Scribd
Upload
23 views6 pages

EECS 224B: Fundamentals of Wireless Communications Spring 2006

This document provides solutions to problems from a problem set on fundamentals of wireless communications. Solution 1 derives an expression for channel capacity when only the I channel is used. It shows capacity is approximately C/2 at high SNR. Solution 2 formulates transmit beamforming as an optimization problem to maximize received SNR. Solution 3 analyzes channel capacity for different MISO channel models. Solution 4 derives an outage probability expression for a Rayleigh fading MISO channel and how it improves with more antennas. Solution 5 calculates minimum Eb/N0 requirements and capacity losses for repetition coding.

Uploaded by

MohamedSalah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
23 views6 pages

EECS 224B: Fundamentals of Wireless Communications Spring 2006

This document provides solutions to problems from a problem set on fundamentals of wireless communications. Solution 1 derives an expression for channel capacity when only the I channel is used. It shows capacity is approximately C/2 at high SNR. Solution 2 formulates transmit beamforming as an optimization problem to maximize received SNR. Solution 3 analyzes channel capacity for different MISO channel models. Solution 4 derives an outage probability expression for a Rayleigh fading MISO channel and how it improves with more antennas. Solution 5 calculates minimum Eb/N0 requirements and capacity losses for repetition coding.

Uploaded by

MohamedSalah
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

EECS 224B: Fundamentals of Wireless

Communications
Spring 2006

Problem set 8: Solutions

Solution 1. (Exercise 5.1) If only the I channel is used, only W2 of the W complex
degrees of freedom are available. The noise affecting the I channel is N02W so we get:
!
!
2P
W
2P
W
=
log2 e
log2 1 +
ln 1 +
CI =
2
N0 W
2
N0 W
At low SNR we have
CI

W 2P
P
=
=C
2 N0 W
N0

At high SNR we have


W
CI
2

!
1
P
W
C
log2
+ log2 2 = C +

N0 W
2
2
2

From Exercise 3.4 we know that to achieve the similar error performance in 2k PAM
and 2k QAM we need a b. If we set the SNR of both modulations to be equal we
get:

Eav 2kQAM

2a2 kQAM
2
1
Eav 2kQAM QAM =
3
2
k

b
Eav 2 P AM P AM =
22kP AM 1
3

QAM = Eav 2kP AM P AM 2 2kQAM 1 = 22kP AM 1


kQAM 2kP AM

(for large SNR, i.e. large k)

This is consistent with the above result, that says that the capacity of the I channel is
roughly half of that of the complex channel at high SNR.
2

Solution 2. (Exercise 5.11) The received SNR is given by EkhN0xk . Hence we need
to maximize this quantity over all x CL such that Ekxk2 P for fixed N0 and some
constant P > 0. The optimization problem is therefore:

max h Kx h
Kx

subject to the constraint that the trace of the covariance matrix Kx is less than P .
The solution of the optimization problem is to make Kx rank 1, with eigenvector h,
i.e. the transmit beamforming strategy.
Solution 3. (Exercise 5.13)

1. Let 1 = [1, 1, , 1]T RL . Then the channel equation is:


y = 1x + z

(1)

where z CN (0, N0 IL ) and x must satisfy the power constraint E[x] P .

We note that we can project the received signal onto the direction of 1 obtaining
the sufficient statistic:

(2)
y = Lx + z
r=
L

where z CN (0, N0 ). Defining x = Lx we see that we have an AWGN


channel

with power constraint LP and noise variance N0 . Therefore C = log 1 + LP


.
N0
We see that there is a power gain of L with respect to the single receive antenna
system.
2. Let h = [h1 , h2 , , hL ]T C L . Then the channel equation is:
y = hx + z

(3)

where z CN (0, N0 IL ) , h is known at the receiver and x must satisfy the power
constraint E[x] P .

Since the receiver knows the channel, it can project the received signal onto the
direction of h obtaining the sufficient statistic:
r=

h
y = khkx + z
khk

(4)

where z CN (0, N0 ). Then the problem reduces to computing the capacity of


a scalar fading channel, with fading coefficient given by khk. It follows that:

khk2 P
LP khk2
C = E log 1 +
= E log 1 +
(5)
N0
N0 L
h 
i
2
In contrast, the single receive antenna system has a capacity C = E log 1 + |h|N0P .
The capacity is increased by having multiple receive antennas for two reasons:
2
has the same
first there is a power gain L, and second khk
L
h mean
i but less varikhk2
2
ance than |h| , and we get a diversity gain. Note that V ar L = 1/L whereas
V ar [|h|2 ] = 1.

