sistema de modulación cofdm en audio digital

8/10/2019 Sistema de modulacin COFDM en audio digital

1/10

The COFDM

modulation system:

the

heart of digital

audio broadcasting

by P. Shelswell

Digital audio.broadcasting offers the potential to give every radio in

Europe the sound quality of a compact disc.

To

accomplish this, it requires

a

rugged method of transmission. The coded orthogonal frequency division

multiplexing (COFDM) modulation system was developed to meet this need.

This paper describes the reasons why a new modulation process was needed, and

explains how the COFDM system has been optimised t o meet the requirements.

1

Introduction

Th e history of sound broadcasting ha s been one of gradual

improvement. In the early days, most people listened to

signals using amplitude modulation, seated round a large

receiver. It was difficult to conside r portable or mobile

reception because receivers were large, and either were

mains powered or required a sizeable battery.

Since then, frequency modulation and stereo

transmissions have been introduced. These were

originally planned to serve only fixed receivers, but

nowadays there

is

a diversity of receivers providing a rang e

of reception, from the conventional fixed hi-fi system,

through the standard transistor portable and ghetto

blasters to the mob ile car and personal radios.

Thus there is a demand for something that was not

originally part of the broadcast plan: mobile reception.

Originally, the FM broadcast chain was developed

assuming fixed reception with a directional rooftop

antenna. The use of small antennas attached to the

receiver, either indoors or on a car, now me ans that many

of the original planning assump tions are inapprop riate. As

a result the receiver h as to cope with a signal that

is

often

weaker than desirable and contains many echoes. Weak

signal strength combined with multipath propagation

lead s to a characteristic degradation of the received signal.

The programme sound is full of unwanted noises, which

are difficult o describ e in words but are recog nised by the

majority of people who listen to their radio in the car. The

problem

is

one of fading. It

is

common to the majority

of

radio systems, whether used for broadcasting, telephony

or general com munications.

In addition, ther e is an increasing dema nd for more and

more services. This h as led to a congested frequency band

with little room for expansion o r improvement.

One solution to these problems

is

to change

the

transmission standard to a digital system. But this is not

enoug h. Simple digital systems do not work well in the

multipath environment either. A major rethink of the

digital transmission system isneeded.

COFDM (coded orthogonal frequency division

multiplex) is a new digital transmission system which can

provide tugg ed reception, even in the fading channel. Th e

work on this system was initiated by

CCEXTL,2

n France

and developed into a major new broadca sting standard by a

collaborative project, Eureka 147. Th e new DAB (digital

audio broadcasting) system provides good reception in a

range of difficult conditions and is ideally suited to the

application. Reference 3 provides an introductory review of

the

DAB

system.

Th e same concept can also be used for transmission

of

any digital information, and it is interesting to note that

many of the digital television systems that are being

studied also make use of similar techniques.

In this paper, an outline

of

the COFDM modulation

system an d its applications

is

given. Most ofthe description

will be based on the Eureka 147 system, but will be

sufficiently general to allow th e theory to be applied to

ELECTRONICS COMM UNICAT ION ENGINEE RING JOURNAL JUNE 1995 127


2/10

amplitude

t

time or position

frequency

Fig. 1

propagation. The frequency response will vary wit h bot h time and p osition

Frequency selective attenuati on is clearly present

Typical fr equency responseof a channel suffering from multipath

othcr systems.The sound codingand multiplexing aspects

of th e DAB signal will notbe discussed in detail to keep the

paper short.

2 T h e p r o b l e m s of m u l t i p a t h p r o p a g a t i o n

Th e main problem with reception of normal radio signals is

fading caused by multipath propagation. This

is

not an

isolated problem. Delayed signals are the result of

reflections from terrain features such as trees, hills or

mountains, or objects such as people, vehicles or

buildings. Some of these reflections can be avoided by

using a good antenna, but there is a trend towards simple

antenn as on radios nowadays, so this isno longer a realistic

solution.

A characteristic of frequency selective fading is that

some frequencies are enhanced whereas others are

attenuated (Fig. 1).When th e receiver and all the objects

giving rise to the reflections remain stationary, then the

effective frequency response of the channel from the

transmitter to the rec eiver will be substantially fixed.

If the wanted signal is relatively narrowband and falls

into part of the frequ ency band with significant attenuation,

then there will be flat fading and reception will be

degraded.

If

on

the other hand, there is som e movement either of

the vehicle containing the receive r or of any of the

surroundings, then t he relative leng ths and attenuations

of

the various reception paths will change with time.

A

narrowband signal will vary in quality as the peaks and

troughs of the frequency response move around in

frequency. There will also be a noticeable variation in

phase re sponse. which will affect all system s using phase

as a means of signalling. For FM reception, thes e channel

distortionsgive rise to distortion of the sound a s well as the

morc obvious d rop outs a s the signal level falls below the

receiver threshold.

