Calculation model for SFN reception and reference receiver characteristics of ISDB-T system

Transcription

1 Report ITU-R BT.9-1 (5/13) Calculaton model for SFN recepton and reference recever characterstcs of ISDB-T system BT Seres Broadcastng servce (televson)

2 Rep. ITU-R BT.9-1 Foreword The role of the Radocommuncaton Sector s to ensure the ratonal, equtable, effcent and economcal use of the rado-frequency spectrum by all radocommuncaton servces, ncludng satellte servces, and carry out studes wthout lmt of frequency range on the bass of whch Recommendatons are adopted. The regulatory and polcy functons of the Radocommuncaton Sector are performed by World and Regonal Radocommuncaton Conferences and Radocommuncaton Assembles supported by Study Groups. Polcy on Intellectual Property Rght (IPR) ITU-R polcy on IPR s descrbed n the Common Patent Polcy for ITU-T/ITU-R/ISO/IEC referenced n Annex 1 of Resoluton ITU-R 1. Forms to be used for the submsson of patent statements and lcensng declaratons by patent holders are avalable from where the Gudelnes for Implementaton of the Common Patent Polcy for ITU-T/ITU-R/ISO/IEC and the ITU-R patent nformaton database can also be found. Seres of ITU-R Reports (Also avalable onlne at Seres BO BR BS BT F M P RA RS S SA SF SM Ttle Satellte delvery Recordng for producton, archval and play-out; flm for televson Broadcastng servce (sound) Broadcastng servce (televson) Fxed servce Moble, radodetermnaton, amateur and related satellte servces Radowave propagaton Rado astronomy Remote sensng systems Fxed-satellte servce Space applcatons and meteorology Frequency sharng and coordnaton between fxed-satellte and fxed servce systems Spectrum management Note: Ths ITU-R Report was approved n Englsh by the Study Group under the procedure detaled n Resoluton ITU-R 1. ITU 13 Electronc Publcaton Geneva, 13 All rghts reserved. No part of ths publcaton may be reproduced, by any means whatsoever, wthout wrtten permsson of ITU.

3 Rep. ITU-R BT REPORT ITU-R BT.9-1 Calculaton model for SFN recepton and reference recever characterstcs of ISDB-T system (1-13) TABLE OF CONTENTS Page Chapter I SFN wth delays less than guard nterval duraton In the case of a sngle SFN wave Mathematcal equatons used n ths text Note to CNR used n theoretcal calculaton Increase n the requred CNR Nose characterstcs n recevers Clppng nose Quantzaton nose of A/D converson Other ampltude proportonal noses Condtons that gves SFN falure SFN non-falure condtons Locaton varaton Locaton correlaton In the case of multple SFN waves Approprate functon to express multple SFN waves SFN falure probablty n the case of multple SFN waves Chapter II SFN wth delays exceedng guard nterval duraton In the case of SFNs wth delays exceedng guard nterval duraton SFN wave wth a large delay exceedng guard nterval duraton SFN wth relatvely small delays (but larger than guard nterval duraton) Alasng effects of scattered plot sgnals Calculaton method for requred DUR Re-consderaton on alasng effects of scattered plot sgnals Settng of FFT wndow... 6

4 Rep. ITU-R BT.9-1 Page Optmum poston of FFT wndow Protecton ratos for analogue to dgtal nterference Recever characterstcs to be specfed GI Mas characterstcs of recevers on the maret... 3 Chapter III Adjacent channel nterference Major factors affectng on nterference performance Recever confguraton Unnecessary components Out-channel nterference mmunty Alasng nose Chapter IV Fadng Appendx 1 Fadng margn for less than 1% of tme (Kumada s law) Appendx Intermodulaton nose A.1 Formulas for ntermodulaton calculatons A. Examples of ntermodulaton nose... 4 A.3 Determnaton of coeffcent Nm... 41

