0% found this document useful (0 votes)
19 views102 pages

Multi 2

WCN

Uploaded by

Chethana Hs
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
19 views102 pages

Multi 2

WCN

Uploaded by

Chethana Hs
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 102

Multi-Carrier Modulation

Dr. Ahmad Gomaa


Assistant Professor
Electronics and Communications department
Faculty of Engineering, Cairo University
http://scholar.cu.edu.eg/gomaa
Contact: [email protected]
Outline
• Single carrier (SC) modulation
• Shortcomings of SC systems
• Multi-Carrier modulation (MCM)
– Main idea
– Implementation using IFFT and FFT
– MCM in multi-path fading
– OFDM parameters
– Bit-Interleaved Coded OFDM
– PAPR
– Subcarrier loading
– OFDM under RF impairments (CFO, PN, IQI, Doppler)
– Single-Carrier Frequency-Division Multiple-Access (SC-
FDMA)
Single Carrier (SC) Modulation
• Bits are modulated according to M-QAM or M-
PSK
• Modulated symbols are passed through a pulse
shaping filter (e.g., Square-Root Raised Cosine
(SRRC) filter)
• Output is up-converted to the carrier frequency
(fc)
Single Carrier Modulation

Pulse Shaping filter

bits M-QAM
(or M-PSK) π/2

Pulse Shaping filter

Pulse shaping filter specifies


 symbol rate & signal shape in frequency-domain
Single Carrier (SC) Modulation
• Transmitted signal passes through multi-path
channel
• Inter-Symbol Interference (ISI) occurs due to
the channel
• The receiver has to apply equalizers to extract
the modulated symbols
• Complex equalizers and not even optimal!
Multi-path channel
Reflecting object
(car, building, ….)

Transmitter Receiver

Reflecting object
(car, building, ….)
Channel impulse response
Multi-path channel
• Multi-path channel causes inter-symbol
interference (ISI)
• Need equalizers at the receiver to reverse the
channel effect
• Baseband model
Channel y[n]
x[n]
h Received samples
Modulated
QAM symbols
z[n]
Receiver noise

y[n]   h[k ]x[n  k ]  z[n]


L

Convolution between x[n] and h


k 0
Multi-path channel
• y[n] is not only function of x[n] but also
function of previous symbols {x[n-1], x[n-2], ….
x[n-L]}
• These previous symbols act as interference on
the current symbol x[n]
• Need to extract x[n] from y[n]
• This is done using equalizers
Equalizers
y[n] xˆ[n]
Channel Equalizer
x[n]
h w
Modulated
QAM symbols
z[n]

xˆ[n]  y[n]  w[n]   w[k ] y[n  k ]


M

k 0

 x[n]  h
[ w
n] [ n ] 
  z[ n ]  w[ n ]
should be   [n] Noise

If w[n] is designed to make h[n]*w[n] = δ[n], then we force the ISI = 0


 Zero-Forcing equalizer  effect on noise?
Equalizers

h[n]  w[n]   [n]


H ( f )W ( f )  1
Channel frequency response

Channel impulse response


Time-domain H(f)
To
h(n) Frequency-domain
f
n
Dip frequencies

As channel length (rms delay spread) increases, frequency-selectivity increases

One-tap channel has flat frequency response


Shortcomings of SC systems
• For W(f) H(f) = 1, W(f) will have large magnitude at dip
frequencies
• W(f) will enhance the noise at these dip frequencies
• When looking in time-domain, these enhanced noise will
impact the whole time-domain  all symbols are impacted by
noise enhancement
• ZF equalizer  Noise enhancement  Poor performance! 
Solution? (Multi-Carrier Modulation)
Multi-Carrier Modulation - Main Idea
• If we can divide H(f) into N subbands (subchannels or subcarriers), each of
small BW
• Then, channel is approx FLAT over each band
• We equalize every subcarrier alone by dividing by the channel gain at this
subcarrier
• We transmit 1 QAM symbol over every
subcarrier
• In receiver, we divide received signal ……
f
into N subcarriers, divide every subcarrier
by channel gain at this subcarrier to detect
Subcarrier (channel approx flat)
QAM symbol at this subcarrier
• Equalizer design is much easier  No convolution
• Only QAM symbols at dip frequencies will suffer  No effect on other
QAM symbols as they are transmitted on other subcarriers
Multi-Carrier Modulation - Main Idea
• For QAM symbol to be transmitted over a small BW symbol
duration has to be large  small data rate?
• Every symbol will have small symbol rate but we transmit N symbols
in parallel  Effective rate is high!

5us
cos(w1t)
Example
Serial to Parallel

.
.
. cos(w5t)
1us
Bit rate = 5 bits / T
Bit rate = 1 bit / 1us
= 5 bits / 5us
= 1Mbps
= 1 Mbps
T
Multi-Carrier Modulation - Main Idea
• In general, we have QAM symbol (which is complex a+jb) rather
than just BPSK symbols (which are real as in the previous example)
• To modulate Xk = ak+jbk over subcarrier k, we need exp(jwkt) and
rather than just cos(jwkt) 

X k exp jk t 
• So, we can write the total output as

x(t )   X k exp jk t ,


N
0t T
k 1

T  Total Symbol duration

Xk = ak+jbk  QAM symbol


Multi-Carrier Modulation - Main Idea
• How to choose wk’s?
• To easily separate these N subcarriers at the receiver, we need
them to be orthogonal, i.e.,

 X exp jk t  X l exp jl t  dt  0


T

k
t 0

• To satisfy this, we can show that

2
k  k, k  1,2,....., N
T

x(t )   X k exp j
N
 2k 
t , 0t T
k 1  T 
MCM – Implementation with IFFT
• In digital domain, set t = n Ts

x(nTs )   X k exp j
N
 2kn  T
Ts , 0  nTs  T  0  n 
k 1  T  Ts

This equation can be effectively implemented using


Inverse Fast Fourier Transform (IFFT)