Now consider a signal which is of greater bandwidth.

Some parts of the signal may suffer from constructive

interference and be enhanced in level, whereas o thers may

suffer from destructive interference and be attenuated,

some time s to the point of extinction. In general, frequency

comp onent s close together will suffer variations in signal

streng th which are well correlated. Othe rs which are

further apart will be

less

well

correlated.

The

correlation

bandwidth is often used as a

measure of this phenomenon.

There is no standard definition

of th e correlation bandwidth.

Some workers also use the term

coherence bandwidth, One

typical definition is the

frequency separation of signals

which are correlated by a factor

of

0.9 or better. For a

narrowb and signal, distortion is

usually minimised if the

bandwidth is less than the

correlation bandwid th of the

channel. There

is,

however, a significant chance that the

signal will be subject to severe attenuation on some

occasions. A signal which occupies a w ider bandwidth,

greater than the correlation bandwidth, will be subject to

more distortion, but will suffer less variation in total

received power even

if

it

is

subject to significant levels of

multipath propagation.

If we

look

at the temporal response of the channel, we

see a number

of

echoe s present. There ar e many different

types of echo en vironmen t which ar e typical of different

geographical areas. In cities, echoe s come from reflections

from buildings; there are many separately identifiable

echoes, with a large range of delays observable.

In

the

countryside, the echoes ar e usually less distinct and have a

smaller range

of

delay, especially if there are no nearby

hills, although the presence of large hills and mountains

can increase the range

of

delay observed.

This range

of

delay can be measured statistically.

Different studies use the total range

of

delay,

or

the

average delay. Whichever is chosen, the inverse of this

leads to a good approximation for the correlation

bandwidth.

Measurements by the BBC have shown that the

correlation bandwidth depends very much on the

particular surrou ndings of the rec eiver. Typical results in a

built-up city show correlation bandwidths (using the

90

definition)

of

about

0.25

MHz in the VH F band. Changing

the definition of the correlation bandwidth to reflect a

corr elati on of

0.5

or more gives a bandwidth of over 1MHz.

COFDM is a wideband modulation scheme which is

specifically designed to cope with the problems of

multipath reception. It achieves this by transmitting a large

number of narrowband digital signals over a wide

bandwidth.

3

The

importance ofFDM

In COFDM, the data is divided between a large number

of closely-spaced carriers . This accounts for the frequency

division multiplex part

of

the name COFDM. Only a small

amount of the data

is

camed on each carrier, and this

significantly reduces the influence of intersymbol

interference (Fig 2). In principle, many modulation

A

qua l i ta t ive desc r ip t ion of COFDM

128

ELECTRONICS COMMU NICATION ENGINEERING JOURNAL

JUNE

1995


3/10

sche me s could be used to modulate the data at a low bit

rate onto each carrier."

In

DAB, quadrature phase shift

keying

(QPSK*)

is used , with differential encoding of the

data at the transmitter and differential demodulation at the

receiver.

It

is an important part of the COFDM system design that

the bandwidth occupied

is

greater than the correlation

bandw idth of the fading channel. Ag ood under standing of

the propagation statistics

is

needed to ensure that this

condition is met. From the me asurem ents we have made,

the bandwidth occupied is ideally more than a bout

1

MHz.

Then, although some of the carriers are degraded by

multipa th fading, he majority of the carrier s should

still

be

adequately received.

The importance of coding

Th e distribution of the data over many carriers means

that selective fading

will

cause som e bits to be received in

error while others are received correctly. By using an

error-correcting code which adds extra data bits at the

transmitter it is possible to correct ma ny or all of the bits

which were incorrectly received. The information carried

by one of the degraded carri ers is corrected because other

information, which is related

to

it

by the error-correction

code,

is

transmit ted in adifferen t part

of the

multiplex (and,

it is hoped, will not suffer the same deep fade). This

accounts for the 'coded' part of the name COFDM.

Ther e are many types of erro r correcting code that could

be used.',? In the DAB system , the main

channel code

is

a convolutional code,

with a Viterbi receiver. As there is

always a chance

of

residual errors after

a Viterbi decoder, some of the higher

priority data is precoded with a block

code for additional security. These

coding options have been specifically

tailored to the audio and signalling data

that is being broadcast and would

probably not he quite the same if

COFDM were used for other purposes.

The

importance

of

orthogonality

Finally, the 'orthogonal' part

of

the

COFDM nam e indicates that there

is

a

precise mathematical relationship

betwe en the frequencies of the carriers

in thesystem. In anormal FDM system ,

the many carriers are spaced apart in

such a way that the signals can be

received using conventional filters and

demodulators.

