You are on page 1of 9

2/22/2016

Understandingcascadedintegratorcombfilters
designreuse.cn | drembedded.com |

SEARCH IP

NEWS

INDUSTRY ARTICLES

BLOGS

VIDEOS

SLIDES

Login |

EVENTS

SubscribetoD&RSoCNewsAlert

SearchIndustryArticles

Understanding cascaded integrator-comb filters

Understandingcascadedintegratorcomb
filters
ByRichardLyons,CourtesyofEmbeddedSystemsProgramming
Mar312005(14:49PM)
URL:http://www.embedded.com/showArticle.jhtml?
articleID=160400592
ThepreviouslyobscureCICfilterisnowvitaltomanyhighvolume
wirelesscommunicationstasksandequipment.UsingCICfilterscan
cutcosts,improvereliability,andhelpperformance.Here'saprimer
togetyoustarted.

SEARCH SILICON IP
16,000IPCoresfrom450Vendors

EnterKeywords....

RELATED ARTICLES
Revisitingtheanaloguevideodecoder:
Brushinguponyourcombfilters

Cascadedintegratorcomb(CIC)digitalfiltersarecomputationallyefficient
implementationsofnarrowbandlowpassfiltersandareoftenembeddedin
hardwareimplementationsofdecimationandinterpolationinmodern
communicationssystems.CICfilterswereintroducedtothesignal
processingcommunity,byEugeneHogenauer,morethantwodecadesago,

UnderstandingPeakFloatingPoint
PerformanceCalculations
UnderstandinglayersintheJESD204B
specification:AhighspeedADC
perspective,Part2

buttheirapplicationpossibilitieshavegrowninrecentyears.1
Improvementsinchiptechnology,theincreaseduseofpolyphasefiltering
techniques,advancesindeltasigmaconverterimplementations,andthe
significantgrowthinwirelesscommunicationshaveallspurredmuch
interestinCICfilters.
Whilethebehaviorandimplementationofthesefiltersisn'tcomplicated,
theircoveragehasbeenscarceintheliteratureofembeddedsystems.
Thisarticleattemptstoaugmentthebodyofliteratureforembedded
systemsengineers.AfterdescribingafewapplicationsforCICfilters,I'll
introducetheirstructureandbehavior,presentthefrequencydomain
performanceofCICfilters,anddiscussseveralimportantpracticalissues
inbuildingthesefilters.

UnderstandinglayersintheJESD204B
specification:AhighspeedADC
perspective,Part1
UnderstandingsigmadeltaADCs:Anon
mathematicalapproach

NEW ARTICLES
SecuringtheMostImportantGoalofUSB
TypeCTechnology:ABetterUser
Experience

CICfilterapplications

FromHereto5G:ARoadmapof
Challenges

CICfiltersarewellsuitedforantialiasingfilteringpriortodecimation
(sampleratereduction),asshowninFigure1aandforantiimaging
filteringforinterpolatedsignals(samplerateincrease)asinFigure1b.
Bothapplicationsareassociatedwithveryhighdataratefiltering,such
ashardwarequadraturemodulationanddemodulationinmodernwireless
systemsanddeltasigmaA/DandD/Aconverters.

LeveragealwaysonvoicetriggerIPto
reachultralowpowerconsumptionin
voicecontrolleddevices
DeliveringHighQualityAnalogVideo
SignalsWithOptimizedVideoDACs
CustomASICsforInternetofIndustrial
Things(IoIT)

See New Articles >>

Figure1:CICfilterapplications
Viewfullsizedimage
Becausetheirfrequencymagnituderesponseenvelopesaresin(x)/xlike,
CICfiltersaretypicallyeitherfollowedorprecededbyhigherperformance
linearphaselowpasstappeddelaylineFIRfilterswhosetasksareto
compensatefortheCICfilter'snonflatpassband.Thatcascadedfilter
architecturehasvaluablebenefits.Forexample,withdecimation,youcan
greatlyreducecomputationalcomplexityofnarrowbandlowpassfiltering
comparedwithifyou'dusedasinglelowpassfiniteimpulseresponse(FIR)
filter.Inaddition,thefollowonFIRfilteroperatesatreducedclockrates
minimizingpowerconsumptioninhighspeedhardwareapplications.

MOST POPULAR
1.

DynamicMemoryAllocationand
FragmentationinCandC++

2.

HowtocalculateCPUutilization

3.

UsingSystemVerilogAssertionsinRTL
Code

4.

SecuringtheMostImportantGoalof
USBTypeCTechnology:ABetterUser
Experience

5.

