Multicarrier Communications

-Lecture 7: OFDMA Systems

Jian (Andrew) Zhang

Andrew.Zhang@nicta.com.au

Wireless Signal Processing Program

National ICT Australia
http://users.rsise.anu.edu.au/ jian/Course mc.html

Canberra, Australia, 2007

MC 2007 Jian
c (Andrew) Zhang 1
What we covered in Lecture 6

MIMO Basics
Capacity Analysis
Space-frequency Coding
MIMO-OFDM Transceiver Design
Beamforming for MIMO-OFDM

MC 2007 Jian
c (Andrew) Zhang 2
What will be covered in this Lecture

Basics of Multiuser Access

OFDMA Basics
OFDMA Synchronization in uplink
Synchronization mechanism in practical systems: Ranging
channel
Synchronization schemes (of Academic interest) [1]
Cross-layer design
Multiuser Diversity
Mobility-dependent Traffic Channels
Multi-cell Frequency Planning

MC 2007 Jian
c (Andrew) Zhang 3
Multiuser Access

Multiple access schemes are used to allow many users to share

simultaneously a finite amount of spectrum.
The sharing of spectrum is required to achieve high capacity by
simultanously allocating the available bandwidth to multiple
users.
For high quality communications, this must be done without
severe degradation in the performance of the system.
There are many access techniques , some of them are
Frequency Division Multiple Access (FDMA)
Time Division Multiple Access (TDMA)
Space Division Multiple Access (SDMA)
Spread Spectrum Multiple Access (SSMA): e.g., CDMA, FHMA

MC 2007 Jian
c (Andrew) Zhang 4
Illustration of MA Schemes

MC 2007 Jian
c (Andrew) Zhang 5
Multiple Access in OFDM Systems
OFDM-FDMA (OFDMA)
OFDM-TDMA
OFDM-CDMA (Multicarrier + CDMA)

MC 2007 Jian
c (Andrew) Zhang 6
Advantages of OFDMA Systems

Better data rate granularity based on both time and frequency

domain resource assignment
Smaller Link budget for low rate users
Receiver Simplicity with multiuser-interference free detection
Multiuser diversity capability

MC 2007 Jian
c (Andrew) Zhang 7
OFDMA Optimality

The capacity is approached in a multi-user multi-carrier SISO

systems when the following conditions are satisfied:
1 Each subcarrier is assigned to only one user, i.e., OFDMA;
2 The assigned user on subcarrier n has the highest channel gain
over K users;
3 The power over subcarriers is allocated using a water-filling
solution with respect to channel gains.

Similarly, OFDMA optimality can also be shown for MIMO systems.

MC 2007 Jian
c (Andrew) Zhang 8
Downlink and Uplink

DL: One transmitter and multiple receivers

UL: Several transmitters and one receiver
In UL, all transmitters have unique time and frequency offset,
thus, UL system design is more difficult than DL.
Asymmetric traffic distribution between UL and DL is always
expected.

MC 2007 Jian
c (Andrew) Zhang 9
OFDMA System Design

Factors to consider
Different users can have different mobile speed, different
channel delay spread – number of subcarriers, subcarrier
interval, length of guarding interval, pilots allocation
Different users may have different data rate and QoS
requirements – subchannel length, subchannel structure,
adaptive modulation and coding
Different users may have different channel impulse response –
Multiuser diversity
Optimization between average system performance and
spectrum efficiency (system capacity) – Intelligent Frequency
planning

MC 2007 Jian
c (Andrew) Zhang 10
Subcarrier Allocation Strategy in OFDMA
Subband SAS
A group of adjacent subcarriers are assigned to each user;
Poor freq. diversity
Interleaved SAS
subcarriers of each user are uniformly spaced;
Good freq. diversity, but not best multiuser diversity
Random SAS
Allow dynamic resource assignment and provide best multiuser
potential

MC 2007 Jian
c (Andrew) Zhang 11
PN Sequence Design in Random SAS

N available data subcarriers are usually first grouped in blocks with

equal length Q, a permutation sequence (PN) and its cyclic versions
are then applied to assign subcarriers to users.
N = PQ, basic PN sequence {cm }, 1 ≤ cm ≤ Q
The period of PN sequence can be equal to P, which is generally
the length of a subchannel.
{cm } should have good circulant auto-correlation and
cross-correlation properties, to minimize intra-cell and inter-cell
interference while maximizing spectrum efficiency and frequency
reuse ratio.

