Orthogonal Frequency

Division Multiplexing (OFDM):

A Primer
Dr. Miguel Rodrigues
Laboratory for Communication Engineering
Department of Engineering, University of Cambridge
E-mail: mrdr3@eng.cam.ac.uk
 Agenda
ƒ The wireless propagation channel: A case for OFDM?

ƒ Elements of a typical OFDM communications system

ƒ OFDM generation/detection

ƒ Coding/decoding and mapping/demapping techniques

ƒ Channel estimation and synchronisation operations

ƒ Advantages and disadvantages of OFDM

ƒ Applications of OFDM

ƒ Recent developments
The Wireless Propagation Channel
Transmitted Signal Received Signal

Dispersion in time
t0 t t0+τ0 t0+τ1 t0+τ2 t0+τ3 t
The Wireless Propagation Channel
Transmitted Signal Received Signal

Dispersion in frequency
fc f fc+cos(θ)v/λ f

v
 The Wireless Propagation Channel
Transmitted Signal 1 Transmitted Signal 2
f1=1/T’=1/(2T’’) f2=1/(2T’)=1/(4T’’)

... ... ... ...

t t

Two-
Two-path channel at time t1 Received Signal 1 at t1 Received Signal 2 at t1

Time Selective Fading

T’ T’
path 2

Transmitter path 1 Receiver

... ... ... ...
Two-path channel
t t
relative delay = T’

Two-
Two-path channel at time t2 Received Signal 1 at t2 Received Signal 2 at t2

path 2 T’’ T’’

Transmitter path 1 Receiver

... ... ... ...
Two-path channel
t t
relative delay = T’’=T’/2

Frequency Selective Fading

 The Scattering Function
ƒ The scattering function measures the power spectrum
of the channel at delay τ and frequency offset λ with
respect to the carrier frequency.
Typical Scattering Function
 The Delay Power Spectrum
and Delay Spread

ƒ The delay power spectrum measures the power of the

channel at delay τ.

ƒ The range of values τ over which the delay power

spectrum is essentially non-zero is defined as the
delay spread Tm.

ƒ The channel coherence bandwidth Bcoh is the inverse

of the delay spread Tm, i.e., Bcoh=1/Tm.

ƒ The channel coherence bandwidth measures the width

of the band of frequencies over which the fading is
highly correlated.
Typical Delay Power Spectrum
Typical Delay Power Spectrum
 The Doppler Power Spectrum
and Doppler Spread

ƒ The Doppler power spectrum measures the power of

the channel at frequency offset λ.

ƒ The range of values λ over which the Doppler power

spectrum is essentially non-zero is defined as the
Doppler spread Bd.

ƒ The channel coherence time Tcoh is the inverse of the

Doppler spread Bd, i.e., Tcoh=1/Bd.

ƒ The channel coherence bandwidth measures the width

of the interval of time over which the fading is highly
correlated.
Typical Doppler Power Spectrum
Typical Doppler Power Spectrum
 Fading Statistics
ƒ The fading characteristics of the wireless channel are
described by probability distributions.

ƒ Assuming that there are a large number of

scatterers/reflectors in the medium, application of the
central limit theorem leads to a complex Gaussian
distribution model for the channel response.

ƒ If there is no dominant scatterer in the medium (e.g.

only NLOS components), the envelope of the channel
response is described by the Rayleigh distribution.

ƒ If there is a dominant scatterer in the medium (e.g.

an LOS component), the envelope of the channel
response is described by the Ricean distribution.
 Channel Models: Classification
and Characteristics
Channel Models Classification and Characteristics

Channel
Models

Model based on Model based on

delay spread Doppler spread

Frequency Non- Frequency

Selective Fading Selective Fading
Fast Fading Slow Fading

W<B
W<Bcoh or T>Tm W>B
W>Bcoh or T<Tm High Bd or T>T
T>Tcoh Low Bd or T<T
T<Tcoh
The Frequency Non-Selective,
Slow Rayleigh Fading Channel
Pb vs. SNR for a BPSK signal in an AWGN
channel and in a Rayleigh fading channel
 Techniques to Overcome Time
and Frequency Selectivity
Techniques to Overcome Time and/or Frequency Selectivity

Techniques

Coding and Spread Multi

Diversity Equalisation
Interleaving Spectrum Carrier

Time Frequency Space Linear Nonlinear

Selection Equal Gain Max Ratio

Combining Combining Combining DFE MLSE
Diversity, Coding and Interleaving
ƒ Diversity techniques are often used to mitigate the
the effects of the wireless channel.
ƒ The principle of diversity is to transmit the same
information on two or more sub-channels that fade
independently from one another.
ƒ Independently fading sub-channels are realised either
in the time, frequency or space domains.
ƒ Selection combining, equal gain combining or maximal
ratio combining techniques recombine the information
in the various sub-channels at the receiver.
ƒ Coding and interleaving are also often used to
mitigate the effects of the wireless channel. Coding
and interleaving are an efficient form of (time)
diversity.
 Equalisation
ƒ Equalisation is generally used to mitigate intersymbol
interference in wireless channels.

ƒ Linear equalisers are less complex than non-linear

equalisers, but generally produce noise enhancement
in severely distorted channels.

ƒ Non-linear equalisers perform better than non-linear

equalisers, but are generally more complex.

ƒ Maximum likelihood sequence estimation (MLSE)

equalisers are optimum, but highly complex.

ƒ Decision feedback equalisers (DFE) are sub-optimum,

but less complex.
 Spread-Spectrum Techniques
ƒ In spread spectrum modulation the transmitted spectrum is
spread over a range much greater than the message
bandwidth.

ƒ In direct sequence spread spectrum (DS-SS) the transmitted

spectrum is spread by multiplying the signal by a wide-band
pseudo-noise (PN) sequence.

ƒ In frequency hopped spread spectrum (FH-SS) the

transmitted spectrum is spread by modulating the signal onto
a wide-band series of frequencies generated by a frequency
synthesiser driven by a pseudo-noise (PN) sequence.

