You are on page 1of 16

Chapter2

Problemm2.1
Usethebuiltinconv( ) functionofMATLABtocomputetheconvolutioninProblem2.35(a)onpage
79ofDTSP.HandinaMATLABscriptthatmakesst em( ) plotsoftheinputandoutputofProblem
2.35(a).(Thetimeaxisofyourplotshouldbecorrectlylabeled.)
Problemm2.2
WehaveshownthatiftheinputandoutputofacausalLTIsystemsatisfythedifferenceequation
y[n]=ay[n1]+x[n],
thentheimpulseresponseofthesystemish[n]=a
n
u[n].
(a)Forwhatvaluesofaisthissystemstable?
(b)ConsideracausalLTIsystemforwhichtheinputandoutputarerelatedbythedifferenceequation
y[n]=ay[n1]+x[n]a
N
x[nN],
whereNisapositiveinteger.Determineandsketchaplotoftheimpulseresponseofthissystem.Hint:
Uselinearityandtimeinvariancetosimplifythesolution.
(c)Isthesysteminpart(b)anFIRoranIIRsystem?Explain.
(d)Forwhatvaluesofaisthesysteminpart(b)stable?Explain.
(e)WritetwobriefMATLABprograms(reallyjustsinglestatements)thatimplementthesysteminpart
(b)fora=0.8.Oneprogramshouldusef i l t er ( ) andtheothershoulduseconv( ) .Testyour
programswithanimpulseinputtoverifythattheyproducethesameimpulseresponse.
Problemm2.3
Usethebuiltinfunctionsf i l t er ( ) andf r eqz( ) ofMATLABtocompute51samplesoftheimpulse
responseandfrequencyresponseofthesystemdefinedbythedifferenceequation
y[n]=1.7163y[n 1] 1.1724y[n 2]+0.2089y[n 3]
+0.5264x[n]1.5224x[n 1]+1.5224x[n 2] 0.5264x[n 3].
(Tocomputetheimpulseresponse,makeaninputvectorforf i l t er ( ) consistingofoneunitsample
followedby50zerosamples.)Handina st em( ) plotoftheimpulseresponse.Usesubpl ot ( ) and
pl ot ( ) tomakeatwopanelplotofthemagnitudeandphaseofthefrequencyresponse.
Problemm2.4
Usethebuiltinfunctionsf i l t er ( ) andf r eqz( ) ofMATLABtocompute51samplesoftheimpulse
responseandfrequencyresponseofthesystemdefinedbythedifferenceequation
y[n]=1.556y[n1]1.272y[n2]+0.398y[n3]+0.0798x[n]+x[n1]+x[n2]+x[n3].
(Tocomputetheimpulseresponse,makeaninputvectorforf i l t er ( ) consistingofoneunitsample
followedby50zerosamples.)Handinast em( ) plotoftheimpulseresponse.Usesubpl ot ( ) and
pl ot ( ) tomakeatwopanelplotofthemagnitudeandphaseofthefrequencyresponse.
Chapter3
Problemm3.1
ConsideranLTIrunningsumsystemdefinedbythedifferenceequation

