Chapter 6

1. (a) For sinc pulse, B =

2N
0
_
+ 0.3Q
_
d
min

2N
0
_
= Q
_
d
min

2N
0
_
d
min
= 2A
= Q
_
_

2A
2
N
0
_
_
(b)
p( m = 0|m = 1)p(m = 1) = p( m = 1|m = 0)p(m = 0)
0.7Q
_
_
A+a
_
N
0
2
_
_
= 0.3Q
_
_
Aa
_
N
0
2
_
_
, a > 0
Solving gives us a for a given A and N
0
(c)
p( m = 0|m = 1)p(m = 1) = p( m = 1|m = 0)p(m = 0)
0.7Q
_
_
A
_
N
0
2
_
_
= 0.3Q
_
_
B
_
N
0
2
_
_
, a > 0
Clearly A > B, for a given A we can nd B
(d) Take
E
b
N
0
=
A
2
N
0
= 10
In part a) P
e
= 3.87 10
6
In part b) a=0.0203 P
e
= 3.53 10
6
In part c) B=0.9587 P
e
= 5.42 10
6
Clearly part (b) is the best way to decode.
MATLAB CODE:
A = 1;
N0 = .1;
a = [0:.00001:1];
t1 = .7*Q(A/sqrt(N0/2));
t2=.3*Q(a/sqrt(N0/2));
diff = abs(t1-t2);
[c,d] = min(diff);
a(d)
c
3. s(t) = g(t) cos 2f
c
t
r = r cos
where r is the signal after the sampler if there was no phase oset. Once again, the threshold that
minimizes P
e
is 0 as (cos ) acts as a scaling factor for both +1 and -1 levels. P
e
however increases
as numerator is reduced due to multiplication by cos
P
e
= Q
_
d
min
cos

2N
0
_
4.
A
2
c
_
T
b
0
cos
2
2f
c
tdt = A
2
c
_
T
b
0
1 + cos 4f
c
t
2
= A
2
c
_

_
T
b
2
+
sin(4f
c
T
b
)
8f
c
. .
0 as f
c
1
_

_
=
A
2
c
T
b
2
= 1
x(t) = 1 +n(t)
Let prob 1 sent =p
1
and prob 0 sent =p
0
P
e
=
1
6
[1.p
1
+ 0.p
0
] +
2
6
[0.p
1
+ 0.p
0
] +
2
6
[0.p
1
+ 0.p
0
] +
1
6
[0.p
1
+ 1.p
0
]
=
1
6
[p
1
+p
0
] =
1
6
( p
1
+p
0
= 1 always )
5. We will use the approximation P
e
(average number of nearest neighbors).Q
_
d
min

2N
0
_
where number of nearest neighbors = total number of points taht share decision boundary
(a) 12 inner points have 5 neighbors
4 outer points have 3 neighbors
avg number of neighbors = 4.5
P
e
= 4.5Q
_
2a

2N
0
_
(b) 16QAM, P
e
= 4
_
1
1
4
_
Q
_
2a

2N
0
_
= 3Q
_
2a

2N
0
_
(c) P
e

23+32
5
Q
_
2a

2N
0
_
= 2.4Q
_
2a

2N
0
_
(d) P
e

14+43+42
9
Q
_
3a

2N
0
_
= 2.67Q
_
3a

2N
0
_
6.
P
s,exact
= 1
_
1
2(

M 1)

M
Q
_
_
3
s
M 1
__
2
Figure 1: Problem 5
P
s,approx
=
4(

M 1)

M
Q
_
_
3
s
M 1
_
approximation is better for high SNRs as then the multiplication factor is not important and P
e
is
dictated by the coecient of the Q function which are same.
MATLAB CODE:
gamma_db = [0:.01:25];
gamma = 10.^(gamma_db/10);
M = 16;
Ps_exact=1-exp(2*log((1-((2*(sqrt(M)-1))/(sqrt(M)))*Q(sqrt((3*gamma)/(M-1))))));
Ps_approx = ((4*(sqrt(M)-1))/sqrt(M))*Q(sqrt((3*gamma)/(M-1)));
semilogy(gamma_db, Ps_exact);
hold on
semilogy(gamma_db,Ps_approx,b:);
7. See gure. The approximation error decreases with SNR because the approximate formula is based
on nearest neighbor approximation which becomes more realistic at higher SNR. The nearest neighbor
bound over-estimates the error rate because it over-counts the probability that the transmitted signal
is mistaken for something other than its nearest neighbors. At high SNR, this is very unlikely and this
over-counting becomes negligible.
8. (a)
I
x
(a) =
_

