OFDM

EBU TECHNICAL REVIEW July 2003 1 / 12

R. Brugger and D. Hemingway
Roland Brugger
IRT
David Hemingway
BBC
This article offers a general overview of the possible strategies for FFT window
synchronization in OFDM receivers. These strategies are equally applicable to the T-
DAB and DVB-T broadcasting systems.
The digital broadcasting systems DVB-T and T-DAB commonly exploit the single-frequency network (SFN)
technique. In such networks, the treatment of inter-symbol interference and the synchronization strategy of
the receivers is a crucial aspect in planning the networks and achieving the desired coverage. This is valid in
two senses.
! Firstly, the performance of the receiver is strongly dependent on the way in which it positions the FFT
window relative to the several received signals that can be present in a multipath environment or in an
SFN. The position of the FFT window affects the receivers behaviour with regard to inter-symbol inter-
ference (ISI).
! Secondly, the modelling of receiver behaviour in network coverage simulations has to be in line with the
synchronization strategies and the treatment of inter-symbol interference in the receivers, in order to give
reliable coverage predictions. However, the synchronization strategies of individual manufacturers are
commercially sensitive and not publicly available. Therefore, predictions have to be carried out on the
basis of assumptions.
1. Inter-symbol interference
1.1. General
In OFDM, the information is carried via a large number of individual carriers in a frequency multiplex. Each
carrier transports only a relatively small amount of information, and high data capacities are achieved by using
a large number of carriers within a frequency multiplex. The individual carriers are modulated by means of
phase-shift and amplitude-modulation techniques. Each carrier has a fixed phase and amplitude for a certain
time duration, during which a small portion of the information is carried. This unit of data is called a symbol;
the time it lasts is called the symbol duration. After that time period, the modulation is changed and the next
symbol carries the next portion of information. Modulation and demodulation are accomplished by the use of
Inverse Fast Fourier Transformation (IFFT) and Fast Fourier Transformation (FFT) respectively. The symbol
duration time is the inverse of the carrier spacing, ensuring orthogonality between the carriers.
OFDMreceivers
impact on coverage of inter-symbol interference
and FFT window positioning
OFDM
EBU TECHNICAL REVIEW July 2003 2 / 12
R. Brugger and D. Hemingway
In general, signals arriving at a receiver by different paths show different time delays which result in inter-
symbol interference (ISI), a degradation in reception. An OFDM system with a multipath capability allows
for the constructive combination of these signals. This is achieved by inserting a guard interval a cyclic
prolongation of the useful symbol duration of the signal. The FFT window, i.e. the time period for the OFDM
demodulation, is then positioned in such a way that a minimum of inter-symbol interference occurs. This
mechanism as far as it is of interest for coverage predictions and network planning is described in more
detail in the following paragraphs.
In order to demodulate the signal and
looking at only one carrier the receiver
has to evaluate the symbol during the
symbol duration. Three consecutive sym-
bols in time, denoted by n1, n and n+1,
and the setting of the FFT window such
that symbol n is evaluated by the receiver,
are shown in Fig. 1. No guard interval is
used in this example, and the FFT win-
dow has the same duration as the symbol.
In an environment where several useful
signals either from multipath echoes or
from other transmitters in an SFN are
available to the receiver, things become
more complex. Usually, the signals arrive
at different times at the receiver which, in
the absence of a guard interval, makes
correct synchronization to all the signals
impossible. Such a situation, with two
signals as an example, is depicted in
Fig. 2. Synchronization to symbol n of
signal 1 leads to an overlap of the FFT
window with the preceding symbol n1 of
the delayed signal 2. Since this symbol
n1 carries different information from
symbol n, the overlap acts as inter-symbol
interference to the evaluation of symbol n.
In order to overcome the inter-symbol interference problem in DVB-T and T-DAB, part of the symbol is cop-
ied from the beginning of the symbol to the end, increasing its duration by a certain amount of time called the
guard interval. This cyclic prolongation of the original symbol is shown in Fig. 3. The guard interval is
denoted by .
The new increased symbol duration is denoted by T
s
and the original symbol duration is often called the useful
symbol duration T
u
. The duration of the FFT window during which the symbol is evaluated is kept at the orig-
inal value T
u
. The orthogonality relationship is kept with the original symbol duration T
u
, not the extended T
s
.
The improvement that is achieved by the insertion of the guard interval can be seen from Fig. 4 with two sig-
nals as an example. The guard interval now allows for the FFT window to be positioned so that there is no
overlap with a preceding or subse-
quent symbol, thus avoiding ISI.
The fact that the duration of the FFT
window is now smaller than the
symbol duration allows for a variety
of different possible FFT window
positions for the evaluation of a sym-
bol. This is indicated in Fig. 5 for
the simple case of synchronization to

