You are on page 1of 31

6.

02Fall2012
Lecture#9
Asmallpostscriptonmultiplerandomvariables
Introductiontomodulationanddemodulation
Input/outputdescriptionsofsystems
Lineartimeinvariant(LTI)models
6.02 Fall 2012 Lecture 9, Slide #1
DealingwithMultipleRandomVariables
PDFofarandomvariableX:
fX(x) 0 and fX(x) dx = 1
(thisintegralisovertheentirerealline)

Thenaturalextensiontothecaseoftworandom
variables XandYisthejointPDFofX and Y:
fX,Y(x,y) 0 andfX,Y(x,y) dx dy = 1
(2Dintegralcoverstheentirex,y plane)
ExpectedvalueofafunctionofX,Y:
E[g(X,Y)] = g(x,y) fX,Y(x,y) dx dy

Andsimilarlyformorerandomvariables
6.02 Fall 2012 Lecture 9, Slide #2
InourSignalDetectionSetting
Lastlecturewediscussedaveragingmultiplerandom
variables:A = (w[1] + w{2] + .. + w[M]) / M
andwantedthemeanandvarianceofA.
Hereeachw[n] wastheadditivenoisecomponentofa
receivedsampleinafixedbitslot,andassumedtobea
zeromeanGaussianofvariance2,independentofall
otherw[.] .Thesew[.] constituteadditivewhite
Gaussiannoise(AWGN)whitehere=zeromeaniid

Strictlyspeaking,weshouldhavebeenworkingwith
thejointPDFoftheM randomvariables,inan M
dimensionalspace.However,thefollowingfactssuffice
togetusthroughwithjust1DPDFs:
6.02 Fall 2012 Lecture 9, Slide #3
TwoImportantFacts
WewritethesefortworandomvariablesX andY, buttheresults
extendtoM randomvariables.Also,g(.) and h(.) beloware
arbitraryfunctions.
1. Expectationisalwaysadditive,i.e.,
E[g(X,Y) + h(X,Y)] = E[g(X,Y)]+ E[h(X,Y)]
Followsfromthefactthatintegrationisadditive;needsno
assumptions(apartfromexistenceoftheexpectedvalue)
Inparticular,E[g(X) + h(Y)] = E[g(X)]+ E[h(Y)]
TheRHSonlyneeds1DPDFs,notjointPDFs!

2. ForINDEPENDENTrandomvariables,expectationis
alwaysmultiplicative.Infact,X andY areindependentifand
onlyif E[g(X)h(Y)] = E[g(X)].E[h(Y)]
forallchoicesoffunctionsg(.) andh(.).
Again,theRHSneedsonly1DPDFs,notjointPDFs!
6.02 Fall 2012 Lecture 9, Slide #4
ASingleLink
codeword
bitsin
generate
1001110101 digitized modulate DAC
symbols

NOISY & DISTORTING ANALOG CHANNEL

demodulate sample&
ADC &filter threshold
1001110101
codeword
bitsout

6.02 Fall 2012 Lecture 9, Slide #5


DigitizedSymbols

1001110101

Sample
value

Samplenumbern
(discretetimeindex)
Distortednoisefreesignaly[n]atreceiver

6.02 Fall 2012 Lecture 9, Sli


Slide #6
ASingleLink
codeword clocked,discretetime
bitsin
generate
1001110101 digitized modulate DAC
symbols

continuoustime
NOISY & DISTORTING ANALOG CHANNEL

demodulate sample&
ADC &filter threshold
1001110101
codeword
clocked,discretetime bitsout

DAC:Digitaltoanalogconverter
ADC:Analogtodigitalconverter
6.02 Fall 2012 Lecture 9, Slide #7
ASingleLink
codeword
bitsin
generate
1001110101 digitized modulate DAC
symbols

NOISY & DISTORTING ANALOG CHANNEL

demodulate sample&
ADC &filter threshold
1001110101
codeword
bitsout

bitratesamples

6.02 Fall 2012 Lecture 9, Slide #8


FromBasebandtoModulatedSignal,and
Back
codeword
bitsin
generate x[n]
1001110101 digitized modulate DAC
symbols

NOISY & DISTORTING ANALOG CHANNEL

demodulate sample&
ADC &filter threshold
1001110101
y[n]
,PDJHE\0,72SHQ&RXUVH:DUH codeword
bitsout

6.02 Fall 2012 Lecture 8, Slide #3


Modulation(attheTransmitter)
Adaptsthedigitizedsignalx[n]tothecharacteristicsof
thechannel.
e.g.,Acousticchannelfromlaptopspeakertomicrophoneis
notwellsuitedtotransmittingconstantlevels v 0 andv
1 to
represent0and1.Soinsteadtransmitsinusoidal
pressurewavesignalsproportionaltospeakervoltages
v cos(2 f t) and
0 c v1 cos(2 fc t)
where f c isthecarrier frequency(e.g.,2kHz;wavelength
at340m/s =17cm,comparablewithspeakerdimensions)and

v0 = 0 v1 = V > 0 (onoffor
amplitude keying)
oralternatively
v = V v = V > 0(bipolaror
0 1
phase-shiftkeying)
Couldalsokeythefrequency.
6.02 Fall 2012 Lecture 9, Slide #10
FromBrantRocktower,radioagewassparked
ByCarolynY.Johnson,GlobeStaff| July30,2006