5 Rep. ITU-R BT Summary Ths Report provdes detaled consderatons on recever characterstcs under sngle frequency networ (SFN) condtons for ISDB-T system. It ntroduces new techncal parameters that domnate recever performances, n addton to the conventonal plannng parameters. The new parameters are ampltude proportonal nose (APN), FFT wndow settng margn and nterpolaton flter characterstcs used for reference carrer recovery. Usng these parameters, the overall recever characterstcs can be expressed by a sngle parameter called guard nterval mas characterstcs, whch s useful n estmatng whether or not the sgnal s receved correctly. Furthermore, the Report gves a reference recever characterstc that would be appled n frequency plannng and/or networ desgn of ISDB-T based broadcastng systems. The calculaton method for SFN recepton and the reference recever characterstcs have been establshed n ARIB TR-B14, whch s successfully appled n plannng and desgnng of broadcastng networs n Japan. It s shown that SFN does not wor uncondtonally and wors well only when the recepton sgnals are ept under certan condtons n terms of recepton voltages, desred to undesred ratos (DURs), delays between man and SFN sgnals and so on. Chapter I SFN wth delays less than guard nterval duraton Frst we wll derve the condton that gves sngle frequency networ (SFN) falure when a sngle SFN wave exsts. Then, we wll dscuss about the condtons when multple SFN waves exst. Also we wll gve some consderatons on the recever characterstcs that are necessary to estmate the area of falure. We call SFN wth delays less than guard nterval duraton as nner-gi SFN n short. 1 In the case of a sngle SFN wave The receved sgnal exhbts rpples n frequency response when a desred sgnal s receved wth SFN waves. In ths case, the bt error rates (BER) of the OFDM carrers postoned at peas n frequency response becomes better, because the nput sgnal levels are hgh for those carrers. On the other hand, the BER of the carrers postoned at dps n frequency response become worse because the nput levels are low. We can estmate the occurrence of SFN falure by calculatng the BERs for every carrer and summng up them to chec whether or not the total BER s worse than the requred value. Fgure 1 shows an example of frequency response of receved sgnal. Fgure 1a s the case where the desred sgnal s just at the level that gves the requred carrer-to-nose rato (CNR). In ths example, we assume the total BER beng worse than the requred value, as there are many carrers of whch BER s worse than the reference value. If we ncrease the levels of both desred and SFN sgnals, the number of carrers havng worse BER s decreased, and then the sgnal s correctly receved, as shown n Fg. 1b. The SFN falure taes place dependng not only on the desred to undesred ratos (DUR) but also on the levels of the receved sgnal tself.

6 4 Rep. ITU-R BT.9-1 FIGURE 1 Example of receved sgnal a) n the case of reference sgnal level b) n the case of sgnal level ncreased 1.1 Mathematcal equatons used n ths text The relatonshp between BER and CNR are as follows: for QPSK for 16-QAM for 64-QAM where: BER = Erfc BER = Erfc 8 BER = Erfc 4 C p : C a : N p : N a : 1 C N 1 p Ca p = = Erfc 1 and p Na N p C 3 1 p a p 1 and = = 1 Erfc 1 N p Na N p C 7 1 p a p 1 and = = 4 Erfc 4 N p Na N p average power of sgnal r.m.s. ampltude of sgnal nose power r.m.s. ampltude of nose and Erfc( x) = exp( t ) π x C dt. C C C C ( BER) 8 BER 3 4 BER 7 (1) () (3)

