Sei sulla pagina 1di 8

KEEE494: 2nd Semester 2009 Week 6

Review: Wireless Channels and Performance over Fading


Channels
1 Wireless Channel
Wireless channel can be characterized by the three independent factors:
Large scale channel variation
(1) Path loss
(2) Shadowing
Small scale channel variation
(3) Multi-path fading
Depending on the Delay spread, the multi-path fading can be categorized as
_
(frequency) Flat fading
Frequency selective fading
Depending on the Doppler spread, the multi-path fading can be categorized as
_
Slow fading
Fast fading
Path loss: The received power P
r
at a distance d from the transmitter can be written as
P
r
=
P
t
A
(4d)

where is the wavelength, is the path loss exponent (which is two in the free space), and A is the constant
reecting the transmit and receive antenna gains and some additional loss due to the hardware implementation.
Shadowing: The long-term average power P
r
is a deterministic value when we consider only the path loss.
However, in reality, P
r
is different when we measure the long-term average power at the different locations but
with the same distance. The shadowing models this phenomena such as
P
r
= P
t

or
P
r,(dB)
= P
t,(dB)
+
(dB)
where
(dB)
is modeled as zero mean Gaussian with a variance
2

, that is,
(dB)
N(0,
2

).
Multi-path fading
Doppler spread: The Doppler frequency denoted as f
D
is dened as
f
m
=
v cos

=
v cos
c
f
c
= f
D
cos
where c is the speed of light, f
c
is the carrier frequency, f
D
is the maixmum Doppler frequency, is the
angle of arrival, and v is the vehicle speed.
1
If v 0 or equivalently, f
D
0, the channel is slowly varying. On the other hand, if v >> 0, the channel
is fast varying.
Now let us dene the coherence time, T
coh
, of which duration the channel does not change fast and is rather
static. Of course, T
coh
is proportional to the Doppler spread f
D
and the usual denition of the coherence
time is given as
T
coh
=
9
16f
D
Figure 1: Doppler frequency where is the wavelength, f
m
is the maximum Doppler frequency and is the angle of
arrival.
Hence, if a symbol duration, T, is less than T
coh
, that is, T < T
coh
, it is slow fading. Otherwise, it is called
fast fading.
Delay spread The channel impulse response over the wireless channel can be written as
Figure 2: Multi-path fading environment.
h(t) = c
1
(t
1
) +c
2
(t
2
) + +c
N
(t
N
),
and it frequency response H(f) is
H(f) =
N

n=1
c
n
e
j2f
1
The channel coherence bandwidth is dened as
B
coh
=
1

N

1
We can observe that if |
N

1
| is larger than a symbol duration T, the intersymbol interference (ISI) can
occur.
If |
N

1
| is very small compared with a symbol duration T, or equivalently, B
coh
>
1
T
, ISI effect is
negligible and in this case, the channel is called frequency at fading. On the other hand, if B
coh
<
1
T
, the
channel is called frequency selective fading. There are some denitions related to the delay spread.
2
Power delay prole: The set of mean power prole of each path is called power delay prole as shown
in Fig. 3.
Figure 3: Power delay prole where c
n
is the amplitude of nth path, which is random process, and its local mean
power is given as E[c
2
n
] =
n
.
Mean delay spread is dened as
=

N
n=1

n

N
n=1

n
Root Mean Square (RMS) delay spread is dened as
RMS delay spread =
_

2
( )
2
where

2
=

N
n=1

n

2
n

N
n=1

n
Flat fading channel model can be simplied as shown in Fig. 4.
Frequency selective fading channel can be modeled as shown in Fig. 5.
Figure 4: Block diagram of the at fading channel model, where

is the fading coefcient with E[|

|
2
] = 1.
In Fig2. 4 and 5, the fading channel coefcient

n
= |

n
|e
j
n
is normalized to have a unit energy such as
E[|

n
|
2
] = 1.
Statistical model of

and SNR
Note that

n
= |

n
|e
j
n
=

n,I
+j

n,Q
For at fading channel, the time difference between the rst and the last arriving path signals is negligible, say

n
= for all n. Hence, we can write the channel impulse response as

=
N

n=1

n
(t tau
n
)
N

n=1

n
(t tau)
Or, for simplicity we can just remove the delay factor yielding

=
N

n=1

n
=
I
+j
Q
where

I
=

N
n=1

n,I
and

Q
=

N
n=1

n,Q
3
Figure 5: Block diagram of the selective fading channel model.
Rayleigh channel: In the Rayleigh channel,

I
and

Q
is modeled as Gaussian random process such as

I
,

Q
N
_
0,
1
2
_
Note that in the above the correlation between

I
and

Q
is zero, that is, E[

Q
] = 0. Then, |

| follows
the Rayleigh distribution and its power |

|
2
follows the exponential distribution.
Let us dene =

. Then, is also Rayleigh distributed with the PDF given by


p

(x) =
2x

exp
_

x
2

_
Also, its power
2
is exponential distributed random variable of which PDF is given by
p

2(x) =
1

exp
_

_
Let us dene the received signal-to-noise ratio (SNR) as
=

2
E
s
N
0
Then, the PDF of is found to be given as
p

(x) =
1

e
x/
where
=
E
s
N
0
Ricean channel: In the Ricean channel model, there exists the line-of-sight (LOS) component. Let
0
denote the LOS component which is deterministic value. Then, =

N
n=0

n
is the sum of the LOS
component (
0
) and the non-LOS components (
n
for n = 1, ...N). Note that the sum of non-LOS
component follows the Gaussian random variable with zero mean. Let
I
=

