ÐÓ°ÉÔ­´´

Imagine driving the length of Britain, over the Channel and across Europe,
listening all the time to the same radio station. The sound is in digital
stereo, which gives it the same quality as that from a compact disc. There
is no interference, and none of the fading and fluttering that normally
blemish reception as you drive past tall buildings, over hills and down
valleys. There is no need to keep retuning the radio because the chosen
station remains on the same frequency throughout Europe – although, of course,
you could retune to alternative national, international or local stations
if you wanted to.

All this and more will be possible with a new telecommunications system
called digital audio broadcasting. DAB is a jigsaw of broadcasting technologies
that is nearing completion after just four years of collaborative research
by Europe’s telecommunications industry. Next week in Birmingham, at the
annual Radio Festival of the Radio Academy, the BBC will demonstrate the
latest working prototype, the result of Eureka research project 147.

Companies participating in EU-147 include electronics manufacturers
AEG, Bosch, Grundig, ITT and Philips, along with broadcasting and telecommunications
organisations such as the BBC in Britain, the Centre National d’Etudes des
Telecommunications in France and the Bundespost in Germany. The European
group claims that its DAB system is unrivalled, and points to the fact that
the National Association of Broadcasters in the US wants to buy rights and
make the system a standard in North America. The only remaining obstacles
are administrative.

‘We have solved all the technical problems,’ says Egon Meier-Engelen,
manager of the project, who works for the German Aerospace Research Establishment
(DLR) in Cologne, which coordinates EU-147. ‘The only remaining challenge
is to find a suitable radio spectrum to carry the service.’

Work on DAB began in 1987 and has already cost $50 million as researchers
have rushed to prove the system in time for the next World Administrative
Radio Conference, to be held in Spain next February. The International Telecommunications
Union, based in Geneva, organises the conference every four years to allocate
radio frequencies. The DAB team wants to use the lowest radio frequency
bands still unallocated, which are around 1500 megahertz, because the atmosphere
disrupts terrestrial signals transmitted at higher frequencies. If the WARC
administrators agree, the system could be in service by the mid-1990s. If
the Eureka researchers miss their deadline, the frequencies they want will
be allocated for something else, such as satellite radio telephones or missile
guidance systems, and the DAB system will be much harder to implement.

DAB is a development of the technology that has enabled broadcasters
to transmit digital stereo into homes to improve the quality of television
sound. The technology, known as NICAM (Near-Instantaneous Companded Audio
Multiplex), was developed by the BBC in the mid-1980s. In NICAM broadcasts,
the digital code for stereo sound is transmitted along with conventional
analogue picture and mono sound signals, but at slightly higher frequencies.
Existing TV receivers ignore the digital signal while new NICAM receivers
decode it to reproduce stereo sound. Satellites that broadcast direct to
homes in Europe also transmit digital stereo.

But the NICAM system does not overcome the problem of poor reception
that dogs conventional analogue broadcasts to receivers in valleys and built-up
areas, or to radios in cars on the move. The problem is caused by a phenomenon
known as multipath. This occurs when a receiving aerial picks up direct
signals from the transmitter plus replicas of them delayed by reflection
from a building or a hill. Unless the direct and reflected waves are in
step, or ‘in phase’, they combine to reduce the strength and quality of
the signal.

On an analogue TV picture, multipath causes ghost images on the screen;
on an analogue radio, fading of the sound; and on analogue car radios, the
sound flutters as the phase relationship changes as the vehicle moves towards
or away from the sources of the reflections. The effect is much more noticeable
with digital transmissions: the receiver becomes so confused by the mixture
of direct and delayed information that it is unable to produce any useful
sound at all. Instead, it makes very unpleasant crackling noises, or goes
silent.

For stationary audio receivers, such as TVs and radios in the home,
auxiliary relay transmitters and aerials pointed at a transmitter can enhance
reception. These measures are of no practical use for mobile receivers,
however. Auxiliary transmitters must operate on a different frequency from
that of the main transmitter, and car radios, for instance, cannot easily
and cheaply change frequency to stay tuned to the same programme as they
move between the reception areas of different transmitters. Also, an aerial
on a moving car will not remain pointed in the right direction for long.

