,

Dr. Fayazur Rahaman Mohammad,

,
Introduction Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.

4G (∼ 2010’s)

Gen. Technology ≈ Data Rate

4G LTE 100 − 200 Mbps
2G (∼ 1990’s) 3G (∼ 2000’s) 4G WiMax 2000 100 Kbps
4G LTE-Adv 0.5 − 1 Gbps
Gen. Technology ≈ Data Rate Gen. Technology ≈ Data Rate
2G GSM 10 Kbps / user 3G WCDMA/UMTS 384 Kbps LTE : Long Term Evolution
2G CDMA 10 Kbps 3G CDMA 2000 384 Kbps WiMax : Worldwide Interoperability for Microwave Access
2.5G GPRS ≈ 50 Kbps 3.5G HSDPA / HSUPA 5 − 30 Mbps
2.5G EDGE ≈ 200 Kbps 3.5G 1 EVDO (Rev A,B,C) 5 − 30 Mbps Evolution in Applications in 3G/4G

Gen. ≈ Data Rate Applications

GSM : Global System for Mobile WCMD : Wideband CDMA
2G / 2.5G 10 − 100 Kbps Voice + Basic data
CDMA : Code Division for Multiple Access UMTS : Universal Mobile Telecomm. Standard
3G / 3.5G 300 Kbps −30 Mbps Voice, High speed Data,
GPRS : General Packet Radio Service HSDPA : High Speed Downlink Packet Access
Video calling
EDGE : Enhanced Data for GSM Evolution EVDO : Evolution Data Optimized
4G > 100 Mbps Online Gaming, HDTV

I) Increase in data rates and realibility made possible by

MIMO: Multiple Input Multiple Output
CDMA: Code Division for Multiple Access
OFDM: Orthogonal Frequency Division Multiplexing

Received Signal

i) Received signal yp (t) = sum of the various multipath

components is
PL−1 n
=⇒ yp (t) = i=0 < ai S(t − τi )ej2πfc (t−τi )
o

L−1 −j2πfc τi e j2πfc (t) ,

nP  o
=< i=0 a i S(t − τ i )e
PL−1
where y(t) = i=0 ai S(t − τi )e−j2πfc τi
is the complex baseband received signal
Wireless Channel Model
Fading
Challenges for Wireless ii) Under Narrowband assumption i.e., fm ≤ fc , we have
x(t) y(t) S(t − τi ) ≈ S(t)
Wireless Channel
Scatterers Input Signal h(t) Output Signal P= h × S(t), where
=⇒ y(t)  channel coefficient is
L−1
h, a
i=0 i e −j2πfc τi
time, t 
Power of Fading Coefficient
LOS NLOS Impulse Response
Wireless
Interference depends on the attenuation and delay of different
multipath components
NLOS Deep Fade
R Channel i) i th path of wireless env. is characterized by
delay τi and attenuation ai
BS
Mobile
Example: h with L = 2
Multi-path
=⇒ can be modeled as ai δ(t − τi )
ii) The impulse response of a system with multipath i) Then h = a0e−j2πfc τ0 + a1e−j2πfc τ1
i) Presense of scatterers lead to multipath propagation
Let a0 = a1 = 1 and τ0 = 0, τ1 = 2f1 , then
PL−1
components is h(t) = i=0 ai δ(t − τi )  ii)
→ Leads to superposition of multiple signals c
h = 1e0 + 1e−jπ = 0 (Destructive interference)
Further leads to constructive or destructive interference iii) Transmitted
n Signal o
→
y(t) = h × s(t) = 0
ii) Constructive interf. amplifies received signal amplitude
Sp (t) = < S(t)ej2πfc t , where Sp (t) is Passband signal, iii) Let a0 = a1 = 1 and τ0 = 0, τ1 = f1 , then
→ Destructive interf. attenuates received signal amplitude c
S(t) is the complex Baseband signal and fc is carrier freq. h = 1e0 + 1e−j2π = 2 (Constructive interference)
iii) Develop a model for multipath propagation
Where for GSM fc ≈ 900 MHz, for 3G fc ≈ 2.1 GHz and y(t) = h × s(t) = 2s(t) ( Enhanced signal amplitude)
for 4G fc ≈ 2.5 GHz iv) The power of channel coefficient h varies with time
Fading
Fading Channel Coefficient (Exponential Model)
x(t) y(t)
Wireless Channel
jφ
i)Input
WeSignal
have h = x + j y = ae , where
p h(t) Output Signaly
Amplitude a = x 2 + y 2 and phase φ = tan−1 x
time, t
ii) Then joint distribution FA,Φ(a, φ) = π1 e
Impulse Response −(x 2+y 2)
|JXY | where
Power of Fading Coefficient

 ∂x ∂y   cos φ sin φ  Deep Fade

Det. of Jacobian |JXY | =  ∂a ∂a  =   = a.
 
 ∂φ ∂φ  −a sin φ a cos φ
∂x ∂y
v) This Fading process causes the received power to vary
FA,Φ(a, φ) = πa e −a 2
∴  and is a key aspect of wireless channels
Fading Channel Coefficient (Real and Img. Model)
iii) Marginal distribution of amplitude A is Fading Example:
i) We have real and imaginary parts h = x + j y, where
y(t) FA(a) = −π FA,Φ(a, φ)dφ
Rπ
PL−1 PL−1
x(t)
x = i=0 ai cos(2πfc τi ) and y = − i=0 ai sin(2πf Wireless Channel
c τh(t)
i) a 2Rπ 2 i) To compute the probability that the channel
Input Signal Output Signal = πe −a
−π dφ = 2ae
−a for 0 ≤ a ≤ ∞ 
ii) Note ai , τi are random variables and attenuation is worse than 20 dB
is a Rayleigh distribution, and
Power of asCoefficient
Fading result h is also termed as
time, t
x, y are sum of large number of random variables ii) We have channel gain = a 2
Impulse Response
rayleigh fading channel. Deep Fade
=⇒ by CLT x, y are assumed to be Gaussian distributed
FA (a)
=⇒ 10 log10 a 2 < −20 dB,
iii) Assuming x, y to be i.i.d ∼ N (0, 12 ), we have FA (a) Rayleigh
FΦ (φ) =⇒ a < 0.1, then
1
2
FX ,Y (x, y) = FX (x)FY (y) Uniform R 0.1
2π
Pr(a < 0.1) = 0 2ae da −a
1 2 1 −y 2 2 0.1
= √π e −x √ e
π = −e |0 = 1 − e−0.01 ≈ 0.01
−a
1 2+y 2)
= √π e −(x  a≈0 a −π π φ
Deep fade
a≈0 a a
deep fade region
iv) Marginal distribution of phase Φ is
FΦ(φ) = −∞ FA,Φ(a, φ)da
R∞
1 2 1 for −π < φ ≤ π
= 2π 0 2ae da = 2π
R∞ −a 
is a Uniform distribution.

v) Note that FA,Φ(a, φ) = FA(a)FΦ(φ),

=⇒ A and Φ are independent RVs.

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
, FA (a) Φ

Dr. Fayazur Rahaman Mohammad,

Rayleigh 1
Uniform

Performance Analysis (BER)

2π

Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.

,
a −π π φ
a≈0
Deep fade
WirelinedeepCommunication
a≈0
fade region
system a(AWGN)
a

Numerical Example: Exact BER

i) Consider AWGN wireline system
noise at n
i) To compute the BER at SNR = 10dB
→ 10 log10 SNR
√ = 10dB, =⇒ SNR = 10
receiver
 √

x y → Pe = Q SNR = Q 10 ≈ 7.82 × 10−4
Transmitted Received
symbol symbol
Approximate BER
AWGN Wireline system
2
we have, y = x + n 1e 2 ,
−x
2
i)Using the approximation Q (x) ≤ 2
1 n
− 2 √
1 SNR
where n is Gaussian noise with FN (n) = e 2σ

√ → Pe = Q SNR ≤ 2 e 2
−
√ 2πσ 2 √
ii) Consider x = − P, then at receiver y = x + n = − P + n
√ Numerical Example: Approx SNR
=⇒ error occurs if y ≥ 0, =⇒ n ≥ P
√
=⇒ Pe = Pr(n > P) i) To compute the SNR required for Pe = 10−6
2
1 n 1 −t2 1 e − SNR
Z ∞ Z ∞
− 2 2 = 10−6, =⇒ SNR = 26.24
= √ √ e 2σ dn = q √ e 2 dt → 2
P 2πσ 2 P
2
2π → SNRdB = 10 log10(26.24) = 14.19dB
σ √
=⇒ Pe = Pr N (0, 1) ≥ P2 = Q P =Q
 q  q  
2 SNR
σ σ
Numerical example: Exact BER
Single antenna wireless communication system
i) To compute the BER at SNR = 20 dB
i) Considering wireless system → 10 log10 SNR = 20dB, =⇒! SNR = 100
1 100
r
Wireless System
→ BER = 1− ≈ 4.92 × 10−3
Fading h noise at n 2 2 + 100
coefficient receiver

x y BER Approximation
Transmitted Received
symbol symbol i) BER on SNR using approximation.,
!
1
BER = 12 1 − 1 1 1 − (1 + 2 )− 2
r
=
 
we have, y = h x + n 2 2 SNR
1+ SNR
where h is fading coefficient and n is Gaussian noise
≈ 12 1 − (1 − 12 SNR2 ) = 1
 
ii) Then, received power = |h|2P = a 2P 2 SNR
a 2P
=⇒ Fading SNR, SNRF = 2 = a 2 SNR i) Note that while the BER is decreasing exponentially
σ 1 SNR
1
(i.e., 2 e 2 ) in a wireline system,
iii) From BERderived for BPSKmodulation, −
√ 1
BER = Q a 2SNR it is decreasing only by 2 SNR in case of wireless system.