7 Rep. ITU-R BT The relatonshp between the Erfc above and Ndst (normal dstrbuton functon) generally used n mathematcs s as follows: where: Ndst Erfc 1 x ( x) = exp( t )dt π ( x) = exp( t ) = π π x = Ndst u: t. dt du 1 exp( u ) = exp( u ) x π x ( x) The relatonshp between the nverse functons of the above s: P P Erfc 1 = = Erfc Ndst 1 ( x) = Ndst( x) ( x) x = P 1 ( P) = Ndst du 1 Ndst 1 P (4) (5) 1. Note to CNR used n theoretcal calculaton Recommendaton ITU-R BT.1368 defnes reference CNRs for varous modulatons and forward error correcton (FEC) code rates, where the carrer level s expressed by the power of the transmsson sgnal that contans several nds of nformaton, such as scattered plots (SP), system controls, n addton to content sgnals. In the followng theoretcal calculaton, we wll deal wth content sgnals only and wll apply the reference CNRs beng a lttle dfferent from those n the Recommendaton. To eep consstency among varous modulatons, we defne the reference symbol error rate (SER) that gves the requred BER for each FEC, and we assume the reference SER s unchanged among modulatons. The reference CNRs are then derved from the reference SERs under addtve whte Gaussan nose envronment. Table 1 shows the reference SERs and CNRs appled n ths Report. TABLE 1 Reference SER and CNR appled n theoretcal calculaton FEC=1/ FEC=/3 FEC=3/4 FEC=5/6 FEC=7/8 Reference SER QPSK 4.3 db 5.8 db 6.9 db 8. db 8.8 db 16-QAM 1.5 db 1.3 db 13.4 db 14.6 db 15.5 db 64-QAM 15.9 db 17.9 db 19. db.5 db 1.4 db

8 6 Rep. ITU-R BT Increase n the requred CNR In the case of a sngle SFN wave, the frequency response of the receved sgnal s gven by equaton (6). where: a ( ω) = 1+ U + U cos ( ω τ) F (6) a U a : ampltude of SFN wave relatve to desred wave τ: delay of SFN sgnal. Fgure shows examples of the ncrease n the requred CNR when an SFN wave exsts. The horzontal axs of the graphs denotes the power of SFN wave relatve to the desred sgnal, that s, the nverse fgures of DUR. The vertcal axs denotes the ncrease n the requred CNR. The actual values of CNR necessary for correct recepton are obtaned by addng these ncreases to the reference CNR such as 19. db for 64-QAM-FEC3/4, 1.4 db for 64-QAM-FEC7/8, and so on. The curves n Fg. are obtaned by calculatng the CNR that gves BER of for FEC3/4 or for FEC7/8 aganst a number of SFN waves of whch ampltude, delay and phase are gven randomly. The relatonshps gven by these curves are called CNR Increase Functons n ths document. Snce CNR Increase Functon cannot be expressed by smple mathematcal formula, we ntroduce approxmated functon for t, as below: where: UdB: CN up γ ( UdB) = α exp [ β UdB ] ( UdB γ = α exp [ β UdB ] UdB + A UdB ( UdB > ) denotes power of SFN wave expressed (db) CN up (UdB): denotes ncrease n requred CNR expressed (db). The values of the coeffcents, α, β, γ and A are gven n Table. a ) (7)

9 Rep. ITU-R BT FIGURE Example of ncrease n the requred CNR a) FEC = 3/4 b) FEC = 7/8 TABLE Coeffcent for CNR Increase Functon Coeffcent Modulaton FEC:7/8 FEC:5/6 FEC:3/4 FEC:/3 FEC:1/ α β γ A 64-QAM QAM QPSK QAM QAM QPSK QAM QAM QPSK QAM QAM QPSK NOTE A = for sngle SFN wave condton (see.1 for multple SFN waves condton). The value of α corresponds to the maxmum ncrease of requred CNR, whch taes place at DUR = db. A large dfference s found n the maxmum CNR ncrease between FEC 3/4 and 7/8, comparng to -3 db n the case wthout SFN waves. Ths dfference must be noted n broadcastng networ desgn, n other words, FEC 7/8 could not be appled n the actual world where SFN s more or less used.