ƒ The ratio of the transmitted signal spectrum to the message

spectrum is the known as the bandwidth expansion factor or
the processing gain.
Direct Sequence Spread Spectrum
DS-SS Transmitter

Data Modulator DS-SS Signal

PN Generator
Carrier

DS-SS Receiver

DS-SS Signal Demodulator LPF Data

PN Generator
Carrier
 DS-SS Transmitter Operation
Waveforms Spectra
≈1/T
... ... Data

0 T 2T
t
0
≈1/Tc
... ... PN sequence

0 Tc T 2T t
0

... ... Spread data ≈1/Tc

0 Tc T 2T t
0
... ... DS-
DS-SS signal ≈1/Tc

0 2T t
Tc T fc
 DS-SS Receiver Operation
Waveforms Spectra
≈1/Tc
... ... DS-
DS-SS signal

0 2T t
fc
Tc T
≈1/Tc
... ... Spread data

0 Tc T 2T t
0

... ... ≈1/Tc

PN sequence

0 Tc T 2T t
0

... ... Data

≈1/T

0 T 2T
t
0
Frequency Hopped Spread Spectrum
FH-SS Transmitter

Data Modulator BPF FH-SS Signal

Frequency
2k frequencies
Synthesiser
Carrier 1 ... k

PN Generator

FH-SS Receiver

FH-SS Signal BPF Demodulator Data

Frequency
2k frequencies
Synthesiser
1 ... k
Carrier

PN Generator
 Spread Spectrum Techniques
on Multipath Channels
path 2
Two-path channel path 1
Transmitter Receiver
relative delay = T’
ƒ Time domain interpretation
path 1 path 1
... ... ... ...
0 T 2T 0 2T
Path 2 contributes Path 2 contributes
T
path 2 with equal power path 2 with lower power
... ... ... ...

0 T’ T+T’ 2T+T’ 0 T’ 2T+T’

T+T’
Ordinary modulation case Spread Spectrum modulation case

ƒ Frequency domain interpretation

Two-path channel Two-path channel
frequency response frequency response

Significant Negligible
fc distortion fc distortion
Ordinary modulation case Spread Spectrum modulation case
 Spread Spectrum Techniques
on Multipath Channels
Spread spectrum modulation with Rake receiver

weight 1

DS-SS Signal Demodulator LPF Data

weight 2

.
..

...
Carrier

Information in multipath
components is collected weight M
...

Delay Delay
1 2
... Delay
M
PN Generator
DIVERSITY M
 Multi-Carrier Techniques
ƒ In a single carrier modulation scheme each data symbol
is transmitted sequentially on a single carrier ⇒
signalling interval equal to data symbol duration.

ƒ In a single carrier modulation scheme the modulated

carrier occupies the entire available bandwidth.

ƒ In a multi-carrier modulation scheme N sequential data

symbols are transmitted simultaneously on N multiple
carriers ⇒ signalling interval equal to N times data
symbol duration.

ƒ In a multi-carrier modulation scheme each modulated

carrier occupies only a small part of the entire available
bandwidth.
 Multi-Carrier Techniques
Single carrier ≈1/TSC

0 TSC 2TSC 3TSC 4TSC

fc
Carrier
Spectrum
TSC: single carrier signal symbol duration

Multi-carrier Carrier 1
≈1/TMC ≈1/TMC ≈1/TMC

Carrier 2
...

s/p ..
0 TSC 2TSC 3TSC 4TSC .
Carrier N fc1 fc2 fcN
Spectrum
TMC=NTSC: multi-carrier signal symbol duration
 Multi-Carrier Techniques
on Multipath Channels
path 2
Two-path channel path 1
Transmitter Receiver
relative delay = T’
ƒ Time domain interpretation
path 1 path 1
... ... ... ...
0 TSC 2TSC 3TSC 4TSC 0 TMC 2TMC
Significant Negligible
path 2 ISI path 2 ISI
... ... ... ...
0 T’ TMC +T’ 2TMC +T’
0 T’ TSC+T’ 2TSC+T’ 3TSC +T’ 4TSC +T’

Single-
Single-carrier case Multi-
Multi-carrier case

ƒ Frequency domain interpretation

Two-path channel Two-path channel
frequency response frequency response
...

Significant Negligible
fc distortion fc1 fc2 fcN distortion
Single-
Single-carrier case Multi-
Multi-carrier case
 Multi-Carrier Techniques
on Multipath Channels
Multi-carrier modulation with coding and interleaving

frequency

Information conveyed in different

time and frequency slots is linked

DIVERSITY

TMC Tcoh
time
∆f Frequency separation greater than Bcoh

Bcoh Time separation greater than Tcoh

Orthogonal Frequency Division
Multiplexing
ƒ OFDM is a multi-carrier modulation scheme.

ƒ In OFDM the frequency spacing between adjacent

sub-carriers is ∆f= 1/TMC=1/(NTSC).

ƒ ∆f=1/TMC is the minimum frequency separation that is

necessary to ensure orthogonality between the sub-
carriers over the signalling interval of length TMC.

ƒ In OFDM the frequency spectrum of each sub-carrier

overlaps the frequency spectrum of adjacent sub-
carriers.
 Orthogonal Frequency Division
Multiplexing
OFDM Transmitter

Converter

Converter

Converter
Converter

IFFT

D/A
P/S
S/P

Serial Data OFDM signal

Up
...

...
Input

OFDM Receiver
Converter

Converter
Converter
Converter
Down

A/D

FFT

P/S
S/P

OFDM signal Serial Data

...

...
Output
Orthogonal Frequency Division
Multiplexing
ƒ Advantages of OFDM:
ƒ Good performance under delay spread/frequency
selective fading conditions;
ƒ Bandwidth efficiency;
ƒ Efficient digital signal processor based generation/
detection techniques.
ƒ Disadvantages of OFDM:
ƒ Poor performance under Doppler spread/time selective
fading conditions;
ƒ Sensitive to non-linear distortion;
ƒ Sensitive to timing and frequency offsets as well as
phase noise.
Binary Output

Binary Input
Decoder

Coder

Deinterleaver

Interleaver

Demapper

Mapper
Channel
Estimation
Pilot Insertion
P/S
Converter
... S/P
Converter
...
FFT
... IFFT

S/P Converter ...

