Noise in AM PDF

Chapter 5
Effect of Noise on Analog Communication

Systems
NCCU
Wireless Comm. Lab.
Outline
) Chapter 5 Effect of Noise on Analog Communication Systems
A 5.1 Effect of Noise on linear-Modulation Systems
¾ 5.1.1 Effect of noise on a baseband system
¾ 5.1.2 Effect of noise on DSB-SC AM
¾ 5.1.3 Effect of noise on SSB AM
¾ 5.1.4 Effect of noise on conventional
A 5.2 Carrier-Phase Estimation with a phase-Locked Loop (PLL)
¾ 5.2.1 The phase-locked loop (PLL)
¾ 5.2.2 Effect of additive noise on phase estimation
A 5.3 Effect of noise on angle modulation
¾ 5..3.1 Threshold effect in angle modulation
¾ 5.3.2 Pre-emphasis and De-emphasis filtering
A 5.4 Comparison of analog-modulation systems
CCU
2 Wireless Access Tech. Lab.
Outline
A 5.5 Effects of transmission losses and noise in analog
communication systems
¾ 5.5.1 Characterization of thermal noise sources
¾ 5.5.2 Effective noise temperature and noise figure
¾ 5.5.3 Transmission losses
¾ 5.5.4 Repeaters for signal transmission
A 5.6 Further reading
CCU
5 Effect of Noise on Analog Communication Systems
) The effect of noise on various analog communication systems
will be analysis in this chapter.
) Angle-modulation systems and particularly FM, can provide
a high degree of noise immunity, and therefore are desirable
in cases of severe noise and/or low signal power.
) The noise immunity is obtained at the price of sacrificing
channel bandwidth because he bandwidth requirements of
angle-modulation systems is considerably higher than
amplitude-modulation systems.
CCU
5.1 Effect of Noise on linear-Modulation Systems
(1) Baseband system
) (1) Baseband system
N
u(t) r(t)
m(t)
N0
Sn ( f ) = LPT
2
+w N0
Pn0 = ∫ df = N 0W
−w 2 -W W
PR : received power
⎛S⎞ PR PR
Then ⎜ ⎟ = =
⎝ N ⎠b PN0 N 0W
baseband
CCU
(2) DSB-SC AM
) (2) DSB-SC AM
N
u(t)
r(t)
m(t) DSB-SC DSB-SC

modulation Demodulation
u ( t ) = Ac m ( t ) cos ( 2π f ct + φc )
r (t ) = u (t ) + n (t )
= Ac m ( t ) cos ( 2π f c t + φc )
+ nc ( t ) cos ( 2π f ct ) − ns ( t ) sin ( 2π f ct )
CCU
(2) DSB-SC AM
) Coherent demodulation
r ( t ) cos ( 2π f ct + φ )
= Ac m ( t ) cos ( 2π f ct + φc ) cos ( 2π f ct + φ )
+ nc ( t ) cos ( 2π f ct ) cos ( 2π f ct + φ )
− ns ( t ) sin ( 2π f ct ) cos ( 2π f ct + φ )
1 1
= Ac m ( t ) cos (φc − φ ) + ⎡⎣nc ( t ) cos(φ ) + ns ( t ) sin (φ ) ⎤⎦
2 2
1
+ Ac m ( t ) cos ( 4π f ct + φ + φc )
2
1
+ ⎡⎣nc ( t ) cos(4π f ct + φ ) − ns ( t ) sin ( 4π f ct + φ ) ⎤⎦
2
CCU
(2) DSB-SC AM
) y t = 1 A m t cos φ − φ + 1 n t cos φ + n t sin φ
() c ( ) ( c ) ⎡⎣ c ( ) ( ) s ( ) ⎤⎦
2 2
) If PLL is employed ⇒ φ = φc ⇒ coherent or synchronon demod
1 1
) Assume φ = φc = 0 ⇒ y ( t ) = Ac m ( t ) + nc ( t )
2 2
2
A
) Message power : P0 = c E ⎡⎣ M 2 ( t ) ⎤⎦
4
2
Ac
= R M (0 )
4
2
Ac
= PM
4
CCU
(2) DSB-SC AM
1
) Noise power : Pn0 = Pnc
4
Note that : Pnc = Pns = Pn

nw(t) n(t)
BPF
∵ Sn ( f ) = SW ( f ) H ( f )
2
⎧ N0
⎪ , f - fc < W
=⎨ 2
-fc fc
⎪⎩0, otherwise
∞ N0
∴ Pn = ∫ Sn ( f ) df = ⋅ 4W = 2WN 0
−∞ 2
CCU
(2) DSB-SC AM
2
) Ac
P 2
⎛ S ⎞ P0 4
M Ac PM
∴⎜ ⎟ = = =
⎝ N ⎠ 0 Pn0 1
2WN 0 2WN 0
4
2 2
Ac A c PM
∵ PR = E ⎡⎣ M ( t ) ⎤⎦ =
2
2 2
⎛ S ⎞ PR ⎛ S ⎞
∴⎜ ⎟ = =⎜ ⎟
⎝ N ⎠ 0, DSB N 0W ⎝ N ⎠ b
CCU
(3) SSB-AM
) (3) SSB-AM
u ( t ) = Act m ( t ) cos ( 2π f ct ) ± Ac mˆ ( t ) sin ( 2π f c t )

∴r ( t ) = u ( t ) + n ( t )
= ⎡⎣ Act m ( t ) + nc ( t ) ⎤⎦ cos ( 2π fct ) + ⎡⎣± Acmˆ ( t ) − ns ( t ) ⎤⎦ sin ( 2π fct )
with ideal-phase coherent demodulator