FFT window
Time
Symbol n1 Symbol n+1 Symbol n Signal
FFT window
Time
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n Signal 1
Signal 2
ISI

Extended symbol Original symbol

T
u
T
s
= T
u
+
Figure 1
OFDM symbol duration and FFT window (no guard interval)
Figure 2
Inter-symbol interference with a delayed signal (no guard
interval)
Figure 3
Increase of the symbol duration by the guard interval
Figure 4
Guard interval utilisation

T
u
Signal 1
FFT window
Time
Signal 2
Symbol n
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1
OFDM
EBU TECHNICAL REVIEW July 2003 3 / 12
R. Brugger and D. Hemingway
a single signal. Three possible FFT
window positions are indicated as
examples. Here, all positions are
equivalent with regard to evaluation
of the symbol, because all the FFT
window positions shown include sam-
ples from only one symbol.
The insertion of the guard interval
reduces the data capacity because not
all of the symbol duration T
s
is used for useful data.
In a multipath or SFN environment, where many potentially useful signals are available to the receiver, the
choice of the FFT window position becomes more complex. A number of different strategies that can be
applied are discussed in Sections 2 and 3.
All signals with time delays that cannot be absorbed by the guard interval, in the way described above, intro-
duce a degradation of reception, similar to that shown in Fig. 2. Any part of a received signal that falls outside
the guard interval has an interfering characteristic which is different for T-DAB and DVB-T due to the dif-
ferent demodulation methods applied (differential in the case of T-DAB and coherent in DVB-T).
1.2. T-DAB
For T-DAB network planning, the power of
all the echoes received within a window of
duration (guard interval width) is consid-
ered as useful, and contributes positively to
the total available signal power. Outside the
guard interval, a part of the echo power is
associated with the same OFDM symbol as
the primary signal, and therefore contributes
positively to the total useful signal power.
Another part of the echo power is associated
with the previous or subsequent OFDM
symbol and produces ISI, which has a simi-
lar effect to uncorrelated Gaussian noise
interference. Therefore, as the echo delay is
progressively increased beyond the guard
interval, the useful contribution decreases
and the ISI increases with a quadratic law.
The echo power becomes fully interfering (i.e. it contains no useful power) when the delay is larger than or
equal to one OFDM symbol (see Fig. 6).
Mathematically, the rule for splitting the signal power into a useful component and an interfering component is
expressed by the Equation 1, at the top of the next page.
It must be borne in mind that I, the total effective interfering power, is weighted by the established T-DAB-to-
T-DAB protection ratio, when being regarded as a source of interference in a coverage calculation.
1.3. DVB-T
In the case of DVB-T, because of the pilot carriers that are needed for coherent demodulation, the total loss of
constructive signal components occurs beyond a relative delay of T
p
= T
u
/3. The failure of the equalisation
algorithm after T
p
(rather than T
u
+ as in the case of T-DAB) produces a more rapid performance degrada-
tion versus the echo delay, than in T-DAB (see Fig. 7). The T-DAB model (see Fig. 6) that splits the echo
FFT window position 1
Time
Signal Symbol n1 Symbol n+1 Symbol n
FFT window position 2
FFT window position 3
Figure 5
Three possible FFT window positions
t
-T
u
0 T
u
+
Useful component
Interfering component
Relative weighting, w
i
Figure 6
T-DAB model splitting of the signal power into useful
and interfering components
OFDM
EBU TECHNICAL REVIEW July 2003 4 / 12
R. Brugger and D. Hemingway
power into a useful and an interfering contribution is thus only applicable in the case of DVB-T for echo
delays up to T
p
, with echoes outside T
p
contributing only to the interfering power, independent of their delay.
The corresponding formula is given by
Equation 2, at the bottom of this page.
Again, as for the T-DAB case, it should be
remembered that I, the total effective inter-
fering power, is weighted by the appropriate
DVB-T-to-DVB-T protection ratio, when
being regarded as a source of interference in
a coverage calculation.
A value of T
u
/3 is regarded as a theoretical
limit for T
p
and would require an interpola-
tion filter with an infinite number of taps.
The formula T
p
= 7T
u
/24 is often quoted and
this gives a sensible practical limit in the
case of real filter designs. At the present
time, many DVB-T receivers do not even
reach this performance.
t
Useful component
Interfering component
Relative weighting, w
i