These factors persuaded broadcasters that the NICAM system used to transmit
audio signals for TVs was unsuitable for radio transmissions, because so
many people listen to the radio only when they are travelling by car. If
listeners to car radios are to appreciate the benefits of digital stereo
broadcasts, transmissions must be made immune to multipath. It was this
analysis that led the European telecommunications industry to pursue the
development of DAB.

The only way to make a digital radio system immune to multipath is to
reduce the rate at which the digital information is transmitted – the ‘bit
rate’ – to very low speeds so that there are long gaps between consecutive
bits. Any bits that arrive during what is supposed to be a gap in transmission
are recognised by the receiver as reflections, and it rejects them. But
reducing the bit rate risks poorer sound quality, because less information
is being transmitted and received each second. The task of the Eureka researchers
was to find a way of compressing the digital signal – that is, selectively
reducing the bit rate – in such a way that sound quality is not impaired.

To produce digital stereo sounds for studio and CD recordings, the analogue
waveform of each of the two mono channels is chopped up or ‘sampled’ over
44 000 times a second. Each sample is then described in a 16-bit digital
word. The result is a very rapid stream of data, around 1.5 million bits
a second, which is too much to broadcast. The NICAM system copes by halving
the rate and still manages to produce good quality sound. But, as tests
by the BBC have shown, NICAM’s data rate remains far too high for decoders
on the move, in car receivers, to handle.

Digital audio broadcasting achieves a much more drastic reduction in
the bit rate in two ways. First, it takes advantage of a phenomenon of human
hearing known as masking, which allows the system to discard sounds that
a listener would not notice anyway. Secondly, it transmits the reduced signal
over many channels to reduce the bit rate of each one.

When two sounds of similar frequency exist together, the ear hears only
the louder one; the quieter one is masked. This is what happens in a musical
chorus where one strong singer can steal the attention from weaker singers.
Analogue tape noise reduction systems, such as Dolby, already exploit this
effect. Tape hiss, which is random high frequency noise, is masked by louder
musical sounds of similar frequency, such as violins. As a result, Dolby’s
electronic filters need to mute the hiss only when no music is playing,
which means there is then no risk of the filters dulling high frequency
music.

Five years ago the Institut Fur Rundfunktechnik, the German radio research
centre in Munich, applied masking to digital signals and developed a system
known as MASCAM (Masking-pattern Adaptive Sub-band Coding and Multiplexing).
The general principle is that quiet sounds on their own are accurately coded;
but where the quiet sounds are masked by louder sounds of similar frequency,
only the louder sounds need to be coded. The result is a reduction in the
number of bits by as much as a factor of 10.

The system depends on a computer working in ‘real time’ to predict what
sounds will be masked to the human ear and what sounds will be heard. As
everyone’s ears are different, and the sounds of music are infinitely variable,
there can be no cast iron rules to guide the computer. The technology is
continually developing as researchers develop new algorithms and programs
that more closely reflect how we perceive sounds.

Philips’s new digital compact cassette (DCC), which is due to go on
sale next year, relies on this same masking trick to reduce the bit rate
needed for hi-fi stereo to a level that can be recorded on a simple tape
cassette. Philips acknowleges that DCC’s coding system, known as PASC (Precision
Adaptive Sub-band Coding), derives from the DAB research pro-ject. It claims
that PASC gives the compact cassettes the quality of CDs for a quarter of
the bit rate. Sony has developed the system further for its Mini Disc recorder,
which compresses data by a factor of five. But even the reduced volume of
bits on the compact cassettes or Mini Discs is still too great for moving
decoders to cope with.

For DAB, the Eureka researchers looked at two rival systems that have
developed from MASCAM. The French government’s radio and communications
research centre in Rennes, CCETT (Centre Commun d’Etudes de Telediffusion
et de Telecommunications), offers MUSICAM (Masking-pattern Adaptive Universal
Sub-band Integrated Coding And Multiplexing). ASPEC (Adaptive Spectral Entropy
Coding) is a joint proposal from AT&T Bell Laboratories, Thomson Consumer
Electronics, Germany’s Fraunhofer Society and CNET (France’s Centre National
d’Etudes des Telecommunications). Both systems split the audible frequency
range into 32 narrow sub-bands, analyse their sound content and try to predict
which sounds will be masked to the human ear and which sounds need coding.
The differences between the systems are in the methods of analysis and in
the software’s assumptions about what the ear will hear.