{X1,X2,….XN} N-point {x(0), x(1),…. x(N-1)}


IFFT

x(n)   X k exp j
N
 2kn 
, 0n N
k 1  N 
MCM – Implementation with IFFT
• For x(nTs) = x(n), we need
T T
N  Ts 
Ts N
• We choose the sampling time (Ts) to be total symbol time (T ) over
number of subcarriers (N)
• Since we choose the subcarrier frequencies as:
2
k  k, k  1,2,....., N
T
• Then, subcarrier bandwidth is the difference between any two
subcarriers which is:
 k 1  k 1
Subcarrier BW  f   
2 2 T
MCM – Implementation with IFFT
• Hence, the total BW of the system is

Total BW  Number of subcarriers  subcarrier BW


N 1
 Nf  
T Ts
MCM – Waveform in time-domain
x(n)   X k exp j
N
 2kn 
, 0n N
k 1  N 
Time-domain signal = weighted summation of sinusoidal signals of
frequencies {k/N, k=1,2,…,N}

Total symbol duration = T

Freq= 1/N
Multiplied
by X2 Multiplied by X1

nTs
0
Ts N-1

Every subcarrier carries


One QAM symbol Freq = 2/N
MCM – Waveform in frequency-domain
Signal in Frequency domain
Subcarrier spacing = 1/T

At the peak of any subcarrier,


contributions from all other subcarriers = zeros  Orthogonal
MCM – Subcarriers orthogonality
• From the previous figure, we clearly notice the subcarriers are
orthogonal to each other
• This system is called OFDM (Orthogonal Frequency Division
Multiplexing) where orthogonal frequency comes from the fact that
the subcarriers are orthogonal to each other
• In the receiver, how to recover Xk from x(n) ?
• Exploit that the subcarriers are orthogonal as follows:


ˆ 1 N 1
 2kn 
Xk  x(n) exp  j , 1 k  N
N n 0  N 

• We multiply x(n) by the conjugate of the kth subcarrier and run the
summation.
• This will null out all subcarriers except for the kth one, it will give NXk
• We divide by N to get Xk
MCM – IFFT/FFT implementation
• This previous equation is exactly what FFT
(Fast Fourier Transform) does
• So, in Tx we run IFFT and in Rx we run FFT

X k 
N N N N
X̂ 
k
bits

Serial to Parallel

Parallel to Serial
bits
Serial to Parallel

N-point FFT
Parallel to Serial
N-point IFFT

QAM QAM
mod demod
MCM in multipath fading
• This works great if channel is AWGN with no
multipath, i.e., channel is flat with no frequency
selectively
• Even the single-carrier case works great with flat
channel (no multi-path) without the headache of
FFT and IFFT
• Now, what if there’s multi-path channel?
MCM in multipath fading
• Consider a multipath channel with only 2 paths:
A direct path with zero delay and another path
with some delay
• Focus on the subcarriers 1/N and 2/N
• In the next page, the solid curves are the two
subcarriers of the direct path (no delay)
• The dashed curve is the delayed version due to
the delayed path
• We plot only the delayed path of subcarrier 2/N
to avoid crowding
MCM in multipath fading

Now, to get X1, we multiply the whole received signal by the blue signal and integrate (sum)
This will cancel out the solid red signal BUT not the dashed red because it’s not a COMPLETE
Sinusoidal.
So, we will have inter-carrier interference (ICI) from X2 and other subcarriers on X1
MCM in multipath fading

 So, multi-path fading  ICI in OFDM

 How to solve that?

 Solution is to make the dashed sinusoidal a COMPLETE wave, i.e.,

have an integer number of periods in the integration (summation) time

 This is done using Cyclic prefix, i.e., copy the last portion of the total

OFDM symbol and put it before the beginning of the OFDM symbol
MCM in multipath fading

Copy
MCM in multipath fading

Now, the delayed version of subcarrier 2 will have an integer number of periods4
within the OFDM symbol duration (summation interval), so it will be cancelled out
when we multiply the received signal by subcarrier 1 and integrate  No ICI
MCM in multipath fading
• How to choose the length of cyclic extension?
• It should be chosen such that it’s greater than
effective channel length
• Cyclic extension converts the linear convolution
between Tx signal and channel into Cyclic Convolution
• Cyclic convolution in time-domain is equivalent to
multiplication of FFTs in Frequency-domain
• FFT of transmitted signal is simply the QAM symbols
(Xk)
• FFT of the channel is the channel frequency response
MCM in multipath fading
Circular convolution
Received signal Thanks to Cyclic extension!

y (t )  x(t )  h(t )  z (t )
Taking FFT we get
Yk  X k H k  Z k , 1  k  N
Assuming channel knowledge,
ˆ Yk
Xk  , 1 k  N
Hk
Here, equalizer is ONLY 1-tap and it’s Maximum Likelihood (ML)  Optimal
 ML minimizes error rate
MCM – Cyclic Extension

IFFT output IFFT output


Cyc. Cyc.
…..
Exten (N samples) Exten (N samples)
time
OFDM symbol OFDM symbol
1 2

• Cyclic extension  Prevents multipath channel from


introducing ICI after taking FFT at the receiver
• Cyclic extension is also called Cyclic Prefix
• Does it have another function ?
MCM – Cyclic Extension [Guard Interval]
• When OFDM signal passes through multi-path channel, every
symbol will leak into its successor (the symbol coming after it)
• To prevent inter-symbol interference (where symbol here
refers to OFDM symbol), we need to have Guard Interval
between OFDM symbols
• Thanks to its position lying between successive OFDM
symbols, Cyclic Extension acts also as Guard Interval

IFFT output IFFT output


(N samples) (N samples)
time

OFDM symbol 1 OFDM symbol 2


Leakage of
OFDM symbol 1
Does NOT reach
OFDM symbol 2
MCM – Pilot and Guard subcarriers

ZEROS Nused  Number of used subcarriers ZEROS


OFDM parameters
Parameter Definition Notes

Ts Sampling time Ts = 1/B = 1/bandwidth (seconds)


Guard interval length Tcp = NcpTs > Effective channel length
Tcp
(Cyclic Prefix length) (seconds)
T IFFT output length T = NTs = 1/Δf (seconds)

Ttot OFDM symbol length Ttot = T+Tcp= (N+Ncp)Ts (seconds)


Subcarriers frequency spacing =
Δf Δf = 1/T = B/N (Hz)
frequency of 1st subcarrier

B OFDM bandwidth B = 1/Ts = N Δf (Hz)


Number of subcarriers
N N = B/Δf = T/Ts
(IFFT size) = # samples in IFFT output

Bused Occupied OFDM bandwidth Bused = Nused Δf (Hz)

Nused Number of used subcarriers


OFDM parameters (Cnt’d)
Parameter Definition Notes

Bguard Guard subcarriers bandwidth B = Bguard + Bused (Hz)

Nguard Number of guard subcarriers N = Nguard + Nused

Ncp # samples in cyclic prefix Ncp = Tcp/Ts

fs Sampling rate fs = 1/Ts = N Δf (Hz)

How to choose T = 1 /Δf ?

