| TV TRANSMISSION | |||
|
THE
MAGICS OF TERRESTRIAL DIGITAL TV ABSTRACT Among the five terrestrial digital broadcast systems standardised during the last five years (i.e.: DAB-T, DVB-T, ISDB-T, ATSC-8VSB and the forthcoming DRM), four are based on a variant of the Coded Orthogonal Frequency Division Multiplex modulation (COFDM). The European DVB-T standard includes a large number of transmission modes to cover a wide variety of broadcast situations:
This presentation intends to highlight how these features have been made possible by the DVB-T standard and its COFDM modulation scheme. Following a tutorial introduction of the DVB-T COFDM signal, SFN operation, Hierarchical Modulation & Mobile capability are presented and commented. Why so much countries have chosen COFDM based systems? The author intends to highlight the magic's of the COFDM, which have driven these choices. INTRODUCTION The
European standard for Terrestrial Digital Video Broadcast (DVB-T)
[1], has defined a system suitable for a wide range of broadcast applications. This
versatility comes from the possibility given to adjust the modulation
parameters, then to implement up to 120 regular modulation modes and
up to 1200 hierarchical ones. The hierarchical
modulation capability has been a long time mixed up with the hierarchical
video source encoding. As a modulation scheme, it has to be considered
as an additional tool given to RF network planners, to introduce new
possibilities in the use of the scarce radio frequency spectrum. Furthermore,
the possibility given to the DVB-T to operate Single Frequency Network
(SFN) provides a tremendous simplification on the RF network planning
exercise. Moreover,
even it has not been designed to broadcast TV contents to mobile receivers,
the DVB-T has publicly shown (Broadcast Asia 98, IBC 98, NAB 99) its
suitably for this particularly difficult propagation environment. These
possibilities are being explored in this paper which, as a tutorial
introduction, presents the basic concepts of the COFDM then highlights
its assets for such various application fields. COFDM: THE FOUNDATIONS Early
in the 60's, the US's Bell Laboratories discovered the spread spectrum
techniques and the particular Orthogonal Frequency Division Multiplex
(OFDM) one, which since have been used for military applications.
Early in the 80's, the French research laboratory CCETT - Centre Commun d'Etudes en Télédiffusion et Télécommunication ( ) - has studied a modulation system sufficiently robust and efficient to carry digital data : the "Coded OFDM" (COFDM). COFDM:
WHAT DOES IT MEANS ? The basic
idea of the COFDM comes from the observation of the impairment occurring
during the Terrestrial channel propagation. The response
of the channel is not identical for each of its frequency sub-bands:
due to the sum of received carriers (main + echoes), no energy or
more than the one transmitted is sometimes received. To overcome
this problem, the first mechanism is to spread the data to transmit
over a large number of closely spaced frequency sub-bands. Then, as
some data will be lost during the terrestrial propagation, to reconstruct
them in the receiver, the data flows are encoded (i.e. : protected)
before transmission. The «
Coded » and « Frequency Division Multiplex » abbreviations
come from these two clean and simple concepts. COFDM:
HOW TO ORGANISE THE CHANNEL? The characteristics
of the transmission channel are not constant in the time domain. But,
during a short interval of time, the terrestrial channel propagation
characteristics are stable. CHANNEL
PARTITIONING
GUARD
INTERVAL INSERTION
During
the guard interval period, corresponding to an inter-symbol interference
one, the receivers will ignore the received signal. CHANNEL
SYNCHRONISATION The DVB-T
system uses « pilot » sub-carriers, regularly spread in
the transmission channel, as synchronisation markers. This is illustrated in the following figure 4.
These
different features (channel partitioning, data encoding, guard interval
and synchronisation markers insertions) constitute the basic characteristics
of the COFDM modulation. Unfortunately,
most of these features imply a lost of the channel payload or a reduction
of its useful bitrate. A contrario, they provide channel robustness. Playing
with such parameters allow to manage various trade-off between "Bitrate
& robustness". To give
as much liberty as possible to the broadcasters, then to adapt the
terrestrial transmission to each specific situation, the DVB-T standard
has defined a range of value for these parameters : their combinations
constitute the DVB-T modes. COFDM:
HOW TO CARRY DATA? The COFDM
modulation spreads the transmitted data in the time & frequency
domains, after protecting data bits by convolutional coding. As the
frequency fading occurs on adjacent frequency sub-bands, contiguous
data bits are spread over distant sub-carriers inside each OFDM symbol.