In

such receivers, guard

bands h q e to be introduced between

thc different carriers, and the

introduction of these guard bands in

* It

is

a minor detail. hut the modulation

system is actually n/4 D-QPSK In this

system, the phases of

the reference signals

are

ncreased by

45

each symbol

period. In

the rest

of

this p aper, for simplicity,

normal

QPSK will he assumed.

the frequency dom ain results in a lowering of the spectrum

efficiency.

It is possible, however, to arrange the carriers in a

COFDM signal so that the sidebands of the individual

carriers overlap and the signals can still be received

without adjacent carrier interference. In order to do this

the carriers m ust be m athematically orthogonal.

Th e receiver acts as a bank of demodulators, translating

each carrier down to DC, the resulting signal then being

integrated over a sym bol period to recover t he raw d ata. If

the o ther carriers all beat down to frequencies which, in the

time domain, have a whole number of cycles in the sym bol

period

( T ) ,

then the integration process results in zero

contribution from all these other carriers.

Thus

the

carriers a re linearly independent (i.e. orthogonal)

if

the

carrie r spacin g is a multiple of l r

Mathematically, suppose we have a set of signals Y,

where YD s the pth element in the set. The signals are

orthogonal if

0

1)

Yo t) Yz(t)d t

= K

for p

= q

O

f o r p e q

where the indicates the complex conjugate. It is fairly

simple to show, for example, that the series sin (mw)for m

=

1 , 2 ,

..

isorthogonal over the interval TI ton .

impulsehannelesponse

time

.

1 carrier

V

V

V

\

V v v \

n

n

carriers

V

n

8 c a r r

n

V

I

\rz/ r

Fig.

2

data rate, increasing the number of carriers reduces the data rate that each

indivi dual carrier must convey, and hence (for a given modulation system)

lengthens the symbol period . This means that the in tersymbo l interference

affects a smaller percentage of each symbol as the num ber of carriers and

hence the symbol period increases

The effect of adopting a multicarr ier system. For a given overall

ELECTRONICS COMMUNICATION ENGINEERING JOURNAL JUNE 1995

129


4/10

Much of transform theory makes use of orthogonal

series. Fourier series are a well known example of an

orthogonal series, although they are by no means th e only

example.

Otherfeaturesof

COFIIM

adopted to ensure the maximum robustness:

In the Eureka 147 system, some further refinements are

the data is interleaved, both in frequency and time. Th e

error correction process works hest

if

the errors in the

incoming data are random. To ensure this

is

true, the

transmitted data is interleaved ove r

all

the carriers and

over

a

range of time.

the addition

of

a guard interval allows the system to

cope with echoes

of

moderate d uration, and with small

inaccuracies in the receiver (for example small timing

errors) . It isdiscussed in more detail below.

Tlze

sound coding associated with

COFDM

When transmitting digitally-coded sound signals it is

particularly importan t to ensur e high efficiency in the use

of spectrum. Bit-rate reduction of the sound

is

therefore

used. Th e choice of sou rce coding for DAB

is

essentially

independent of the choice of COFDM for the modulation

scheme.

The Eureka 147 group has developed

a

system called

MUSICAM, which has now been adopted by the

International Standard s Organisation in their

IS0

11172-3

Layrr I1 standard.

This

offers a variety

of

options, in which

bit-rate can he traded for quality. For high-quality stereo

signals, a hit-rate of about 256 kbit/s is needed, hut the

standard offers a range starting at

32

kbit/s and rising

to

384 kbit/s.

In order to occupy sufficient bandwidth to gain the

advantages of the COFIIM system, it isnecessary to go up

a

number of programme s together to form awideb and system.

4 Mathematical description of

COFDM

After the qualitative descrip tion of th e system it isvaluable

to discuss the mathematical definition of the modulation

systcm. This allows

us

to

see

how the signal is generated

and how the receiver must operate, and it gives u s a tool to

understand the effectsof imperfections n th e transmission

channel.

As noted above, COFDM transmits a large num ber of

narrowband carriers, closely spaced in the frequency

domain. In order to avoid a large number of modulators

and filters at the transmitter and c omplem entary filters and

demodulators at the receiver, it is desirable to be able to

use mod ern digital signal processing techniq ues.

Mathematically, each carrier can be described as a

complex wave:

s, t)

=

A,(t )e

l , cf

W ' l

e )

Th e real signal is the real part

ofs,(t).

Both A,(t ) and @


5/10

This is the same condition that was

required for orthogonality (see Section

3). Thus,

one consequence

of

maintaining orthogonality is that the

COFDM signal can be defined by using

Fourier transform procedures.

The

Fourier

transform

The Fourier transform allows u s to

relate events in the time domain to

events in the frequency domain. The re

are several versions of the Fourier

transform , and the choice of which one

to use depends on the particular

circumstances of the work.

?he conventional transform relates

continuous signals which are not limited

in either the time or frequency domains.

