IEEE TRANSACTIONS ON BROADCASTING, VOL. 56, NO. 3...

IEEE TRANSACTIONS ON BROADCASTING, VOL. 56, NO. 3, SEPTEMBER 2010 331

On the Methodology for Calculating SFN Gain inDigital Broadcast Systems

David Plets, Wout Joseph, Member, IEEE, Pablo Angueira, Member, IEEE, José Antonio Arenas,Leen Verloock, and Luc Martens, Member, IEEE

Abstract—For broadcast networks, the Single-Frequency Net-work (SFN) mode is an alternative to the well-known Multi-Fre-quency Network (MFN) mode, where instead of transmitters op-erating at different frequencies, all base stations use the same fre-quency. Besides the optimal frequency reuse, it is usually expectedthat the more homogeneous distribution of received signal strengthreception in an SFN will improve the quality of service. Neverthe-less, it should be noted that not all the locations within the servicearea will benefit from the SFN configuration. Some areas will showa degraded quality caused by the SFN echoes. In this paper, theSFN gain is defined as a parameter describing potential gain or in-terference. An unambiguous methodology to obtain the actual SFNgain is presented and the variation of the gain is investigated for aDVB-H network as a function of the signal strength difference re-ceived from different transmitters. This SFN gain can be used forcoverage planning of future broadcast networks.

Index Terms—Digital video broadcasting-handheld, DVB,DVB-H, methodology, network, physical layer, SFN, SFN gain.

I. INTRODUCTION AND OBJECTIVES

W HEN operators deploy a broadcast network, they wantto keep the total cost as low as possible. This can be

achieved by keeping the number of base stations and their trans-mitting power as low as possible. From a theoretical perspec-tive, a Single Frequency Network (SFN) will deliver the samequality with a certain amount of reduction in transmitted power.This assumption is based on the fact that the field strength willbe homogeneously distributed due to the spatial diversity asso-ciated to SFN networks. Nevertheless, the quality of service inthe SFN area will depend on the number and nature of the re-ceived signals (number of transmitters, relative delays, and rela-tive amplitudes). Furthermore, the equalization method and syn-chronization stages of the reference receiver will also be relevantto the final shape of the service area. The SFN gain parameterwas proposed at the same time that the SFN concept was de-veloped (in practical terms) for digital broadcasting in the early

Manuscript received February 17, 2010; revised May 04, 2010; accepted May05, 2010. Date of publication June 21, 2010; date of current version August 20,2010. This work was supported by the IBBT-MADUF project, co-funded by theIBBT (Interdisciplinary institute for BroadBand Technology), a research insti-tute founded by the Flemish Government in 2004, and the involved companiesand institutions.

D. Plets, W. Joseph, L. Verloock, and L. Martens are with the Department ofInformation Technology, Ghent University/IBBT, 9050 Gent, Belgium (e-mail:[email protected]).

P. Angueira is with the University of the Basque Country (UPV/EHU), 48015Bilbao, Spain.

J. A. Arenas is with Telefónica Móviles España, 28013 Madrid, Spain.Color versions of one or more of the figures in this paper are available online

at http://ieeexplore.ieee.org.Digital Object Identifier 10.1109/TBC.2010.2051176

nineties. A review on the literature related to SFN gain calcu-lations [1]–[12] shows that most references have dealt with theSFN gain optimistically, just considering the improvement inthe homogeneity of the behavior of the field strength, leavingaside potential self-interference and both synchronization andequalization problems associated to the receiver. For most ofthe cases studied in the literature, the SFN gain calculation hasbeen based on field strength measurements, trying to obtain thestatistical improvement in the field strength distribution. Thispaper will show that this should not always be the case, and willprovide a methodology to calculate a meaningful value of theSFN gain for network planning purposes [13]. Also, it shouldbe noted that previous work has been carried out for traditionalbroadcasting scenarios to fixed receivers. In cases where the re-ceivers are portable (either indoor or outdoor) or mobile, thenumber of SFN echoes and the extension of the area where theyare present will increase.

The proposed procedure will be applied on a real DVB-Hnetwork in Ghent, Belgium and the SFN gain at a certain lo-cation will be calculated as a function of the difference in signalstrength received from the different transmitters (overlappingdegree). Also, the influence of the reception quality in MFN op-eration on the SFN gain will be analyzed.

The DVB-H network, the measurement equipment and thecollected data are discussed in Section II. Section III defines theSFN gain and Section IV discusses the processing of the mea-surement samples. In Section V, quality categories and trans-mitter overlapping are defined, Section VI discusses the results,followed by a comparison with related literature in Section VIIand finally, the conclusions are presented in Section VIII.

