channel lecture important 2
TRANSCRIPT
-
8/6/2019 Channel Lecture Important 2
1/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
1
ChannelAmplitudeResponseH(f)
Achannelischaracterizedbyeitherits:
1. ImpulseResponseh(t):Thisisthetimedomainrepresentationofthechannel.Itrepresentstheoutputofthechannelwhentheinputofthechannelisadeltafunction(t),whichisanimpulse
at t = 0. The impulse response indicates how a channel spreads the input signal. An ideal
channeldoesnotspreadthesignalatallwhileanonidealchannelspreadstheinputsignal.The
spreadingoftheinputsignalisequaltothewidthoftheimpulseresponse.
2. Transfer function H(f): This is the frequency domain representation of the channel. Itrepresentstheamplitudeof theoutputof thechannelwhenasinusoidwithamplitude1V is
andfrequencyf isinputtothechannel.
Intimedomain,anidealchannelwithinfinitebandwidthisonethathasan impulseresponsethat isa
delta function at time 0 or later. If it has an impulse response of delta at 0, the channel does not
introduceanydelayinsignal(thereiszerodelaybetweentransmittingthesignalandreceivingitatthe
outputofthechannel).Iftheimpulseresponseisadelayeddeltafunction,thechannelisstillidealbut
there is a nonzero delay between transmitting the signal and receiving it at the other side of the
channel.Infrequencydomain,anidealchannelwillhaveatransferfunctionwithaflatmagnitudeforall
frequencies.Thephaseofthetransferfunctioniszeroifthechannelhaszerodelayandwillbeastraight
line(linearcurve)ifthechannelhasanonzerodelay.
Inreality,nochannelhasinfinitebandwidthbutallchannelshavesomebandwidthWthatisfinite(less
than infinity). A channelwith finite bandwidth can still be considered ideal if the response of that
channelisflat(constant)oversomefrequencyrange(forexamplefrom0toWforalowpasschannels)
asseeninpart(a)ofthefigurebelow.Asignaltransmittedoversuchachannelwillexperienceanideal
channelifthebandwidthofthesignalislessthanW.IfthebandwidthofthesignalisgreaterthanW,the
channelwillnotactasanidealchannelforthissignal,butsomedistortionwilloccurinthesignalasitis
transmittedthroughthechannel.
Furthermore, in reality,channelsarenever ideal (somechannelsmayonlybeapproximated tobeing
idealforspecificapplications).Atypical lowpasschannel isshown inpart(b)ofthefigurebelow.The
bandwidthofthechannelinthiscasemaybeoneofseveralvaluesdependingontheapplication,but
themostwidelyusedisthe3dBbandwidth.
-
8/6/2019 Channel Lecture Important 2
2/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
2
(a) (b)
Now consider thatwe have a digital signal thatwewould like to transmit over a specific channel.
Assumethat:
BandwidthofDigitalSignal=s
W
BandwidthofTransmissionChannel= cW
Wecaneasilyverifythefollowingtwocases(assumingnonoiseexists):
Forthecaseof sW W :The input signalwill completelypass through the channel andwillnotexperienceanydistortion
giving an output signal that is similar to the input signal with the exception of
amplification/attenuationanddelay.
Forthecaseofs
W W> :
The input signalwillonly partially pass through the channel and thereforewill experience some
distortion that results from filtering out high frequencies. The signal may also experience
amplification/attenuationandpossiblysomedelay.
-
8/6/2019 Channel Lecture Important 2
3/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
3
NyquistSignalingRate
Nyquistcameupwithaformulathatgivesthemaximumnumberofpulsesyoucantransmitthrougha
channelthathasbandwidthc
W .TheNyquistSignalingRatetheoremstatesthatthemaximumrateat
whichpulsescanbetransmittedoverachannelwithBandwidthc
W is
Maximumpulsetransmissionrate maxR = 2 cW pulses/s
Note:theNyquestSignalingRatetheoremgivesthemaximumPULSERATEnotthemaximumBITRATE.