This feature, known as frequency interleaving, is illustrated in figure
5.
BASIC
CONSTELLATION Depending
on the constellation chosen, 2 bits (4QAM), 4 bits (16QAM) or 6 bits
(64QAM) are carried at a time on each sub-carrier.
But each
constellation has a dedicated robustness, in regard to the minimum
C/N tolerated for viable demodulation : roughly, 4QAM is 4 to 5 times
more tolerant to noise, than 64QAM. HIERARCHICAL
CONSTELLATION As shown
in figure 7, hierarchical modulation can be viewed as a separation
of the RF channel in two virtual circuits, each having a specific
bitrate capacity, a specific roughness and accordingly, covering two
slightly different areas.
HIERARCHICAL
MODULATION The characteristics
of the two virtual channels follow the constellation & coding
rate combinations applied. The second
one, less rugged, either in the 4QAM or 16QAM cases, is named Low
Priority stream (LP). THE
HIGH PRIORITY STREAM (HP) The companion
LP stream, introduced as an over modulation of the HP one, can be
viewed by the receiver, as an additional noise in the quadrant of
the received constellation. Then, the HP stream suffers a penalty,
in term of admissible C/N, in comparison to a regular 4QAM. There
are two ways to compensate (or to mitigate) the HP's C/N penalty:
The choice
between these two strategies will depend on the penalty the broadcaster
accepts to report on the LP stream (there will be one whatever the
chosen solution). THE
LOW PRIORITY STREAM (LP) As far
as robustness is concerned, as the LP's 4QAM or 16QAM modulation is
applied over the HP's 4QAM one, the C/N required to demodulate it
is far more important than it will be in non-hierarchical 4QAM &
16QAM modes. In fact,
the C/N required for the hierarchical LP is comparable with the one
needed for the whole constellation (i.e.: regular 16QAM or 64QAM). HIERARCHICAL
MODULATION: WHY? In many
countries, the introduction of digital TV services is performed sharing
the UHF/VHF bands with existing analogue TV services, that means using
the so-called "taboo" channels. This
situation drives to introduce digital TV services at an average power
of 15..20 dB below the analogue vision carrier one. The network
planning exercise is then focused on the optimisation of the DVB-T
channel capacity. Accordingly,
the network planners tend :
That's
why generally 64QAM constellation and 2/3 coding rate protection are
chosen. The network
type and the field coverage topologies drive to the selection of the
remaining parameters :
On top
of these choices, the hierarchical modulation allows a further refinement
for planning. In practice, the choice of hierarchical modes parameters allows to manage a range of situation between the two extreme ones shown in figure 8.
During
the early days of the DVB-T experimentation, the hierarchical modulation
has been only viewed has a way to define two coverage areas for a
given transmitter. That
is essentially true, but only essentially : two coverage areas have
to be considered only if a single category of Service is envisaged. But if
two categories of Services have to be implemented, then the hierarchical
modulation heavily facilitates their introductions in a spectrum already
occupied by the traditional analogue ones. The flexibility
offered by the hierarchical modulation can be customised in different
ways by the broadcasters, some examples are reported hereafter. BROADCAST
TO FIXED & PORTABLE RECEIVERS Fixed
receivers take benefit of the roof top antenna gain, whereas the portable
receiver are penalised by the building penetration loss. As figure
9 symbolised it, in comparison to a regular modulation mode, HP &
LP provide two distinguished coverage areas. But, as far as portable
& fix receivers are concerned, such coverage areas will not have
the same behaviour.
Practically,
what appears if a regular mode 64QAM-2/3 is converted in a hierarchical
one as HP : 4QAM-1/2 and LP : 16QAM-2/3 ?
The useful
bitrate capacity moves from 24,13 Mbps to 22,12 Mbps. The global capacity
is then reduced from 2,01Mbps. But in
terms of C/N, in a Gaussian channel, the HP protection is far better,
whereas the LP one is quasi-identical to the regular transmission
mode. Accordingly,
the overall bitrate is less reduced than the HP's robustness gain
is increased, whereas the LP's robustness (i.e. : coverage) is quasi-identical.