However, signal processing is made

easier

if

the signals are sampled.

Sampling of signals with an infinite

spectrum causes all sorts

of

problems

because the sampling process leads

to

aliasing, and th e pr ocessing of signals

which are not time limited can lead to

interesting problems with signal storage.

10 void this, the majority of signal

processing uses a version of the

discrete Fourier transform (DET). Th e

DFT is a variant on the normal

transform in which the signals are

sampled in both the time and the

frequency domains. By definition, the

time waveform mus t repeat continually,

and this leads to a frequency spectrum

which repeats continually in the

frequency domain.

7 b e fast Fourier transform (FFT) is

frequency,

0 3

MHzIdivision

merelv a raoid mathematical method for calculatine the

in a domestic system

I

DIT . I t is the availability of this tec hniq ue, and technolog y

that allows it to be im plemented on integrated circuits at a

reasonable price. that has permitted COFDM to be

developed as far as it has. The rapid progress in the

development of FET chip sets has led to many exciting

opportun ities in this field.

The process of transforming from the time domain

representation to th e frequency domain uses the Fourier

transform itself, whereas the reverse process uses the

inverse Fourier transform.

T ~

se

ofthe FFTin COFDM

The main reason that the CO FDM technique has taken

so long to come to prominence has been practical. It has

becn difficult to generate such a signal, and even ha rder to

receive and demodulate the signal. Th e hardware solution

which makes use of multiple modulators and

demo dulators in parallel was somewh at impractical for use

Now, the ability to define the signal in the frequency

domain, in software, and to generate the signal using the

inverse Fourier transform is the key to its current

popularity.The use of the reverse process in the receiver is

essential

if

chea p and reliable receivers are

to

be readily

available. Although the original proposals were made

som e time ago. it has taken som e time for technology to

catch up.

At the transmitter, the signal isdefined in software in the

frequency domain. It is a sampled digital signal, and it is

defined such tha t the discrete Fourier s pectrum exists only

at discrete frequencies. Each COFDM c am er corresponds

to one element of this discrete Fourier spectrum. The

amplitudes and phasesof the cam ers depend on the data to

be transmitted. As each carrier isQPSKin this example, all

the carrier am plitudes are unity, but this

is

not necessary in

the more general case.The phase of each cam er

is

defined

for each transmitted symbol. At the carriers have their data

ELECTRONICS COMMUNICATION ENGINEERING JOURNAL JUN E

1995

131


6/10

Fig.

5

Generation of an

OFDM signal. The incomin g

data is converted from a

single serial bi t stream to a

multip le parallel stream,

which is modulated

o n to

a

large number of carriers.

Generation of the OFDM

signal itself could use a large

bank of oscillators and

multipliers. n reality, that

function (the part of the

system outl ined in th e

coloured box) is replaced by

an inverse

FFT

chip

spectrum

of

compositesignal

_ -

transitions synchronised, and can be processed together,

symbol by symbol. Fig.

5

is a schematic diagram of the

transmitter signal processing.

Using VLSl it is possible to carry out the inverse

FFT.

This provides a series of samples which are the time

domain representation of the signal. Thes e samples can be

applied to a conventio nal digital-to-analogue converter

(DAC) to give the real electrical signal.

To

enable the signal to be generated using an inverse

FFT, it

is

preferable that the number of carriers considered

in

the calculationis a po wer of 2. In practice, it is not always

desirable to have the number of real camers restricted in

this way. H owever, it is convenient to m ake up the actual

number chosen to a power of 2 by setting the amplitudes of

thos e not wanted to zero. This feature also simplifies the

design of the anti-alias filter after th e DAC.

In

the receiver, the reverse process is applied.Assuming

that the receiver has som e means of synchronisatio n, then

the signal is converted from an analogue format to a

sampled digital representation. The samples

corresponding to each symbol are than Fourier

transformed to the frequency domain. This gives the

amplitude and phase of each transmitted carrier.

In the Eureka 147 system, it

is

the change in phase of

each carrier from one symbol to the next which

communicates the information.

Orthogonality

A natural consequence of this method is that it allows us

to generate carriers which are orthogonal. The memb ers

of an orthogon al set are linearly independent.

Consider that our set of transmitted carriers, Y, is an

orthogonal set , such t hat

If

this

is truely orthogonal, then the orthogonality

relationship in eqn. 1 hould hold, that is

h

h

=

(b - a fo rp

=

q

= O f o r P t q a n d ( b - a ) = s ( 1 2 )

(remember thatp and

q

are integers)

Thus

the carrie rs, which are separated in frequency by

l/z,

meet the requirements

of

orthogonality provided that they

are correlated over a period z This is the formal derivation

of the result quoted earlier.

If the ana lysis is extended to include the phase of each

carrier, then this

is

recovered by the process defined in

eqn. 12. This is exactly the computation that is needed in

the receiver.

