Topic 3.

Description of continuous-time signals in

time-domain. Convolution, correlation, sampling. Basic
signal operations-time shifting and reversal- Even and
odd signals,
Different signals are function of either time (e.9., music and
voice signals), space (e.9., images), or of time and space (e.9.,
video signals). ln this course we will concentrate on timedependent
signals.
The description of a signal in time-domain, broadly speaking,
is what we see on the screen ofthe oscilloscope (Fig. 3.1).
1

o.a
o.6
0.4
o2

42 fii i .fl Elr ."

-o-4 f
0_6

-o-8
I
o 2 4 6 a l0

Fiq. 3.1 A segment of a speech signal

T!*s. Ae ildepelds! vr:anl6 B tiB6 r, ud rlo Eluo ol :::€ f*.;! .t s:: r:G !o, r'{ro), ;s a r.J (or a
oadu) rale aAtlld4d i. !:a@_- s;srEls 6e ra] n i! u*as r ;h€ory b 52tJE -JE .ptim .i @trpld
r€lued sis!d:.t h js asdd .h:r b.rh liia.. ad agraloplil*:. rd e. r::r m:i 6F:1: it!€ed,lrm

According to lhr dimensior of dreir srurDrt, 3ig.ats ..n bc oflinie or of lifuii. suplon. Sur!
pon dbe und6.@d as rhe time intdEl of th{ signt outsid€ of $hic}: rhe senal is il&ars

Phasor reoresentations ofsine and cosine siqnals

.1 si'usojd l{t :s a p€dodi( signal represented by

-rO:Acosiool+*)
wherc -,r is rh. amdnude (}o = 2'lo c tbe frcquenq in .iF.., ..d 1' G tlt tLm in ndir6 ''ne
sisnalr(r) is deined tbr lll !al!$ of1, rnd it tepetrb ?€r'odiallj shh a tEiod To : 1/o (Rc), so
,hitfo n (h€ t€quc.qv nr .]deslf. or i. uenz irl, ) {iD honor of I L R scn } civ.n dr^t lhe uDirs
of.h c.rd/*c rhe aorh$6d,i6t /s) r G()= kad), \rhi(h.oindds r'n\ ihe unnsofrne
plrr* ,r', rnd pe..!tu di.onrliur:rio! oalh...sin? lf1, : 0, rhen nO is, cosnF. a nd al !' = tl1,
,hi i u irJ \in.
Ifore knows the fiequencyo{ (E vs) in rhe above eq@riol! lhe cosine is chraclerzed by its
lf.n. tnoH (rirr.} in liqrrrio' (o.r,r}, !!€ rosinc is .i}ddizd br ir.
LI{ tieqR.ct co
inrpllild. and phs. ftis p.nris us to d.6nr plna e *.omrld r{nbeE chrt ftiad b-v dr
.nnphude ind rh. ph e ofi osnro signrl oia.eflrin ftqurnq Ao 'llrn. ir ao! a vohagc sianrl
(): A.o{eor + r) nre @rsrondins plrtrsor is
v =,+;r = A.as1lt + j|tinll!') : AZ,!

v(t) = lP'el\uz.tl:lk1^e)(%r+t'I : A<os(eo. + {,)

On. .an rln's rnrl oldre wh.Ac sienal {t a. th. poidiotr o*hc phi$r l'ont. rh. rcrl aris rnd
unitrscoun(c.(l{lTis. anrofaond/sr\timcr=orh(Jns'coirlrphlsoikg.Cr.rrLt
rhc phanr..6nnion h rtu.fordnlt 6nc fr.qurD.,v, h lht.rc Oo. rnLl ir is nlsn'i.orncd.i tJ1

ih.ingl< 9 (J bc (*d b ,Ji6i retr(ilr. .osnr6 rnd snrcs. fttr i'xrmcq tehdr
lntercsrinsly enouBh,
rdvins irou .' a hk oal4, Ccr@r6 rs r proi{rio. on rhc r.i $k $c
{, = 0, rhc phiior v
volhs. sisffl rds(rrro, whil. $lrn I = -i12, rh. plsor v .r . .dc of
'novirg.Rnftl $ain
Oo g.nchr6isinueidAs'n{l'roo = Arc(o.! -:/1) $n is proj{&d onlo thc rql !xis.'ihGstrb.
lishs thc wcll-hnowD li'( tlr sin. lilr ih. osnr by.' ?/1 6di s or 90 ds*res, o! rhat re(snic
lcids rhr sine l' 'lE or90 <ji:R!-ft!! dr F.oentio. ind dado ofstrEr ind cosi.cs
Ediins
'/2
can be easily oblained using the plor in Fig. 3.2.