This
constitutes an interesting benefit for the penalised portable receivers,
as it was verified during field experimentation performed in UK by
the BBC Research Department [3]. INCREASE
THE NET BITRATE CAPACITY OF THE CHANNEL
Instead
of the regular 64QAM-2/3, the hierarchical HP : 4QAM 3/4 and LP :
16QAM 3/4 is used.
Compared
to the regular constellation, the HP's C/N becomes slightly better
(+2,8 dB) and the LP's C/N slightly worse (-2,1 dB). Then the LP's
coverage will be slightly smaller and HP's coverage slightly larger
than in a regular mode. But the
bitrate capacity moves from 24,13 Mbps to 27,15 Mbps, which represents
an noticeable gain of 3,02 Mbps. In short,
at the expense of a distortion in the overall coverage area (5 dB
penalty between HP & LP) and then a larger LP's sensitivity to
the co-channel interferers, the payload capacity of the channel will
be enlarged by several Mbps ! BROADCAST
TO MOBILE RECEIVERS The 4QAM
and 16QAM constellations seem viable if they are supported by a strong
protection. Moreover, as mobile applications are generally forecast
for Urban area (i.e.: public transport), the propagation will be characterised
by short echoes, allowing the use of a short Guard Interval. Accordingly,
DVB-T modes offering a transport capacity of 5 to 15 Mbps (1..3 TV
programmes) seem suitable to broadcast Services to Mobile receivers.
Then :
If two
RF channels are available, one can be allocated to the Mobile receivers
(roughly 2 programmes) and the other one to the traditional fix &
portable receivers (roughly 5 to 6 programmes). A further
analysis, considering the hierarchical modulation will drive to another
conclusion… Let us
choice a hierarchical DVB-T mode having a 1/16 guard interval and
a HP : 4QAM-1/2, LP : 16QAM-3/4. The bitrate capacities offered by this hierarchical mode are HP: 5,85 Mbps and LP: 17,56 Mbps, as resumed in Table 3.
This
hierarchical DVB-T mode gives, per RF channel, the capacity for one
strongly protected programme (i.e.: for Mobiles in HP stream) and
roughly, four programmes, for static reception, in the LP stream.
As two
RF channels are available, if hierarchical modulation is applied on
them, the broadcast capacity will be TWO programmes for mobile and
EIGHT programmes for fixed receivers. The conclusion
is crystal clear:
Again, it looks like the hierarchical modulation gives a significant advantage to an efficient use of the RF spectrum. SIMULCAST
OF HD AND SD DIGITAL TV FORMATS This
objective is justified by the expectation to have High Definition
screens at affordable prices, in the time scale of the Digital TV
deployment. Nevertheless,
in the interim period of the DTV services introduction, all the receivers
will not have the High Definition screen : it is then necessary to
simulcast the digital programmes both in the High Definition and the
Standard Definition formats. A hierarchical
mode using 1/16 guard interval, HP : 4QAM-3/4 and LP : 16QAM-3/4 will
offer sufficient bitrate capacities (HP : 8,78 Mbps and LP : 17,56
Mbps) to perform such simulcast. It remains
possible to decrease the HP's coding rate to a 4QAM-1/2, at the expanse
of its bitrate (HP : 5,85 Mbps) but with a gain in C/N (HP : 8,9 dB)
to improve its portable reception even to mobile reception situation. This
possibility has been publicly demonstrated, by the DVB project, during
the recent NAB 2000 convention in Las Vegas. It has
shown to the US broadcasters that the transition to the digital TV
era, is not necessarily confined to HD receivers, but can address
simultaneously two populations of receivers (HD & SD sets), then
two ranges of cost for the customers, without demanding additional
RF resources ! ANOTHER
COFDM MARVEL : THE "SFN" The advantages
of the sophisticated COFDM digital modulation are numerous, but one
of the major is its immunity against echoes. Such
echoes can be produced either by reflections on the environment (i.e.