T  4TCP  Reasonable CP overhead (rate/power efficiency)


f  Coherence bandwidth  Flat channel over subcarrier
1
Coherence bandwidth α al

channel length
4G LTE OFDM parameters
Licensed BW 3 MHz

= B (OFDM BW)
(N)

IFFT output
= 1 /subcarrier spacing (Δf) = 1/ 15000 = 67 us
length (T)
Nused = Bused/Δf 72+1(DC) 180+1 300+1 600+1 900+1 1200+1
Nguard = N-Nused 55 75 211 423 535 847
Bused = Nused Δf
1.08 2.7 4.5 9 13.5 18
(MHz)
BW efficiency =
Bused /licensed 86 % 90 %
BW x 100 %
4G LTE spectrum
Used BW (Bused)

ZEROS
Used subcarriers ZEROS

Left Guard Right Guard


Subcarriers Subcarriers freq

Licensed BW  Purchased spectrum

B = fs = sampling frequency

We transmit zeros on part of licensed BW  In air, it accommodates side lopes


of used subcarriers  side lopes will not get outside licensed BW 
So we do not interfere with neighboring bands

We transmit outside licensed BW but we just transmit zeros  No


interference on neighboring bands
4G LTE spectrum

Side lopes of modulated subcarriers


WiFi OFDM parameters

IFFT size (N) = 64 (52 used subcarriers + 1 DC + 11 guard subaccreirs)


IFFT output duration (T) = 1 / subcarrier spacing = 1/(312.5x103)= 3.2 us
Cyclic prefix length = Tg = 0.8us = T/4
Sample time (Ts) = 1/(20x106) = 0.05 us
Used BW (Bused)= 53 (52 + 1 DC) subcarriers x 312.5 kHz = 16.6 MHz = OBW = Occupied BW
Comments on OFDM parameters
• Cyclic prefix (CP) length in LTE > CP in WiFi, why?
• Because WiFi works in indoor environment while LTE works outdoor.
• In outdoor, channel paths can be reflected from far objects so will
come to Rx after long delay  Need longer cyclic prefix
• What if we use too long CP, much longer than channel length?
• Then, we need IFFT output length (T > 4Tcp) to be very long as well
– Channel can vary during OFDM symbol (because it’s too long in
duration)  Doppler effect  ICI See slide
– Large latency  receiver will need to wait too long before it can receive
OFDM symbol  OFDM detection cannot start before the whole OFDM
symbol is received  Large latency not suitable for real-time
applications (voice/video chatting, gaming, …)
• What if we use too short CP, shorter than channel length?
• Interference between successive OFDM symbols (every OFDM
symbol will leak into its successor)  ISI
• Subcarriers orthogonality will be destroyed leading to Inter-Carrier
interference (ICI)
MCM – Deep fading and NBI
• What about subcarriers with DEEP fading?
• What if a Narrow-Band Interference (NBI) hits
OFDM signal? (e.g., BLUETOOTH into WLAN)

Narrow-band Interference (NBI)

……
OFDM signal

Deep Fading
MCM – Deep fading and NBI
• In case of deep fading, Hk is very small, so dividing by it will amplify
the noise  Gives wrong QAM symbol after dividing by Hk

• In case of NBI, noise power seen at impacted subcarriers is high 


Gives wrong QAM symbol after dividing by Hk

Yk  X k H k  Z k , 1 k  N

Small in case of Deep fading Large in case of NBI

ˆ Yk Zk WRONG detection
Xk   Xk  Large in deep fading
Hˆ k Hˆ k and NBI Xˆ k  X k
Channel estimate
Bit-Interleaved Coded OFDM
• How to solve this problem?
• Answer : Channel coding
• Channel coding: Introduce redundancy of information bits
(e.g., Repetition code, Convolutional code, Turbo code)
• Then, distribute the resulting bits over subcarriers
• If a subcarrier is hit by deep fading or NBI, we lose its bits but
we use the redundancy bits distributed to other subcarriers to
correctly decode information bits

Serial to Parallel N N

Parallel to Serial

Add Cyclic Prefix


N-point IFFT
bits
Channel QAM
Interleaver
Coding modulation
Bit-Interleaved Coded OFDM
• Example, Use repetition code, every information bit is
repeated 3 times

101 111000111 101101101


Channel BPSK
Interleaver
Coding modulation

NBI

1 0 1 1 0 1 1 0 1

Subcarrier Subcarrier Subcarrier freq


1 4 6
Bit-Interleaved Coded OFDM
• Subcarriers 4-6 are hit by NBI  their data is wrongly detected
• We can still detect the original information bits after decoding
NBI

1 0 1 1 0 1 1 0 1

After detection freq

1 0 1 0 1 0 1 0 1 WRONG!
After De-Interleaving
1 0 1 0 1 0 1 0 1

After Decoding (Majority Rule)


1 0 1 CORRECT!
Interleaver design
• Note the importance of the Interleaver!
• It distribute code bits among subcarriers, so code bits
corresponding to any information bit are assigned to NON-
ADJACENT subcarriers, so if one is lost, others are not lost
• One possible Interleaver design: Matrix Interleaver
• Write code bits row by row in a matrix and read them out
column by column
c1 c2 c3 ……. c10
c1 And c2 separated
c11 c12 c13 ……. c20 by 10 bits!
c21 c22 c23 ……. c30
.
{c1 c2 c3 …… c10 {c1 c11 c21 …… c91
.
c11 c12 c13 …… c20 c2 c12 c22 …… c92
c21 c22 c23 …… c30 c91 c92 c93 ……. c100 c3 c13 c23 …… c93
….….…. .…. c100} ….….…. .…. c100}
OFDM Transmitter and Receiver

Add Cyclic Prefix


Serial to Parallel
N N

N-point IFFT

Parallel to Serial
bits
Channel QAM
Interleaver
Coding modulation

Channel is estimated Freq domain Channel equalization (1 tap/subcarrier)


using pilots

1 / Hˆ 1
.

