Orthogonal Frequency Division Multiplexing

Cihan Tepedelenlioglu Arizona State University

Overview
Intro to digital communications Multipath fading and ISI Elimination of the ISI: Equalization Basic idea in OFDM History and context Advantages and disadvantages of OFDM Channel estimation and equalization Bit loading and DMT Diversity issues Conclusions
2

Digital Communications
Transmission of 0s and 1s (as opposed to waveforms)

bits

Tx

Channel

Rx

bit estimates

Tx turns bits into waveforms Channel is the propagation medium that causes distortions and noise Rx turns the received distorted waveforms to bits again It is easier to manipulate and store bits than waveforms; thats why digital is better.
3

Wireless Multimedia Communications

Current Research Trends

Bottleneck: The wireless channel

5

Multipath Fading

Reflections cause vectorial combination of received signal This results in frequent fading At high data rates multipath fading is frequency-selective If the receiver is in motion we also have Doppler Hence
Mobility causes Doppler (time-variations) High-rate signaling causes intersymbol interference (ISI)
6

Discrete-Time Model for ISI

In digital comm. multipath channel can be modeled in discrete-time

s(n)

p(t)

h(t)

p(-t)

x(n)

Hence x(n) and s(n) can be related by discrete-time convolution:

x(n) = h(l ) s (n l ) + noise

l =0

Discrete-time baseband equivalent model for the ISI channel Carrier modulation and demodulation is bypassed in baseband eq. ISI is also present in wired communications (e.g., ADSL)
7

ISI Example
x(n) = s (n) + s (n 1)
s(n) = +1, -1, +1, +1, -1, -1 +1 x(n) = , 0, 0, 2, 0, -2, 0 x(n) is Nothing like what was transmitted!!! x(n) can be viewed as the output of an LTI system

How to Eliminate ISI: Equalization

w(n) s(n) h(l) g(l)

( n ) s

Problems with linear equalization in SC systems:

Requires channel estimation (with training sequences) Channel inversion Suffers when channel zeros on the unit circle (no stable inverse) When the channel is long (high rates) the equalizer must be long (complexity)

The OFDM Concept

zk ( n ) = H ( 2k / N ) sk ( n ) + wk ( n ) k = 1,, N z( n ) = D H s( n ) + v ( n )

Duality: convolution in time is multiplication in frequency There is no ISI anymore! The channel is experienced in the frequency domain
10

Cyclic Prefix
N CP h Data

Cyclic prefix aids in removing the ISI and diagonalizes the channel
CP length should be greater than the channel length Introduces redundancy Hence greater the channel length implies more redundancy Ratio of redundant CP length to N can be minimized by making N large OFDM symbol length (N) is limited by other considerations such as
9 Complexity (FFT length) 9 ICI effects 9 PAR 11

History and Context

Basic idea is to use carriers with orthogonal frequencies Chang66, Saltzberg67 looked into orthogonal carriers (analog) Weinstein and Ebert, 1971, thought about using the DFT Kalet89 proposed assigning more bits to stronger subcarriers to increase data rates for a given performance (Water-Filling) for wired channels Hybrids of OFDM and CDMA for wireless communications Fazel97 The OFDM literature is rich Used in wireless LAN standards (laptops in a hotspot) DAB/DVB in Europe

12

Advantages of OFDM

zk ( n ) = H ( 2k / N ) sk ( n ) + wk ( n ) k = 1,, N
ISI channel can be handled as if we had a narrow-band system (no ISI)
Reduced complexity equalization

Higher-order constellations can be used on stronger subcarriers

Higher data rates for a given BER Requires channel feedback to the Tx, more feasible for wired comm.
13

Disadvantages of OFDM

zk ( n ) = H ( 2k / N ) sk ( n ) + wk ( n ) + ICI ( k ) k = 1,, N
Presence of channel zeros on the unit circle
Requires coding across subcarriers to recover transmitted symbols Even if channel zeros are close to the UC, performance degrades due to noise amp.

Time variations
Channel variations within one OFDM symbol Presence of a frequency offset Oscillator phase noise

High peak to average power ratio (dynamic range of transmitted signal)

14

Channel Estimation

zk ( n ) = H ( n;2k / N ) sk (n ) + wk ( n ) k = 1,, N
Pilot symbol insertion in time and frequency Interpolation can take advantage of channel correlation in time and freq.
15

Channel Equalization
zk ( n ) = H (n;2k / N ) sk ( n ) + wk ( n ) k = 1,, N k (n ) = zk (n ) / H ( n;2k / N ) = sk (n ) + wk (n ) / H ( n;2k / N ) s
Noise is amplified when there are channel nulls Error correction coding is necessary Channel estimation and equalization is much simpler than SC systems

16

Bit Loading and DMT for ADSL

zk ( n ) = H ( 2k / N ) sk ( n ) + wk ( n ) k = 1,, N

In wired comm. channel is time-invariant

Feasible to feedback the CSI to the transmitter This enables using higher-order constellations (higher rates) at stronger subcarr Exploitation of all the channel resources Similar to water-filling of Shannons Information Theory

17

DMT Block Diagram

18

DMT Bit Loading

Stronger channels get higher order constellations, thus maximizing rate

19

Relation between SNR and Bit Loading

Stronger channels get higher order constellations, thus maximizing rate

20

10

Time Domain Equalizer in DMT

The TEQ shortens the effective channel length, allowing a shorter CP

Implying less bandwidth waste on CP length
21

ICI in OFDM
By duality, time variations cause ICI in OFDM for wireless

22

11

PAR in OFDM
In SC comm. the dynamic range of the trans. signal is small In OFDM, N point FFT of symbols are transmitted:
x (n ) = s(n )e j 2kn / N
k =0 N 1

Peak value of |x(n)| is proportional to N Large dynamic range for transmitted signal
Expensive power amplifiers at the Tx

Solutions:
Coding to reduce PAR Use variants of OFDM without IFFT at the transmitter
23

Diversity Issues in Wireless OFDM

Diversity improves performance over fading channels The ISI channel offers multiple copies of the Tx signal OFDM does not exploit multipath diversity

To get diversity with OFDM

Hybrids of OFDM and CDMA Multi-antenna reception in OFDM Space-time coding combined with OFDM

24

12

Conclusions
Multipath fading and ISI Elimination of the ISI: Equalization Basic idea in OFDM History and context Advantages and disadvantages of OFDM Channel estimation and equalization Bit loading and DMT Diversity issues
25

13