10 8 Rep. ITU-R BT Nose characterstcs n recevers The nose that affects recepton performance can be summarzed n two categores; one s the nose of whch ampltude s ndependent of the nput sgnal, such as thermal nose, and the other s the nose of whch ampltude ncreases/decreases accordng to the nput sgnal level. The former s called fxed nose and the latter ampltude proportonal nose (APN) n ths text. Fxed nose conssts of thermal nose, man-made nose, nose fgure of recevers, and so on. An example of ln budget reported by the Natonal Councl on Informaton and Telecommuncaton n Japan apples thermal nose of 3 K, man-made nose of 7 K, and nose fgure of 3 db. The value of fxed nose s estmated to be 8.5 db(μv) (@75Ω) n ths case. Ampltude proportonal nose s manly determned by recever characterstcs, such as quantzaton nose of A/D converson, clppng nose, phase jtter of local oscllator. The words fxed and ampltude proportonal come from the features of nose when expressed equvalently at the recever nput termnal Clppng nose Snce the ampltude dstrbuton of OFDM sgnal exhbts a normal dstrbuton, an enormous dynamc range s requred for recevers to treat the sgnal wthout dstortons. Normal recevers clp the sgnals exceedng a certan level, and generate clppng nose n some extent. Fgure 3 explans clppng. The clpped waveform s equvalent to the nput waveform wth addng the sgnal components that exceed the clppng level n the opposte polarty. The component that exceeds the clppng level has a pulse waveform, as shown n the fgure. The probablty that an OFDM sgnal taes a level of x s wrtten by Gauss(x/S rms ), where S rms denotes r.m.s. ampltude of the sgnal, and Gauss(*) probablty densty functon of a normal dstrbuton. The ampltude of the pulse component that exceeds the clppng level s wrtten by (x CL), where CL denotes the clppng level. Usng the above, we can estmate the power of the pulses or clppng nose as below: NP clp = CL S rms ( x CL) Gauss ( x S ) The spectrum of clppng nose can be regarded as flat nose, because the ntervals of these pulses are consderably large and solated pulse has flat spectrum. rms dx (8)

11 Rep. ITU-R BT FIGURE 3 Clppng nose 1.4. Quantzaton nose of A/D converson It s assumed for normal recevers that the clppng level s set equal to the full scale of the A/D converter. Then, the quantzaton nose s gven by the well-nown equaton (9). where: LSB: NP adc = LSB = 1 N N ( CL ) 1 = CL 1 denotes the ampltude of the least sgnfcant bt N: denotes the bt-length of A/D converter. The nose power gven by equatons (8) and (9) s n proporton to square of clppng level, that s, the ampltude of the nose s n proporton to the clppng level. As the recevers usually adjust the sgnal ampltude by AGC to a certan level correspondng to the clppng level, we can regard these noses to be n proporton to the nput sgnal level. Fgure 4 shows examples of ampltude proportonal nose. It s seen from the fgure that the clppng level should be set at approxmately 4 tmes (+1 db) the r.m.s. sgnal ampltude and that the optmum clppng level depends on the bt-length of A/D converson. (9)

12 1 Rep. ITU-R BT.9-1 FIGURE 4 Ampltude proportonal nose Other ampltude proportonal noses There are a lot of ampltude proportonal nose sources other than descrbed above, such as phase jtter of PLL, computatonal errors, and so on. In addton, transmtted sgnals nclude APN such as ntermodulaton components generated n the transmtter power amplfers, noses ncluded n the transmtter nput sgnal (n the case of relay statons wth broadcastng wave relay), etc. We can treat these transmtter noses to be equvalently generated n the recever, and we wll express the APN ncludng all nose sources n a sngle value relatve to the r.m.s. ampltude of the nput sgnal. 1.5 Condtons that gves SFN falure Now we wll express the factors that are used n the dgtal recepton estmaton. They are: where: CN: CN : CN = CN NP = NP S rms = fx D CN rms up + NP + U ( Urms Drms ) ( S ) amp rms requred CNR when SFN wave exsts requred CNR wthout SFN waves CN up (*): CNR Increase Functon defned by equaton (7) U rms : D rms : S rms : NP: NP fx : NP amp (*): r.m.s. ampltude of SFN wave r.m.s. ampltude of desred wave r.m.s. ampltude of the nput wave composed of desred and SFN waves rms sum of fxed nose and ampltude proportonal nose nose power of fxed nose nose power of ampltude proportonal nose expressed as a functon. When the receved sgnal s hgher by CN tmes the nose power calculated by equaton (1), the dgtal sgnal s correctly receved, as the CNR of the receved sgnal beng hgher than the requred (1)