In an effort to determine which system is the better one, the International
Standards Organization asked the Swedish Broadcasting Corporation to run
comparative tests. Last July in Stockholm, a panel of 60 listeners heard
the same pieces of music, through headphones and loudspeakers, after different
degrees of compression by both MUSICAM and ASPEC. The music used in the
tests was deliberately chosen to try to confuse the electronic systems.
For instance, not even humans can find a logical, predictive thread in the
playing of the American avant-garde saxophonist Ornette Coleman.

For the tests, each of the two mono channels sent data at a rate of
768 kilobits per second, without reduction, into a prototype DAB encoder.
With the data compressed by a factor of 12, to 64 kilobits per second, most
people judged that both systems degraded the sound. At 128 and 96 kilobits
per second, listeners heard differences that varied from system to system,
depending on the music. This led the Eureka researchers to try to merge
the best features of the two rival systems and stage a new series of tests,
which were completed in May. The results of the latest trials are due to
be incorporated in the demonstration of digital audio broadcasting at the
Radio Festival next week.

Meier-Engelen pledges that whatever compression system the Eureka team
finally adopts, it will be flexible enough to allow the telecommunications
industry to take advantage of future developments in coding technology without
making existing consumer products obsolete. This should mean that the same
receiver will be able to cope with incoming data rates of 192, 128, 96 and
64 kilobits per second.

For the moment, however, the minimum bit rate for unimpaired sound quality
is not low enough to make DAB immune to multipath interference. So the Eureka
researchers have found a way of achieving a further reduction that is more
apparent than real. Instead of transmitting all the digital bits needed
for a sound signal in a single radio channel, they spread the bits over
a large number of channels of closely spaced frequencies. Known as COFDM
(Coded Orthogonal Frequency Division Multiplex) and devised by the French
research centre CCETT, the technique is analogous to sending a signal down
a large number of thin parallel wires, instead of a single large cable.

Each channel carries only a small number of digital bits, so there are
long gaps between them. After receiving one directly transmitted bit, the
receiver ignores any signals, such as unwanted reflections, that arrive
in the long gap before the next directly transmitted bit is due. Additionally,
some digital bits are sent by more than one signal path. So even if one
path suffers serious interference, the lost bits still get through on another.

The Eureka researchers originally planned to spread 16 stereo radio
programmes over 500 separate channels. After masking compression, which
would reduce the bit rate of the signal by at least one-sixth, this plan
would further reduce the volume of data per channel to around 10 kilobits
per second. However, the researchers have now found a way of overlapping
adjacent channels of similar frequency so that the programmes can be spread
over 2000 of them. This reduces the data stream to a few kilobits per second
and renders the DAB system immune to reflections from buildings or hills
up to 5 kilometres away.

For DAB, broadcasters want to group at least 16 stereo radio programmes
together in slices of the radio spectrum, each with a frequency bandwidth
of 7 megahertz. This bandwidth is equivalent to that allocated for a single
analogue TV channel. Each stereo sound channel would be accompanied by at
least 4 kilobits per second of extra data that could be displayed on small
liquid crystal screens close to the radio receivers. The additional information
could provide news, traffic flashes, or details about the radio station
or music being played.

Immunity to multipath brings an unexpected bonus. Car radios are liable
to interference that sounds like multipath when they receive broadcasts
from two transmitters operating on the same frequency. As the car moves,
changing its relative distance to the two transmitters, the received signals
move in and out of phase, causing fluttering and fading of the reception.
This is why broadcasters must use different frequencies for neighbouring
radio transmitters and why, in turn, motorists must continually retune their
radios as they drive from one region to another.

Because the DAB system is so resistant to multipath interference, it
can also cope with the areas of overlap between different transmitters operating
on the same frequency. This means that a DAB network can operate across
the entire country on a single frequency. The receiver locks onto the strongest
signal and rejects any weaker signals from other transmitters as if they
were reflections.

This opens up another exciting possibility for DAB, which is out of
the question for analogue radio transmissions. DAB signals could be broadcast
from a satellite focused on individual countries, or on states right across
a continent. Receivers on the ground would pick up signals either direct
from the satellite or from relay stations on the ground, which would rebroadcast
them on the same frequency. With current analogue systems, relay stations
must always operate on different frequencies from the main signal to prevent
interference.