OFDM Receiver
OFDM Transmitter

P/S Converter
Remove Cyclic
Prefix
Insert Cyclic
Prefix/Window
Time/Frequency
Synchronisation
D/A Converter

A/D Converter
Communications System

RF Transmitter
Elements of a Typical OFDM

RF Receiver
To Channel

From Channel
 Oscillator Based OFDM
Generation
Oscillator Based OFDM Generation

1/√T·e-j2π(N/2)t/T

S0

..
.
Equation:
Serial-to-Parallel

Up--Converter
1/√T
Modulation

Converter
Symbols

SN/2 s(t)

..

To Up
.
1/√T·ej2π(N/2-1)t/T
SN-1

N is the number of sub-carriers

T is duration of signalling interval
Oscillator Based OFDM Generation

ƒ In OFDM the frequency separation between adjacent

sub-carriers is 1/T.

ƒ This is the minimum frequency separation between

adjacent sub-carriers necessary to achieve
orthogonality and hence detectability.

ƒ In OFDM the frequency spectrum of each sub-carrier

overlaps the frequency spectrum of adjacent sub-
carriers.
Oscillator Based OFDM Generation

Spectrum of an OFDM Symbol

1/T 1/T 1/T

Spectrum of an OFDM symbol

consists of several sinc functions

The maximum of each sinc

function corresponds to the zero-
zero-
crossings of all the others

f
 Oscillator Based OFDM
Detection
Oscillator Based OFDM Detection

1/√T·ej2π(N/2)t/T

R0
Integrator
..
Down--Converter

.
Equation:

Parallel-to-Serial
1/√T

Modulation
Converter

Symbols
r(t) RN/2
Integrator
From Down

..
.
1/√T·e-j2π(N/2-1)t/T

RN-1
Integrator
Oscillator Based OFDM Detection

ƒ The operation of an OFDM receiver can be viewed from two

different perspectives.

ƒ From one view point, the receiver correlates the OFDM

symbol with a local version of each complex sub-carrier.

ƒ Orthogonality implies that the correlation between any two

different complex sub-carriers is zero but when the complex
sub-carriers have the same frequency.

ƒ Therefore, the correlation operation yields the information

conveyed by each sub-carrier without ICI.
Oscillator Based OFDM Detection
ƒ From another view point, the receiver evaluates the Fourier
transform of an OFDM symbol at specific frequencies.

ƒ The spectrum of an OFDM symbol consists of a series of

sinc() functions where the maximum of each function
corresponds to the zero-crossings of all the other sinc()
functions.

ƒ Or, the spectrum of an OFDM symbol fulfils Nyquist’s

criterion for an ISI free pulse shape, where in the OFDM
case the pulse shape is present in the frequency domain
instead of the time domain.

ƒ Therefore, the Fourier transformation operation yields the

information conveyed by each sub-carrier without ICI.
Oscillator Based OFDM Detection

Information Recovery

R0 R1 R2 RN-3 RN-2 RN-1

..
.
..
.

-(N/2)/T -(N/2-1)/T -(N/2-2)/T (N/2-3)/T (N/2-2)/T (N/2-1)/T

f
 FFT Based OFDM
Generation/Detection
ƒ OFDM generation/detection is achieved using the IDFT and
the DFT, or, the IFFT and the FFT, respectively.

ƒ The scaled samples sk=√(T/N)·s(kT/N),k=0,…,N-1, of the

transmit OFDM symbol are generated by taking the IDFT of
the “re-ordered” modulation symbols,

N −1 2πkn
1 j
sk =
N
∑ S(n+N 2) N+
e N
, k = 0,1,..., N − 1
n =0

ƒ The “re-ordered” modulation symbols are detected by

taking the DFT of the scaled samples
rk=√(T/N)·r(kT/N),k=0,…,N-1, of the receive OFDM symbol,
N −1 2πkn
1 −j
R(n+ N 2 )N =
N
∑ rk e N
, n = 0,1,..., N − 1
k =0
 Modulation
From Down
Down--Converter
Symbols

D/A Converter S/P Converter

.
..
.
..

S1
S0

SN/2-
SN-1
SN/2

N/2-1
SN/2+1

S/P Converter
IFFT

.
..
.
..
.
..
.
..

IFFT
P/S Converter

.
..
.
..

R1
R0

RN/2-
RN-1
RN/2

N/2-1
RN/2+1

P/S Converter
FFT Based OFDM Detection

D/A Converter
FFT Based OFDM Generation

Modulation
Symbols To Up
Up--Converter
FFT Based OFDM Transceiver
 Zero Padding
ƒ With an N point IDFT, it is only possible to generate
an N sub-channel OFDM symbol sampled at the
Nyquist rate.
ƒ In this case, it is difficult to recover the continuous
time signal from the sampled signal with filters with
realisable passband-to-stopband transition regions.
ƒ With an N’>N point IDFT, it is possible to generate an
N sub-channel OFDM symbol sampled at a rate higher
than the Nyquist rate.
ƒ In this case, it is easier to recover the continuous time
signal from the oversampled signal using filters with
realisable passband-to-stopband transition regions.
ƒ The zero padding technique achieves oversampling.
 Zero Padding
Spectrum of a Non-Oversampled IFFT Output
SN/2
Signal is not
SN/2+1 filter
.. .. recovered
. .
SN-1
IFFT

… …
N

S0
S1
.. ..
. . -2N/T -N/T 0 N/T 2N/T f
SN/2-
N/2-1

N/T

Spectrum of an Oversampled IFFT Output

SN/2
SN/2+1
.. .. filter Signal is recovered
. .
SN-1
0
IFFT

..
N’--N

… …
N’

.
N’

0
S0
S1 -N’/T 0 N’/T f
.. ..
. .
SN/2-
N/2-1 N/T
 Cyclic Prefix
ƒ To eliminate ISI in OFDM a guard time is inserted with a
duration longer than the multipath channel maximum
delay.
ƒ Moreover, to eliminate ICI in OFDM the guard time is
cyclically extended.
ƒ Note that in a multipath channel an appropriate guard
time avoids ISI but not ICI, unless it is cyclically
extended.

