Lecture 3

Fading (radio) channels: DC13a

In this lecture
• Introduction: DC13.1
• Time-variant channel filters: DC13.2
• Correlation properties: DC13.1.1, DC13.1.2
• Flat Rayleigh fading: DC13.3
a Proakis , Digital Communications

Adv. Digital Communications 3.1 c M. Skoglund, R. Thobaben

 Models for Radio Channels

• The basic course: the AWGN model,

AWGN

transmitted signal received signal

• Radio channels, • We will look at:

– attenuation – Models
– multi-path propagation – Performance
– fading – Counter-actions,
– time-variations ∗ diversity
– limited bandwidth ∗ coding

Adv. Digital Communications 3.2 c M. Skoglund, R. Thobaben

 Radio Transmission

speed v > 0

transmitter
receiver

reflections, multi-path, interference, attenuation, time-variation,. . .

Adv. Digital Communications 3.3 c M. Skoglund, R. Thobaben

 • Transmitter:
– Signal: s(t) = Re[sl (t) exp(jωc t)]
– Carrier modulation at ωc = 2πfc ,
complex baseband representation sl (t)
• Channel:
– Large number of paths
– Potential movement → time-variation
• Receiver:
– Received signal: x(t) = Re[rl (t) exp(jωc t)]
– The receiver moves relative to the transmitter
• The nth path: delay τn (t), amplitude αn (t), and phase θn (t),
– time-varying and random

Adv. Digital Communications 3.4 c M. Skoglund, R. Thobaben

 j2πfc t

• x(t) = Re rl (t)e , with
X
rl (t) = αn (t)e−jθn (t) sl (t − τn (t))
n

with θn (t) = −2πfc τn (t)

R∞
• We can write rl (t) = −∞ c(τ ; t)sl (t − τ )dτ , with
X
c(τ ; t) = αn (t)e−jθn (t) δ(τ − τn (t))
n

• Impulse response c(τ ; t0 ) at time t = t0

• Many terms ⇒ c(τ ; t) is a Gaussian stochastic process

Deterministic impulse δ(t − t0 ) received as a Gaussian random

waveform x(t) = c(t; t0 )

Adv. Digital Communications 3.5 c M. Skoglund, R. Thobaben

 Time-variant Linear Filter Channel
Time-variant channel impulse response c(τ ; t),
• input sl (t) ⇒ output
Z ∞
rl (t) = c(τ ; t)sl (t − τ )dτ
−∞

t ↔ time-variation
τ ↔ different paths, delay
• time-variant transfer function
Z ∞
C(f ; t) = c(τ ; t)e−j2πf τ dτ
−∞

Adv. Digital Communications 3.6 c M. Skoglund, R. Thobaben

 Correlation Properties (Time Domain)

• Autocorrelation function of c(τ ; t)

Rc (τ2 , τ1 ; t2 , t1 ) = E[c∗ (τ1 ; t1 )c(τ2 ; t2 )]

• Standard assumption: c(τ ; t) is wide-sense-stationary (WSS)

Rc (τ2 , τ1 ; ∆t) = E[c∗ (τ1 ; t)c(τ2 ; t + ∆t)]

• Additional assumption: uncorrelated scattering (US)

Rc (τ2 , τ1 ; ∆t) = Rc (τ1 ; ∆t)δ(τ2 − τ1 )

with Rc (τ ) = Rc (τ ; 0): multipath intensity or power delay profile

Adv. Digital Communications 3.7 c M. Skoglund, R. Thobaben

 Power Delay Profile – Delay Spread
Rc (τ )

Tm

• Tm = delay spread
• RMS delay spread,
2
R
2 τ Rc (τ )dτ
στ = R
Rc (τ )dτ
• The distribution of power over different delays (paths)

Adv. Digital Communications 3.8 c M. Skoglund, R. Thobaben

 Correlation Properties
(Frequency Domain)

• Autocorrelation function of C(f ; t)

RC (f2 , f1 ; t2 , t1 ) = E[C ∗ (f1 ; t1 )C(f2 ; t2 )]

