CAP 7 PAR1

The Broadcast Side
Wolfram Titze, Gemid Chouinard and Stephen Baily
7.1 General
This chapter explains the broadcasi side of DAB without touching upon the issue of signal deliverv to the different transmission sites of the network since this was covered in the previo us chapter. DAB transmission networks can ha ve national, regional or local coverage. Depending upon specific national. geographical and financiai require-ments, Band III or L-band frequencies may be used for these networks.
The rest of this chapter is structured as follows: section 7.2 deals with the trans-_mission channel limitations due to RF signal propagation, illustrates the reasons behind the choice of 1.5 MHz bandwidth for the DAB system and gives a simplified propagation model for the DAB system. Next, section 7.3 gives an introduction to DAB networks, highlighting the major differences of DAB networks cornpared to conventiónaJ FM networks and explaining the concept of a single frequency network (SFN), The particiiiariíies of SFNs are then described in section 7.4. The equipmení needed on the transmitter site to set up SFNs and the associated specifics are presentad in section 7.5. Section 7.6 treats the aspects to be considered when planning networks with SFN coverage and section 7.7 discusses the issues of coverage evalu-aíion and monitoring for SFNs. Finally, aspects of frequency management and frequency allocaüon fcr DAB networks are covered in section 7.8.
7,2 Radio Frequency Propagation Aspects
The design. the complexity and, as a cousequence. the cost of a DAB system is slrongJy dependen t on the faclors affecting the propagation characteristics of the

transmission channel to the vehicular receiver, the indoor receiver, the portable receiver and the stationary receiver; in decreasing order of difficulty. The propaga-tion path is subject to attenuation by shadowing due to buildings, trees and other foliage and to multipath fading due to specular reflections and diffuse scattering from the ground and nearby obstacles such as trees and buildings (see section 2.1). The degree of impairment to the received signal depends on the operating frequency, the receiving antenna height and the type of environment in which the receiver is operating: whether it is an open, rural, wooded, mountainous, suburban or dense urban environment.
7.2.1 The Impaired RF Channel
The usual channel response is illustrated by a two-dimensional representation of the time delay spread of the channel, where multipath results in peaks of various amplitude, present at specific excess delays relative to the original channel excitation impulse, as illustrated in Figure 7.1. Each delayed peak corresponds to a reflected signal.
When the receiver is in motion, various Doppler shifts appear for each of the signáis received. The Doppler shift on an individual multipath component depends on the angle of its arrival with respect to the direction of the vehicle displacement. The cumulative effect of these shifts is called the channel Doppler spread. The procedure involved in analyzing such spread is to line up a series of successive impulse response snapshots, take a slice through them at a given time delay, and then do a Fourier transform on the resulting samples. As a result, the broadcast channel can be illustrated by a three-dimensional representation of its impulse response.
The effect of the spreading of the received signal in both time and frequency is illustrated as scattering diagrams in Figures 7.2 and 7.3 where the channel impulse response is given on a three-dimensional plot of amplitude versus time and fre¬quency. This is a very useful tool for visualizing the effects of the surrounding environment on the signal received by a mobile receiver. Figure 7.2 shows an example of a scattering diagram obtained from a measurement run that took place in the downtown área of a large city. The máximum Doppler shift that can occur is a
Figure 7.1 Simplified 2-dimentional channel impulse response


Figure 7.2 Scatter'mg diagram - downtown/large city
function of the radio frequency and the vehicle speed. In terms of the normalized frequency scale used here, the máximum shift at l.SGHz is + 5Hz/m. This corres-ponds to a máximum shift of 139 Hz at a typical highway speed of lOOkm/h.
In an urban área such as the one depicted here, the scattering diagram tends to be limited to small excess delays but is rather complex due to the multitude of echoes. In this case, the scattering is from objects cióse to the receiver, and the scattered signal arrives from many different angles. Particularly in dense urban environments, the arrivals tend towards a uniform angular distribution, which results in a "U" shaped Doppler spectrum. At larger excess delays in such an environment, the components with small Doppler shift (nearly perpendicular to the direction of travel) tend to disappear. In urban environments, these components are frequently blocked by the buildings along the street. The Doppler spectrum then divides into two groups, one corresponding to the signal components arriving from the direction in which the vehicle is travelling (positive Doppler shift), and the other from components arriving behind the vehicle (negative Doppler shift).
Figure 7.3 shows a scattering diagram derived from a measurement in the downtown área of a small city. The división of the Doppler spectra into two groups near the máxima is more evident here. The diagram clearly shows the funnelling of the signal by the street along which the vehicle is moving, with many components lined up near the máximum Doppler shift. There is a sizeable component having about 5 IJLS excess delay that seems to be caused by a reflection from a building located some 750m (2.5 JJLS*XC) behind the transmitter and reaching the receiver from