Cyclic Extension of an OFDM Symbol

copy

t
TCP TCP

T
 Cyclic Prefix

OFDM Signal with “Empty” Guard Time

Two-path channel
relative delay = T’
path 2
path 1
Transmitter Receiver
ISI is eliminated but ICI is not

Path 1 T+TCP
Path 2

T’ TCP T
 Cyclic Prefix

OFDM Signal with Cyclic Extended Guard Time

Two-path channel
relative delay = T’
path 2
path 1
Transmitter Receiver
Both ISI and ICI are eliminated

Path 1 T+TCP
Path 2

T’ TCP T
 Cyclic Prefix
ƒ With a cyclic prefix, the received symbol in frame k
and sub-channel n, Rk,n, is related to the transmitted
symbol in the same frame and sub-channel, Sk,n, by

Rk,n=Hk,nSk,n+Nk,n

where Hk,n is the channel frequency response in frame

k and sub-channel n and Nk,n is the noise.

ƒ Insertion of an appropriate cyclic prefix eliminates ISI

and ICI in a multipath channel but it also introduces a
loss in the SNR and data rate.

ƒ The SNR loss is equal to 10×log10(1+TCP/T) (dB)

ƒ The data rate loss is equal to (1+TCP/T)-1 (×100%)

 Windowing
OFDM Signal with Rectangular Window

Sharp phase transitions High out- of-band radiation … Delay spread tolerance=TCP
out-of-
TCP T

OFDM Signal with Raised Cosine Window

Smooth phase transitions Low out- of-band radiation … Delay spread tolerance=TCP-βT
out-of-
TCP T

t
βT βT
 Windowing
Power Density Spectrum of
“Windowed” OFDM Signals

20
β=0
β = 0.025
β = 0.05
β = 0.1
0

−20
PDS (dBr)

−40

−60

−80

−100
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
(f−fcentre)/B3dB
 Coding and Interleaving
Techniques
ƒ In an OFDM system, the received symbol in frame k
and sub-channel n, Rk,n, is related to the transmitted
symbol in the same frame and sub-channel, Sk,n, by
Rk,n=Hk,nSk,n+Nk,n

where Hk,n are the channel transfer factors and Nk,n is

the noise.
ƒ Occasional deep fades in the channel cause groups of
adjacent frames/sub-channels to be less reliable than
other groups and hence errors to occur in burst rather
than independently.
ƒ Coding and interleaving provides a link between
independently fading frames/sub-channels, so that
strongly received ones correct for weakly received
ones.
 Coding and Interleaving
Techniques
OFDM with coding and interleaving

frequency

Information conveyed in different

time and frequency slots is linked

DIVERSITY

T Tcoh
time
1/Τ
Frequency separation greater than Bcoh

Bcoh Time separation greater than Tcoh

 Coding Methods
ƒ A number of coding methods have been proposed for
OFDM systems including block, convolutional,
concatenated and turbo coding.

ƒ Trellis-coded modulation methods have also been

proposed for OFDM systems, where the coding and
modulation operations are merged together.

ƒ Hard and soft decision decoding techniques can be

used.

ƒ Hard decision decoding performs worse than soft

decision decoding. However, the former is less
complex than the later.
 Interleaving Methods
ƒ A number of interleaving methods have been
proposed for OFDM systems including block and
convolutional interleaving.

ƒ In block interleaving, the bits/symbols are written into

a matrix column by column.

ƒ These are subsequently read out from the matrix row

by row to produce the interleaved bits/symbols.

ƒ In convolutional interleaving, the bits/symbols are

cyclically written into one of K shift registers that
introduce a delay of 0 to K-1.

ƒ These are subsequently read out cyclically to produce

the interleaved bits/symbols.
Mapping/Demapping Techniques
ƒ At the transmitter, the bits at the output of the
encoder and interleaver are mapped to modulation
symbols, which will constitute the input to the IFFT.
ƒ The transmitter uses either non-differential or
differential encoding to map bits to modulation
symbols.
ƒ At the receiver, the modulation symbols at the output
of the FFT are demapped to bits or “soft” bits, which
will constitute the input to the deinterleaver and the
decoder.
ƒ The receiver uses coherent of differential detection to
demap modulation symbols to bits/soft bits,
depending on the mapping scheme used at the
transmitter.
 Non-Differential Encoding
ƒ For mappings with no differential encoding, the encoded and
interleaved bits are directly mapped to modulation symbols
Sk,n.

ƒ Examples of this mapping technique include M-ary phase

shift keying (M-PSK) or M-ary quadrature amplitude
modulation (M-QAM).

No Differential Encoding
Gray Coded
QPSK Mapping
Q

P/S Converter
Binary Input

Interleaver

(01) ej3π
j3π/4 ejπ/4 (11)
Converter
Mapper

j5π/4 ejπ/4 j7π

ej5π
Coder

… 00 1110 … … ej7π/4…
IFFT …
S/P

...

...
I

(00) e j5π
j5π/4 ej7π
j7π/4 (10)
 Differential Encoding
ƒ For mappings with differential encoding, the encoded
and interleaved bits are mapped to the quotient Bk,n of
two successive modulation symbols.

ƒ If performed in the time direction, the modulation

symbol Sk,n=Sk-1,nBk,n., and each sub-carrier of the first
OFDM symbol conveys a known/reference value.

ƒ If performed in the frequency direction, the modulation

symbol Sk,n=Sk,n-1Bk,n, and the first sub-carrier of each
OFDM symbol conveys the known/reference value.

ƒ Examples of this mapping technique include M-ary

differential phase shift keying (M-DPSK), where
Bk,n∈{ej2πm/M; m=0,…,M-1}.
 Differential Encoding
Differential Encoding in Time
DQPSK Mapping
Q

P/S Converter
N symbols
Binary Input

ejπ/2 (01)
Interleaver

Converter
Mapper
… ejπ/2ej0… ejπ/2 ejπ/2…
Coder

… 1110 …

IFFT
… (00) ejπ ej0 (11)

S/P

...