SNRF = Q
p

→ a depends on the fading coefficient h which is a RV,

Wireless System Numerical example: Exact and Approx. SNR
→ =⇒ BER in wireless is a random quantity Deep fade
i) To compute the SNR required for BER 10
Fading h noise at n −6
iv) We coefficient
then have average BER in a Rayleigh fading channel as
receiver
Tx ! Rx =
√
Bit-Error Rate, BER 2SNR F (a) da SNR
 r
x BERavg = Q a
R∞
0 A y 1
→ 2 1− Single link
= 10−6, =⇒ SNR = 4.99 × 105
√
2 2 2 + SNR
i) BER is a metric to characterize the channel performance
 
= 0 Q a SNR 2aesymbolda
Transmitted R∞ −a
Received
symbol
→ SNRdB = 10 log10(4.99 ×Rx105) = 56.98 dB
ii) Bits of information undergo errors when transmitted 1 SNR
r !
Tx
ii) Using approx formulae, we have SNR = 2 BER 1 ,
iii) BER is the average number of bits that are in error = 1−
2 2 + SNR
further we get
linksSNR = 57dB for BER = 10−6
iv) As bits are RV, BER is frequently expressed as
Multiple
T1 Link-1 R1
iii) =⇒ a wireless channel requires an extra SNR of 43 dB
Probability of Bit error Pe , where 0 ≤ Pe ≤ 0.5
Tx Link-2 Rx (i.e., 104.3 times more) to acheive the same BER when
i) Bits are modulated prior to transmission over channel R T1compared to AWGN
Multiple receiver antenna wireless
2
communication system Link-1
h1 R1
ii) Binary Phase shift keying (BPSK) is one such scheme,
√ √ Rx BER for the given SNR and L = 2
i) Considering wireless system with antenna diversity and Tx Numerical hexample: Exact
y1
where two phases are used 0 → x = P and 1 → x = − P x 2
Link-2
y2
=⇒ Avg. power of BPSK is P optimal MRC combining i) The
Transmitted average BER can
R2
be obtained as,
Received
symbol symbols 
iii) At the receiver, if h1
2
(1 − λ) P1 2+l −1

1+λ
l
BER = ,
h2 y1
received symbol y ≥ 0, map → 0 and Tx x y2 Rx 2 l=0 l 2
2
received symbol y < 0, map → 1 1−λ
hL yL 
= {2 + λ}
2
Then the received SNR after MRC is given by, ii) To compute BER when SNRdB = 20dB, we have
P 2 2 2 P 2 P SNRdB 100
r
SNRm = 2 {|h1| + |h2| + · · · + |hL | } = 2 k h k = 2 g , SNR = 10 10 = 100 and λ = = 0.9901
σ σ σ 2 + 100
ii) Assuming hi , ∀i are IID Rayleigh fading channel coefficients BER = 7.2564 × 10−5
with E{|hi |2} = 1, ∀i, we have
→ g is also a RV with Chi-Squared distribution χ22L with Numerical example: Approx BER for the given SNR and L = 2
1
2L degrees of freedom and FG (g) = g L−1e−g i) The approx average BER for L = 2 can be obtained as,
√ (L − 1)! 2
√
iii) We now have BER = Q( SNRm ) = Q( g SNR) 3 1 3 1
 
BER = =
iv) We then have average BER in a Rayleigh fading channel as 2 2SNR 4 SNR2
BERavg = 0 Q
R∞ √
gSNR FG (g) dg

ii) To compute approx BER when SNRdB = 20dB, we have
SNRdB
√ 1 SNR = 10 10 = 100
= 0 Q gSNR g L−1e−g dg
R∞

(L − 1)! 3 1
L
1 − λ PL−1 L + l − 1 1+λ
l BER = = 7.5 × 10 −5
4 1002
  
= l=0 , where
q 2 l 2
SNR n
n! Numerical example: Approx SNR for the given BER and L = 2
λ = 2+SNR and k = k!(n−k)!
i) Given BER = 10−6, the SNR can be obtained as,
v) At high SNR using approx. we have 3 1
r
1−λ 1 1+λ SNR = , =⇒ SNR = 866.0254 = 29.37dB,
≈ and ≈ 1; 4 10−6
2 2SNR
L 2 
1 PL−1 L + l − 1 Conclusions

=⇒ BER ≈ l=0
2SNR l

2L − 1

1
L i) Huge amount of additional power is required to obtained the
≈
L 2SNR desired BER in a single antenna wireless system when
compared to AWGN
ii) As the antennas are increased to L = 2, the required transmit
BER of different communication systems
power P is decreased by 27.63 dB
i) SNR required to achieve BER= 10−6 for same noise power iii) Increase in antennas leads to significant decrease in BER.
Channel Type SNR Required This is the advantage of Diversity.
AWGN 14.19 dB
Rayleigh Fading with L = 1 57 dB Intuition behind the decrease in BER
L
Rayleigh Fading with L = 2 29.37 dB i) For L rx antennas , we have BER ≈ 2L−1 1


L 2SNR
Saving in SNR due to diversity (57 − 29.37) = 27.63 dB 1 1
for L = 1, BER ≈ 2SNR ∝ SNR
for L = 2, BER ≈ 34 1 2 ∝ 1 2
=⇒ SNR SNR
for L = 3, BER ≈ 54 1 3 ∝ 1 3
SNR SNR
in general for L rx antennas, BER ∝ 1 L
SNR
ii) Deep fade analysis will help understand this behaviour

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
,

Dr. Fayazur Rahaman Mohammad,

,
Performance Analysis (Deep Fade) Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.

Deep Fade Analysis

i) We have received power = a 2P and noise power = σ 2

→ System is in deep fade, if the received signal power is
Single antenna wireless system lower than noise power
2 1 , =⇒ a < 1
=⇒ a 2P < σ 2, =⇒ a 2 < σP = SNR
q
i) Recall, for a single antenna system , we have, y = h x + n  SNR

1
q
Wireless System ii) Probability of deep fade, PDF = Pr a < SNR
Fading h noise at n
1
q
coefficient receiver
SNR
=⇒ PDF = 0 FA(a) da
R
x y 1
q
R SNR 2 1
Transmitted
symbol Wireless System
Received
symbol
= 0  2ae da = 1 − e
−a − SNR

Deep fade≈ 1 − 1 −
1 at high SNR = 1
Fading h noise at n SNR SNR
ii) Andcoefficient
received power = 2P = a 2P
receiver
|h| Rx
Tx
1
x a 2P iii) Note, that
Single Bit
link Error Rate (≈ 2SNR ) ∝ Deep Fade
=⇒ Fading SNR, SNRF = 2 = a 2 ySNR
Transmitted σ Received → The destructive interference in wireless channel results
symbol symbol
in the deep fades Rx
Tx
→ The performance degrade due to deep fades can be overcome
by theMultiple links of diversity
principles
T1 Link-1 R1

Tx Rx Deep Fade Analysis with diversity

Link-2
R2 i) TWe1 have R1 power = g P and noise power = σ 2
h1received
Link-1

Multiple receive antenna Wireless system → System is in deepy1fade, if the received signal power is
Tx x h2 Rx
lower than noise ypower
Link-2
2
i) Recall that for a system with multi rx antenna R22 σ2 = 1
Transmitted
=⇒
symbol gP < σ , =⇒ g<
Received
symbols P SNR
h1
ii) Probability of deep fade, PDF = Pr g < SNR 1
 
h2 y1
Tx x y2 Rx 1
R SNR
Deep Fade
yL
=⇒ PDF = 0 FG (g) dg
hL

Wireless System 1 1
g L−1e−g dg
R SNR
i) The power of each channel coefficient hi , ∀i varies with time = 0
we have the received SNR after MRC as, Fading h noise at n (L − 1)! Deep fade

P P P
coefficient
At high SNR, we have SNR
receiver 1 ≈ 0, Tx Rx
Fading 2 2 2 2
SNRm = 2 {|h1| + |h2| + · · · + |hL | } = 2 k h xk = 2 g ,
σ σ σ and note that consideredy 1 , =⇒ e −g ≈Single
limit is 0 ≤ g ≤ SNR 1 link
ii) Assuming hi , ∀i are IID Rayleigh fading channel coefficients
Transmitted 1
RReceived 1 1 g L 1
2
symbol
=⇒ PDF = symbol SNR
g L−1 dg = SNR
with E{|hi | } = 1, ∀i, we have 0 (L − 1)!
|
1)! L 0
(L − Tx Rx
→ g is also a RV with Chi-Squared distribution χ22L with 1 1 1
= ∝
time, t 1 L! SNR L SNRL
g L−1e−g
Power of Fading Coefficient
2L degrees of freedom and FG (g) = Multiple links
(L − 1)! T1 Link-1 R1
Deep Fade
iii) With increasing number of antennas L, probability of
Tx Deep
Link-2 fade decreases
Rx and this leads to
ii) This Fading process causes the received power to vary
significantR2 decrease in BER T1
and is a key aspect of wireless channels
Link-1
h1 R1

iii) The poor performance of a wireless comm. system can be Inference Tx x h2

Link-2
y1
Rx
explained based on the Deep Fade events
y2
R2
i)
This form of diversity is called spatial diversity Transmitted
symbol
Received
symbols
h1
ii)
One keyh assumption
y1
2 is that all the coefficients hi , ∀i
are Independently Rxfading
Tx x y2
h yL
L

→ For this to hold true, antennas have to be placed sufficiently

far apart

Intuitive explanation d
Wireless System
Deep fade
Rx
Fading h noise at n
coefficient receiver Rx
Tx
x y Single link
Transmitted
symbol
Received iii) Let d denote the spacing between the uniformly
symbol
Rx spaced antennas
Tx
→ For independently fading channels, the condition is given by
Multiple links dmin = λ2 , where dmin is the min separation between antennas
and λ(= C f ) is the wavelength of the carrier
i) In a system with single tx and single rx antenna, c

we have a single link, and the

1 Numerical example (GSM)
probability of Deep Fade ∝
SNR Q) To compute the dmin for a given fc = 900MHz (GSM)
ii) In a system with single tx antenna and L rx antennas
3 × 108m/s
we have L links between the txer and rxer Sol) We have, λ = 6
= 33.33cm
900 × 10 /s
→ Let Ei denote the event that link i is in deep fade ∴ dmin = 16.66cm, i.e., distance for independent fading
1
we know Pr(Ei ) = , ∀i → Note, that this distance is not possible in a regular phone,
SNR
→ For a system to be in deep fade, all links however, possible in large devices like routers etc.
have to be in deep fade, i.e.,
Numerical example (3G)
Pr(Deep fade) = Pr(E1 ∩ E2 ∩ · · · ∩ EL )
= Pr(E1) × Pr(E2) × · · · × Pr(EL ), Q) To compute the dmin for a given fc = 2.3GHz (3G)
for IID channel coefficient hi , ∀i 3 × 108m/s
1 Sol) We have, λ = 9
= 13.04cm
= 2.3 × 10 /s
SNRL ∴ dmin = 6.5cm, i.e., distance for independent fading
→ Note that this distance is within dimensions of a
smart phone (3G / 4G)

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
,

Dr. Fayazur Rahaman Mohammad,

,
Diversity Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.