The method can be extended to show the effect

of

frequency and timing erro rs in the sys tem, but there is

insufficient space to d o so here.

Synchronisation and the guar d interval

In the receiver it is necessary to sample the incoming

signal and perform a transform to recover the carriers.

If

the system is all locked up, and the sampling frequency is

correct and in the right phase, there is no problem.

However this ideal case will be difficult to achieve,

especially when the receiv er is first switched on. 'Ihere

is

132

ELECTRONICS COMMU NICATION ENGINEERING JOURNAL JUNE 1995


7/10

therefore a need for acquiring timing lock.

In the DAB system, coarse

synchronisation is provided by a simple

technique: all the carriers are switched off

on a regular basis. By using a simple

amplitud e detection circuit, it is possible to

generate an approximation to th e timing.

However, the timing will not be perfect,

and all of the samples could be displaced

by a k e d time offset, giving rise to

intersymbol interference (see Fig. 6).

In DFT transform theory, the assum ed

waveform is a continuously repetitive

sequence, rather like wallpaper patterns

repeating across a wall. The minimum

information required

is

one cycle

of

this

pattern. To avoid the timing problems,

more than one complete symbol

is

transm itted the part of the symbol that is

repeated has been named the guard

intcrval (see Fig.

7).

With the guard interv al, he initial timing

accuracy only needs

to

ensure that the

samples are taken from one symbol (see

Fig. 8).Th e longer the guard interval, the

more rugged the syste m, but at a penalty of

the power needed to transmit the guard

interval. Ther e is obviously a compromise

to

be reached, but more of that later. In

practice, it is convenient to think

of

the

transm itted sym bol in two parts: the guard

interval precedes th eactive symbol period,

which is so called because in a correctly

aligned receiver the F l T window is in that

time

slot.

This gives ruggedness in the

uresence of echoes.

symbol

M

symbol

M

1

ymbol

M i

w w w

w

w w w

a Correctly timed samples

w w

w

w w

b

Incorrectly timed samplesgiving ntersymbol

interference

Fig. 6

must be decoded by sampling withi n the symbol period and then

perf orm ing a fast Fourier transform t o calcu late the phases of the

individual carriers. With a timi ng error, th ere can be considerable

intersymbol interference

The effect of timi ng errors o n reception of a signal. The signal

symbol

M 1

symbol symbol

M 1

uard active

period

interval

Fig.

7

parts. The whole signal is contained in the acti ve symbol (shown

highlig hted or the symbol

M),

the last part of which (shown in a lighter

colour ) is also repeated at the st art of the symbol and is called the

guard interval

Example of the guard in terval. Each symbol is made up of t wo

Once coarse synchronisation has beenobtained, there

is

then the question of how to improve it. After the null

symbol, a reference signal

is

transmitted. Correlation of

Another use of the guard interval

is

to minimise the

effectof echoes. If the echo isshort compared with the total

symbol period, then any energy conveyed from one

the received version of this symbol with the known

transmitted signal provides the impulse response of the

channel. From this, a much more accurate timing can be

obtained.

Similarly, the same signal also provides a

measure of the frequency error of both the sampling

frequency and the actual frequency

of

the signal (and

hence allows accurate automatic frequency control to be

implemented).

symbol

to

the next by the echo only degrades the gua rd

interval. Th e active symbol period contains direct ene rgy

and reflecti ons whose delay is less than the guard interval,

and istherefore derived from the sa me symbol period. This

is not intersymbol interference, but a form of linear

distortion. In the receiver the calculated phase of the

symbol for each spectral component is distorted by th e

multipath signal. If the channel is not changing rapidly

I

I I

I

I

symbol

M - i i:zA j

symbol

M ~

symbol

M +1

w w

w w w w w

aCorrect lyt im ed samples

w w

b

Incorrectly timed but decodable samples

w w w w w w

c Incorrectly timed samples giving intersymbol interference

Fig.

8

tolerance of adding a guard

interval. With a guard interval

included in the signal, the

tolerance on timi ng th e samples

i s considerably more relaxed

The effect on the timing

ELECTRONICS COMMUN ICATION ENGINEERING JOURNAL JUNE 1995

133


8/10

then successive symbols of any one carrier will be

perturbed in a similar manner. Ifwe use differential coding,

the receiver looks at the difference in phase from one

symbol o the next, and the e rrors cancel out (Fig. 9), Thus,

provided that the echo ha s a delay less than the additional

guard interval, there is no degradation to reception of the

active symbol period.

