Lecture 3

2
Outline
Signal fluctuations fading
Interference model detection of signals
Link model
3
Small scale propagation effects
Presence of reflectors, scattering and terminal motion
results in multiple copies being received at the mobile
terminal
Distorted in amplitude, phase and with different angle of arrivals
They can add constructively or destructively -> fluctuations in the
received signal

If there is no direct line of sight (NLOS), the received signal is




( )
( ) )) ( cos( ) (
rem) limit theo (Central r.v. Gaussian - ) ( , ) (
) sin( ) ( ) cos( ) (
t f t t C t r
t B t A
t t B t t A t r
s
s
+ =
+ =
e
e e
( ) ( )
on distributi Uniform -
) (
) (
tan f(t)
d distribute Rayleigh - ) (
1 -
2 2
t A
t B
t B t A t C
=
+ =
4
Envelope: Rayleigh distributed
( ) 0 s , exp
1
>
|
|
.
|

\
|
=
p
s
p
s f
S
( ) 0 c ,
2
exp
2
>
|
|
.
|

\
|
=
p
c
p
c
c f
C
p average power measured over
a time interval in the order of 1 sec
(lognormal r.v.)
Instantaneous power: exponential distributed
Instantaneous phase: uniformly distributed with pdf
t
t

2 0 ,
2
1
) ( s s = f f f
If there is a line of sight: LOS: ( ) ( ) t r t r t r
D s
+ = ) (
Direct component
( ) )) ( cos( ) ( t f t t C t r + = =
The envelope is a Rice r.v.
( )
( ) ( ) ( ) 0 order of function Bessel modified cos exp
1
0 ,
2
exp
0
0
0
2 2
=
s s
|
|
.
|

\
|
|
|
.
|

\
| +
=
}
t
t
dq q x x I
c
p
cD
I
p
D c
p
c
c f
C
( ) ( ) q t D t r
D
+ = e cos
5
p is the power in the scattered component; the long term average
power in r(t) is p+D
2
/2
The instantaneous received power chi-square distributed:



( ) 0 ,
2
2
2
exp
1
0
2
>
|
|
.
|

\
|
|
|
.
|

\
|
+
= s
p
s D
I
p
s D
p
s f
s
If D=0 -> exponential
6
Large and small time scale fading: summary
Fading effects - different at different time scales

the instantaneous signal envelope (short time scales (ms)) is
Rayleigh distributed (NLOS)
Rice distributed (LOS)

the mean value of the Rayleigh (or Rice) distribution can be considered
a constant for the shorter time scales, but in fact it is a random variable
with a lognormal distribution (large time scales (seconds))
caused by the changes in scenery (occur on a larger time scale)

the mean of the Lognormal distribution varies with the distance from the
transmitter according to the path loss law
If the mobile moves away or towards the transmitter (e.g. base station) the
received signal will also vary in time, according to the appropriate power law
loss model (e.g. free space: decreases proportional with the square of the
distance, etc.)

7
Large scale fading (shadow fading)
Described by a lognormal distribution, determined by
empirical measurements
No underlying physical phenomenon is modeled
Small scale fading underlying physical
phenomena
Multipath
Multiple copies of the signal arrive at destination
Doppler shift of the carrier frequency
relative motion of the receiver and transmitter causes
Doppler shifts
yields random frequency modulation due to different
frequency shifts on the multipath components

8
Doppler effect
Can be caused by
the speed of mobile
speed of surrounding objects
If the surrounding objects move at a greater speed than the mobile,
this effect dominates, otherwise it can be ignored
Doppler shift and Rayleigh fading
Mobile moving towards the transmitter with speed v: a maximum
positive Doppler shift


The n-th path, moving within an angle o
n
,

has a Doppler shift of

v
f
d
=
max
( )
n
n
d
v
f o

cos =
v
o
n

n-th path
The random phase for the n-th path:
n n n
t f | t u + = 2
9
( ) ( ) ( )
( )
( ) ) 2 sin(
) 2 cos(
) 2 sin( ) 2 cos(
1
0
1
0
n n
N
n
n s
n n
N
n
n c
c S c c
t f C E t T
t f C E t T
t f t T t f t T t E
| t
| t
t t
+ =
+ =
=

=
=
It can be shown that the E-field can be expressed as the in-phase
and quadrature form (Doppler shift very small compared to the carrier
frequency narrow band process):
Gaussian r.v.
C
n
does not change significantly over small spatial distances, so fading is
primarily due to phase variations caused by the Doppler shift.
Using Clarkes model (waves arrive with equal probability from all directions),
the spectrum of the signal can be determined to be
( )
2
1
5 . 1
|
|
.
|

\
|

=
m
c
m
E
f
f f
f
f S
t
( )
antenna vertical , 4 /
ow , 0
2 0 ,
2
1

t o
t
o

s s
= p
when
( ) ( ) Rayleigh - ) (

t T t T t r
S c
+ =

10
Therefore, the power spectral density of the received signal can be
represented as in the following figure:
m c
f f
m c
f f +
c
f
S
E
f
Doppler spread leads to frequency dispersion and time selective fading
However, there is another phenomenon related to the multipath propagation,
which introduces time dispersion and frequency selective fading:
multipath delay spread
The small scale fading considered up to now, assumes that all the frequencies in
the transmitted signal are affected similarly by the channel (flat fading).
11
Multipath delay spread
Multiple copies of the signal arrive with different delays
May cause signal smearing, inter-symbol interference (ISI)
The power delay profile gives the average power (spatial
average over a local area) at the channel output as a function of
the time delay.

