Chapter 15

1. City has 10 macro-cells

each cell has 100 users
total number of users = 1000
Cells are of size 1 sqkm
maximum distance traveled to traverse =

2km
time =

2
30
= 169.7s
In the new setup
number of cells = 10
5
microcells
total number of users = 1000 100
2
users
time =

210
3010
3
=1.69s
number of users increases by 10000 and hando time reduces by 1/100
2. See Fig 1
(0,0)
(0,1)
v
u
(1,1)
(1,0)
60
Figure 1: Problem 2
D
2
= (j20)
2
+ (i20)
2
2(j20)(i20) cos(2/3)
D = 2a
_
i
2
+j
2
+ij =

3R
_
i
2
+j
2
+ij
3. diamond shaped cells, R= 100m
D
min
= 600m
D = 2KR
K =
D
2R
=
600
2100
= 3
N = K
2
= 9
(a) number of cells per cluster = N = 9
(b) number of channels per cell = total number/N = 450/9 = 50
4. (a) R=1km
D=6km
N =
A
cluster
A
cell
=

3D
2
/2
3

3R
2
/2
=
1
3
(D/R)
2
=
1
3
6
2
= 12
number of cells per cluster = N = 12
D
SHADED CELLS USE
SAME FREQUENCY
Figure 2: Problem 3
(b) number of channels in each cell = 1200/12 = 100
(c)
_
i
2
+j
2
+ij = 2

3 i = 2, j = 2
5. R=10m
D=60m

I
= 2

0
= 4
M = 4 for diamond shaped cells
SIR
a
=
R

I
MD

0
=
R
2
4D
4
= 32400
SIR
b
=
R
4
4D
4
= 324
SIR
c
=
R
2
4D
2
= 9
SIR
a
> SIR
b
> SIR
c
6. = 2
BPSK
P
b
= 10
6
P
b
= Q(

2
b
)
b
= SIR
0
= 4.7534
B = 50MHz
each user100KHz = B
s
SIR =
1
M
_
D
R
_

M=6 for hexagonal cells

a
1
= 0.167
a
2
= 3
N >
1
a
2
_
SIR
0
a
1
_
2/
N 9.4879 N = 10
C
u
= 50
7. G = 100
= 1
= 1.5
With no sectorization
SIR =
1

3G
(N
c
1)(1 +)
= 4.7534
N
c
= 26.2450 = 26
With sectorization, interference is reduced by a factor of 3
N
c
= 76.7349 = 76
8. SINR =
G

N
c
1
i=1
X
i
+N
= p(X
i
= 1)
N G(0.247N
c
, 0.078N
c
)
P
o
ut = p(SIR < SIR
0
)
(a) P
out
= p
_
G

N
c
1
i=1
X
i
+N
< SIR
0
_
= p
_

N
c
1
i=1
X
i
+N >
G
SIR
0
_
(b) X =

N
c
1
i=1
X
i
then X Bin(, N
c
1)
p(x +N > G/SIR
0
) =

N
c
1
N=0
p(n +N > G/SIR
0
|x = n)p(x = n)
p(x = n) =
_
N
c
1
n
_

n
(1 )
N
c
1n
p(x +N > G/SIR
0
) =
N
c
1

N=0
p(N > G/SIR
0
n|x = n)p(x = n)
=
N
c
1

N=0
p
_
N 0.247N
c

0.078N
c
>
G
SIR
0
n 0.247N
c

0.078N
c
|x = n
_
=
N
c
1

N=0
Q
_
G
SIR
0
n 0.247N
c

0.078N
c
_
p(x = n)
(c) N
c
= 35
= 0.5
SIR
0
= 5
G = 150
p = 0.0973
MATLAB
for i = 1:length(n)
pn(i) = (factorial(Nc-1)./(factorial(n(i)).*factorial(Nc-1-n(i))))...
*alpha.^n(i)*(1-alpha).^(Nc-1-n(i));
end
sump = 0; for i = 1:length(n)
f = ((G/sir0)-n(i)-.247*Nc)/(sqrt(.078*Nc));
sump = sump + .5*erfc((f)/sqrt(2))*pn(i);
end
(d) If x can be approximated as Gaussian then
x G((N
c
1), (N
c
1)(1 ))
x +N G(0.247N
c
+ (N
c
1), 0.078N
c
+ (N
c
1)(1 ))
p(x +N > G/SIR
0
) = Q
_
G
SIR
0
(0.247N
c
+ (N
c
1))
_
0.078N
c
+ (N
c
1)(1 )
_
(e) p= 0.0969 (very accurate approximation!)
9. dene

k
=
g
k
P
k
n
k
+

k=j
g
kj
p
j
k, j {1, . . . K}
where,
g
k
is channel power gain from user k to his base station n
k
is thermal noise power at user ks base
station
is interference reduction factor ( 1/G)
g
kj
is channel power gain from j
th
interfering transmitter to user ks base station
p
k
is user ks Tx power
p
j
is user js Tx power
dene a matrix F such that
F
kj
=
_
0 k = j

k
g
kj

g
k
k = j
k, j {1, . . . K}
u =
_

1
n
1
g
1
,

2
n
2
g
2
, . . . ,

K
n
K
g
K
_
If Perron Ferbinius eigenvalue of F is less than 1 , then a power control policy exists. The optimal
power control policy is given to be P

