ELE751 — Solutions to Homework # 1

1. Hexagonal geometry

(a) Prove that among the three shapes that tessellate, the hexagon has the largest area for a
given distance between the center to its farthest perimeter points (called vertices).
√
(b) (3.1 in your textbook) Prove that Q = 3N .

Proof

(a) The areas as a function of radius for equilateral triangle, square and regular hexagon are
repectively
√
3 2
AT riangle = R
4 T riangle
2
ASquare = RSquare
√
3 3 2
AHexagon = RHexagon
2
If we set RT riangle = RSquare = RHexagon , clearly, we have AT riangle < RSquare <
RHexagon . Hence done.
(b) Using cosine law for triangles, we have
√ √ √ √
D2 = ( 3iR)2 + ( 3jR)2 − 2( 3iR)( 3jR) cos 120o = 3N R2
√
Hence Q = D/R = 3N .

2. Reverse channel SIR analysis

For certain N and n, find the worst case SIR for reverse channel using first tier approximation. In
your analysis, assume equal transmitting power for all MS.

D=2R

Figure 1: Worst case scenario for reverse channel co-channel interference N=3.

Answer
 R

Figure 2: Worst case scenario for reverse channel co-channel interference N=7.

(a) N = 3 (i = j = 1).
From Figure 1, we see clearly that in the worst case scenario, all MS in the first tier co-
channel cells sit at the vertex that’s closest the the BS. It is straightforward to show that they
are all equidistant to the desired BS with distance 2R. Thus the exact first tier interference
for reverse channel is
S R−n 2n
= =
I 6(2R)−n 6
This is worse than the forward link SIR using first tier approximation.
(b) N=7. From Figure 2, it is straightforward to show that they are all equal distant to the
desired BS with distance
√ √
R2 + (2 3R)2 = 13R
Thus the exact first tier interference for reverse channel is
S R−n 1 √
= √ = ( 13)n
I 6( 13R)−n 6

Clearly, the SIR loss in dB between the two cases is:

2n 1 √
10log10 ( ) − 10log10 ( ( 13)n ) = 2.57ndB
6 6

3. Second-tier approximation on C/I

(a) Using hexagonal geometry, find the number of second-tier co-channel cells (co-channel cells
obtained by repeating the shifting procedure twice). Is it a function of cluster size?
(b) Use N = 7 and n = 4 as an example, find the approximate SIR by considering both first
and second tier co-channel cells. Notice that the second tier cells are not equidistant to the
reference cell. Compare this to that use first tier only.

2
 Answer

(a) There are a total of 12 second tier co-channel cells and is independent of the cluster size.
(b) In the approximation, we approximate the distance between co-channel BS to the desired
MS using the distance between the BS’s. Using hexogonal geometry, we can figure out easily
that there are two groups of second tier co-channel cells: 6 of them have a distance of 2D
√
to the desired the BS while the other 6 have a distance of 3D. Therefore
√
S R−n ( 3N )n
= √ = √ = 62.63 = 17.97dB
I 6[D−n + (2D)−n + ( 3D)−n ] 6(1 + 2−n + ( 3)−n
which is slightly worse than the first tier approximation.

4. Noisy channel
Field measurement was conducted and it was found that the average receive power at d = 20m
is 1W while at d = 200m it is 1mW .

(a) What is the path loss exponent?

(b) Assume that the radius of a cell is 1km. The channel bandwidth is 30kHz and the additive
white noise has a one-sided power spectral density of 10−9 W/kHz. What should the cluster
size be if the desired signal to noise and interference ratio (S/(N + I)) is at 15dB? In your
analysis, you need only consider first tier interfering BS. Compare the result with noiseless
case.