0
e
at
2
x
2
+t
2
dt
since the integral converges we can interchange integral and derivative for a0
I
x
(a)
a
=
_

0
te
at
2
x
2
+t
2
dt
x
2
I
x
(a)
I
x
(a)
a
=
_

0
(x
2
+t
2
)e
at
2
x
2
+t
2
dt =
_

0
e
at
2
dt =
1
2
_

a
0 5 10 15 20 25 30
10
300
10
200
10
100
10
0
P
s
Problem 2 Symbol Error Probability for QPSK
Approximation
Exact Formula
0 1 2 3 4 5 6 7 8 9 10
10
3
10
2
10
1
10
0
SNR(dB)
P
s
Problem 2 Symbol Error Probability for QPSK
Approximation
Exact Formula
Figure 2: Problem 7
(b) Let I
x
(a) = y, we get
y

x
2
y =
1
2
_

a
comparing with
y

+P(a)y = Q(a)
P(a) = x
2
, Q(a) =
1
2
_

a
I.F. = e

P(u)u
= e
x
2
a
e
x
2
a
y =
_

1
2
_

a
e
x
2
u
du
solving we get
y =

2x
e
ax
2
erfc(x

a)
(c)
erfc(x

a) = I
x
(a)
2x

e
ax
2
=
2x

e
ax
2
_

0
e
at
2
x
2
+t
2
dt
a = 1
erfc(x) =
2x

e
ax
2
_

0
e
at
2
x
2
+t
2
dt
=
2

_
/2
0
e
x
2
/sin
2

d
Q(x) =
1
2
erfc(x/

2) =
1

_
/2
0
e
x
2
/2sin
2

d
9. P = 100W
N
0
= 4W, SNR = 25
P
e
= Q(

2) = Q(

50) = 7.687 10
13
data requires P
e
10
6
voice requires P
e
10
3
so it can be used for both.
with fading P
e
=
1
4
b
= 0.01
So the system cant be used for data at all. It can be used for very low quality voice.
10. T
s
= 15sec
at 1mph T
c
=
1
B
d
=
1
v/
= 0.74s T
s
outage probability is a good measure.
at 10 mph T
c
= 0.074s T
s
outage probability is a good measure.
at 100 mph T
c
= 0.0074s = 7400s > 15s outage or outage combined with average prob of error can
be a good measure.
11.
M

(s) =
_

0
e
s
p()d
=
_

0
e
s
1

e
/
d
=
1
1 s
12. (a) When there is path loss alone, d =

100
2
+ 500
2
= 100

6 10
3
P
e
=
1
2
e

b
= 13.1224
P

N
0
B
= 13.1224 P

= 1.3122 10
14
P

P
t
=
_

G
4d
_
2
4.8488W
(b)
x = 1.3122 10
14
= 138.82dB
P
,dB
N(P

, 8),
dB
= 8
P(P
,dB
x) = 0.9
P
_
P
,dB
P

8

x P

8
_
= 0.9
Q
_
x P

8
_
= 0.9

x P

8
= 1.2816
P

= 128.5672dB = 1.39 10
13
13. (a) Law of Cosines:
c
2
= a
2
+b
2
2ab cos C with a,b =