FPGAImplementationofAES
EncryptionandDecryption

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

See the Top 20 >>

1/9

2/22/2016

Understandingcascadedintegratorcombfilters

AcrucialbonusinusingCICfilters,andacharacteristicthatmakesthem
popularinhardwaredevices,isthattheyrequirenomultiplication.The
arithmeticneededtoimplementthesedigitalfiltersisstrictlyadditionsand
subtractionsonly.Withthatsaid,let'sseehowCICfiltersoperate.

EmailThisArticle

PrinterFriendlyPage

Figure2:Dpointaveragingfilters
Viewfullsizedimage

Recursiverunningsumfilter
CICfiltersoriginatefromthenotionofarecursiverunningsumfilter,
whichisitselfanefficientformofanonrecursivemovingaverager.Recall
thestandardDpointmovingaverageprocessinFigure2a.Therewesee
thatD1summations(plusonemultiplyby1/D)arenecessarytocompute
theaverageroutputy(n).
TheDpointmovingaveragefilter'soutputintimeisexpressedas:

Equation1
wherenisourtimedomainindex.Thezdomainexpressionforthis
movingaverageris:

Equation2
whileitszdomainH(z)transferfunctionis:

Equation3
Iprovidetheseequationsnottomakethingscomplicated,butbecause
they'reuseful.Equation1tellsushowtobuildamovingaverager,and
Equation3isintheformusedbycommercialsignalprocessingsoftwareto
modelthefrequencydomainbehaviorofthemovingaverager.
ThenextstepinourjourneytowardunderstandingCICfiltersistoconsider
anequivalentformofthemovingaverager,therecursiverunningsum
filterdepictedinFigure2b.Thereweseethatthecurrentinputsamplex(n)
isadded,andtheoldestinputsamplex(nD)issubtractedfromthe
previousoutputaveragey(n1).It'scalled"recursive"becauseithas
feedback.Eachfilteroutputsampleisretainedandusedtocomputethe
nextoutputvalue.Therecursiverunningsumfilter'sdifferenceequation
is:

Equation4
havingazdomainH(z)transferfunctionof:

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

2/9

2/22/2016

Understandingcascadedintegratorcombfilters

Equation5
WeusethesameH(z)variableforthetransferfunctionsofthemoving
averagefilterandtherecursiverunningsumfilterbecausetheirtransfer
functionsareequaltoeachother!It'strue.Equation3isthenonrecursive
expressionandEquation5istherecursiveexpressionforaDpoint
averager.Themathematicalproofofthiscanbefoundinmybookon
digitalsignalprocessing,butshortlyI'lldemonstratethatequivalencywith
anexample.2
Here'swhywecareaboutrecursiverunningsumfilters:thestandard
movingaveragerinFigure2amustperformD1additionsperoutput
sample.Therecursiverunningsumfilterhasthesweetadvantagethat
onlyoneadditionandonesubtractionarerequiredperoutputsample,
regardlessofthedelaylengthD.Thiscomputationalefficiencymakesthe
recursiverunningsumfilterattractiveinmanyapplicationsseekingnoise
reductionthroughaveraging.Nextwe'llseehowaCICfilteris,itself,a
recursiverunningsumfilter.

CICfilterstructures
Ifwecondensethedelaylinerepresentationandignorethe1/Dscalingin
Figure2bweobtaintheclassicformofa1storderCICfilter,whose
cascadestructureisshowninFigure2c.ThefeedforwardportionoftheCIC
filteriscalledthecombsection,whosedifferentialdelayisD,whilethe
feedbacksectionistypicallycalledanintegrator.Thecombstagesubtracts
adelayedinputsamplefromthecurrentinputsample,andtheintegrator
issimplyanaccumulator.TheCICfilter'sdifferenceequationis:

Equation6
anditszdomaintransferfunctionis:

Equation7

Figure3:SinglestageCICfiltertimedomainresponseswhenD=5
Viewfullsizedimage
ToseewhytheCICfilterisofinterest,firstweexamineitstimedomain
behavior,forD=5,showninFigure3.Ifaunitimpulsesequence,a
unityvaluedsamplefollowedbymanyzerovaluedsamples,wasapplied
tothecombstage,thatstage'soutputisasshowninFigure3a.Nowthink,
whatwouldbetheoutputoftheintegratorifitsinputwasthecombstage's
impulseresponse?Theinitialpositiveimpulsefromthecombfilterstarts
theintegrator'sallonesoutput,asinFigure3b.Then,Dsampleslater,the
negativeimpulsefromthecombstagearrivesattheintegratortozeroall
furtherCICfilteroutputsamples.
ThekeyissueisthatthecombinedunitimpulseresponseoftheCICfilter,
beingarectangularsequence,isidenticaltotheunitimpulseresponsesof
amovingaveragefilterandtherecursiverunningsumfilter.(Moving
averagers,recursiverunningsumfilters,andCICfiltersareclosekin.
Theyhavethesamezdomainpole/zerolocations,theirfrequency
magnituderesponseshaveidenticalshapes,theirphaseresponsesare
identical,andtheirtransferfunctionsdifferonlybyaconstantscale
factor.)Ifyouunderstandthetimedomainbehaviorofamovingaverager,
thenyounowunderstandthetimedomainbehavioroftheCICfilterin
Figure2c.

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

3/9

2/22/2016

Understandingcascadedintegratorcombfilters

Figure4:CharacteristicsofasinglestageCICfilterwhenD=5
Viewfullsizedimage
ThefrequencymagnitudeandlinearphaseresponseofaD=5CICfilteris
showninFigure4awherethefrequency sistheinputsignalsamplerate
inHz.
WecanobtainanexpressionfortheCICfilter'sfrequencyresponseby
evaluatingEquation7'sHcic(z)transferfunctiononthezplane'sunitcircle,
bysettingz=ej2,yielding:

Equation8
UsingEuler'sidentity2jsin()=ejej,wecanwrite:

Equation9
IfweignorethephasefactorinEquation9,thatratioofsin()termscanbe
approximatedbyasin(x)/xfunction.ThismeanstheCICfilter'sfrequency
magnituderesponseisapproximatelyequaltoasin(x)/xfunctioncentered
at0HzasweseeinFigure4a.(ThisiswhyCICfiltersaresometimes
calledsincfilters.)
Digitalfilterdesignersliketoseezplanepole/zeroplots,soweprovide
thezplanecharacteristicsofaD=5CICfilterinFigure4c,wherethe
combfilterproducesDzeros,equallyspacedaroundtheunitcircle,and
theintegratorproducesasinglepolecancelingthezeroatz=1.Eachof
thecomb'szeros,beingaDthrootof1,arelocatedatz(m)=ej2m/D,
wherem=0,1,2,...,D1,correspondingtoamagnitudenullinFigure4a.
Thenormallyriskysituationofhavingafilterpoledirectlyontheunit
circleneednottroubleusherebecausethereisnocoefficientquantization
errorinourHcic(z)transferfunction.CICfiltercoefficientsareonesand
canberepresentedwithperfectprecisionwithfixedpointnumberformats.
Althoughrecursive,happilyCICfiltersareguaranteedstable,linearphase
showninFigure4b,andhavefinitelengthimpulseresponses.At0Hz(DC)
thegainofaCICfilterisequaltothecombfilterdelayD.Thisfact,whose
derivationisavailable,willbeimportanttouswhenweactuallyimplement
aCICfilterinhardware.2

Figure5:SinglestageCICfiltersusedindecimationandinterpolation
Viewfullsizedimage
Again,CICfiltersareprimarilyusedforantialiasingfilteringpriorto
decimationandforantiimagingfilteringforinterpolatedsignals.With
thosenotionsinmindweswaptheorderofFigure2c'scomband
integratorwe'repermittedtodosobecausethoseoperationsarelinear
andincludedecimationbyasampleratechangefactorRinFigure5a.(You
maywishtoprovethattheunitimpulseresponseoftheintegrator/comb

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

4/9

2/22/2016

Understandingcascadedintegratorcombfilters

combination,priortothesampleratechange,inFigure5aisequaltothat
inFigure3c.)InmostCICfilterapplicationstheratechangeRisequalto
thecomb'sdifferentialdelayD,butwe'llkeepthemasseparatedesign
parametersfornow.

Figure6:Magnituderesponseofa1storder,D=8,decimatingCIC
filter:beforedecimationaliasiingafterR=8decimation
Viewfullsizedimage
ThedecimationoperationRmeansdiscardallbuteveryRthsample,
resultinginanoutputsamplerateof s,out=s,in/R.ToinvestigateaCIC
filter'sfrequencydomainbehaviorinmoredetail,Figure6ashowsthe
frequencymagnituderesponseofaD=8CICfilterpriortodecimation.
Thespectralband,ofwidthB,centeredat0Hzisthedesiredpassbandof
thefilter.AkeyaspectofCICfiltersisthespectralfoldingthattakesplace
duetodecimation.
ThoseBwidthshadedspectralbandscenteredaboutmultiplesofs,in/Rin
Figure6awillaliasdirectlyintoourdesiredpassbandafterdecimationbyR
=8asshowninFigure6b.Noticehowthelargestaliasedspectral
component,inthisexample,isroughly16dBbelowthepeakofthebandof
interest.OfcoursethealiasedpowerlevelsdependonthebandwidthB
thesmallerBis,thelowerthealiasedenergyafterdecimation.