Wireless System
Deep fade
Fading h noise at n
coefficient receiver Rx
Tx
x y Single link
Transmitted Received
symbol symbol
Tx Rx

Multiple links Receiver

T1 Link-1 R1
i) Processing at the receiver using weighted
" # combination
Tx Rx
Multiple Antenna System i y
1
= wTy, where
Link-2
y = w1 y1 + w2 y2 = w1 w2
h
R2 y2
e
T1 Link-1
h1 R1

y1 w1, w2 are combining weights,

y = wT (h x + n) = wT h x + wT n
Tx x h2 Rx
Link-2
y2 =⇒ e
Transmitted
R2
Received P |wT h|2
symbol symbols ii) SNR at the receiver SNR = T 2
where
|w n|
y1 = h1 x + n1 → expected value of noise power:
i) System model:
y2 = h2 x + n2, where n1, n2 are E{|wT n|2} = E{(w1n1 + w2n2)2}
Principle of Diversity independent zero-mean Gaussian noise, each with power σ 2. = w12E{n12} + w22E{n22} + 2w1w2E{n1n2}
i.e., E{|n1|2} = E{|n2|2} = σ 2 and E{n1n2} = 0 = σ 2(w12 + w22) = σ 2 k w k2
i) Diversity is used to combat the effects of Deep Fades
ii) Vector # (assuming
model " # h1, h2 as real) → signal power at receiver:
ii) Single link system: Communication fails if the single link is in
y1 h1 n1 wT h
" # "
=w·h
deep fade = x+ ; y=hx +n
y2 h2 n2 |w · h|2 =k w k2 k h k2 cos2 θ
P k w k2 k h k2 cos2 θ
Deep fade
P 2 cos2 θ
Tx Rx → ∴ SNR = = k h k
σ 2 k w k2 σ2
y Single link

Received
symbol Example of a MRC

Given, h1 = √1 + √1 j and h2 = √1 − √1 j and

Rx
Tx
i)
2 2 2 2
Multiple links noise power σ 2 = 12 = −3dB 
1 1
√ +√ j
y1 n1
" # " #
iii) Multiple link system: Communication is not disrupted even Sol) We have = 1 2 2 x+
y2 √ −√ j1 n2
 
if two links are deep fade because of many alternative paths 2 2
iv) Ex: Antenna diversity 1 + √1 j
 
1  2√
2 
→ Multiple receive antennas introduces receive diversity Optimal Receiver (MRC) → then, optimal MRC vector w = √
2 √1 − √1 j
2 2
→ Other forms: Transmit and receive diversity, time diversity, i) SNR is maximum when θ = 0,
→ Futher we obtain, |wH h|2 = 2 and k w k2= 1,
freq. diversity, user diversity etc. i.e., w is in the same direction of h (or w ∝ h) P |wT h|2 2P
P(h12 + h22) 1 h1 → ∴ SNR = = 1 = 4P
" #
T
|w n| 2
=⇒ SNRmax = 2
, when w = q 2
σ h 2 + h 2 h2 1 2
Generalizing the MRC for L receiver antennas
→ When channel coefficients
" #are complex, we use
i y 1 h1
" #
y1 h1 n1
     
H 1
and w = p
h
y = w y = w1 w2
∗ ∗
y2 |h1|2 + |h2|2 h2 2 2
E{|ni | } = σ , ∀i  y2   h2   n2 
e      
i) Given ,  =  x +  
Diversity → This receiver is called Maximal Ratio Combiner (MRC) E{ni nj } = 0, ∀i 6= j  .. 
  .. 
  
 .. 
 
yL hL nL
Analytical Example with L = 1
and e
y = w1∗y1 + w2∗y2 + · · · + wL∗ yL
Q) Compute the diversity order of a wireless system with =⇒ e y = wH (h x + n) = wH h x + wH n
L = 1 antenna h
ii) For maximum SNR, we choose MRC with w =
1 k h
k
Sol) In this case, we have Pe (SNR) ≈ under high SNR k h k2 P |h1|2 + |h2|2 + · · · + |hL |2 P
2SNR ∴ SNR =
1 σ 2
=
σ 2

log
2SNR log 2
d = − lim = lim 1 + =1
SNR→∞ log SNR SNR→∞ log SNR

Analytical Example with multiple antennas

Q) Compute the diversity order of a wireless system with

L antennas
Diversity Order, d
Sol) In this case, we have
i) Is a measure to characterize the diversity of a system
L
log 2L−1 1


L 2SNR
log Pe (SNR) d = − lim SNR ,
d = − lim , SNR→∞ log
SNR→∞ log SNR −L log SNR+log 2L−1
1
L 2L
where Pe (SNR) denotes BER as a function of SNR = − lim log SNR
SNR→∞
2L−1
1
log L L
= lim L − log SNR2 = L
SNR→∞

Analytical Example for wireline system

Q) Compute the diversity order of a wireline system

Sol) In this case, we have
1 − 12 SNR
log 2 e log 12 − 12 SNR
d =− lim log SNR = − lim SNR ,
SNR→∞ SNR→∞ log
1 SNR
= lim SNR
2= ∞, using L‘Hospital’s rule
SNR→∞ log
=⇒ A AWGN channel is a combination of ∞ independently
fading links

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
,

Dr. Fayazur Rahaman Mohammad,

,
Wireless Channel Characterization Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.

RMS Delay Spread

g0
g1
g2 g0
Example:
g3 g1
Noise floor
g2 g3
τ0 τ1 τ2 τ3 τ0 Spurious components
Power Profile Q) Compute RMS delay, given L = 4, τi = {0, 1, 3, 5}
Noise floor
µs ∀i,
τ3 τ0
i) Practically, there might be large number of spurious paths with
Spread over time τ0 τ1 τ2
gi = {0.01, 0.1, 1, 0.1} = {−20, −10, 0, −10}dB ∀i,
Scatterers negligible power below the noise floor Sol) Given,
Spread over time

In such case, max. delay spread is not an appropriate metric Multipath power profile
LOS NLOS Interference
ii) RMS Delay spread (root mean square) is considered in 0 dB g2
g1 g3
NLOS
such scenario. We now consider
−10 dB
+ dx g0
R gi −20 dB
BS
Mobile bi = PL−1 , the fraction of power in i th path g0
j=0 gj
Multi-path g1 0 1 2 3 4 5 τ (µs)
g2
Average delay by weighing each delay by fraction of power 0.01 × 0 + 0.1 × 1 + 1 × 3 + 0.1 ×
g3
5 floor
Noise

i) Channel response is
PL−1
gi τi → ττ0= τ1 = 2.9752µs
PL−1
PL−1
τ = i=0 bi τi = PL−1i=0  0.01 s
τ2
+ 0.1 + 1 + 0.1
τ3 τ0 Spurious compo

h(t) = i=0 ai δ(t − τi ), where ai is attenuation and i=0 gi

Spread over time 0.01 × (0 − 2.98)2 + · · ·
→ RMS Delay στ = = 0.8573µs
τi is the delay of i th multipath Root Mean square delay spread Multipath power profile 0.01 + 0.1 + 1 + 0.1
ii) Then Power profile of the i th multipath is στ = b0(τ0 − τ )2 + · · · + bL−1(τL−1 − τ )2 τmax = 5 − 0 g= 2 5 µs
q
→
0 dB Transmitted Sig
PL−1
φ(t) = i=0 |ai |2 δ(t − τi )
1 g g3
u PL−1 u PL−1 −10 dB S0 S3 S
v v
u i=0 gi (τi − τ ) 2 u i=0 |ai |2(τi − τ )2
PL−1 2 is gain of i th path −20 Inter Symbol
g0
Interference (ISI)
= i=0 gi δ(t − τi ), where
dB
g
g0 i = |a i | =t PL−1 =t PL−1 
2
i=0 gi
S1 S2
i=0 |ai |
g0 g1
Example: Consider an L = 4 multipath g2 channel
g1 i) The
0
two1 delayed
2 signals
3 4 5arriving
τ (µs)at the receiver add-up,
g3 T
g2
g0 multipath, i gain, ggi3 delay, τi Noise floor
Noise floor
g0 Signal delayed by τ0
τ3 τ0 g1
τ0 0 τ1
τ0 τ2
τg20 τ3
g1 τ1 Spurious components
S0 S3 S4 S6
τ0 τ0
Impact of channel power profile g2
Spurious components
g2 g3 g3
1Spread g1
Spread over time
over time τ1 S1
Noise floor
τ2 S 2 τ3 S5 τ0
2 powerg2 i) Consider a transmitted signal with symbols {S0, S1, S2, · · · } τ0 τ1
τ3 Spurious com
τ0 τ1 τ2 Multipath profile τ2
0 dB Multipath power profile
τ0
3 g g
Spread over time g2 Transmitted Signal Spread
S0 over time S3 S4 S6
g2 3
0 dB g
τ3 Transmitted Signal
1 3
−10 dB g1 g3 S0 S3 S4 S6
−10 dB g0 S0 S3 S4 S6 Multipath power profile
S1 S2
i) The spread of delay −20
overdBwhich0
different signal components are
−20 dB g
0 dB g2
S5
Transmitted S
S1 S2
arriving is called Delay spread 0 τ1
S5 g1 g3
1 S S S5 −10 dB Signal delayed by τ1 S0 S3
2 3 4 5 τ (µs) 1 2
g0
ii) Maximum Delay spread στ max
0 1 2
characterizes the delay spread,
3 4 5 τ (µs) T t
t −20 dB τ 1 − τ0
T S1
στmax = τL−1 − τ0, where
S2
Signal delayed by τ0 where T is the symbol time Because
0 1of the2 difference
3 4 5in delays
τ (µs) the adjacent symbols interfere
S0 S3 S4
τ0 is the delay of the first arriving component and ii) Futher consider 2-path wireless channel (i.e., L = 2) where
S6
ii) Inter symbol inteference (ISI) occurs
Signal delayed by τ0
when (τ 1 − τT)>T
0
τL−1 is the delay of the last arriving component S1 S2 S5 h(t) = a0δ(t − τ0) + a1δ(t − τ1) i.e, delay spread
S0 is greater than
S3 symbol
S4 time S6 or στ > T
iii) Ex: For L = 4, let τ0 = 0µs and ττ03 = 5µs, then στmax = 5µs = δ(t − τ0) + δ(t − τ1), where a0 = a1 = 1 → ISI is undesirable since it leads to distortion of the original signal
S0 S3 S4 S6 S1
→ To avoid ISI, we need the symbol time ST5 to be larger than
S2