[t does not matter whether th e echo

is

passive, such as

one from a hill o r a building, or active, such a s one from

another transmitter. This is a major feature which allows

DAB to use the s am e frequency from all of its transmitt ers

in the network. Provided that th e transmitters radiate the

sam e signal at the sam e time, then reception will he good,

even in the overlap zones between the transmitters. This

reuse of the spectrum is a major saving over conventional

systems which would need different frequencies in

adjacent service areas. CO FDM, is, for a single channel,

only as efficient as the underlying modulation system.

However when the planning of a complete service is

considered. The single-frequency network (SFN)

operation is a major advantage." Th e spectral reus e leads

to major spectrum efficiencies.SeeSection 5for details.

5

Let us use the

DAB

system as an example of a COFDM

system and consider the choice

of

parameters.

First there

is

the question of what bandwidth to use for

The

appl icat ion

of

COFDM

t o

DAB

symbol

echo 1

resultant

1

symbol

2

77

esultant

1

echo 2

symbol 2

a b

Fig. 9

signal in th e presence of a single echo. Here, two

adjacent symbols are depicted, bot h affected by a single

echo. Arbitrarily, the

QPSK

signal has been coded such

that there i s a

90

phase shift fro m one symbol to th e

next. When there are echoes present, the received phase

i s distorted. In

a)

he symbol and t he echo have a fixed

relationship. This occurs when the echo contains energy

which comes exclusively from t he same symbol period .

An example would be if t he samples for t he signal are

timed as in Fig. 83 and

b .

The phase relationship

between the two resultant symbols is therefore

maintain ed. As differen tial decodi ng is used, the

constant phase shift is of no consequence. b )When

echoes are delayed by more than the guard inter val, as is

th e case in Fig.8c. matters are different. The phase of

the signal in the previous symbol does not have a fixed

relationship wi th th e current symbol, and so the

relationship between the phases of t he two resultant

symbols changes, and this distortion could degrade the

signal

Vector diagram of the di stortion caused to a

the COFDM system. The wider the bandwidth, the more

likely that he system exceeds the correlation bandwidth of

the channel. Short delay ec hoes are the main problem to

overcome, and as hese are always present, there is no hard

bound. The narrower the bandwidth, the m ore likely it is

that th e whole signal will he affected. Ther e is

a

trade-off

between bandwidth and transmitter power.

The original experiments were carried out with a

bandwidth of about 7 MHz, and sho wed few problems."

Then t he bandwidth was successively reduced to below

2

MHz, at which point, there is a degradation equivalent

to about 1dB in performance. This is not much, but the

degradation starts to increase quite rapidly when the

bandwidth

is

reduced below 1.5MHz. If the bandwid th is

reduced to th e

200

kHz used for

FM

sound, then the

margin required would b e an additional

6

dB or so.

T h u s

a figure of about

1.5

MHz for the bandwidth of the system

is

a good compromise for the type

of

propagation

conditions that apply to mobile and portable radio

reception.

One of the parameters that is directly affected by the

bandwidth is the available bit-rate. Th e modulation sy stem

on each carrier is QPSK. The carriers are separated in

frequency by about the inverse

of

their symbol period.

' h u s

the maximum bit-rate available is

2

bit/s/Hz of

bandwidth.This figure

is

reduce d by th e inefficiency of the

guard interval, the null symbol and t he er ror coding. For

DAB.

this brings the useful bit-rate down to about

1bit/s/Hz of bandwidth.

Thus a DAB system will provide just unde r

1.5

Mbit /s of

useful data. Thi s is considerably more than t he

256

kbit/s

that is needed for a high-quality stereophonic progr amm e,

so the implication is that several broadcast program mes

will sh are th e same multiplex.

Now consider the number

of

carriers. The more there

are,

he greater is the resolution

of

the diversity offered by

the system. Ther e is, however, a relationship between th e

symbol period and the carrier spacing. Th e carrier spacing

is I/s. For the differential demodulatio n to work properly,

the multipath environment must change slowly from

symbol to symbol. Thus there is a limit to the symbol

period and hence the number of camers. For static

reception, this is not a major problem. For mobile

reception, however, the motion

of

the vehicle leads to

changes in the multipath environment. Over a symbol

period, a vehicle moving at a velocity v m/s will travel

UT

x f / c wavelengths. This ishrw avelength s, where f,

is

the maximum Doppler shift. If this is to introduce

negligible phase distortion, then the function

f ~

ust he

small. A figu re of f ,7


9/10

significant echoes would all be relatively short. Surveys

indicate that agu ard interval of the ord er of 10

ks

would be

satisfactory for the majority of locations, n th e UK at least.

The use of several transmitters puts a limit on the

minimum guard interval that should be used. The

transmissions from areas so me distance away can reach

quite high levels on occasions of abnorm al propagation. O n

the days that th e weatherman announce s that television

reception may be subject to interference, ther e is also a

strong possibility that interference from remote DAB

transmitters may cause a problem. It

is

possible to plan a

service without worrying too mu ch about this if there is

suftlcient signal from the local transmitters. Then it

becomes a simple matter of deciding the spacing of the

main transmitters. This is a compromise between a small

number of high-power transmitters spaced by about 50 km,

or a much larger number of lower-power transmitters.