1 1
y ( t ) = LPF ⎡⎣ r ( t ) cos ( 2π f c t ) ⎤⎦ = Ac m ( t ) + nc ( t )
2 2
CCU
(3) SSB-AM
1 2
⇒ P0 = Ac PM
4
1 1
Pn0 = Pnc = Pn
4 4
∞ N0
Pn = ∫ Sn ( f ) df = ⋅ 2W = WN 0
-∞ 2
2
⎛ S ⎞ P0 Ac PM
∴⎜ ⎟ = =
⎝ N ⎠ 0, SSB Pn0 WN 0
2 2
Ac Ac
but PR = PM + PMˆ = Ac PM
2
2 2
⎛S⎞ ⎛S⎞
∴⎜ ⎟ =⎜ ⎟
⎝ N ⎠ 0,SSB ⎝ N ⎠ b
CCU
(4) Conventional AM
) (4) Conventional AM
u ( t ) = Ac ⎡⎣1 + amn ( t ) ⎤⎦ cos ( 2π f c t )
r ( t ) = ⎡⎣ Ac ⎡⎣1 + amn ( t ) ⎤⎦ + nc ( t ) ⎤⎦ cos ( 2π f c t ) − ns ( t ) sin ( 2π f c t )
for synchronous demodulation (similar to DSB, except using

⎡⎣1 + am n ( t ) ⎤⎦ instead of m(t) )
1 1
∴ y ( t ) = Ac ⎡⎣1 + a m n ( t ) ⎤⎦ + n c ( t )
2 2
by a dc block
1 1
y ( t ) = Ac a m n ( t ) + n c ( t )
2 2
CCU
(4) Conventional AM
1 2 2 1 1 N 1
∴ P0 = Ac a PM n , Pn0 = pnc = ⋅ 0 ⋅ 4W = N 0W
4 4 4 2 2
1 2 2
Ac a PM n A 2a 2 P
⎛ ⎞
S
∴⎜ ⎟ = 4 =
c Mn
⎝ N ⎠0 1
N 0W 2 N 0W
2
2
Ac
but P R = 2 ⎡⎣ 1 + a P M n ⎤⎦
2
⎛ S ⎞ a 2
P A
2
⎡
⎣ 1 + a 2
PM n ⎦⎤
∴⎜ = ⋅
M n c
⎟
⎝ N ⎠ 0,AM 1 + a PM n
2
2 N 0W
a 2 PM n PR a 2 PM n ⎛ S ⎞
= ⋅ = ⋅ ⎜ ⎟
1 + a 2 PM n N 0W 1 + a 2 PM n ⎝ N ⎠ b
⎛ S ⎞ ⎛ S ⎞
=η ⎜ ⎟ < ⎜ ⎟
⎝ N ⎠b ⎝ N ⎠b
CCU
(4) Conventional AM
In practical application : a ∼ 0.8 ↔ 0.9

PM N ∼ 0.1
⎛S⎞ ⎛S⎞
⎜ ⎟ ≈ 0.075 ⎜ ⎟ ,
⎝ N ⎠0, AM 1 ⎝ N ⎠b
⎛S⎞ ⎛S⎞
i.e. ⎜ ⎟ − ⎜ ⎟ ≈ 11dB( a lost of 11dB )
⎝ N ⎠0,dB ⎝ N ⎠0, AM 1,dB
CCU
(4) Conventional AM
Envelope detector
{ }
r ( t ) = Ac ⎡⎣1 + amn ( t ) ⎤⎦ + nc ( t ) cos ( 2π f ct ) − ns ( t ) sin ( 2π f c t )
∴Vr ( t ) = ⎣ Ac ⎡⎣1 + amn ( t ) ⎤⎦ + nc ( t ) ⎦ + ns ( t )

2
⎡ ⎤ 2
CCU
(4) Conventional AM
A Case 1 Ac (1 + amn (t ) ) n ( t )
P (n (t ) Ac (1 + amn (t ) ) ) ≈ 1,
P ( ns ( t ) Ac (1 + amn (t ) ) ) ≈ 1, (i.e. A (1 + am (t ) )
c n n (t ))
⇒ Vr ( t ) ≈ Ac (1 + amn (t ) ) + nc ( t )
After removing the dc component,
y ( t ) = Ac ( amn (t ) ) + nc ( t )
The same as y(t) for the synchronous demodulation, without the ½
coefficient
⎛S⎞ ⎛S⎞
∴ under high SNR ⇒ ⎜ ⎟ =⎜ ⎟
⎝ N ⎠0, AM 2 ⎝ N ⎠0, AM 1
CCU
(4) Conventional AM
A Case 2 Ac (1 + amn (t ) ) n ( t )
Vr ( t ) = ( A (1 + am n ( t ) ) + nc ( t ) ) + ns ( t )
2 2
c
Ac (1 + amn ( t ) ) + nc ( t ) + ns ( t ) + 2 Ac nc ( t ) (1 + amn ( t ) )
2
=
2 2 2
⎡ 2 Ac nc ( t ) ⎤
≈ ⎣ nc ( t ) + ns ( t ) ⎦ ⎢1 + 2
⎡ 2 2
⎤ (1 + amn ( t ) )⎥
⎢⎣ nc ( t ) + ns ( t )
2
⎥⎦
⎡ Ac nc ( t ) ⎤
≈ Vn ( t ) ⎢1 + 2
Vn ( t )
(1 + amn ( t ) )⎥
⎢⎣ ⎥⎦
Ac nc ( t )
≈ Vn ( t ) +
Vn ( t )
(1 + amn ( t ) )
= Vn ( t ) + Ac (1 + amn ( t ) ) cosθ n ( t )
CCU
(4) Conventional AM
Notes: i) Signal and noise components are no longer additive.
ii) Signal component is multiple by the noise and is no
longer distinguishable.
iii) no meaningful SNR can be defined.
Æ It is said that this system is operating below the threshold

