Announcement

The TA prefers to have the paper copy of

the HW. From next time, please bring
paper copy of your HW solution.

Review Ch.2
Power decays with distance
Shadowing
Path-loss
+?
Multiplath

PL+S =Outage Probability

& Cell Coverage Area

Pout (d, Pmin )

Pmin,dB Prx,dB (d)

=Q

C = 1/2 + e2/b Q(2/b),

d :=

10 log10

dB

dB

Prx,dB (d) = Ptx,dB + KdB

10 log10 (e)

d
d0

Pmin,dB = Prx,dB (R)

Statistical Multipath Model

Random number of multipath components, each with

!
!
!
!

Random amplitude
Random phase
Random delay
Random Doppler shift (which well consider later on)

These components change with time (can either add

constructively or destructively)

Leads to time-varying channel impulse response

Statistical Multipath Model

1/B
1 2

t
Delay spread = 3- 1

Time Varying Channel

!

Each path is associated with ergodic and stationary:

! attenuation (t) (path-loss and shadowing),
! delay (t) = r(t)/c,
! phase (t) = 2 [ fc (t) - u<t v/ cos((u)) du ]
with: c = speed of light,
r(t) = path length,
fc = carrier frequency,
= wavelength,
v = velocity,
(t) = angle of arrival.

Signal Model (Appendix A)

!

Bandpass real-valued signal

Transmit signal: s(t) bandpass, u(t) baseband

Channel: hl(t) baseband:

Received signal:

Time Varying Channel

Baseband signals: u(t) sent, v(t) received:

i.e., time varying linear system.

How the transmitted signal u(t) is affected by the
channel depends on how its bandwidth B
compares with the channel delay spread.
Two cases to consider:
! narrowband channel and
! wideband channel.

Narrowband Channel
1/B
123

t
Delay spread = 3- 1 << 1/B

Wideband Channel
1/B
1

3
Text

t
Delay spread = 3- 1 >> 1/B

Time Varying Channel

!

Response of channel at t to impulse at t-:

! t is time when impulse response is observed

! t- is time when impulse got into the channel
! is how long ago impulse was put into the

channel for the current observation

! n(t) = 2 fc n - Dn
! fc n can be quite large --> huge phase variations
over small time

Time Varying Channel

Delay Spread
!

Delay spread := maxm `not= n|n(t)-m(t)|

Delay spread << 1/bandwith: narrowband fading,

non-resolvable fading = no inter-symbol interference

Delay spread >> 1/bandwith: wideband fading,

or resolvable fading = inter-symbol interference

Delay spread is a random variable --> needs statistical

characterization

If path attenuation is too low then that path component is

buried in noise and should not be counted in the delay
spread --> power delay profile

Indoor: fraction of microseconds;

Outdoor: several microseconds;

Narrowband Channel
1/B
123

t
Delay spread = 3- 1 << 1/B

Narrowband Channel
!

Delay spread = maxm,n|n(t)-m(t)|<<1/B,

then u(t)u(t- n(t)) for all n (non-resolvable multipath)

Channel input response becomes

c(, t) =

n (t)e

n (t)

( )

No signal distortion/spreading in time

Multipath = complex scale factor attenuation

Multipath = random variable

Wideband Channel
1/B
1

3
Text

t
Delay spread = 3- 1 >> 1/B

Wideband Channel
Delay spread = maxm,n|n(t)-m(t)|>>1/B,
delays are a continuum (resolvable multipath)
Channel response given by

!
!
!

A continuum of multipath components

Signal distortion/spreading in time
Multipath = random process

Narrowband Channel
c(, t) =

X
n

n (t)e

n (t)

( )

c(, t) =

n (t)e

n (t)

Narrowband Channel
n

Delay spread << 1/B

u(t)=1 (unmodulated carrier)
X
j2fc t
r(t) = <{u(t)e
n (t)e
n

= rI (t) cos(2fc t)

in-phase component

n (t)

rQ (t) sin(2fc t)

quadrature component

( )

In-Phase and Quadrature

under CLT Approximation
!

In phase and quadrature signal components:

Without a dominant LOS, for N(t) large, we model

rI(t) and rQ(t) as jointly Gaussian (because sum of
large number of random variables -- by CLT).