-T
u
0 T
u
+ T
p
Figure 7
DVB-T model splitting of the signal power into useful and
interfering components.
i i
i
i
i i
u
u
u
u
u
u
u
u
i
C w I
C w C
T t if
T t if
T
t T
t if
t T if
T
t T
T t if
w
) 1 (
0

) (
0 1
0
0
2
2
!
!
=
=
"
"
"
"
"
#
"
"
"
"
"
$
%
+ >
+ <
&
&
'
(
)
)
*
+ +
<
<
&
&
'
(
)
)
*
+ +

=
Where:
C
i
= the power contribution from the i-th signal at the
receiver input
C = the total power of the effective useful signal
I = the total effective interfering power
w
i
= the weighting coefficient for the i-th component
T
u
= the useful symbol length
= the guard interval length
t = the signal arrival time
Equation 1
Where:
C
i
= the power contribution from the i-th signal at the
receiver input
C = the total power of the effective useful signal
I = the total effective interfering power
w
i
= the weighting coefficient for the i-th component
T
u
= the useful symbol length
= the guard interval length
t = the signal arrival time
T
p
= the interval during which signals usefully con-
tribute
i i
i
i
i i
p
p
u
u
p
u
u
p
i
C w I
C w C
t T if
T t if
T
t T
t if
t T if
T
t T
T t if
w
) 1 (
0