13 Rep. ITU-R BT one. On the contrary, when the sgnal s lower than CN tmes the nose power, the dgtal sgnal cannot be correctly receved, as the CNR of the receved sgnal s lower than the requred one. Hence, the condton that gves SFN falure can be obtaned by resolvng the followng equaton: D rms = CN NP = CN CN up ( U D ) rms rms NP fx + NP amp Drms + U Fgure 5 shows examples of SFN falure condtons calculated by equaton (11) n the case of APN of 4 db. The area surrounded by the pn curve represents the condtons that gve SFN falure. In the case of FEC 3/4, SFN falure does not tae place even at the worst case of DUR = db when the desred sgnal s hgher than a certan level, whle n the case of FEC 7/8, t occurs around DUR = db regardless of the desred sgnal level. rms (11) FIGURE 5 Example of SFN falure condtons (64-QAM, APN = 4 db) SFN sgnal voltage (db( μ V)) Mod: 64 QAM FEC: 3/4 APN: 4 db SFN sgnal voltage (db( μ V)) Mod: 64 QAM FEC: 7/8 APN: 4 db Desred sgnal voltage (db( μv)) a) FEC = 3/ Desred sgnal voltage (db( μv)) b) FEC = 7/8 Report BT.9-5 Fgure 6 shows examples of SFN falure n the case of APN of 8 db. In ths case, the falure taes place just on the condton of DUR = db even wth FEC 3/4. The range of DUR that generates SFN falure becomes wder wth FEC 7/8. The SFN falure condtons thus depend on the FEC appled as well as on the APN, whch s one of the recever characterstcs to be specfed.

14 1 Rep. ITU-R BT.9-1 FIGURE 6 Example of SFN falure condtons (64-QAM, APN = 8 db) SFN sgnal voltage (db( μ V)) Mod: 64 QAM FEC: 3/4 APN: 8 db SFN sgnal voltage (db( μ V)) Mod: 64 QAM FEC: 7/8 APN: 8 db Desred sgnal voltage (db( μv)) a) FEC = 3/ Desred sgnal voltage (db( μv)) b) FEC = 7/8 Report BT SFN non-falure condtons It s attractve that there exst condtons wthout SFN falure, as shown n Fg. 5. We wll gve consderatons on whch condtons SFN falure does not occur. We wll tae nto account APN only, snce the condtons wthout SFN falure are lmted n the hgh sgnal voltage range n whch fxed nose can be neglected. The maxmum ncrease n the requred CNR taes place at DUR = db, as shown n Fg., and the value s equal to α n Table. For example, the requred CNR s calculated to be 8.1 db by addng the CNR ncrease of 8.9 db to the reference CNR of 19. db n the case of 64-QAM-FEC3/4, and t s 41.1 db (ncrease of 19.7 db + reference of 1.4 db) n the case of 64-QAM-FEC7/8. The nput sgnal level s 3 db hgher compared to the desred sgnal when DUR = db, and hence, the APN s also larger by 3 db. Then we wll obtan the condtons for SFN non-falure when the desred sgnal s hgher than ths ncreased CNR. NPamp ( CN + CNup(max) + 3) = ( CN + α + 3) db (1) It can be derved by resolvng equaton (1) that SFN non-falure condtons requre the APN beng less than 3.3 db for 64-QAM-FEC3/4 and 44.8 db for 64-QAM-FEC7/8. It s seen n Fg. 4 that the APN of 35 db can be realzed wth an 8-bt A/D converter, and that 1-bt or more s requred to obtan APN of 45 db. It may be dffcult to realze APN to be less than 45 db when tang nto account the other nose sources such as jtter of PLL, etc. Thus, t depends on the recever performance whether or not the SFN falure taes place and we have to specfy the reference recever characterstcs to be used n the broadcastng networ desgn. 1.7 Locaton varaton If we can predct accurate values of feld strength of the desred and SFN waves, we can estmate whether or not SFN falure taes place at the locaton concerned, by applyng equaton (1) and/or Fg. 5. However, we ntroduce a statstcal method, because t s mpossble to predct accurate feld strengths.