: terrestrial channel propagation) or by several transmitters operating
in the same RF channel. In the new broadcast world made possible by the COFDM, the natural echoes produced by reflection or refraction are voluntary reinforced by active echoes sources issued from co-channel transmitters or repeaters.
That's
because the COFDM is able to use the « positive echoes »
(i.e.: the ones which increase the received power), and is able to
bypass the negative effects of the others. Accordingly,
COFDM offers to the broadcasters a new way to operate their terrestrial
networks : to increase the number of co-channel signal sources to
improve the quality of services, inside the transmission cell. With
COFDM, it becomes more efficient to use several low power transmitters
or repeaters than using a highly powered single transmitter, as this
last will never be able to avoid shadowed zones in the service area.
SFN
HOW ? To summarise
such conditions, the following « SFN Golden Rules » can
be used : Each
transmitter involved in an SFN shall radiate:
These
constraints have a direct consequence on the way to setup the primary
distribution network and the transmitters, as presented in the following
chapters. FREQUENCY
DOMAIN CONTRAINSTS But,
in the case of COFDM SFN operations, the stability and the accuracy
of the working frequency shall ensure that each radiated sub-carrier
has an absolute position whatever the RF channel frequency one. Practically,
a global frequency reference issued from GPS receivers is used to
synchronise the SFN networks. This
case is illustrated in the figure 12.
TIME
DOMAIN CONTRAINSTS The guard
interval value chosen to operate SFN has a major implication on the
network topology : as its duration governs the maximum echoes delay
admissible by the system, the guard interval value influences the
maximum distance between SFN transmitters. This
is illustrated in figure 13.
The receiver
has to setup a time window to sample the on-air signal only during
its useful period. That allows excluding the guard interval period
during which the signal is made of a mixture of two consecutive COFDM
symbols. Accordingly,
the guard interval has to be considered as a budget : it has to be
consumed on-air and not used to compensate a wrong time synchronisation
of the SFN transmitters. Practically,
network operators use the one pulse per second signal (1 pps) issued
from a GPS receiver, as an Universal time reference available in any
point of the SFN network. This
common time reference is used, at the front-end of the primary distribution
network, to insert "time-stamps" in the multiplex. As it
is common, such time reference is used, in each transmission site,
by the COFDM processor to delay the incoming multiplex until a common
launching time instant occurs. To investigate
the practical and theoretical performance limits of the DVB-T standard
for mobile reception, a consortium lead by T-NOVA, and grouping 17
broadcasters, network operators, equipment manufacturers and research
centres, founded the Motivate project. Few Motivate's
results are reported here-after. MOBILE
DVB-T : THE RECEIVERS POINT(S) OF VIEW The Motivate's
laboratory tests have demonstrated that, until a given Doppler limit
(or inter-carrier interference level), the receivers are able to perform
sufficient channel equalisation to demodulate the DVB-T signal. When
the Doppler (i.e. : the speed of the mobile) further increases, the
recovery performance decreases drastically until a point where no
demodulation remains possible. This receiver behaviour is illustrated
in figure 14.
The Mobile
behaviour curve is then characterised by a « C/N floor »,
(C/N)min, giving information about the minimum signal requirement
for a good mobile reception, and by an upper Doppler limit, giving
information on the « maximum speed » reachable by the
receiver. LABORATORY
TEST CAMPAIGNS Up to
eight receivers have been evaluated. These receivers are representative
of three generations of design (using Discrete components, 1st chipset
generation and 2nd chipset generation) and three application domains
(consumer, professional products and experimental receivers). The following figure 15 shows, as a representative example, the C/N vs Doppler characteristics obtained from the various receivers when using the "2K - 16QAM - CR : 1/2 - GI : 1/4" DVB-T mode.
The
dispersion of characteristics comes essentially from two factors :
Inside
the group of single front-end receivers, the second generation showed
better results than the first one, but an experimental receiver, which
makes use of a deep time domain filtering to perform the channel estimation,
shows also excellent results. In the
group of receivers using "antenna diversity" techniques,
two different methods have been tried. Nevertheless, whatever the
method, the diversity receivers are using two synchronised front-ends
supplied by two distinct antennas.
