Beruflich Dokumente
Kultur Dokumente
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
88 T R AIN
R SEA C T I O N S ON COMMUNICATIONS
SYSTEMS June
there is no intersymbol dependence. At the receiver it is
known that at a particular time interval one of the two
signals has been transmitted. The receiver will then decide
which of thetwowastransmittedduringthistime
interval. In the presence of interference such as noise or
fading, the decision is based on the digit tvhich is more
likely to have produced the received signal.
11. PsK SYSTEM
= 0, elsewhere.
I F S I G N A L
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
19.60 Glenn: Comparison of Binary Coded Transmission Systems 89
H SENT
TIME
\
\/ OUADRATURE
= 0, elsewhere,
and for the space frequency as
(14)
bl,- RESPONSE AMPLITUDE
DETECTED
POWER
I 1I SIGNAL POWER
I \1 A
= 0 elsewhere. (16)
The Fourier transform of these time functions shows a NOISE POWER
sin 8 / 8 form of spectrum centered a t t h e mark and space
frequencies, respectively. The noise is assumed to be white
Gaussian distributed. The spectral densities of the signal
and noise before and after detection are shown in Figs.
, 4 and 5, respectively.
IV. IMPROVEMENT
THRESHOLDS
The curves shown in Fig. 6 illustrate the variation of
theoutput signal-to-noise ratio vs theinput signal-to-
noise ratio for the differentmodulationsystems. The
I
;,f !y -eD+
I
I
4- 'f
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
90 IRE T R A N S A C T I O N S ON COMMUNICATIOXS S Y S T E M S June
rapiddeteriorationinperformance if this condition is applied to a high-& circuit which is tuned to the second
not met.3 harmonic of frequency and independent of the phase of
In thecoherent PSK system, where a noise-free reference the input signal. This double frequency signal is used to
is assumed to be present in the receiver, there is no phase lock a divide by two oscillator. The output of this
threshold.However, when the phase reference must be oscillator is then used as a phase-stable reference signal,
obtained from thesignal, such as for differentially coherent as its phasecharacteristic is notaffectedbythephase
PSI< or phase comparison systems, the threshold character- shifts in the transmitted signal. I n this way, a n improve-
isticsapproximately follow theamplitudemodulation ment in SNR is obtained for the referencesignal. The
case. input signal-to-noise ratiomust be above the threshold
The PSK system is the optimum binary system using conditions for this circuit. This SNR mill be dependent.
two signals which havea cross-correlation coefficient of upon the frequency pull in range andpull in time.
minus one. Or, the two signals are the negatives of each
other. I n a carrier system, this means that thetwo signals
di.ffer fromeach other only by aphasereversal of the
PHASE DELAY
carrier. In the ideal system, it is assumed that the phase DETECTOR LINE
. reference is known at the receiver. Since the phase refer- I
ence needed to distinguish between the two signals is not
available at the receiver in the practical case, this PSI<
systemfallssomewhatshort of the theoretical limit, in
rLyLRis
-7- OUTPUT
RECEIVER
I
decision process depends upon the previously transmitted PRODUCT
signalfor the phasereference,thereforeresultingin DEMOD
degrade the SNR. Thus,performance of this type of PSK Since the receiver is, in the last analysis,a decision
detection technique degrades as the SNR decreases over device, a decision rule is needed which may or may not
the optimum where the noise-free reference voltageis’ Staff, Mass. Inst. Tech., Cambridge, Mass., lecturenoteson
assumed. Reliable Long Range Communicat~ons;1959.
Another method of detecting the phase of the signal M. Schwartz, “Information Transmission,Modulation, and
Noise,” McGraw-Hill Book, Co., Inc., New York, N. Y., sec. 6-2;
is shown in Fig. 7(b). The input signal is passed through 1959.
a conventionalfull-wave rectifier. Theoutput is then J. G. Lawton,“Comparison of BinaryData Transmission,”
Proc. 1958 Conf. on Medical Electronics.