II. TRANSMITTING NETWORK AND AVAILABLE

MEASUREMENTS

A. Transmitting Network

The transmitting network is located in the city of Ghent (Bel-gium), a mixture of a suburban and an urban environment. TheSFN consists of three base station antennas (BS) and operatesat a frequency of 602 MHz. The channel bandwidth is 8 MHz.Fig. 1 shows a map of Ghent with the location of the threebase stations marked with black circles with white dots in it.All transmitting antennas (Tx) are omnidirectional and verti-cally polarized. The heights of these Tx are ,

, and , respectively. The EIRP(Equivalent Isotropically Radiated Power) used for these Tx is36.62 dBW, 39.93 dBW, and 40.90 dBW, respectively. The con-stellation used for the tests is 16-QAM 1/2 with an MPE-FEC(Multi-Protocol Encapsulation-Forward Error Correction) rateof 7/8, corresponding with a useful bit rate of 9.68 Mbps. This

0018-9316/$26.00 © 2010 IEEE

332 IEEE TRANSACTIONS ON BROADCASTING, VOL. 56, NO. 3, SEPTEMBER 2010

Fig. 1. Map of Ghent with the 3 BS (black circles with white dots) and indica-tion of the measurement route.

constellation is preferred because of its satisfactory behaviorregarding both bit rate and coverage area [14]–[16]. The usedguard interval is 1/8 and the FFT mode is 4K [17]–[22].

B. Measurement Equipment

The measurements are performed with a tool implementedon a PCMCIA (Personal Computer Memory Card InternationalAssociation) card with a small receiver antenna Rx. The gain ofthe antenna is 5 dBi [15]. The PCMCIA card is plugged intoa laptop, which is used to perform the measurements.

Every 0.5 s, a sample is recorded, while the receiver is ei-ther locked or unlocked. A locked receiver can receive DVB-Hframes, which are either correct or incorrect. Incorrect tables can(sometimes) be corrected by the MPE-FEC code. The tool logsparameters as MER (modulation error ratio), FER (Frame ErrorRate), MFER (Multi-Protocol Encapsulation FER), and elec-tric-field strength. MFER is the ratio of the number of residualerroneous frames (i.e., not recoverable) and the number of re-ceived frames [23]. FER is the ratio of the number of erroneousframes before MPE-FEC correction and the number of receivedframes [23]. Location and speed are recorded with a GPS device.Measurements are performed inside a small van at a height ofabout 1.5 m above ground level.

C. Available Data

The methodology proposed in this paper has been developedusing measurements taken along a 50 km route. The routestretches from the very center of Ghent to the municipalitiesthat surround Ghent (see Fig. 1). The SFN gain behavior hasbeen analyzed using four network configurations, which, inone scenario comprises all the transmitters being active (andsynchronized) as a SFN network, and in the rest of scenarios,each one of the transmitters will be the active transmitter,

TABLE IDESCRIPTION OF THE FOUR INVESTIGATED SCENARIOS

being the rest switched off. The four scenarios (each with ascenario ID) that have been investigated, are summarized inTable I. Each scenario provides us with a collection of samplesrecorded along the track. Each sample consists of positiondata (GPS coordinates), signal data (electric-field strength E

), and Modulation Error Ratio (MER) [dB]. Whenan MPE (Multi-Protocol Encapsulation) table [17] is received,it is also recorded whether this table is correct, incorrect, orcorrected after MPE-FEC correction. About 10,000 samplesare collected for each scenario.

III. DEFINITION OF SFN GAIN

This paper proposes to evaluate the SFN gain SFNG basedon the actual performance of a standard receiver and also on thespecific digital broadcast standard. The parameter to be used ineach case might differ (in some cases there will be more thanone candidate). In any case this figure must reflect the actualquality of service being received. This paper will use the MER[dB] as an indication of the quality. Obviously, the MER willbe closely related to field strength levels E but itwill reflect much more accurately all reception conditions, in-cluding the SFN effect, positive or negative (and not only signalstrength variations). SFNG is then defined as the MER in SFNmode minus the MER in MFN mode. It should be noted thatthe field strength values have been used in this paper to evaluatethe real SFN effect, especially when obtaining statistical resultsfrom the measurements. SFN gain effect will be significant onlyin locations where the field strength values received from morethan one transmitter are comparable. Locations where one trans-mitter is dominant, even in the case of having all the transmit-ters on, have not been considered as spots with SFN gain (ei-ther positive or negative). The general SFNG definition has beenempirically analyzed using the experimental data described inSection II-C. The data sets were formed by measurements alonga route, with a strong spatial variation associated to mobile re-ception in an urban environment. In order to remove the fastvariation, the measurement samples were grouped in spatial seg-ments of a certain length (see Section IV). The definition of theSFNG then becomes:

(1)

with the median MER in a segment when all threetransmitters are active and the maximum ofthe median MER values of , with (only onetransmitter active). Thus, for each segment, only two of thefour scenarios will be used for calculation: and theone-transmitter scenario with the highest median MER in thatsegment. The latter scenario represents the ‘best server’ case.

PLETS et al.: ON THE METHODOLOGY FOR CALCULATING SFN GAIN IN DIGITAL BROADCAST SYSTEMS 333

Fig. 2. Process of creating �� data set.