=
=
1
0
] [ ] [
L
k
k n x n y
Further,supposethattheinputisaBernoullirandomsignalwhosevaluesareeither+1or1withequal
probability,andwhoseautocorrelationfunctionis
xx
[m]=[m];i.e.,theinput
iswhitenoisewithmeanzeroandaveragepowerequaltounity.
(a) Makeplotsofthefirstorderprobabilitydensityfunctionoftheinputsignal,p
x
(x);the
autocorrelationfunctionoftheinputsignal,xx[m]=[m];and
xx
(e
j
),thepowerspectrumof
theinputsignal.
(b) Determineageneralexpressionfor
yy
[m],theautocorrelationfunctionoftheoutputofthe
system.Sketch
yy
[m]forL=5.
(c) Determineexpressionsforthemeanandaveragepoweroftheoutputsequencey[n].
(d) NowconsiderthecaseL=5andstateallpossiblevaluesthattheoutputsignaly[n]cantakeon.
Also,enumerateallthedifferentwaysthateachoutputvaluecanbeachieved(i.e.,thepattern
ofinputsamplesthatproducesit)andfromthisdeterminetheprobabilityofeachoutput
samplevalue.Plottheprobabilitydensityfunctionoftheoutput,p
y
(y).
(e) Notethaty[n]isthesumofLindependentrandomvariables.Thereforetheprobabilitydensity
functionoftheoutputcanbeobtainedbyconvolutionofthep
x
(x)withitselfL1times.Use
thisfacttocomputetheprobabilitydensitiesforL=2,3,4,5.Howwouldyouranswerchangeif
thesystemwereanLpointmovingaverageinsteadofjustasum?
(f) WriteaMATLABMfilethatwillcomputetheprobabilitydensityoftheoutputforanyvalueofL.
Yourfunctionshouldhavethefollowingcallingsequence:[ p, x] =out _pr ob( L) ,wherepisa
vectorofprobabilities, xisavectorofcorrespondingamplitudes,andListhelengthofthe
runningsum.Therefore,youshouldbeabletoobtainaplotoftheprobabilitydensitywith
st em( x, p) .HandinaplotforL=5.(Hint:Youcanusetheconv( ) functiontocomputethe
requiredconvolutionsasdiscreteconvolutions.)
(g) Determineaclosedformexpressionforthefrequencyresponseofthesystem.Usethisto
determineanexpressionforthepowerspectrumoftheoutput.PlotthisfunctionforL = 5.
Problemm3.2
WorkProblem3.32(c)onpage145inDTSP.Firstdothepartialfractionexpansionthatisrequiredby
hand.ThencheckyourresultusingtheMATLABfunctionr esi duez( ) .
Problemm3.4
WorkProblem3.33(c)onpage145inDTSP.Firstdothepartialfractionexpansionthatisrequiredby
hand.ThencheckyourresultusingtheMATLABfunctionr esi duez( ) .
Problemm3.5
Downloadthefilel t i sys_demo. zi pandplacethetwofilescontainedthereininthesamedirectory.
Asitissetup,theprogramonlyusesamovingaveragefilter.Iwantyoutolookattheprogramand
figureouthowtodothefollowing:
1. Changetheprograml t i sys_demo. zi pintoanewprogrami i r sys_demo. zi psothatthe
callingsequenceis
f unct i on i i r sys_demo( N, b, a, Ncor r , Nbi ns, Npl t )
%si mul at i on of ber noul i r andompr ocess wi t h i i r f i l t er i ng
%USAGE i i r sys_demo( N, b, a, Ncor r , Nbi ns, Npl t )
%N=Number of sampl es of i nput and out put
%b=vect or of f eedf or war d or numer at or coef f i ci ent s
%a=vect or of f eedback or denomi nat or coef f i ci ent s
%Ncor r =number of cor r el at i on l ags t o show ( must be even i nt eger )
%Nbi ns=number of hi st ogr ambi ns
%Npl t =number of sampl es of i nput and out put pl ot t ed
Inotherwords,IwantyoutochangetheprogramsothatitcanuseeitheranIIRorFIRsystem.You
shouldonlyhavetochangeafewlinesofcode.
2. Inl t i sys_demo( ) theimpulseresponseisdefinedbythelengthparameterM.ForIIRsystems
youwillhavetoplotonlyafinitelengthportionoftheimpulseresponse.Forconsistency,lets
everyoneplot51samples.Youcanusef i l t er ( ) tocomputetheimpulseresponse.
3.LateintheprogramthereisaplacewhereIsuperimposethetheoreticalautocorrelationofamoving
averagefilter.Takethisout.Youdonotneedtoreplacethiswiththetheoreticalautocorrelationforthe
generalIIRcase.Thatwouldrequirealotofwork.Useyourprogramwithacausalsystemdefinedbythe
systemfunction
1 1
1
8 . 0 5 . 1 1
1
) (