...
I
N symbols
ej3π
j3π/2 (10)

Differential Encoding in Frequency

DQPSK Mapping
Q

P/S Converter
Binary Input

ejπ/2 (01)
Interleaver

Converter
Mapper

… ej0 ej0 ejπ/2…

Coder

… 1110 …
IFFT … (00) ejπ ej0 (11)
S/P

...

...
I

ej3π
j3π/2 (10)
 Coherent Detection
ƒ For mappings with no differential encoding coherent
detection is used at the receiver, whereby the decision
is based on the quotient Dk,n given by

Dk,n=Rk,n/Ĥk,n=(Hk,nSk,n+Nk,n)/Ĥk,n=Sk,n+Nk,n/Ĥk,n

where Ĥk,n is an estimate of the channel transfer

factor Hk,n.

ƒ Note that in OFDM systems an equaliser corresponds

to a bank of complex multipliers.

ƒ The principal advantage of OFDM systems follows

from this simple equalisation operation.
 Coherent Detection
Coherent Detection Followed by Hard Decoding Q

(01) ej3π
j3π/4 ejπ/4 (11)

S/P Converter
Binary Output

Deinterleaver

noise?

Estimation
Demapper

Converter
Channel
Decoder

noise? Dk,n=Rk,n/Ĥk,n
… 11 … … Rk,n …
…

FFT
P/S

...

...
I
noise?
noise?
(00) ej5π
j5π/4 ej7π
j7π/4 (10)

QPSK
Channel Estimates Demapping

Coherent Detection Followed by Soft Decoding Q

|N1|2 (01) ej3π

j3π/4 ejπ/4 (11)

S/P Converter
Binary Output

Deinterleaver

|N2|2 N2 N1
Estimation
Demapper

Converter
|N3|2
Channel
Decoder

|N4|2 Dk,n=Rk,n/Ĥk,n
… … … Rk,n …
FFT …
P/S

...

...
N3 I
N4

(00) ej5π
j5π/4 ej7π
j7π/4 (10)

QPSK
Channel Estimates Demapping
 Differential Detection
ƒ For mappings with differential encoding differential
detection is used at the receiver.

ƒ If differential encoding is performed in the time

direction, the decision is based on the quotient:
Dk,n=Rk,n/Rk-1,n=(Sk-1,nBk,nHk,n+Nk,n)/(Sk-1,nHk-1,n+Nk-1,n)

ƒ If differential encoding is performed in the frequency

direction, the decision is based on the quotient:
Dk,n=Rk,n/Rk,n-
k,n-1=(Sk,n-
k,n-1Bk,nHk,n+Nk,n)/(Sk,n-
k,n-1Hk,n-
k,n-1+Nk,n-
k,n-1)

ƒ In the absence of noise, the symbol containing the

information is recovered provided that Hk,n≈Hk-1,n or
Hk,n≈Hk,n-1, i.e. signalling interval is smaller than Tcoh
or frequency separation is smaller than Bcoh.
 Differential Detection
Differential Detection (in the Time Direction) Q
Rk,n-
k,n-1 ejπ/2 (01)

Rk,n

S/P Converter
Binary Output

Deinterleaver

Demapper

Converter
Dk,n=Rk,n/Rk,n-
Decoder

k,n-1
… 11 … … Rk-1,n … Rk,n …
…

FFT
P/S
(00) ejπ ej0 (11)

...

...
I
N symbols

ej3π
j3π/2 (10)

DQPSK
Demapping

Differential Detection (in the Time Direction) Q

Rk,n-
k,n-1 ejπ/2 (01)

Rk,n

S/P Converter
Binary Output

Deinterleaver

Demapper

Converter

Dk,n=Rk,n/Rk,n-
Decoder

k,n-1

… 11 … … Rk,n- Rk,n … (00) ejπ ej0 (11)

…
FFT
P/S

k,n-1
...

...

ej3π
j3π/2 (10)

DQPSK
Demapping
 Merits/Demerits of Differential
and Non-Differential Schemes

ƒ Differential schemes require less complex receivers

than non-differential schemes, as channel estimation
is not necessary.

ƒ Differential schemes are more robust than non-

differential schemes to residual phase offsets caused
by imperfect synchronisation.

ƒ However, non-differential techniques require up to 3

dB less power than differential techniques to achieve a
specific target error rate in noise.
Channel Estimation Operations

ƒ In coherent detection schemes, information recovery

requires an estimate of the channel response.
Estimation of the channel consists of two steps:

ƒ The first step involves the insertion of known symbols

or a pilot structure into the OFDM signal, that yield
point estimates of the channel frequency response.

ƒ The second step involves an interpolation operation,

that yields the remaining points of the channel
frequency response from the point estimates.

ƒ The performance of channel estimation operations

depends both on the pilot structure and on the
interpolation method.
 Channel Estimation Operations
Example of a Pilot Structure to Track Channel Variations
in Time and Frequency

…
Frequency (OFDM sub-carriers)

High density of pilots gives good channel

estimation accuracy, but higher loss in
bandwidth efficiency/SNR
…
Low density of pilots gives lower loss in
bandwidth efficiency/SNR, but poorer
channel estimation accuracy

Maximum pilot separation in frequency=

=Channel coherence bandwidth, Bcoh

Maximum pilot separation in time=

=Channel coherence time, Tcoh
Time (OFDM Symbols)
 Synchronisation Operations

ƒ Information recovery also requires accurate symbol

timing and carrier frequency synchronisation between
the OFDM transmitter and the OFDM receiver.

ƒ Synchronisation in OFDM systems is performed before

detection and involves two phases: acquisition and
tracking.

ƒ In the acquisition phase, the frequency and timing

errors are coarsely estimated and corrected.

ƒ In the tracking phase, only small short-term

deviations are estimated and corrected.
 Synchronisation Operations
ƒ A number of techniques to perform symbol timing and
carrier frequency synchronisation have been proposed.

