanHd

unuonHd
Mjeth
n e uwhutb the
all0wf Hhe
AFilter
4 allow
rom
ttu ilp siqnals Jequilo
the
Aeq
dosisad tlaq ? to
0btcun

Phaps) elp eignal

TYPA 0 _ F i l d
analog fla
Filte
dig'tal
Coninusus hnJion,
ol
analog iler -ilp &elp oli CLLe
-
Hm fr nak
Diqiral filer -

9euilally
diqjtat hilk RA aa u Jupenke
JupOme
A appoxima
shith
Pye em
wth i bern digita!
ip beng digital
ioltsilod auo drq'tal Paumpay
faupa 2 eP
Diqital F i l
Hpes of

IIA FR
Ininite inLpute Rspontk) inin JmpuRerp on
04
Ol non uukiu
pubtnt elp dep encl
ilp
p.hue ntilp, pokt pu nt 0lp duundy
Blp sapl
-Po t p faupe
157
 Module -5

DIGITAL SIGNALPROCESSORS

210
 Basic Architectural Features
 PDSP’s should provide instructions similar to
microprocessors.
 Basic computational capabilities provided by the way of
instructions should include the following:
1. Arithmetic operations
2. logical operations
3. MAC operations
4. Signal scaling operations

 To perform all these operations a dedicated high speed H/W

must be provided.

 The architecture should include the following H/W features

also:

211
 1. on chip registers – storage of intermediate results
2. on chip memories – signal samples (RAM)
3. on chip pgm memory – pgms & fixed data such as
filter coefficients (ROM).

1. Investigate the basic features that should be provided in DSP

architecture to be used to implement the following Nth order
FIR filter.

Where x(n) denotes the i/p samples

y(n) the o/p samples
x(n-i) is the i/p sample I samples earlier than x(n)
and h(i) the ith filter coefficients

212
 DSP Computational Building Blocks
The basic building blocks that are essential to carry out DSP
computations are as follows :
1. multiplier
2. shifter
3. MAC unit
4. ALU
MULTIPLIER
 Earlier multiplication schemes relied either on
S/W or Micro coded controllers
 Both these options require several processor cycles to
complete the multiplication

 The advances made in VLSI technology in speed & size made

possible the H/W implementation of parallel multipliers

213
Before designing an actual multiplier, the specifications such as
speed , accuracy and dynamic range must be clear.

accuracy and dynamic range – is decided based on the

number of bits used to represent the multiplication operands
and whether they are represented in fixed point or floating
point format.

Speed – is decided by the architecture employed.

214
 Parallel Multiplier
Let us consider the multiplication of 2 unsigned numbers A & B

Let
A --- represents m bits multiplicand [Am-1, Am-2…A0]
B --- represents n bits multiplier [Bn-1, Bn-2…B0]
P ---product of A & B . Max (m+n) bits

215
The multiplication operations using 4 bit s for A & B are shown
below.

The fig below shows the H/W structure of the multiplier for
this example and is called as Braun multiplier.

216
The structure of 4 × 4 Braun multiplier

217
 For a n× n multiplier we require n(n-1) adders.

 The Structure requires : 12 [3 i/p & 2 o/p]

 Braun’s multiplier does not take in account the signs of the

numbers that are being multiplied.

 Additional H/W is required before & after the multiplication

when signed numbers represented in 2’s complement are
used

 Let us consider A& B represented in 2’s complement format

A & B having m & n bits respectively

218
 -------1

-----2

The product p = pm+n-1 is written as

-----3

the 2 subtractions in eqn 3 can be expressed as addition of

2’s complement number.

 This can be implemented through a structure similar to the

Braun multipliers using only adders.

 This modified structure is called as Baugh Wooley multiplier.

219
220
 shifters
 Is an essential component of any DSP architecture.

 Required to scale down or scale up the operands & results to

avoid errors resulting from overflows & underflows during
computation.

Let us consider the following cases.

1. Let us consider the sum of ‘N’ numbers each represented by

n bits.
 As accumulated sum grows the number of bits required to
represent it also increases

 The max no of bits to which the sum can grow is

(n + log 2 N ) bits

221
If each of the N no is scaled down by log 2 N bits prior to
addition, the loss of the result due to overflow can be
avoided.

 The accumulator will then hold the sum scaled down by

log 2 N bits.

 Accuracy of sum is lost but the summation would be

completed with out overflow error.

The actual sum can be obtained by scaling up the result

by log 2 N bits when required.

222
2. When 2 no’s each of ‘n’ bits are multiplied, the product can
have a max of ‘2n’ bits

 when this product is saved in memory which is ‘n’ bit wide,

the lower order ‘n’ bits are generally discarded resulting in
loss of accuracy

 in case of multiplication of 2 signed no’s, the accuracy can be

slightly improved

 By shifting the product by one bit position to the left before

saving ‘n’ higher order bits.

 The accuracy improves because instead of discarding n dits

we now discard (n-1) bits

223
3. When carrying out floating point additions

 the operands should be normalized to have same exponent.

 This is done by shifting one of the operands by required no of

bit positions

Problems:
1.It is required to find the sum of 64 no’s each represented by
16 bits. How many bits should the accumulator have so that the
sum can be computed with out the occurrence of overflow error
or loss of accuracy?

