OFDM Orthogonal Frequency Division Multiplexing

Dr. Jean Armstrong Associate Professor Department of Electronic Engineering La Trobe University
1

Overview
!

Introduction
! !

applications multicarrier systems multipath transmission

Why use OFDM?

!

! ! ! !

How OFDM works Applications of OFDM Problems with OFDM Research in OFDM
2

Applications of OFDM
!

Digital Television
!

European and Australian standard Hiperlan 2 High speed data transmitted along existing telephone lines

Wireless Local Area Networks (LANs)

!

ADSL (asymmetric digital subscriber loop)

!

Future mobile telephony?

3

Multicarrier systems
!

Single carrier system

!

signal representing each bit uses all of the available spectrum available spectrum divided into many narrow bands data is divided into parallel data streams each transmitted on a separate band
4

frequency !

Multicarrier system
!

!
frequency

W N

What is OFDM?
!

OFDM is a multicarrier system

!

!
frequency

W N
!

uses discrete Fourier Transform/Fast Fourier Transform (DFT/FFT) sin(x)/x spectra for subcarriers

Available bandwidth is divided into very many narrow bands

! !

Frequency W/N

~2000-8000 for digital TV ~48 for Hiperlan 2

Data is transmitted in parallel on these bands

5

Why is OFDM so popular for new broadband systems?

!

Most broadband systems are subject to multipath transmission Conventional solution to multipath is an equalizer in the receiver
!

high data rates - equalizers too complicated

With OFDM there is a simple way of dealing with multipath

!

relatively simple DSP algorithms

6

What is Multipath?
!

More than one transmission path between transmitter and receiver Received signal is the sum of many versions of the transmitted signal with varying delay and attenuation

Effect of Multipath on Received Baseband Signal

1 2 3 4 5

Signal on Path 1 Received Signal on Path 2 Signal on Path 3 Signal

Received signal depends on bits 2 - 4

!

Received signal at any time depends on a number of transmitted bits

!

Intersymbol Interference (ISI)

8

Need equalizer to recover data

ISI gets worse as data rate increases

1 2 3 4 5

Signal on Path 1 Received Signal on Path 2 Signal on Path 3 Signal

Received signal depends on bits 1 - 4

! !

ISI covers more symbol periods Equalizer becomes too complicated

9

How does OFDM solve the multipath problem?

!

Data is transmitted in parallel

! !

longer symbol period e.g. for N parallel streams, symbol period is N times as long trick to avoid residual ISI

Cyclic prefix
!

10

How are signals transmitted in parallel without interference?

First three subcarriers
! !

Symbol period
T

2 lt 2 kt dt = 0, k l sin sin T T 0

Each subcarrier has a different frequency Frequencies chosen so that an integral number of cycles in a symbol period Signals are mathematically orthogonal
11

How is data carried on the subcarriers?

!

Data is carried by varying the phase or amplitude of each subcarrier QPSK, 4-QAM, 16-QAM, 64-QAM

Two possible subcarrier values

12

Baseband OFDM system

Discrete frequency domain Each input controls signal at one frequency Discrete Time Domain Samples of modulated and multiplexed signals

High speed data (complex)

Serial to Parallel

IFFT

Parallel to Serial

D/A Conver -ter

LowPass Filtering

Transmitter
Parallel to Serial Serial to Parallel LowPass Filtering

Received high speed data

FFT

A/D Conver -ter

Receiver
13

How are OFDM signals generated?

Typical IFFT Output Samples
!

Signal values at the output of the IFFT are the sum of many samples of many sinusoids - looks random

Parallel data streams are used as inputs to an IFFT IFFT output is sum of signal samples IFFT does modulation and multiplexing in one step Filtering and D/A of samples results in baseband signal
14

Modulation
!

Varying the complex numbers at the IFFT input results in modulation of the subcarriers

16-QAM

8-PSK

15

Signals at Input and Output of Transmitter IFFT

High speed data (complex) Serial to Parallel

IFFT

Parallel to Serial

D/A Conver -ter

LowPass Filtering

Transmitter
Complex value representing data is input to IFFT IFFT output gives samples of modulated multiplexed signal

16

OFDM in a multipath environment

- effect on one subcarrier
First symbol Second symbol

Signal on Path 1

Received signal in one symbol period is not a sinusoid Causes intercarrier interference (ICI)

Signal on Path 2

delay

ICI

17

Cyclic Prefix
Cyclic Prefix Symbol without prefix

Signal transmitted on one subcarrier for one symbol

! !