ƒ Synchronisation techniques based on the cyclic prefix

rely on the computation of the correlation of the
received OFDM signal with a delayed version of the
same received OFDM signal over an interval equal to
the cyclic prefix interval.

ƒ Synchronisation techniques based on the transmission

of special OFDM training symbols rely on the
computation of the correlation of the received OFDM
signal with the known OFDM training symbols.

ƒ The peaks of the correlation function yield the timing

information and the phase of the peaks of the
correlation function yield the frequency information.
 Synchronisation Operations

Synchronisation Based on the Cyclic Prefix

Correlation main peak power Correlation sidelobes

AND Use special training symbols
varies from symbol to symbol are not negligible
T
TCP TCP

OFDM signal … …
t

T
Delayed … …
OFDM signal
t
TCP
Average over TCP
… …

Correlation … …

t
Choice of OFDM Signal Parameters
ƒ How do we choose the parameters of an OFDM signal (e.g.,
T, Tcp, N) given specific data rate and bandwidth constraints
and a channel with a specific delay and/or Doppler spread?

ƒ Channel delay spread Tm dictates length of the OFDM cyclic

prefix TCP…length of the OFDM cyclic prefix is approx. two to
four times channel delay spread to limit ISI/ICI.

ƒ Length of the OFDM cyclic prefix TCP dictates total length of

the OFDM symbol T+TCP…total length of the OFDM symbol is
approx. five times length of OFDM cyclic prefix to limit SNR
loss to approx. 1dB.

ƒ The number of sub-carriers N, code rate and modulation are

selected to meet the data rate and bandwidth constraints.

ƒ Channel Doppler spread Bd may dictate a maximum number

of sub-carriers N.
Choice of OFDM Signal Parameters
ƒ Design an OFDM system with the following requirements:
data rate=6, 9, 12, 18, 24, 36, 48 and 54Mbps, bandwidth<
20MHz and delay spread=200ns.

ƒ Choose TCP=4×200ns=800ns and T+TCP=5×800ns=4000ns,

so T=3200ns and ∆f=1/T=312.5kHz.

ƒ Note that the maximum number of sub-carriers that can be

fit into the allocated bandwidth is 64.

ƒ Choose size of IFFT (FFT) to be equal to 64 (power of 2)

and number of (data) sub-carriers to be equal to 48.

ƒ Note that the remaining IFFT (FFT) values are used for
pilots and/or zero-padding.

ƒ Finally, choose code rate and modulation to meet data rate

requirements.
Choice of OFDM Signal Parameters
Choice of OFDM Signal Parameters

Data Rate Bandwidth N Code Rate Modulation

6 Mbps 15 MHz 48 1/2 BPSK

9 Mbps 15 MHz 48 3/4 BPSK

12 Mbps 15 MHz 48 1/2 QPSK

18 Mbps 15 MHz 48 3/4 QPSK

24 Mbps 15 MHz 48 1/2 16-QAM

36 Mbps 15 MHz 48 3/4 16-QAM

48 Mbps 15 MHz 48 2/3 64-QAM

54 Mbps 15 MHz 48 3/4 64-QAM

 Advantages/Disadvantages of
OFDM
ƒ Advantages of OFDM:
ƒ Good performance under delay spread/frequency
selective fading conditions;
ƒ Bandwidth efficiency;
ƒ Efficient digital signal processor based generation/
detection techniques.
ƒ Disadvantages of OFDM:
ƒ Poor performance under Doppler spread conditions/time
selective fading conditions;
ƒ Sensitive to non-linear distortion;
ƒ Sensitive to timing and frequency offsets as well as
phase noise.
 Timing and Frequency Offsets
ƒ Timing offsets originate due to uncertainties to
establish the OFDM symbol boundaries.

ƒ Timing offsets give rise to intersymbol interference

and interchannel interference or simply a phase offset
in desired data.

ƒ Carrier frequency offsets originate from frequency

differences in the local oscillators at the transmitter
and the receiver used to convert the baseband signal
to a bandpass signal and vice versa.

ƒ Frequency offsets give rise to interchannel

interference and a reduction in power in the desired
data.
 Timing Offset

Effects of Timing Offset

Timing offsets where FFT observation
Intersymbol and intercarrier interference
time overlaps adjacent symbols

FFT observation time 1 FFT observation time 2

∆T1 ∆T2

Nominal FFT
observation time

TCP T
 Timing Offset

Effects of Timing Offset

Timing offsets where FFT observation

Phase offset ϕn=2π
=2πfn∆t, n=0,…,N-
n=0,…,N-1
time does not overlap adjacent symbols

Earliest possible FFT

observation time

Latest possible FFT

observation time

TCP T
 Frequency Offset

Effects of Frequency Offset

∆f ∆f ∆f ∆f ∆f ∆f

Reduction of desired power

..
.
Intercarrier interference

..
.

-(N/2)/T -(N/2-1)/T -(N/2-2)/T (N/2-3)/T (N/2-2)/T (N/2-1)/T

f
 Phase Noise

ƒ In practice, the up-converter and down-converter

oscillators in the OFDM system do not produce a
carrier at exactly one frequency.

ƒ Instead, these oscillators in the OFDM system produce

a carrier at a nominal frequency with a time-varying
frequency offset, i.e., phase noise.

ƒ Phase noise introduces a phase offset common to all

sub-carriers as well as intercarrier interference.
 The PAPR
ƒ OFDM signals are sensitive to non-linear distortion
due to its high peak-to-average power ratio (PAPR)
or, its multi-carrier nature.

ƒ From one view point, the saturation region of any

non-linear element in the system occasionally clips
the signal due to its high PAPR.

ƒ From another view point, any non-linear element in

the system introduces severe intermodulation
distortion (IMD) due to the multi-carrier nature of the
signal.