Even
the "antenna diversity" receivers tested were at an early
stage of design, they shown better results than the single front-end
receivers. Furthermore,
the "carrier combining" technique obtained noticeable better
results than the "packet selection", even if it has been
penalised by the use of a first generation chipset. The work
accomplished on receivers, during the last years reveals that considerable
improvements have been achieved in the channel estimation and channel
correction techniques; nevertheless room remains to further enhance
the synchronisation algorithms. Moreover,
the performance dispersion between receivers strongly highlights the
room for reception improvements offered, by the DVB-T standard, to
the manufacturers. MOBILE
DVB-T : THE NETWORK POINT OF VIEW The response
to this second question is not immediate : "Mobile reception
of a DVB-T signal is influenced by the characteristics of the terrestrial
propagation channel, which itself depends on the geographical environment
and on the robustness of the DVB-T modes used". The combination
of these factors will allow the reception of various numbers of programmes
at various maximum speeds ! DVB-T
MODES FFT
SIZE INFLUENCE
Roughly,
2K modes can cope with 4 times higher Doppler shift than 8K modes,
due to the 4 times larger carrier-spacing it produces. As the
DVB-T 2K & 8K modes offer exactly the same bitrate ranges, it
is expected that the choice of the FFT size will be mainly driven
by the network planning criteria. Then,
it is anticipated that broadcasters will firstly identify the nature
of the mobile Service area (ie : rural / urban / SFN), will then deduce
the maximum speed in this area and will choose finally the 8K or 2K
mode as a function of the transmission cell size (in MFN or SFN cases). But,
whatever the broadcaster choices, the receivers will have to provide
a hand-over function to deal with transmission cell hopping. GUARD
INTERVAL INFLUENCE In the
mobile situation, where the receivers have to cope with fast variations
of the transmission channel characteristics, the guard interval constitutes
another disadvantage : it stifles the receiver's channel estimation
with the information needed for tracking the fast channel variations. This
constitutes an additional argument in regard to the 2K vs 8K choice
: as 8K modes provides symbol having four times the duration of the
2K modes, the channel estimation is performed less often in 8K modes,
making it difficult for the receivers to compensate the fast channel
variations… but 8K modes allow greater transmission cells. Motivate
choose to focus laboratory tests on a single value of the guard interval
: 1/4. As this value is the most stringent one, the results obtained
have to be considered as the worst possible case. In other
words, networks using a shorter guard interval will enjoy an increase
of available channel bitrate whilst the receivers will achieve slightly
better mobile performance. CODING
RATE OR PROTECTION INFLUENCE Even
if it consumes useful bitrate, a strong protection is definitively
required to help the receivers to cope with the degradation experienced
in a time varying multipath channel, as is the mobile one.
This
example shows that decreasing the coding rate from 1/2 to 2/3 brings
two penalties both in terms of maximum admissible Doppler (~50 Hz)
and minimum C/N required (~5 dB)… but it provides a 1.7 Mbps
bitrate capacity gain ! Accordingly,
as usual the network planners will be faced with a trade-off between
robustness and bitrate, but in the case of the Services to mobile
receivers, this trade-off needs to be carefully weighed-up because
it will make the difference between an operational and a non-operational
system. REGULAR
CONSTELLATION INFLUENCE Accordingly,
4QAM and 16QAM modulations will give better mobile reception performance
than 64QAM which provides high bitrate capacity. Nevertheless, with
improved receivers, the use of 64QAM constellation remains possible. The figures
18A, 18B and 18C illustrate, in the CR:1/2 cases, the behaviour of
three receivers faced with the 4QAM, 16QAM and 64QAM constellations.
These
figures clearly show that a ~5 dB C/N floor penalty, accompanied by
a ~100 Hz loss in the maximum admissible Doppler frequency, occurs
when constellation density is increased. Nevertheless,
the inadequacy of the regular 16QAM or 64QAM constellations for the
mobile situation can be mitigated somewhat by the introduction of
the hierarchical modulation schemes. HIERARCHICAL
CONSTELLATION INFLUENCE In other
words, using hierarchical modulation, a single RF channel carries
two transport streams (i.e. : two MPEG-TS) having dedicated properties
and different robustness. That
means the hierarchical modes can offer the way to introduce simultaneously
Digital TV Services for static receivers and for mobile receivers.