Figure7:1storder,D=R=8,interpolatingCICfilterspectra:input
spectrumoutputspectralimages
Viewfullsizedimage
Figure5bshowsaCICfilterusedforinterpolationwheretheRsymbol
meansinsertR1zerosbetweeneachx(n)sample,yieldingay(n)output
samplerateof s,out=R s,in.(InthisCICfilterdiscussion,interpolationis
definedaszerosinsertionfollowedbyfiltering.)Figure7ashowsan
arbitrarybasebandspectrum,withitsspectralreplications,ofasignal
appliedtotheD=R=8interpolatingCICfilterofFigure5b.Thefilter's
outputspectruminFigure7bshowshowimperfectfilteringgivesriseto
theundesiredspectralimages.
Afterinterpolation,unwantedimagesoftheBwidthbasebandspectrum
resideatthenullcenters,locatedatintegermultiplesofs,out/R.Ifwe
followtheCICfilterwithatraditionallowpasstappeddelaylineFIRfilter,
whosestopbandincludesthefirstimageband,fairlyhighimagerejection
canbeachieved.

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

5/9

2/22/2016

Understandingcascadedintegratorcombfilters

Figure8:3rdorderCICdecimationfilterstructure,andmagnitude
responsebeforedecimationwhenD=R=8
Viewfullsizedimage

ImprovingCICattenuation
ThemostcommonmethodtoimproveCICfilterantialiasingandimage
rejectattenuationisbyincreasingtheorderMoftheCICfilterusing
multiplestages.Figure8showsthestructureandfrequencymagnitude
responseofa3rdorder(M=3)CICdecimatingfilter.
Noticetheincreasedattenuationats,out/RinFigure8bcomparedwiththe
1storderCICfilterinFigure6a.BecausetheM=3CICstagesarein
cascade,theoverallfrequencymagnituderesponsewillbetheproductof
theirindividualresponsesor:

Equation10
Thepricewepayforimprovedantialiasattenuationisadditionalhardware
addersandincreasedCICfilterpassbanddroop.Anadditionalpenaltyof
increasedfilterordercomesfromthegainofthefilter,whichis
exponentialwiththeorder.BecauseCICfiltersgenerallymustworkwith
fullprecisiontoremainstable,thenumberofbitsintheaddersis
Mlog2(D),whichmeansalargedatawordwidthpenaltyforhigherorder
filters.Evenso,thismultistageimplementationiscommonincommercial
integratedcircuits,whereanMthorderCICfilterisoftencalledasincM
filter.

Figure9:SinglestageCICfilterimplementations:fordecimationfor
interpolation
Viewfullsizedimage

BuildingaCICfilter
InCICfilters,thecombsectioncanprecede,orfollow,theintegrator
section.It'ssensible,however,toputthecombsectiononthesideofthe
filteroperatingatthelowersampleratetoreducethestorage
requirementsinthedelay.SwappingthecombfiltersfromFigure5with
theratechangeoperationsresultsinthemostcommonimplementationof
CICfilters,asshowninFigure9.Noticethedecimationfilter'scomb
sectionnowhasadelaylength(differentialdelay)ofN=D/R.That's
becauseanNsampledelayafterdecimationbyRisequivalenttoaD
sampledelaybeforedecimationbyR.Likewisefortheinterpolationfilter
anNsampledelaybeforeinterpolationbyRisequivalenttoaDsample
delayafterinterpolationbyR.
ThoseFigure9configurationsyieldtwomajorbenefits:first,thecomb
section'snewdifferentialdelayisdecreasedtoN=D/Rreducingdata
storagerequirementssecond,thecombsectionnowoperatesata
reducedclockrate.Bothoftheseeffectsreducehardwarepower
consumption.