Figure 7.3 Scattering diagram - downtown/small dty
behind (negative Doppler shift). These two figures are each the result of a group of 128 successive channel impulse response snapshots representing some 5m of travel.
7.2.2 Propagation Models
The probability distribution functions relevant to the reception of DAB signáis were found to correspond to a number of statistical distribution models related to the specific environment. These distribution models are generally different in so-called "small áreas" and "large áreas". The large áreas are usually defined as locations extending over a number of wavelengths (Á.), usually 200 m by 200 m which corres-ponds to 1000 X. by 1000 X at 1.5 GHz.
7.2.2.1 Large Área Distribution
On large áreas, once the signal levéis have been averaged to remove the signal variations within the "small áreas", it has been found experimentally that the probability distribution function of the mean received signal power takes the log-normal form and corresponds to the model described in Recommendation ITU-R P.1546 [P.1546] (superseding the former Recommendation ITU-Rec P.370). It was also found empirically that, for wideband transmission systems such as DAB, the



standard deviation of location variation is signifícantly lower than that for narrow-band analogue systems. The widely accepted valué for the standard deviation for DABis5.5dB.
7.2.2.2 Small Área Distribution
Results of extensive measurements carried out in United States and in Europe with narrow-band signáis or CW signáis indícate that the small área behaviour of the received signal can be modelled by a Rician distribution (constant vector plus Rayleigh distributed vectors). American and European researchers concluded that the probability density function of the received power should combine log-normal and Rayleigh distribution in order to take account of both large-area variations and small-area variations. The distribution of instantaneous valúes in a small área is obtained by considering a Rice or Rayleigh variable whose mean valué is itself a random variable having a log-normal distribution.
Figure 7.4 confirms these conclusions and depicts the variation of a CW signal received by a mobile receiver in an urban área. The relative field strength (normalized to the mean) is shown as a function of the receiver location along the measurement route. It can be seen that multipath causes fast and very deep fades, while a rather

Figure 7.4 Relative signal variation (dB) with the corresponding cumulative distribution function ofa CW signal as seen by the mobile unit in a large área and a small área


slow variation of the signal envelope in the large área plot in Figure 7.4 reveáis the presence of shadowing due to tall buildings.
A magnified view of the received signal as a function of location, presented as the small área, shows that multipath causes large signal variations resulting from signal cancellation between the various scattered signal components. This variation of the resulting signal field strength usually corresponds to a Rayleigh distribution. The corresponding cumulative distribution functions (CDF) of the measured signal levéis are presented on the right side of the figures. Due to the time interleaving used by DAB (see section 2.2.5), these small scale fluctuations will not affect reception in vehicles moving at sufficiently high speed.
7.2.2.5 Effect of Channel Bandwidth
The effect of the transmitted signal bandwidth on the variability of the received signal is illustrated by comparing Figures 7.4 and 7.5. As can be seen from these figures, the level of the 1.47MHz bandwidth signal shown in Figure 7.5 exhibits much less multipath fading than the CW signal from Figure 7.4, and the CDF of the wide-band signal is closer to the reference Gaussian CDF. This is due to the fact that the power is integrated over the 1.47 MHz bandwidth, thus averaging out most of the