+

=
z z
z
z H .
Handinafullsetofplotsandaprintoutofyourmodifiedprogram.Forconsistency,callyourprogramas
follows:
b=[ ???] ;
a=[ ???] ;
i i r sys_demo( 8000, b, a, 30, 41, 151)
Problemm3.6
WorkProblem2.28onp.75ofDTSP.Thereisanerrorintheproblemstatement.The+signinthe
definingequationshouldbeasign;i.e.,thesequencew[n]shouldbedefined
as
( )
( ) ] [ ) / 2 cos( 1 5 . 0
, 0
0 ) / 2 cos( 1 5 . 0
] [
n r M n
otherwise
M n M n
n w

where


=
otherwise
M n
n r
, 0
0 , 1
] [
Thesequencer[n]isoftencalledarectangularwindow,andthesequencew[n]=0.5[1cos(2n/M)]r[n]
iscommonlyknownastheHanningwindow.TheybothhavemanyusesinDSPincludingfilterdesign
andspectrumanalysis.Itisthereforeworthwhiletoknowtheirpropertiesandcharacteristics.
(d)AlsodeterminethephaseofR(e
j
)andW(e
j
).
(e)UseMATLABtographR(e
j
)andW(e
j
)forforthecaseM=32.Handinonlyyourplotsof
w[n]and20log|W(e
j
)|forgradingalongwithalistofMATLABinstructionsthatyouusedtogenerate
thesetwoplots.
HintsonMATLABSolution
Typehel p hanni ngtofindoutaboutMATLABsbuiltinHanningwindow,orgenerateityourself
directly.Therectangularwindowr[n]canbecreatedinMATLABwiththeMATLABfunctionones( ) or
thefunctionboxcar ( ) .
ThediscretetimeFouriertransform(DTFT)isafunctionofthecontinuousvariable,andacomplete
graphwouldshowallvaluesofW(e
j
)oversomeintervaloflength2.Acomputergeneratedplotmust
necessarilyinvolveonlyafinitenumberofpoints,butifthecurveissmoothenough,thenconnecting
thepointsbystraightlineswillgiveanadequategraphicalrepresentation.Asetofvaluesofthecomplex
DTFTatequallyspacedvaluesoffrequencycanbeobtainedusingtheMATLABfunctionforany
f r eqz( ) polynomialorrationalDTFTfunction.Youcanspecifythenumberoffrequencydomainpoints
(L)asaninputtof r eqz( ) .ExperimenttoseehowlargeLshouldbecomparedtoM=32sothatthe
plotissmooth.
Thefunctionf r eqz( ) willreturnvectorsofboththevaluesoftheDTFTandthefrequenciesatwhich
theDTFTvalueswerecomputed.Usehelpf r eqz( ) tofindouthowtousethisveryusefulMATLAB
function.Notethatthevectoroffrequenciesreturnedbyf r eqz( ) rangesbydefaultfrom0to,but
youcanchangethis.IfHisavectorofvaluesofW(e
jk
),andWisthevectoroffrequenciesk,thentry
pl ot ( W, l og10( abs( H) ) ) orpl ot ( W, db( H, 80) ) ) .(Seehel p db.)
Whathappensifyoutrytoplotthephase?(Seehel p angl e.)
Chapter4
Problemm4.1
Considertheimpulseresponsesequence
h[n]=Ar
n
cos(
0
n+)u[n] (1)
InthisproblemyouwilluseMATLABtostudythepropertiesoftheztransformofthissequence.
(a) Usehel p f i l t er tounderstandhowtheMATLABfunctionf i l t er ( ) works.Now,writea
MATLABmfilethatwillcomputethecoefficientvectorsthatcanbeusedtoimplementthesystem
withimpulseresponsegivenbyEquation(1)usingtheMATLABfunctionf i l t er ( ) .Yournewm
fileshouldhavethecallingsequence[ b, a] =coef ( A, r , omega0, phi ) .(Todothisyoumayuse
theresultsofProblem3.6onpage139ofDTSPtodeterminethenecessaryvaluesofthecoefficients
inH(z)=B(z)/A(z).)Handinacopyofyourmfile.
(b) Demonstratethatyourmfileworksbyusingthenewfunctioncoef ( ) inconjunctionwiththe
MATLABfunctionf i l t er ( ) tocomputetheimpulseresponseh[n]overtheinterval0n50for
A=20,r=.975,
0
=2/17,and=/4.Compareyourresulttowhatyouobtainbydirect
evaluationofEquation(1)for0n50.Thefunctioni mpl ot ( ) canbeusedtoplotthetwo
sequences.Handinaprintoutofthenumeratoranddenominatorcoefficientsusedandaplotof
thesesequenceseithersuperimposedonthesamegraphwithdifferentsymbolsorinatwopart
plotusingthesubpl ot ( ) function.
(c) Nowusehel p r esi dueztofindoutaboutthemfilethatdoespartialfractionexpansionsinthe
formusedinthetext.Usingthisfunctionandtheparametervaluesofpart(b),obtainapartial
fractionexpansionforH(z)intermsoffirstorderterms.WriteouttheequationforH(z)inthisform
andhanditinaspartofyoursolution.Usethevalues[ R, P, K] returnedbythefunction
[R, P, K] =r esi duez( b, a) tocomputeandplottheimpulseresponseovertheinterval0n50,
andcompareittotheresultsofpart(b).WritedowntheMATLABstatement(s)thatdothis
computationtohandinassolutiontothispart.
(d) Nowusehel p f r eqztorefreshyourmemoryaboutthisfunctionthatcanbeusedtocomputethe
frequencyresponseofasystemwitharationalsystemfunction.Usethisfunctiontorelatethe
location,amplitude,and3dBbandwidthoftheresonancepeaktothepoleradiusrandangle
0
.
ForA=20,
0
=2/17,and=0,computethefrequencyresponseatN=501pointsforeachofthe
casesr=0.9,r=0.95,r=0.975,andr=0.99.Foreachcaseplotthemagnitudeorlogmagnitudeof
thefrequencyresponse.(Youmayfinditusefultoplotallthecasesonthesamegraph.TheMATLAB
functiondb( ) isusefulforthis.)Fromyourcomputation,fillintheentriesinthefollowingtable
andhanditin.
Radiusr 1r Peaklocation Peakheight 3dBbandwidth
0.9
0.95
0.975
0.99
Whatistherelationshipbetweenandr?Isthepeakalwaysat
0
?
(e) Now,youshouldtryholdingr=.975,A=20,and=0whilevarying
0
.Plotthefrequencyresponse
forseveralcasesincluding
0
=2/17and
0
=2(18)/17.Howdoyouaccountfortheresultof
theselattertwocases?
Chapter5
Problemm5.1
ConsiderthesystemofExample5.8,p.299inDTSP.Itssystemfunctionis
( )( )
( )( )
2 1 1
2 1 ) 1
7957 . 0 4461 . 1 1 683 . 0 1
0166 . 1 1 1 05634 . 0
) (