TABLE IINUMBER OF SEGMENTS RETAINED, MEDIAN NUMBER OF SAMPLES IN A RETAINED SEGMENT, NUMBER OF KILOMETERS RETAINED, 90% �� , AND AVERAGE

STANDARD DEVIATION OF MER AND E IN SEGMENTS FOR �� AND �� , FOR DIFFERENT SEGMENT LENGTHS

As mentioned previously, being the reception mobile and con-sidering a significant amount of Non Line-of-Sight (NLOS) sit-uations, the dominant transmitter will be changing from loca-tion to location. In order to smooth this variation and have areference transmitter for each segment, same segments of dif-ferent drives were compared, and a new ‘best server’ route rep-resenting the set of segments with best reception was created.This route is called in the paper . This dataset will represent the ‘ideal MFN scenario’. The process is ex-plained in Fig. 2. The aim of considering a maximum scenario isto include the fact that the best server for one location in a MFNscenario will not necessarily be the closest transmitter, speciallyin the case of mobile and portable reception.

IV. PROCESSING METHODOLOGY OF THE SAMPLES

The field tests were based on mobile DVB-H measurementsalong a 50 km route. The measurement route was repeated fourtimes, with different transmitter configuration (Scenarios 1, 2, 3and ). Each time, the measurement car drove the sameroute, but obviously, the instantaneous measurements were nottaken at exactly the same location. In order to make a propercomparison between all routes, a spatial alignment procedureis applied using the position data from the GPS device. Thealignment procedure starts with a division of the route intosmaller segments of a certain well-chosen segment length, e.g.100 m. One of the four drives is taken as reference and dividedinto segments. The sample at the end of each segment is called“border sample”. Then, the corresponding border samples for theother three trajectories are obtained by picking the sample that

is located the closest to the border sample of the first trajectory,only considering samples starting from the first sample after theprevious border sample. Excellent trajectory synchronizationwas obtained using this procedure (average difference betweensegments lengths is approximately 5 m for a segment length of100 m).

Not all segments were retained for further processing: firstly,because of lack of statistical relevance at certain segments. Datawere discarded where or did not containmore than 5 samples. When short segment lengths are chosen,a lot of segments will therefore be discarded (see Table II, lesskilometers are retained for shorter segment lengths). Secondly,when the resulting actual segment length for differsmore than 20% from the segment length for , the seg-ment is again discarded, because of a possible alignment error.The remaining segments are assumed to be valid segments.

The segment length is varied from 20 m up to 150 m andresults are shown in Table II. This table is used to decide whichsegment length will be used for the SFNG calculation. Thesegment length selection will depend on the number of seg-ments retained and the correctness of the median field strength

within each segment. In [24], a segmentlength of 40 is considered to be the proper length to usein smoothing out Rayleigh fading. For our analysis however,40 corresponds with a segment length of about 20 m whichappears to be an inappropriate choice for our analysis, due tothe limited number of retained segments (94, see Table II) orthe limited number of kilometers (1.9 km, Table II) retained.For the same reason, segment lengths of 50 m and 75 m will


be excluded. From a segment length of 100 m on, the numberof retained kilometers stabilizes (38.7 km). Assuming that theelectric-field strength is Rayleigh distributed within a segment,and its (linear) average within the segment Gaussian, it isshown in [24] that a 90% confidence interval in dB is

with N the number of sampleswithin a segment. Table II shows the 90% confidence intervalfor the average electric-field strength within one segment (basedon the median number of samples in a segment for ).Except for segment lengths of 20 m and 50 m (2.31 dB and 2.09dB, respectively), the differences are rather limited, indicatingthat with respect to this criterion, a segment length of 75 mor large is feasible, with a preference for shorter segmentlengths(closer to 40 ). Adding to all this the fact that in ETSI[25] segments of 100 m 100 m are used to classify DVB-Hcoverage in a small area, we select a segment length of 100 m( , 387 segments retained).

To summarize, we conclude that shorter segment lengths areexcluded from our analysis, since a larger part of the measure-ment track would then be discarded. Larger segment lengths areexcluded because they smooth out the local mean informationmore, which is not desired. Nonetheless, Table II shows that thedifferences between the average standard deviation within a seg-ment are limited (for both MER and E), indicating that evenfor segment lengths above 40 , the variations are rather lim-ited. Once the segment lengths are chosen, a database is createdwith for each segment the information necessary to perform thegain calculation: median E and MER [dB] withinthe segment for and .

V. DEFINITION OF QUALITY CATEGORIES AND

OVERLAPPING DEGREE

The SFNG will reflect the SFN coverage improvement withrespect to the MFN situation. This evaluation is not easy toconvey: some of the areas might present degradation whereasothers might be improved. Also, the SFNG should be evaluatedusing data from areas (segments in this paper) where the SFNeffect is relevant. The comparison study will provide the changesin quality of service (classified as perfect, good, doubtful, andlow) as a function of the intensity of the SFN overlapping. Basedon our observations of the behavior of the SFNG, we assumethat a relevant SFN effect will occur provided two (or more)transmitters are received with a relative field strength differencelower than 9 dB. We observed that if at least one transmitter wasreceived 9 dB above the rest, the effect was less relevant in mostcases (at least if the SFN is composed of 3 transmitters only).Applying these statements to DVB-H, four quality categorieswere defined (perfect, good, doubtful, and low quality), based onthe MFER values [23]. These MFER values are the percentage oflocations with valid tables (method explained in [15]). Table IIIshows these four quality categories based on MFER limits,and their associated upper and lower MER. The overlappingdegree between transmitters will be analyzed as [dB].is defined (for each segment), as the difference between theelectric-field strength due to the dominant transmitter and theelectric-field strength due to the second strongest transmitter.