Power
Delay
Power
Delay
Average delay
RMS delay spread
( ) t |
( )
k
P t
( )
k
t
( ) t
( ) ( )
( )
}
}

=
0
0
2
t t |
t t | t
o
t
d
d
T
( )
( )
}
}

=
0
0
t t |
t t t|

t
d
d
( )
2
2
t t o =
T
( )
( )

=
k
k
k
k k
P
P
t
t t
t
2
2
( )
( )

=
k
k
k
k k
P
P
t
t t
t
RMS delay spread Average delay
12
Interpreting the delay spread in the frequency domain
While the delay spread is a natural phenomenon, we can define the
coherence bandwidth as a measure derived from the RMS delay spread
Coherence bandwidth B
c
= statistical measure of the range of
frequencies over which the channel can be considered to be flat (i.e.,
the channel passes all the spectral components with approx. equal gain
and phase)


o
T
and B
c
describe the nature of the channel in a local area; they offer
no information about the relative motion of the transmitting and the
receiving mobile terminals.

Doppler effect interpretation

Spectral broadening B
D
is a measure for the rate of changes of the
mobile radio channel due to Doppler effects
If the bandwidth of the baseband signal is much greater than B
D
, the effect
of doppler shift is negligible

is the time duration over which the channel impulse response is
essentially invariant

T
Bc
o
1
~
D
C
B
T
1
~
13
Small scale fading: classification
Flat Fading: the channel has a constant response for bandwidth
greater than the transmitted signal bandwidth




Frequency Selective Fading


T S
C S
T
B B
o >>
<<
S(f)

R(f)
T S
C S
T
B B
o <
>
S(f)

R(f)
C(f)
C(f)
Rule of thumb: frequency selective if
S T
T 1 . 0 > o
Needs channel equalization
14
Fast fading channel impulse response changes rapidly within the
symbol duration



Slow fading channel impulse response changes at a rate much
slower than the transmitted symbol bandwidth
Small scale fading: classification
C S
D S
T T
B B
>
<
C S
D S
T T
B B
<<
>>
Flat
slow
Freq sel.
slow
Freq sel.
fast
Flat
fast
T
o
S
T
S
T
S
T
Freq. sel.
Fast
Freq. sel.
slow
Flat
Fast
Flat
Slow
D
B
C
B
S
B
S
B
Summary of channel fading characteristics
C
T
15
Fading and time scales
Time scales for analysis are important for selecting the
correct fading model
If lots of averaging ignore Rayleigh fading
If analysis looks at the bit level: Rayleigh fading counts
To combine the effects, consider the averaging of the conditional pdf
(Y/X) obtain the marginal pdf of Y

( ) ( ) ( )dx x f x X y f y f
X
b
a
X Y Y
}
= = /
/
| | ( ) x f b a
X
on distributi of support ,
16
Physical Layer: Link Model
A B
p
Link probability = probability that a
link is going to be available for
transmission, i.e., meet target SIR
requirements
p affected by:
- path loss (depends on the distance to the receiver) - mobility
- Lognormal fading (depends on the location and environment)
- mobility
- Rayleigh fading mobility
- Interference may dynamically vary
- mobility
- traffic burstiness
- arrival/departure statistics
17
Dynamic adaptation algorithms
Fading affects useful signal strength
Power control
Adaptive modulation
Adaptive coding
Antenna Diversity
Adaptive MAC MAC Layer
Route diversity
Adaptive channel allocation

Interference: determines the equivalent noise level SINR
Power control
Adaptive modulation
Adaptive coding
Smart Antennas beamforming
Interference cancellation
Adaptive MAC MAC Layer
Interference aware routing
Admission control
Adaptive channel allocation (frequency, time slot, code)


}
Physical layer
}
Physical layer
}
Network Layer
} Network Layer
Not
adaptive
18
General model of signals and interference in a
multi-user wireless system
H1
H2
H3
H4
+
Receiver
( ) t s
0
( ) t r
0
( ) t s
2
( ) t s
M 1
( ) t s
M
) (t n (AWGN)
If the channel response is flat: multiply the signal with an attenuation factor
- this factor is a random variable (pdf selected according to the
appropriate fading model)
Transfer function of the channel
incorporates the fading effects
Desired signal
Received signal
19
Detection of signal in noise
Consider that in the previous general model, we want to detect the
information bit a
0
carried by the signal s
0
(t):



the detection problem is illustrated for the simplest case for which no
interferers are present and the channel does not introduce any
fading
At the receiver, we need to estimate , such that the probability of
error would be minimized. We denote our decision estimate by .
We know that the information bit transmitted was either +1 or -1,
with equal probability: this is called a priori probability