Remove Cyclic Prefix


Serial to Parallel
N
Parallel to Serial

N-point FFT
bits
Channel De- QAM
.
decoding Interleaver demod 1 / Ĥ k
.
.
1 / Ĥ N
Peak-to-Average Power Ratio
and
Subcarrier Loading
Subcarrier Loading

• If the channel has slow variation over time (e.g., in xDSL)


• We can adapt the modulation of each subcarrier
• Subcarriers with deep fades  loaded with few bits (e.g., BPSK
or QPSK)  Low power
• Subcarriers with good channel  loaded with many bits (e.g.,
64-QAM)  High power
• We can design the loading of every subcarrier to maximize
overall throughput (bit rate)  Use waterfilling technique
Subcarrier Loading

Subcarrier
Subcarrier
With good
With bad
Channel 
Channel 
Assigned
Not assigned
many bits
Any bits (not used)
(high QAM)
 Large
power
Subcarrier Loading

• Can we use adaptive loading in Single-Carrier systems?

• No because by definition adaptive loading is used in multi-


carrier systems

• This is one advantage of OFDM over Single-Carrier


Peak-to-Average Power Ratio (PAPR)
• One disadvantage of OFDM is that it has high PAPR

• This means that the peak power is much higher than the
average power

OFDM symbol power


Peak power

Average power

time
Peak-to-Average Power Ratio (PAPR)
• OFDM has high PAPR, so what?
• After generating OFDM signal, we feed it into Power Amplifier (PA) to
amplify it
• Any PA has a linear region and a saturation (non-linear) region
• To avoid saturation and generation of non-linear components, we
need to work in the linear region
• Hence, we need to push the OFDM signal such that both its peak and
average are within the linear region
• This means that most of the time (where average power lies), we will
work in low gain  Low efficiency of PA
Peak-to-Average Power Ratio (PAPR)
Output
Linear region
signal
power
Non-linear (Saturation) region

High gain region is rarely utilized!

Input signal power

Most
Of time,
We work
In low
Gain region
Peak-to-Average Power Ratio (PAPR)
• Why does OFDM have high PAPR?

x(n)   X k exp j
N
 2kn 
, 0n N
k 1  N 
Random variable

• OFDM signal = Summation of N random variables

• Central Limit Theory  Summation of many random variables has


a Gaussian PDF regardless of PDF of every random variable

• x(n) is Gaussian distributed for large N (number of subcarriers)

• Gaussian random variables have high PAPR, why?


Peak-to-Average Power Ratio (PAPR)

Average power (Most of time)


Peak Power (rare!)

Peak power is much higher than average power for large variance
Peak power is less likely to happen than average power
Peak-to-Average Power Ratio (PAPR)

• What about PAPR of Single-Carrier systems?

• It’s actually lower than OFDM! Why?

• In single-carrier (SC), transmitted signal = modulated


QAM signal directly  No IFFT

• Hence, PAPR of SC = PAPR of QAM symbol

• Consider QPSK  What’s PAPR?


Peak-to-Average Power Ratio (PAPR)
Quadrature
s2 s1
A
d2 d1
QPSK Constellation
A
Inphase
d3 d4
PAPR of QPSK s3 s4

   2 A   2 A
1 M
1 M
1 4
Average Power  Pi  di 
2 2 2

M i 1 M i 1 4 i 1

Peak Power  Maximum Power  max Pi  max d i  2 A 2


2
i i

Peak Power 2 A2
PAPR   1
Average Power 2 A M  Number of
2

Peak Power QAM symbols


PAPR in dB  10 log 10  0 dB
Average Power
Peak-to-Average Power Ratio (PAPR)

PAPR of SC with QPSK = 0 dB  Very good for PA efficiency

If we used OFDM with QPSK, PAPR will be much higher

In terms of PAPR, Single Carrier is better than OFDM!

Exercise: Compute PAPR of Single Carrier with 16-QAM


OFDM UNDER RF IMPAIRMENTS
OFDM under RF impairments
• Effect of Carrier Frequency Offset (CFO)
• Effect of Phase Noise (PN)
• Effect of Doppler [Not RF impairment but has
similar impact]
• Effect of I/Q imbalance (IQI)
OFDM under CFO
• CFO means that carrier frequency of receiver local
oscillator is different from that of the transmitter

fc,Tx fc,Rx
π/2
π/2

CFO ≠0 due to inaccuracies


f c , Rx  f c ,Tx  f  CFO of crystal oscillators
used at Tx and Rx
OFDM under CFO
• Frequency of Crystal oscillator (XO) has manufacturing error
measured as part per million (ppm)
• ppm = error (in Hz) in every 1 million Hz (1 MHz)
• Specifications of XO tells its frequency and expected error in ppm
• Example: XO has frequency = 2 GHz with 0.1 ppm
This means we have 0.1 Hz maximum error in every 1 MHz
• XO frequency = 2 GHz = 2000 MHz
• Maximum error will be 2000*0.1 = 200 Hz
• XO frequency = 2 GHz ± 200 Hz
• We expect XO frequency to be any where in the range
2 GHz -200 Hz < XO frequency < 2 GHz + 200 Hz
• If both Tx and Rx XOs have 0.1 ppm, then maximum carrier
frequency offset (CFO) between them is when one of them is
2 GHz -200 Hz while the other is 2 GHz + 200 Hz  Δfmax= 400 Hz
OFDM under CFO
• At Rx, taking FFT of received signal is equivalent to
sampling the received signal in frequency-domain at the
locations of subcarriers as follows

