Scribd
Upload
136 views

Assignment 3 Sol

This document provides solutions to an assignment with 4 questions. It includes: 1) Summaries of solutions to parts (a)-(c) of Question 1 regarding entropy calculations. 2) Summaries of solutions to parts (a)-(c) of Question 2 involving SVD and determinants. 3) Summaries of solutions to parts (a)-(e) of Question 3 about outage probability and throughput calculations for different data rates. 4) Summaries of solutions to parts (a)-(f) of Question 4 involving SVD, multiplexing gain, water-filling, beamforming, and capacity calculations for a MIMO channel.

Uploaded by

Chuah Chian Yeong
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
136 views

Assignment 3 Sol

This document provides solutions to an assignment with 4 questions. It includes: 1) Summaries of solutions to parts (a)-(c) of Question 1 regarding entropy calculations. 2) Summaries of solutions to parts (a)-(c) of Question 2 involving SVD and determinants. 3) Summaries of solutions to parts (a)-(e) of Question 3 about outage probability and throughput calculations for different data rates. 4) Summaries of solutions to parts (a)-(f) of Question 4 involving SVD, multiplexing gain, water-filling, beamforming, and capacity calculations for a MIMO channel.

Uploaded by

Chuah Chian Yeong
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

ENGN4536/ENGN6536 Assignment Three Solutions

The assignment will be marked out of 20 and is worth


5% of the nal mark
Solution to Question One (3 marks):
(a)
This is a Bernoulli RV with probability parameter 0.3. Using the denition
of entropy, we can compute it to be 0.8813. If the probability parameter changes
to 0.7, the entropy stays the same.
Ans: entropy is 0.8813 bits and it does not change as the probability changes
from 0.3 to 0.7. 1 mark
(b)
Entropy does not change with mean value of Gaussian distribution.
Ans: no change. 1 mark
(c)
Using I(X; Y ) = H(X) H(X|Y ) and the fact that H(X|Y ) = 0 (there
is no residual uncertainty in X if the value of Y is given), we get the mutual
information to be 1.
Ans: 1 bit 1 mark
1
Solution to Question Two (4 marks):
(a)
Perform SVD of H, we obtain the singular values to be 5.2645, 3.2713,
0.7637.
Ans: 5.2645, 3.2713, 0.7637 1 mark
(b)
In Matlab, hermitian transpose of H is computed as H

.
H

1 j 3 + 2j 1
1 +j 3 j 1
2 j 2 1

The eigenvalues of HH

is the square of the singular values of H, hence are


27.7154, 10.7014, 0.5833.
Ans: See above 2 marks
(c)
In Matlab, we perform the command det(eye(3)+H*H

).
Here is the mathematical derivation if you are interested in it (not required
to be mastered by you): Denote the SVD decomposition of H as H = USV

.
Then, HH

= US
2
U

, where U is a unitary matrix. Also, for any identity


matrix, we have I = UIU

. Then, we have the following


I +HH

= UIU

+US
2
U

= U(I +S
2
)U

Computing the determinant:


|I +HH

| = |U(I +S
2
)U

|
= |U

U(I +S
2
)|
= |(I +S
2
)| (1)
where we have used properties of determinant to move matrices around and also
the fact that U

U = I to arrive at the last equation. Now, note that (I + S


2
)
is a diagonal matrix and its determinant is given by the product of its diagonal
elements, i.e.,
|(I +S
2
)| = (1 + 27.7154)(1 + 10.7014)(1 + 0.5833) = 532. (2)
Ans: 532 1 mark
2
Solution to Question Three (6 marks):
(a)
With R
0
= 191940 bps and B = 30000, the outage probability is given by
p
out
= P(2
R
0
/B
1 > SNR) = P(SNR < 83.33),
which is the CDF of SNR evaluated at 83.33. The CDF of SNR is given by
P(SNR < ) = 1 exp
(