We assume n(t) uniform and independent of n(t).

The received signal is thus completely characterized
by its mean, autocorrelation, and cross correlation.

Auto and Cross Correlation

!

Assume n~U[0,2]

Recall that n is the multipath arrival angle

Autocorrelation of in-phase/quad signal is

Cross-correlation of in-phase/quad signal is

P = average rx power = 1/2 sum_{n} E[^2_n].

Uniform AOAs
!

Under uniform scattering, in phase and quad comps

have no cross-correlation and auto-correlation is
Decorrelates over roughly half a wavelength

The PSD of received signal is

SrI(f)

Used to generate simulation values

-fD

fD

J_0 (not the same as I_0!)

!

The zeroth-order Bessel function of the

first kind

25

Signal Envelope Distribution

!

CLT approx. leads rI and rQ to be iid Gaussian; the

the envelope / modulo of rI +j rQ has a Rayleigh
distribution (power is exponential, phase is uniform,
and they are independent)

When LOS component present, Ricean distribution

Measurements support Nakagami distribution in

some environments
!
!

Similar to Ricean, but models worse than Rayleigh

Lends itself better to closed form BER expressions

Rayleigh Fading

pdf

cdf

27

Rice Fading

pdf

cdf

\nu = mean of G
\sigma = std of G

28

Nakagami Fading

pdf

cdf

29

Level crossing rate and

Average Fade Duration
!

LCR: rate at which the signal crosses a specific value

AFD: How long a signal stays below target R/SNR

!

Derived from LCR

R
!

t1

t2

t3

For Rayleigh fading

!
!

Depends on ratio of target to average level ()

Inversely proportional to Doppler frequency

Wideband Channel

Wideband Channels
Individual multipath components resolvable
! True when time difference between
components exceeds signal bandwidth
!

Narrowband

Wideband

Scattering Function
!

Typically characterize its statistics, since

c(,t) is different in different environments

Underlying process WSS and Gaussian, so

only characterize mean (0) and correlation
! Autocorrelation is Ac(1,2,t)=Ac(,t)
! Statistical scattering function:
!

S(,)=Ft[Ac(,t)]

Multipath Intensity Profile

Ac()
!

Defined as Ac(,t=0)= Ac()

!
!

TM

Determines average (TM ) and rms () delay spread

Approximate max delay of significant m.p.

Coherence bandwidth Bc=1/TM

!

Maximum frequency over which AC(f)=F[Ac()]>0

AC(f)=0 implies signals separated in frequency by

f will be uncorrelated after passing through channel

Ac(f)

Bc

Doppler Power Spectrum

Sc()
!

Sc()=F[Ac(=0,t)]= F[Ac(t)]

Doppler spread BD is maximum Doppler freq.

for which Sc ()>0.
Coherence time Tc=1/BD

Bd

Maximum time over which Ac(t)>0

Ac(t)=0 implies signals separated in time by t will

be uncorrelated after passing through channel

Visual Representation

36

Main Points: narrowband

!

Statistical multipath model leads to a time-varying channel

impulse response

Two multipath models: narrowband and wideband

Narrowband model and CLT lead to in-phase/quad

components that are jointly stationary Gaussian processes
! Processes completely characterized by their mean, autocorrelation, and cross-correlation.
! Zero-mean
! Auto and cross-correlation depends on AOAs of
multipath.

Main Points: narrowband

!

Uniform scattering makes autocorrelation of in-phase and

quad follow Bessel function
! Signal components decorrelate over half wavelength ca.
! If cross correlation is zero in-phase/quadrature are indep.
The power spectral density of the received signal has a bowel
shape centered at carrier frequency
! PSD useful in simulating fading channels
Narrowband fading distribution depends on environment
! Rayleigh, Ricean, and Nakagami all common

Main Points: wideband

!

Wideband fading
! Scattering function characterizes rms delay spread and
Doppler spread. Key parameters for system design.
! Delay spread defines maximum delay of significant
multipath components. Inverse is coherence bandwidth of
channel.
! Doppler spread defines maximum nonzero Doppler, its
inverse is coherence time.