1. SENIOR PROJECT
Dominick Marciano
Advisor: Michael Knox
Department of Electrical Engineering
Five Metrotech Center, Brooklyn, NY
MULTIPATH DETECTION USING SPREAD
SPECTRUM RAKE RECEIVERS
MAY 17, 1997

2. ABSTRACT
The initial goal of myprojectwas to show the benefitof usingarake receiverwithadirect
sequence spreadspectrum(DSSS)communicationsystem. Due totime andequipmentconstraints,I
was able tocompile limitedresultsthatshow intuitivelythatarake receiverwill improve signal
receptionandoverall reliability. More conclusive dataachievedbyotherswillbe presented.
While recordingthe resultsof myexperimentation,Irealizedthatthisdatacan also be usedto
prove that the autocorrelationfunctiondoesworkexactlyastheorypredicts. Autocorrelationisthe
basicelementof DSSSsystems. Withoutautocorrelation,DSSSsystemscouldnotexist. Therefore,Iwill
alsodiscussDSSSsystemsandthe autocorrelationfunction.
THEORY
A systemof communicationwasneededthatwouldbe hardtodetectand alsoensure the
privacyof the data. One systemthatwas developedtomeetthese requirementsiscalleddirect
sequence spreadspectrum(DSSS). The primaryconceptbehindDSSSisspreadingof the signal. Digital
data ismultipliedbyapseudo-randomnoisecode (PN code). ThisPN code isgeneratedbya shift
registerandwill repeat. There are particularconditionsthatmustbe metin orderto ensure a goodPN
code1
. One of the propertiesisthat itmustlookrandom. The PN code musthave a muchhigher
frequencythanthe datathat is to be transmitted.
Whentwo binarypulsesare multipliedinthe time domain,we know thatthe bandwidth(BW) of
the resultantsignal inthe frequencydomainwillbe twice the BWof the of the multiplier(PN code) in
the time domain2
. So,if our data wasa voice signal,itwouldnormallyoccupyapproximately4KHz
bandwidth. But,if the voice signal ismultipliedbyaPN code thatis operatingatclock speedof 1MHz,
the resultantsignal inthe frequencydomainwillhave a2 MHz bandwidth. Therefore,the originalvoice
data has beenspreadsignificantly. Inordertoconserve energy,the amplitude of the signal must

3. decrease asthe bandwidthof the signal increasestherebymaintainingthe areaunderthe signal curve
constant.
By significantlyspreadingthe signal,the amplitude of the signal maygolow enoughthatitis less
than backgroundnoise and thereby makingitdifficulttodetect. Hidingthe signal belowbackground
noise resultsinincreasingthe privacyof the signal. Thisspreadsignal canthenbe modulatedonan RF
carrier andtransmittedbycable or wirelessly.
On the receiverend,the processisreversedinordertoretrieve the original signal (data). The
firststepisto multiplythe receivedsignalbycarrierfrequency. Thiswill removethe carriercomponent
fromthe signal sowe can work ondecodingthe original data. The nextstepisto de-spreadthe signal
by multiplyingthe signalbythe same exactPN code that was usedtospreadthe signal. Itiscritical that
the PN code on the receiverendisnotonlyidentical tothe PN code onthe transmitterend,butthey
mustproperlyalignintime.
Since the PN codes are binaryand identical,we getthe autocorrelationof the twosignalswhen
theyare multiplied. Fromknowledge of FourierTransformswe know thattwopulses(digital 1’s)
multipliedwill givearesultanttriangle shape. The autocorrelationfunctiontellsusthatwe will see this
resultwhenthe twoPN codesare perfectlyaligned. The more the PN codesare misaligned,the less
correlationwill existmakingitmore difficulttorecoverthe original data. If the PN codes are completely
misaligned,nocorrelationwill existandthere will be nodatarecovery. Throughpreviousworkitis
knownthat the lowerlimitof the signal willequal:
(# agreements − #disagreements)
PN Code Length
= −
1
15