Gives QAM symbol carried by red subcarrier


with ZERO contributions of other subcarriers
OFDM under CFO
• If we have CFO, then this sampling is SHIFTED

Gives QAM symbol carried by red subcarrier


with NON-ZERO contributions of other subcarriers

Inter-Carrier Interference (ICI)


OFDM under CFO
• CFO results in  Inter-Carrier Interference
• Every carrier sees interference from NEIGHBOURS
• Considering any subcarrier, it will see more interference
from close subcarriers than from far subcarriers. This is
clear from previous figure
• How to solve this problem?
• Use pilots to estimate CFO and compensate it in time-
domain BEFORE taking FFT
OFDM under CFO
x(t )  x(t ) exp j 2 f t ,  f  f c , Rx  f c , Rx

x(nTs )  x(nTs ) exp j 2 f nTs   x(n) exp  j


 2 f n  1 1
, Ts  
 N f  B N f
After taking FFT @ Rx, we get
2 f n 

1 N 1
  2kn 
Xˆ k  x ( n ) exp
  j 
 exp  j 
N
n 0  N f   N 
f
X k   x(n) exp  j
1 N 1
 2 (  k )n 
ˆ , 
N n 0  N  f
Xˆ  g X   g X
k 0 k l l
l k

  exp  j
2n 
  exp j 
N 1

g 0 large N
x(n)   X k exp j
 N  N
 2kn 
n 0

k 1  N 
OFDM under CFO
• Effect of CFO:
1. Desired symbol is rotated by phase = πα
2. ICI from neighboring subcarriers
f CFO in Hz
 
f subcarrier spacing in Hz

ICI effect Rotation effect


OFDM under CFO
• For α to be small, we need to choose XO specs
such that CFO is small compared to subcarrier
spacing
• This means that 4G LTE (subcarrier spacing =
15 kHz) will have tougher spec on XO than
WLAN (subcarrier spacing = 312.5 kHz)
• To compensate for CFO, we estimate α (and
hence δf) using pilots and multiply received
signal by exp(jδfTsn), n=0:N-1 BEFORE taking
FFT
OFDM under Phase Noise
• Phase Noise (PN) means that oscillator has a random
phase that changes with time

cos2fct tx t  π/2


π/2
cos2fct rx t 
Rx PN
Tx PN

Linewidth

Ideal XO
Practical XO
No PN
With PN
fc f fc f
OFDM under Phase Noise
• Phase Noise (PN) causes energy to leak around carrier

frequency

• Good oscillators have their energy leaking in a narrow

bandwidth (they have small linewidth)

• PLL is much better than free-running oscillators in terms of PN

• Effect of PN on OFDM (similar to CFO):

– Inter-Carrier Interference from neighboring subcarriers

– Phase rotation of every subcarrier


OFDM under Phase Noise
x(t )  x(t ) exp j (t ) , x(nTs )  x(nTs ) exp j (nTs ) 

  exp j (n)   g 0 X k   g l X l
ˆ 1 N 1
 2kn 
Xk  x(n) exp  j
N n 0  N  lk

 exp j (n)  Average PN over OFDM symbol


1 N 1
g0 
N
 
n 0

All Xˆ k will have same g 0


K
k 1

For practical oscillators with slow PN variation over time,

 g 0  exp j avg ,  avg    (n)  Average PN in OFDM symbol


1 N 1
N n0
g 0  Called ' Common Phase Error (CPE)'
Common because it' s the same for all subcarriers
OFDM under Phase Noise
• How to solve PN problem?
– Estimate PN samples and multiply received signal by
exp(jφ(n)), n=0:N-1 BEFORE taking FFT
– Difficult solution because PN change from sample to
sample and difficult to track
• Practical solution
– Use pilots to estimate average PN φavg and multiply all
subcarriers by conj(g0) = exp(jφavg) to reverse rotation
induced by PN
– This removes CPE and is called CPE compensation
– Here, we ignore ICI caused by PN
– Practical oscillators have small PN linewidth  Less
leakage  Light ICI that can be ignored
OFDM under Doppler
• High Doppler means that channel rapidly changes over
time due to Transmitter and/or Receiver movement
2fd
Transmitted Time-varying
channel Received
signal
Signal

fc f fc f Doppler spread

Doppler frequency = fd = v fc/c


c  Speed of light = 3x108 m/s
v  Relative speed between Tx and Rx

If Tx and Rx are not moving (e.g., WiFi), No Doppler spread


Doppler increases as carrier frequency and/or velocity increases
OFDM under Doppler

• 4G LTE will have mobile phones moving in high


speed
• LTE has higher Doppler than WiFi
• Doppler effect is similar to PN effect  Both
have energy leakage in frequency domain
• Effect of Doppler on OFDM:
– ICI and rotation of every subcarrier
OFDM under Doppler
• Coherence time = 1/Doppler frequency
• Coherence time  Time over which channel can be
considered static
• To avoid Doppler effect, make sure that OFDM
symbol duration is smaller than Coherence time
• What’s the Coherence time if fc = 2 GHz and v = 300
Km/hr?
• Coherence time = 1/fd = c/ (v fc) = 1.8 ms
• OFDM symbol duration in LTE = 67 us << 1.8 ms
• So, Doppler spread can be safely ignored in LTE
OFDM under I/Q imbalance
• I/Q imbalance (IQI) means  Inphase and Quadrature paths
have different gains and non-90 degrees phase differences

hI,tx(t) hI,rx(t)

1 1
εt εr
π/2+θt π/2+θr

hQ,tx(t) hQ,rx(t)