’S. Reiger, “Error probabilities of binarydata transmission
M. G. Crosby,“Frequencymodulation
a noise characteristics,” systemsin the presence of random noise,” 1953 I R E CONVEN-
PROC.
I R E , vol. 25, pp. 417-572; April, 1937. TION RECORD, pt. 8, pp. 72-i9.
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
1960 Comparison
Glenn: of Binary Coded
Systems
Transmission 91
involve a threshold. The receiver must decide whether or If f ( t ) is applied to a filter whose impulse response is
not a signal is present and it must make thisdecision with h(t), then the output at t = T ois
some requiredprobability of error. The decision is a TO
“yes”or L‘nolloperation which may be determined by fo(6 = [ f(W(T0 - t> d t . (20)
means of a threshold device. The problem of decision
then resolves itself into the question of whetherornot Comparing this expression with the preceding expres-
the signal exceeds the threshold. If noise alone is present, sion, the decision rule can be realized. Connect two linear
the threshold should not be exceeded, and, if signal plus matched filters havingimpulse responses s1 ( T o - t )
noise is present, the thresholdshouldbe exceeded. As and sz ( T o - t ) to synchronous detectors and connect the
the noise can have any amplitude with a finite probability, outputs to a comparison circuit as shown in Fig. 8. The
it is evident that noise alone will occasionally exceed the synchronous detectors will be operated so that the signal
threshold and cause triggering of the systemoutput from the filters is sampled a t t = To.If the output of the
circuits. This event causes an error and the probability first detector exceeds the second detector, announce that
of this happening canbe decreased by raising the threshold s l ( t ) is present;otherwise,report that s2(t) is present.
sufficiently high. Another way is to have a positive pulse if s,(t) is present
The receiver will utilize the received waveform and a and a negative pulse if s,(t) is present.
knowledge of the transmitted waveforms. For a binary-
coded data transmissionsystem, the twotransmitted
signals areeither s,(t) or s,(t). Assume thatthe two
symbols have the same probability of occurrence and that
there is no intersymbol dependence. The received symbol
consists of one known signal s l ( t ) , or anotherknown
signal sz(t), andineither case additiveGaussian noise
n 0).
The signals sl(t) and s2(t) may represent a mark and
space and be carriers of different frequencies as in FSK. A
device is needed for each interval in the message to decide u -
which of the two possible signals mas transmitted. Fig. 8-Decision circuit with matched filters.
If f ( t ) is the received signal in the interval 0 < t < To,
s l ( t ) will beannounced, if the meansquare difference
between f ( t ) and sl(t) is less than the mean square differ- Now consider the condition for noncoherent detection.
ence between f ( t ) and s2(t). If the oppositecondition is This is for the random phase FSIi type of signal. That is,
satisfied, then the signal s2(t) is present. This is expressed the synchronous detectors are replaced by envelope
mathematically as detectors and the two filters are tuned to the mark and
space frequencies. Then it ban be shown that a mark or
s l ( t ) is announced if
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
92 IRE TRANSACTIONS O N COI"UNICA1'IONS S Y S T E M S June
input to the filter with impulse response The difference betlveen the coherent PSK and non-
coherent FSK system is about 4 db for low error prob-
h(t) = Sl(2', - t ) - s,(T, - 1 ) . (22) abilities. The 3-db difference is due t o the difference in the
If the output at t = To exceeds zero, announce s,(l); correlation coefficient of the two transmitted waveforms
otherwise, announce s2(t).Since the filter is linear and the for the PSK and FSK signals. The additional 1-db differ-
input noise Gaussian, the sample of the output is Gaus- ence is explained as follows.