==> threshold effect
CCU
)
CCU
CCU
CCU
5.2 Carrier-Phase Estimation with a phase-Locked Loop (PLL)
) Consider DSB-SC AM
r ( t ) = u ( t ) + n ( t ) = Ac m ( t ) cos ( 2π f c t + φc ) + n ( t )
assume E ⎡⎣ m ( t ) ⎤⎦ = 0 , zero-mean (i.e. no dc component)
⇒ the average power at the output of a narrow band filter

tuned to the carrier frequency fc is zero.
CCU
r ( t ) = Ac m 2 ( t ) cos2 ( 2π f c t + φc ) + noise terms

2
Let 2
1 2 2 1 2 2
= Ac m ( t ) + Ac m ( t ) cos ( 4π f ct + 2φc ) + noise terms
2 2
∵ E ⎡⎣ m 2 ( t ) ⎤⎦ = Rm ( 0 ) > 0
∴ There is signal power at the frequency 2π f c , which can

be used to drive a PLL.
CCU
) The mean value of the output of Bandpass filter is a sinusoid

with frequency 2 f c , phase 2φc , and amplitude
H (2 fc )
Ac E ⎡⎣ m ( t ) ⎤⎦
2 2
2 .
CCU
5.2.1 The phase-locked loop (PLL) )
)
) If input to the PLL is cos ( 4π f c t − 2φ ) and output of the

( )
VCO is sin 4π f c t − 2φˆ , φˆ represents the estimate of φ ,
(
then e ( t ) = cos ( 4π f t − 2φ ) sin 4π f t − 2φˆ
c c )
= sin 2 (φ − φ ) + sin( 2π f t −2φˆ −2φ)
1 ˆ 1
c
2 2
CCU
) Loop filter is a lowpass filter, dc is reserved as 4 f c is removed.
1 + τ 2s
) It has transfer function G ( s ) = (τ 1 τ 2 )
1 + τ 1s
) n(t) provide the control voltage for VCD (see section 5.3.3).
) The VCO is basically a sinusoidal signal generation with an
instantaneous phase given by
4π f ct − 2φ = 4π f c t − kv ∫ v (τ ) dτ
t
ˆ
−∞
where kv is a gain constant in radians/ volt-sec.
CCU
) The carrier-phase estimate at the output of VCO is :
∫ v (τ ) dτ
t
2φˆ = k v
−∞
kv
and its transfer function is .
s
) Double-frequency terms resulting from the multiplication of

the input signal to the loop with the output of the VCD is
removed by the loop filter
⇒ PLL can be represented by the closed-loop system model
as follows.
CCU
)
CCU
) In steady-state operation, when the loop is tracking the phase
of the received carrier, φˆ − φ is small.
1
( )
∴ sin 2 φˆ − φ ≈ φˆ − φ
2
) With this approximation, PLL is represented the linear model
shown below.
CCU
⇒ closed-loop transfer function
kG ( s ) / s
H (s) =
1 + kG ( s ) / s
where the factor of ½ has been absorbed into the gain
parameter k
1 + τ 2s
∵G ( s) =
1 + τ 1s
1 + τ 2s
⇒ H (s) =
⎛ 1 ⎞ τ1 2
1 + ⎜τ 2 + ⎟ s + s
⎝ k⎠ k
CCU
) The closed-loop system function of the linearized PLL is
second order when the loop filter has a signal pole and signal
zero.
) The parameter τ 2 determines the position of zero in H(s),
while K, τ 1 and τ 2 control the position of the closed-loop
system poles.
CCU
) The denominator of H(s) may be expressed in the standard
form
D ( s ) = s + 2ζ wn s + wn
2 2
where ζ : loop-damping factor

wn : natural frequency of the loop
ζ = wn ⎛⎜τ 2 + ⎞⎟ 2
k 1
wn = ,
τ1 ⎝ k⎠
∴ H (s) =
( 2ζ w n − wn / k s + wn
2
) 2
s + 2ζ wn s + wn
2 2
CCU
) The magnitude response 20log H ( jw ) as a function of the
normalized frequency w wn is illustrated, with the damping
factor as a parameter and τ 1 1 .
CCU
)
ζ = 1 → a critical damped loop response
ζ < 1 → a under-damped loop response
ζ > 1 → a over-damped loop response
) The one-side noise equivalent bandwidth
Bneq =
(
τ 2 1/ τ 2 + K / τ 1
2 2
) = 1 + (τ w )
2 n
2
4 (τ 2 + 1
K ) 8ζ / wn
) Trade-off between speed of response and noise in the phase
estimate
CCU
5.2.2 Effect of additive noise on phase estimation
) PLL is tracking a signal as
s ( t ) = Ac cos ( 2π f c t + φ ( t ) )
which is corrupted by additive narrowband noise

n ( t ) = nc ( t ) cos ( 2π f c t ) − ns ( t ) sin ( 2π f c t )
nc ( t ) , ns ( t ) are assumed to be statistically independent
stationary Gaussian noise.
n ( t ) = xc ( t ) cos ( 2π f ct + φ ( t ) ) − xs ( t ) sin ( 2π f ct + φ ( t ) )
where xc ( t ) = nc ( t ) cos (φ ( t ) ) + ns ( t ) sin (φ ( t ) )
xs ( t ) = -nc ( t ) sin (φ ( t ) ) + ns ( t ) cos (φ ( t ) )
xc ( t ) + jxs ( t ) = ⎡⎣ nc ( t ) + jns ( t ) ⎤⎦ e
− jφ ( t )
CCU
) Problem 4.29 Æ a phase shift does not change the first two
moments of nc(t) and ns(t).
i.e. xc(t) and xs(t) have exactly the same statistical
characteristics as nc(t) and ns(t).
{ ( )}
e ( t ) = LPF ⎡⎣ s ( t ) + n ( t ) ⎤⎦ 2sin 2π f c t + φˆ ( t )
= LPF {( A + x ( t ) ) sin ( φ ) + ( A + x ( t ) ) sin ( 4π f t + φ ( t ) + φˆ ( t ) )