= (I F)
1
u
10. Matlab
D = 2:.01:10;
R = 1;
gamma = 2;
Pdes = R^(-gamma);
for i = 1:length(D)
Pint = 6*(.2*(D(i)-R)^(-gamma)+.2*(D(i)-R/2)^(-gamma)+.2*(D(i))^(-gamma)...
+.2*(D(i)+R/2)^(-gamma)+.2*(D(i)+R)^(-gamma));
Pintbest = 6*((D(i)+R)^(-gamma));
Pintworst = 6*((D(i)-R)^(-gamma));
ASE(i) = log(1+Pdes/Pint)/(pi*(.5*D(i))^2);
ASEbest(i) = log(1+Pdes/Pintbest)/(pi*(.5*D(i))^2);
ASEworst(i) = log(1+Pdes/Pintworst)/(pi*(.5*D(i))^2);
end
2 3 4 5 6 7 8 9 10
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
D
A
S
E
(
D
)
5 Case ASE
Best Case ASE
Worst Case ASE
Figure 3: Problem 10
11. P
t
= 5W
B = 100KHz
N
0
= 10
16
W/Hz
P
r
= P
t
K
_
d
0
d
_
3
d
0
= 1, K = 100
(a) D=2R
2 users share the band available
Each user gets 50KHz
BASESTATION
USER
R=1Km
Figure 4: Problem 11a
(b) P
b
=
1
4
b
10
3
=
1
4
b

b
= 250
If D(n) = 2nR, number of users that share band = 2(2(n-1)+1)
each user gets
100KHz
2(2n1)
= B
u
(n)
interference is only from rst tier
SIR(n) =
P
t
K
_
d
0
R
_
3
N
0
2
B
u
(n) + 2
_
P
t
K
_
d
0

R
2
+D(n)
2
_
3
_ > 25
using Matlab , n = 4, SIR = 261.9253, D = 8R
Matlab
Pt = 5;
R = 1000;
sigma_2 = 1e-16;
n = 1;
D = 2*n*R;
Bu = (100/(2*(2*n-1)))*1e3;
K = 100;
d0 = 1;
Pdes = Pt*K*(d0/R)^3;
Pint = 2*(Pt*K*(d0/sqrt(R^2+D^2))^3);
Npower = sigma_2*Bu;
sir = Pdes/(Npower+Pint);
while sir < 250
n = n+1;
D = 2*n*R;
Bu = (100/(2*(2*n-1)))*1e3;
K = 100;
d0 = 1;
Pdes = Pt*K*(d0/R)^3;
Pint = 2*(Pt*K*(d0/sqrt(R^2+D^2))^3);
Npower = sigma_2*Bu;
sir = Pdes/(Npower+Pint);
end
(c) ASE =
(R
1
+R
2
)/B
2km2km
R
1
= R
2
= B
u
(1) log(1 +SIR(1))
B
u
(1) = 50KHz, SIR(1) = 5.5899
ASE = 0.6801bps/Hz/km
2
12. B=100KHz
N
0
= 10
9
W/Hz
K = 10
P = 10mW per user
(a) 0 1 is channel gain between cells.
See Matlab
If is large, interference can be decoded and subtracted easily so capacity grows with as high
SNRs (beyond an value) .
For low SNR values ( less than a value) c decreases with increase in as interference is increased
which cannot be easily decoded due to low SNR.
MATLAB CODE:
B = 100e3;
sigma_2 = 1e-9;
P = 10e-3;
K = 10;
ss = .001;
alpha = 0:.01:1;
theta = 0:ss:1;
for i = 1:length(alpha)
capvec = log2(1+(K*P*(1+2*alpha(i)*cos(2*pi*theta)).^2)/(sigma_2*B));
C(i) = (1/K)*sum(capvec)*ss;
end
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.82
0.84
0.86
0.88
0.9
0.92
0.94
0.96
0.98
1
1.02

C
(

)
/
B
B = 100 KHz
Figure 5: Problem 12a
(b) C(K) as K because as the number of mobile per cell increases system resources get shared more
and so per user capacity C(K) has to fall.
MATLAB CODE:
BB = 100e3;
sigma_2 = 1e-9;
P = 10e-3;
K = 1:.1:30;
ss = .001;
alpha = .5;
theta = 0:ss:1;
for i = 1:length(K)
capvec = log2(1+(K(i)*P*(1+2*alpha*cos(2*pi*theta)).^2)/(sigma_2*B));
C(i) = (1/K(i))*sum(capvec)*ss;
end
0 5 10 15 20 25 30
0
1
2
3
4
5
6
K
C
(
K
)
/
B
B = 100 KHz
Figure 6: Problem 12b
(c) as transmit power P , capacity C but gets saturated after a while as the system becomes
interference limited.
MATLAB CODE:
B = 100e3;
sigma_2 = 1e-9;
P = [0:.1:100]*1e-3;
K = 10;
ss = .001;
alpha = .5;
theta = 0:ss:1;
for i = 1:length(P)
capvec = log2(1+(K*P(i)*(1+2*alpha*cos(2*pi*theta)).^2)/(sigma_2*B));
C(i) = (1/K)*sum(capvec)*ss;
end
0 10 20 30 40 50 60 70 80 90 100
0
0.2
0.4
0.6
0.8
1
1.2
1.4
P (in mW)
C
(
P
)
/
B
B = 100 KHz
Figure 7: Problem 12c