Beruflich Dokumente
Kultur Dokumente
3.1 Introduction
Consider the baseband communication system shown in Fig. 3.1. The baseband
message signal X(t) is to be transmitted over a baseband channel.
The channel has a transfer function Hc (f ), and the linear distortion introduced by
the channel is removed by an equalizer with a transfer function
so that the signal component at the destination point is distortionless. That is,
The channel also corrupts the signal with additive noise ni (t) that produces an addi-
tive noise component no (t) at the destination point.
We will assume that the signal X(t) and the front end noise ni (t) are random
processes with the following properties:
2. ni (t) is a stationary, zero mean Gaussian random process with a power spectral
density function Gni (f ).
With these assumptions, we can now proceed to analyze the effects of noise in base-
band communication systems.
The signal quality at the output of analog communication systems is usually mea-
sured by the average signal power to noise power ratio defined as
E{Xo2 (t)}
S
= (3.3)
N d E{n2o (t)}
NOISE IN BASEBAND SYSTEMS 43
In systems designed for transmitting audio signals, this ratio ranges from 10 dB for
barely intelligible voice signals to 30 dB for telephone quality voice signals, and
60 dB for high fidelity audio signals.
For the system shown in Fig. 3.1, if we assume
(
K exp(−j2πf td ), for |f | < fx
Heq (f ) Hc (f ) = (3.4)
0, elsewhere
then we have
and
Z fx
E{Xo2 (t)} = K 2 E{X 2 (t − td )} = K 2 Gx (f ) df (3.6)
−fx
and
η
Gni (f ) = (3.12)
2
the output signal–to–noise ratio is given by
Rf
K 2 −fx x Gx (f ) df
S Sr
= = (3.13)
N d η fx η fx
44 NOISE IN ANALOG COMMUNICATION SYSTEMS
where Sr is the average received signal power. Now, the average transmitted signal
power in this case is given by
Z fx
2
ST = E{X (t)} = Gx (f ) df (3.14)
−fx
K 2 ST
S ST
= = K2 (3.15)
N d η fx η fx
The denominator in Eq. (3.15) is often called the inband noise power and it rep-
resents the noise power in the message bandwidth. The ratio given in Eq. (3.15)
is generally (but not always) an upper bound for analog baseband performance that
may or may not be achieved in practical systems due to the nonideal nature of the
channel itself and other elements in the system such as nonideal filters. Nevertheless,
we will use the ratio given in Eq. (3.15) as a basis for comparing the performance of
other systems.
Z ∞
E{n2o (t)} = Gni (f ) |HR (f )|2 df (3.19)
−∞
and
Z ∞
ST = E{XT2 (t)} = Gx (f ) |HT (f )|2 df (3.20)
−∞
substituting for |HT (f )|2 is from Eq. (3.16), we have the following expression
R∞ R∞
ST E{n2o (t)} −∞
[Gx (f )/|HC (f )HR (f )|2 ]df −∞ Gni (f )|HR (f )|2 df
= R∞ (3.21)
E{Xo2 (t)} −∞ x
G (f ) df
Finally, from Eq. (3.21) we obtain the maximum value for the output signal–to–noise
ratio as
R∞
ST
G (f ) df
S −∞ x
= R 2 (3.26)
N dmax ∞ p
−∞ Gni (f )Gx (f )/|Hc (f )| df
The limits on the integrals may be changed to ±fx , since Gx (f ) = 0 for |f | > fx .
EXAMPLE 3.1
(c) Assume the channel characteristics specified in part (b) and compute (S/N )dmax
for the system that uses optimum preemphasis/deemphasis filters.