εt ≠ 1  Gain imbalance @ Tx εr ≠ 1  Gain imbalance @ Rx


θt ≠ 0  Phase imbalance @ Tx θr ≠ 0  Phase imbalance @ Rx
hI,tx(t) ≠ hQ,tx(t)  Filter imbal @ Tx hI,rx(t) ≠ hQ,rx(t)  Filter imbal @ Rx
OFDM under I/Q imbalance
• Gain and Phase imbalance  Frequency-flat (Frequency-
Independent) I/Q imbalance (FI-IQI)
• Filter imbalance  Frequency-selective (Frequency-
dependent) I/Q imbalance (FD-IQI)
• Effect of FI-IQI on baseband signal:

xFI  IQI (t )   t x(t )   t x* (t )  Effect of Tx IQI on Tx signal x(t)


y FI  IQI (t )   r y (t )   r y * (t )  Effect of Rx IQI on Rx signal y(t)

1   t exp j t 
t  ,  t  1   t* If no IQI,
2
1   r exp j r  ε = 1, θ = 0
r  ,  r  1   r*
2 α =1, β = 0
OFDM under I/Q imbalance
• FD-IQI has a similar effect except that α and β are filters
convolved with y(t) and y*(t), respectively
• Assume y(t) = exp(j2πf1t)  1 Complex sinsoid
• What’s yFI-IQI (t) ?  α exp(j2πf1t) + β exp(-j2πf1t)

Y( f ) 1 YFI-IQI( f )
With IQI α
β

0 f1 f -f1 0 f1 f

IQI creates an image of every subcarrier


OFDM under I/Q imbalance - Example
cos(2πf1t) Let f1 = 1 MHz, fc = 700 MHz
Without I/Q imbalance, RF signal x(t)
1 x(t)
Should be cos(2π(fc+f1)t), i.e.,
fc =700 MHz εt
@ 701 MHz
π/2

sin (2πf1t)

x(t) = cos(2πf1t) cos(2πfct)- εt sin(2πf1t) sin(2πfct)


= (1+ εt)/2 cos(2π(fc+f1)t) + (1- εt)/2 cos(2π(fc-f1)t)

Due to IQI, we get frequency component @ fc-f1=699 MHz in


addition to the expected 701 MHz. This 699 MHz is the image
of 701 MHz around the carrier frequency fc
OFDM under I/Q imbalance - Example
1 t 1 t
x(t )  cos(2t  701MHz)  cos(2t  699MHz)
2 2
Original Component Image Component

If εt = 1, i.e., no I/Q imbalance, we see only 701 MHz component

and there will be no frequency component @ 699 MHz

Signal analyzer Output Signal analyzer Output


(no IQI) (With IQI) 1 t
1 2
1 t
image 2

fc=700 701 f(MHz) 699 fc=700 701 f(MHz)


OFDM under I/Q imbalance
• With IQI, subcarrier k will create an image for itself
• This image interferes with subcarrier –k
• IQI causes Inter-Carrier Interference (ICI) from image
subcarriers not neighboring subcarriers as CFO, PN and
Doppler
• Signal-to-image ratio is called Image Rejection Ratio (IRR) and
is defined for a mixer as:
 Power @ original frequency
2

IRR  10 log10 
 Power @ image frequency
• Good mixers have high IRRs (IRR > 30 dB). IRR = ∞ if no IQI
exists
• Baseband processing is used to estimate and compensate IQI
to increase IRR to +45 dB
OFDM under I/Q imbalance
• Phase imbalance is given in degrees, e.g., 2o
• Gain imbalance is given in dBs or % as follows:
 Inphase Gain  1
Gain Imbalance  20Log10    20Log10   dB
 Quad Gain   

Inphase Gain - Quad Gain 1 


Gain Imbalance   100 %   100 %
Inphase Gain  Quad Gain 1 

• If ε=0.9  Gain imbalance = 0.9 dB or 5%


• If ε=1 (No imbalance)  Gain imbalance = 0 dB or 0%
• Good mixers have Gain imbalance close to 0 dB or 0%
OFDM under I/Q imbalance
• An RF mixer has Gain imbalance = 3 % and phase
imbalance = 2 degrees, what’s IRR?

• ε = 0.94 and θ = 2 *180/pi  Convert from deg to rad


  
1  0.94 exp  j 2  
r   180   0.97  j 0.016
2
 r  1   r*  0.03  j 0.016
r
2

IRR  10 log10  29 dB
r

IRR is the same if we used eqns of αt and βt instead of αr and βr


OFDM under Rx I/Q imbalance
Baseband received signal Baseband received signal
Before Mixer After IQI-impaired Mixer
1 α
0.5 0.5α

0 f 0 f

β Image conjugated
β/2

0 f
Add both waveforms
To get baseband signal under IQI
Baseband compensation of I/Q Imbalance
Use Calibration to estimate Gain and Phase imbalance
Calibration means sending single tone through the mixer and
using received samples to estimate ε and θ

Baseband I/Q compensation circuit

hI,rx(t)

1
εr
π/2+θr

hQ,rx(t) 1/εr 1/cos(θr) +


Single-Carrier Frequency-Division
Multiple-Access (SC-FDMA)
Shortcomings of OFDM
• Large PAPR
• Vulnerable to RF impairments (CFO, PN, IQI)
– These impairments cause ICI which reduces effective
SNR of each subcarrier
– ICI can cause strong carriers (with good channels) to
interfere on weak carriers (with deep fading)
==causing==> low carrier to noise ratio
– Effect of these impairments on Single-Carrier does
not include inter-symbol interference (ISI) in AWGN
channel. Only symbol rotation and/or scaling but No
ISI in AWGN channel
– Hence, Single-Carrier is more robust to these
impairments than OFDM!
Shortcomings of OFDM
• RF impairments can be compensated in baseband using fairly
low-complexity methods
• Hence, we can live with RF impairments
• However, large PAPR of OFDM is a persistent problem that
impacts PA efficiency
• High PA efficiency is a key factor for long-lasting battery in
cell phones
• 4G LTE standard adopted Single-Carrier for Uplink where cell
phone is transmitting
• OFDM is chosen for 4G LTE downlink where eNodeB (Base
station) is transmitting
• There’s no battery constraints in base stations
Single-Carrier
• Single-carrier can be implemented using OFDM platform
but with FFT introduced before IFFT to cancel its effect

Serial to Parallel

QAM
modulation FFT IFFT
Single-Carrier Frequency-Division
Multiple-Access (SC-FDMA)
• In 4G LTE uplink, multiple user equipments (UEs) (cell phones) can
transmit at the same time but need to be frequency-multiplexed