S1 S2 S5 delay τspread
0
i.e.,
S 0 T > στ S3 S4 S6

τ1 Signal delayed by τ1
iii) Typical outdoor delay spread in a wireless channels is 2µs
S1 S2 S5
g0 τ 1 − τ0 Signal Bandwidth Bs < Channel Coherence Bandwidth Bc (difference in LOS and NLOS path in a cell is 1KM , )
g1 τ1 Signal delayed by τ1
g2
Signal Bandwidth
τ1 − τ0 Bs > Channel Coherence Bandwidth Bc
g3 X(f ) X(f )
Noise floor
τ0 τ1 τ2 τ3 τ0 Spurious components
Bs Bs X(f ) X(f )
Spread over time
H(f ) Flat Fading H(f )
Freq. Selective Bs Bs
Multipath power profile Bc Bc
0 dB g2 Transmitted Signal H(f ) Flat Fading
Freq. Selective
g1 g3 Bc Bc
−10 dB Y (f ) S=0 H(f )X(f ) S3 NoSDistortion
4 S6 Y (f ) = H(f )X(f )
g0 Distortion
−20 dB
S1 S2 S5 Y (f ) = H(f )X(f ) No Distortion Y(
0 1 Distortion
2 3 4 5 τ (µs)
i) Consider that the bandwidth
t of the signal Bs , is less than
Channel T X(f )
the coherence bandwidth of the channel Bc . I.e., Bs < Bc
Signal delayed by τ0
S0 S3 S4 S6 Bs i) Consider that the bandwidth of the signal Bs , is greater than
→ In this scenario, there is no distortion in the the coherence bandwidth of the channel Bc . I.e., Bs > Bc 
X(
S1 S2 S5 received signal Y (f ) H(f ) → In this scenario, there is distortion in the
Bs
Frequency Domain Analysis → Note that the flat portion
Bc of channel freq. response H (f ) is
τ0
S0 S3 S4 S6 received signal Y (f )
i) We have the response of the channel in time domain as called the Coherence bandwidth
Y (f ) = H(f )X(f ) ii) This scenario is termed as a Frequency selective fading Bc
S1 S
y(t) = h(t) ∗ x(t), where x(t)2 is transmitted Ssignal,
5 ii) This scenario is known as a Flat fading channel, as the signal
y(t) is received
τ1
signal and Signal
h(t) isdelayed
the impulse
by τ1 response BW is within the flat portion of the channel freq. response Typical Component Y(

ii) Then in the frequency

τ 1 − τ0 domain, we have response as i) Focusing on a typical channel component e−j2πf τi ,
Y (f ) = H (f ) X (f ), X(f ) X(f ) At f = 0 the phase = 0 and at f = 2τ1 the phase = π
i
where X (f ), Y (f ) and H (f ) are the Fourier transforms → Note that the phase (freq. response) has changed significantly
within f = 2τ1 , i.e., within a bandwidth = 2 × 2τ1 = τ1
Bs Bs

Flat Fading i i i
∴In general, approximately the BW of the channel = σ1τ = Bc
Freq. Selective H(f ) H(f )
Bc Channel
Bc frequency response
ii) =⇒ The larger the delay spread, the smaller
Y (f ) = H(f )X(f ) No Distortion i) ToYobtain the)X(f
(f ) = H(f freq.) response of the channel, we have is the coherence BW
Distortion PL−1
h(t) = i=0 ai δ(t − τi ), and iii) Typically στ = 2µs, =⇒ Bc = 500KHz
δ(t − τi ) ↔ e−j2πf τi
PL−1 PL−1 −j2πf τ
X(f ) i=0 ai δ(t − τi ) ↔
Interpretation from freq. and time domain analysis
i=0 ai e
=⇒ i = H (f )

Doppler Effect
Bs
Example Freq. domain Time domain Implication
BS If Bs > Bc =⇒ στ > T Freq. Selective Fading and ISI
Q) Given
H(f f)c = 1.85GHz, V = 60mph and θ = 30◦, compute the
If Bs < Bc =⇒ στ < T Freq. Flat Fading and no ISI
Bc doppler shift and received freq.

Sol) YWe 5
(f )have
= H(fV)X(f
= 60
) × 1.6 × 16 = 26.8m/s
θ V
Mobile Mobile velocity
fd = 26.8×cos 30◦
× 1.85 × 109 = 143Hz and
is moving 8 3×10
BS
fr = 1.85 × 109 + 143Hz
i) In a scenario where the mobile is moving, will causes the freq.
of the received signal to change Impact of Dopper on Wireless channel
→ This is known as the Doppler effect Pi=L−1 −j2πf τ
i) Recall fading coefficient is h = i=0 ai e c i
→ It is the change in freq. of the received signal due to motion ii) When the mobile is moving, the delay τi changes with time
between Txer and the Rxer V cos θt
=⇒ τi (t) = τi ± and
ii) Assuming the mobile is moving at an angle θ with the BS, C 
Pi=L−1 −j2πfc τi ± V cos

θt
V cos θ h = i=0 ai e C
we have the doppler freq. fd = × fc , where  Example:
C Pi=L−1 −j2πf τ ∓j2πf t
V is velocity of mobile, C is speed of light and fC is carrier freq. = i=0 ai e c i e d = h(t) 
Note, the channel coefficient is time-varying and hence known i)Considering fd = 143Hz,
iii) Frequency of rxed signal fr = 1 + V cos θ f .

C  1 = 1.7ms
the coherence time Tc = 4×143
C
as time selective channnel
If mobile is → The channel is approx. constant for (or changing after) 1.7ms
iii) Mobility → Dopper → Time-varying phase → Time selective
→ moving towards BS (i.e., 0 < θ ≤ π2 ), then fr ≥ fc
iv) Considering the phase factor
→ moving away from BS (i.e., π2 < θ ≤ π), then fr ≤ fc
at t = 0 the phase = 0, and at t = 4f1 the phase = π2
d
has changed significatly
→ Coherence time Tc (= 4f1 ) is defined as the time
d
over which channel is approx. constant 

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
,

Dr. Fayazur Rahaman Mohammad,

,
Code Division Multiple Access Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.
Code Division Multiple Access Chip Time and Bandwidth expansion in CDMA
i) CDMA is a key technology used for multiple access in various Example: 2-user multiple access system
i) Considering an example with symbol rate of 1 kbps, i.e.,
3G standards like WCDMA, HSDPA / HSUPA, CDMA 2000 and i) Consider codes C0 and C1, given by 1000 sym/sec, we have time per symbol Ts = 10001 sec.
1×EVDO C0 = 1 1 1 1 and C1 = 1 −1 −1 i.e., Bandwidth required BW = T1 = 1 KHz.
h i h i
1
s
Common Radio Channel
and transmitted symbol of user-0 and user-1 as a0 and a1 ii) However in CDMA (with 4 chips/sym),
ii) Before transmission each symbol chip time Tc = TNs = 1 ms
4 = 0.25 ms
i is multiplied with resp. code as
a0 , =⇒ new bandwidth = T1 = 4 KHz.
h
User 1 a0 C0 = a0 a0 a0 c
Original bandwidth has spread in CDMA.
h i
User 0
a1 C1 = a1 −a1 −a1 a1
BS
and the combinedhsum is transmitted
ii) Unlike wireline where each user is allocated a dedicated channel, i 4 KHz
CDMA signal

in multiple access, users share a common radio channel a0 C0 + a1 C1 = a0 + a1 a0 − a1 a0 − a1 a0 + a1

iii) TDMA and FDMA are other examples of multiple access As a result 4 chips are transmitted in a symbol duration. User 0 User 1 User 2

iii) To extract the individual symbols at the receiver, each user BW 0

BW 2
technologies
BW 1 f 1 KHz f
Original Signal
→ TDMA: In time division multiple access, different users are correlates the received symbol with its code i.e.,
allocated different time slots user-0: 1(a0 + a1) + 1(a0 − a1) + 1(a0 − a1) + 1(a0 + a1) = 4a0 → Spectrum of the original signal is spread by a factor of N ,
→ FDMA: In frequency division multiple access, different users are user-1: 1(a0 + a1) − 1(a0 − a1) − 1(a0 − a1) + 1(a0 + a1) = 4a1 where N is the length of the code. As a result, it is also called
allocated different bands. FDMA was used in 1G comm. systems There by multiple users are able to simultaneously receive signals spread spectrum system
iv) Similarly, in CDMA different users are allocated different codes on a common channel → CDMA codes are also known as the spreading codes. And N is
→ CDMA uses the concept of codes for multiple access termed as spreading factor.