Infact,differentpulsesmaycarrydifferentnumberofbitsperpulse(1bitforpulsesthattakeoneof
twovalues5Vand+5V,or2bitsforpulsesthattakeoneof4values0V,1V,2V,and3V,or3bitsfor
pulsesthattakeoneofthe8values0V,1V,2V,3V,4V,5V,6V,and7V).
Ifeachpulserepresents1bit,thenbitrateinthiscaseis 2 cW bits/s Usingmultileveltransmissionwithnumberof levelsof 2mM = (where m isnumberofbits),we
getamaximumtransmissionbitrateof
max
pulses bits bits2 2
second pulse secondc c
R W m W m= =
SignaltoNoiseRatio(SNR)andMulti LevelTransmissions
Onemeasureof thequalityofasignal is theSignal toNoiseRatio (SNR).Thesignaltonoiseration is
often defined as shown in the following equation. Sometimes the SNR becomes useful if it is
representedindBformasshownbelow.
-
8/6/2019 Channel Lecture Important 2
4/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
4
( ) ( )10
Average Signal Power(Linear)Average Noise Power
dB 10log (Linear)
SNR
SNR SNR
=
=
ShannonChannelCapacity
Shannon(anAmericancommunicationengineerandmathematician)provedthatforaspecificchannel
withaspecificsignaltonoiseratioSNR,youcantransmitamaximumnumberofbitspersecondequal
toCthatisgivenbytheformula(thesignaltonoiseratioisthelinearSNR)
( )2bits
log 1+SNR(Linear)second
cC W=
Where Wc is the channel bandwidth in (Hz), and SNR(Linear) is the linear signal to noise ratio. The
ShannonChannelCapacitystatesthatforanoisychannel,youcantransmitreliablyamaximumnumber
ofbitspersecondequalto C . Ifyou trytoexceedthechannelcapacity foraparticularchannel,the
noisewillcauseuncontrollablehighratesoferrors.
BandwidthofAnalogSignals
One characteristic of an analog signal is its bandwidth. The bandwidth of analog signals is usually
computedusingoneofthefollowingtwomethods:
Bandwidth isequal to thehighest frequency component in the signal: theproblemwith thismethod of computation of the bandwidth is that there is a possibility that the frequency
componentsofthesignalaboveaspecificfrequencyareverysmallthattheycontainnegligible
power,yet,thebandwidthbecomessolargeduetotheseverylowpowercomponents
Bandwidthisequaltothefrequencyatwhichthepowerofthesignalcomponentsdropbelowthe power of the signal components at zero frequency: this bandwidth is less than the
bandwidthaboveand itconsidersthesignalcomponentsabovethatbandwidthnottobepart
ofthesignal(theycanbecompletelyremovedwithoutmuchdistortiontothesignal.
SamplingofAnalogSignals(NyquistSamplingTheorem)
Beforeananalogsignalcanbetransmittedoveradigitalcommunicationsystem,theanalogsignalmust
besampledandthenquantized,whichconvertstheanalogsignalintoasequenceofbits.Accordingto
-
8/6/2019 Channel Lecture Important 2
5/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
5
NyquistSamplingTheorem,topreservethequalityofananalogsignal,itmustbesampledatarateATLEASTdouble thebandwidthof thesignal (called theNyquistsamplingrate). Ifsampling isdoneatalowerrate,aliasingoccurswhichcausesthereconstructedsignaltobedistorted.Sincesignalsmayhave
componentswith frequency above the bandwidth, the proper thing todo before sampling to avoid
aliasing from occurring is to pass the analog signal through a lowpass filter that removes any
componentsaboveonehalftheNyquistsamplingrate.Thisisillustratedinthefollowingfigure:
-
8/6/2019 Channel Lecture Important 2
6/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
6
Quantization
Onceananalogsignalissampled,thenextstepistoquantizeit,meaningthatafiniteprecessionvalueis
used todescribe thevalueofeachsample. Inquantization,ananalogsamplewithanamplitude that
maytakeanyvalueinaspecificrangeisconvertedtoadigitalsamplewithanamplitudethattakesone
ofaspecificpredefinedsetofquantizationvalues(withfinitenumberofdigitsneededtodescribethe
value
of
each
sample).