MARSHFIELD,MA Acenturyago,radiopioneer
ReginaldA.Fessendenusedamassive420footradio
towerthatdwar fedBrantRocktosendvoiceandmusicto
shipsalongtheAtlanticcoast,inwhathasbecomeknown
astheworld'sfirstvoiceradiobroadcast.Thisweek,
Marshfieldwilllayclaimtoitslittleknownradioheritage
withathreedayextravaganzatocelebratethefeat
includingpilgrimagestothebaseofthelongdismantled
tower,acocktailtobenamedtheFessendenFizz,anda
dramaticreenactmentofthehistoricmoment,called
``MiracleatBrantRock.

AmplitudeModulation(AM)

6.02 Fall 2012 Lecture 9, Slide #11


codeword
Modulation
bitsin
generate x[n]
1001110101 digitized
symbols
t[n]
c=2/16
cos(cn)
16samplespercycle

1001110101

6.02 Fall 2012 Lecture


L cture
Le t 9, Slide #13
IdeasforDemodulation
Foronoffkeying,itsufficestodetectwhentheres
signalandwhenthereisnt,sincewereonlytrying
todistinguish v0 = 0 v1 = V > 0

Manywaystodothat,e.g.,takeabsolutevalueand
thenlocalaverageoverhalfperiodofcarrier

Forbipolarkeying,weneedthesign:

v0 = V v1 = V > 0

6.02 Fall 2012 Lecture 9, Slide #14


Assumingno Demodulation
distortionor
noiseon
channel,so
t[n] z[n]
whatwas
transmitted
isreceived cos(cn)

z[n] = t[n]cos( c n)

z[n] = x[n]cos(c n) cos(c n)

z[n] = 0.5x[n](1+ cos(2c n))

z[n] = 0.5x[n] + 0.5x[n]cos(2 c n)

6.02 Fall 2012 Lecture 9, Slide #15


codeword
Demodulation
bitsin
generate
1001110101 digitized modulate DAC
symbols

x[n] t[n]
NOISY & DISTORTING ANALOG CHANNEL c=2/16

t[n] 16samplespercycle
z[n]
ADC demodulate

z[n]

1001110101

6.02 Fall 2012 Lecture


Lecttu
urre 9,
9, Slide
Slide #16
Averagingfilter
codeword
bitsin
generate
1001110101 digitized modulate DAC
symbols
x[n] t[n]
c=2/16
NOISY & DISTORTING ANALOG CHANNEL

t[n] z[n] 16samplespercycle


r[n]
ADC demodulate filter

r[n]=z[n]++z[nL],L+1lengthoftheaveragingfilter

ForL+1=8,2ccomponentisat2/8,whichis8samplespercycle

So,the2ccomponentgetsaveragedout

Attransitions,thereisabitofdegradation,butwemakedecisionson
themiddlesamples

6.02 Fall 2012 Lecture 9, Slide #17


Filtering:Removingthe2
ccomponent
codeword
bitsin
generate
1001110101 digitized modulate DAC
symbols
x[n] t[n]
c=2/16
NOISY & DISTORTING ANALOG CHANNEL

t[n] z[n] 16samplespercycle


r[n]
ADC demodulate filter

r[n]

1001110101

6.02
.02 Fall 2012 Lecture
Lecturre 9,
9 Slide #18
ModelingChannelBehavior
codeword
bitsin
generate x[n]
1001110101 digitized modulate DAC
symbols

NOISY & DISTORTING ANALOG CHANNEL

demodulate sample&
ADC &filter threshold
1001110101
y[n]
codeword
bitsout

6.02 Fall 2012 Lecture 9, Slide #19


TransmissionoveraChannel
Signalx[n]fromdigitizedsymbolsattransmitter

Distortednoisefreesignaly[n]atreceiver

6.02 Fall 2012 Lecture 9, Slide #20


SystemInputandOutput
input response

x[n] S y[n]

Adiscretetimesignalsuchasx[n]ory[n]isdescribedby
aninfinitesequenceofvalues,i.e.,thetimeindexntakes
valuesinto+.Theabovepictureisasnapshotata
particulartimen.

Inthediagramabove,thesequenceofoutputvaluesy[.]is
theresponseofsystemStotheinputsequencex[.]

Question:WhydidntIwrite:

Inthediagramabove,thesequenceofoutputvaluesy[n]is
theresponseofsystemStotheinputsequencex[n] ??