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

6/9

2/22/2016

Understandingcascadedintegratorcombfilters

Figure10:CICdecimationfilterresponses:forvariousvaluesof
differentialdelayN,whenR=8fortwodecimationfactorswhenN=2
Viewfullsizedimage
Thecombsection'sdifferentialdelaydesignparameterNistypically1or2
forhighsamplerateratiosasisoftenusedinup/downconverters.N
effectivelysetsthenumberofnullsinthefrequencyresponseofa
decimationfilter,asshowninFigure10a.
AnimportantcharacteristicofaCICdecimatoristhattheshapeofthe
filterresponsechangesverylittle,asshowninFigure10b,asafunctionof
thedecimationratio.ForvaluesofRlargerthanroughly16,thechangein
thefiltershapeisnegligible.ThisallowsthesamecompensationFIRfilter
tobeusedforvariabledecimationratiosystems.
TheCICfiltersuffersfromregisteroverflowbecauseoftheunityfeedback
ateachintegratorstage.Theoverflowisofnoconsequenceaslongasthe
followingtwoconditionsaremet:
therangeofthenumbersystemisgreaterthanorequaltothe
maximumvalueexpectedattheoutput,and
thefilterisimplementedwithtwo'scomplement(nonsaturating)
arithmetic.
Becausea1storderCICfilterhasagainofD=NRat0Hz(DC),M
cascadedCICdecimationfiltershaveanetgainof(NR)M.Eachadditional
integratormustaddanotherNRbitswidthforstages.InterpolatingCIC
filtershavezerosinsertedbetweeninputsamplesreducingitsgainbya
factorof1/Rtoaccountforthezerovaluedsamples,sothenetgainofan
interpolatingCICfilteris(NR)M/R.Becausethefiltermustuseinteger
arithmetic,thewordwidthsateachstageinthefiltermustbewideenough
toaccommodatethemaximumsignal(fullscaleinputtimesthegain)at
thatstage.
AlthoughthegainofanMthorderCICdecimationfilteris(NR)Mindividual
integratorscanexperienceoverflow.(TheirgainisinfiniteatDC.)Assuch,
theuseoftwo'scomplementarithmeticresolvesthisoverflowsituation
justsolongastheintegratorwordwidthaccommodatesthemaximum
differencebetweenanytwosuccessivesamples(inotherwords,the
differencecausesnomorethanasingleoverflow).Usingthetwo's
complementbinaryformat,withitsmodularwraparoundproperty,the
followoncombfilterwillproperlycomputethecorrectdifferencebetween
twosuccessiveintegratoroutputsamples.
Forinterpolation,thegrowthinwordsizeisonebitpercombfilterstage
andoverflowmustbeavoidedfortheintegratorstoaccumulateproperly.
So,wemustaccommodateanextrabitofdatawordgrowthineachcomb
stageforinterpolation.Thereissomesmallflexibilityindiscardingsome
oftheleastsignificantbits(LSBs)withinthestagesofaCICfilter,atthe
expenseofaddednoiseatthefilter'soutput.Thespecificeffectsofthis
LSBremovalare,however,acomplicatedissueyoucanlearnmoreabout
theissuebyreadingHogenauer'spaper.1
WhiletheprecedingdiscussionfocusedonhardwiredCICfilters,these
filterscanalsobeimplementedwithprogrammablefixedpointDSPchips.
Althoughthosechipshaveinflexibledatapathsandwordwidths,CIC
filteringcanbeadvantageousforhighsampleratechanges.Largeword
widthscanbeaccommodatedwithmultiwordadditionsattheexpenseof
extrainstructions.Evenso,forlargesampleratechangefactorsthe
computationalworkloadperoutputsample,infixedpointDSPchips,may
besmall.

Compensationfilters
Intypicaldecimation/interpolationfilteringapplicationswewant
reasonablyflatpassbandandnarrowtransitionregionfilterperformance.
ThesedesirablepropertiesarenotprovidedbyCICfiltersalone,withtheir
droopingpassbandgainsandwidetransitionregions.Wealleviatethis
problem,indecimationforexample,byfollowingtheCICfilterwitha
compensationnonrecursiveFIRfilter,asinFigure1a,tonarrowtheoutput
bandwidthandflattenthepassbandgain.