c c c c c
− x ( t ) cos ( φ ) + x ( t ) cos ( 4π f t + φ ( t ) + φˆ ( t ) )}
s s c
= ( Ac + xc ( t ) ) sin ( φ ) − xs ( t ) cos ( φ )
φ = φˆ − φ ,
e (t ) x c (t ) x s (t )
= s in ( φ )+ s in ( φ )− cos ( φ)
Ac Ac Ac
CCU
) The equivalent model is shown as below
CCU
2
A
) When the power Pc = c of the incoming signal is much
2
larger than the noise power, φˆ ≈ φ .
) Then we may linearize the PLL shown as bellow.
CCU
xc ( t ) xs ( t )
n1 ( t ) = sin ( φ ) − cos ( φ )
Ac Ac
∵ n1 ( t ) is additive at the input to the loop, the variance of the
phase error φ , which is also the variance of the VCO
output phase is
N 0 Bneq
σ φˆ =
2
2
Ac
Bneq : one-sided noise equivalent bandwidth of the loop, given by
Bneq =
(
τ 2 2 1/ τ 2 2 + K / τ 1 ) = 1 + (τ w )
2 n
2
4 (τ 2 + 1
K ) 8ζ / wn
CCU
σ
2 2
) Note that Ac
2
is the power of the input sinusoid, and φˆ is
simply proportional to the total noise power with the
bandwidth of the PLL divided by the input signal power,
hence
1
σ φˆ =
2
ρL
where ρ L is defined as the SNR
Ac 2
ρL = 2
Bneq ⋅ N20
Thus, the variance of φˆ is inversely proportional to the SNR.
CCU
) Note : the variance of linear
model is close to the exact
variance for ρ L > 3
CCU
CCU
y ( t ) = r ( t ) = ( u ( t ) + n ( t ) ) = u 2 ( t ) + 2u ( t ) n ( t ) + n 2 ( t )
2 2
noise terms
) By computing the autocorrelation and power spectral density

of these two noise component, one can show that both
components have spectral power in frequency band centered
at 2 fc.
) Let’s select Bneq << Bbp, then total noise spectrum at the input
to PLL may be approximated by a constant with the loop
bandwidth.
CCU
) This approximation allows us to obtain a simple expression
for the variance of the phase error as
1
σ φˆ =
2
ρ LSL
where SL is called the squaring loss and is given as
1
SL =
1+
B bp 2 B neq
ρL
since SL<1 , we have an increase in the variance of the phase