+
+ +
=
z z z
z z z
z H
TogetfamiliarwithMATLABstoolsforanalyzingsystems,IwantyoutoduplicatetheplotsinExample
5.8.(Youwillneedthegr pdel ay( ) functionintheMATLABSignalProcessingToolboxtodothis
problemcompletely.)
Usingtheconv( ) function,youcanexpandthenumeratoranddenominatorintopolynomialsof
degree3.Thesewillbeyouraandbvectorsforuseinthefollowingparts.
(a) Usethefunctionr oot s( ) todeterminethepoleandzerolocations.
(b) Usethefunctionsf r eqz( ) andgr pdel ay( ) toduplicatetheplotofFigure5.15.
(c) Usethefunctionzpl ane( ) toduplicatetheplotofFigure5.14.
(d) Usethefunctionr esi duez( ) tomakeapartialfractionexpansionofH(z).
(e) Drawthesignalflowgraphofanimplementationofthesystemasacascadeofasecondorder
systemwithafirstordersystem.Markthevaluesofthecoefficientsontheflowgraph.
(f) DrawadirectformIIsignalflowgraphforimplementingthesystem.Usetheexpandedpolynomials
aandbtoobtainthevaluesofthecoefficientsontheflowgraph.
(g) DrawthetransposeddirectformIIimplementationflowgraph.
(h) Usetheresultsofthepartialfractionexpansiontodrawthesignalflowgraphofanimplementation
ofthesystemasaparallelcombinationofasecondordersystem,afirstordersystem,anda
constantgain.Markthevaluesofthecoefficientsontheflowgraph.
Problemm5.2
Acausallineartimeinvariantsystemhassystemfunction
( )( )
( )( )
2 - 1 - 1 -
-2 -1 -1
0.7957z + 1.4461z - 1 0.683z - 1
z + 1.0166z - 1 z + 1 0.05634
) ( = z H
(a) Determinethepolesandzerosofthissystemfunctionandplottheminthezplane.Whatisthe
regionofconvergenceforthissystemfunction?
(b) Fromthepolezeroplot,sketchthelogmagnitudeofthefrequencyresponseofthissystemas
accuratelyasispossible.
(c) UsetheMATLABconv( ) toobtainthirdorderpolynomialsforthenumeratoranddenominator
andthenusef r eqz( ) toevaluateandplotthelogmagnitudeofthefrequencyresponse.
Comparetoyouranswerinpart(b).
Chapter6
Problemm6.1
AcausalLTIdiscretetimesystemhassystemfunction
( )( )
( )( )
1 1
1 1
1 1
1 1
) (