• Standard assumption: channel is wide-sense-stationary

RC (f2 , f1 ; ∆t) = E[C ∗ (f1 ; t)C(f2 ; t + ∆t)]

• Additional assumption: uncorrelated scattering

RC (f2 , f1 ; ∆t) = RC (∆f ; ∆t)

with RC (∆f ; ∆t): frequency-time correlation function

Adv. Digital Communications 3.9 c M. Skoglund, R. Thobaben

 Group Task
Go through the following chain of equations, and verify/try to understand
it. Do not go to the next line unless you have understood the previous
ones. (3min)
∞
Z ∞
Z
∗ j2π(f1 τ1 −f2 τ2 )
RC (f2 , f1 ; ∆t) = E[c (τ1 ; t)c(τ2 ; t + ∆t)]e dτ1 dτ2
−∞ −∞
∞
Z ∞
Z
j2π(f1 τ1 −f2 τ2 )
= Rc (τ1 ; ∆t)δ(τ2 − τ1 )e dτ1 dτ2
−∞ −∞
∞
Z
= Rc (τ1 ; ∆t)ej2π(f1 −f2 )τ1 dτ1
−∞
∞
Z
= Rc (τ1 ; ∆t)ej2π∆f τ1 dτ1 = RC (∆f ; ∆t)
−∞

Now explain it to your neighbor/verify your neighbors explanations!

(3min)

Adv. Digital Communications 3.10 c M. Skoglund, R. Thobaben

 Frequency Correlation – Coherence
Bandwidth
|RC (∆f )|

∆f

Bm

• Frequency correlation:
Z ∞
RC (∆f ) , RC (∆f ; ∆t = 0) = Rc (τ )e−j2π∆f τ dτ
−∞

• Coherence bandwidth: Bm ≈ 1/Tm

• The channel is ≈ constant in f over the bandwidth Bm

Adv. Digital Communications 3.11 c M. Skoglund, R. Thobaben

 Frequency-Selective/Flat Fading

• Transmitted signal s(t) with bandwidth W ,

– W ≪ Bm ⇒ frequency-flat fading (only scaling and
phase-shift, no “filtering”)
– W ≫ Bm ⇒ frequency-selective fading (linear filtering, ISI)

Adv. Digital Communications 3.12 c M. Skoglund, R. Thobaben

 Doppler Spread and Coherence Time
SC (λ)

Bd

• Doppler spectrum:
Z ∞
SC (λ) = RC (∆f = 0; ∆t)e−j2πλ∆t d∆t
−∞

– signal intensity as a function of the Doppler frequency λ

– relate Doppler effects to variations in time
• Doppler spread: Bd
• Coherence time: Td ≈ 1/Bd

Adv. Digital Communications 3.13 c M. Skoglund, R. Thobaben

 Fast/Slow Fading

• Doppler spread: A frequency impulse (sinusoid) is broadened to

bandwidth ≈ Bd
• The channel is ≈ constant in time for Td seconds
• Transmitted signal s(t) with bandwidth W ,
– W −1 ≪ Td (W ≫ Bd ) ⇒ slow fading (no Doppler spread)
– W −1 ≫ Td (W ≪ Bd ) ⇒ fast fading (Doppler spread)

Adv. Digital Communications 3.14 c M. Skoglund, R. Thobaben

 Scattering Function
• Average power output as a function of the time delay τ and the
Doppler frequency λ
Z∞
S(τ ; λ) = Rc (τ ; ∆t)e−j2πλ∆t d∆t
−∞
Z∞ Z∞
= RC (∆f ; ∆t)e−j2πλ∆t ej2πτ ∆f d∆t d∆f
−∞ −∞

Adv. Digital Communications 3.15 c M. Skoglund, R. Thobaben

 Summary: WSS-US Channel Model

• Frequency–time correlation function: RC (∆f ; ∆t)

• Power delay (or intensity) profile: Rc (τ )
• Doppler spectrum: SC (λ)
• Scattering function: SC (τ ; λ)

Adv. Digital Communications 3.16 c M. Skoglund, R. Thobaben

 Example: Doppler Power Spectrum of
Mobile Radio Channels

αn e−jθn (t) sl (t − τn (t))