2
a.s. 1, so it follows that C log 1 + LP
As L , khk
for large L.
L
N0
3. With full CSI, the transmitter knows the channel, and for a given realization of
the fading process {h[n]}N
n=1 the channel supports a rate:

N
kh[n]k2 P [n]
1 X
(6)
R=
log 1 +
N n=1
N0
2

and the problem becomes that of finding the optimal power allocation strategy.
We note that the problem is the same as the one corresponding to the case of a
single receive antenna, replacing |h[n]|2 by kh[n]k2 . It follows that the optimal
solution is also obtained by waterfilling:

1
N0 +

P (khk ) =
(7)
khk2

where is chosen so that the power constraint is satisfied, i.e. E[P (khk2 )] = P .
The resulting capacity is:

khk2 P
(8)
C = E log 1 +
N0

At low SNR, when the system is power limited, the benefit of having CSI at the
transmitter comes from the fact that we can transmit only when the channel
is good, saving power (which is the limiting resource) when the channel is bad.
The larger the fluctuation in the channel gain, the larger the benefit. If the
channel gain is constant, then the waterfilling strategy reduces to transmitting
with constant power, and there is no benefit in having CSI at the transmitter.
When there are multiple receive antennas, there is diversity and khk2 /L does not
fluctuate much. In the limit as L we have seen that this random variable
converges to a constant with probability one. Then, as L increases, the benefit
of having CSI at the transmitter is reduced.
4.

Pout



khk2 P
N0
2
R
= P r log 1 +
< R = P r khk < (2 1)
N0
P

(9)

We know that we can approximate the pdf of khk2 around 0 by:


f (x)

1
xL1
(L 1)!

(10)

where Rayleigh fading was assumed, and hence the distribution function of khk2
evaluated at x is approximately given by:
1 L
x
(11)
L!
for x small. Thus, for large SNR we get the following approximation for the
outage probability:


1
N0 L
Pout
(12)
(2R 1)
L!
P
F (x)

We see that having multiple antennas reduces the outage probability by a factor
of (2R 1)L /L! and also increases the exponent of SNR1 by a factor of L.
3

Solution 4. (Exercise 5.20)


1. The low-SNR -outage probability approximation of the parallel channel is given
by (see Exercise 5.19, part 3):

LC
2
P khk <
= ,
SNR log2 e
where C denotes the per-channel -outage capacity, i.e., the largest rate achievable while maintaining outage probability below . Let F (x) = P(khk2 > x) be
the complementary CDF of h. Then we have that
C =

1 1
F (1 )SNR log2 e.
L
1

2. For Rayleigh i.i.d. fading branches, F 1 (1 ) (L!) L L , and so


C =

1
1
1
(L!) L L SNR log2 e,
L

is the per-channel outage capacity.


3. The delay-spread of the channel is 1s. Hence, from equation (2.47), page 33,
1
we know that the coherence bandwidth is 210
6 = 0.5 MHz. But the available
bandwidth is 1.25 MHz. Hence, if we exploit the frequency coherence, we can have
two independent, parallel channels in frequency. Also, since our time constraint
is 100ms and the coherence time is 50ms, we have two parallel channels in time.
This makes a total of four parallel channels that we can exploit. Consequently,
we let L = 4 in our calculations.
Since the SINR per chip is 17 dB and the processing gain is G = W/R =
1.25MHz100ms = 125000, the SINR per bit per user is roughly 34 dB. Plugging
in these values into the formula given in part (2) of this question, we get that
C0.01 is roughly 631 bps/Hz/user.
On the other hand, the capacity of the unfaded AWGN channel with the same
SNR is roughly 3607 bps/Hz. Thus the 1%-outage capacity of this parallel channel is roughly 17.5% of the unfaded AWGN channel with the same received SNR.
Solution 5. (Exercise 5.23)
1. For the AWGN channel the maximum achievable rate is given by:



P
Eb R
= W log 1 +
R = W log 1 +
N0 W
N0 W
where we used P /R = Eb .
4

(13)

Then, the minimum required Eb /N0 for reliable communication is:


 

W R/W
Eb
2
1
=
N0 req
R

For the IS-95 system we get NEb0 req = 0.695 = 1.58dB.

(14)

At low SNR R/W is small, and we can approximate 2R/W = exp[(R/W ) ln 2]


1 + (R/W ) ln 2, to get
 
Eb
ln 2 = 1.59dB
(15)
N0 req
and we see that as the SNR goes to zero, the minimum Eb /N0 requirement is
-1.59dB.
2. Since we are forced to repeat each transmitted symbol 4 times, we consider the
received signal in a block of length 4:
y = 1x + z

(16)

and use 3.a) to conclude that I(x; y) log(1 + 4P/N0 ) where the upper bound
can be achieved by choosing the input distribution to be CN (0, P ) i.i.d.. Then
the maximum achievable rate (in bits/s/Hz) of this strategy is:


1
4P
Rmax = log 1 +
(17)
4
N0
which is strictly smaller than log(1 + P/N0 ), the capacity of the AWGN channel.
The loss is due to the concavity of the log() function. For small x, log(x) is
approximately linear and the loss due to concavity is small for low SNR. On the
other hand, repetition coding has a large loss for high SNR.
3. Loss is greater at high SNR where the loss of d.o.f. is felt more.
4. For repetition coding the minimum Eb /N0 required for reliable communication is
given by:
!
 
W 24R/W 1
Eb
=
(18)
N0 req
R
4
The increase in Eb /N0 requirement is:
E
b

E 0

req(rep)

N0 req(AW GN )

24R/W 1
=
4(2R/W 1)

For the IS-95 system this loss is only 0.035dB.


5

(19)

You might also like