Because the first option is likely to be the che apest, the

guard interval is set to about 250 ms quivalent to a

maximum difference of about

80

km in transmission

distance.

The symbol period need not be directly related to the

guard interval. It

is

just a qu estion of how much of the

symbol period is repeated in the gua rd interval. This

is

a

straigh t question of efficiency, as the power transm itted in

the guard interval does not form a useful part of the

information in the receiver unless there are substantial

echoes. To minimise the power loss by this me chanism, it

is

desirable to keep the guard interval to as low a

perce ntage a s possible of the sym bol period. In practice a

guard interval of the order of 25%of the symbol period h as

been found to be agoo d compromise.

This leads to a symbol period of 1ms and hence , using

eqn. 10, to a carrier spacing

of

about 1 kHz. Thus about

1500 carriers are accomm odated in the minimum

bandwidth desirable for on e COFD M transmission.

In Section 4, it was suggested that the DAB system is

spectrum efficient. We now have the data to expand on this

staiement. Within the multiplex we are achieving about

one programme p er 250 kHz of bandwidth. It

is

possible to

transmit the sa me signal through out the full exten t of any

area requiring the sa me se t of prog rammes, thu s offering

four programmes per megahertz in that area. To permit

separate regional or national requirements, by the m a p

colouring theorem, four frequency blocks should be

adequate to provide one single-frequency network (SFN)

in ta ch country. This leads to a minimum overall spectrum

requirement of 1 MHz per stereo prog ramme to provide

flexibility of progra mm es w orldwide.

Hy comparison. at least 2.2 MHz

is

needed to provide

VHF FM programmes in the UK, and this extends to

3.3

MHz in Europe, where the countries are not

surrounded by sea. Optimistic estimates to provide

separa te service s of conventional digital radio, for examp le

using the NICAM system used for television sound,

indicate that the minimum required spectrum would be 3.4

MHz per programme.

A

figure closer to 10 MHz per

programme

is

thought m ore appropriate.

T h u s he SFN a pproach usin g COFD M offers significant

economies in spectrum.

When the signal

is

generated, it is defined in the

frequency domain, and then transformed into its time

domain representation. Sampled digital signals of course

suffer from aliasing. In a real system , the spec tral repeats

are not wanted, and

so

must b e filtered

off.

To make the

filtering easier, the spec trum is not defined such th at allthe

avai lable carr iers a re used. Th e outer c am ers

are

defined

to be zero and this leaves a gap in the spectrum which can

be used to minimise the complexity of the output filter.

When deciding exactly how many carriers are used, and

how many are processed, the decision is guided by th e fact

that the

FFT

works most effectively if the number of

samples in either domain isapo wer of 2. Thu s for DAB, the

number of carriers defined

is

1536, and th e processing is

based on 2048 carriers in the system. When DAB was first

being developed, the need to include a 2048 point FFT in

the receiver was seen as a disadvantage. Now, however,

technology has advanced to the point where higher orde r

FFTs

are bein g pro posed for digital television applications.

6 Other systems

Th e DAB system discussed here, as defined by the Eu reka

147 partners,

is

not the only system that could be defined.

The Eureka group itself has other options which are

tailored for use with systems operating in the top e nd o fthe

UHF range, and which have satellite transmission as a

major objective.

If the syste m is for futed reception, as is usually the cas e

with television, then many of the pro blem s with multipath

propagation are less severe. Th ere is less variation in the

channel, both because the antenna is capable of being

selective and because the channel is naturally not as

variable as it would be in a moving vehicle. As a

consequence it is possible to use shorte r guard intervals,

or even

no

guard interval at all.

The system used to modulate the individual carriers

does not need to be QPSK. It could be a higher order

system such as 16,6 4or even 256 QAM. Indeed a mixture

of modulation systems

is

possible,

so

long as they are

orthogonal. This restriction is usually met

if

the symbol

periods are the same . Th e use of coherent demodulation

will give better performan ce if the channel

is

not varying

too much. For mobile reception of DAB, the channel

response may vary rapidly in phase, and

so

the potential

benefits of coherent demodulation are lost in the

implementation. For the more static channel, these

benefits can be realised, and

so

it

is

useful to ad opt the

technique. Th e main challenge of using the se higher o rder

systems

is

to be able to cope with the changing

environment,

so

synchronisation and equalisation

strategie s pay a key part in their design.

7 Conclusions

In this short paper, it has only been possible to give an

outline of COFDM. This modulation system

is

being

increasingly used fo r digital transmission in environments

where multipath propagation can cause significant signal

distortions.