sharp frequency selective fading occurring in small áreas. At the limit, such averaging would have the same effect as the averaging over a small área, therefore eliminating the Rayleigh portion of the combined channel propagation model.
Normally, this averaging of the small área frequency fading could not be done because the transmission system would suffer from mese narrow-band fades but the Eureka 147 DAB system has the advantage of having been designed to opérate quite well in a multipath environment. As long as the signal echoes fall within the symbol guard interval, the assumption of integrating the power within the channel bandwidth, thus averaging out the sharp frequency selective fades, is a useful approximation, although the underlying mechanisms (channel coding and interleaving) have to be considered in detail for a precise analysis. (In fact, DAB reception in a Rayleigh fading environment suffers an apparent raise of the threshold C/N of about 2 dB (for QPSK modulation) compared to a Gaussian channel but once it is in a Rayleigh environment, the number and relative level of echoes becomes irrelevant.)
Based on this assumption, an actual field measurement was conducted to charac-terize the sensitivity of the received signal variation as a function of channel band¬width. Figure 7.6 shows the increasing multipath fade margin as the channel bandwidth is increased from lOOkHz to 5MHz in a dense urban environment. The fade margin can be interpreted as the possible saving in transmit power relative to that needed for a lOOkHz channel bandwidth system, for an equivalent service availability objective.
Figure 7.6 shows that, for service availability objectives lower than 50%, the improvement in fade margin remains in the order of 1.5dB in a dense urban área. Significant improvement is observed for service availability objectives of 90% or greater. Each curve can be divided into two sections, the first part being from lOOkHz to a bandwidth valué that corresponds to a knee in the curve, the second part being from the knee position to the 5 MHz bandwidth valué.
It seems that the position of the knee on the curves falls between 1 and 2 MHz, confírming the validity of the choice of 1.5 MHz bandwidth for the DAB system. Below 1 MHz, the multipath fading increases abruptly while above 2 MHz the improvement in fade margin is generally not very signifícant. This also means that most of the small área Rayleigh fading is removed when a channel bandwidth of more than about 1 MHz is used in combination with the DAB modulation.
7.2.3 Simplifica Propagation Model for the DAB System and its Range of Validity
As seen in the previous section, except for a cost of about 2 dB for the operation of the DAB system in a multipath environment, it can be assumed that the performance of the DAB reception will correspond to the averaged field strength found over the "small área". This means that, instead of using the combined log-normal and Rayleigh propagation model, the reception performance for the DAB system can be modelled following the simpler "large área" log-normal model. Recommendation ITU-R P.1546 describes the log-normal model used by the ITU-R in predicting

The original ITU-R prediction model ITU-R P.370 [P.370] had to be extended to include the case for vehicular reception. The reduction in receive antenna height resultad in a correction factor which depends on the distance from the transmitter and varíes from 9 dB to 7 dB. Although prediction for Band III propagation could be interpolated, the model had to be extended to allow prediction at l.SGHz. This resulted in a correction factor of around 1 to 2 dB from the 600 MHz curves. These corrections were included in the latest versión of the model [P.1546].
For the calculation of the fíeld strengths of a digital sound broadcasting signal, in particular in rural áreas, the propagation prediction method of Recommendation ITU-R P.1546 can therefore be used. In built-up áreas, the method given in Rec. ITU-R P.1546 is also appropriate, giving cióse agreement with the well-known Okumura/Hata propagation model [Okumura, 1968], [Hata, 1980]. However, a model based on more precise terrain cover information would be more accurate.
The validity of these models is illustrated by Figures 7.7 and 7.8 which compare the results of actual field measurement conducted at 1.5GHz in the Montreal área in Canadá with these propagation models for a gíven effective isotropic radiated power (e.i.r.p) and height above average terrain (HAAT).
The range of validity of this simplified log-normal propagation model is defined as a function of the type of multipath encountered at the receiver. If there are some signal echoes falling outside the range that can be corrected by the DAB receiver, a more complex model including local "small áreas" fading would need to be used. This can happen in three cases:
1) Micro-reflections with excess delays below about the reciprocal of the DAB channel bandwidth will result in apparent fiat fading over the channel. If the receiver is not moving to take advantage of the time interleaving, this could result
Figure 7.7 Comparisons of measured data with Rec. ITU-R P.370 model and free-space loss curve in Montreal for 50% of locations (e.i.r.p. = 41.1 dB(W), HAAT = 235.5m)
Figure 7.8 Comparisons of measured data with the different Okumura models in Montreal for 50% of locations (e.i.r.p. = 41.1 dB(W), HAAT = 235.5m)