=
z e r z e r
z e r z e r
K z H
j
p
j
p
j
z
j
z

whereK=0.95,r
z
=1,r
p
=0.99995,and=0.4.Thissettingoftheparameterswillproduceavery
sharpnotchfilterfrequencyresponse.
(a) Atwhatfrequencywillthenotchinthefrequencyresponseoccurifinfiniteprecisioncoefficientsare
usedintheimplementation?
(b) UseMATLABoryourcalculatortodeterminethecoefficientsoftheDirectFormIIimplementation
ofthissystem;i.e.,findthecoefficientsin
2
2
1
1
2
2
1
1 0
1
) (



+ +
=
z a z a
z b z b b
z H .
(a) ObtaintheQ15representationofthesecoefficients.Giveboththeintegerformandtheequivalent
fractionalformandcomparethefractionalformtotheunquantizedcoefficients.AlsouseMATLABs
r oot s( ) functiontofindthezerosandpolesoftheQ15quantizedsystem.Comparethemtothe
unquantizedzerosandpoles.Istheunquantizedsystemstable?IstheQ15systemstable?Atwhat
frequencywillthenotchinthefrequencyresponseoccurfortheQ15system?
(b) ObtaintheQ12representationofthesecoefficients.Giveboththeintegerformandtheequivalent
fractionalformandcomparethefractionalformtotheunquantizedcoefficients.AlsouseMATLABs
r oot s( ) functiontofindthezerosandpolesoftheQ12quantizedsystem.Comparethemtothe
unquantizedzerosandpoles.Istheunquantizedsystemstable?IstheQ12systemstable?
Chapter7
Problemm7.1
WorkProblem7.12ofDTSP.UsetheMATLABfunctionsconv( ) andr esi duez( ) tosolveparts(b)
and(c).
Problemm7.2
Considerthefollowingsystemforinterpolatingsequencex[n]x
c
(nT)toasamplingratethatisL=16
timesashighastheinputsamplingrate.

Assumethattheinputsequencex[n]=x
c
(nT)wasobtainedbysamplingabandlimitedcontinuoustime
signalwhoseFouriertransformsatisfiesthefollowingcondition:|X
c
(j)|=0for||(3600).Assume
thattheoriginalsamplingperiodwasT=1/8000.
(a) MakeasketchoftheFouriertransformX
c
(j)ofatypicalbandlimitedinputsignalandthe
correspondingdiscretetimeFouriertransformsX(e
j
)andX
e
(e
j
).
(b) DetermineandplotcarefullyanideallowpassfilterH(e
j
)thatwillexactlyinterpolatethe
upsampledsignaltogivey[n]=x
c
(nT/16).
(c) Inordertoimplementtheinterpolationsystem,wemust,ofcourse,useanonidealfilter.Useyour
plotofX
e
(e
j
)obtainedinpart(a)todeterminethepassbandandstopbandcutofffrequencies(
p