)
,
where = 333.33 is the average SNR.
The outage probability is computed as p
out
= 0.221.
Ans: p
out
= 0.221 1 mark
(b)
Following the steps in (a) with R
0
= 150000 bps, the outage probability is
computed as p
out
= 0.089.
Ans: p
out
= 0.089 1 mark
(c)
The average throughput is given by
C
out
= (1 p
out
)R
0
.
With R
0
= 191940 bps, we have C
out
= 149.48 kbps
Ans: C
out
= 149.48 kbps 1 mark
(d)
The average throughput is given by
C
out
= (1 p
out
)R
0
.
With R
0
= 150000 bps, we have C
out
= 136.68 kbps.
Ans: C
out
= 136.68 kbps 1 mark
(e)
This is an advanced question.
For a given R
0
, the outage probability is given by
p
out
= P(2
R
0
/B
1 > SNR) = 1 exp
(

2
R
0
/B
1

)
,
and hence the average throughput is given by
C
out
= (1 p
out
)R
0
= R
0
exp
(

2
R
0
/B
1

)
,
3
where B = 30000 and = 333.33. The optimal R
0
is found by maximizing C
out
above. This can be either solved analytically (by setting the rst derivative of
C
out
to zero and nd the root) or numerically using Matlab (by plotting C
out
v.s. R
0
over a large range of R
0
). Here, the numerical plot is presented in Fig. 1,
from which the optimal value of R
0
can be found.
0 0.5 1 1.5 2 2.5 3 3.5 4
x 10
5
0
5
10
15
x 10
4
the value of R
0
a
v
e
r
a
g
e

t
h
r
o
u
g
h
p
u
t
Figure 1: Average throughput vs. data rate
Ans: The optimal R
0
= 187.88 kbps 2 marks
4
Solution to Question Four (7 marks):
(a)
The SVD of H is computed using Matlab as [U, S, V ] = svd(H). Round
each entry to the second decimal, we have
U =

0.91 0.12 0.41


0.31 0.47 0.83
0.29 0.87 0.39

S =

1.63 0 0 0
0 0.40 0 0
0 0 0 0

V =

0.42 0.46 0.51 0.59


0.61 0.52 0.47 0.37
0.51 0.31 0.54 0.59
0.43 0.65 0.48 0.41

Ans: The three matrices shown above. 1 mark


(b)
The multiplexing gain is 2, since there are only 2 non-zero diagonal entries
in S. In other words, the rank of H is 2.
Ans: 2 1 mark
(c)
Water-lling policy should be used in this case. Dene the following SNR
terms
SNR
1
=

2
1
P

2
= 26.57
SNR
2
=

2
2
P

2
= 1.6
where
1
and
2
are the two diagonal entries in S.
Now we determine the cut-o value SNR
0
in the water-lling expression,
which depends on the number of active channels. Firstly, lets assume both
channels are active, hence
2

i=1
(
1
SNR
0

1
SNR
i
)
= 1. (3)
Solving the above equation gives SNR
0
= 1.2 which is smaller than both SNR
1
and SNR
2
. Hence, the assumption of both channels being active is valid.
5
Then, we compute the power distribution:
P
1
P
=
1
SNR
0

1
SNR
1
= 0.79
P
2
P
=
1
SNR
0

1
SNR
2
= 0.21
In other words, P
1
= 7.9 W and P
2
= 2.1 W.
The capacity is then given by
C
CSIT
=
2

i=1
Blog
2
(
SNR
i
SNR
0
)
= 24.42 (4)
Ans: Capacity is 24.42 bits per second 2 marks
(d)
If beamforming is used, only the strongest channel is active with full power
allocated. Then, the capacity is given by
C
BF
= Blog
2
(
1 +

2
1
P

2
)
= 23.92 (5)
Ans: Capacity is 23.92 bits per second 1 mark
(e)
If only the receiver has the CSI, uniform power allocation is used at the
transmitter. The capacity is given by
C
CSIR
= B
2

i=1
log
2
(
1 +

2
i
P
4
2
)
= 17.1 (6)
Note that there are 4 antennas at the transmitter, hence the total power is
divided by 4 in the above expression.
Ans: Capacity is 17.1 bits per second 1 mark
(f)
The capacity expression with total transmit power P

is given by
C
CSIR
= B
2

i=1
log
2
(
1 +

2
i
P

4
2
)
, (7)
which must be equal to 24.42. Solving this equality, we get P

= 21.33.
Ans: The new power value should be roughly 21.33 Watts. (any answer
between 20 - 23 will do) 1 mark
6

You might also like