error caused by the added noise power that results from the
squaring operation.
) E.g. if 2 Bneq = ρ L , S L = 12 the loss is 3dB or equivalently, the
Bbp
variance in the estimate increase by a factor of 2.
CCU
) Costas loop
CCU
r ( t ) = Ac m ( t ) cos ( 2π f ct − φ ) + n ( t )
yc ( t ) = ⎡⎣ Ac m ( t ) cos ( 2π f ct − φ ) + nc ( t ) cos ( 2π f ct )
(
− ns ( t ) sin ( 2π f c t ) ⎤⎦ cos 2π f ct − φˆ )
Ac 1
= m ( t ) cos ( φ ) + nc cos φˆ
2 2
()
1
()
− ns ( t ) sin φˆ + double frequency terms
2
CCU
)
y s ( t ) = ⎡⎣ Ac m ( t ) cos ( 2π f c t + φ ) + nc ( t ) cos ( 2π f c t )
(
− ns ( t ) sin ( 2π f c t ) ⎤⎦ sin 2π f ct − φˆ )
Ac 1
= m ( t ) sin ( φ ) + nc sin φˆ
2 2
()
1
()
+ ns ( t ) cos φˆ + double frequency terms
2
CCU
φ = φˆ − φ
e ( t ) = yc′ y s′
Ac m 2 ( t )
2
= sin ( 2 φ )
4
Ac m ( t ) ⎡
+
4 ⎣
()
nc ( t ) cos φˆ − ns ( t ) sin (φˆ ) ⎤ sin ( φ )
⎦
Ac m ( t ) ⎡
+
4 ⎣
()
nc ( t ) sin φˆ + ns ( t ) cos (φˆ ) ⎤ cos ( φ )
⎦
1
()
+ ⎡ nc ( t ) cos φˆ − ns ( t ) sin ( φˆ) ⎤ ⎡⎣ nc ( t ) sin ( φˆ) + ns ( t ) cos (φˆ ) ⎤⎦
4⎣ ⎦
CCU
) These terms (signal × noise and noise × noise) are similar to
the two noise terms at the input of the PLL for the squaring
method.
) In fact, if the loop filter in the Costas loop is identical to that
used in the squaring loop, the two loops are equivalent.
) Under this condition the pdf of the phase error, and the
performance of the two loops are identical.
) In conclusion, the squaring PLL and the Costas PLL are two
practical methods for deriving a carrier-phase estimation for
synchronous demodulation of a DSB-SC AM signal.
CCU
5.3 Effect of noise on angle modulation
) RX for a general angle-modulated signal
u ( t ) = Ac cos ( 2π f c t + φ ( t ) )
⎪ c
=⎨
( c
−∞ )
⎧ A cos 2π f t + 2π k t m (τ ) dτ , FM
f ∫
⎪ Ac cos ( 2π f ct + k p m ( t ) ) , PM
⎩
CCU
r ( t ) = u ( t ) + nc ( t ) cos ( 2π f c t ) − ns ( t ) sin ( 2π f ct )
n ( t ):bandpass Gaussian noise
) Assume: signal power is much higher than noise power
P ⎡⎣Vn ( t )
i.e.
Ac ⎤⎦ ≈ 1
⎛ ⎛ ns ( t ) ⎞ ⎞
) n ( t ) = nc ( t ) + ns ( t ) cos ⎜ 2π f c t + tan ⎜
2 2 −1
⎟ ⎟⎟
⎝ nc ( t ) ⎠ ⎠
⎜
⎝
⎛ ⎞
= Vn ( t ) cos ⎜ 2π f c t + Φ n ( t ) ⎟
⎜ ⎟
Envelope ⎝ phase ⎠
CCU
CCU
∴ r ( t ) = ⎡⎣ Ac + Vn cos ( Φ n ( t ) − φ ( t ) )⎤⎦ ×
⎛ ⎛ Vn ( t ) sin ( Φ n ( t ) − φ ( t ) ) ⎞ ⎞
cos ⎜ 2π f ct + φ ( t ) + tan ⎜
−1
⎟
⎜ ⎜ Ac + Vn ( t ) cos ( Φ n ( t ) − φ ( t ) ) ⎟⎟ ⎟
⎝ ⎝ ⎠⎠
≈ ⎡⎣ Ac + Vn cos ( Φ n ( t ) − φ ( t ) )⎤⎦ ×
⎛ Vn ( t ) ⎞
cos ⎜ 2π f ct + φ ( t ) + sin ( Φ n ( t ) − φ ( t ) ) ⎟
⎝ Ac ⎠
Note that
⎧k p m ( t ) PM
⎪
φ (t ) = ⎨
⎪⎩2π k f ∫−∞ m (τ ) dτ FM
t
CCU
) The output of the demodulator is
⎧⎪ k p m ( t ) + Yn ( t ) , PM
y (t ) = ⎨
⎪⎩ f
k m ( t ) + 2π dt Yn ( t ) ,
1 d
FM
⎧ k p m ( t ) + VnA(t ) sin ( Φ n ( t ) − φ ( t ) ) , PM
⎪
=⎨
c
⎪⎩k f m ( t ) + 2π dt Ac sin ( Φ n ( t ) − φ ( t ) ) , FM
1 d Vn ( t )
where
sin ( Φ n ( t ) − φ ( t ) )
Vn ( t ) 1
Yn ( t ) Ac ∝
Ac
(noise component)
CCU
) Noise component
sin ( Φ n ( t ) − φ ( t ) )
Vn ( t )
Yn ( t ) = Ac
= 1
Ac
⎡⎣Vn ( t ) sin ( Φ n ( t ) ) cos (φ ( t ) ) − Vn ( t ) cos ( Φ n ( t ) ) sin (φ ( t ) )⎤⎦
= 1
Ac
⎡⎣ ns ( t ) cos (φ ( t ) ) − nc ( t ) sin (φ ( t ) )⎤⎦
) The autocorrelation function

E ⎡⎣Yn ( t + τ ) Yn ( t ) ⎤⎦ = 2 E ⎣⎡( Rns (τ ) cos (φ ( t ) ) cos (φ ( t + τ ) )
1
Ac
+ Rnc (τ ) sin (φ ( t + τ ) ) sin (φ ( t ) )⎤⎦
Rnc (τ ) E ⎡⎣cos (φ ( t + τ ) ) − φ ( t ) ⎤⎦
1
= 2
Ac
CCU
) Assume m(t) is a sample function of a zero-mean stationary
Gaussian process M(t) with autocorrelation RM (τ )
then
⎧⎪ k p M (t ) PM
Φ (t ) = ⎨
⎪⎩2π k f ∫−∞ M (τ ) dτ FM
t
) Φ ( t ) is a zero-mean, stationary Gaussian process
CCU
) At any fixed time t, the random variable Z ( t ,τ ) = Φ ( t + τ ) − Φ ( t )
is the difference between two jointly Gaussian random
variables.
) Z ( t ,τ ) is a Gaussian R.V. with zero mean and variance
σ Z 2 = E ⎡⎣Φ 2 ( t + τ ) ⎤⎦ + E ⎡⎣Φ 2 ( t ) ⎤⎦ − 2 RΦ (τ )
= 2 ⎡⎣ RΦ ( 0 ) − RΦ (τ ) ⎤⎦
CCU
) Autocorrelation function
R Y n (τ )=E ⎡⎣ Y n ( t + τ ) Y n ( t ) ⎤⎦
R n c (τ ) E ⎡⎣ c o s ( Φ ( t + τ ) )⎤⎦
1
= 2 )− Φ (τ
Ac
=
1
R n c (τ ) R e ⎡⎣⎢ E ⎡ e j ( Φ ( t + τ )− Φ (τ ) ) ⎤ ⎤
Ac
2
⎣ ⎦ ⎦⎥
1
R n c (τ ) R e ⎡⎣ E jZ ( t ,τ )⎤⎤
= ⎡e
Ac
2 ⎣ ⎦⎦
1 ⎡ − 12 σ z ⎤
R n c (τ ) R e ⎢⎣ e
2
= 2 ⎥⎦
Ac
) R e ⎡⎣ e − ( R ( 0 ) − R Φ (τ ) ) ⎤
1
= R n c (τ Φ
Ac
2
⎦
) e − (R ( 0 ) − R Φ (τ ) )
1
= 2
R n c (τ Φ
Ac
CCU
) Power spectral density
SY (f )= F ( RY (τ ) )
⎡ 1 − ( R Φ ( 0 ) − R Φ (τ ) ) ⎤
= F ⎢ 2 R n c (τ ) e ⎥
⎣ Ac ⎦
- RΦ ( 0 )
e
= 2
F ⎡⎣ R n c (τ ) g (τ ) ⎤⎦
Ac
- RΦ ( 0 )
e
= 2
S nc ( f )∗G ( f )
Ac
where g (τ )= R Φ (τ )
e and G(f) is its Fourier transform.
CCU
) The bandwidth of g (τ ) is half the bandwidth Bc of the angle-
modulated signal, which for high modulation indices is much
larger than W, the message bandwidth.
) Since the bandwidth of the angle-modulated signal is defined