) (
0 1
0
0
2
2
!
!
=
=
"
"
"
"
"
#
"
"
"
"
"
$
%
<
<
&
&
'
(
)
)
*
+ +

<
&
&
'
(
)
)
*
+ +

=
Equation 2
OFDM
EBU TECHNICAL REVIEW July 2003 5 / 12
R. Brugger and D. Hemingway
DVB-T
DVB-T
T-DAB
T-DAB
2. FFT window synchronization
The synchronization of an OFDM receiver
is performed in two stages:
! initial synchronization in which the
receiver is aligned with the symbol
rate, and;
! secondary synchronization in which
the receiver positions the FFT win-
dow to demodulate the signal.
The initial synchronization is normally
done by correlating samples taken T
u
apart
in time. When the waveform repeats, as
shown in Fig. 8, the correlator output
exceeds a threshold value. From this, the
receiver can detect the start of a new symbol period.
In a real multipath environment, the receiver encounters a multitude of echoes which complicates the task of
the second-stage synchronization process, i.e. finding the best position for the FFT window. As a conse-
quence, various strategies can be applied in order to optimize the receiver performance.
A difference arises from the distinction between direct signals and echoes. In an MFN, where each transmitter
acts independently on its own channel/frequency, the receiver may get one direct signal and a number of scat-
tered echoes. The direct signal is not necessarily the strongest signal nor is there necessarily a direct signal at
all, particularly in the case of portable or mobile reception. On the other hand, there are also cases where there
is only the direct signal present. In an SFN, all transmitters in the network use the same channel/frequency. In
this case, the receiver gets a number of direct signals
and a number of scattered echoes.
It is necessary to distinguish between different imple-
mentations of synchronization strategies in real
receivers. These strategies are outlined in Section 3.
Most coverage prediction methods use two dimen-
sional prediction models, taking into account only the
direct path. Therefore in an MFN, the modelling of the FFT window positioning is simple and unique since
there is only one direct path present. In an SFN, receiver synchronization modelling is no longer unique since
there are usually several direct-path signals present.
In some three-dimensional prediction models, a multipath propagation environment for each transmitter is
considered. Therefore the FFT window positioning for an MFN becomes as complex as that for an SFN.
In planning simulation tools, a natural way to describe the reception situation would be to model real receiver
behaviour. Unfortunately, the receiver FFT window positioning is not prescribed in detail in either the T-DAB
or DVB-T system specifications. This means that
all manufacturers have their own solutions and,
moreover, they regard these various solutions as
confidential making a single description of
receiver FFT window positioning difficult.
A further difference arises from the fact that real
receivers have to account for the time variation of
the transmission channel, whereas software modelling of the receiver FFT window positioning usually
assumes a static reception situation. (This, to some extent, is justified by the different time scales of succes-
sive synchronization instants and the time variation of shadow fading in a transmission channel.) This means
that a real receiver will not show exactly the same synchronization behaviour as that described in the simple
model cases below. Real receiver synchronization has to deal with all existing echoes in the multipath envi-
ronment, while receiver modelling usually does not.
Figure 8
Initial receiver synchronization (image courtesy of Philips Ltd)
OFDM
EBU TECHNICAL REVIEW July 2003 6 / 12
R. Brugger and D. Hemingway
3. Synchronization strategies
3.1. General
This section describes five different strategies for second-stage synchronization (i.e. positioning of the FFT
window) that are commonly used in receiver modelling. Four of them are relatively simple and straightfor-
ward strategies, while the fifth is an idealised optimal strategy.
The strategy employed by a receiver determines:
! which peak, in the time-domain impulse response of the received signal, the receiver uses for synchroni-
zation;
! where the receiver sets the FFT window relative to this peak.
In a single-signal environment, the synchronization configuration is simple and clear. The principle was
already explained in Section 1 and can be seen from, for example, Fig. 4.
FFT window synchronization is of particular importance for mobile and portable reception, when the
receiver will need to be able to synchronize in a rapidly changing environment and in the presence of pre-
and post-echoes.
3.2. Strongest signal
A natural approach for the FFT window positioning is to
synchronize to the strongest signal, in a similar way to
that shown in Fig. 4 for a single signal. In order to dem-
onstrate the principle, a configuration with four signals is
chosen as an example. Fig. 9 shows the channel response
function for the configuration, where the peaks represent
a characteristic time instant of the signals, such as the
start of symbol n.
Signal 3 is the strongest signal. Accordingly, the FFT
window is synchronized to signal 3. Since relevant con-
tributions of further signals may be found preceding
signal 3 or following signal 3, it seems reasonable to
locate the centre of the FFT window at the centre of symbol n of signal 3. This is depicted in Fig. 10. In the
example, signals 3 and 4 contribute fully to the evaluation of symbol n, whereas the FFT window exhibits an
overlap with symbol n+1 of the signals 1 and 2, which results in a certain amount of ISI.
A more sophisticated synchronization strategy, based on the strongest signal approach, would not be fixed to the
centre of the symbol duration but would check for better positions within the symbol duration of the strongest
signal. In the chosen example, it would be advantageous to move the FFT window a tiny bit backwards in time
to avoid the small amount of ISI arising
from the overlap with symbol n+1 of
signal 2. Also, the inter-symbol inter-
ference from signal 1 would be
reduced.
3.3. First signal above a
threshold level
This strategy takes the first signal of
the time impulse response as a refer-
ence for the FFT window. Normally, a
minimum threshold level is necessary
for a signal in order to be accepted as a
Signal 1
Signal 4
R
e
l
a
t
i
v
e