( )

( ) ( ) 1 0 , 2 cos
2
carrier
bit n informatio
0 0
T t t f
T
E
a t s
c
s s =

t
0
a
0
a
( ) ( )
2
1
1 1
0 0
= = = = a P a P
For notation simplicity, we denote
( )
( )

= =
+ = =
1
1
0 2
0 1
a P P
a P P
The bit period
20
To understand digital modulation and demodulation is important to know
that a signal can be represented equivalently both in time domains and in
signal space domain

For the example considered, formula (1) is the time representation of the
signal
.
If we denote by , as the basis function that

describes the signal space for this example, the signal constellation can be
represented as in the following figure

( ) t s
0
( ) ( ) T t t f
T
t
c
s s = 0 , 2 cos
2
t
0
( ) 1
0
= a
E
( ) 1
0
=

a
E
( ) t
21
Thus, in the signal space domain, the received signal can be expressed as
n E a n s r + = + =
0 0
n is a Gaussian random variable with zero mean and variance
2
0
2
N
= o
We will decide that was transmitted if the a posteriori conditional

probability (conditioned on the received r) is larger for than that for


1
0
+ = a
1
0
+ = a
1
0
= a
( ) ( ) 1 / 1 / 1
0 0 0
+ = = > + = a r a P r a P
This is called the maximum a posteriori probability rule: MAP rule
Using Bayes rule, we express
( )
( )
( ) r p
P a r p
r a P
1 0
0
1 /
/ 1
+ =
= + =
( ) 2
( ) 3
( ) 4
22
Then, from (3) and (4), we have
( ) ( ) 1 1 / 1 /
0 2 0 1 0
+ = = > + = a P a r p P a r p
From (2), we see that
( )
0
/ a r p is a Gaussian random variable, with mean
( ) E a a r E
0 0
/ = and variance
2
o
( ) 5
( ) 5
becomes
( )
( )
1 1
2
exp
2
1
2
exp
2
1
0
1
2
2
2
2
2
+ = = >
|
|
.
|

\
|
+

|
|
.
|

\
|

a
P
P
E r
E r
o o t
o o t
After simplification, we take logarithm on both sides and we obtain after computation:
1 0
1 0
1 0 4
0
0
0
= <
+ = >
+ = >
a r
a r
a E r
Decision regions
23
The probability of error can be computed as
( ) ( )
( )
( ).
2
exp
2
1

) (
0
1 / 0 1 / 1
2
2
2
0 0 0 1 /
0
SNR Q
E
Q
dn
n
E n P
E n P
a r P a a P
E
a e
= |
.
|

\
|
=
|
.
|

\
|
=
= > =
> =
= > = = + =
}

+ =
o
o o t
2
1
when ,
2 1 1
0
/
1
0
/ 2 1
0
/ 1
= = =
+ =
+ =
= + =
P P P
P P P P P
a e
a e a e e
( )
}

|
|
.
|

\
|
=
x
dt
t
x Q
2
exp
2
1
2
t
24
The structure of the detector is
( ) t r
+
( )dt
T
}
0
.
r
( ) . sgn
1 0
1 0
0
0
= <
+ = >
a
a

correlator

b) If an interferer

is present, then a similar derivation shows that
n E a E a r + + =
1 1 0
( ) ( ) T t t f
T
E
a t s
c
s s + = 0 , 2 cos
2
1
1 1
| t
and
( ) . cos
2
1
1
2
, 1 /
1 0
|
.
|

\
|
+ =
+ =

o o
E
a
E
Q P
a a e
( ) t
25
If ( ) 0 cos = the signals are orthogonal and there is no interference
(the signals are completely separated)
The signals can be separated in
- frequency : FDMA (frequency division multiple access)
- time: TDMA (time division multiple access)
- using different signature codes: CDMA (code division multiple access)
If the signals are orthogonal, the simple correlation receiver (or the equivalent
matched filter) is optimal for detection in Gaussian noise
Disadvantage of orthogonal signals: require additional bandwidth:
The number of orthogonal waveforms N of duration T that exist in a bandwidth W
Is limited by:
- for coherent detection (a phase reference is available)

- for non-coherent detection (without a phase reference)

WT N 2 s
WT N s
26
If fading is also considered:
n E a h E a h r + + =
1 1 1 0 0
where are random variables (e.g. Rayleigh, lognormal, etc)
1 0
, h h
The probability of bit error or bit error rate (BER) is a key measure for the performance
of the physical layer. In general, computing the BER can be quite complex, and in
practice, the link quality can be measured using a mapping for the BER
performance requirement into a signal to interference ratio (SIR) requirement.

Thus SIR constitutes a key performance measure for the link quality.
Sometimes the link performance is measured using SINR (signal to interference
and noise ratio). Many times, the use of the SIR acronym is used to denominate
In fact the signal to interference and noise ratio.
27
Reading assignment for next class

V. Kawadia and P.R. Kumar, A cautionary
Perspective on Cross Layer Design, University of
Illinois at Urbana Champaign, preprint;
http://decision.csl.uiuc.edu/~prkumar/psfiles/cross-layer-
design.pdf