• This is achieved by making FFT and IFFT sizes different

• FFT size = M and IFFT size = N where M < N

• User 1  Take FFT output (M entries) and assign to a part of IFFT


input and pad remaining part (N-M entries) by zeros

• User 2  Assign your FFT output to the part where User 1 put
zeros and put zeros in the part where User 1 put its data

• See next slides for Two-Users example


SC-FDMA Transmitters @ Cell phones
User 1
Transmitter M

Add Cyclic prefix


Parallel to Serial
M

Serial to Parallel

N-point IFFT
QAM M-point
modulation FFT N N
M QAM
symbols
N-M zeros

User 2
Transmitter N-M zeros

Add Cyclic prefix


Parallel to Serial
Serial to Parallel

N-point IFFT
QAM M-point
N N
modulation FFT
M QAM M
symbols M
SC-FDMA Receiver at eNodeB (Base station)
To User 1

Parallel to Serial
decoder
1

M-point IFFT
QAM Select U1
demod
M Hˆ k ,U 1 M
subcarriers

User 1 Channel

Remove Cyclic prefix


estimate

Serial to Parallel
point FFT
Channel estimation

N-point
N N
User 2 Channel
To User 2 estimate
decoder
Parallel to Serial

1
M-point IFFT

QAM Select U2
demod
M
Hˆ k ,U 2 M
subcarriers
SC-FDMA
• Different users can be assigned different number of subcarrier
(M1, M2, ….)
• Subcarrier allocation  localized (adjacent) or distributed
(Interleaved) 
• eNodeB tells every user its allocation size M and location (indices)
of these subcarriers
• Difference between OFDM and SC-FDMA receivers:
– In OFDM, we get (QAM symbol x Channel after FFT), so
division by channel is Maximum-Likelihood (ML) optimal
– In SC-FDMA, we get (FFT of QAM symbols x Channel after FFT),
so division by channel is NOT ML-optimal
• Subcarriers in OFDM carriers QAM symbols, while in SC-FDMA,
they carry FFT of QAM symbols
• Cyclic prefix is used in both OFDM and SC-FDMA to prevent ICI
and maintain orthogonality among subcarriers
SC-FDMA
• SC-FDMA PAPR depends on QAM order & subcarrier
allocation  Interleaved PAPR < Localized PAPR
• In 4G LTE, no pilots are multiplexed with data subcarriers in
order to maintain low PAPR as pilots are transmitted in
higher power than data
• Instead, pilots are transmitted alone in a separate symbol
every 6 data symbols
• In WiFi, OFDMA is used at both Tx and Rx. No high PA
efficiency is needed because
– Tx-Rx distance in WiFi < Tx-Rx distance in LTE
– More restrictions on output power in WiFi than in LTE because WiFi
transmits in unlicensed bands  To limit interference to others
– Hence, WiFi PA gain < LTE PA gain  No high efficiency needed in
WiFi
References (1)
1) J. A. C. Bingham, "Multicarrier modulation for data transmission: an idea whose time
has come," in IEEE Communications Magazine, vol. 28, no. 5, pp. 5-14, May 1990
2) Nee, Richard van, and Ramjee Prasad. OFDM for wireless multimedia communications.
Artech House, Inc., 2000.
3) Heiskala, Juha, and John Terry Ph D. OFDM wireless LANs: A theoretical and practical
guide. Sams, 2001.
4) Van de Beek, J-J., et al. "On channel estimation in OFDM systems." Vehicular
Technology Conference, 1995 IEEE 45th. Vol. 2. IEEE, 1995.
5) Wong, Cheong Yui, et al. "Multiuser OFDM with adaptive subcarrier, bit, and power
allocation." IEEE Journal on selected areas in communications 17.10 (1999): 1747-
1758.
6) Rhee, Wonjong, and John M. Cioffi. "Increase in capacity of multiuser OFDM system
using dynamic subchannel allocation." Vehicular Technology Conference Proceedings,
2000. VTC 2000-Spring Tokyo. 2000 IEEE 51st. Vol. 2. IEEE, 2000.
7) Armstrong, Jean. "Peak-to-average power reduction for OFDM by repeated clipping
and frequency domain filtering." Electronics letters 38.5 (2002): 1.
8) Cimini, Leonard J., and Nelson R. Sollenberger. "Peak-to-average power ratio reduction
of an OFDM signal using partial transmit sequences." IEEE Communications letters 4.3
(2000): 86-88.
9) Pollet, Thierry, Mark Van Bladel, and Marc Moeneclaey. "BER sensitivity of OFDM
systems to carrier frequency offset and Wiener phase noise." IEEE transactions on
communications 43.2/3/4 (1995): 191-193.
References (2)
10) Jang, Jiho, and Kwang Bok Lee. "Transmit power adaptation for multiuser OFDM
systems." IEEE Journal on selected areas in communications 21.2 (2003): 171-178.
11) Zhao, Yuping, and S-G. Haggman. "Intercarrier interference self-cancellation scheme
for OFDM mobile communication systems." IEEE Transactions on
Communications 49.7 (2001): 1185-1191.
12) Li, Ye, and L. J. Cimini. "Bounds on the interchannel interference of OFDM in time-
varying impairments." IEEE Transactions on Communications 49.3 (2001): 401-404.
13) Morelli, Michele, and Umberto Mengali. "A comparison of pilot-aided channel
estimation methods for OFDM systems." IEEE Transactions on signal processing 49.12
(2001): 3065-3073.
14) Jiang, Tao, and Yiyan Wu. "An overview: peak-to-average power ratio reduction
techniques for OFDM signals." IEEE Transactions on broadcasting54.2 (2008): 257.
15) Bahai, Ahmad RS, Burton R. Saltzberg, and Mustafa Ergen. Multi-carrier digital
communications: theory and applications of OFDM. Springer Science & Business
Media, 2004.