4. for a four(4) bitshiftregister(Figure 1). PN code lengthisdeterminedsimplybythe following function:
2n
-1 where n = # of stages in the shift register
Figure 1. Autocorrelation Function - 4 Bit Shift Register
Whendealingwithwirelesscommunications,the propagationof electromagneticwavesmust
be considered. When asignal istransmittedwirelessly,itwill propagateinall directions(basicdipole
antenna). The propagatingwave maybe inline of sight(LOS) of the receiverorit may getreflectedor
refractedpriorto gettingtothe receiver. The LOSsignalswill arrive atthe receiverpriortothe
reflected/refractedsignals. The twoor more signalsare identical exceptthe reflected/refractedwaves
are a delayedversionof the LOSsignal3
. These non-LOSsignalsare referredtoasmulti-pathsignals.
Since the multipathsignalsare delayed,we now know thatwhentheyare multipliedbythe PN
code there will be little ornocorrelationandthe signal canbe discarded. However,there isnoreason
whywe can’t use these signalstoimprove ourreceptionandreliability. Theyare still exactreplicasof
the original andLOS signals. Thiscan be accomplishedbyuse of a rake receiver.
A rake receivercandetectand assistindecodingmultiple delayedsignalsbysplittingthe
receivedsignal intoasmanybranches asdesired. Eachbranch signal will getmultipliedbythe original
PN code shiftedtodifferentpointsintime. Assumingthere isasingle multipathsignalthatisdelayedby
0.5µsec (approximately150 metersof extratravel),inadditional tothe LOSsignal,the delayedsignal
can be recoveredif itismultipliedbythe original PN code delayedby0.5µsec. Now the receivercan

5. processtwo(2) copiesof the original signal. The twosignals,ormore ina real worldapplication,canbe
combinedtoincrease the signal strength4
andimprove signal clarity. Anotheroptionistoassigna
weightingfactortoeachof the decodedsignalsbasedontheirsignal level. Once thisisaccomplished,a
decisioncircuitcanbe usedto selectthe strongestsignal. If the selectedsignal islost ordegraded due
to interference orfading,the nextstrongestsignal canbe chosen. Utilizingthismethodwill
undoubtedlyimprovesystemreliability.
EXPERIMENT
Since the experimentwasconductedinalaboratoryenvironment,propagationdelayhadtobe
simulated. Inordertoachieve a1µsec delayedsignal,the transmittedsignalwouldhave totravel
approximately300 metersbefore beingreceived. Thistype of delaycouldnotbe accomplishedina
small laboratory,soI useda Hewlett-Packard11759BRF channel simulatorbetweenthe transmitterand
receiverandhardwiredthe systemtogetherwithcoaxialcable (Figure 2).
Figure 2. System used to simulate multipath signals and their recovery

6. The channel simulatoriscomputercontrolledtosplitanincomingsignal andtointroduce signal
delay. The delay toa signal issetbythe operatorandcan be variedinstepsof 0.1µsec. Due to the
designof the simulator,adelayof 0.5µsec will be introducedto anyinputsignal. Thisinternal delayof
the simulatorhadto be accountedfor inthe assemblyof the PN code generator. The simulatoralso
introducedsignificantsignalattenuation. Inorderto account forthis attenuation,anamplifierwas
addedat the outputof the simulator.
The PN code generatorusedisshowninFigure 3. It was clockedat1MHz inorderto ensure the
data was spreada significantamount. The code comingfromregister5 was usedtospreadthe data.
The spreadsignal wasinputto the channel simulatorwithadelayof 0µsec forpath #1, whichresultedin
a 0.5µsec delayinthe outputsignal due tothe internal systemdelay. Forease of calculation,an
additional 0.5µsecdelaywasintroducedtothe path#1 signal givingita total 1µsec delay. Thiswasused
as the reference signal (LOS) where t0 =0.1µsec. The same PN code generatorwasusedto decode the
receivedsignals. Thiswasdone forsimplicitybyavoidingthe needtotrack the signal andobtain
alignment. AnotherPN code generatorwasusedtosimulate the transmitteddata. ThissecondPN code
generatorwasclockedat 6KHz whichrepresentsaslightlygreaterbandwidththanthatof voice data.

7. Figure 3. PN code generator
To ensure the transmitterwasfunctioningproperly,aspectrumanalyzerwashookeduptothe
inputof the channel simulator. InFigure 4 itcan be seenthatthe bandwidthof the signal is2MHz;
whichistwice the bandwidthof the spreadingPN code andis exactlywhatisexpected. Several other
measurementswere takenandFigure 5showstwosignalscomingoutof the channel simulatorwhere
one of the signalsisdelayedby 2µsec.