(2)

TABLE IIIFOUR QUALITY CATEGORIES, CORRESPONDING MFER LIMITS, AND THEIR

UPPER AND LOWER MER

Fig. 3. Overlapping degree � between the transmitters for each valid seg-ment (red: � � � ��, orange: � �� , light green:� �� , and dark green: � �� ).

When is small (i.e., high overlapping degree) and thesignal strengths thus similar, we expect to observe an SFNG.Fig. 3 shows the overlapping degree between the trans-mitters for each valid segment. The color code is explained inthe figure caption. The overlapping degree is mostly only small(red) for segments far away from all transmitters. Also smallparts between transmitters 1 and 2, and transmitters 2 and 3 areorange or red. The parts of the route that are colorless (onlya black trail) in Fig. 3 (and all further route figures) are thesegments that were discarded due to bad synchronization. InFig. 3 five zones are indicated where SFNG has a particularbehavior.

VI. RESULTS

A. SFN Gain and

Two different effects are associated to SFN operation. Somelocations will show a contribution of the SFN effect to the cov-erage (positive SFNG) and at other locations the SFN effect willbe degrading the coverage (negative SFNG). Fig. 4 shows theSFNG along the route. Figs. 3 and 4 show that segments thatare not served (further away from the transmitter), show rele-vant overlapping degrees (low ). In those areas SFNG ismostly positive (yellow and green). This happens because theMER in the MFN cases is too low because the received fieldstrength from any of the transmitters is too weak. When oper-ating the network in SFN mode, the MER increases, due to the


Fig. 4. SFNG in the different segments along the route.

TABLE IVCOMPARISON OF THE QUALITY OF THE RETAINED SEGMENTS FOR ��

AND �� FOR THREE DIFFERENT RESTRICTIONS WITH REGARD TO

� : NO LIMIT ON � (ALL SEGMENTS), AND � LIMITED TO A

MAXIMUM OF 3 dB AND 6 dB

contribution of the three transmitters. Close to an active trans-mitter in the SFN mode (zones 3a, 3b, and 3c in Fig. 3), ismostly larger than 9 dB and the gains there tend to be negative(red and orange). For the segments ‘in the middle’ between thethree transmitters (low ), the gain is positive. This indicatesagain that low values of correspond with positive SFNGs.

Table IV lists the number and the percentage of retainedsegments in the different quality categories for and

and the difference when switched from to. The data is classified for statistical analysis into three

different categories of overlapping degree: no limit on(All segments), and limited to a maximum of 6 dB and3 dB. Table IV shows that areas with significant overlapping[this happens in segments in the middle between two or three

transmitters (zones 1 and 4 in Fig. 3) and in the segments faraway from all three transmitters (zone 5 in Fig. 3)], there is apositive gain. For an overlapping degree corresponding to an

lower than 3 dB, the number of perfect segments increasesfrom 4 to 20 when the operation mode is changed from MFNto SFN (compared to 124 vs. 127 when there is no upper limiton ). As expected, this effect is higher when is lower.

If the whole set of data is studied, no relevant SFNG isobserved (as shown in Table IV). This is a major conclusionfrom the methodology presented in this paper, being even morerelevant in the case of dense broadcast networks for mobile andportable services in urban areas, where the number, relativedelay, and relative amplitudes of signals from different trans-mitters in the SFN will be very variable. In these scenarios,in order to take advantage of the SFN gain, and also in orderto avoid coverage degradation caused by the SFN, it will benecessary to have an accurate planning tool suitable for portableand mobile reception in urban environments. These tools havebeen widely used in cellular network planning during the lastdecade and will be necessary for broadcast SFN planning forportable devices [26], [27].

Table V shows for different ranges of the median SFNGin the different quality categories, the number of segments avail-able, and the standard deviation. Segments of perfect quality( 2.87 dB) and segments with ( 3.4 dB) mostlyhave a negative gain. This corresponds with our previous find-ings that gains are negative close to each of the three transmit-ters. Segments of good and doubtful quality are likely to havepositive gains (0.91 dB and 2.2 dB, respectively). Low qualitysegments (mostly far from all transmitters, ) havepositive gains as well (2.2 dB), but are not really representativefor a real network. The standard deviations are mostly around 3to 4 dB. Generally, Table V shows that the larger the overlap-ping degree (or the lower ), the higher SFNG.