The European Broadcasting Union in Geneva envisages an integrated broadcasting
system in which satellites serve large areas by direct transmission, with
areas between buildings and in valleys that are ‘shaded’ from the satellite
served by relay stations. Cars would carry omnidirectional aerials that
would pick up the strongest signal available, whether from satellite or
relay transmitter.

Europe’s telecommunications industry is considering the possibility
of splitting up the 7-megahertz slices of bandwidth, which it expects to
be allocated to the DAB system, into four separate blocks. This would let
neighbouring countries draw radio boundaries. Just as only four colours
are needed to separate a jigsaw of territories, so only four blocks of frequencies
are needed to ensure that no two adjoining countries share the same radio
bandwidth.

While American and European broadcasters agree that the ideal frequency
range for DAB of those still unallocated is in the L-band at around 1500
megahertz, the National Association of Broadcasters in the US is vehemently
opposed to the idea of broadcasting digital radio from satellites, which
Europe favours. The NAB represents the 11 000 local radio stations that
currently serve the US, most of which cover only small areas of the country
and rely on advertising to stay in business. If satellites blanket large
areas, many local stations are likely to go out of business as advertisers
switch their allegiance to the broadcasters with the biggest audiences.
Meier-Engelen says that satellite transmission is inevitable: ‘NAB cannot
avoid this. It will happen. The Canadians will use satellites anyway.’

The European Space Agency is promoting the idea of a DAB service fed
by satellite transmissions. However, conventional geostationary spacecraft
will not be used, because they orbit the equator and thus appear to hang
very low in the sky the farther from the equator you go. In northern Scandinavia,
for instance, geostationary satellites appear just 10 degrees above the
horizon and signals from them are often shielded by hills, buildings and
even trees.

Two years ago ESA commissioned a report from British Aerospace on the
use of satellites with a completely different kind of orbit for DAB. These
are the highly elliptic inclined orbit satellites that were pioneered in
the Soviet Union to serve its northernmost territories. HEO satellites behave
as if they are hovering almost vertically above selected regions and are
thus ideal for broadcasting directly into cities and mountainous terrain.

The ESA-BAe plan, code-named Archimedes, is for four spacecraft following
highly elliptical paths to orbit the Earth at six-hour intervals. Like balls
tossed in the air by a juggler, each spacecraft rises very steeply into
space over the targeted land area and then falls, equally as steep, back
towards the Earth. The craft skirt the other side of the Earth at low altitude
before starting another orbit. At any given time there is always at least
one satellite in a position to broadcast signals almost vertically down
to the targeted area. At worst, predicts British Aerospace, the angle of
elevation is 60 degrees. Tests with transmitters in aircraft, commissioned
by Britain’s Department of Trade and Industry and carried out by the University
of Bradford, indicate that terrestrial relays will be needed in only a very
few city areas.

Like any digital system, the DAB receiver need only distinguish digits
from background noise to produce a perfect music signal. This means the
transmitted signal can be weaker, which suits satellite systems that rely
on solar panels for power. BAe estimates that the DAB transmitters on an
HEO spacecraft will need to draw around 6 watts of power for each digital
stereo channel, compared with the 50 watts required for an analogue satellite
TV channel. At that rate of power consumption, each Archimedes satellite
could broadcast up to 100 stereo programmes simultaneously.

The EBU hopes that ESA will be able to launch the first pair of HEO
satellites in 1995 with two more following a year later. This is an optimistic
target; the EBU’s fallback position is that experimental DAB services should
begin as soon as possible, using spare frequencies in the terrestrial VHF
and UHF radio and TV bands. Tests by the CCETT at Rennes show it is possible
to transmit DAB signals in the frequency gaps that exist between terrestrial
TV services.

For once, Japan is streets behind Europe. Japan’s Ministry of Posts
and Telecommunications has licensed six consortia to provide a national
digital radio service. But the Japanese broadcasters use conventional digital
coding, similar to that used in Europe to provide the sound for satellite
TV. Programmes can only be received with an accurately aligned satellite
dish and the Japanese system will not work in cars. If, as looks increasingly
likely, Europe and North America adopt a common standard for DAB, Japan
will be obliged to swallow national pride, accept the European system and,
for a change, pay royalties to the West.