It can also be considered to improve the delivery to portable indoor
ones. Moreover,
as soon as the HP delivery is guaranteed, the LP bitrate delivery
capability can be viewed as a gift to the network planners. This point is illustrated in the figures 19A, 19B & 19C, for three mode using a CR:1/2 cases which offer the same bitrate capacity, but are respectively obtained with a regular 4QAM, a 4QAM-HP in 16QAM and a 4QAM-HP in 64QAM.
RF
CHANNEL INFLUENCE The following
figure 20 gives a view of the variation of the speed limit (i.e. :
at C/N floor+3dB) as a function of the RF frequency channel, in the
different DVB-T modes experienced.
Figures related to the DAB (Digital Audio Broadcasting - ETS 300 401) have been included as this system also takes advantage of the COFDM modulation scheme. The laboratory
tests made use of the UHF channel 40 (626 MHz). This channel is roughly
situated in the middle of the DVB-T Services band. This
implies that the C/N performances reported in this document are average
in regard to the various usable RF channels. Better
performances are generally obtained when the lower part of the broadcast
bands is used, whilst worse performance occurs when the upper part
of the broadcast bands is used. DVB-T
MOBILE RECEPTION PERFORMANCE They
highlight :
The figure
21 shows the maximum Doppler frequency, for the three channel profiles,
in each DVB-T mode.
For a given bitrate, it is then possible to comprehend the merit of each DVB-T mode in regard to the maximum Doppler (that means maximum speed) the receivers will be able to accept, in the environments modelled by the channel profiles.
The figure
22 shows the maximum speed, as it will be when using CH40 (626 MHz),
for the three profiles, in each DVB-T mode. The histogram shows the
various bitrate values, in a given range, as a function of the Guard
Interval chosen (N.B. : the GI:1/4 is the worst case both for the
receiver and for the channel bitrate capacity). The superimposed
segment gives the indication of the maximum speed as a function of
the three channel profiles, when GI:1/4 is used. The higher
speed value referred to the "two echoes" profile whilst
the lower one is related to the typical rural area one; the point
shows the maximum speed obtained for the typical urban profile. This last figure allows to compare the relative merit of each DVB-T mode, in terms of maximum mobile speed and allows to weight-up the acceptable speed range, in relation with the geographical environment where the mobile digital TV Service is delivered. At the
end of the Motivate's project life-time, the following conclusions
can be drawn :
In the
early days of the DVB-T standard deployment around the world, the
hierarchical modulation feature has not very much attracted the broadcasters.
Currently, things and thoughts evolve to place the hierarchical modulation
in an attractive situation. Hierarchical
modulation constitutes one of the numerous assets of the DVB-T standard,
allowing another kind of spectrum efficiency usage by addressing various
categories of receivers in various situations, without demanding additional
RF resources. This
liberty given to the broadcaster to start digital TV by introducing
various categories of services, without demanding additional spectrum,
is viewed as a new definitive advantage for the DVB-T system. SFN operation
is definitively not a laboratory-interesting feature; it is an efficient
way to operate broadcast networks on the field. Motivate
did not intend to prove that the DVB-T standard is today the only
international system able to broadcast digital TV to mobile receivers
- obviousness has not to be demonstrated (!) - but evaluated the limit
of the DVB-T standard to deliver Digital TV to mobile receivers. It appears
clearly today that the new generation of receivers improves spectacularly
the mobile DVB-T usability. In addition, the receivers using antennas
diversity techniques bring further capability to DVB-T mobile reception.
On the
network side, the usability limits of the DVB-T modes have been characterised
: providing a strong coding rate (i.e. : 1/2, 2/3) a wide range of
bitrates (i.e. : 5 to 15 Mbps) can be broadcast to receivers in motion
at various speeds (i.e. : from 50 to 500 km/h). The traditional trade-off
between "robustness and bitrate" must now include "robustness,
bitrate, area type and speed". At least,
the author hopes that this presentation has definitively convinced
the broadcasters that using the DVB-T standard allows a tremendous
variety of Digital TV applications to be delivered to a large range
of receivers in a wide range of situations. BIBLIOGRAPHY Gerard FARIA (gfaria@harris.com) - 2001 |
|
||
| Copyright 2000, Digital TV Australia |