B. Comparison of SFNG for Segments With DifferentReception Qualities

In the following we will investigate ‘the shift between qualitycategories’ (defined in Table III) when the operation mode isswitched from MFN to SFN. Table VI shows for all four qualitycategories the number of segments that are lost to other qualitycategories when the network operation mode is switched fromMFN , left part of Table VI) to SFN .Firstly, Table VI shows that low quality segments are unlikelyto obtain much improvement (84% of low quality remains lowquality). These are mostly segments far away from all three basestations, so despite the positive gain there, the quality remainslow. Secondly, ‘doubtful’ and ‘good’ segments are more likelyto be improved. 42% of the good segments are improved toperfect segments, while 27% (2% of good to doubtful 25% ofgood to low) have a decrease in quality due to interference. 58%(25% 33%) of the doubtful tables are improved in quality(to perfect and good), while 33% gets worse (to low). Finally,18% of ‘perfect’ segments become low quality segments. Forthe most part, these are the segments in zone 2 of Fig. 3 (darkred in Fig. 4). This might be due to the difference in path lengthbetween these segments and transmitters 2 and 3 (or 2 and 1,respectively). As the environment is suburban, it is possible


TABLE VTHE MEDIAN SFNG IN THE DIFFERENT QUALITY CATEGORIES IN MFN OPERATION �� FOR DIFFERENT RANGES OF �

TABLE VINUMBER OF SEGMENTS MOVED FROM ONE QUALITY CATEGORY TO ANOTHER

WHEN OPERATION MODE CHANGES FROM MFN �� TO SFN��

Fig. 5. Indication of quality category the valid segments belong to in ��(red = low, orange = doubtful, light green = good, dark green = perfect).

that the reflected signal from transmitters 1 and 3 arrive toolate compared to signals received from transmitter 2, this wayimpairing the reception quality.

Thus, we can conclude that SFN operation improves goodand doubtful reception and decreases perfect quality reception.However, the MER mostly remains high enough to maintain per-fect reception quality. Thirdly, a lot of zones west of transmitter2 have a high decrease in reception quality with a highly nega-tive SFNG (see Fig. 4, zone 2 in Fig. 3).

C. MFN Versus SFN

Fig. 5 shows for each segment to which quality category itbelongs in (MFN mode), while Fig. 6 shows the same,but in (SFN mode). The color code is explained in thefigure captions. Firstly, when we compare the figures, we see

Fig. 6. Indication of quality category the valid segments belong to in ��(red = low, orange = doubtful, light green = good, dark green = perfect).

that reception quality has improved in the middle between trans-mitters 1 and 2 (more dark green instead of light green and or-ange). As indicated in Fig. 3 (zone 4), is low there too andSFNG is positive (see Fig. 4). Secondly, the parts of the routeclose to each of the transmitters have a small decrease in recep-tion quality (zones 3a, 3b, and 3c in Fig. 3). Fig. 4 indeed showsthat the gains are highly negative there, but the signal strength ishigh enough to maintain sufficient reception quality. Thus, wecan conclude that SFN operation improves good and doubtfulreception and decreases perfect quality reception. However, theMER mostly remains high enough to maintain perfect receptionquality. Thirdly, a lot of zones west of transmitter 2 have a highdecrease in reception quality with a highly negative SFNG (seeFig. 4, zone 2 in Fig. 3, see Section VI-B). From Figs. 3–6, wecan conclude that the SFNG depends both on the location rela-tive to the transmitters and the overlapping degree between thetransmitters.

D. SFNG in Link Budget and Summary

Finally, we calculate a value for SFNG which can be used inlink budget calculations. We should only look at segments onthe border between two transmitters (where the signal strengthreceived by both transmitters is similar) and therefore we willretain the segments where is lower than 9 dB and where thereception quality in is doubtful or good. Low qualitysegments are excluded in order to exclude zone 5 (not realisticfor an actual network) and zone 2 (unrealistic positioning, othertransmitters would surround this zone in a real network). With


these restrictions and from Table V, the median SFNG equals1.1 dB with a standard deviation of 3.3 dB (good and doubtfulsegments with in Table V are retained). This gainof 1.1 dB is compared to an ideal MFN, which is already betterthan a realistic MFN. A significant advantage of the use of anSFN, confirmed by experimental results is that the standard de-viation on the MER decreases significantly: when the entire tra-jectory is considered, the standard deviation drops from 5.21 dBfor to 2.69 dB for .

VII. COMPARISON OF RESULTS WITH OTHER STUDIES

Only limited research about SFNG is available [1]–[5] inprevious works. In [1] and [2], SFN gains have been studiedusing different methods. In [1], measurements have been per-formed for an SFN with two transmitters with 600 W and 1200W Equivalent Radiated Power (ERP) transmitter power, respec-tively. No actual gains were calculated in [1], but the benefit ofSFN networks was demonstrated. It was shown that an SFN im-proved the coverage, but the results obtained with this approachwere entirely dependent on the chosen route. This work con-cluded that an extra transmitter would improve the coverage,but not necessarily thanks to the SFNG.