This
is
performed
by
dividing
the
range
of
possible
values
of
the
analog
samples
intoL different levels, and assigning the center value of each level to any sample that falls in thatquantization interval. The problemwith this process is that it approximates the value of an analog
samplewiththenearestof thequantizationvalues.So, foralmostallsamples, thequantizedsamples
willdifferfromtheoriginalsamplesbyasmallamount.Thisamountiscalledthequantizationerror.To
getsomeideaontheeffectofthisquantizationerror,quantizingaudiosignalsresultsinahissingnoise
similartowhatyouwouldhearwhenyouplayarandomsignal.
Assume thatasignalwithpowerPsis tobequantizedusingaquantizerwithL = 2n levelsranging in
voltagefrommp to mpasshowninthefigurebelow.
Quantizer Output Samplesq
x
Quantizer Output Samples x
Wecandefinethevariable v tobetheheightoftheeachoftheLlevelsofthequantizerasshown
above.Thisgivesavalueof v equalto
-
8/6/2019 Channel Lecture Important 2
7/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
7
2 pmv
L = .
Therefore,forasetofquantizerswiththesamemp,thelargerthenumberoflevelsofaquantizer,the
smaller the size of eachquantization interval, and for a setofquantizerswith the samenumber of
quantizationintervals,thelargermpisthelargerthequantizationintervallengthtoaccommodateallthe
quantizationranges.
Aftersimplemanipulation,wecanwritetheSignaltoNoise(orQuantizationNoise)Ratio(SNRorSQNR)
as
2
2
Signal Power
Noise Power
3.
s
q
s
p
PSNR
P
LP
m
= =
=
IngeneralthevaluesoftheSNRaremuchgreaterthan1,andamoreusefulrepresentationoftheSNR
canbeobtainedbyusing logarithmicscaleordB.WeknowthatLofaquantizer isalwaysapowerof
twoorL = 2n.Therefore,
( )
( )
22
10 10 102 2
10 102
3 310 log 10 log 10 log 2
3
10 log 20 log 2
6 dB.
n
dB s s
p p
sp
LSNR P P
m m
P nm
n
= = +
= +
= +
Note that shown in the above representation of the SNR is a constant for a specific signalwhen
differentquantizerswiththesamempareused.
ItisclearthattheSNRofaquantizerindBincreaseslinearlyby6dBasweincreasethenumberofbits
that thequantizerusesby1bit.Thecost for increasing theSNRofaquantizer is thatmorebitsare
-
8/6/2019 Channel Lecture Important 2
8/8
EE 400: Communication Networks (091) Dr. Wajih A. Abu-Al-Saud
Ref: A. Leon Garcia and I. Widjaja, Communication Networks, 2nd Ed. McGraw Hill, 2006Latestupdateofthislecturewason17 12 2009
Lecture13:DigitalTransmissionFundamentals
8
generatedand thereforeeitherahigherbandwidthor longer timeperiod isrequired to transmit the
outputdigitaldataofthequantizer.
DigitalTransmissionofAnalogSignals
Now,oncethesignalhasbeenquantizedbythequantizer,thequantizerconvertsittobits(1sand0s)
andoutputsthesebits.Lookingattherepresentationofthequantizershownabove,weseethateachof
the levelsof thequantizer isassignedacode from 000000 for the lowestquantization interval to
111111 forthehighestquantization intervalasshown inthecolumn totherightofthefigure.The
outputbitsarecalledaPulseCodeModulation.ThePCMsignalisobtainedbyoutputtingthebitsofthe
differentsamplesonebitaftertheotherandonesampleaftertheother.Thiswouldbethecodethat
would be transmitted in a digital communication system. So, it is clear that either increasing the
samplingrateorthenumberofquantizationlevelsofthequantizerwouldincreasetheinformationthat
wewouldhavetotransmit.
t4T
sT
s3T
s5T
s2T
s0
mp
mp
L = 2n
L levels
n bits
0.000
0.001
0.010
1.111
.
.
.
.
PCM Code
n bits/sample
0
v
Quantizer Output Samplesq
x
Quantizer Output Samples x
A quantization interval Corresponding quantization value