Properties of the CDMA PN sequence

i) Balance Property:
Code generation of CDMA Example: LFSR for N = 4 → The number of (−1)’s (=8) u the number of (1)’s (=7)
i) The code is typically a Pseudo Noise (PN) sequence (random). ii)
StatesRun
→ Xi−1length
Xi−2 Property:
Xi−3 Xi−4 Xi (= Xi−4 ⊕ Xi−3 )
State Variables → Xi−1 Xi−2 Xi−3 Xi−4 Xi (= Xi−4 ⊕ Xi−3 ) Runs ↓ 1
A PN seq. is generated by a linear feedback shift register (LFSR)
1 1 1 1 0
States ↓ 1 1 1 1 1 0 → A2 Run0 is defined
1 1 as 1the string 0of continuous values

ii) A LFSR with D states can go through 2D − 1 states 1

3 0 0 1 1 0
States → Xi−1 2Xi−2 0Xi−3 1Xi−4 X
1i (= Xi−4
1 ⊕ Xi−3 ) 0
This
4 property
0 0 states,
0 1 that n of
2 1 total runs are of length n
Runs ↓ 1 1 31 0 1 01 1 10 0
(except all zeros [0 · · · 0]) There0 are 1a total
0 of 08 runs in the above example and
5 1 0 0 0 0
2 0 41 0 1 01 0 10 1 6 0

→ A LFSR which goes through 2D − 1 states is

3 0 50 1 1 01 00
18 1 0run of
1 n = 4 (Note # of state variables = 4)
0 0 7 0 0 1 0 1
4 0 0 1
60 0 0 11 0 01 0 1 1
called maximal length LFSR 1 = 0( 3 ×1 8) 1run of 1
0 n = 3, 21 = ( 2 × 8) run of n = 2
5 1 0 0 0 0 9 1 0 0 0
71 0 0 00 1 00 1
6 0 10
2 2
iii) Example: Consider a LFSR with 4 state variables and 7 0 80 1 1 00 0 11 1 11
4=(
1 1
0 1 1
0
× 8) run of n = 1
1 1 0
8 1 90 1 0 11 0 01 0 12
2 0 1 1
Xi = Xi−4 ⊕ Xi−3. 9 1 10 1 0 0 10 1 00 1 ii) Auto Correlation Property:
13
14
1
1
0
1
1
0
0
1
1
1
Xi Xi−1 Xi−2 Xi−3 Xi−4 Output 10 0 11 1 1 1 00 1 11 0 1 P
15 . . .1. . . . . . .1. . . . . . .1. . . . . . .0. . . . . . . . . . . . . . 1. . . . . . . . . . .
N
D D D D 11 1 12 0 0 1 11 0 10 1 → Auto correlation r(d) ,
All 1’s 16 1 1
n=0 C (n)C (n − d)
1 1 0
12 0
13 1 1 0 0 1 01 N
1 1 1 PN 2 N
Note that, r(d = 0) = C (n) = =1
Feedback 13 1 0 1 0 1
14 1 14 1 1 0 11 0 11 1 
C(n) = −1 −1 −1 −1 1 1 N
n=0
1 −1 1 1 −1 −1N 1 −1 1


This has 15(= 24 − 1) states, as a result is a maximal length

15 . . .1. . . .15 1 . . .1. . .1. . . .1. .0. . . . .1. . . . . . .01. . . . . . . . . . . . . . 1. . . . . . . . . . .
1 PN −1 −1
However, r(d = 2) = C (n)C (n − 2) =
..........................................
All 1’s 16All 1’s
1 16 1 1 1 11 1 10 0 n=0 =
LFSR N N 15
The CDMA PN sequence for this case is, C(n) = [ −1 −1 −1 −1 1 1 1 −1 1 1 −1 −1 1 −1 1 ]
→ The code is obtained by mapping the output Xi−4 as,   C(n − 2) = [ −1 1 −1 −1 −1 −1 1 1 1 −1 1 1 −1 −1 1 ]
C(n) = −1
−1 −1
−1 −1−1−1 −11 1 1 −1 1 1 −1 −1 1 −1 1
1C(n) × C(n − 2) = [ 1 −1 1 −1 −1 1 −1 1 −1 −1 −1 −1
0 → 1 and 1 → −1
1 1 1 ]

C(n) = 1 1 1 −1 1 1 −1 −1 1 −1

1 if d = 0,
(
=⇒ r(d) =
C(n) = [ −1 −1 −1 −1 1 1 1 −1 1 1 −1 −1 1 −1 1 ]
−1 (→ 0)
N if d =
6 0 (and largeN ).
System model of a CDMA (single-user) system C(n)
C(n − [ [−1−1 1 −1
2) = = −1 −1
−1 −1 −11 1 1 1 1 1 −1
−1 1 1 11 −1 −1 1 1]
−1 −1 −1 1 ]
C(n) × C(n
C(n−−2)2)= =
[ [ 1−1−1 1 1 −11 −1 −1
−1 1 −1−1 1 1 −1
1 1−1 −1
−1 −1
1 11 1−1] −1 1 ]
i) Consider a single-user CDMA system, i.e., User-0 transmits SNR
C(n) ×of a−
C(n single-user
2) = [ 1 CDMA
−1 1 system
1 −1 −1 1 −1 1 −1 −1 −1 BER 1of 1a single-user
−1 ] CDMA system
symbol a0, uses C0(n) ∈ {−1, +1} ∀n with length N , and let Further, analyzing the noise component i) Using SNR = SNRBPSK = P2 , we have
1 PN σ
Transmitted symbol x(n) = a0C0(n) E{e w} = N  n=0 E{w(n)}C0(n) =0 SNRCDMA = N |h|2SNR
CDMA
Received symbol y(n) = hx(n) + w(n), where e ∗} = E{ N1 n w(n)C0(n) N1 m w ∗(m)C0∗(m) }
  
E{e ww N SNR
P P r !

h is Rayleigh fading coeff 1 =⇒ BER = 12 1− (Exact) 

= 2 n m E{w(n)w ∗(m)}E{C0(n)C0∗(m)} 2 + N SNR
P P
N
w(n) is IID Gaussian noise N (0, σ 2) 2 1
= 12 n E{|w(n)|2}1 =
P σ ≈ (Approx. for high SNR) 
=⇒ E{w(n)} = 0 ∀n and N N
 2N SNR
σ2 if n = m, Note that, Noise power decreased by N
(
1
E{w(n)w ∗(m)} = Q) Consider CDMA system with code length N = 256 and
0 if n =
6 m Signal Power E{|h|2|a0|2} |h|2P
iii) SNRCDMA = = 2
=N 2  SNR = P2 = 15dB, compute BER over Rayleigh fading channel
ii) At receiver, y(n) is correlated with C0(n) Noise Power σ /N σ σ 
1.5
s
1 PN 1 PN This important factor N is called spreading gain or processing 256 × 10
ry0 = N n=0 y(n)C0(n) = N n=0 [hx(n) + w(n)] C0(n) Sol. BER = 12 1 −  = 6.175 × 10−5
1 PN 2 1 PN gain of CDMA system 2 + 256 × 101.5
= N n=0 ha0C0 (n) + N n=0 w(n)C0(n)
1 PN
= ha0 + w e , where we = N n=0 w(n)C0(n)
1
CDMA system with Frequency selective channel (or ISI)

1 i) Consider an L−tap channel with {h(0), h(1), · · · , h(L − 1)}

and the transmitted symbols as {x(0), · · · , x(n)}. Then,
y(n) = h(0)x(n) + h(1)x(n − 1) + · · · + w(n)
PL−1
= l=0 h(l)x(n − l) + w(n)
i.e., a clear case of inter symbol interference
SINR of multi-user CDMA system ii) Considering a single-user CDMA system x(n) = a0C0(n), then
PL−1
Further to analyze the interference component, define r01 , y(n) = l=0 h(l)a0C0(n − l) + w(n)
1 PN C (n)C (n)
N n=0 0 1 iii) Performing correlation operation at the receiver,
E{r01} = N1  n E{C0(n)} E{C1 1 PN −1
=0 r(0) = N n=0 y(n)C0(n)
P
(n)}
2 } = E{ 1 1 1 1
 
E{r01 C m C0(m)C1(m) } = N n l h(l)a0C0(n − l)C0(n) + N n w(n)C0(n)
P P P P P
N n 0 (n)C 1 (n) N
1 1
= l h(l)a0 N n C0(n − l)C0(n) +(w e , we have
= 2 n m E{C0(n)C1(m)}E{C0(m)C1(n)}
P P P P
N
1 2 2 1 1 1 if l = 0,
= 2 n E{C0(n) }E{C1(n) } = autocorrelation, N n C0(n − l)C0(n) =
P

P
N N −1 ≈ 0 if l 6= 0
N
1 if n = m,
(
As E{C0(n)C1(m)} = =⇒ r(0) = h(0)a0 + w 
0 if n 6= m
e
iv) Similarly, performing correlation with C0(n − 1), we get
i) Signal to interference plus noise ratio in this case is
r(1) = h(1)a0 + w 
Signal Power
e
System model of CDMA (Multi-user) system SINR = v) Vector model can be written as, r = ha0 + w,
Interference Power + Noise Power
e
|h|2|P0|2 N |h|2P0 r(0) h(0) w
     
e (0)
i) Consider a system with {user-0, user-1} transmitting symbols = 2 = 2 2
 . . .
2
σ + |h| |P1| 2
σ + |h| P1 . = .  a0 +  .
     
{a0, a1}, using {C0(n), C1(n)} with length N , and let N N
 
N |h|2P0 r(L − 1) h(L − 1) we (L − 1)
Transmitted symbol x(n) = a0C0(n) + a1C1(n) ii) Generalizing for K + 1 users, SINR = 
K
Received symbol y(n) = hx(n) + w(n) σ 2 + k=1 |h|2Pk Rake Receiver: Carrying out Optimal processing using MRC on r,
P

= ha0C0(n) + ha1C1(n) + w(n) hH

Example: 2-user CDMA system i.e., r, we obtain
khk 2
ii) At receiver, y(n) is correlated with C0(n) Q) Consider P0 = P1 = 15dB, σ 2 = 3dB, N = 256 and
k h k2 P
h = √1 + j √1 . Compute the SINR for user-0.
1 PN 2 SNR
ry0 = N n=0 y(n)C0(n) i) SNRrake = 2
= N k h k 
PN 2 2 σ /N
1
= N n=0 [ha0C0(n) + ha1C1(n) + w(n)] C0(n) 256 × ( 12 + 12 ) × 101.5 k h k2 is also known as multi-path diversity in CDMA
1 PN Sol. SINR = = 233.81 ≈ 23.68dB
= ha0 + ha1 N n=0 C0(n)C1(n) + w 2 + ( 12 + 12 ) × 101.5 L
1

2L−1

e
Note, the multi-user interference (MUI) component above ii) As a result BER = L , where L is the # of
2NSNR
1 PN
= ha1 N n=0 C0(n)C1(n) arises because of simultaneous Near-Far effect
h i
taps of freq. selective channel and coefficients h(0) · · · h(L − 1)
transmission in CDMA i) The interference of User-1 who is close to the BS than the are IID rayleigh with unity avg power
desired user-0 by d1 = √d0 equals the signal power.
N Example: Given L = 3 taps for FSC, N = 256 and SNR = 15dB.
3
This leads to disruption 1