For tests in Finland, the SFNG was tested in three SFN testroutes, which were selected to be situated roughly in the middleof the two SFN transmitters [2]. An SFNG between 1.4 dBand 5.4 dB was obtained. This approach only used electric-field strength values, and did not consider the possible degrada-tion due to echoes from different transmitters. Also, in outdoorpedestrian and indoor corridor use cases, measurements withonly one transmitter were carried out at the University of Turku[2]. The results show better MFER at a certain signal level whenusing two transmitters. The SFNG is 1 2 dB. In [2], results offield tests in Barcelona to study SFNG have also been presented.The SFNG was calculated and treated as a statistical variable.The variable showed a lognormal behavior with a mean valueof 6.1 dB and a standard deviation of 6.4 dB. The range of thevalues is very wide so the study concluded that it is not recom-mendable to use SFNG values in link budgets.

In [3], SFN gains around 3 dB are obtained. The consid-ered locations were far from the transmitter. These are segmentscomparable to zone 5 (low quality, ) in this paper(see Fig. 3). The value of 3 dB corresponds well with our SFNGvalues of 1.98 dB and 3.61 dB as shown in Table V for lowquality segments with an or .

In [4] the MER was measured at locations where the receptionquality was perfect in MFN mode (MER of around 27 dB withone transmitter active). When the second transmitter is activatedso that the received signal strengths from two transmitters aresimilar the MER becomes about 7 dB lower(SFNG of about 7 dB). This matches very well the value of

7.92 dB obtained in this paper (see Table V).In [5] SFN gains have already been calculated for the network

considered in this paper. A value of 0.3 dB was obtained, butthe approach was too general: no distinction was made betweenthe behavior close to the transmitter (negative SFNG) and thebehavior on the border between network cells (positive SFNG).

The studies mentioned show that the determination of theSFN gain is very dependent on the choice of the measurement

route. Routes only in the vicinity of one transmitter will lead tolimited SFNGs: the contribution of the nearby transmitter willbe dominant compared to the one of the other (further) trans-mitter (1.4 dB in [2]). This paper has proven that even nega-tive gain values are possible in segments that are really closeto one transmitter. SFNGs will also be higher in zones betweentransmitters.

This paper proposes a SFNG calculation methodology thatmakes use of quality categories and that is thus independent ofthe chosen route. We have proposed an unambiguous method-ology to analyze the SFNG and we have applied it to an ac-tual DVB-H network. This analysis provides more insight intothis so far rather unclear phenomenon. Comparison with relatedwork shows that results that seemed contradictory in previousworks (SFNG of 3 dB in [3] vs. SFNG of 7 dB in [4]) are ex-plained if the results in this paper are taken into account.

VIII. CONCLUSION

In this paper we have proposed a new methodology to obtainunambiguously the SFN gain. The method has been explainedusing experimental work from a DVB-H network in Ghent,Belgium. Four quality categories (perfect, good, doubtful, low)have been defined and their influence on the SFN gain has beenanalyzed. The relationship between the SFN gain and the geo-graphical location has also been investigated. The results showthat gain is negative on locations very close to one transmitter,but since the recorded MER there is high enough, it has no realnegative influence on the reception quality. In general, when thewhole set of data is studied, no relevant SFNG is observed, buton locations where at least two transmitters provide the receiverwith a similar signal strength , the SFN gain ispositive and the reception quality is improved. An SFN gainof 1.1 dB with a standard deviation of 3.3 dB is obtained forlocations on the border between network cells. Future studiesmay include the investigation of the SFN gain in networks withmore transmitters in urban environments. It is expected that ina large city the number of transmitters received at a locationmight be much higher than three, affecting the behavior of thereceiver and increasing the complexity of planning.

REFERENCES

[1] “DVB-H Field Trials, Testing Mobile TV,” Instinct, Metz, France, 2005[Online]. Available: http://dea.brunel.ac.uk/instinct/PublicDocs.htm#Deliverables

[2] P. Talmola et al., “Wing TV, Services to Wireless, Integrated, No-madic, GPRS-UMTS and TV handheld terminals,” Tech. Rep., Nov.2006 [Online]. Available: http://www.celtic-initiative.org/Projects/WING-TV/default.asp

[3] G. Santella, R. D. Martino, and M. Ricchiuti, “Single frequency net-work (sfn) planning for digital terrestrial television and radio broadcastservices: The Italian frequency plan for t-dab,” in Veh. Technol. Conf.,May 17–19, 2004, vol. 4, pp. 2307–2311.

[4] J. Boveda, G. Marcos, J. M. Perez, S. Ponce, and A. Aranaz, “MERdegradation in a broadcast mobile network,” in 2009 IEEE Int. Symp.Broadband Multimedia Syst. Broadcast., Bilbao, Spain, May 13-15,2009mm09-71.

[5] D. Plets, L. Verloock, W. Joseph, L. Martens, E. Deventer, and H.Gauderis, “Weighing the benefits and drawbacks of an SFN by com-paring gain and interference caused by operation,” in 2009 IEEE Int.Symp. Broadband Multimedia Syst. Broadcast., Bilbao, Spain, May13-15, 2009mm09-34.


[6] S. G. Tanyer, T. Yucel, and S. Seker, “Topography based design of theT-DAB SFN for a mountainous area,” IEEE Trans. Broadcast., vol. 43,no. 3, pp. 309–319, Sep. 1997.