SOLUTION
(a) The only filter needed for this case is an ideal lowpass filter at receiving end to
limit the out of band noise. The filter cutoff frequency should be 10 kHz, and
Z 104
E{Xo2 (t)} = (10−8 ) df = 2 × 10−4 watt
−104
Z 104
E{n2o (t)} = (10−14 ) df = 2 × 10−10 watt
−104
S
= 106 = 60 dB
N d
ST = E{Xo2 (t)} = −7 dBm
NOISE IN BASEBAND SYSTEMS 47
(b) For this case (S/N )d is given by Eq. (3.10). With K=1
Hence,
S
= 44.5 dB
N d
ST = E{X 2 (t)} = 2 × 10−4 watt = −7 dBm
(c) With optimum filters, (S/N )d is given by Eq. (3.26). The transfer function of
the optimum terminal filters are (with c = 1)
p p
2 Gx (f ) 2 Gni (f )
|HR (f )| = p , |HT (f )| = p
|Hc (f )| Gni (f ) |Hc (f )| Gx (f )
and
fx
Z p
Gx (f ) Gni (f )
ST = df
−fx |Hc (f )|
and
104
" 4 #
fx
p
Gx Gni
Z Z
−11 f
df = 10 1+ df
−fx |Hc | −104 fc
= 8.4 × 10−7
Hence,
2 × 10−4
S
= = 23.8 dB
N dmax 8.4 × 10−7
and
Now, if ST is raised to 2 × 10−4 , as in case (a) and (b), then (S/N )dmax will
be raised by a factor of 2 × 10−4 /8.4 × 10−7 or by a factor of 23.8 dB. Thus,
for the same transmitter power we have:
48 NOISE IN ANALOG COMMUNICATION SYSTEMS
(c) (S/N )d for a nonideal channel with optimum filters = 47.6 dB.
We are now ready to analyze the performance of linear modulation schemes (DSB,
SSB, AM, and VSB) in the presence of additive noise. In linear modulation schemes,
we have bandpass transmission as opposed to baseband transmission. Models of
narrowband (bandpass) noise developed in the previous chapter will prove to be quite
useful in analyzing the noise performance of CW modulation schemes.
The detector (demodulator) responds to Y (t) and its response is modeled by the
following idealized characteristics:
k1 Yc (t), synchronous detector
k R (t),
2 Y envelope detector
Z(t) = (3.32)
k3 θY (t), phase detector
k dθY (t) ,
frequency detector
4 dt
In Eq. (3.32), k1 , k2 , k3 , and k4 are the detector gains that are often assumed to be
unity.
The detector output is lowpass filtered to remove out of band noise and harmonic
signal terms. The lowpass filter, also referred to as a baseband or post–detection
filter, has a bandwidth of fx Hz.
The analysis we present below for computing the output signal–to–noise ratio is
valid for any given form of Gni (f ). In most practical systems the front–end noise
can be assumed to be white with a psd of
η
Gni (f ) = watt/Hz (3.33)
2
White noise assumption simplifies the calculations considerably and allows us to
look at the conceptually important aspects of analysis.
With additive noise at the input to the detector, it is reasonable to anticipate that
the output of the system will consist of a signal component Xo (t) and an additive
noise component no (t). While this is not the case in general, it is true in most cases
and the output signal–to–noise ratio is obtained by computing the ratio of E{Xo2 (t)}
and E{n2o (t)}.
The output signal quality will depend on the quality of the signal at the detector
input. The input signal quality is measured by the signal–to–noise power ratio at the
detector input, which is defined as
50 NOISE IN ANALOG COMMUNICATION SYSTEMS
K 2 E{Xc2 (t)}
S
= (3.34)
N i BT η
The numerator in the preceding equation is the average signal power at the receiver
input and is denoted by Sr .
and ac = K Ac .
where nc (t) and ns (t) are lowpass Gaussian random processes with power spectral
densities (see Fig. 3.5).