6.02 Fall 2012 Lecture 9, Slide #21

Notation,Notation!
Wewanttobeclear,butbeingoverlyexplicitaboutthingsleads
toalotofnotationalclutter.Sowetakeshortcutsandliberties,
abusingandoverloadingthenotation,inthehopethatcontext
andotherfactorswillmakeourmeaningclear.
Butpoornotationcanalsoimpede,mislead,confuse!Soonehas
drawthelinecarefully.

Example:ourhardworkingdiscretetimeindexn
(incontinuoustime,itst).Specifically,x[n]candenote
(a) thevalueofthesignalxataparticulartimen
(b) thesequenceofvaluesforninto+,i.e.,the
entiresignalx.
For(b),itsoftenclearertowritex[.]orjustxparticularlyif
therearemultiplesignalsinvolved,becausethesamedummy
indexnshouldntbeusedforboth.

Ontheotherhand,ifyouwanttousex[n]foraspecificvalueof
time,itssometimesclearertowritex[n0]
6.02 Fall 2012 Lecture 9, Slide #22
UnitStep
Asimplebutusefuldiscretetimesignalistheunit step
signalorfunction,u[n],definedas

0, n < 0
u[ n] =
1, n 0

6.02 Fall 2012 Lecture 9, Slide #23


UnitSample
Anothersimplebutusefuldiscretetimesignalistheunit
sample signalorfunction,[n],definedas
0, n 0
[n] = u[n] u[n 1] =
1, n = 0

6.02 Fall 2012 Lecture 9, Slide #24


UnitSample
Anothersimplebutusefuldiscretetimesignalistheunit
samplesignalorfunction,[n],definedas
0, n 0
[n] = u[n] u[n 1] =
1, n = 0

Notethatstandardalgebraicoperationsonsignals
(e.g.subtraction,addition,scalingbyaconstant)
aredefinedintheobviousway,instantbyinstant.
6.02 Fall 2012 Lecture 9, Slide #25
UnitSampleandUnitStepResponses

Unitsample Unitsampleresponse

[n] S h[n]

Theunit sample response ofasystemSistheresponseof


thesystemtotheunitsampleinput. Wewillalways
denotetheunitsampleresponseash[n].

Similarly,theunit step response s[n]:

Unitstep Unitstepresponse

u[n] S s[n]

6.02 Fall 2012 Lecture 9, Slide #26


UnitSample
Decomposition

Adiscretetimesignalcanbedecomposed
intoasumoftimeshifted,scaledunit
samples.

Example:inthefigure,x[n]isthesumof
x[2][n+2]+x[1][n+1]++x[2][n2].

Ingeneral:


x[n] = x[k][n k]
k=

Foranyparticularindex,only
l i
onetermofthissumisnonzero
6.02 Fall 2012 Lecture 9, Slide #27
UnitStep
Decomposition
Digitalsignalingwaveformsare
D
e
easilydecomposedintotime
s
shifted,scaledunitsteps(each
t
transitioncorrespondstoanother
s
shifted,scaledunitstep).

I
Inthisexample,x[n]isthe
t
transmissionof1001110using4
s
samples/bit:
x = u[n] u[n 4]+ u[n 12] u[n 24]
x[n]

6.02 Fall 2012 Lecture 9, Slide #28


TimeInvariantSystems

Lety[n]betheresponseofStoinputx[n].

Ifforallpossiblesequencesx[n]andintegersN



x[nN] S y[nN]


thensystemSissaidtobetime invariant(TI). Atime
shiftintheinputsequencetoSresultsinanidentical
timeshiftoftheoutputsequence.

Inparticular,foraTIsystem,ashiftedunitsample
[n N ] attheinputgeneratesanidentically
function
[n N ] attheoutput.
shiftedunitsampleresponseh
6.02 Fall 2012 Lecture 9, Slide #29
LinearSystems

Lety1[n]betheresponseofStoanarbitraryinputx1[n]
andy2[n]betheresponsetoanarbitraryx2[n].

If,forarbitraryscalarcoefficientsa andb, wehave:


ax [n]+ bx [n]
1 2 S ay1[n]+ by2 [n]


thensystemSissaidtobelinear. Iftheinputisthe
weightedsumofseveralsignals,theresponseisthe
superposition(i.e.,weightedsum)oftheresponseto
thosesignals.

Onekeyconsequence:Iftheinputisidentically0fora
linearsystem,theoutputmustalsobeidentically0.
6.02 Fall 2012 Lecture 9, Slide #30
OurfocuswillbeonLTIModels
LTI=Linearand TimeInvariant

Gooddescriptionoftimeinvariantsystemsfor
smalldeviationsfromanominaloperating
equilibrium

Lotsofstructure,detailedanalysispossible,
amenabletodevelopmentofgoodcomputational
tools,

Majorarenaforengineeringdesign

6.02 Fall 2012 Lecture 9, Slide #31


MITOpenCourseWare
http://ocw.mit.edu

Introduction to EECS II: Digital Communication Systems


Fall201

ForinformationaboutcitingthesematerialsorourTermsofUse,visit:http://ocw.mit.edu/terms.

You might also like