5

Then BER = 3 = 2.36 × 10 −12
ii) Thereby the concept of power control is used to avoid this. 2 × 256 × 101.5

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
,

Dr. Fayazur Rahaman Mohammad,

,
Multiple Input Multiple Output Fading
Fading Channels
Channels Dept. of ECE, Mahatma
I I
Gandhi Institute
1
of Technology,
1
Hyderabad.
Special
I2 ICases
2
System Model
Tx Rx Rx Rx Rx
Tx Tx
Tx
SIMO: Single Input Multiple Output system ,
i) symbols, T = x , x , · · · , x can be
h i
Transmitter: t x y1 h1 w1
     
t t r r 1 2 t t t r r
. .  . 
transmitted onantennas
Tx the t transmit
Tx antennas antennas. Where x is the
Rx antennas
Rx antennas i.e., t = 1 and  .  =. x + . 
t dimensional transmit vector. yr r×1 hr r×1 wr r×1
x1 x1 y1
MISO: Multiple Input Single Output system ,
y1 Hr×tHr×t
x2 y2 y2 xr×1
t×1xx yyt×1
xxx
x1
x2 r×1 t×1
MIMO yyr×1
t×1 r×1
 
Rx Rx MIMO
Tx Tx Channel
Channel .
i.e., r = 1 and y = h1 · · · ht
h i
Fading Channels xxrt yr yr . +w
xxrt 1×t
I1
xt t×1
t Transmit
t Transmit r Receive
r Receive
I2
antennas
antennas
antennas antennas
Rx Rx SISO: Single Input Single Output system ,
Multiple Input Multiple Output
Tx Tx
ii) Receiver: r samples, yT =
y1, y2, · · · , yr are received
h i
i.e., r = t = 1 and y = hx + w
i) This is a key technology for 3G/4G wireless systems t r
across r receive antennas. y is the tr dimensional r
receive vector.
Multiple Inputs =⇒ Multiple Transmit Antennas (t)
Tx antennas
iii) Channel is represented by matrix Hr×t , which is defined by
Rx antennas Noise Vector analysis
Multiple Outputs =⇒ Multiple Receive Antennas (r) yr×1 = Hr×t xt×1 + wr×1 i) Each wi is Gaussian with zero-mean and σ 2 variance ,
MIMO =⇒ System with r × t antennas x1 y1 Hr×t
E {wi } = 0 and E |wi | = E wi wi = σ 2
 2
 ∗

h11 h12 · · · h1t
 
x2 y2 xxx yyt×1
ii) Further, wi are IID,
 i.e.,
Fading Channels t×1
r×1
MIMO r×1
Tx Rx
h21 h22 · · · h2t 
I1 Channel

H, yr  , σ 2 if i = j,
xxrt I2 ..
 n o
E wi wj =∗

E {wi } E wj∗ = 0 if i 6= j
  n o
h h hrt
Tx Rx Tx r Receive Rx
t Transmit r1 r2 · · ·
antennas antennas r×t
where h is the channel coefficient between i th receive → As a result noise Covariance matrix R can be obtained as 
t r t ij r
and j th transmit antenna σ2 0 · · · 0
 w1 h
  
Tx antennas Rx antennas
0 σ 2 0
iv) In general, in a r × t system we have
 i
  
H  .  ∗
n o
ii) MIMO leads to E ww = E  .  w1 · · · wr ∗ = . .. 
.
 
yl = hl1x1 + hl2x2 + · · · + hlt xt + wl , where yl is the sample 
wr

→ Increase in Data Rates by transmitting several information σ2
   
received at l th antenna. 0 ···
streams in parallel also termed Spatial Multiplexing.
R = σ 2Ir×r , where I is an identity matrix
I1 Example: 3 × 2 MIMO system
I2
3 receive antennas and 2 transmit antennas Least Square Cost Function
Rx Rx
Tx
y1 h11 h12 w1
     
x1 f (x) , k y − H x k2= (y − H x)T (y − H x)
" #
y2 = h21 h22 + w2
     
t r
x2 = (yT − xT HT )(y − H x)
y3 3×1 h31 h32 3×2 2×1 w3 3×1
= yT y − xT HT y − yT H x + xT HT H x
as
→ Increase in Reliability due to increased diversity y1 = h11 x1+ h12 x2+ w1
= yT y − 2xT HT y + xT HT H x
=⇒ y2 = h21 x1+ h22 x2+ w2 Differentiating f (x) and equating to 0, we have
y3 = h31 x1+ h32 x2+ w3 ∂f (x)
= 0 − 2HT y + 2HT H x = 0
∂x
Least Square problem (when r > t) =⇒ HT H x = HT y
=⇒ x̂ = (HT H)−1HT y 
i) However, if r > t (i.e., # of Rx antennas > # of Tx antennas),
then H is not a square matrix, =⇒ inverse of H does not exist. Zero Forcing Receiver
→ r > t =⇒ # of Equations > # of unknowns
i) x̂ = (HT H)−1HT y, is called as the Zero Forcing receiver
ii) Alternatively, define e , y − H x, now find x that minimizes
→ Generalizing it to the case where H is complex, we have
the error, i.e.,
x̂ = (HH H)−1HH y 
min k e k2= min k y − H x k2
ii) Note that HH H is a t × t square matrix and
iii) To find the x̂, differentiate the RHS w.r.t vector x and
(HH H)−1HH is the Pseudo (also called left) inverse of H
MIMO receiver (when r = t) set equal to zero. Further we use the below two properties,
i.e., (HH H)−1(HH H) = I
∂(cT x) ∂(xT c)
i) Objective is to recover the transmit vector x at the receiver a) = = c and
∂x ∂x
∂(xT Px) Example: ZF receiver for 3 × 2 MIMO system
from the receive vector,
Fading Channels
y = Hx + w b) = 2Px, where P = PT
ii) In a simple scenario, where r = t, Hr×r is a square matrix I1 ∂x i) Consider a MIMO system with r = 3 and t = 2
(i.e., # of Rx antennas = # of Tx antennas). BER for
I2 BPSK with ZF receiver y1 2 3 w1
     
x1
" #
→ Then, Tx if H is invertible, we get x
^Rx
= H−1y Rx y2
 
= 1 3
 
+ w2
 
i) Considering transmit symbol power E{|xi |2} = P and each hij
Tx
x2
using approx. that y ≈ Hx, i.e., system of linear equations y3 3×1 4 2 3×2 2×1 w3 3×1
t as IID complex Gaussian with unity avg. power,
MIMO t r r
−0.0405 −0.1676 0.3121
" #
L
1 H
ii) ZF receiver = (H H) H =
−1 H

Tx antennas Rx antennas 2L−1
BER = L , where L = r − t + 1 0.1676 0.2659 −0.1503


2SNR
i.e., equivalent to MRC system with r − t + 1 receive antennas =⇒ xˆ1 = −0.0405 y1 − 0.1676 y2 + 0.3121 y3
x1 y1
xˆ2 = 0.1676 y1 + 0.2659 y2 − 0.1503 y3
Hr×t
y2
x2
Rx
xxx
Q)
t×1
r×1
Given,
MIMO3 × 2 MIMO
yyt×1
r×1
system, SNR = 25dB and ZF receiver,
Tx Channel
yr compute the BER.
xxrt 2
Sol) Avg BER = 32 1

7.5 10−6


2.5 = ×
MISO: Transmit Beamforming (r = 1)
t Transmit r Receive 2×10
antennas antennas

i) Considering a 2 × 1 MISO system, we have SNR due to Transmit Beamforming

h1 k h k2 E{|x|2} k h k2 P BER of Transmit Beamforming

i) We then have SNR = 2
=
x1
σ σ2
Tx y Rx =⇒ Trasmit Beamforming is able to acheive SAME SNR i) Trasmit beamforming can be extended
h to a general scenario
x2 h2
with 1 × L MISO system, i.e., h = h1 · · · hL
i
as the MRC
ii) However, major CHALLENGE is that Channel State Information h ∗
 
i 1
i x . 1
" #
(CSI) is to be known as the transmitter x + w and at high SNR
h
1
+ w, then What is the optimal tx scheme x? y = h1 · · · hL  . 
h
y = h1 h2 khk
x2 hL∗
1 h1∗ Q.) Given a 1 × 2 MISO system, with CSI h1 = 2 + j and L
" #
1

ii) Transmit Beamformer uses, x = x, 2L−1
h2 = 1 − 2j available at the transmitter, compute the optimal BER = L , where L = # of Tx antennas


k h k h2∗
2SNR
where x is the transmit beamforming vector." #
" symbol and
Q.) Given 1 × 3 MISO with SNR= 25dB, compute avg. BER.
1 h1∗ 1 2−j
" #
1 h1∗
#
the unit vector is the transmit beamformer Sol) We have x = x=√ x and 3
k h k h2 ∗ 10 1 + 2j 1

k h k h2 5

∗ Sol) We have L = 3 and BER = 3 = 6.3 × 10−8
i 2−j 1 2 × 102.5
" #
√
iii) At the receiver,
" we # then have
h
y = 2 + j 1 − 2j √ x + w = 10x + w
i h∗ 1 + 2j 10
h
1 1
y = h1 h2 x +w =khkx +w
h2 k h k
∗
Alamouti Code BER
Alamouti Receiver (Orthogonal STBC)
2
3 1 3
i) We have L = 2 and BER =