(
η, for |f | < fx
Gnc (f ) = Gns (f ) = (3.41)
0, elsewhere
and
2Y (t) cos(ωc t) = [ac X(t) + nc (t)] + [ac X(t) + nc (t)] cos(2ωc t) − ns (t) sin(2ωc t)(3.43)
Since the cutoff frequency of the post-detection filter is fx and fx << fc , the double
frequency terms are removed by the filter and the output is given by
Substituting a2c Sx/2 = Sr where Sr , is the average power at the receiver input we
have
S Sr
= (3.46)
N d ηfx
52 NOISE IN ANALOG COMMUNICATION SYSTEMS
The reader can verify that the signal–to–noise ratio at the input of the detector is
S Sr
= (3.47)
N i 2ηfx
and
(S/N )d
= αd = 2 (3.48)
(S/N )i
A SSB coherent demodulation system is shown in Fig. 3.6. In this case the modu-
lated signal can be expressed as
where m̂(t) is the Hilbert transform of the message signal m(t). Therefore, the input
to the demodulator is
Expressing the channel bandpass noise in terms of quadrature components, the signal
at the detector input, Y (t), is
Y (t) = [Ac m(t) + nc (t)] cos(2πfc t) + [Ac m̂(t) + ns (t)] sin(2πfc t) (3.51)
Hence,
and, therefor
(S/N )d
=1 (3.56)
(S/N )i
Coherent Demodulation of AM
In AM system, the modulated signal is
where m is the modulation index and x(t) is normalized so that its minimum
value is −1. If a coherent demodulator is employed, the situation is basically
similar to the DSB case, except that we have 1 + m x(t) instead of x(t). There-
fore, in this case, after mixing and lowpass filtering, we have
1
Ỹo (t) = {Ac [1 + m x(t)] + nc (t)} (3.59)
2
However, in this case, the desired signal is x(t), not 1 + m x(t). The dc com-
ponent in the demodulated waveform is removed by a dc blocking device and,
hence, the lowpass filter output is
1 1
Yo (t) = Ac m x(t) + nc (t) (3.60)
2 2
In this case, the received signal power Sr is
A2c
Sr = [1 + m2 Sx ] (3.61)
2
54 NOISE IN ANALOG COMMUNICATION SYSTEMS
where we have assumed that the message signal is zero mean. Now we can
derive the output SNR for the coherent demodulator as
0.25 A2c m2 Sx
S
=
N d 0.5 η fx
A2c m2 Sx
=
2 η fx
m2 Sx A2c /2(1 + m2 Sx )
=
1 + m2 Sx η fx
m2 Sx Sr
=
1 + m2 Sx η fx
2 × m2 Sx
S
=
1 + m2 Sx N i
S
= 2η̂ (3.62)
N i
where η̂ denotes the modulation efficiency.
Envelope Demodulation and Threshold Effect
The input to the detector is then
where
p
RY (t) = {Ac [1 + m X(t)] + nc (t)}2 + [ns (t)]2 (3.65)
and
ns (t)
θY (t) = tan−1 (3.66)
Ac [1 + m X(t)] + nc (t)
The output of the envelope detector will be proportional to RY (t). The expres-
sion for RY (t) given in Eq. (3.65) can be considerably simplified if we assume
that the signal power is either very large or very small compared to the noise
power.
Now, we assume that the signal component in Y (t) is much stronger than the
noise component. With this assumption, we have
which is basically the same as Yo (t) for the coherent demodulation without the
1/2 coefficient. This coefficient, of course, has no effect on the final SNR, and
therefore we conclude that, under the assumption of high SNR at the receiver
input, the performance of coherent and envelope demodulators is the same.
Now, let us assume that at the receiver input the noise power is much stronger
than the signal power. This means that
p
RY (t) = {Ac [1 + m X(t)] + nc (t)}2 + [ns (t)]2
p
= A2 [1 + m X(t)]2 + n2c (t) + n2s (t) + 2Ac nc (t)[1 + m X(t)]
s c
2 2
2Ac nc (t)
≈ [nc (t) + ns (t)] 1 + 2 [1 + m X(t)]
nc (t) + n2s (t)
Ac nc (t)
≈ Rn (t) 1 + [1 + m X(t)]
Rn2 (t)
Ac nc (t)
= Rn (t) + [1 + m X(t)] (3.69)
Rn (t)
where we have used the fact that [1p + m X(t)]2 is small compared to the other
components and we have denoted n2c (t) + n2s (t)√by Rn (t), the envelope of
the noise process, and have used the approximation 1 + ≈ 1+/2, for small
, where
2Ac nc (t)
= [1 + m X(t)] (3.70)
nc (t) + n2s (t)
2
From Eq. (3.69), it is observed that at the demodulator output, the signal and
the noise components are no longer additive and, in fact, the signal component
is multiplied by noise and is no longer distinguishable. In this case, no mean-
ingful SNR can be defined. It is said that this system is operating below the
threshold. The name comes from the fact that there is some value of (SNR)i
above which signal distortion due to noise is negligible and below which the
system performance deteriorates rapidly. The threshold effect does not occur
when coherent demodulation is used.