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

7/9

2/22/2016

Understandingcascadedintegratorcombfilters

Figure11:CompensationFIRfilterresponseswitha1storder
decimationCICfilterwitha3rdorderdecimation
Viewfullsizedimage
ThecompensationFIRfilter'sfrequencymagnituderesponseisideallyan
invertedversionoftheCICfilterpassbandresponsesimilartothatshown
bythedashedcurveinFigure11aforasimplethreetapFIRfilterwhose
coefficientsare[1/16,9/8,1/16].Withthedottedcurverepresentingthe
uncompensatedpassbanddroopofa1storderR=8CICfilter,thesolid
curverepresentsthecompensatedresponseofthecascadedfilters.If
eitherthepassbandbandwidthorCICfilterorderincreasesthecorrection
becomesgreater,requiringmorecompensationFIRfiltertaps.Anexample
ofthissituationisshowninFigure11bwherethedottedcurverepresents
thepassbanddroopofa3rdorderR=8CICfilterandthedashedcurve,
takingtheformof[x/sin(x)]3,istheresponseofa15tapcompensation
FIRfilterhavingthecoefficients[1,4,16,32,64,136,352,1312,352,
136,64,32,16,4,1].
Awidebandcorrectionalsomeanssignalsnears,out/2areattenuatedwith
theCICfilterandthenmustbeamplifiedinthecorrectionfilter,adding
noise.Assuch,practitionersoftenlimitthepassbandwidthofthe
compensationFIRfiltertoroughly1/4thefrequencyofthefirstnullinthe
CICfilterresponse.
ThosedashedcurvesinFigure11representthefrequencymagnitude
responsesofcompensatingFIRfilterswithinwhichnosampleratechange
takesplace.(TheFIRfilters'inputandoutputsampleratesareequaltothe
s,outoutputrateofthedecimatingCICfilter.)IfacompensatingFIRfilter
weredesignedtoprovideanadditionaldecimationbytwo,itsfrequency
magnituderesponsewouldlooksimilartothatinFigure12,where>s,inis
thecompensationfilter'sinputsamplerate.

Figure12:Frequencymagnituderesponseofadecimateby2
compensationFIRfilter
Viewfullsizedimage

Advancedtechniques
Here'sthebottomlineofourCICfilterdiscussion:adecimatingCICfilter
ismerelyaveryefficientrecursiveimplementationofamovingaverage
filter,withNRtaps,whoseoutputisdecimatedbyR.Likewise,the
interpolatingCICfilterisinsertionofR1zerosamplesbetweeneachinput
samplefollowedbyanNRtapmovingaveragefilterrunningattheoutput
samplerates,out.ThecascadeimplementationsinFigure1resultintotal
computationalworkloadsfarlessthanusingasingleFIRfilteralonefor
highsampleratechangedecimationandinterpolation.CICfilterstructures
aredesignedtomaximizetheamountoflowsamplerateprocessingto
minimizepowerconsumptioninhighspeedhardwareapplications.Again,
CICfiltersrequirenomultiplicationtheirarithmeticisstrictlyadditionand
subtraction.Theirperformanceallowsustostatethat,technically
speaking,CICfiltersarelean,meanfilteringmachines.
Inclosing,therearewaystobuildnonrecursiveCICfiltersthateasethe
wordwidthgrowthproblemofthetraditionalrecursiveCICfilters.Those
advancedCICfilterarchitecturesarediscussedinmybookUnderstanding
DigitalSignalProcessing,2E.2
RichardLyonsisaconsultingsystemsengineerandlecturerwithBesser
AssociatesinMountainView,Ca.HeistheauthorofUnderstandingDigital
SignalProcessing2/EandanassociateeditorfortheIEEESignal
ProcessingMagazinewherehecreatedandeditsthe"DSPTips&Tricks"

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

8/9

2/22/2016

Understandingcascadedintegratorcombfilters

column.Youcanreachhimatr.lyons@ieee.org.

Endnotes
1.Hogenauer,Eugene."AnEconomicalClassofDigitalFiltersFor
DecimationandInterpolation,"IEEETransactionsonAcoustics,Speech
andSignalProcessing,Vol.ASSP29,pp.155162,April1981.
2.Lyons,Richard,UnderstandingDigitalSignalProcessing,2ndEd.,
PrenticeHall,UpperSaddleRiver,NewJersey,2004,pp.556561.

Partnerwithus

ListyourProducts

VisitournewPartnershipPortalfor
moreinformation.

Suppliers,listyourIPsforfree.

Partnerwithus

ListyourProducts

DesignReuse.com
ContactUs
Aboutus
D&RPartnerProgram
AdvertisewithUs
PrivacyPolicy

http://www.designreuse.com/articles/10028/understandingcascadedintegratorcombfilters.html

2016DesignAndReuse
AllRightsReserved.
Noportionofthissitemaybecopied,retransmitted,
reposted,duplicatedorotherwiseusedwithoutthe
expresswrittenpermissionofDesignAndReuse.

9/9

You might also like