N
n=1

n,I
and
Q
=
4

N
n=1

n,Q
, then
I
and
Q
are modeled as zero mean Gaussian random variable with variance /2.
Hence, in the fading coefcient =
I
+ j
Q
, the mean values of
I
and
Q
are not zero but certain
deterministic values. Let us denote those mean values are
E[
]
= m
I
E[
Q
] = m
Q
Then, in the Ricean channel model,
I
and
Q
are non-zero Gaussian random variable such as

I
N
_
m
I
,
2
_

Q
N
_
m
Q
,
2
_
As in the Rayleigh fading case, the correlation between
I
and
Q
is zero, E[

Q
] = 0. Then, total
power from the channel is
E[||
2
] = = m
2
I
+m
2
Q
+ 2
2
.
The Ricean factor K is dened as the ratio of the powers between the LOS component and the non-LOS
component such as
K =
s
2
2
2
where s
2
= m
2
I
+ m
2
Q
. Since the ||
2
follows the non-central chi-square distribution, we can nd the
PDFs of || and ||
2
given, respectively as,
p
||
(x) =
2x(K + 1)

exp
_
K
(K + 1)x
2

_
I
0
_
2x
_
K(K + 1)

_
p
||
2(x) =
K + 1

exp
_
K
(K + 1)x

_
I
0
_
2
_
K(K + 1)x

_
Also, the PDF of the SNR is
p

(x) =
(K + 1)

exp
_
K
(K + 1)x

_
I
0
_
2

K(K + 1)x

_
Nakagami-m channel: The PDFs of || and ||
2
given, respectively as,
p
||
(x) =
2m
m
x
2m1
(m)
m
exp
_

mx
2

_
, m
1
2
p
||
2(x) =
_
m

_
m
x
m1
(m)
exp
_

mx

_
The Ricean distribution can be closely approximated by using the following relation between the Ricean
factor K and the Nakagami shape factor m:
K =

m
2
m
m

m
2
m
m =
(K + 1)
2
(2K + 1)
5
2 Performance
The bit error rate of coherent digital modulation is generally given as
P
b
(E|) aQ(
_
b)
where is the received SNR and a and b are the certain values depending on the modulation schemes.
For example for BPSK, a = b = 2, so that P
b
(E|) = 2Q(

2). It can be shown that the Gaussian Q


function can be equivalently written as
Q(x) =
1

_
2
0
exp
_

x
2
2 sin
2

_
d.
We are interested in average probability of error performance over fading channel which can be obtained
as
P
b
(E) =
_

0
P
b
(E|)p

() d
=
_

0
aQ(
_
b)p

() d
=
_

0
a
1

_
2
0
exp
_

b
2 sin
2

_
p

() dd
=
a

_
2
0
__

0
p

() exp
_

b
2 sin
2

_
d
_
d
=
a

_
2
0
M

b
2 sin
2

_
d
where M

(s) is the moment generating function (MGF) of dened as


M

(s) =
_

0
p

()e
s
d
Example: Consider the Rayleigh fading channel in which case the PDF of is given as
p

() =
1

exp
_


_
, 0,
and the MGF of is found to be
M

(s) =
_

0
1

e
/
e
s
d = (1 s)
1
On the other hand, for the Nakagami-m fading, we can obtain the MGF of as
M

(s) =
_
1

m
_
m
.
Let us dene the integral
I
n
(c) =
1

_
2
0
_
sin
2

sin
2
+c
_
n
d,
where n is the integer. Then, it can be known as
I
n
(c) =
1
2

_
1
2
A(c)
_
n1

i=0
_
2i
i
_
[A(c)]
i
[1 A(c)]
i
6
where A(c) =
1
2
_
1
_
c
1+c
_
. If n = 1,
I
1
(c) =
1
2
_
1
_
c
1 +c
_
.
For Rayleigh, the average probability of error is
P
b
(E) =
a

_
2
0
_
1 +
b
2 sin
2

_
1
d
=
a

_
2
0
2 sin
2

2 sin
2
+b /2
d
= aI
1
(b /2) =
a
2
_
1

b /2
1 +b /2
_
Example: For BPSK, a = 1 and b = 2. Hence, the average BER over Rayleigh fading channel is
P
b
(E) =
1
2
_
1
_

1 +
_
For binary DPSK, the conditional BER is given as P
b
(E|) =
1
2
e

and the average BER is


P
b
(E) =
1
2
M

(1)
For M-PSK, the conditional symbol error rate (SER) given is given as
P
s
(E|) =
1

_ M1
M

0
exp
_

g
psk

sin
2

_
d
where g
psk
= sin
2
_

M
_
. Then, the average SER over fading channel is
P
s
(E) =
_

0
P
s
(E|)p

() d
=
1

_ M1
M

0
M

sin
2
(/M)
sin
2

_
d
For M-QAM, the conditional SER given is given as
P
s
(E|) = 4
_
1
1

M
_
Q(
_
2g
QAM
) 4
_
1
1

M
_
2
Q
2
(
_
2g
QAM
)
where g
QAM
=
3
2(M1)
. Recall that Q(x) =
1

_
/2
0
exp
_

x
2
2 sin
2

_
d. Then we can show that
Q
2
(x) =
1

_
4
0
exp
_

x
2
2 sin
2

_
d.
Then, the average SER over fading channel is
P
s
(E) = 4
_
1
1

M
__
2
0
M

g
QAM
sin
2

_
d
4
_
1
1

M
_
2
_
4
0
M

g
QAM
sin
2

_
d
7
For MSK, the conditional SER given is given as
P
b
(E|) = 2Q(
_
2) 2Q
2
(
_
2)
=
2

_
2
0
exp
_


sin
2

_
d
2

_
4
0
exp
_


sin
2

_
d
=
2

_
2

4
exp
_


sin
2

_
d
Then, the average BER over fading channel is
P
s
(E) =
2

_
2

4
M

1
sin
2

_
d.
8

Potrebbero piacerti anche