and
s
)requiredtopreservetheoriginalbandoffrequenciesmoreorlessunmodifiedwhile
significantlyattenuatingtheimagesofthebasebandspectrum.Assumingthatthiscanbeachieved
withpassbandapproximationerror
1
=.01(forfilterpassbandgainof1)andstopband
approximationerror
2
=.01,plotthespecificationsforthedesignofthefilterH(e
j
)for.
(d) UseMATLABsel l i por d( ) andel l i p( ) functionstodesignanIIRellipticfilterthatmeetsthe
specifications.Initially,designthefiltertohaveunitygaininthepassband.Plotitsfrequency
response(indB)vs/for00.1.Usehol d ontokeepthisplottoaddsubsequentplotsto
it.Makeaprintoutofthecoefficientsofthefilter.Howwouldyoumodifythecoefficientsofthe
ellipticfiltertochangethegainto16asrequiredforinterpolationby16?
(e) UseMATLABskai ser or d( ) andf i r 1( ) functionstodesignanFIRfilterthatmeetsthe
specifications.Addaplotofitsfrequencyresponse(indB)vs/for00.1totheplot
obtainedin(d).Designthefilterassumingpassbandgainofunity.Howwouldyoumodifythefilter
coefficientstochangethegainto16?
(f) UseMATLABsr emezor d( ) andr emez( ) functionstodesignanFIRfilterthatmeetsthe
specifications.Designthefilterassumingpassbandgainofunity.Howwouldyoumodifythefilter
coefficientstochangethegainto16?Addaplotofitsfrequencyresponse(indB)vs/for0
0.1totheplotobtainedin(d).Handinacopyofyourplotshowingthethreefrequencyresponses
superimposed.
(g) Discusstherelativeadvantagesofthethreefiltertypesforthisapplication.Whichfiltercouldbe
implementedwiththeleastnumberofmultiplicationsandadditions?
Problemm7.3
Considerthefollowingsystemforinterpolatingsequencex[n]x
c
(nT)toasamplingratethatisL=15
timesashighastheinputsamplingrate.

Assumethattheinputsequencex[n]=x
c
(nT)wasobtainedbysamplingabandlimitedcontinuoustime
signalwhoseFouriertransformsatisfiesthefollowingcondition:|X
c
(j)|=0for||(3600).Assume
thattheoriginalsamplingperiodwasT=1/8000.
(a) MakeasketchoftheFouriertransformX
c
(j)ofatypicalbandlimitedinputsignalandthe
correspondingdiscretetimeFouriertransformsX(e
j
)andX
e
(e
j
).
(b) DetermineandplotcarefullyanideallowpassfilterH(e
j
)thatwillexactlyinterpolatethe
upsampledsignaltogivey[n]=x
c
(nT/15).
(c) Inordertoimplementtheinterpolationsystem,wemust,ofcourse,useanonidealfilter.Useyour
plotofX
e
(e
j
)obtainedinpart(a)todeterminethepassbandandstopbandcutofffrequencies(
p

and
s
)requiredtopreservetheoriginalbandoffrequenciesmoreorlessunmodifiedwhile
significantlyattenuatingtheimagesofthebasebandspectrum.Assumingthatthiscanbeachieved
withpassbandapproximationerror
1
=.01(forfilterpassbandgainof1)andstopband
approximationerror
2
=.01,plotthespecificationsforthedesignofthefilterH(e
j
)for.
(d) UseMATLABsel l i por d( ) andel l i p( ) functionstodesignanIIRellipticfilterthatmeetsthe
specifications.Initially,designthefiltertohaveunitygaininthepassband.Plotitsfrequency
response(indB)vs/for00.1.Usehol d ontokeepthisplottoaddsubsequentplotsto
it.Makeaprintoutofthecoefficientsofthefilter.Howwouldyoumodifythecoefficientsofthe
ellipticfiltertochangethegainto15asrequiredforinterpolationby15?
(e) UseMATLABskai ser or d( ) andf i r 1( ) functionstodesignanFIRfilterthatmeetsthe
specifications.Addaplotofitsfrequencyresponse(indB)vs/for00.1totheplot
obtainedin(d).Designthefilterassumingpassbandgainofunity.Howwouldyoumodifythefilter
coefficientstochangethegainto15?
(f) UseMATLABsr emezor d( ) andr emez( ) functionstodesignanFIRfilterthatmeetsthe
specifications.Designthefilterassumingpassbandgainofunity.Howwouldyoumodifythefilter
coefficientstochangethegainto15?Addaplotofitsfrequencyresponse(indB)vs/for0
0.1totheplotobtainedin(d).Handinacopyofyourplotshowingthethreefrequencyresponses
superimposed.
(g) Discusstherelativeadvantagesofthethreefiltertypesforthisapplication.Whichfiltercouldbe
implementedwiththeleastnumberofmultiplicationsandadditions?
Chapter8
Problemm8.1
Considerthefollowingtwostagesystemforinterpolatingsequencex[n]=x
c
(nT)toasamplingratethat
is25timesashighastheinputsamplingrate;i.e.,wedesirey[n]=x
c
(nT/25).