16) Heiskala, Juha, and John Terry Ph D. OFDM wireless LANs: A theoretical and practical
guide. Sams, 2001.
17) Morelli, Michele, and Umberto Mengali. "An improved frequency offset estimator for
OFDM applications." Communication Theory Mini-Conference, 1999. IEEE, 1999.
18) Falconer, David, et al. "Frequency domain equalization for single-carrier broadband
wireless systems." IEEE Communications Magazine 40.4 (2002): 58-66.
19) Horlin, François, and André Bourdoux. Digital compensation for analog front-ends: a
new approach to wireless transceiver design. John Wiley & Sons, 2008.
References (3)
20) Schenk, Tim. RF imperfections in high-rate wireless systems: impact and digital
compensation. Springer Science & Business Media, 2008.
21) Holma, Harri, and Antti Toskala, eds. LTE for UMTS-OFDMA and SC-FDMA based radio
access. John Wiley & Sons, 2009.
22) Berardinelli, Gilberto, et al. "OFDMA vs. SC-FDMA: performance comparison in local
area IMT-A scenarios." IEEE Wireless Communications 15.5 (2008): 64-72.
23) Myung, Hyung G., Junsung Lim, and David J. Goodman. "Single carrier FDMA for uplink
wireless transmission." IEEE Vehicular Technology Magazine 1.3 (2006): 30-38.
24) Myung, Hyung G., Junsung Lim, and David J. Goodman. "Peak-to-average power ratio of
single carrier FDMA signals with pulse shaping." 2006 IEEE 17th International
Symposium on Personal, Indoor and Mobile Radio Communications. IEEE, 2006.
25) Huang, Gillian, Andrew Nix, and Simon Armour. "Decision feedback equalization in SC-
FDMA." 2008 IEEE 19th International Symposium on Personal, Indoor and Mobile Radio
Communications. IEEE, 2008.
26) Zhang, Chao, et al. "Frequency domain decision feedback equalization for uplink SC-
FDMA." IEEE transactions on broadcasting 56.2 (2010): 253-257.
27) Gomaa, Ahmad, and Naofal Al-Dhahir. "A sparsity-aware approach for NBI estimation in
MIMO-OFDM." IEEE Transactions on Wireless Communications10.6 (2011): 1854-1862.
References (4)
28) Gomaa, Ahmad, and Naofal Al-Dhahir, "SC-FDMA Performance in Presence of
Oscillator Impairments: EVM and Subcarrier Mapping Impact," Global
Telecommunications Conference (GLOBECOM 2011), 2011 IEEE, Houston, TX,
USA, 2011, pp. 1-5.
29) Gomaa, Ahmad, and Naofal Al-Dhahir. "Multi-user SC-FDMA systems under IQ
imbalance: EVM and subcarrier mapping impact." Global Telecommunications
Conference (GLOBECOM 2011), 2011 IEEE. IEEE, 2011.
30) Gomaa, Ahmad, and Naofal Al-Dhahir. "Phase noise in asynchronous SC-FDMA
systems: Performance analysis and data-aided compensation." IEEE
Transactions on Vehicular Technology 63.6 (2014): 2642-2652.
31) Gomaa, Ahmad, and Louay MA Jalloul. "Data-aided I/Q imbalance estimation
and compensation in OFDM systems." IEEE Communications Letters 18.3
(2014): 459-462.
32) Gomaa, Ahmad, and Naofal Al-Dhahir. "Blind phase noise compensation for SC-
FDMA with application to LTE-uplink." 2012 IEEE International Conference on
Communications (ICC). 2012.
33) Gomaa, Ahmad, Ayman Elezabi, and Mohamed Eissa. "A Subspace Method for
I/Q Imbalance Estimation in Low-IF Receivers." arXiv preprint
arXiv:1411.2199 (2014).
34) Gomaa, Ahmad, and Louay Jalloul. "Receiver Architecture for Frequency Offset
Correction and I/Q Imbalance Compensation in Equal Bandwidth Contiguous
Carrier Aggregation."
References (5)
35) Mokhtar, Mohamed, Ahmad Gomaa, and Naofal Al-Dhahir. "OFDM AF relaying
under I/Q imbalance: Performance analysis and baseband compensation." IEEE
Transactions on Communications 61.4 (2013): 1304-1313.
36) ElSamadouny, Ahmed, Ahmad Gomaa, and Naofal Al-Dhahir. "Likelihood-based
spectrum sensing of OFDM signals in the presence of Tx/Rx I/Q
imbalance." Global Comm. Conference (GLOBECOM), 2012 IEEE. IEEE, 2012.
37) Priyanto, Basuki E., et al. "Initial performance evaluation of DFT-spread OFDM
based SC-FDMA for UTRA LTE uplink." 2007 IEEE 65th Vehicular Technology
Conference-VTC2007-Spring. IEEE, 2007.
38) Raghunath, K., and A. Chockalingam. "SC-FDMA versus OFDMA: Sensitivity to
large carrier frequency and timing offsets on the uplink." Global
Telecommunications Conference, 2009. GLOBECOM 2009. IEEE. IEEE, 2009.
39) Berardinelli, Gilberto, et al. "Improving SC-FDMA performance by turbo
equalization in UTRA LTE uplink." Vehicular Technology Conference, 2008. VTC
Spring 2008. IEEE. IEEE, 2008.
References (6)
40) Yoshida, Yuki, et al. "Analysis and compensation of transmitter IQ
imbalances in OFDMA and SC-FDMA systems." IEEE Trans on
Signal Processing 57.8 (2009): 3119-3129.
41) Tarighat, Alireza, Rahim Bagheri, and Ali H. Sayed. "Compensation
schemes and performance analysis of IQ imbalances in OFDM
receivers." IEEE Transactions on Signal Processing 53.8 (2005):
3257-3268.
42) Tarighat, Alireza, and Ali H. Sayed. "Joint compensation of
transmitter and receiver impairments in OFDM systems." IEEE
Transactions on Wireless Communications 6.1 (2007): 240-247.
43) Zou, Qiyue, Alireza Tarighat, and Ali H. Sayed. "Compensation of
phase noise in OFDM wireless systems." IEEE Transactions on
Signal Processing 55.11 (2007): 5407-5424.
Thank You!

You might also like