ƒ These effects result in signal error probability

degradation and signal spectral spreading.
 The PAPR
Effects of Non-Linearities

Non-Linear Element output

Distorted OFDM Signal

input t

Non-Distorted OFDM Signal

Large signal excursions hit saturation

region of non-
non-linear element

Output signal is a distorted replica of

input signal
t
 The PAPR
Distorted OFDM Signal
Effects of Non-Linearities

Non-Distorted OFDM Signal

t t

Non-
Non-Linear
t Element t

Non-
Non-linear element introduces IMD
t
t
In-
In-band IMD associated with signal
error probability degradation

Out-
Out-of-
of-band IMD associated with
signal spectral spreading t
 The PAPR

ƒ A number of techniques have been proposed to

improve the performance of non-linearly distorted
OFDM signals, e.g., PAPR reduction and pre-distortion.

ƒ PAPR reduction techniques rely on the reduction of the

OFDM signal variability.

ƒ Pre-distortion techniques rely on an appropriate prior

distortion of the OFDM signal such that the effect of
the non-linearity is undone.
 Applications of OFDM
ƒ OFDM has been proposed for a number of systems
including wired and wireless applications.

ƒ For example, OFDM has been selected for assymetric

digital subscriber line (DSL) systems under the
acronym of discrete-multitone (DMT).

ƒ OFDM has also been selected for digital audio

broadcasting (DAB) and digital video broadcasting
(DVB) as well as wireless local area networks (LAN) –
in IEEE 802.11/HIPERLAN- and wireless metropolitan
area networks (MAN) – in IEEE 802.16/HIPERMAN.
 Digital Subscriber Lines
ƒ OFDM under the acronym of DMT has been adopted
for ADSL.

ƒ ADSL is a scheme for high speed communication in

the telephone access network or the subscriber line,
where the bit rate offered in the downstream direction
(to the subscriber) is larger than the bit rate offered in
the upstream direction (to the central office).

ƒ For example, in the USA the ADSL standard supports

downstream bit rates from 1.54 to 6.1 Mbit/s and
upstream bit rates from 9.6 to 192 kbit/s.

ƒ ADSL is suitable for applications like video on

demand, games, virtual shopping and internet surfing.
 Digital Subscriber Lines
ƒ Copper wire pairs are the dominating medium type in
telephone access networks or subscriber lines.

ƒ Impairments in this environment include the “highly

spectrally shaped” frequency response of a typical
copper wire pair and various noise sources.

ƒ Noise sources include crosstalk from other copper wire

pairs in the same cable, RF noise and impulse noise.

ƒ Near-end crosstalk occurs at the central office when

the weak upstream signal is disturbed by strong
downstream signals.

ƒ Far-end crosstalk is crosstalk from one transmitted

signal to another in the same direction and appears
both at the central office and the subscriber.
 Digital Subscriber Lines
Near-End Crosstalk (NEXT)

Cable

Wire pair
Transmitter
NEXT
Wire pair
Receiver

Far-End Crosstalk (FEXT)

Cable

Wire pair
Transmitter
FEXT
Wire pair
Receiver
 Digital Subscriber Lines
ƒ OFDM in conjunction with bit loading techniques has
been selected for ADSL for it efficiently combats the
adverse effects in the telephone access network or the
subscriber line.

ƒ Bit loading is a scheme whereby higher-order

constellations are assigned to high signal-to-noise
ratio sub-channels and lower-order constellations are
assigned to low signal-to-noise ratio sub-channels.

ƒ Essentially, OFDM in conjunction with bit loading

techniques offers a practical solution to achieve
transmission rates close to the capacity of a “highly
frequency shaped” linear channel.
 Digital Broadcasting
ƒ OFDM systems have been proposed to assist the
migration from old analogue broadcasting systems to
new digital broadcasting systems.

ƒ For example, in Europe OFDM has been adopted for

DAB and terrestrial DVB.

ƒ OFDM systems have been proposed for digital

broadcasting systems mainly due to their capability to
combat multi-path propagation and narrowband
interference as well as their capability to allow for the
implementation of single frequency networks.
 Digital Broadcasting

ƒ Narrowband interference originates from the

coexistence of analogue and digital broadcasting
systems.

ƒ Narrowband interference can be combated efficiently

using spectrum shaping techniques.

ƒ Essentially, sub-carriers experiencing high signal-to-

interference ratios are transmitted whereas sub-
carriers experiencing low signal-to-interference ratios
are not transmitted.
 Digital Broadcasting

ƒ In a conventional broadcasting network,

geographically adjacent transmitters transmit the
same signal/program on distinct frequencies.

ƒ In a single frequency broadcasting network several

geographically dispersed transmitters transmit the
same signal/program synchronously on the same
frequency.

ƒ A receiver therefore observes several replicas of the

same signal possibly attenuated and delayed with
respect to one another.
 Digital Broadcasting
ƒ A single frequency network can greatly enhance the
spectrum and power efficiency of a broadcasting
system.
ƒ Spectrum efficiency is improved because only a single
frequency instead of multiple frequencies is used to
cover an entire region or country.
ƒ Power efficiency is improved because the coverage
area served for example by two transmitters
operating simultaneously is greater than the sum of
the coverage areas for each of the two transmitters
operating independently.
ƒ Essentially, the phenomena occurring in a single
frequency network is analogous to the phenomena
occurring in a multipath fading channel and hence it
can be combated efficiently with OFDM.
 Wireless LAN

ƒ OFDM has been selected as the basis for the physical layer
of a number of packet based indoor wireless LAN standards
such as IEEE 802.11a for the 5 GHz band, IEEE 802.11g for
the 2.4 GHz band and HiperLAN2.

ƒ For example, IEEE 802.11a is a standard to interconnect

portable devices to broadband networks.

ƒ IEEE 802.11 offers data rates ranging from 6 Mbit/s to 54

Mbit/s.
 Wireless LAN
ƒ A nominal channel bandwidth of 20 MHz is used.

ƒ A 64-point IFFT/FFT is used where 48 sub-carriers are

allocated for data, 4 sub-carriers are allocated for pilots and
the remaining sub-carriers are used for zero-padding.

ƒ The cyclic prefix duration is set to be equal to 800 ns and

the useful symbol duration is set to be equal to 3200 ns.

ƒ Modulation schemes used include BPSK, QPSK, 16-QAM and

64-QAM.