sian distributed with amean of V , or - V o depending Acoherent PSK systemrequiresa knowledge of the
upon whether sl(l) or s2(t) was present. See Fig. 9. phase as well as the frequency and envelope of the ex-
If sl(t) and s,(t) are equally likely to be transmit.ted and pected waveforms. In a practical FSIi system, the phase
the decision rule follows (19) and (21), then the probability information will not be known at the receiver. When the
of error can be s h o ~ n ' . ~ bet o phase is known, the filter output can be sampled 'at the
precise instant when the expected output reaches a
maximum. However, if the carrier phase is not known, i t
is necessary t o form the envelope of the matched filter
where output signal and sample the signal at its peak value. As
E is the average signal energy per received symbol,
p is the finitetimecorrelation coefficient of the two
waveforms, s,(l) and sz(t) which ranges from
-1 I P 51,
N o is the noise power density. SPACE t NOISE
tI pv
MARK+ NOISE
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
1960 Comparison
Glenn: of Binary Coded Transkission
Systems 93
the formulation of the envelope is affected by both the after demodulationmust besatisfied. I n a low capture
in-phase andquadrature noise components, the error ratio FSK system, if the threshold condition at the input
probability will be increased for the noncoherent system to the limiter is satisfied, the thresholdcondition after
compared to the coherent case where only the in-phase demodulation will be satisfied.
noise components affect the amplitude of the sample.
The analysis of errorprobabilityfor the different VI. ANALYSISOF PSI< AND FSK RECEIVERS
systemsassumes that the receiver islinear up to the The block diagrams of optimum and differentially
demodulator and that matched filters are used. Also, the coherent PSK receivers are shown in Fig. 11. The optimum
cross-correlation of noise in the two filters used for FSK PSK receiver assumes knowledge of the phase of the signal
is negligibly small. When there are nonlinearities such as at the receiver, whereas the differentially coherent PSK
limiters, the receiver first threshold point occurs at the receiver obtains this information from the signal. Block
limiter inputapproximately when the desired and un- diagrams of noncoherent FSK receivers are shown in
desired signal voltages attain the same pea.k amplitudes. Figs. 12 to 14.
If the undesired signal is a random function then there will
be a certain probability that this signal will be between A . Noncoherent Dual-Filter Linear FSK and Differentially
certainvalues. For example, if the undesiredsignalis Coherent PSK Receivers
Gaussian noise, the probability that the noise amplitude The firstcomparison will be betweenanoncoherent
will exceed four times the rms value is about one in ten FSK two-filter linear receiver and differentially coherent
thousand. For t,his case, the so-called crest factor ( C F ) is
equal t o 4. For this example, the SNR before limiting LOW PASS
FILTER F ‘ i ’ O U l i
will be shown in Section VI to be 8 to 1 or 9 db.
PHASE
The analysis of errorprobability uses E / N o the nor- LOCKED
N
and T o is the pulse duration. PRODUCT
DEMO0
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
94 T R A N S AICRTEI O N S ON COMMUNICATIONS
SYSTEMS June
and coherent PSI< receivers. These receivers are shown For the phase comparison or differential PSI< receiver,
inFigs. 14 and 11, respectively. Thebandwidthand the CNR should exceed the threshold level of the phase
CNR at point a inFig. 14 before the mark and space detector (CNR > 3), or the output CNRwill be degraded
filters are B,F and C‘/N, respectively. It is obvious that from the input CNR. The PSK receivers shown in Fig.
the “ideal”pass band for each filter would need to be 11 can tolerate only relatively small frequency shifts such
B, + 2 A F , where AF is the increased bandwidth re- that A F / B l < 3. dbovethisvalue,thesecircuitsare
sultingfromfrequencyinstability. As previously men- inadequate.However, PSK demodulators have been
tioned, the mark and spacefrequenciesshouldbe suffi- developed that can tolerate moderately larger frequency
ciently separated to insure isolation between “practical” errors.
filters with finite attenuation rates. The noise bandwidth Eqs. (32) and (33) give the relationship of the output
of each practical filter is B , +
AP. Therefore, the approxi- SNR as a function of the CNR, and bandwidths before
mate total noise bandwidth is B,F w 2(B, AF). + andafterdetection. It is now necessary t o relate the
The SNR, after either filter, a t point b is average SNR to the normalized SNR so that the error
probability curves can be used. For a matched filter,
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
1960 Comparison
Glenn: of Binary Coded Transmission
Systems 95
Substituting (35) and (36) into (38), weget for an error probability of 1 in 10,000 is
C
- = 8. (43)
B,, = 20 + B1 + AF (41)
fm = bandwidth of the low-pass filter (fm NN 11) .