a
m
p
l
i
t
u
d
e
Time
Signal 2
Signal 3
Figure 9
Synchronization to the strongest signal
(signal 3); impulse response
Signal 1
Signal 2
FFT window
Time
Symbol n1
Signal 3
Symbol n+1
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Signal 4
ISI
Figure 10
Synchronization to the strongest signal (signal 3); FFT window
position
OFDM
EBU TECHNICAL REVIEW July 2003 7 / 12
R. Brugger and D. Hemingway
trigger. Again the 4-signal configuration of
the previous section is taken as an example.
The impulse response is given in Fig. 11
with the threshold value indicated by a hor-
izontal dashed line.
The first signal above the threshold is
signal 2. It serves here as the trigger for the
FFT window. If the threshold is chosen
reasonably, it can be expected that there is
no significant signal preceding signal 2
and, therefore, it is logical to align the end
of the FFT window with the end of the
symbol n of signal 2. This is indicated in
Fig. 12.
With this synchronization strategy,
signals 2, 3 and 4 contribute fully con-
structively, whereas signal 1 adds a
certain amount of ISI.
The choice of the threshold value is a
specific issue of this synchronization
strategy. It may be taken as the power
corresponding to the minimum field
strength or, more pragmatically, as a
value, say 6 to 10 dB, below the
strongest signal.
In a recent workshop, EICTA has indi-
cated that various manufacturers apply
the first signal above a threshold
level, or a strategy similar to it.
3.4. Centre of gravity
In this case the receiver looks at the impulse response, calculates the centre of gravity of the impulse
response spectrum and centres the FFT window on that point in time:
The impulse response of the chosen example, with the corresponding centre of gravity indicated by a dashed
line, is given in Fig. 13.
In this example, signals 2 and 3 fully contribute constructively. Signals 1 and 4 show a small amount of inter-
symbol interference arising from an overlap of the FFT window with symbol n+1 of signal 1 and with
symbol n1 of signal 4. This is depicted in Fig. 14.
The centre of gravity approach responds well to pre-echoes and delayed signals of similar amplitude, since it
does not fix the FFT window to a particular signal but takes into account the average behaviour of the impulse
response of the transmission channel. On the other hand, it can lead to ISI in cases where other strategies may
not lead to ISI: for example, most two-echo cases, separated by virtually the whole guard interval, would cause
difficulties for this strategy unless the two echoes were of equal power.
Signal 1
Signal 4
R
e
l
a
t
i
v
e

a
m
p
l
i
t
u
d
e
Time
Signal 2
Signal 3
Threshold
Figure 11
Synchronization to the first signal above a threshold
(signal 2); impulse response
Signal 1
Signal 2
FFT window
Time
Symbol n1
Signal 3
Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Signal 4
ISI
Figure 12
Synchronization to the first signal above a threshold (signal 2);
FFT window position
where
t
c
= centre of gravity
p
i
= power of the i-th signal of the impulse response
t
i
= time of the i-th signal of the impulse response
!
!
=
i
i
i
i i
c
p
t p
t
OFDM
EBU TECHNICAL REVIEW July 2003 8 / 12
R. Brugger and D. Hemingway
3.5. Quasi-optimal
This strategy builds on that described
in Section 3.3., in an attempt to
approach the Maximum C/I
described in the next section.
The first signal of the impulse response
above a minimum threshold level is
taken as a reference for the FFT win-
dow. The process is described in the
flowchart shown in Fig. 15.
3.6. Maximum C/I
Whereas the previously-discussed
strategies all offer means of quickly
finding a good FFT window position,
an optimal choice would be a position
where the effective C/I is maximized.
This position, however, is not easily
found and would in general take too
much time to be calculated. Therefore,
normally one of the above simpler
strategies, or a combination of them, is
applied.
Such simpler approaches can be justi-
fied by the fact that the optimum C/I
will often show a relatively flat maxi-
mum, i.e. errors introduced by sub-
optimal synchronization are small. But
there are also difficult configurations
possible, e.g. in a two-echo case, if the
difference in delay is close to the guard
interval, there is only one position that
will result in no ISI, so the optimum
here would be very sharp.
Note that the method described in Sec-
tion 3.5. does not attempt to find a posi-
tion for the FFT window that gives the
best C/I. It merely seeks to find a posi-
tion for the window at which the C/I is
good enough to allow demodulation
and decoding with an acceptable error
rate.
Receiver manufacturers indicate that
the evaluation of C/I is by no means
trivial for a DVB-T receiver and, for a
DVB-T mode with a large guard inter-
val of T
u
/4, there seem to exist theoret-
ical limits for the evaluation of C/I
which would prevent the application of
a maximum C/I synchronization
strategy in this case.
Signal 1
Signal 4
R
e
l
a
t
i
v
e