[7] S. I. Park, S. R. Park, H. Eum, J. Young Lee, Y.-T. Lee, and H. M.Kim, “Equalization on-channel repeater for terrestrial digital multi-media broadcasting system,” IEEE Trans. Broadcast., vol. 54, no. 4,pp. 752–760, Dec. 2008.

[8] R. Rebhan and J. Zander, “On the outage probability in single fre-quency networks for digital broadcasting,” IEEE Trans. Broadcast.,vol. 39, no. 4, pp. 395–401, Dec. 1993.

[9] A. Ligeti and J. Zander, “Minimal cost coverage planning for singlefrequency networks,” IEEE Trans. Broadcast., vol. 45, no. 1, pp. 78–87,Mar. 1999.

[10] S. O’Leary, “Field trials of an MPEG2 distributed single frequencynetwork,” IEEE Trans. Broadcast., vol. 44, no. 2, pp. 194–205, Jun.1998.

[11] T. Prosch, “The digital audio broadcast single frequency networkproject in southwest Germany,” IEEE Trans. Broadcast., vol. 40, no.4, pp. 238–246, Dec. 1994.

[12] A. Mattsson, “Single frequency networks in DTV,” IEEE Trans. Broad-cast., vol. 51, no. 4, pp. 413–422, Dec. 2005.

[13] D. Kateros, D. Zarbouti, D. Tsilimantos, C. Katsigiannis, P. Gkonis, I.Foukarakis, D. Kaklamani, and I. Venieris, “DVB-T network planning:A case study for Greece,” IEEE Antennas Propag. Mag., vol. 51, no. 1,pp. 91–101, Feb. 2009.

[14] D. Plets, W. Joseph, L. Martens, E. Deventer, and H. Gauderis, “Evalu-ation and validation of the performance of a DVB-H network,” in 2007IEEE Int. Symp. Broadband Multimedia Syst. Broadcast., Orlando, FL,Mar. 2007.

[15] D. Plets, W. Joseph, L. Verloock, E. Tanghe, L. Martens, E. Deventer,and H. Gauderis, “Influence of reception condition, MPE-FEC rateand modulation scheme on performance of DVB-H networks,” Spe-cial Issue on Quality Issues in Multimedia Broadcasting, IEEE Trans.Broadcast., vol. 54, no. 3, pp. 590–598, Sep. 2008.

[16] D. Plets, W. Joseph, L. Verloock, E. Tanghe, L. Martens, H. Gauderis,and E. Deventer, “New method to determine the range of DVB-H net-works and the influence of MPE-FEC rate and modulation scheme,”presented at the EURASIP Journal on Wireless Communicationsand Networking, 524163 [Online]. Available: http://www.hin-dawi.com/journals/wcn/2009/524163.html, unpublished

[17] G. Faria, J. A. Henriksson, E. Stare, and P. Talmola, “DVB-H: Digitalbroadcast services to handheld devices,” Proc. IEEE, vol. 94, no. 1, pp.194–209, Jan. 2006.

[18] J. Paavola, H. Himmanen, T. Jokela, J. Poikonen, and V. Ipatov, “Theperformance analysis of MPE-FEC decoding methods at the DVB-Hlink layer for efficient IP packet retrieval,” IEEE Trans. Broadcast., vol.53, no. 1, pp. 263–275, Mar. 2007.

[19] M. Poggioni, L. Rugini, and P. Banelli, “DVB-T/H and T-DMB:Physical layer performance comparison in fast mobile channels,”IEEE Trans. Broadcast., vol. 55, no. 4, pp. 719–730, Dec. 2009.

[20] D. Gomez-Barquero, D. Gozalvez, and N. Cardona, “Application layerFEC for mobile TV delivery in IP datacast over DVB-H systems,” IEEETrans. Broadcast., vol. 55, no. 2, pp. 396–406, Jun. 2009.

[21] H. Shirazi, J. Cosmas, T. Owens, Y.-H. Song, and A. Centonza, “A QoSmonitoring system in a heterogeneous multi-domain DVB-H platform,”IEEE Trans. Broadcast., vol. 55, no. 1, pp. 124–131, Mar. 2009.

[22] Y. Zhang, J. Cosmas, M. Bard, and Y.-H. Song, “Diversity gain forDVB-H by using transmitter/receiver cyclic delay diversity,” IEEETrans. Broadcast., vol. 52, no. 4, pp. 464–474, Dec. 2006.

[23] “Digital Video Broadcasting (DVB); Transmission to handheld termi-nals (DVB-H); validation task force report,” ETSI, TR 102 401 v1.1.1,May 2005.

[24] W. C. Y. Lee, Mobile Communications Design Fundamentals. NewYork: Wiley and Sons, Inc., 1993.

[25] “Digital Video Broadcasting (DVB); DVB-H implementation guide-lines,” ETSI, TR 102 377 v1.1.1, Feb. 2005.