-y
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
96 IRE TRANSACTIONS ON COMMUNICATIONS SYSTEMS June
crossings 0 ill become moreuniform andapproachthe two-filter nonlinear FSK receivers is approximately
desired signal frequency asthe SNR increases. These
characteristics mill have a sharpthreshold which, in
TDPSK
turn, mill dependupon the SNR before the detector." N
TFSK-NL
I n a low capture ratio receiver, the threshold condition
after detection will alwaysbe satisfied if the threshold
condition before the limiter is satisfied. For CNR before detection of less than 3, the degradation
Eq. (37) shows that the range will be dependent upon of SNR in the detection process must then be accounted
the bandwidth of the receiver before limiting. This band- forindeterminingthe necessary E / N o t o satisfy the
width, in turn, depends upon the frequency instability. threshold criterion.
Increasing the bandwidth will increase the noise power
N , so that the signal power C must be thereby increased D. Operating Range lis A F I B , for PSI< and FXK Systems
t o satisfy the CNR thresholdcondition. The effect of the ratio of the increased bandwidth A F ,
which may be due t o frequency instability to theinforma-
C.,Noncoherent Nonlinear Dual-Filter F S K Receiver tionbandwidth B , before demodulationonoperating
range performance for the various types of FSK systems
The next receiver to consider is shown in Fig. 13, where and the differentiallycoherent PSK system is shown in
the single broad-band filter is replaced by two narrow- Fig. 15. Thesecurvesareplotted for equal effective
band filters. The bandwidth before limiting is radiated transmitted power, effective area of the receiving
antenna, receiver noise figure, a,ntenna temperature,
system margin, and the same bit error probability. For a
bit error probability of 1 part in 10,000,
.For the samesignal power at the limiter input, the
ratio of the CNR for the single and dual filter receivers
are, on the basis of their passband ratio,
where
(GIN), is theCNR of the single broad-bandfilter
receiver at the limiter input,
(GIN), is theCNR of thedual narrow-bandfilter
receiver at the limiter input.
AF/B, -; ,,
Fig. 15-Operating range vs AF/B1.
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
1460 Comparison
Glenn: of Binary Coded Transmission
Systems 97
A typical example of these systems is given for the follow- 4) The signal experiences a time delay and a loss, both
ing conditions: of which may be time-varying.
biterrorprobability P, = A careful consideration shows that time-varying
a
mark or space duration To = 1.33 milliseconds, linear network with two or more terminal paths may be
“IF” filter bandwidth B , = 750 cps, used to characterize the variouschannels.Trammission
postdetection filter f m = 375 cps. through a two-terminal pair channel may be represented
by the following equation
Table I shows the performance as a function of A F I B , .
It is seen that this follows the general Characteristics. Edt) = K(O E i d t ) + E,(t). (51)
This equation indicates that the channel may operate
TABTX I
P~ = 10-4 on the signal in a multiplicative fashion, by K(t), and it
may also add to the signal by a certain amount E,(t).
Therefore, there aretwo major categories into which these
effects of the channel may be divided, that is,multi-
plicative and additive disturbances.