as the frequencies that contain 98%-99% of the signal power,
G(f) is very small in the neighborhood of f = B2 and
c
⎧ N0 , f < B2c
Snc ( f ) = ⎨
⎩ 0, otherwise
CCU
) A typical example of G(f), Snc ( f ) and the result of their
convolution is shown below.
CCU
) Because G(f) is very small in the neighborhood of f = 2 ,
B c
the resulting SY(f) has almost a flat spectrum for f < W , the
bandwidth of the message.
− RΦ ( 0 ) − RΦ ( 0 ) Bc
e e
SY ( f ) = Snc ( f ) ∗ G ( f ) = N0 ∫ G ( f ) df
2
2 2 −
Bc
Ac Ac 2
− RΦ ( 0 ) − RΦ ( 0 )
e ∞ e
≈ 2
N 0 ∫ G ( f ) df = 2
N 0 g (τ ) τ =0
Ac −∞ Ac
− RΦ ( 0 )
e RΦ ( 0 ) N0
= 2
N 0e = 2
Ac Ac
i.e. for | f |<W , the spectrum of the noise components in the PM
and FM. ⎧⎪ AN , PM 0
Sn0 ( f ) = ⎨ N 2
2
c
⎪⎩ Ac2 f , FM
0
CCU
) The power spectrum of the noise component at the output of
the demodulator in the frequency interval | f | < W for PM
and FM is shown below.
CCU
) PM has a flat noise spectrum and FM has a parabolic noise
spectrum.
) The effect of noise in FM for higher-frequency components is
much higher than the effect of noise on low-frequency
components.
) The noise power at the output of the lowpass filter is the noise
power in the frequency range [ -W, +W ].
+W
Pn0 = ∫ Sn0 ( f ) df
−W
⎧ +W N 0 ⎧ 2WN 0
⎪ ∫−W A 2 df , PM ⎪ A 2 , PM
⎪ ⎪
=⎨ =⎨
c c
+W N 3
⎪ ⎪ 2W N0
∫ 0 2
f df , FM , FM
⎪⎩ − W Ac
2
⎪⎩ 3 Ac 2
CCU
) Output signal power
⎧⎪ k p 2 PM , PM
Ps0 = ⎨ 2
⎪⎩ k f PM , FM
) The SNR
⎧ k p 2 Ac 2 PM
⎪ ⋅ , PM
⎛S⎞ Ps0 ⎪ 2 N 0W
⎜ ⎟ = =⎨ 2 2
⎝ N ⎠0 Pn0 ⎪ 3k f Ac PM
⎪ 2W 2 ⋅ N W , FM
⎩ 0
CCU
) Note that Ac 2 is the received signal power, denote by PR.
2
) ⎧ β p = k p max m ( t ) , PM
⎪
⎨ k f max m ( t )
⎪β f = . FM
⎩ W
) The output SNR
⎧
( )
2
βp PM
⎪ PR max m ( t )
, PW
⎛S⎞ ⎪ N 0W
⎜ ⎟ =⎨
⎝ N ⎠ 0 ⎪ 3P
( )
2
βf PM
max m ( t )
, FM
⎪⎩ R
N 0W
CCU
⎛S⎞ PR
) Let ⎜ ⎟ =
⎝ N ⎠b N 0W
) The output SNR of a baseband system with the same received

power
⎧
( ) ⎛S⎞
2
βp
⎪ PM max m ( t ) ⎜ ⎟ , PW
⎛S⎞ ⎪ ⎝ N ⎠b
⎜ ⎟ =⎨
⎝ N ⎠0 ⎪
( ) ⎛S⎞
2
βf
⎪⎩
3PM max m ( t ) ⎜ ⎟ , FM
⎝ N ⎠b
CCU
PM
) Note that in the above expression ( max m(t ) )2 is the average-to-
peak power ratio of the message signal (or, equivalently, the
power content of normalized message, PM ).
) Therefore,
⎧ 2 ⎛S⎞
⎪ β p PM n ⎜ N ⎟ , PW
⎛S⎞ ⎪ ⎝ ⎠b
⎜ ⎟ =⎨
⎝ N ⎠0 ⎪ ⎛S⎞
3β f PM n ⎜ ⎟ , FM
2
⎪⎩ ⎝ N ⎠b
CCU
) Carson’s rule Bc = 2 ( β + 1) W
) The ration of the channel bandwidth to the message
bandwidth
Bc
Ω= = 2 ( β + 1)
W
) We can express the output SNR in terms of the bandwidth