MC 2007 Jian
c (Andrew) Zhang 12
Timing and Frequency Synchronization

General progress is:

1 In downlink, MT performs freq. and timing estimation and adjust
itself for uplink transmission;
2 BS estimates freq. and timing for all the users;
A challenging task as the BS receives a mix of signals each
affected by exclusive synchronization errors;
A multi-parameter estimation problem where each user must be
separated from others.
3 Timing and freq. correction in BS

MC 2007 Jian
c (Andrew) Zhang 13
Timing in Uplink

1 BS broadcasts timing pilots and k-th MT estimates initial timing

difference τp (k) + ∆T (k ), τp (k ): propagation delay; ∆T (k):
basic timing difference between transceiver
2 k -th MT synchronizes to BS, and transmits uplink stream;
3 Signals from different users will arrive at the BS with timing
difference τp (k) − τp (m);
4 Guarding period (CP) needs to be larger than the sum of channel
delay spread and the propagation delay max(|τp (k) − τp (m)|).
For systems with large covering area, propagation delay can be
comparable to the symbol period. Unacceptable!

MC 2007 Jian
c (Andrew) Zhang 14
Quasi-synchronous Network
Some mechanisms, e.g., a Ranging channel, can be adopted to
establish an ISI-free quasi-synchronous system such that the timing
error becomes small and can be incorporated in channel estimates.

One example:
1 BS broadcasts timing pilots and k-th MT estimates initial timing
difference τp (k) + ∆T (k ), τp (k ): propagation delay; ∆T (k):
basic timing difference between transceiver
2 k -th MT synchronizes to BS, and transmits unique timing pilots
to BS;
3 BS estimates τp (k) by using correlator with known template for
each user, and send the value of τk to user k ;
4 User k then adjust his transmitting time by −τk , so that signals
from all users will arrive at the BS roughly at the same time.

MC 2007 Jian
c (Andrew) Zhang 15
Ranging Channel

Ranging channel is a control channel allowing users to build

connection with BS. Functions include
letting MT adjust its parameters (CFO, Timing, etc.) to join a
quasi-synchronized network;
Resource request and assignment
Parameter tracking
Power measurement/Power control and Handover
Research problems:
Optimal resource configuration of ranging channel
Timing and CFO estimation based on pilots in ranging channel
Ranging code design

MC 2007 Jian
c (Andrew) Zhang 16
Ranging Channel - An Example

MC 2007 Jian
c (Andrew) Zhang 17
Synchronization with Subband SAS
If CFO is smaller than the subcarrier guard intervals, users’
signals can be separated by passing the received samples
through a bank of digital band-pass filters, each selecting one
subband.
General sync algorithms can then be applied independently for
each user.
Perfect users’ separation cannot be achieved in practice as this
would require ideal brickwall filters and/or large guard intervals.

MC 2007 Jian
c (Andrew) Zhang 18
CFO Estimation with Interleaved SAS

Each user’s block has a periodic structure. For m-th user

occupying P subcarriers {im + pL; 0 ≤ p ≤ P − 1}, at distance L,

sm (k) = ej2πn(im +m )/L sm (k + `P), 0 ≤ ` ≤ L − 1. (1)

This signal model motivates the use of Spectrum Analysis

algorithms, e.g., MUSIC, ESPRIT and Matric pencil, to estimate
the CFO term m .