a
m
p
l
i
t
u
d
e
Time
Signal 2
Signal 3
Centre of gravity
Figure13
Synchronization to the centre of gravity (between signal 2 and 3);
impulse response
Signal 1
Signal 2
FFT window
Time
Symbol n1
Signal 3
Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Symbol n1 Symbol n+1 Symbol n
Signal 4
ISI
ISI
Figure 14
Synchronization to the centre of gravity (between signal 2 and 3);
FFT window position
No
Yes
Yes
No
FFT window aligned at the beginning
of a symbol of the first signal
above the threshold value
Is the C/I good
enough to allow
demodulation?
FFT window aligned at the
beginning of a
symbol of this signal
Are there any
other signals in the time
impulse response above
the threshold?
FFT window aligned at the
beginning of a symbol of
the signal that allows the
greatest C/I
FFT window aligned with the next
signal in the impulse response
to exceed the threshold value
Figure 15
Flowchart describing the quasi-optimal strategy
OFDM
EBU TECHNICAL REVIEW July 2003 9 / 12
R. Brugger and D. Hemingway
With regard to receiver modelling in computer simulations (e.g. for coverage calculations), the detection of the
maximum C/I position of the FFT window is not a major problem. A simple but time-consuming approach
would be to scan the time period of interest with an appropriate step size, calculate the C/I for each sampling
point and use the time position with the maximum C/I as the reference.
A more sophisticated strategy to find the maximum C/I position is based on the observation that the maximum
C/I is always found at a position where the FFT window is aligned with the start, or the end, of one of the
incoming signals for the symbol under consideration. A check of all these possible positions, which amounts
to 2N evaluations of C/I for N signals, then gives the maximum C/I position. Experience with practical soft-
ware implementations shows that the computational effort is about twice that of the basic strategies described
in Sections 3.2. to 3.4..
4. Studies
4.1. Theoretical regular T-DAB network
As described above, the FFT window synchronization strategy has an effect on the receiver performance in the
presence of more than one signal (coming from either one transmitter or from transmitters in an SFN). In
order to demonstrate how FFT window synchronization influences the receiver performance, a regular 7-trans-
mitter hexagonal network, having characteristics similar to that of the Wiesbaden 95 Band III T-DAB refer-
ence network, has been used. The network is open, and the central transmitter can be given a time offset. T-
DAB Mode I ( = 246 s) is used in all the simulations shown in this example. With no time offset on the
central transmitter, the area inside the hexagon is virtually free from inter-symbol interference. By applying a
time offset of 300 s to the central transmitter, a situation where ISI occurs has been deliberately created.
It must be emphasized that this theoretical network is used solely for the comparison of receiver synchroniza-
tion methods and is not intended to represent a network that would actually be constructed. However, in real
networks, time delays do occur.
A study carried out in Germany, to investigate the above synchronization modes in this theoretical network,
indicates that the synchronization strategy is a crucial feature for the performance of a receiver.
The results for four synchronization strategies are given in Figs 16 - 19.
The results show that, for very critical reception situations, the difference between the predicted coverage
probabilities for the various synchronization strategies may amount to as much as 20%. For the high probabil-
ity range, the differences are found to be between 5 and 10%.
Moreover, the investigation showed that knowledge about the synchronization behaviour of the receivers is
important for reliable T-DAB network planning.
4.2. Real network simulation
In a case study in Switzerland, a realistic DVB-T SFN simulation in the Zrich region was performed. The
network consisted of three transmitters, the characteristics of which were optimized in terms of ERP and time
Abbreviations
C/I Carrier-to-Interference ratio
CEPT European Conference of Postal and
Telecommunications Administrations
DAB Digital Audio Broadcasting (Eureka-147)
DVB Digital Video Broadcasting
DVB-T DVB - Terrestrial
EICTA European Information, Communications and
Consumer Electronics Technology Industry
Associations
ERP Effective Radiated Power
FFT Fast Fourier Transform
IFFT Inverse Fast Fourier Transform
ISI Inter-Symbol Interference
MFN Multi-Frequency Network
OFDM Orthogonal Frequency Division Multiplex
SFN Single-Frequency Network
T-DAB Terrestrial - Digital Audio Broadcasting
OFDM
EBU TECHNICAL REVIEW July 2003 10 / 12
R. Brugger and D. Hemingway
delays in order to maximize coverage. Different receiver synchronization algorithms were compared and the
resulting population coverage was calculated. Table 1 gives the results. The trigger threshold used in simula-
tions for the first signal strategy was 20 dB above the noise level. However, it should be mentioned that only
signals inside the guard interval were regarded as wanted, whereas signals outside the guard interval were
treated as interference (cliff-edge transition), which is an approximation to the response in Fig. 7.

