Vous êtes sur la page 1sur 5

Tutorial #8: AM Modulation with Noise

• In the following, an interesting problem about VSB modulation system with noise will be discussed. This
may assist you in your preparation to the final exam from many aspects.

Example 1. Communication engineers are given the job of designing a VSB communication system for
the transmission of speech signals with bandwidth W = 4 kHz over an additive white Gaussian noise
channel with Sn (ω) = N0 /2. They consider two configurations for this system, one where the vestigial
filtering is done at the transmitter, and the other where the vestigial filtering is done at the receiver.

System A

HT (ω) HR (ω)
A 1 B C 1 D E
m(t) × • • + • • × • LPF m(t) + n1 (t)
ab c d ab c d
cos(2πfc t + θ) n(t) 4 cos(2πfc t + θ)

transmitter channel receiver

System B

HT (ω) HR (ω)
A 1 B C 1 D E
m(t) × • • + • • × • LPF m(t) + n2 (t)
ab c d ab c d
cos(2πfc t + θ) n(t) 4 cos(2πfc t + θ)

transmitter channel receiver

where a = ωc − 2π103 , b = ωc , c = ωc + 103 , d = ωc + 4π103 .

(a) Find the output noise power of each of these systems (in terms of N0 ).
(b) Assuming m(t) has the power spectrum
(
C, |ω| < 8π × 103 ;
Sm (ω) =
0, otherwise,

sketch the power spectrum of the signal at the input to the channel in each of these systems and
find the mean power PT of these signals.
(c) Find the signal to noise ratio at the output of each of these systems, expressing the results in terms
of PT and N0 W . State whether one of these systems is superior to the other in that less power must
be input to the channel to achieve the same output SNR. State any advantage in decibles (dB).

1
• Before solving the above problem, I would like to introduce to you the following important relationships
that are frequently used in our solution. I encourage you to add them to your “cheat sheet”.
If x(t) is a band-pass signal with power spectral density function Sx (ω) centered around ωc , then
one can express x(t) in terms of its inphase and quadrature components, xI (t) and xQ (t), respec-
tively, as
x(t) = xI (t) cos(ωc t + θ) − xQ (t) sin(ωc t + θ)
Few properties are worth mentioning regarding the signals xI (t) and xQ (t):

1. Both signals are low-pass type.

2. The power spectral density function of both components are equal, i.e., SxI (ω) = SxQ (ω).

3. To find SxI (ω) or SxQ (ω), you do the following: Shift the +ve frequency component of Sx (ω)
to the left by ωc , then shift the −ve frequency component of Sx (ω) to the right by ωc , and
then add them up. To illustrate, here is an example:

Example 2. Suppose that x(t) is a band-pass signal with Sx (ω) given by (assuming a, b > 0)

Sx (ω)

K
2

ω
−ωc − b −ωc −ωc + a ωc − a ωc ωc + b

then SxI (ω) or SxQ (ω) is given by

SxI (ω) = SxQ (ω)

K
2

ω
−b −a a b

If x(t), with power spectral density function Sx (ω), is the


input to an LTI system with frequency response H(ω),
LTI then the power spectral density function of the output
x(t) y(t)
H(ω) y(t), denoted by Sy (ω), is given by

Sy (ω) = Sx (ω)|H(ω)|2

If x(t) has a power spectral density function Sx (ω), then


the power spectral density function of y(t), denoted by
y(t) = ax(t) Sy (ω), is given by

Sy (ω) = Sx (ω)|H(ω)|2

2
If x(t) has a power spectral density function Sx (ω), then
the power spectral density function of y(t), denoted by
y(t) = ax(t) cos(ωc t + θ) Sy (ω), is given by

a2 a2
Sy (ω) = Sx (ω − ωc ) + Sx (ω + ωc )
4 4

• Let us go back now to our main example:


Solution:

(a) We begin with System B: Denote the noise signal at node D by nD (t). Then, the mean power
spectral density function of nD (t) is given by:

N0
SnD (ω) = Sn (ω)|HR (ω)|2 = |HR (ω)|2
2
where the following figure shows a sketch of SnD (ω):
SnD (ω)

N0
2

ω
−d −c −b −a a b c d

Now, since nD (t) is a band-pass signal, we can express the signal using the the inphase and quadra-
ture form as follows:
nD (t) = nD,I (t) cos(ωc t + θ) − nD,Q (t) sin(ωc + θ)
where nD,I (t) and nD,Q (t) are the corresponding inphase and quadrature components with mean
power spectral density function given by

N0
2

ω
−8π103 −2π103 2π103 8π103

SnD,I (ω) = SnD,Q (ω)

N0
2

ω
−8π × 103 −2π103 2π103 8π × 103

Now, denote the noise signal at point E by nE (t), then we have


nE (t) = 4nD (t) cos(ωc t + θ)
= 4nD,I (t) cos2 (ωc t + θ) − 4nD,Q (t) cos(ωc t + θ) sin(ωc t + θ)
= 2nD,I (t) + 2nD,I (t) cos(2ωc t + 2θ) − 2nD,Q (t) sin(2ωc t + 2θ)
| {z }
High frequency components centerd around 2ωc

3
After low-pass filtering, the “high frequency” components that appear in the above equation are
blocked. Since nD,I (t) is a low-pass signal with bandwidth equal to the message bandwidth (see
SnD,I (ω)), then the signal 2nD,I (t) will pass through the LPF filter, i.e.,
n1 (t) = 2nD,I (t)
Therefore, the average noise power of n1 (t) is (note that PnD,I = PnD )
1
Pn1 = 4PnD,I = 4PnD = 4 × [Area under SnD (ω)]

  
2 1 N0 N0 44
= 2× × 4π × 103 × + 6π × 103 × = × 103 N0
π 3 2 2 3
For system A: following the same approach that has been used to analyze system B, one can
show that the average power of the noise signal n2 (t) is

Pn2 = 20 × 103 N0

(b) Consider again system B: The message signal m(t) has a power spectrum Sm (ω) that is shown in
the following figure
Sm (ω)

ω
−8π103 8π103

where the average power of m(t) can be easily found to be Pm = 8 × 103 C.


Now, at point A, the signal is given by
mA (t) = m(t) cos(ωc t + θ)
In this case, we have that
1 1
SmA (ω) = Sm (ω − ωc ) + Sm (ω + ωc )
4 4
SmA (ω)

C/4

ω
−ωc − 8π103 −ωc −ωc + 8π103 ωc − 8π103 ωc −ωc + 8π103

Finally, assuming the signal at point B is mB (t), then


SmB (ω) = SmA (ω)|HT (ω)|2

SmB (ω)

C/4

ω
−d −c −b −a a b c d

4
The average power at the input to the channel, i.e., mB (t) is given by
 
1 1 C 5
PT = × [Area unde SmB (ω)] = × 2 × (d − a) × = × 103 C
2π 2π 4 2

For system A: following the same approach as above, one can show that

11
PT = × 103 C
6

(c) The signal to noise ration at the output of the system is defined as

Pm
SNRA =
Pn1

Pm
SNRB =
Pn2
It is required that we express the SNR in terms of PT and the product N0 W , where W = 4 × 103
(the bandwidth of the message). First, we need to express Pm in terms of PT . In this case, it can
be easily deduced that
48


 × PT , for system A
 11

Pm =
 16 × PT ,


for system B

5
In this case,
48
× PT W 48 × 4 × 103

PT

48

PT

SNRA = 11 × = = = SNRB
20 × 103 N0 W 11 × 20 × 103 N0 W 55 N0 W

System A is better since it generates the same SNR for 1.5 dB less transmitted power.