15 Rep. ITU-R BT Here we assume the mean and standard devaton of a feld strength dstrbuton n the area concerned to be μ x and σ x respectvely for the desred wave, and those for the SFN wave to be μ y and σ y. Further assumng the correlaton between the desred wave and the SFN wave to be ρ, then we can express the probablty densty of feld strengths as follows: 1 F( Ex, Ey ) = exp[ { a11( Ex μx ) + a1( Ex μx )( Ey μ y ) + a( Ey μ y ) }] π A (13) where, A = σ x σ y ( 1 ρ ), a = σ A, a = ρ σ σ A, a = σ A 11 y We can calculate the probablty of SFN falure occurrence at a partcular locaton by ntegratng equaton (13) over the condtons that gve SFN falure. Then the probablty of SFN falure occurrence s wrtten by equaton (14). P μ, μ = F E, E de de (14) ( x y ) ( x y ) x y 1 SFN falure condtons Fgure 7 shows the calculaton results of equaton (14) for 64-QAM-FEC3/4 and for 64-QAM-FEC7/8, respectvely. The probablty of the falure for 64-QAM-FEC3/4 s almost zero n the regon of receved voltage beng hgher than 4 db(μv), whle t remans at 1-% even n hgh recepton voltage regons for 64-QAM-FEC7/8. We apply equaton (14) as the basc algorthm of dgtal recepton smulator, whch s used n the channel plan calculaton, especally for small-scale transmsson statons. x y x FIGURE 7 Example of probablty of SFN falure occurrence (64-QAM) a) FEC = 3/4 b) FEC = 7/8 1.8 Locaton correlaton The feld strength dstrbutons of the desred and SFN waves can be assumed to exhbt the same statstcs n each other, even f they come from dfferent drectons. Recommendaton ITU-R P.1546 descrbes the standard devaton of feld strengths for dgtal broadcastng waves to be 5.5 db. Consderng that the feld strength dstrbuton s caused manly by clutters near the recepton pont, such as terran, trees, buldngs, etc. feld strengths seem to have correlaton between the desred and SFN waves. For example, feld strengths would be lower at a hollow n the ground regardless of

16 14 Rep. ITU-R BT.9-1 wave drectons, or they would be hgher at a top of hll. The correlaton would be stronger wth closer the drectons. Fgure 8 shows changes n the probablty of SFN falure occurrence by the correlaton ρ. We wll apply ρ = as the most general condton, although the approprate values dffer from one recepton locaton to another. FIGURE 8 Changes n probablty of SFN falure by locaton correlaton a) correlaton = b) correlaton =.5 c) correlaton =.7 d) correlaton =.9 In the case of multple SFN waves In the case of a sngle SFN wave, the ncreases of requred CNR can be unquely defned by equaton (7), because the frequency rpples at a recever nput are decded by sole parameter of the ampltude of SFN wave,.e. U a n equaton (6). It cannot be unque n the case of multple SFN waves, because the rpples tae varous forms dependng on the ampltude, delay and phase of each SFN wave.

17 Rep. ITU-R BT Fgure 9 shows examples of frequency rpples generated by three waves (a man and two SFNs) at a recever nput. The total power of SFN waves s set at db (same as man wave) n both examples, but each SFN wave has a dfferent ampltude as ndcated n the fgures. The ncreases of requred CNR for 64-QAM-FEC7/8 are calculated to be 3 db and 14 db for Fgs 9a and 9b respectvely. Thus we cannot fnd a unque relatonshp between the total power of SFN waves and the ncrease of requred CNR, and hence we wll apply the statstcal approach to analyse multple SFN waves. FIGURE 9 Examples of frequency rpples a) Lttle ncrease n requred CNR b) Large ncrease n requred CNR Fgure 1 shows examples of the ncrease n requred CNR, where each dot (orange coloured) represents the ncrease aganst a set of SFN waves havng random ampltudes, delays and phases. Each graph contans 5 random trals. Fgure 1a s for the case of SFN waves, Fg. 1b for 3 SFN waves and Fg. 1c for 1 SFN waves. Snce the strongest wave n the receved sgnal s treated as the man wave, the dots do not exst n the range beyond the number of SFN waves (3 db for SFN waves, 4.8 db for 3 SFN waves and so on). The maxmum value n the horzontal axs taes place only when all receved waves have an dentcal ampltude. FIGURE 1 Examples of ncrease n requred CNR (64-QAM-FEC7/8) a) SFN waves b) 3 SFN waves c) 1 SFN waves