in loss of service. This would not be predicted by the log-normal model. This will happen mainly in cases of closely-spaced large buildings (downtown core) and with indoor reception conditions since the physical difference of signal paths would need to be small (below 200 m). This would come in addition to the log-normal model but is not typical of suburb and rural reception.
2) Multipath reflections which fall beyond the symbol guard interval and créate inter-symbol interference. The reception performance will rapidly be impacted by this intra-system interference and reduce the service availability. This comes in addition to the log-normal prediction.
3) If the vehicle velocity is higher than recommended for the given DAB transmis-sion mode and carrier frequency, Doppler spread will affect the performance of the DAB reception (see section 7.3.2). Such apparent signal fades will come in addition to the log-normal model.
Although the ITU-R prediction model is widely accepted, it is preferable to augment it with more precise prediction methods based on topographic databases and land occupation data [Voyer, 2000], [Voyer, 1996], [Whitteker, 1996]. In this way, a DAB system more precisely tailored to the given service área can be planned and specific needs, such as repeaters to cover specific hard to reach áreas, can be predicted.
7.2.4 Building Penetration Losses at 230 MHz and 1.5 GHz
Field strength measurements were perforrned to derive typical figures for the building penetration losses in the frequency bands relevant to DAB.
Around 230 MHz, building penetration loss was measured in the UK [Green, 1992] and Germany [Schramm, 1996]. The UK results show that the building penetration loss varies between 2dB and 18dB on the ground floor of domestic buildings. Measurements on the first floor gave about 6dB more field strength. The average loss was found to be 8dB + 1.2dB. The Germán results basically support these figures. The penetration loss measured ranged from 3dB to 20 dB and the median valué for typical Germán buildings was found to be 9 dB at 220 MHz and 8.5 dB at 223 MHz with a standard deviation of 3.5 dB. The attenuation caused by a building located between an outdoor receiver and the transmitter was found to be 13 dB.
At 1.5 GHz, measurements in Australia ha ve shown that the average building penetration loss for DAB in domestic dwellings averages 6.7 dB (ranging from 6.1 dB to 9.4 dB, depending on construction materials used) and is approximately 18.6 dB in reinforced concrete commercial buildings [DSB, 2002]. Measurements in a 1.5 GHz SFN were perforrned in the DAB pilot project at Dresden, Germany [Michler, 1998]. The field strengths in rooms at different floors of seven different buildings were measured. The buildings were all located in a zone where two or three transmitters contributed to reception. In most buildings the level difference between outdoor and indoor measurements was found to be O dB to 5 dB at upper floors and 8 dB to 15 dB at the ground floor. In a modern office building (a concrete-steel construction with metal coated windows), however, the corresponding valúes were 20 and 30 dB, respectively.

7.3 Introduction to DAB Networks
7.3.1 Difference Between FM and DAB Networks
Planning of transmission networks for FM broadcasting is traditionally based on the concept of múltiple frequency networks (MFNs). In an MFN, adjacent transmitters radiate the same programme but opérate on different frequencies to avoid interfer-ence of the signáis where the coverage áreas of different transmitters overlap. Basic FM receivers cannot cope with interfering signáis from other transmitters of the same network using the same or nearby frequencies. Coverage planning for an FM network requires frequency planning for the different transmitter sites, to optimise use of the scarce resource: RF frequencies.
DAB in contrast allows single frequency networks (SFNs), where all transmitters of the network transmit exactly the same information on the same frequency. The main condition for a working SFN is that all transmitters are synchronised to each other in frequency and fulfil certain time delay requirements which wül be explained later in this chapter. Coverage planning for a DAB network requires time delay planning between the different transmitters instead of frequency planning as in the case of FM. The SFN capability of DAB allows complete coverage of very large regions without the receiver having to tune to a different frequency while moving around in the área.
In contrast to FM broadcasting, DAB transmits typically five to seven different programmes in one single ensemble on one frequency and all programmes contained in that multiplex share the same coverage área. Distinction by coverage área is therefore not possible for radio stations whose programmes share the same multi¬plex. It is also not advisable in an SFN to introduce local windows, that is áreas where some transmitters of the SFN radiate a slightly different multiplex to achieve local programme variation. By definition, local windows cause problems for the receiver in the overlap área of the differing programmes of the multiplex since it cannot determine which programme to select.
7.3.2 Why SFNs Are Possible with DAB
DAB is a digital broadcasting system which was developed especially for the challen-ging transmission characteristics of the mobile radio channel. Typical phenomena of this channel like Doppler shift and multipath propagation with the resulting time and frequency selective fading had to be taken into account while developing DAB. To cope with these problems, the guard interval was introduced between consecutive data symbols, time and frequency interleaving techniques were applied to the data stream, a choice of sub-carrier spacings in the multicarrier modulation scheme was introduced and channel coding techniques were applied to correct for transmission errors. Table 7.1 shows the most important problems of mobile radio transmission systems and how they are alleviated in DAB.
The DAB system is a very robust and frequency economical transmission system which enables correct decoding of information despite Doppler spread and multipath



Tafe/e 7.1 Problems of mobile radio transmission systems and their solution in DAB



Problems
DAB Solution




Time-dependent fading (multipath while driving) Frequency-dependent fading (stationary
multipath) Doppler spread (speed dependen!, while driving)
Delay spread (due to multipath) Transmission errors

Time interleaving
Broadband system with frequency interleaving
Sub-carrier spacing increased as a function of the
transmission frequency Guard ¡nterval (allows SFNs) RCPC (Rate Compatible Punctured
Convolutional) codes and Viterbi decoding to
reconstruct the original bit stream