Each symbol is cyclically extended Some loss in efficiency as cyclic prefix carries no new information

18

Effect of multipath on symbol with cyclic prefix

Signal on Path 1 Signal on Path 2

Path delay

Cyclic Prefix

If multipath delay is less than the cyclic prefix

! !

no intersymbol or intercarrier interference amplitude may increase or decrease

19

Frequency selective fading

1 Amplitude

Transmitted Signal Main signal + Delayed signal

0 0.5 1 Symbol Duration 1.5 2

-1 2 Amplitude

Transmitted Signal Main signal + Delayed signal

-2

0.5

1 Symbol Duration

1.5

20

Spectrum of Received Signal

!

W
!

Multipath fading causes some frequencies to be attenuated Fading is approximately constant over narrow band This is corrected in the receiver
21

Amplitude and phase change

!
1 Amplitude

Transmitted Signal Main signal + Delayed signal

0 0.5 1 Symbol Duration 1.5 2

-1 2 Amplitude

!
Transmitted Signal Main signal + Delayed signal

-2

0.5

1 Symbol Duration

1.5

Multipath delay causes change in amplitude and phase of each subcarrier Change depends on subcarrier frequency Corrected in receiver by one complex multiplication per subcarrier
22

Multipath fading corrected by single tap equalizer

Received high speed data Parallel to Serial One Tap Equalizer

FFT

Serial to Parallel

A/D Conver -ter

LowPass Filtering

Change in phase and amplitude corrected by complex multiplication Receiver structure suited to DSP implementation
23

Digital Video Broadcasting (DVB) OFDM is used in the Australian

! !

digital television system 2048 point IFFT

!

1705 subcarriers used variable error coding variable cyclic prefix variable constellation
!

Flexible standard
! ! !

TV

4QAM, 16QAM, 64QAM

Broadcast system
!

mode determined by broadcaster

24

DVB - single frequency network DVB designed to allow the

! ! !

same frequency to be used for the same channel throughout a region Single Frequency Network More than one received signal
!

like extreme multipath

TV
Antenna Antenna

Reason for large number of subcarriers

!

8000 subcarrier option allows greater distance between transmitters

25

OFDM in ADSL
!

OFDM used in ADSL is usually called Discrete Multitone (DMT) Two way transmission
!

transmission can be tailored to the particular channel only real (not complex signal can be transmitted)

Baseband system
!

26

Frequencies used for ADSL

Power Spectral Density

Upstream

POTS

ADSL with Echo Cancelling (EC)

Frequency

0-4 kHz
Power Spectral Density

Downstream 25-1104 kHz

Upstream

ADSL with Frequency Division

Downstream 138-1104 kHz
Frequency

POTS

Duplexing (FDD)

27

OFDM/DMT in ADSL
Power Spectral Density

Upstream

POTS

Downstream 138-1104 kHz

Frequency

! !

256 subcarriers Test signals transmitted

!

received signal and noise level of each tone measured

Large constellations used on good tones

28

Hiperlan-2 - Wireless LAN

! !

64 point FFT, 52 subcarriers used Different modes

!

signal constellation, error coding, cyclic prefix feedback be used to determine transmission mode

Two way channel

!

29

OFDM Problems
!

High peak-to-average power ratio

!

peak signals power much greater than average signal power need very linear amplifiers with large dynamic range tight specifications for local oscillators Doppler limitation

Very sensitive to frequency errors

! !

30

High peak-to-average power

!

OFDM signal is sum of many separate sinusoids In worst case may all add constructively High peaks occur rarely

31

Solutions to peak-to-average power

High speed data (complex) Serial to Parallel

IFFT

Parallel to Serial

D/A Conver -ter

LowPass Filtering

Transmitter
! ! !

Coding to avoid the peaks - Monash Clip the peaks - La Trobe Predistort the signal to compensate for the amplifier nonlinearity - Victoria University
32

Frequency Sensitivity

Frequency

! !

! !

Individual subcarriers have sin(x)/x spectrum Large sidelobes result in sensitivity to frequency offset Subcarriers no longer orthogonal Tight specifications on local oscillators
33

W/N

Research at La Trobe University

!

Peak-to-average power reduction

! ! !

clipping effect on signal constellation clipping noise added at transmitter

Alternative modulation schemes based on OFDM

!

polynomial cancellation coded OFDM (PCC-OFDM)

34

Peak-to-Average power of OFDM

! !

High PAPR Linear amplifiers with large dynamic range required Peaks after D/A conversion may occur between Inverse Fast Fourier Transform (IFFT) output values
35

Amplitude distribution in OFDM

Histogram of signal amplitude Probability amplitude exceeds given power

!

Power >8.3dB above average for 0.1% of time

36

Previous clipping techniques

!

Clip at IFFT output

!

may miss peaks non-linearity causes out-of-band power

Clip after interpolation

!