Ans:- sum grows by a max of log2 64 = 6 bits

to avoid over flow the number of bits the accumulator
should have is 16+6 = 22

224
2. If for problem 1 it is decided to have an accumulator with
only 16 bits but shift the numbers before addition to prevent
overflow . By how many bits should each number be shifted?

By 6 bits to right since the sum grows by 6 bits

3. If all the numbers in problem 2 are fixed point integers , what

is the actual sum of the numbers

The actual sum = (contents of accumulator) × 26

4.What is the error in computation of the sum in problem 3

Since 6 lower bits are lost the sum could be off by as much as
26 -1 = 63

225
 Barrel Shifter
 In µps shifting is implemented by a operation similar to one
performed in a shift register .

 The operation takes one clock cycle for every single bit shift.

 In DSP’s shifting of several bits in a single cycle is possible by

a combinational ckt known as Barrel Shifter.

 Barrel shifter connects the i/p lines representing a word to a

group of output lines with the required shift which is
determined by its control inputs as shown below

 Control i/ps also determines the direction of the shift (left or

right)

226
 If the input word has ‘n’ bits, shift is from 0 to n-1 bits

 The control i/p requires log2 n lines to determine the no. of

bits to be shifted.
 One more line is required to indicate the direction of the
shift.

 Left shift:- bits shifted out of the i/p word are discarded and
new bits positions are filled with zeros.

 Right shift:- the new bit positions are replicated with the MSB
to maintain the sign of the shifted result.

227
 3IS

33tar 9T40
4 4

S3 |9 Co89+j sin
14

H8)
-e) Cr-s)
Ba(t): t:) (t-)
S

h-3) (? )

H CA)
172
 1. A barrel shifter is to designed with 16 inputs for left shift
form 0 to 15 bits. How many control lines are required to
implement the shifter .

ANS: 4 control lines are required.

Multiply & Accumulate (MAC ) Unit

 The configuration of a multiply & accumulate unit is commonly
known as MAC unit as shown below.

The MAC unit is used to implement functions of the type A +BC.

 Multiplication & accumulation are 2 distinct operations, each

normally require separate instruction execution cycle.

But they can work in parallel.

229
230
Pipelined operation of MAC

 The pipelined operation makes use of the fact that the

multiplier and adder can work separately and simultaneously.

 Consider a example

y(n) = x1h1+ x2h2 + x3h3

Cycle time number Multiplier Adder

1 x1h1 -
2 x 2 h2 A+x1h1 A

3 x 3 h3 A+x2h2 A
4 - A+x3h3 A

231
175
1. If a sum of 256 products is to be computed using a pipelined
MAC unit, & if the MAC execution time of the unit is 100 nsec,
what will be the total time required to complete the
operations?

ANS: total time required = 257 ×100×10-9 sec =25.7µsec.

Over flow & under flow

When designing a MAC attention has to be paid for
1. word size at the i/p of the multiplier
2. the sizes of the add/ subtract unit &
3. The accumulator
as over flow & under flow conditions can be encountered .

233
 Techniques to prevent over flow & under flow conditions are
1. Barrel shifters at i/p & o/ps of MAC unit
2. Guard bits in the accumulator
3. Saturation logic

1. Barrel shifters
 shifters are normally provided at i/p & o/ps of MAC unit.

 the i/p shifters helps to normalize the data samples as they

are fed to the multiplier.

 The shifter at the output are used to denormalize the result

after the sum of products before storing in memory.

 o/p shifters can also be used to discard the redundant sign

bits or to shift the o/p by required no. of positions.

234
2. Guard bits:
 If accuracy is preserved, the i/ps are not normalized.

 When repetitive MAC operations are performed the

accumulator sum grows with each MAC operation.

 This increases the no. of bits required to present the result

w/o loss of accuracy.

 So extra bits must be provided in accumulator called as

guard bits or extension bits.

 After completion of computation, the required sum of

product , the extension bits may be saved as a separate
word if required.

 Or. The sum along with the guard bits may be shifted by
required amount & be saved as a single word.

235
2. consider a MAC units whose inputs are 16 bits nos. if 256
products are to be summed up in this MAC , how many guard
bits should be provided for the accumulator to prevent
overflow condition from occurring.

Ans: 16 ×16 multiplication has 32 bits. Since 256 such products

are summed, the sum can grow by max of log2 256 = 8 bits
guard bits should be provided for the accumulator to prevent
overflow condition from occurring = 8 bits.

236
3. Saturation logic: -
 With or w/o guard bits ,an over flow condition occurs when
the accumulated result becomes larger than the largest no. it
can hold.

 When handling a –ve no. an under flow will occur if the

accumulator becomes smaller than the smallest no. it can
hold.

 So it is better to limit the accumulator contents to most +ve

(most –ve) to avoid an error known as wrap around error.

 Limiting the accumulator contents to its saturation limit is

achieved with a simple logic ckt called as saturation logic as
shown below.

 This ckt detects the overflow / underflow condition &

accordingly loads the accumulator with to most +ve or most
–ve value

237
 the overflow / underflow condition is detected by monitoring
the carry into the MSB & the carry out of MSB.

 If carry-in ≠ to carry out , the overflow / underflow occurs

 The selection bet’ the most +ve or most –ve value is made
based on the sign bit of the no.

238
 239
Scanned by CamScanner
 240
Scanned by CamScanner
 241
Scanned by CamScanner
 242
Scanned by CamScanner
 243
Scanned by CamScanner