The multiplicative type of disturbance is due to a
10 21.6 8 1.0 0.8 8 0.6 5.2
10.8 12.5 fadingtransmissionmedium andantennafluctuations
5 11.6 1.3 8 1.1 8 0.8 3.8
5.8 6.3
2 5.6 8 2.6 2.3 2.8 0.8 8 2.6 2.5 (important for airborne-type a n t e n 1 d 5 ) . T h effect
e of the
0 1.6 8 8.0 8.0 8 0.8 1.4
0.8 0 fading medium can be divided into long and short-term
variations. The long-termvariationsarepractically in-
dependent of frequency and location. Diurnal and yearly
Curve 3 shows the ratio of the range performance of the variations are considered long-term variations. The short-
two-filter linear FSIC tothe two-filternonlinear FSK term variations of the received signal, which may take
systems. This curve shows that the performance is ap- place in a fraction of a second, are generally uncorrelated
proximately equal for A F I B , < 1. Practical considerations a t different frequencies and a t different receiving locations.
may dictate one type of system over another. Theshort-termvariationsare usuallyassociatedwith
Curves 4 and 5 show the ratio of the range performance the changing multipath propagation over a circuit.
of the two-filter nonlinear FSK receiver (Fig. 13) compared The long-term variations
are
associated with the
tothe nonlinear one-filter FSK receiver (Fig. 12) as a seasonal variations in temperature and moisture content
function of A F / B , and for modulation indexes of 1 and 4. of the troposphere or with the changes in the ionosphere
For A F / B , > 2, there is insignificant difference between due to seasonalordaily(night andday)variations.
the two types of receivers. The other type of disturbance experienced in a channel
is the additive type which is present in the absence of an
VII. CHANNEL CHAR.4CTERIZBTION4 input signal. Thesedisturbancesmaybenatural, or
The communication engineer is interested in character- intentionalorunintentionalman-madetype of inter-
izing a communication channel by a mathematical model. ference. Theadditivetype of disturbance thatmay
This model can then be used to determine the effects of be
described by
Gaussian-distributed noise may be
radiopropagation,antenna fluctuations, andadditive receiver noise or cosmic noise. Impulse-type noise is
interference, which may be intentional or unintentional. due t o lightning or
man-made ignitioninterference.
The simplest type of channelis the two-terminalpair Intentional or . nonintentionalinterferencefrom other
which has an input into which signals are sent and an man-made signals is also classified asadditivetype of
output from which signals are received. There are many interference. It is necessary to understand the character-
types of channels such as ground wave systems, line of istics of the different types of interference in order to
sightsystems,systemsemploying reflection from the analyze the performance of the communicationsystem.
ionosphere, ionospheric scatter systems, meteor trail The previousanalysis of the PSK and FSK systems
reflection systems, tropospheric scatter systems, systems was on the basis of only additive Gaussian random noise.
employing scattering
from artificial satellites,
space An excellent discussion is given by Turin“ on the coherent
communications through ionized gas plasmas, etc. and noncoherent receivers for “heavy” fading, (‘moderate’’
The transmission characteristics of some of these fading, and no fading. His analysis suggests that a practical
channelshaveenormous differences. Therefore, it seems system which is to operate through both fixed and non-
questionablewhether it is possible to representthese selective fading paths will be essentially optimal if it is
different channels by a single mathematicnl model.
However, the various channels havethe following common
characteristics: I5 A. B. Glenn and G. Lieberman, “Effect of propagation fading
and antenna fluctuations on communication systems in the presence
of jamming,” Proc.Natl.AeronauticalElectronics Conf., Dayton,
I) They are usually linear. Ohio; May, 1960.
16 G. L. Turin, “E.rror probabilities for binary symmetric ideal
2) They have input(s) and output(s). reception through nonselective slow fading and noise,” PROC.IRE,
3) There are additive disturbances in the output(s). vol. 46, pp. 1603-1619; September, 1958.
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
98 T R AI N
R SEA C T I O N S ON COMMUNICATIONS
SYSTEMS June
designed to useuncorrelated signals and noncoherent
reception exclusively.
The phase comparison system does not havea knowledge
of the phase of the signal at the receiver. Therefore, in
addition t o fading, any additionalperturbation of the
phase of the signaldue toantennafluctuations(par-
ticularlyairborneantennas), or any type of man-made
interference will degrade the performance of the phase
comparison systemmorerapidly thanthe noncoherent
FSK system.