ALAMOUTI Space Time Block Code ii) By rearranging y1 and y2∗ and process them jointly, we have 2 2× 12 SNR
= SNR2

y1 h1 h2 x1 w1 h1 h2 w1
" # " #" # " # " # " # " #
i) Does NOT require knowledge of CSI at Transmitter. = ∗ + = ∗ x1 + x2 + Q.) Alamouti processing "given # CSI h1 = 2 + j and h2 = 1 − 2j, " We# have
Very useful from practical viewpoint y2∗ h2 −h1 x2
∗ w2∗ h2 −h1 ∗ w2∗ i x
1
i −x ∗
2
Sol) y1 = 2 + j 1 − 2j x + w1 , y2 = 2 + j 1 − 2j
h h
x + w2
h1 h2 x x1
" # " #
∗
ii) Consider a 2 × 1 MISO"system transmit
# symbols x1 and x2 =⇒ y = c1 x1 + c2 x2 + w, where c1 , ∗ and c2 , # " #2 " #
y1 2 + j 1 − 2j x1 w1 2+j 1 − 2j w1
" # " " # " # " #
x1 h2 −h1∗
# "
−x2∗ = + = x1 + x2 +
using transmit vectors and at time instants Note that c1 and c2 are ORTHOGONAL !. i.e., cH
y2∗ 1 + 2j −2 + j x2 w2∗ 1 + 2j −2 + j w2∗
x2 x1
∗ 1 c2 = 0
y1
" #
cH
iii) To recover x1 from y, we process as Extract x1 using , kc11k y = √110 2 − j 1 − 2j f1, similarly
h i
=k c1 k x1 + w
1 and 2 respectively. y2
∗
cH c H cH cH
iii) The, the symbol 1 y= 1 c x + 1 c x + 1 w e where,
=k c1 k x1 + 0 + w
" y# that we receive at time instant#1 and 2 is y1
" #
kc1k 1 1 kc1k 2 2 kc1k cH 1
Extract x2 using , kc2k y = 10 1 + 2j −2 − j
h i
kc1k 2 √ =k c2 k x1 + w f2
i x H H y2
"
i −x ∗ c c
∗
1 2 +w e ∗} = E kc1 k wwH kcc1k = kc1 k E wwH kcc1k = σ 2
n o
+ w1 and y2 = h1 h2 E {wew
h h
y1 = h1 h2 2
x2 x1∗ 1 1 1 1
Q.) Given SNR = 25dB, compute the BER
k c1 k2 E{|x1|2} k h k2 P2 1 SNR of MRC !
iv) Then, SNR = = = 2 Sol) BER = 3 2 = 23 2 = 3 × 105
σ2 σ2 SNR (10 .5)

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
,

Dr. Fayazur Rahaman Mohammad,

,
Characterizing the MIMO Channel Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.

Spatial Multiplexing

Singular Value Decomposition (SVD) SVD in MIMO Systems i) At the tx, by performing precoding i.e., x = Ve
x, we have
y H Ve
i) The channel matrix Hr×t is SV decomposed for r ≥ t when i) SVD can be employed in MIMO comm. systems as a tool.    x+w
e = Σ V e = Σx e+ w
e 
ye1 σ1 0 · · · 0 xe1 w

f1
Hr×t = Ur×t Σt×t VH t×t Consider the MIMO model
ye2  0 σ2
  
 xe2 w
   
2
yr×1 = Hr×t xt×1 + wr×1, now substituting H = U Σ VH   +   ,
f 
 ..   ..   ..   .. 
   =
σ1 0 0  H
v1
v2H  y = U Σ VH x + w       
yet 0 · · · 0 σt xet wet
 
0 σ2 
ii) At the receiver, multiply the received vector with UH , we have
 
Hr×t = u1 u2 ··· ut r×t 
 
  .. 
   . 
0 
=⇒ yei = σi xei + wei , for 1 ≤ i ≤ t
U H y = U H U Σ VH x + U H w

0 0 σt t×t vtH t×t
ii) This indicates, there are t parallel channels and t information
ii) Columns of U are y = Σ VH x + w e , where y e , UH y and w e , UH w
( Orthonormal symbols can be transmitted in parallel
e
k u k 2= 1 if i = j,
i
i.e., uH
i uj = Examples: Using SVD, SPATIAL MULTIPLEXING is obtained !
0 if i 6= j
1 h
uH
 
1 0 ··· 0
   
1
" #
√ √
1
I) SVD of H = is =  12  2 1 Example: Consider 3 × 3 MIMO Channel
ih i
 H h
u  0 1 · · · 0 1
 
H 2 √
i
U U=  u1 u2 · · · ut = 
 ..   ..
 = It×t 2
 √
where UH U = 1, σ1 = 2 > 0 and V = [1] is unitary matrix i) Given H , and observing that its 
columns are orthogonal
uH 2 √
   
0 ··· 0 1 √ 0
−6

2 −6 0 √ 13 0 0
  
t # " # " √ # " # "√ 13 52  √
→ Similarly VH V = VVH = It×t , i.e. V is UNITARY matrix 1 0 0 1 0 5 0 1 5 0 0 1   √3 √4  
" #" #
H = 3 4 0 =  0  0 52 0

II) H =
 
√ = = 13 52 
iii) Σt×t is a diagonal matrix with σ1 ≥ σ2 ≥ · · · ≥ σt ≥ 0, 0 5 1 0 1 0 1 0 0 1 1 0 0 0 2 0 0 1 0 0 2
→ Singular values are non-negative and are arranged in decreasing where UH U = I, VH V = VVH = I Toget the singular
√ √   values√in decreasing order
order and σ1(= 5) > σ2(= 1) > 0 −6 √2 0
√ 0 52 0

→ The # of non-zero singular values indicate the rank of  452 313  √
√ 0  13 0 0
1 2 1 1 1 0 1 1 0 2
" # " #" # " #" #
= √

the channel matrix III) H = = =
 52 13 
1 −2 1 −1 0 2 −1 1 1 0 0 0 1 0 0 2
MIMO-II 
1 1
" √ 
−6 √2 0
 √
√ 52 0 0 0 1 0
 
1 1 2 0 0 1 2 2 0 0 1
" #" #" # #" #
√ √
= = 2 2 √  452 313   √ H
−1 1 0 1 1 0 1 1 0 2 1 0 √ 0  0 13 0 1 0 0 U V
 
= √ = Σ
 
−√ √  52 13 
2 2
  
0 0 1 0 0 2 0 0 1
Solution using Lagrange multipliers → Note, UH U = I, VH V = VVH = I and
Capacity of a MIMO wireless system √ √
σ1(= 52) > σ2(= 13) > σ3(= 2) > 0
i) Need to maximize the !
below equation
i) Considereing spatial multiplexing in MIMO i.e., ii) Applying transmit preprocessing (Precoding) at the transmitter,
Pt σi2Pi  Pt 
yei = σi xei + wei where σi is gain of i th channel, i=0 log2 1 + + λ P − i=0 Pi x1 0 1 0 e x1
    
σ 2
E{|xei |2} = Pi is power of i th transmitted symbol and x = Ve x, i.e., x2 = 1 0 0 e x2
    
2P ii) From partial differentiation w.r.t Pi and equating to 0
σ i x3 0 0 1 e x3
wei is Gaussian with var σ 2, we have SNRi = i 2 ∂ σi2/σ 2
σ = 2 2
+ λ(0 − 1) = 0 iii) At the receiver, y e = Σe x+w
σi2Pi ∂Pi 1 + σi Pi /σ   √
! e
y1 52 0 0 e x1 w
   
ii) Then we have Shannon rate = log2 1 + 2 2
σi /σ 2 
2
+
√
e1
1
e
σ =⇒ λ = 2 2
, =⇒ Pi = λ − 2 σ
i.e., ey2  =  0
  
13 0 ex2 + w
   
e2
Pt 1 + σ P i /σ σi
The sum rate = i=0 log2 (1 + SNRi ) i
y3 0 0 2 e x3 w
e3
 1 − σ if 1 − σ2 ≥ 0,
2
 e
iii) Considering transmit power as P, the ! capacity of MIMO Pi = λ σi2 λ σi2  → 3 decoupled channels resulting in spatial multiplexing.
1σ 2P 0 otherwise
C = max.
Pt
1 i i
i=0 log 2 +
Example: Optimal Power Allocation
P σ2 iii) To obtain the value of Lagrange 
multiplierλ, need to solve
Pt
sub.to i=0 Pi = P
Pt Pt 1 2
i) To achieve Shannon capacity
i=0 Pi = P, i=0 λ − 2 = P
=⇒ σ
which is a constrained optimization problem σi 
2
+ 
2 +

2
+ 
2
+
→ This algorithm is termed as the Water Filling Algorithm P1 = λ1 − σ2 = λ1 − σ52 , P2 = λ1 − σ2 = λ1 − σ13

σ1 σ2
+ 
2 2
 +
and P3 = λ1 − σ2 = λ1 − σ4
σ3
ii) Considering σ 2 = 0dB = 1 and total Tx power P = 3dB ≈ 2
→ P1 + P2 + P3 = 2, =⇒ λ1 = 0.7821
P1 = −1.1755dB, P2 = −1.517dB and P3 = −2.74dB
→ Power allocation decreases as the channel gain σi2 decreases

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]
,

Dr. Fayazur Rahaman Mohammad,

,
Orthogonal Frequency Division Multiplexing Dept. of ECE, Mahatma Gandhi Institute of Technology, Hyderabad.

Basic Principle of OFDM

Principle to Overcome ISI

Single carrier
Single carrier subcarriers subbands
=N
Orthogonal Frequency Division Multiplexing B
N
i) Key enabler of high data rates (BROADBAND Technology) in
4G wireless systems and WiFi (WLAN) systems Broadband B = 10M Hz
Broadband B = 10M Hz B = 10M Hz
→ LTE: Long Term Evolution i) Consider a broadband channel with single-carrier and
i) B each
Divide the bandwidth B into N subbands of bandwidth N
→ LTE-A: LTE Advanced bandwidth B = 10 MHz,
WiMAX: Worldwide Interoperability for Microwave Access Use a sub-carrier for each subband
→ → Symbol duration T = B1 = 10MHz 1 = 0.1µs
WLAN: 802.11a, 802.11g, 802.11n, 802.11ac ii) Example: B = 10MHz and N = 1000
→ Typical delay spread of a channel Td ≈ 2 − 3µs
=⇒ BW of each subband = 10MHz 1000 = 10KHz
=⇒ T  Td (i.e., Symbol Time  Delay spread), 1
Symbol time in each subband T = 10KHz = 100µs  Td (= 2µs)
results in ISI or Inter symbol interference
=⇒ NO ISI in new system with Multi-Carrier Modulated
Single carrier As bandwidthsubbands
→ subcarriers B increases symbol time decreases and leads to ISI
(MCM) system
ii) It is a significant challenge
=N in a broadband wireless system to
overcome ISI B Alleviate the need for large number of subcarriers
Single carrier subcarriers
N subband
B
i) Since the signal is bandlimited to {−fmax , fmax } i.e., {− 2 , 2 }, B B =N
Extraction of symbols at receiver
it can be sampled at the Nyquist rate = 2fmax = B
N
Broadband B = 10M Hz B = 10M Hz
i) To extract symbol Xl on the l th subcarrier ii) Sampling interval = B 1 = T B = 10M Hz
B = 10M
Broadband
Transmission in MCM System R f10 −j2πlf t R f10 −j2πlf t P
f0 0 e 0 x(t)dt = f
0 0 e 0 X e j2πkf0t dt l th sampling instant = lT = Bl
k k
1
f0 −j2π(k−l)f
iii) we have Single carrier system
subcarriers0 t subbands
subcarriers = k Xk f0 0 e = k Xk δ(k − l) = Xl Psubcarriersj2πkf
Single carrier
R
subbands
P P
subbands x(t) = X e 0t = N
k k
ii) At receiver by correlating (coherent demodulation) with e−j2πlf0t
=N
=N
j2πk B l l T = 1/B
B

x(l) = x(lT ) =
P
X e j2πkf
B
0 lT P
X e N B
Total tim
k k = k k
N

or matched filtering with e−j2πlf0(T −t), one can extract symbol

N

B P−B 0 Bj2π2Bkl
th
on l subcarrier = kN Xk eN NN (which is IDFT or IFFT)
N Broadband B = 10M Hz B = 10M Hz