8. Figure 4. Spread signal from transmitter
Figure 5. Two signals, one with a 2 microsecond delay
Usinga digital oscilloscope,itiseasytoconfirmthe operationof the entire system. InFigure 6,
the original signal isshownalongwiththe decodedsignal. Itisquite obviousthatthe signal isdecoded
and demodulatedasexpected. InFigures7through9 the delaywasprogressivelyincreaseduntilthe PN
codeswere misalignedbyexactly1bit(1µsec). It can be clearlyseen thatthe autocorrelationfunctionis

9. Figure 6. One signal recovered
Figure 7. A 0.2 microsecond delayed signal recovered

10. Figure 8. A 0.7 microsecond delayed signal recovered
Figure 9. Attempted recovery of a 1 microsecond delayed signal

11. performingasexpected. Asthe bitsbecome lessaligned,the decodedsignalbecomesmore degraded.
Whenthe two PN codesare completelymisaligned,the signalisnotdecodedatall.
To testthe abilitytodecode multipleversionsof the same signal,Isetpath#1 to a 1µsec delay
and path#2 toa 2µsec delay. Usinga 1µsec and a 2µsec delayedPN codes (registers6and 7 of the PN
code generator),bothsignalswere successfullydecoded (Figure10). It isclear that the signalsare the
same representingidentical data. The onlyitemtonote isthat the signal frompath #2 isinverteddue
to the logicimplementedandisinsignificant. Inthe final step,the signal inpath#2 wasadjustedto
preventPN alignment. Thisresultedinthe lossof signal frompath#2 withno effectonthe receptionof
the signal onpath #1 (Figure 11).
Figure 10. Two signals successfullyrecovered

12. Figure 11. Attempted detection of two signals, with one lost
RESULTS OF OTHERS
There are several differentmethodsof implementingrake receiversinaDSSSsystem. One
methodthathas beenemployedistohave a multiple stage rake receiver5
. Inthismethod,the channel
impulse response isestimatedaswell asmultipleaccessinterference. Byknowingthisinformation,it
enablesmore reliable detectionof databits. Thishelpsimprove systemsthatemploycode division
multiple access(CDMA). InFigure 12,it can be seenthatwhena 2 stage rake receiverisusedas
opposedtoa single stage receiver,the biterrorrate (BER) fora givensignal tonoise ratio(SNR) is
decreased.

13. Figure 12. BER as a function of SNR for 1 and 2 stage rake receivers in the presence of 2 multipath components
Anothermethodusesmultiplesub-carriersinthe transmitteraswell asthe receiverandis
referredtoas multiple carrierspreadspectrum(MCSS)6
. The datais splitintoparallel pathsafteritis
codedby normal DSSSmethods. Each pathis thenmultipliedbyanindividual chipof the PN code and
subsequentlymodulatedontosub-carriers. These individual pathsare thensummedintoaserial multi-
carrier signal thatwill thenbe modulatedontothe maincarrierandtransmitted. Inthe receiver,the
processisreversedtodecode the signal. InFigure 13 it can be seenthe BER vs SNR forthe MCSS system
improvesasthe numberof multipathchannelsincreases. Whenthere are onlytwo(2) multipath
signals,astandardrake receiveractuallyprovidesslightlybetterperformance. Usingastandard rake
receiver,the probabilityof errordecreasesfora givenSNRasthe numberof branchesisincreased

14. (Figure 14). Aninterestingfactisthatas the PN sequence isincreased,there isalsoadecrease inthe
probabilityof error7
.
Figure 13. Performance of DSSS and MCSS rake receivers with L multipath signals

15. Figure 14. Probability of error based on # of rake branches and PN code length
CONCLUSION
In thisexperiment,aswellasthose conductedbyothers,itisclearthat DSSS systemsbehave as
predicted. Fromthe bandwidthspreadingtothe performance of the autocorrelationfunction,thistype
of systemprovides reliablecommunication. Reliabilityisimprovedasexpectedwhenarake receiveris
usedand ithas beenshownthatBER and probabilityof errorare reduced. These resultsare both
intuitive andconclusive fromthe experimentsconducted. There are many otherimplementationsof
rake receiverswaitingtobe built. Asmore researchandexperimentationisconducted,signal reception
and reliabilitywill undoubtedlyimprove.