P
• Channel: rl (t) = n
with θn (t) = 2π[ (fc + fD,n (t))τn (t) − fD,n (t)t ] where
– τn (t): delay of nth path
– fD,n (t): Doppler shift of nth path
• Doppler shift: Since v > 0 (mobile user), the carrier frequency fc
is received as fc + fD,n , with

fD,n = fd cos(ϕn )

where fd = v/λw = fc v/c is the maximal shift

Adv. Digital Communications 3.17 c M. Skoglund, R. Thobaben

 • Assume isotropic scattering
– correlation properties,

RC (0; ∆t) = P · J0 (2πfd ∆t)

– Doppler spectrum,

P 1

πfd ·√ |λ| < fd
1−(λ/fd )2
SC (λ) = F[RC (0; ∆t)] =
0 |λ| ≥ fd

– Doppler spread Bd = fd , coherence time Td ≈ 1/fd

Adv. Digital Communications 3.18 c M. Skoglund, R. Thobaben

 Time Correlation and Doppler Spectrum
1 5

4.5
√ 1
J0 (2πx) 4
1−x2
3.5
0.5
3

2.5

2
0
1.5

x 0.5
x
−0.5 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

Adv. Digital Communications 3.19 c M. Skoglund, R. Thobaben

 More on Flat Fading

• Channel input (complex baseband) sl (t), bandwidth W

• Channel output rl (t)
• Flat fading, C(f ; t) ≈ C(0; t) , g(t), for |f | ≤ W ⇒
Z ∞
rl (t) = Sl (f )C(f ; t) exp(j2πf t)df
−∞
Z ∞
≈ C(0; t) Sl (f ) exp(j2πf t)df = C(0; t)sl (t) = g(t)sl (t)
−∞

• Attenuation/scaling |g(t)|, phase-shift arg{g(t)}

Adv. Digital Communications 3.20 c M. Skoglund, R. Thobaben

 • rl (t) = g(t)sl (t) with

g(t) = x(t) + jy(t) = a(t) exp(φ(t))

– x(t) and y(t) are independent and Gaussian

• No “line-of-sight” component ⇒ E[x(t)] = E[y(t)] = 0 ⇒
Rayleigh fading,
– a(t) is Rayleigh distributed
a −a2 /(2σ2 )
f (a) = 2 e
σ
and φ(t) is uniform over [0, 2π)
• “Line-of-sight” component ⇒ E[g(t)] 6= 0 ⇒ Ricean fading

Adv. Digital Communications 3.21 c M. Skoglund, R. Thobaben

 Rayleigh
0.14

0.12 0.5

0.1

0.08
2

0.06

0.04
6

0.02

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

a −a2 /2σ 2
f (a) = σ2 e , for σ 2 = 0.5, 2, 6.

Adv. Digital Communications 3.22 c M. Skoglund, R. Thobaben

 Shadowing

• We have focused on “small scale fading”

• Large scale fading ↔ signal attenuated by (large) objects in the
radio channel:
– Variation on much larger time-scales
– Common model, log-normal fading. . .
• Slowly time varying multiplicative random process

r(t) = A0 g(t)s(t)

with mean path loss A0 , transmitted signal s(t), and shadowing

process g(t).
• The mean path loss depends on a huge number of factors, e.g.

Adv. Digital Communications 3.23 c M. Skoglund, R. Thobaben

 Hata model for a large city in urban area

Loss in dB = 69.55 + 26.16 log10 fc − 13.82 log10 ht −

3.2(log10 11.75hr )2 − 4.97 + (44.9 − 6.55 log10 ht ) log10 d

with carrier frequency fc (150 < f < 1500 MHz), transmit

antenna height 30 < ht < 200 m, receive antenna height
1 < hr < 10 m, distance 1 < d < 20 m.
• The shadowing process is modelled statistically as lognormally
distributed, i.e.
1
p(g) = √ exp(−(log g − µ)2 /2σ 2 ) g ≥ 0.
2πσ 2 g

Adv. Digital Communications 3.24 c M. Skoglund, R. Thobaben