E
s
, c = d
min
, C = = 22.5
c = d
min
=
_
2E
s
(1 cos 22.5) = .39

E
s
Can also use formula from reader
(b) P
s
=
m
Q
_

s
_
= 2Q
_
_
d
min
2
2N
o
_
= 2Q(

.076
s
)

m
= 2,
m
= .076
(c) P
e
=

_
0
P
s
(
s
)f(
s
)d
s
=

_
0

m
Q(

s
)f(
s
)d
s
Using alternative Q form
=

m

2
_
0
_
1 +
g
s
(sin)
2
_
1
d with g =

m
2
=

m
2
_
1
_
g
s
1+g
s
_
= 1
_
.038
s
1+.038
s
=
1
.076
s
, where we have used an integral table to evaluate
the integral
(d) P
d
=
P
s
4
(e) BPSK: P
b
=
1
4
b
= 10
3
,
b
= 250, 16PSK: From above get
s
= 3289.5
Penalty =
3289.5
250
= 11.2dB
Also will accept
b
(16PSK) = 822 = 5.2dB
14.
P
b
=
_

0
P
b
()p()d
P
b
() =
1
2
e

P
b
=
1
2
_

0
e
b
p

()d =
1
2
M
But from 6.65
M

(s) =
_
1
s
m
_
m
P
b
=
1
2
_
1 +

m
_
m
For M = 4, = 10
P
b
= 3.33 10
3
15. %Script used to plot the average probability of bit error for BPSK modulation in
%Nakagami fading m = 1, 2, 4.
%Initializations
avg_SNR = [0:0.1:20]; gamma_b_bar = 10.^(avg_SNR/10); m = [1 2 4];
line = [-k, -r, -b]
for i = 1:size(m,2)
for j = 1:size(gamma_b_bar, 2)
Pb_bar(i,j) = (1/pi)*quad8(nakag_MGF,0,pi/2,[],[],gamma_b_bar(j),m(i),1);
end
figure(1);
semilogy(avg_SNR, Pb_bar(i,:), line(i));
hold on;
end
xlabel(Average SNR ( gamma_b ) in dB); ylabel(Average bit error
probability ( P_b ) ); title(Plots of P_b for BPSK modulation in
Nagakami fading for m = 1, 2, 4); legend(m = 1, m = 2, m =
4);
function out = nakag_MGF(phi, gamma_b_bar, m, g);
%This function calculates the m-Nakagami MGF function for the specified values of phi.
%phi can be a vector. Gamma_b_bar is the average SNR per bit, m is the Nakagami parameter
%and g is given by Pb(gamma_b) = aQ(sqrt(2*g*gamma_b)).
out = (1 + gamma_b_bar./(m*(sin(phi).^2)) ).^(-m);
SNR = 10dB
M BER
1 2.3310
2
2 5.5310
3
4 1.0310
3
16. For DPSK in Rayleigh fading, P
b
=
1
2
b

b
= 500
N
o
B = 3 10
12
mW P
target
=
b
N
0
B = 1.5 10
9
mW = -88.24 dBm
Now, consider shadowing:
P
out
= P[P
r
< P
target
] = P[ < P
target
P
r
] =
_
P
target
P
r

1
(.01) = 2.327 =
P
target
P
r

P
r
= 74.28 dBm = 3.73 10
8
mW = P
t
_

4d
_
2
d = 1372.4 m
17. (a)

1
=
_
0 w.p. 1/3
30 w.p. 2/3

2
=
_
5 w.p. 1/2
10 w.p. 1/2
In MRC,

=
1
+
2
. So,

=
_

_
5 w.p. 1/6
10 w.p. 1/6
35 w.p. 1/3
40 w.p. 1/3
(b) Optimal Strategy is water-lling with power adaptation:
S()
S
=
_
1

,
0
0 <
0
Notice that we will denote

by only hereon to lighten notation. We rst assume

i=1
_
1

i
_
p
i
= 1

0
= 1 +
4

i=1
p
i

0
= 0.9365 < 5
So we found the correct value of
0
.
C = B
4

i=1
log
2
_

0
_
p
i
C = 451.91 Kbps
(c) Without, receiver knowledge, the capacity of the channel is given by:
C = B
4

i=1
log
2
(1 +
i
)p
i
C = 451.66 Kbps
Notice that we have denote

by to lighten notation.
18. (a)
s(k) = s(k 1)
z(k 1) = g
k1
s(k 1) +n(k 1)
z(k) = g
k
s(k) +n(k)
From equation 5.63 , the input to the phase comparator is
z(k)z

(k 1) = g
k
g
(
k 1)

s(k)s

(k 1) +g
k
s(k 1)n

k1
+
g
(
k 1)