MC 2007 Jian
c (Andrew) Zhang 19
CFO Estimation with random SAS

Research still in very early stage.

ML estimation is developed based on training blocks;
Complexity forbids its implementation in practice.

MC 2007 Jian
c (Andrew) Zhang 20
Timing and CFO Compensation in uplink

1 Estimates are returned to MTs which adjust its parameters

accordingly.
2 Compensation in BS directly by applying advanced algorithms
Independent compensation for each user in Subband SAS
Frequency compensation through Interference Cancellation
Frequency compensation through Linear Multiuser Detection

MC 2007 Jian
c (Andrew) Zhang 21
Cross-layer Design

Cross-layer design: Joint optimization multiple layers in system

design. Why cross-layer design?
In broadband networks, traffic is highly diverse with distinct QoS
parameters, delay sensitivity and error-sensitivity, channel may
vary dramatically, and user pattern presents high dynamics in
mobility.
Decoupled layer design which copes with the worst case
condition will lead to very inefficient usage of spectrum and
energy.
Protocols with multi-layer adaptability can adapt to the service
variations and achieve high spectrum efficiency.

MC 2007 Jian
c (Andrew) Zhang 22
Cross-layer Design Issues in OFDMA

Multiuser diversity
Mobility-dependent Traffic Channels
fixed-portable applications: slow fading; low signalling overhead
with occasional feedback ⇒ Multiuser diversity suitable
Mobile applications: fast fading; intensive overhead for resource
allocation ⇒ Frequency diversity suitable
Adaptive Modulation and Coding
Scheduling

MC 2007 Jian
c (Andrew) Zhang 23
Multiuser Diversity

In a wireless system with many users, the utility value (data rate,
channel) of a given resource unit varies from one user to another.
Such fluctuation allow the overall system performance to be
maximized by allocating each radio resource unit to the user that
can best exploit it.
It can be proven that intelligent resource allocation renders
performance improvement (e.g., expected transmission rate per
subcarrier), and the improvement increases with the number of
users.
The system performance improvement due to the increase in the
number of users is referred as multiuser diversity gain.

MC 2007 Jian
c (Andrew) Zhang 24
Where is Multiuser Diversity available

Service-enabled: difference in data rate and QoS - Granularity

related
Mobility-enabled multiuser diversity in relay networks
Space-enabled multiuser diversity in SDMA
Frequency-selectivity enabled multiuser diversity in OFDMA

MC 2007 Jian
c (Andrew) Zhang 25
Multiuser Diversity in OFDMA

Channel fading is statistically independent for different MTs, as long

as their receive antennas are separated considerably. – Similarity
between MIMO-OFDM and MISO-OFDMA
Water-filling algorithms
Finite Tones Water-filling
Loading algorithms
Algorithms to realize multiuser diversity in OFDMA will be discussed
in Lecture 8.

MC 2007 Jian
c (Andrew) Zhang 26
Mobility-dependent Traffic channels

Traffic channels (subchannels)

Multiuser diversity (MD) and Frequency diversity (FD) is a
tradeoff in OFDMA systems.
MD provides more performance improvement than FD;
MD and FD have different requirement on the distribution of
subcarriers in a subchannel.

MC 2007 Jian
c (Andrew) Zhang 27
Subchannel configuration for fixed/portable
applications

Clusters should be grouped as tight as possible to enable higher

multiuser diversity, leading to higher aggregated rate;
Small subchannel sizes are preferred - the multiuser diversity
gain can be readily captured with simple progressive channel
allocation scheme.

MC 2007 Jian
c (Andrew) Zhang 28
Subchannel configuration for mobile applications

Larger subchannels are preferred to provide better frequency

diversity and, thus, higher outage capacity;
For applications with small outage probability requirements, the
clusters should be distributed to enable higher frequency
diversity; the opposite is true for applications that can tolerate
high outage probabilities.
For mixed applications, the ideal system platform should be able
to support both “small and tight” traffic channels with maximum
multiuser diversity and “large and loose” traffic channels with
maximum frequency diversity.
The partition can be optimized based on the ration of low
mobility and high mobility users.