Fig. 3.2. Generation ofsinusoids from phasors ofa frequency Oo

Transformations ofthe independent variable of the sionals

A number of important operations ale often p€dormed on signab. Most ol ttres€ op€r_
arions involve tiansfornations of the indePetdert variable. It is imponant that the
reader know how to pcrform such operations ed understand rhe phvsicrl E€aning of
each one. The lhree oPeradons we disuss in thiE section rrc shiftidg, renecling' and
time seling.
The shifting operation
Sisnal r(r - 6) repre"senh a tine-shifted ve6ior of r(t); s.e Frer're 33 . The
shift in
d;e is to. If to > 0, ther lhe signal i5 dclay€d by qo seconds. PhJsicaly- ro sonot late
on negaiive values, but ftoD lhe aDa$i@l viewloi , rG .J, 4D < 0, repre€ent! an
-
advanced replica of a(t). Sisnals rhat de relat€d in this fashion aris€ in applicrtions
such as radar, sond, @Municatior sy$ens, and s€ismic siSral Foc€$ing.
 O ,o-r,
Fig. 3-3. lllustration of shiffing operation
The reflection oPeration
The sienal r(-.) is obtained from .he signal ..{t} by a rece.*id lbctr t = O (i.e- by
.eve6insr(r)), a! sho}! irFrglrl€ 3-4 ThE,if'(r) rcpr66tsa sisual olt of* vid€o
{eodec llen r{-t) is t}e siAnal dt of a vid.o plays x&eo the lewind $.i.ch is
puhed on (assurl1iu rhat tne reriDd and ptay sp.rds d thc iee).

zl++3)

t 5-1 -1 -2 -r
(d)
tc)
Fig. 3.4. lllustrations of reflection operation
 Even and Odd Signals
'Itr addiiion to its lse in reprseriing ph]Eid pnenoEeDa such 6 rhat i. the video
r@rder elmple. renedior is exrebely usefll in exdininS rhe syMtry piopeitiB
thar the signal my possess. A siglalr(r) is refered io s M et6 sigDal, o, issid ro
be €ven srrrnetric, irit is idertic2l io its reflection about the oigiF{hat is. if
x(- t) = x(t)
A signal is refered to as odd sydnetric if
r(-d = -x(')
An dbiEary Sgtral t(t) cm always be expressed as a su of eyer dd odd sisrals as
r(r):r.(r)
+ r,(r)
i'here r.(t) is €aled the eveD pan of r(r) dd is given by
1
r_fr)=.tr{r)+r(-r)l
'.' 2'
ad re(t) is drbd the odd part of r (r) ed is expre*sed a
.,(l)=;1.(r) -.(-01

Fig. 3.5. Illustrations oieven and odd signals

The time-scaling operation
Consider &e sisals r(r), r(3t), ard r(l), as sho*a in F:slre 3.6 , As is seetr in the
fisure, jr(3t) cBn b€ d€scdbed as r(r) coDtracted by a fa.tor of 3. sistilaru r(/2) .in
be de.sibed 6 r(.) erpaaded by a factor of 2. Both r(3r) atrd r(./2) arc said ro be
rime-scaled versions ofr(r). ID gereral, if the indep€nded variabl€ js scaled by a para-
rnerer r then x(.t) is a compressed ve$ion o:;G) if lnl > 1 (the signal exists in a
snsxer tine interval) ed is an expaded versioE ofr(r) it ln | < 1 (the signal ensts in
a larger tiEe int€rval)- ff se think of r(t) as tbe output of a videotap€ recorder, tfien
r(3r) i! &e sigral obtained wheD the recording js played bact at lhre€ tim6 the sp€ed
at which it was recorded, and r(.2) is the signal obtained when th€ recording is played
bad ar b.lt sped.
 Fig- 3.6. lllustrations of time-scaling operation

Basic siqnal operations

The following are basic signal operations used in the representation and
processing ofsignals (for some of these opelationswe indicate the system
that is used to realize the operation)i
1. signal addition-:fwo signals x{t) and y(t) are added to obtain their
6um z(t). An €ddelis used.
2. Constant multiplicatlotr-A signalx(0 is mullipli€d by a constant q. A
co nstant nult iplier is used.
3. nme and lrequency srriftirgF.The signal x(l) is delayed r seconds to
got x(f-r) end advancod by r to get x(f+r). A del€y shifts right a time
signal.
4. Time wlndowing-A signal x(f) is multiplied by a window signalof tv(l)
so that x(f) is available in th€ suppod of lt(t)

x(t)
-----+
+
(b)