ELECTRONICS COMMU NICATION ENGINEERING JOURNAL J U N E 1995 135


10/10

Now that the Eureka 147

DAB

system has been fully

defined and accepted

as a

European Standard,13 many

broadcasters

are

preparing to start services.

In

the UK, the

BBC is planning

to

start broadcasting using DAB in

September 1995. Although, at the time of writing this

paper, we have yet to

see

the first dom estic receivers,

the

rapid pace of development m eans that the complexities of

COFDM are likely to

be

implemented

in

domestic

equipment before the paper

is

published.

The

future

is

exciting.

Acknowledgments

Th e author would like to thank the many eng ineers in the

Eureka 147 project who have contributed to his

understanding

of

the system, and to the BBC for

permission t o publish this paper.

References

1 POMMIER. D.. and WU. Y.: 'Interleaving or spectrum

spreading

in

digital radio intended for vehicles', EBUReview,

June

1 9 8 6 , 2 1 7 ,pp.12%142

2 ALARD M., and LASSALLE,R: Principles ofmodulation and

channel coding for digital broadcasting for mobile receivers'.

EBLJ Collected Papers on concepts for sound broadcasting

into the 2lst century, August

1988, pp.47-69

3 PRICE. H.M.: 'CD by radio: digital audio broadcasting', IEE

K w i e w . lGAp ril199 2,38, 4), pp.131-135

4

FAILLI, M.: 'COST 207 Digital land mobile radio

communications', Commission of the European

Communities.

1988. p.137

etseq.

5 PROAKIS. J.G.: 'Digital communications' (McGraw Hill, 1989,

2nd edn.), p.708

6

AGHVAMI. AH .: 'Digital m odulation techniques for mobile

and personal communication system s',Electron. Commun.

Eng.J., 1993.5 , (3) , pp .125132

7 BURR,

A.G.:

'Block versus trellis: an introduction to coded

modulation', Electron. Cummun. Eng. J., 1993,

5,

(4),

pp.24&248

8

KRIGHAM,

E.O.:

The

fast Fourier transfomi' (Prentice Hall,

New Jersey, 1974)

9 WEIN jTEIN.S.B., and EBERT, P.M.: 'Data transmission by

frequency-division multiplexing using the discrete Fourier

transform'. IEEE Trans., October 1971. COM-19, (5).

pp.62&534

10 BELL, C.P. , and WILLIAMS, W.F.: 'Coverage aspects of a

single frequency network designed for digital audio

broadcasting'. BBC Research Department Report No.

RD 1993/3

11

SHELSWELL, P., et al.: 'Digital audio broadcasting, the first

UK

field trial'. BBC Research Department Report No.

RD 1991/2

12 LE FLOCH, B., HALBERT-LASALLE,R.. CASTEWN,

R.:

'Digital sound broadcasting to mobile receivers', IEEE TYQW. ,

13 European Telecommunication Standard

EXS

300401, 'Radio

broadcast systems; Digital Audio Broadcasting (DAB) to

mohile, portable and fixed receivers'

August 1989, CE-3 5, (3), pp.493-503

EE:

1995

First received 24th February

1994

and in revised

form

3rd March

1995

The author is with the BBC Research and Development

Department, Kingswood Warren, Tadworth. Surrey, KTZO 6NP,

U K.

IBC 9 5

International

Broadcasting

Convention

14- 18

September

1995

RA/,

Amsterdam

Full programme and

registrat ion forms

available from:

IBC Office

Savoy Place

London

WCZR OBL,

UK

Tel :

+44 (0)171

240

3839

Fax: 44 0)171 497 3633

Telex: 261176 ICC LDN C

September Paper Sessions

1995 Session A Session 6

Panel Sessions

Thursday A M

14

PM

Friday A M

15

PM

Saturday

AM

16

PM

Sunday AM

17

PM

Monday

A M

18

PM

DAB technology and

markets

Digital terrestrial TV

in Europe

Digital satellite cable

and

M M D STV

in Europe

Bit rate reduction for

video

Station automation

Economics and

marketing

Commercial introduction

scenarios for digital

TV and radio services

Measurement and signal DAB

quality

R F

Digital editing

Bit rate reduction for

audio

Desk-top production for Public service

TV

and radio broadcasting

Broadband networks Satellite and

B-ISDN, ATM and the

consumer

Super-highways and full

MPEG-2

service networks implementation

Interactive services Transborder satellite

and multimedia transmission

Digital signal routing Training

and distribution broadcasters

VOD and MVOD

terrestrial services

Future trends in recording. Direct-to-home

tap e and disc satellite broadcasting techniques

Studio production Satellite distribution Picture quality

techniques

Compression

136

ELECTRONICS COMMUNICATION ENGINEERING JOURNAL JUNE 1995

sistema de modulación cofdm en audio digital

Documents