ƒ Forward error correction used includes the industrial

standard convolutional code with rate 1/2, constraint length
7 and generator polynomials (133,171) octal. Higher rates
of 2/3 and 3/4 are obtained by puncturing the convolutional
code.
 Wireless LAN
OFDM Parameters in IEEE 802.11

Cyclic Prefix Duration 800ns

Useful Symbol Duration 3200ns

Total Symbol Duration 4000ns

Sub-carrier Spacing 312.5kHz

Number of data sub-carriers 48

Number of pilot sub-carriers 4

Total number of sub-carriers 52

Bandwidth 16.56MHz
 Wireless LAN
OFDM Parameters in IEEE 802.11

Data Rate Code Rate Modulation

6 Mbps 1/2 BPSK

9 Mbps 3/4 BPSK

12 Mbps 1/2 QPSK

18 Mbps 3/4 QPSK

24 Mbps 1/2 16-QAM

36 Mbps 3/4 16-QAM

48 Mbps 2/3 64-QAM

54 Mbps 3/4 64-QAM

 Wireless MAN
ƒ OFDM has been selected as the basis for the physical layer
of a number of wireless MAN standards such as IEEE
802.16a in the 2 to 11 GHz range and HiperMAN.

ƒ For example, IEEE 802.16a specifies a wireless MAN which

provides an alternative to cable, DSL or T1 level services for
last mile broadband access, and backhaul for 802.11
hotspots.

ƒ IEEE 802.16a supports low latency applications such as

voice and video.

ƒ Moreover, IEEE 802.16a provides broadband connectivity

without requiring direct line of sight between the
subscribers terminal and the base station.
 Wireless MAN
ƒ Flexible channel sizes of e.g. 3.5, 5, 10 MHz, etc. are used.
ƒ A 256-point IFFT/FFT is used where 192 sub-carriers are
allocated for data, 8 sub-carriers are allocated for pilots and
the remaining sub-carriers are used for zero-padding.
ƒ Here the ratio of the cyclic prefix duration to the useful
symbol duration is equal to 1/4, 1/8, 1/16 or 1/32.
ƒ Modulation includes QPSK, 16-QAM or optionally 64-
QAM.
ƒ Forward error correction includes an outer Reed-Solomon
code concatenated with an inner convolutional code, or
optionally block turbo coding or convolutional turbo coding.
Variable forward error correction capability is obtained by
puncturing.
ƒ Space-time coding can be optionally used to further improve
performance.
 Recent Developments

ƒ Recent developments in the area of OFDM include

orthogonal frequency division multiple access
(OFDMA) and multi-carrier code division multiple
access (CDMA).

ƒ OFDMA is a multiple access technique based on OFDM

technology.

ƒ Multi-carrier CDMA is a multiple access technique

based on the combination of OFDM and CDMA.

ƒ The most popular multi-carrier CDMA techniques are

MC-DS-CDMA and MC-CDMA.
 OFDMA

OFDMA Transmitter

User 1 .. ..
. .

D/A Converter
P/S Converter

Up-Converter
User 2 .. ..
IFFT

. . OFDMA signal
.. ..
. .
User U .. ..
. .
 OFDMA

OFDMA Receiver

.. .. User 1
. .
Down-Converter

A/D Converter

S/P Converter
.. .. User 2

FFT
OFDMA signal . .
.. ..
. .
.. .. User U
. .
 CDMA
CDMA Transmitter

User 1
Modulation
Symbols 1 Modulator

c1(t)
Carrier
CDMA Signal
…

User U
Modulation
Symbols U Modulator
Different users are allocated
different spreading sequences
cU(t)
(t)
Carrier Spreading sequences are
orthogonal
 CDMA
CDMA Receiver

User 1
CDMA Modulation
Signal Demodulator LPF Symbols 1

c1(t)
Carrier

User U
…
CDMA Modulation
Signal Demodulator LPF Symbols U

cU(t)
(t)
Carrier
 MC-DS-CDMA

MC-DS-CDMA Transmitter
Different users are allocated different
spreading sequences
User U Spreading sequences are orthogonal
cu(t)

cu(t)
Other users

D/A Converter
S/P Converter

P/S Converter

Up-Converter
Modulation
Symbols

IFFT

MC-
MC-DS-
DS-CDMA
..
signal
. ..
.
cu(t) Other users
 MC-DS-CDMA

MC-DS-CDMA Receiver

User U
cu(t)

LPF
cu(t)
Down-Converter

A/D Converter

S/P Converter

P/S Converter

Modulation
LPF

Symbols
MC-
MC-DS-
DS-CDMA

FFT
signal
..
.. . ..
. .
cu(t)

LPF
 MC-CDMA

MC-CDMA Transmitter
Different users are allocated different
spreading sequences
User U Spreading sequences are orthogonal
c1u

c2u
Other users

D/A Converter
P/S Converter

Up-Converter
Modulation
Symbols

Copy

IFFT

MC-
MC-CDMA
..
signal
. ..
.
cNu Other users
 MC-CDMA

MC-CDMA Receiver

User U
c1u

c2u
Down-Converter

A/D Converter

S/P Converter

Modulation
Symbols
MC-
MC-CDMA

FFT
signal
..
.. .
.
cNu
 Summary
ƒ The characteristics of the wireless channel and its
effects on communications systems were reviewed.
ƒ In this context, OFDM is proposed to overcome the
problems associated with wireless propagation.
ƒ The elements of a typical OFDM communications
system were described.

ƒ Emphasis was given to generation/detection of OFDM

signals, coding and interleaving methods, mapping
and demapping techniques as well as channel
estimation and synchronisation operations.

ƒ The effects of timing and frequency offsets, phase

noise and non-linearities on OFDM signals were also
described.
 Summary

ƒ A range of applications of OFDM were reviewed

including digital subscriber lines, broadcasting
systems (DAB and DVB), wireless LAN and wireless
MAN.

ƒ A range of recent developments in the area of OFDM

were also reviewed including OFDMA, MC-DS-CDMA
and MC-CDMA.