The threshold value of (SNR)i is usually defined as that value of (SNR)i which
Rn < Ac with probability 0.99. We know that
r2
rn
fRn (rn ) = exp − n , for rn > 0 (3.71)
No 2No
or
EXAMPLE 3.2
SOLUTION
The bandwidth requirements are easl1y computed as
(
20 kHz, for DSB and AM
BT =
10 kHz, for SSB
m2 Sx
S Sr 1 Sr
= =
N d ηfx 1 + m2 Sx 3 ηfx
Before we proceed to derive expressions for the signal–to–noise ratios at the output
of angle modulated systems, let us restate appropriate signal and system models. The
transmitted signal Xc (t) has the form
where
(
kp X(t) for PM
φ(t) = Rt (3.76)
kf −∞
X(τ ) dτ, for FM
As before, the message signal X(t) is normalized such that |X(t)|max = 1. The
phase deviation constant kp is ≤ π for PM so that the message signal can be demod-
ulated from the PM waveform without ambiguity.
The detector is assumed to be ideal. With an input of Y (t) = RY (t) cos(ωc t +
θY (t)), the output of the detector is
(
kd θY (t) for phase detection
Z(t) = 0 dθ (t)
Y
(3.77)
kd dt , for frequency detection
For convenience, we will assume that the detector gain kd is such that
0
kp kd = kf kd = 1 (3.78)
We will use a common approach for analyzing FM and PM cases together and
separate the results by replacing φ(t) by the proper function.
Figure 3.7 Phasor diagram for an angle modulated signal corrupted by additive noise.
where Rn (t) and θn (t) are the envelope and phase of n(t). Using the relationships
shown in Fig. 3.7, we can express Y (t) in envelope and phase form as
Y (t) = RY (t) cos[ωc t + θY (t)] (3.83)
where the angle θY (t) is
θY (t) = φ(t) + θe (t) (3.84)
In Eq. (3.84), θe (t) is the perturbation of the carrier phase angle (which contains the
information signal) due to noise. This perturbation can be from Fig. 3.7 as
−1 Rn (t) sin(θn − φ)
θe (t) = tan (3.85)
Ac + Rn (t) cos(θn − φ)
With the assumption A2c >> E{n2 (t)}, we can expect Rn (t) << Ac most of the
time and hence θe (t) can be approximated by
−1 Rn (t)
θe (t) ≈ tan sin(θn − φ)
Ac
Rn (t)
≈ sin(θn − φ) (3.86)
Ac
the last step being obtained using the approximation tan−1 (α) = α when α << 1.
Combining Eqs. (3.84) and (3.86), we have the following expression for the angle
of the signal plus noise at the input to the detector:
Rn (t)
θY (t) ≈ φ(t) + sin[θn (t) − φ(t)] (3.87)
A
| c
|{z}
signal term
{z }
noise term
We are now ready to calculate the signal–to–noise ratio at the output of angle
modulation systems.
(S/N)d in PM Systems
Using the detector model stated in Eq. (3.77) the output of the phase detector is
Z(t) = kd θY (t)
Rn (t)
= kd φ(t) + kd sin[θn (t) − φ(t)] (3.88)
Ac
NOISE IN ANGLE MODULATION SYSTEMS 59
If we use np (t) to denote the noise term in the preceding equation, then we can
express the detector output as
To calculate the output noise power, we need to derive the spectral characteris-
tics of
Rn (t)
np (t) = kd sin[θn (t) − φ(t)] (3.91)
Ac
Rn (t)
np (t) = kd sin[θn (t)] (3.92)
Ac
or
kd
np (t) = ns (t) (3.93)
Ac
kd2
Gnp (f ) = Gns (f ) (3.94)
A2c
where
Gns (f ) = Gn (f − fc ) + Gn (f + fc )
(
η, for |f | < BT /2
= (3.95)
0, elsewhere
60 NOISE IN ANALOG COMMUNICATION SYSTEMS
The post–detection filter passes the spectral components of np (t) that lie within
|f | < fx , and hence the output noise power is easily computed as
Z fx
E{n2o (t)} = Gnp (f ) df
−fx
kd2
= 2 η fx (3.96)
A2c
The output signal–to–noise ratio for the PM system is given by (using the results
stated in Eqs. (3.90) and (3.96))
E{Xo2 (t)} A2c Sx
S
= = (3.97)
N d E{n2o (t)} kd2 2ηfx
Substituting kp = 1/kd and Sr = A2c /2, we have
S Sr
= kp2 Sx (3.98)
N d ηfx
(S/N)d in FM Systems
The PM detector output may be obtained as
dθY (t)
Z(t) = kd
dt
d Rn (t)
= kf kd X(t) + kd sin[θn (t) − φ(t)] (3.99)
dt Ac
Once again, setting φ(t) = 0 and recognizing that Rn (t) sin[θn (t)] = ns (t),
we get
kd d
Z(t) = kf kd X(t) + [ns (t)]
Ac dt
= X(t) + n1 (t) (3.100)
since kf kd = 1. In the preceding equation, n1 (t) denotes the noise term. The
psd of n1 (t) is given by
kd2
Gn1 (f ) = Gns (f ) (2πf )2
A2c
( k2
2
A2 (2πf ) η,
d
for |f | < BT /2
= c (3.101)
0, elsewhere
The post–detection filter rejects the spectral components of Gn1 (f ) that lie out-
side the message band and the output noise psd is
( k2
2
d
2 (2πf ) η, for |f | < fx
Gno (f ) = Ac (3.102)
0, elsewhere
NOISE IN ANGLE MODULATION SYSTEMS 61
This spectrum is illustrated in Fig. 3.8. The parabolic shape of the spectrum
results from the differentiation action of the FM discriminator. It is clear from
Fig. 3.8 that low–frequency message components are subjected to lower noise
levels than higher frequency message components.