It is then important to have a knowledge of the multi- - -
To To
plicative and additive type of disturbances for a particular Fig. 16-Spectrum of PSK signal.
channel in order t o decide whether a phase comparison
-
system or a noncoherent FSK system should be used.
MA1N 2 NO Low
CARRIER -----c PRODUCT
IF 4 FILTER
VI11. PSK-AM SYSTEM AM DEMOD
PS DETECTOR
AMP
k
f
Pn
-
The signal of, a phase shift keyed amplitude modulated
(PSK-AM) wave is expressed as I ST PHASE
TUNER
- I FA M P
+ mg(t)] coswot
LOCKED
F osc
f(2) = A[1 FI'
where
m = the modulating factor,
A = the amplitude of the carrier,
wo = the main carrier angular velocity, Fig. 17-Coherent PSK-AM receiver, where
g(t) = the modulating function. E, is the open-circuited antenna voltage,
R. is the antenna radiationresistance,
F 1s the receiver noise figure
Thus, (BIF)lis the bandwidth of the first IF amplifier,
9isdetector,
the carrier-to-noise (power) ratio at the inpnt to the AM
(BIF)2is the bandwidth of the second IF amplifier,
is the output filter bandwidth (adjust for maximum SNR),
= 0 elsewhere in time; c' is the signal-to-noise ratio (power) in the second IF amplifier
-
let
e = 0 for a mark
[a]
N after filtering,
0"
is the signal-to-noise ratio (power) in the output amplifer.
t
= n for a qpace.
The frequency spectrum is The spectrum of the PSI<-AM signal in the first.and
second IF amplifiers and in the outputis shown in Fig. 18.
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
1960 Glenn: Comparison of Binary Coded Transmission Systems 99
BANDPASS
~ T ; D E
carriershould
isolation ofbe
the twosufficiently
filters. large to ensurepractical
I , f
(55)
fo- fI fo+ fl
where
REFERENCE
(a) /CARRIER ( C / N ) a = the subcarrier to noise ratio a t point a,
I--BI.--- (BIF)~ +
= ~ ( B I AFz),
AMPLITUDE B, = bandwidth of markspaceor filter,-
MAIN
CARRIER = 2fm',
\ fm = bandwidth of low-pass filter,
AF, = frequency instability of subcarrier.
0 f , t - f If the CNR (PJP,) before the maincarrier envelope
(b)
detector is above the threshold, then
REFERENCE
\--I
CARRIER
L O W PASS FILTER
AFTER SUBCARRIER DETECTION
P
">3 (56)
where
NOISE ( B I F )=
~ 2[2(B1 + AFJI + AF1.
(C)
i Therefore, from (55) and (56),
pig. 18-Spectral densities of PSK-AM signal and noise, (a) first IF,
(b) second IF, (e) output. (57)
pa
-_ -
- c (B1F)Z
-~ E. > 3 (59)
NOISE
Pn N (BIF)l ' Pn
0 kf.iQ
e o
, ~ , ~ ~ where
, + 0 4
~ ~ , +o I
c4'%k-LAh
(b)
~
(BIF)~
= 2(B1 f AFz) + AF1. (60)
Fig. 19-signal and noise spectral densities for an FSK-AM wave,
(a) Prior to maincarrierdetection, (b) prior to detection of Therefore,
subcarrier.
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.
1.00
- 19
AFI
4 + B ,
This shows that for AF,/B, > 4, the CNR will begin
Eqs. (64)-(66) areplottedinFig. 21 as afunction to show significant degradationinthemaincarrier
of AFIR,. Thesecurvesarepracticallyindependent of envelope detector.
Authorized licensed use limited to: SPAWAR System Center Charleston. Downloaded on May 25,2010 at 17:41:49 UTC from IEEE Xplore. Restrictions apply.