[26] Y. Corre and Y. Lostanlen, “Three-dimensional urban em wave prop-agation model for radio network planning and optimization over largeareas,” IEEE Trans. Veh. Technol., vol. 58, pp. 3112–3123, Sep. 2009.

[27] T. K. Sarkar, Z. Ji, K. Kim, A. Medouri, and M. Salazar-Palma, “Asurvey of various propagation models for mobile communication,”IEEE Trans. Antennas Propag., vol. 45, no. 3, pp. 51–82, 2003.

David Plets was born in 1983 in Torhout, Belgiumon the 26th of May. After an education in mathe-matics and sciences in secondary school, he beganhis engineering study in Ghent. In July 2006, he fin-ished his final year dissertation on the developmentof an iDTV framework for sport coverage on theMultimedia Home Platform (MHP) and he obtaineda Master in Electrotechnical Engineering, with ICTas main subject. Currently, he is a member of theWiCa research group (Department of InformationTechnology - INTEC, Ghent University), where he

mainly focuses on performance and propagation of wireless technologies, suchas DVB-H.

Wout Joseph (M’05) was born in Ostend, Belgiumon October 21, 1977. He received the M. Sc. degreein electrical engineering from Ghent University (Bel-gium) in July 2000.

From September 2000 to March 2005, he was aresearch assistant at the Department of InformationTechnology (INTEC) of the same university. Duringthis period, his scientific work was focused on elec-tromagnetic exposure assessment. His research workdealt with measuring and modeling of electromag-netic fields around base stations for mobile commu-

nications related to the health effects of the exposure to electromagnetic radi-ation. This work led to a Ph.D. degree in March 2005. Since April 2005, he ispostdoctoral researcher for IBBT-Ugent/INTEC (Interdisciplinary institute forBroadBand Technology). Since October 2007, he is a Post-Doctoral Fellow ofthe FWO-V (Research Foundation - Flanders). His professional interests areelectromagnetic field exposure assessment, propagation for wireless communi-cation systems, antennas and calibration. Furthermore, he specializes in wirelessperformance analysis and Quality of Experience.

Pablo Angueira (M’04) received the M.S. and PhDin telecommunication engineering at the Universityof the Basque Country (Spain) in 1997 and 2002respectively. Prof. Pablo Angueira joined the Dpt. ofElectronics and Telecommunications of the Univer-sity of the Basque Country in 1998. Within the TSRResearch Group, he has been involved for 10 yearsin research projects to evaluate digital broadcastingsystems in different frequency bands. His currentresearch activities are related to DRM and newbroadcast technologies (DVB-T2 and NGH). Pablo

Angueira is co-author of a wide list of international peer reviewed journals,conference presentations related to digital terrestrial broadcast planning. Hehas supervised several PhD dissertations in this field. He has also co-authoredcontributions to the ITU-R working groups WP6 and WP3. Dr. Angueira isan Associate Editor of the IEEE Trans. Broadcasting journal and Chairs theStrategic Planning Committee of the Broadcast Technology Society Adminis-trative Committee.

José Antonio Arenas received the M.S. in telecom-munication engineering at the University of theBasque Country (Spain) in 2000. He has workedin Telefónica Móviles as radio engineer for 9 yearsin mobile communications network planning andoptimization. He also has been working part time asProfessor in the Dpt. of Electronics of the Universityof Extremadura at Merida Faculty of Engineering forthe last 6 years. His current research interests includeseveral aspects of digital broadcast planning. He isworking toward his Thesis on mobile multimedia

broadcasting network planning.


Leen Verloock was born in Eeklo, Belgium onNovember 15, 1979. She received the Mastersdegree in electronics engineering from KatholiekeHogeschool Ghent (Belgium) in 2001. She thenjoined the Department of Information Technology(INTEC) of Ghent University where she is currentlyworking as technical and research assistant in theWireless and Cable Research group. She is workingon measuring and modeling the propagation ofelectromagnetics fields around wireless systems.She is also doing measurements of electromagnetic

fields around base stations in order to check their compliance with the exposurelimits.

Luc Martens (M’92) was born in Gent, Belgium onMay 14, 1963. He received the M. Sc. degree in elec-trical engineering from Ghent University (Belgium)in July 1986.

From September 1986 to December 1990 he wasa research assistant at the Department of InformationTechnology (INTEC) of the same university. Duringthis period, his scientific work was focused on thephysical aspects of hyperthermic cancer therapy. Hisresearch work dealt with electromagnetic and thermalmodeling and with the development of measurement

systems for that application. This work led to a Ph. D. degree in December 1990.Since January 1991, he is a member of the permanent staff of the InteruniversityMicroElectronics Centre (IMEC), Ghent, and is responsible for the research onexperimental characterization of the physical layer of telecommunication sys-tems at INTEC. His group also studies topics related to the health effects ofwireless communication devices. Since April 1993 he is Professor in electricalapplications of electromagnetism at Ghent University.

IEEE TRANSACTIONS ON BROADCASTING, VOL. 56, NO. 3...

Documents

Transcript of IEEE TRANSACTIONS ON BROADCASTING, VOL. 56, NO. 3...