MC 2007 Jian
c (Andrew) Zhang 29
802.16e traffic channels
IEEE 802.16e defines three types of subchannel configuration
Fully used subchannelization (FUSC) and Partially used
subchannelization (PUSC) - “loose” and distributed configuration
for mobile channels to maximize FD
Advanced modulation and coding subchannel (AMC) - ”tight” and
adjacent channel configuration for fixed applications to maximize
MD
Zone Switching allows dynamic configuration of PUSC, FUSC
and AMC in a frame.

MC 2007 Jian
c (Andrew) Zhang 30
Multi-cell Frequency Planning [2]

Frequency resue: Same set of frequency bands can be reused by

multiple base stations as long as the cochannel interference is
tolerable.
Classification of Frequency planning schemes
Static/fixed frequency chanell allocation (FCA)
Adaptive/Dynamic channel allocation (DCA)
Combination of FCA and DCA

MC 2007 Jian
c (Andrew) Zhang 31
Fixed Channel Allocation
Total number of channels are divided into disjoint groups and
assigned to cells. Planning in the system setup and installation stage
and slight adjustment in the system operation process.
For Hexagonal cells, we have

Reuse distance √
reuse distance ratio = = 3 × cluster size (2)
Cell radius
where cluster size is the minimum number of neighboring cells
that are assigned the entire set of channels.
In a given area and a fixed cell size, smaller cluster size leads to
higher capacity.
Factors to consider in FCA
Capacity
Desired SIR in user terminals
Traffic loads (e.g., Irregular FCA for unevenly loaded traffic)
MC 2007 Jian
c (Andrew) Zhang 32
Dynamic Channel Allocation

Channel distribution adapts over time during system operation.

Centralized DCA in central controller: Excellent performance
with intensive signaling and computational complexity
Distributed DCA in BSs or MTs, independently or cooperatively
DCA in OFDMA is very challenging because
SINR criterion changed from predetermined to varied
Channels changed from flat fading to frequency selective fading
Computation complexity increased significantly due to the
subcarrier-based planning

MC 2007 Jian
c (Andrew) Zhang 33
OFDMA DCA: an Example
DCA realized by the coordination of central controller (e.g., RNC)
BSs and MTs. Signalling overhead and computation complexity are
both reduced.
Using beacon signals from BSs, each MT determines the
dominant interfering BS, and the achievable rates with and
without the dominant interference, and feedbacks the information
to BSs and then to RNC;
RNC updates all users’ CSI every super-frame, performs
interference avoidance, and determines specific set of
subchannels assigned to each BS and the recommended user
assignment for each BS;
BS makes actual pairing between the traffic bearers and the
users. When a user recommended by RNC has traffic to send,
BS follows RNC’s suggestions, otherwise, BS performs own’s
channel adaptation.
MC 2007 Jian
c (Andrew) Zhang 34
Inspiration on Research work

Feedback channel is usually present in modern systems

Design and exploit feedback channel
How to minimize the feedback information by allowing
preprocessing in MTs?
How to exploit the feedback information with errors?
How to design a feedback channel with multiuser access?
Multiuser diversity + Beamforming for MIMO/MISO OFDMA
Essentially Cross-layer design problems

MC 2007 Jian
c (Andrew) Zhang 35
Reference

M. Morelli, C.-C. J. Kuo, and M. Pun, “Synchronization techniques

for orthogonal frequency division multiple access (ofdma): A
tutorial review,” vol. 95(7), pp. 1394–1427, July 2007.
H. Liu and G. Li, OFDM-Based Broadband Wireless Networks -
Design and Optimization. New Jersey, USA: John Wiley &
Sons, 2005.

MC 2007 Jian
c (Andrew) Zhang 36