From Eqs. (3.100) and (3.102) we obtain the signal power and the noise power
at the output of the post–detection filter as
and
fx
kd2 2kd2
Z
E{n2o (t)} = 2
(2πf )2 η df = (2π)2 η fx3 (3.104)
−fx Ac 3A2c
3A2c Sx
S
= (3.105)
N d 2(2πkd )2 η fx3
In Eq. (3.107), ∆f /fx is the deviation ratio (D) of the FM system. Comparing
(S/N )d for PM with FM, we find that FM is superior to PM insofar as noise
performance is concerned since (S/N )d for FM can be made as large as desired
by increasing ∆f whereas in PM this increase is limited by the requirement that
kp ≤ π.
62 NOISE IN ANALOG COMMUNICATION SYSTEMS
EXAMPLE 3.3
SOLUTION
(a) FM: We are given that the FM system operates with a deviation ratio ∆f /fx =
5. Hence, BT = 2(∆f + fx ) = 60 M Hz. From Eq. (3.107), we have
2
S ∆f Sr
= 3 Sx
N d fx ηfx
3 × (5)2 × 0.5 × Sr
106 =
10−14 × 5 × 106
1
⇒ Sr =
750
with a channel attenuation of 60 dB, ST = Sr × 106 = 1333 watts.
(b) AM: BT = 2fx = 10 M Hz. With 100% modulation (m = 1) and envelope
demodulation we have
m2 Sx
S Sr
=
N d 1 + m2 Sx η fx
1
3 × Sr
106 = −14
10 ×5× 106
15
⇒ Sr =
100
Hence,
ST = Sr × 106 = 150 kW
PREEMPHASIS/DEEMPHASIS FILTERING IN CW MODULATION SYSTEMS 63
ST = Sr × 106 = 50 kW
In the preceding sections, we saw that the output of a CW modulation scheme operat-
ing above threshold consists of undistorted signal component X(t) plus an additive
noise component no (t). Hence, a CW modulation scheme can be modeled by the
baseband–to–baseband model shown in Fig. 3.9. The psd of the noise, no (t), de-
pends on the modulation and demodulation process used in the system. Two cases
of interest are
and
Equation (3.108) represents the noise psd at the output of a linear modulation scheme
using coherent demodulation, and (3.109) represents the noise psd at the output of
an FM system using a discriminator for demodulation.
The reader should note the similarity between the CW system model shown in
Fig. 3.9 and the baseband system model shown in Fig. 3.1. This similarity suggests
the possibility of using preemphasis/deemphasis filtering in CW modulation systems
for improving the output signal–to–noise ratio. The preemphasis/deemphasis filter
arrangement is shown in Fig. 3.10.
64 NOISE IN ANALOG COMMUNICATION SYSTEMS
Equation (3.113) requires that the normalized average power of the baseband sig-
nal X(t) is the same as the normalized power of the preemphasized signal Xp (t).
This constraint assures that for linear modulation schemes the transmitted power is
not altered by preemphasis filtering. For FM systems, equal power requirement en-
sures that the bandwidth of the FM signal remains the same.
The signal power at the output is the same with or without preemphasis/deemphasis
filtering. The noise power at the output without filtering is
Z fx
E{n2o (t)} = Gno (f ) df (3.114)
−fx