Clip after interpolation+filter to remove out-of-band power

!

complicated filters, long impulse response, cause Intersymbol Interference (ISI)

37

New Peak Reduction Technique

Transmitter block diagram
IFFT+ Interpolate
a0,i " aN 21,i
N x (I1-1)+1 " zeroes

Clip

Frequency Domain Filter

c0,i " cN 21,i

Cyclic prefix D/A filter etc

fc

0 0

NxI1 point Inverse DFT Oversampling rate I1

Nonlinear processing Clipping ratio = CR

NxI1 point Forward DFT Oversampling rate I1

N x (I2-1)+10 zeroes

" 0

NxI2 point Inverse DFT Oversampling rate I2

aN / 2+2,i " aN 1,i

cNI1 N 2+1,i " cNI1 1,i

Add cyclic prefix P/S D/A Filter

38

Trigonometric Interpolation usingoversize IFFT

a0,i " aN 21,i
N x (I1-1)+1 " zeroes

!
NxI1 point Inverse DFT Oversampling rate I1

0 0

aN / 2+2,i " aN 1,i

Oversize IFFT Zeros on middle inputs trigonometric interpolation of output

39

Frequency Domain filtering

In-band components
c0,i " cN 21,i
NxI1 point Forward DFT Oversampling rate I1 N x (I2-1)+1 0 zeroes NxI2 point Inverse DFT

Clipped signal

c " cNI1 1,i

Oversampling rate NI1 N 2+1,i I2

" 0

In-band components are passed to inverse DFT Out-of-band components are nulled Operates on blocks - time variant filter Filter does not distort wanted OFDM signal

Out-of-band components
40

Performance
! !

PAPR reduction Out-of-band power

! !

with ideal amplifier with limited dynamic range with non-linear amplifier

In-band distortion

41

PAPR Reduction
! !

PAPR is reduced Filtering causes some peak regrowth Better performance than clipping before interpolation
!

note logarithmic scales

42

Out-of-band power
!
!

No increase in out-of-band power after clipping and filtering

in systems with unused band-edge subcarriers these subcarriers nulled by filter signal outside dynamic range of amplifier amplifier non-linearities

Amplification will result in out-of-band power

! !

43

Signal to clipping noise ratio

!

Even extreme clipping gives moderate signal to clipping noise ratio Main effect of clipping is shrinking of constellation Clipping noise is added at transmitter so fades along with signal

44

Repeated clipping and filtering

!

Peak regrowth can be reduced by repeatedly clipping and filtering

45

PCC-OFDM - solution to frequency sensitivity

1 1

0.5

0.5

0.5

-0.5

-0.5

-0.5

-1 -5

0 Frequncy

10

-1 -5

0 Frequncy

10

-1 -5

0 Frequncy

10

By coding the subcarriers in pairs frequency sensitivity can be reduced Would have been a better basis for DVB
46

ISI/ICI of OFDM and PCC-OFDM

Concentration of subchannels in time and frequency domain reduce ICI and ISI
47

OFDM

PCC-OFDM

PCC with overlapping symbol periods

0
!
! !

T/2

3T/2 2T

Symbols are overlapped

ISI is deliberately introduced equalizer required in receiver to recover data
48

PCC-OFDM receiver structure

exp ( j 2 ( f c + f )t )
LPF and ADC Delay line BPF X

y0,i " yN1,i

DFT

z0,i " Weight and zN1,i sum

v0,i v0,i1 v0,i2 " T/2 " T/2 " vN1,i Delay vN1,i1Delay vN
2
2

1 ,i2

Data clocked in at rate T/N

One DFT operation every T/2

Vi1 Vi Vi+1

" "

TwoDimensional Equalizer

i D

! !

Requires two dimensional equalizer Properties of PCC-OFDM mean that only a few terms along the diagonal are significant
49

Performance in a multipath channel

10
0

10

-1

10 Bit Error Ra te

-2

10

-3

10

-4

!
-5

10

10

-6

four line a r s ta ge s - 10dB four line a r s ta ge s - 15dB OFDM - 10dB OFDM - 15dB 0 5 10 15 20 de la y s pre a d (T/64) 25 30

PCC-OFDM outperforms OFDM Advantages increase as delay spread increases tolerance to delay spread depends on equalizer length, not length of cyclic prefix

N=64, OFDM cyclic prefix length =6T/64

50

Conclusions
!

OFDM is used in many applications

! !

solution to multipath good digital signal processing algorithms

Any questions?

51

Baseband OFDM system

High speed data (complex) Serial to Parallel Parallel to Serial D/A Conver -ter LowPass Filtering

IFFT

Transmitter
Parallel to Serial Serial to Parallel LowPass Filtering

Received high speed data

FFT

A/D Conver -ter

Receiver
52