−B
th−f0 0 f0 2f0 N symbols
→ where x(l) is the l IDFT point of symbols transmitted on
0 B 2B across N
Single carrier system
N subcarriers {X0, X1, · · · , XN −1}
N N N subcarriers subbands subcarriers
−f0 0 f0 2f0
iv) The principle of OFDM is to transmit the sample points
=N
T = 1/B Total time = N/B
B,
B

i) Consdier each subcarrier is transmitted with spacing of f0 = N i.e., IFFT at Transmitter and FFT at Receiver
N

−B 2B
0 B

i.e., at frequencies {· · · , −2f0, −f0, 0, f0, 2f0, · · · }

N N N
−f 0 0 f0 2f0 N symbols
ii) k th subcarrier frequency is kf0 = k N B across N MCM IFFT at Transmitter and FFT at Receiver Total time = N
subcarriers Multicarrier
→ Carrier is ej2πkf0 and let symbol be denoted by Xk
→ The signal transmitted on k th subcarrier is
Transmit symbols: {X0 , X1 , · · · , XN −1 }

Xk ej2πkf0t
N-pt IFFT
sk (t) =subbands
subcarriers
Trasmitted Samples: {x(0), x(1), · · · , x(N − 1)}
iii) Net transmitted =MCM signal
Total time = N/B
N Sampled version
of MCM signal Channel
x(t) =
P
X e j2πkf0t
k k
B

Practical difficulties of MCM system: {y(0), y(1), · · · , y(N − 1)}

N

Note, this is similar to Fourier series

i) Generating a MCM signal is difficult because of the large number Reciver: N-pt FFT
B = 10M Hz
of subcarriers {X0 , X1 , · · · , XN −1 }

N (= 1000) number of local oscillators required, filters etc N

Single carrier system
OFDM Equivalent N Flat fading channels
X(0)
i) Equivalently, Y (k) can be considered received symbol on
Cyclic Prefix X0 , X1 , · · · , XN −1

k th subcarrier, H (k) channel coefficient N-pt IDFT th

of k subcarrier and
T = 1/B Total time = N/B
Cyclic Prefix to avoid IBI X(1)
as the
X (k)x(N − L̄),transmitted
· · · , x(N − 1)symbol on the
x(0), x(1), th subcarrier
· · · k, x(N − 1)
Cyclic Prefix X0 , X1 , · · · , XN −1 Prefix of L̄ L̄ Samples
N symbols Nfrom
samples Flattailfading channels
across N MCM N-pt IDFT
H(0) X(N − 1)
subcarriers Multicarrier Cyclic
x(N − L̄), · · · , x(N − 1)Prefix
x(0), x(1), · X X1 , ·−
· ·0,, x(N · · 1)
, XN −1 X(0) Y (0)

Prefix of L̄ L̄ Samples
N-pt IDFT H(1)Circular Convolution
samples from tail X(1) Y (1)
Frequency Selective Channel scenario Cyclic Prefix x(N −XL̄), · ·1·, ,· x(N
· · , X− 1) x(0), x(1), · · · , x(N − 1)
n=1 h(1) h(0) h(4) h(3) h(2)
0, X
i) In a block of transmitted sysmbols, the L̄ samples from the tail
N −1
n=0
i) Model for Freq. selective channel is given by Prefix of L̄ N-pt IDFT L̄ Samples h(0) h(3) h(2) h(1)
are prefixed at the head samples from tail
H(N − 1)
y(n) = Total
h(0)x(n) h(1)x(n − 1) + · · · + h(L − 1)x(n − L + 1)
+ N/B
time = ii) x(N
The received
− L̄), symbols
· · · , x(N y(1)· · corresponding
y(0),x(1),
− 1) x(0), · , x(N − 1) to x(0), x(1) X(N − 1)
h�x
Y (N − 1)
ii) Consider 2 consecutive OFDM blocks, then aresamples
given by
Prefix of L̄ L̄ Samples
from tail x(N − 3)
Circular Convolution x(0) x(1) x(N − 1)
} �0 , X
�1 , · · · , X
�N −1 y(0) = h(0)x(0) + h(1)x(N − 1) + · · · + h(L − 1)x(N − L + 1) =⇒ OFDM has converted Frequency Selective channel
x(N − 2)into
X X0 , X1 , · · · , XN −1 h(1) h(0)
equivalent N Flat fading channels
n=1
y(1) = h(0)x(1) + h(1)x(0) + · · · + h(L − 1)x(N − L)
h(4) h(3) h(2)
N-pt IFFT N-pt IFFT n = 0 Convolution
Circular h(0) h(3) h(2) h(1)
− 1)}
�(1), · · · , x
�(0), x
x �(N − 1) x(0), x(1), · · · , x(N − 1) h(1) h(0) Tx and Rx Schematic
n=1 h(4) h � xh(2)
h(3)
Transmit samples Transmit samples
n=0 h(0) h(3) h(2) h(1)
of OFDM sym 1 of OFDM sym 2 Transmit Symbols: X(0) X(1) · · · X(N − 1)
− 1)} Channel x(0) x(1) x(N − 3) x(N − 1)
output for x(0) h�x x(N − 2) N-pt IFFT

Symbols to be Transmitted: x(0) x(1) · · · x(N − 1)

x(N − 1) + · · · + h(L − 1)�
y(0) = h(0)x(0) + h(1)� x(N − L + 1)
1} x(0) x(1) x(N − 3) x(N − 1)
Cyclic Prefix

→ Results in Inter Block Interference (IBI)

x(N − 2) Symbols Transmitted : x(N − L̄) · · · x(N − 1) x(0) · · · x(N − 1)
with Cyclic Prefix Prefix
iii) Addition of cyclic prefix (CP) has resulted in a circular Channel
BER of OFDM convolution at the output of the OFDM system, i.e., y = h ~ x
Received Symbols: y(N − L̄) · · · y(N − 1) y(0) · · · y(N − 1)
i) Considering the Gaussian noise at the receiver we have, =⇒ applying FFT, Y (k) = H (k) X (k), where Remove Prefix
y = h ~ x + w, and Y (k) = H (k) X (k) + W (k), where → Y (k) is the k th DFT/FFT coefficient of OFDM received signal Symbols to be processed : y(0) y(1) · · · y(N − 1)
W (k) is the k th FFT pt of [w(0), w(1), · · · , w(N − 1)], i.e.,
→
PN −1 l [y(0) y(1) · · · y(N − 1)] N-pt FFT
W (k) = l=0 w(l)e−j2πk N and
PL−1 −j2πk Nl → H (k) is the k th DFT/FFT of [h(0) h(1) · · · h(L − 1) 0 · · · 0]
→ H (k) = l=0 h(l)e Symbols with Flat fading : Y (0) Y (1) · · · Y (N − 1)
ii) Assuming that noise samples w(l) are IID Gaussian with, Detect symbols

E{w(l)} = 0, E{|w(l)|2} = σ 2 and E{w(l)w ∗(l̃)} = 0 Detected Symbols : X̂(0) X̂(1) · · · X̂(N − 1)
→ We then have Example:
−j2πk Nl
PN −1
E{W (k)} = l=0 E {w(l)} e =0
E{|W (k)|2} = E {W (k)W (k)∗} Q) Given L = 16 channel taps, N = 256 subcarriers and
SNR = 35 dB, compute OFDM BER.
l l̃
∗ o
N −1 N
nP  P
=E l=0 w(l)e
−j2πk N −1
l̃=0
w( l̃)e −j2πk N

u 16 3.5
   v 
(l−l̃)
=E
PN −1 PN −1
w(l) w(l̃)e
∗ −j2πk N u 256 10
l=0 l̃=0 1
Sol) BEROFDM = 2 1 − t  = 2.5 × 10−3
PN −1 PN −1 n o (l−l̃) 16 103.5
2 + 256
= l=0 l̃=0
E w(l)w (l̃) e ∗ −j2πk N
PN −1 2
= l=0 σ = N σ 2 
iii) Similarly, assuming the channel coeffs h(l) as IID complex Gaussian with
E{h(l)} = 0, E{|h(l)|2} = 1 and E{h(l)h ∗(l̃)} = 0 (Uncorrelated Scattering)
→ E{H (k)} = 0 and E{|H (k)|2} = L  (Rayleigh fading with avg. power L)
iv) With E{|X (k)|2} = P, we have SNR at receiver as
|H (k)|2P LP
SNRr = and Avg{SNRr } =
N σ2 N! σ2
L
N SNR
s
=⇒ BEROFDM = 2 1 − 1
L
, where SNR = σP2 (Transmit SNR) References
2 + N SNR
[1] NOC: Principles of Modern CDMA MIMO OFDM Wireless
Communications By Prof. Aditya K. Jagannatham, IIT Kanpur

Aug-Nov 2019, Version 1.1 This work is licensed under a Creative Commons Attribution 4.0 International Suggestions: [email protected], [email protected]