(k 1)n
k
+n
k
n

k1
but s(k) = s(k 1)
s(k)s

(k 1) = |s
k
|
2
= 1 (normalized)
(b)
n
k
= s

k1
n
k
n
k1
= s

k1
n
k1
z = g
k
g

k1
+g
k
n

k1
+g

k1
n
k

x
(s) =
p
1
p
2
(s p
1
)(s p
2
)
=
A
s p
1
+
B
s p
2
A = (s p
1
)
x
(s)|
s=p
1
=
p
1
p
2
p
1
p
2
B = (s p
2
)
x
(s)|
s=p
2
=
p
1
p
2
p
2
p
1
(c) Relevant part of the pdf

x
(s) =
p
1
p
2
(p
2
p
1
)(s p
2
)
p
x
(x) =
p
1
p
2
(p
2
p
1
)
L
1
_
1
(s p
2
)
_
=
p
1
p
2
(p
2
p
1
)
e
p
2
x
, x < 0
(d)
P
b
= prob(x < 0) =
p
1
p
2
(p
2
p
1
)
_
0

e
p
2
x
dx =
p
1
p
2
p
1
(e)
p
2
p
1
=
1
2N
0
[
b
(1
c
) + 1]
+
1
2N
0
[
b
(1 +
c
) + 1]
=

b
+ 1
N
0
[
b
(1
c
) + 1][
b
(1 +
c
) + 1]
P
b
=

b
(1
c
) + 1
2(
b
+ 1)
(f)
c
= 1
P
b
=
1
2(
b
+ 1)
19.
b
0 to 60dB

c
= J
0
(2B
D
T) with B
D
T = 0.01, 0.001, 0.0001
where J
0
is 0 order Bessel function of 1
st
kind.
P
b
=
1
2
_
1 +
b
(1
c
)
1 +
b
_
when B
D
T = 0.01, oor can be seen about
b
= 40dB
when B
D
T = 0.001, oor can be seen about
b
= 60dB
when B
D
T = 0.0001, oor can be seen between
b
= 0 to 60dB
20. Data rate = 40 Kbps
Since DQPSK has 2 bits per symbol. T
s
=
2
4010
3
= 5 10
5
sec
DQPSK
Gaussian Doppler power spectrum,
c
= e
(B
D
T)
2
B
D
= 80Hz
Rician fading K = 2

c
= 0.9998
P
floor
=
1
2
_
1

(
c
/

2)
2
1 (
c
/

2)
2
_
exp
_

(2

2)K/2
1
c
/

2
_
= 2.138 10
5
21. ISI:
Formula based approach:
P
floor
=
_
T
m
T
s
_
2
Since its Rayleigh fading, we can assume that
T
m

T
m
= 100ns
P
floor
10
4
which gives us
_
T
m
T
s
_
2
10
4
T
s


T
m
_
P
floor
= 10sec
So, T
s
10s. T
b
5s. R
b
200 Kbps.
Thumb - Rule approach:

t
= 100 nsec will determine ISI. As long as T
s

T
, ISI will be negligible. Let T
s
10
T
. Then
R
2bits
symbol
1
T
s
symbols
sec
= 2Mbps
Doppler:
B
D
= 80 Hz
P
floor
= 10
4

1
2
_
_
1

_
_

c
/

2
_
2
1
_

c
/

2
_
2
_
_

c
0.9999
But
c
for uniform scattering is J
0
(2B
D
T
s
), so

c
= J
0
(2B
D
T
s
) = 1 (f
D
T
s
)
2
0.9999
T
s
39.79s
T
b
19.89s. R
b
50.26 Kbps.
Combining the two, we have 50.26 Kbps R
b
200 Kbps (or 2 Mbps).
22. From gure 6.5
with P
b
= 10
3
, d =
T
m
/T
s
,
T
m
= 3s
BPSK
d = 5 10
2
T
s
= 60sec
R = 1/T
s
= 16.7Kbps
QPSK
d = 4 10
2
T
s
= 75sec
R = 2/T
s
= 26.7Kbps
MSK
d = 3 10
2
T
s
= 100sec
R = 2/T
s
= 20Kbps