Table 1
Percentage of the population covered by a digital SFN (16-QAM
3
/
4
, =
1
/
8
) in the Zrich region for different
synchronization algorithms at the receiver. ERP of all transmitters is 40 dBW.
FFT Window Fixed
outdoor reception
Portable
indoor reception
Band III Band IV Band III Band IV
Start = the strongest signal 85.8 % 86.1 % 78.0 % 70.7 %
Start = the first signal over threshold 97.1 % 95.3 % 83.8 % 74.2 %
Centre = the centre of gravity of all signals 77.8 % 77.7 % 63.7 % 59.9 %
Coverage Probability:
95%+
85% to 95%
75% to 85%
65% to 75%
54% to 65%
44% to 54%
35% to 44%
25% to 35%
15% to 25%
transmitter
DAB Mode I
7 transmitter network
P = 1.0 kW, nd (peripheral)
P = 0.1 kW, nd (central)
t = 0.3 ms (peripheral)
Rec. P-370, VHF, 50%L/1%T
x (km)
y

(
k
m
)
-60
-40
-20
0
20
40
60
-60 -40 -20 0 20 40 60
Strongest signal
Coverage Probability:
95%+
85% to 95%
75% to 85%
65% to 75%
54% to 65%
44% to 54%
35% to 44%
25% to 35%
15% to 25%
transmitter
DAB Mode I
7 transmitter network
P = 1.0 kW, nd (peripheral)
P = 0.1 kW, nd (central)
t = 0.3 ms (peripheral)
Rec. P-370, VHF, 50%L/1%T
x (km)
y

(
k
m
)
-60
-40
-20
0
20
40
60
-60 -40 -20 0 20 40 60
First signal (above a threshold)
Coverage Probability:
95%+
85% to 95%
75% to 85%
65% to 75%
54% to 65%
44% to 54%
35% to 44%
25% to 35%
15% to 25%
transmitter
DAB Mode I
7 transmitter network
P = 1.0 kW, nd (peripheral)
P = 0.1 kW, nd (central)
t = 0.3 ms (peripheral)
Rec. P-370, VHF, 50%L/1%T
x (km)
y

(
k
m
)
-60
-40
-20
0
20
40
60
-60 -40 -20 0 20 40 60
Centre of gravity
Coverage Probability:
95%+
85% to 95%
75% to 85%
65% to 75%
54% to 65%
44% to 54%
35% to 44%
25% to 35%
15% to 25%
transmitter
DAB Mode I
7 transmitter network
P = 1.0 kW, nd (peripheral)
P = 0.1 kW, nd (central)
t = 0.3 ms (peripheral)
Rec. P-370, VHF, 50%L/1%T
x (km)
y