{a)

rIl x'
x(t) _r) w(t)
------'l---
perav

(c)
w{t)
(d)
Fig. 3.7. Diagrams of basic signal operations: (a) adder, (b) constant
multiplier, (c) delay, and (d) time windowing
 Convolution of two siqnals
'We now tum to a more detailed study ofthe convolution oftwo fimclions, a very
important operation to coNider v,/hen looking at LTI systems- We wite the
convolution oftwo tunctionstrt) and s{t as
te
I f\t\s(t _r)dt
f.s= J-6
The convolution operation
. is conunutative, i.e.,/.g = g',f
. [/'g].h = f'lE'hl
is associative, i.e.,
. /.(g + h)= f's + f'h
is disfibutive, i.e.,

Cross-correlation between two siqnals

Cross-correlation is a way to de3sure the siEilarity betwEen two energy signals.
For example, $€ might wish to measur€ the properties ofatr unl(nown signal by
comparing it to a lmow! signal. The cross-conelation between two signals tl (t)
and rr(l) is a funcrion of time tbat has a form that is somewhat analogous to
convolutior. Specifically, we compare jrr(t) to a time-delayed vqsion oft2(t)
The cross-correlatior finction -RD(.) is given by

R"@:
J: r(t)x20 r)dt (3.1)

We can fbi.lofthe parameter 7 as a time-lag, which acls as a sc'nning parametet

That is, we arc searching or scanning one unloown siSDal ta s€e ifit contains
featur€s that are prcseDt in the knolrD siglal. Ifthe ts/o signals alo coffelate in
some way, then (3. l) will have a finite but nonzero value wet some range of z
To se€ why this is the case think of orthogonal fimctions. Two functions /(r)
and g(t) a& orthoSonal pmvided tha! the integral ofthek prcduct vanishes:

/(t)s0oo d, = 0

Forexample,supposethat/(l) = cosIandg(l) = sin, Weplotthese functions

in Fig. 6-1, showing cost as the solid line and sitl as the dashed line lna
seDse, thesefinctions are opposites. Consider how these functions b€hav€ over
lhe range 0 j I:rr/2. Notice that when cos t is Aecreasing, sin I i3 increasing.
When cost is at its maximun, sinr is at its minimum, and !'1ce v€rsa. These
functions do not seem very similar. If we iDtegrale over a iirll perio4 lhis idea
gots some validatioD, sirce the i$egral vanishes:
 .,/ I
Fig

Frg- 3.9. A plot of /(l) = cos, (lolid lin€) aid p(t) - sir (r - 6'./5) (drshcd linc)

tk sin']t l' sin2 2r

costsinr&= 2 $=o-o=o
,/o L= z

Now lett consider time-shified sine function. This tihe lr]€ continue to take
a
/(t) = cos I but inst€ad set g() = sh (t -
6r!/5). We show a plot of these two
tunctions h Fig. 6-2, agair plottirg /(t) = cos t wilh a solid line.
This time the behavior ofthe two firnctiois seems to be simi,l,f, or coftelated.
Wile s(r) = sin (t 6r!/5) can be viewPd a! leading /(t) = cos t a bit in timg
-
we lee thrt the tno signals ha]€ basically the s3me shape and behavior. This
time the iutegral oftheir produ.t ir noDzero:

t2' rf -
I
Jn\
cos{O sin(r - 6z15) dr =
iV l0 - 2./5

Retuming !o the cmss-corelation finctioD, rotice ioB (6. I) that ifwe exchange
the roles ofthe two sigmls we obtain

fd
R,r(')- r)dr
/*x,(')ri(' -
From this e€ s€e that gaerally speakiDg, the cross-correlatiotr is not cornmu-
tatir€; i.e.,
RDG) + R2t{t)
 Autocorrelation of a siqnal
Autocorelatior is the crcsssrrelation ofa signal with its€lf Agair !€ tim+
delay the signal and sca. for pattems. ln short w ue finding repeating pattems
in a sigral. The fomula ued for utocorr€latior is

,'',1';: (6.3)
/-'1rp1r -
zy ar

The nomalized aersr cort€Dt of a sigDal is r€lated to the autocorelation

ftnction by setting z =0h(6.3),giving

n,,(o) = t''(r)'?d' = E (6.4)

/-
To fnd lhe autocorrelation of a poq/er siglal we compr.& lh€ tim€-average of
the autocorrelation ftnctior. That is, we compute

I tfl2
xll(r) lim;,
= r_e t J -Itz
.rr(t!r(r - r)d/ (6.5)

In th€ case ofa periodic signsl, w€ comput€ thc tim€-averrgc autoconelation

ffnction ovcr lhe period to

t tlr/z
jt,,G) = , *,(,}I,(,-r)d, (6,6)
=ro J_hn
S€tting z = 0 in (6.5) giv$ lhe a\€rage power in the signal

xI(0) =
,. I ffn kr(r)12d.
=P (6.7)

CHECKLIST OF IMPORTANT TERMS

Adder Multiplier
Autocorrelation Phasor
Convolution Reflection
cross-correlation Time.windowing
Delay Time-shifting
Even and Odd Signals Ti me-reversa I