expansion factor
⎧
( ) ⎛ S ⎞
Ω 2
−1
⎪ PM ⎜ ⎟ , PW
2
m ax m ( t )
⎛ S ⎞ ⎪ ⎝ N ⎠b
⎜ ⎟ =⎨
⎝ N ⎠0 ⎪
( ) ⎛ S ⎞
Ω 2
−1
3P ⎜ ⎟ , FM
2
m ax m ( t )
⎪⎩ M ⎝ N ⎠b
CCU
) In both PM and FM, the output SNR is proportional to the
square of the modulation index β .
A Therefore, increasing β increases the output SNR even with
low received power.
A This is in contrast to amplitude modulation where such an
increase in the received SNR is not possible.
) The increase in the received SNR is obtained by increasing
the bandwidth.
A Therefore, angle modulation provides a way to trade-off
bandwidth for transmitted power.
) The relation between the output SNR and the bandwidth
expansion factor, Ω , is a quadratic relation.
A This is far from optimal
CCU
) In fact, if we increase β such that the approximation
( p (Vn ( t ) Ac ) ≈ 1 ) does not hold, a phenomenon known as the
threshold effect will occur, and the signal will be lost in the
noise.
) In both PM and FM the output SNR is proportional to the
square of the modulation index β . Therefore, increasing β
increases the output SNR even with low received power.
CCU
) A comparison of the above result with the SNR in amplitude
modulation shows that, in both case, increasing the
transmitter power will increase the output SNR, but the
mechanisms are totally different.
) In AM, any increase in the received power directly increases
the signal power at the output of the receiver.
) In angle modulation the message is in the phase of the
modulated signal and, consequently, increasing the
transmitter power does not increase the demodulated message
power. In angle modulation what increases the output SNR is
a decrease in the received noise power.
CCU
5.3.1 Threshold Effect in Angle Modulation
) The noise analysis of angle demodulation schemes is based on
the assumption that the SNR at the demodulator input is high.
) This assumption of high SNR is a simplifying assumption that
is usually made in analysis of nonlinear-modulation systems.
) Threshold effect
A There exists a specific signal to noise ratio at the input of the
demodulator known as the threshold SNR, beyond which signal
mutilation occurs.
A The existence of the threshold effect places an upper limit on the trade-
off between bandwidth and power in an FM system.
A This limit is a practical limit in the value of the demodulation
index β f .
A Threshold in an FM system
( NS )b,th = 20 ( β + 1)
CCU
) In general, there are two factors that limit the value of the
demodulation index β .
A The limitation on channel bandwidth which effects β through
Carson’s rule.
A The limitation on the received power that limits the value of β
to less than what is derived from Equation ( NS )b,th = 20 ( β + 1) .
CCU
PM 1
=
( max m ( t ) )
2
2
( NS )0 = 23 β 2 ( NS )b
( NS )b,th = 20 ( β + 1)
( NS )0 = 60 β 2 ( β + 1) PM
CCU
CCU
CCU
5.3.1 Threshold Extension in Frequency modulation
) In order to reduce the threshold, in other words, in order to
delay the threshold effect to appear at lower-received signal
power, it is sufficient to decrease the input-noise power at the
receiver.
A This can be done by increasing the effective system bandwidth
at the receiver.
) Two approaches to FM threshold extension are to employ
FMFB or PLL-FM at the receiver.
) In application where power is very limited and bandwidth is
abundant, these systems can be employed to make it possible
to use the available bandwidth more efficiently.
) Using FMFB, the threshold can be extended approximately
by 5 - 7 dB.
CCU
5.3.2 Pre-emphasis and De-emphasis Filtering
) The objective in pre-emphasis and de-emphasis filtering is to
design a system which behaved like an ordinary frequency
modulator – demodulator pair in the low frequency band of
the message signal and like a phase modulator – demodulator
pair in the high-frequency band of the message signal.
) At the demodulator , we need a filter that at low frequencies
has a constant gain and at high frequencies behaves as an
integrator.
) The demodulator filter which emphasizes high frequencies is
called the pre-emphasis filter and the demodulator filter
which is the inverse of the modulator filter is called the de-
emphasis filter.
CCU
CCU
) The characteristics of the pre-emphasis and de-emphasis
filters depend largely on the power-spectral density of the
message process.
) In commercial FM broadcasting of music and voice, the

frequency response of the receiver (de-emphasis) filter is
1
Hd ( f ) =
1+ j f
f0
where
f 0 = 2π ×751×10−6 ≈ 2100Hz is the 3-dB frequency of the filter
CCU
) The noise component after the de-emphasis filter has a power
SnPD ( f ) = Sn0 ( f ) H d ( f )
spectral density 2
N0 2 1
= 2 f
Ac 1+ f 2
f2
0
) The noise power at the output of the demodulator