Assumethattheinputsequencex[n]=x
c
(nT)wasobtainedbysamplingabandlimitedcontinuoustime
signalwhoseFouriertransformsatisfiesthefollowingcondition:|X
c
(j)|=0for||2(1800).
AssumethattheoriginalsamplingperiodwasT=1/4000.
(a) MakeasketchoftheFouriertransformX
c
(j)ofatypicalbandlimitedinputsignalandthe
correspondingdiscretetimeFouriertransformsX(e
j
)andX
e
(e
j
).
(b) Inordertoimplementtheinterpolationsystem,wemust,ofcourse,usenonidealfilters.Useyour
plotofX
e
(e
j
)obtainedinpart(a)todeterminethepassbandandstopbandcutofffrequencies(
p1

and
s1
)requiredtopreservetheoriginalbandoffrequenciesmoreorlessunmodifiedwhile
significantlyattenuatingtheimagesofthebasebandspectrum.(Thatis,wedesirethatw[n]
x
c
(nT/5).)Assumingthatthiscanbeachievedwithpassbandapproximationerror
1
=.005(forfilter
passbandgainof1)andstopbandapproximationerror
2
=.01,plotthespecificationsforthedesign
ofthefilterH1(e
j
)for.
(c) NowmakeasketchofW
e
(e
j
)[assumingw[n]=x
c
(nT/5)]anduseittodeterminethepassbandand
stopbandcutofffrequencies
p2
and
s2
requiredforthesecondfilter.
(d) UsetheformulaofEq.(7.117)onpage570ofDTSPtodeterminethefilterordersM
1
andM
2
of
ParksMcClellanfiltersthathavethepassbandandstopbandcutofffrequenciesdeterminedinparts
(b)and(c)with
1
=0.005and
2
=0.01forbothfilters.
(e) Howmanymultipliesarerequiredtocompute25samplesoftheoutputforthiscase.

Chapter9
Problemm9.1
Solveproblem9.55onpage787inDTSP.WriteasimpleMATLABfunctiontoimplementyourmethod.
Yourfunctionshouldhavethefollowingcallingsequence:
f unct i on H=f r eekz( b, a, N)
%USAGE H=f r eekz( b, a, N)
%H=z- t r ansf or meval uat ed on uni t ci r cl e at N poi nt s
%b=vect or of numer at or coef f i ci ent s
%a=vect or of denomi nat or coef f i ci ent s
%N=number of equal l y spaced f r equenci es t o eval uat e
Thefirstapproachthatyoumightthinkofwouldbetousethedifferenceequationtocomputethe
impulseresponse.ThenusetheFFTtoevaluateH(e
j(2/512)k)
.However,thiswillnotworkwellifthe
systemisanIIRsystem.TrytothinkofanotherapproachthatmightinvolvemorethanoneFFT
calculationplussomeothercomputation.
Chapter10
Problemm10.1
Solveproblem10.23onpage869inDTSP.Firstsketchhowyouthinkthespectrogramwilllook.Then
usetheMATLABfunctionspect r ogr am( ) inthesignalprocessingtoolboxtomakeanimageofthe
spectrograminMATLAB.Hint:Usethe yaxi s optiontodisplayinatraditionalwayforDSPwork.

You might also like