Answer

(a) Path loss exponent n satisfies

1 0.001
−n
=
20 200−n
Hence
n=3

(b) One sided PSD N0 = 10−9 W/kHz with channel bandwidth 30kHz, hence the noise power
is
Pn = 10−9 W/kHz × 30kHz = 3 × 10−8 W
The signal power at the edge of the cell is
 
1000 −3
S =1× = 8 × 10−6 W
20

Given that the S/(I + Pn ) = 15dB = 31.62, then we can get the maximum allowed
interference is
I = 2.23 × 10−7 W
But we know that interference is a function of cluster size N ,
⎡  √ −3 ⎤
1000 3N
I = 6 × ⎣1 × ⎦
20

3
 From this we can calculate that
N ≥ 11.97
Thus we choose N ≥ 12.
For the noiseless case, setting
S 1 √
= ( 3N )3 = 15dB = 31.62
I 6
we get N = 11.01. Thus in both case we should choose N = 12.

5. 3.7 Answer

(a) Calls are not lost due to weak signal condition during handoff if:

dmin − dH0
≥ 4.5seconds
v
where v is the mobile speed.
dmin corresponds to the distance when power at BS1 reaches Pr,min

Pr,min = −29log10 (dmin ) ⇒ dmin = 10−Pr,min /29 = 1083m

dH0 corresponds to the distance when power at BS1 reaches Pr,H0

1083 − 10−Pr,H0
≥ 4.5seconds ⇒ Pr,H0 ≥ 29log10 (1083 − 4.5v)dBm
v
Thus,
∆ = Pr,H0 − Pr,min ≥ 29log10 (1083 − 4.5v) + 88

(b) If we set ∆ too large, several unnecessary handoffs will be requested and performed, increasing
the signaling traffic between the base stations and mobile switching center (MSC). On the
other hand, if ∆ is too small, that is, Pr,H0 is only slightly greater than Pr,min , there will
not be enough time to complete the handoff (especially for high speed mobiles), and calls
may be lost due to weak signal condition.

6. Answer
Given that the number of channels per cell equals 24 and P [blocking] = 0.5%, from Table 3.4,
we know that the total traffic that it can handle, from Erlang B formular, is 14.2Erlang.
With 60o sectoring, each sector on average will have 24/6 = 4 channels. From Table 3.4, we can
get the traffic per sector is 0.701Erlangs, thus the total traffic is 6 × 0.701 = 4.206Erlangs.
The loss of trunking efficiency is
14.2 − 4.206
= 70.4%
14.2

7. 3.15 Answer

(a) Given GOS = 2%

4
 i. For C = 4 Channels, from the Erlang B chart
Atot
Atotal = 1.1Erlangs ⇒ Aperchannel = = 0.275Erlangs
C
ii. For C = 20 Channels, from the Erlang B chart
Atot
Atotal = 14Erlangs ⇒ Aperchannel = = 0.7Erlangs
C
iii. For C = 40 Channels, from the Erlang B chart
Atot
Atotal = 31Erlangs ⇒ Aperchannel = = 0.775Erlangs
C
(b)
Atotal 31
U= = 1 ≈ 1063users
λH 3600 × 105
8. Power control example in the lecture notes

(a) The S/(I + N ) must satisfy the following

P1
≥ 101.5
0.002P2 + 10−5
0.5P2
≥ 101.5
0.05P1 + 10−5
For minimum total signal power, it is equivalent to solve the above equations assuming
equality (why?). This leads to

P1 = 4.45 × 10−4 W P2 = 0.002W

(b) By increasing P1 and P2 , we can not achieve arbitrarily large SINR. Set the maximum SINR
the two users can jointly achieve as x, we have
P1
=x
0.002P2 + 10−5
0.5P2
=x
0.05P1 + 10−5
⇒ P1 = x(10−5 + 2 × 10−3 P2 ).
0.5P2
Substitue P1 into 0.05P1 +10−5
= x,

(10−6 + 2 × 10−4 P2 )x2 + 2 × 10−5 x − P2 = 0

10−6 x2 + 2 × 10−5 x
⇒ P2 =
1 − 2 × 10−4 x2
Since for the answer to be meaningful, P2 must be positive. Thus we must have

1 − 2 × 10−4 x2 > 0

Hence √
x< 5000 = 70.4848 = 18.48dB
Maximum achievable SINR is 18dB. Thus 20dB is not achievable for the above case.