+W
Pn P D = ∫ −W
S nPD ( f )df
2
N0 +W f
=
Ac
2 ∫ −W
1+ f 2
df
f02
2 N 0 fc ⎡W ⎛ W ⎞⎤
3
−1
= 2 ⎢ − ta n ⎜ ⎟⎥
Ac ⎣ f0 ⎝ f0 ⎠⎦
CCU
) The ratio of the output SNRs
( NS )0 PD Pn0
=
( NS )0 PnPD
2 N 0W 3
3 Ac 2
= 2 N 0 f 03
⎡ Wf − tan −1 ( Wf ) ⎤
Ac 2 ⎣ 0 0 ⎦
( )
3
W
1
=
f0
3 Wf0 − tan −1 ( Wf0 )
CCU
)
CCU
CCU
5.3.4 Comparison of Analog0Modulation System
) Linear modulation system
A DSB-SC
A Conventional AM
A SSB-SC
A VSB
) Nonlinear modulation system
A FM
A PM
CCU
) The bandwidth efficiency of the system
A SSB-SC > SSB > VSB > DSB-SC > DSB
A PM = FM
A The most bandwidth-efficient analog communication system is
the SSB-SC system with a transmission bandwidth equal to the
signal bandwidth.
A PM and particularly FM are least-favorable systems when
bandwidth is the major concern, and their use is only justified
by their high level of noise immunity.
CCU
) The power efficiency of the system
A Angle-modulation scheme and particularly FM provide a high
level of noise immunity and, therefore, power efficiency.
A Conventional AM and VSB+C are least power-efficient system.
CCU
) The case of implementation of system
A The simplest receiver structure is the receiver for conventional
AM, and the structure of receiver for VSB+C system is only
slightly more complicated.
A FM receivers are also easy to implement.
A DSB-SC and SSB-SC require synchronous demodulation and
their receiver structure is much more complicated.
CCU
5.5 Effects of transmission losses and noise in analog
communication systems
r (t ) = α s (t ) + n (t ) , α < 1.
CCU
5.5.1 Characterization of thermal noise sources
CCU
) Based on quantum mechanics, the power-spectral density of
thermal noise is
2 Rh f
SR ( f ) =
(e )
h f
V 2 / Hz
kT
−1
where h is Planck’s constant, k is Boltzmann’s constant, and T
is the temperature of the resistor in degree Kelvin, i.e.,
T=273+C, where C is in degrees Centigrade.
) by h f
= 1+
hf
e kT
kT
⇒ S R ( f ) = 2 RkT V 2 / Hz
CCU
) The power spectral density of noise voltage across the load
resistor is
kT
Sn ( f ) = W / Hz
2
N0
= W / Hz
2
where T0 = 290 K
−21
N 0 = 4 × 10 W / Hz
CCU
5.5.2 Effective noise temperature and noise
figure
) Thermal noise converted to amplifier and load
CCU
figure
) The noise power at the output of the network is
∞ N0 ∞
Pn0 = ∫ Sn ( f ) H ( f ) df = ∫ H ( f ) df
2 2
−∞ 2 −∞
) Noise-equivalent bandwidth of the filter is defined as
1 ∞
∫ H ( f ) df
2
Beq =
2g −∞
where g = H ( f ) max
2
) The output noise power from an idea amplifier that

introduces no additional noise
Pn0 = gN 0 Bneq
CCU
figure
) The noise figure of the amplifier is
⎛ Te ⎞
F = ⎜1 + ⎟
⎝ T0 ⎠
pni
where Te =
gkBneq
and Pni is the power of the amplifier output
due to internally generated noise.
CCU
5.5.3 Transmission losses
) Transmission Loss
L = PT / PR
Ldb=10log L = 10log PT – 10log PR
CCU
)
CCU
)
CCU
)
CCU
5.5.4 Repeaters for signal transmission
CCU
) PR = PT / L
) P0=g PR = g PT / L
) By g = L and P0= PT
) The SNR at the output of the repeater is
⎛ S ⎞ 1 ⎛ S ⎞
⎜ ⎟ = ⎜ ⎟
⎝ ⎠1
N Fa ⎝ N ⎠i
1 ⎛ PR ⎞
= ⎜⎜ ⎟⎟
Fa ⎝ N 0 Bneq ⎠
1 ⎛ PT ⎞
= ⎜⎜ ⎟⎟
Fa L ⎝ N 0 Bneq ⎠
CCU
CCU

Noise in AM PDF

Uploaded by

Copyright:

Available Formats

Noise in AM PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Noise in AM PDF

Uploaded by

Copyright:

Available Formats

Chapter 5

Effect of Noise on Analog Communication

m(t) DSB-SC DSB-SC

) If PLL is employed ⇒ φ = φc ⇒ coherent or synchronon demod

Note that : Pnc = Pns = Pn

u ( t ) = Act m ( t ) cos ( 2π f ct ) ± Ac mˆ ( t ) sin ( 2π f c t )

with ideal-phase coherent demodulator

for synchronous demodulation (similar to DSB, except using

In practical application : a ∼ 0.8 ↔ 0.9

∴Vr ( t ) = ⎣ Ac ⎡⎣1 + amn ( t ) ⎤⎦ + nc ( t ) ⎦ + ns ( t )

Æ It is said that this system is operating below the threshold

assume E ⎡⎣ m ( t ) ⎤⎦ = 0 , zero-mean (i.e. no dc component)

⇒ the average power at the output of a narrow band filter

r ( t ) = Ac m 2 ( t ) cos2 ( 2π f c t + φc ) + noise terms

∴ There is signal power at the frequency 2π f c , which can

) The mean value of the output of Bandpass filter is a sinusoid

) If input to the PLL is cos ( 4π f c t − 2φ ) and output of the

where kv is a gain constant in radians/ volt-sec.

) Double-frequency terms resulting from the multiplication of

where ζ : loop-damping factor

) The one-side noise equivalent bandwidth

which is corrupted by additive narrowband noise

= LPF {( A + x ( t ) ) sin ( φ ) + ( A + x ( t ) ) sin ( 4π f t + φ ( t ) + φˆ ( t ) )

) By computing the autocorrelation and power spectral density

since SL<1 , we have an increase in the variance of the phase

variance in the estimate increase by a factor of 2.

) Assume: signal power is much higher than noise power

) The autocorrelation function

) Φ ( t ) is a zero-mean, stationary Gaussian process

) Z ( t ,τ ) is a Gaussian R.V. with zero mean and variance

) Since the bandwidth of the angle-modulated signal is defined

) The output SNR of a baseband system with the same received

) We can express the output SNR in terms of the bandwidth

) In commercial FM broadcasting of music and voice, the

f 0 = 2π ×751×10−6 ≈ 2100Hz is the 3-dB frequency of the filter

) The noise power at the output of the demodulator

3 Wf0 − tan −1 ( Wf0 )

) Noise-equivalent bandwidth of the filter is defined as

) The output noise power from an idea amplifier that

You might also like