(
k
m
)
-60
-40
-20
0
20
40
60
-60 -40 -20 0 20 40 60
Maximum C/I
Figure 17
Synchronization First Signal above a
threshold
Figure 18
Synchronization Centre of gravity
Figure 16
Synchronization Strongest signal
Figure 19
Synchronization Maximum C/I
OFDM
EBU TECHNICAL REVIEW July 2003 11 / 12
R. Brugger and D. Hemingway
The results show that the choice of the synchronization strategy is crucial for the coverage of the network. The
study concluded that a strategy that aligns the FFT window with the start of the first signal above a certain
threshold is the best choice of the three considered.
4.3. Measurements on DVB-T receivers
Since manufacturers information is not available, laboratory measurements in Italy have been performed in
order to investigate the synchronization strategies that are employed by nine receivers.
The study came to the conclusion that the majority of the available DVB-T receivers uses the threshold algo-
rithm (but with different threshold levels). Only one receiver (based on the 2k mode and therefore not directly
suitable for wide SFNs) adopts the centre of gravity method. No receiver was identified using the
maximum C/I algorithm.
5. Conclusions
As the T-DAB and DVB-T systems differ significantly in their specifications, T-DAB receivers work differ-
ently from DVB-T receivers in some crucial aspects. In particular, T-DAB receivers do not exhibit the time cut-
off that is a feature of all DVB-T receivers, so longer echoes can be used constructively in T-DAB systems.
Different possible strategies of FFT window synchronization in OFDM receivers have been reviewed. Since
the FFT window-positioning strategy may significantly affect the coverage probability in a network:
1) it is important for network planners to know about the synchronization strategies adopted by current and
future receivers, and;
2) it should be of interest to the receiver manufacturers to know which strategy has been assumed for the
planning of OFDM networks.
These considerations should seriously be taken into account by the manufacturers (EICTA) and by the interna-
tional and national bodies responsible for frequency planning (CEPT and member administrations).
In a recent meeting between the EBU and EICTA, it was indicated that the first signal above a threshold
strategy is most commonly used in receivers. This agrees with the findings of the study mentioned in Section
Dr Roland Brugger graduated from the Ludwig-Maximilians University, Munich, in
1981 and obtained a PhD in physics in 1987. In the same year, he joined the Fre-
quency Management Section of the Institut fr Rundfunktechnik (IRT) in Munich,
where he worked on statistical aspects and on combinatorial optimisation problems in
the frequency planning field. Since 1992, he has been involved in frequency and net-
work planning for digital terrestrial broadcasting services.
Roland Brugger participated in the T-DAB and DVB-T Planning Meetings in Wies-
baden (1995), Chester (1997) and Maastricht (2002), and is presently engaged in the
preparation of the RRC04/06. He is a member of ITU-R TG6/8 and of the EBU project
groups B/CP-R and B/CP-T.
David Hemingway joined the BBC's Transmission department in 1988 and worked
on HF, VHF and UHF transmitting stations before becoming responsible for micro-
wave link projects in the mid-1990s. After the privatisation of BBC Transmission in
1997, he helped to establish the Spectrum Planning group within Crown Castle UK
Ltd.
In 2000, Mr Hemingway moved to work for Nokia, where he managed a European
group of 3G radio planners and worked in Sweden and Switzerland as well as the UK
in that capacity. In 2001, he rejoined the BBC and now works as a Senior R&D Engi-
neer in the Spectrum Planning Group at Kingswood Warren. He is currently one of
the leaders of the BBC's preparatory work for the RRC in 2004/06.
OFDM
EBU TECHNICAL REVIEW July 2003 12 / 12
R. Brugger and D. Hemingway
4.3., and corresponds to the technique assumed by most network planners when designing their networks.
Note however that the value of the threshold used is not specified.
Acknowledgement
This article results from work done in EBU Project Group B/CP-T (Conference Preparations Television).
The authors would like to thank all the colleagues who participated in the preparation of the article.
The article is dedicated to Jrn Andersen.
25 July 2003