18 16 Rep. ITU-R BT Approprate functon to express multple SFN waves The green coloured curves n Fg. 1 express the trend curves for the ncreases of requred CNR calculated aganst a number of random trals, and they are reproduced n Fg. 11a to show dfferences by the number of SFN waves. Fgure 11b shows dstrbuton of the ncrease n requred CNR, where the horzontal axs denotes db dfference between the actual ncrease for random SFN waves and the trend curve. The vertcal axs denotes cumulatve percentage. The dstrbutons are almost dentcal except for sngle SFN wave and we wll adopt here the trend curve for 1 SFN waves as the representatve one. It s seen from the fgure that 9% of random trals are covered at a db dfference of 1.1 db, 95% at 1.7 db and 99% at 3. db. FIGURE 11 Trend curve and dstrbuton of ncrease n requred CNR (64-QAM) a) Trend curves (64-QAM) b) Dstrbuton of ncreases n requred CNR (64-QAM-FEC7/8) Trend curves are often used to express typcal characterstcs n the statstcal sense. However, the trend curves shown n Fg. 11a have problem that they exhbt a large dscrepancy between sngle SFN wave and multple waves, and we have to select the approprate one accordng to the number of SFN waves. Such selecton s mpractcal, because the number of SFN waves are unnown n most cases or t may change by ntroducton of new TX statons/gap-fllers and by varatons n propagaton condton such as fadng. To avod the above problem, we wll adopt CNR Increase Functons shown n Fg. 1a as the approprate functons that are to represent the ncreases of requred CNR for multple SFN waves. These functons are defned ndependent of the number of SFN waves. They apply the same coeffcent values as equaton (7) for α, β and γ. The values for coeffcent A are obtaned by least square approxmaton aganst a number of random trals, and are lsted n Table n 1.3. Fgure 1b shows the dstrbuton of dfference (referred to Dfference Dstrbuton herenafter) between actual ncreases of requred CNR and CNR Increase Functon. The graphs are smlar to those for the trend curves, and 9% of random trals are covered at a db dfference of 1.6 db, 95% at.3 db and 99% at 3.7 db. These values n db dfference are regarded as the margns that are added to CNR Increase Functon to cover the desgnated percentage of random trals. Table 3 summarzes the margns for CNR Increase Functon to cover 95% of random trals.

19 Rep. ITU-R BT FIGURE 1 Approprate functon and ncrease of requred CNR a) CNR Increase Functon (64-QAM) b) Dstrbuton of ncrease n requred CNR (64-QAM-FEC7/8) TABLE 3 Margns for CNR Increase Functon to cover 95% of random trals FEC 7/8 FEC 5/6 FEC 3/4 FEC /3 FEC 1/ 64-QAM.3 db 1.3 db.6 db.4 db.3 db 16-QAM.8 db 1.5 db.7 db.5 db.3 db QPSK 3.6 db. db 1. db.5 db.3 db. SFN falure probablty n the case of multple SFN waves Probablty of SFN falure caused by multple SFN waves can be calculated based on equaton (14). In the case of a sngle SFN wave, the ntegral boundary n equaton (14) s unquely determned by resolvng equaton (11). In the case of multple SFN waves, the ncreases of requred CNR dstrbute around CNR Increase Functon, and consequently the ntegral boundary dstrbutes, as shown n Fg. 13a.

20 18 Rep. ITU-R BT.9-1 FIGURE 13 Dstrbuted ntegral boundares and equvalent blurred sgnal ponts 6 db Wthn ± 1 db 6 db Wthn ± 1 db SFN sgnal voltage (db( μ V)) 4 Wthn ± Wthn ± 3 db Sgnal pont ( μ x, μ y ) SFN sgnal voltage (db( μ V)) 4 Wthn ± Wthn ± 3 db Blurred sgnal ponts 4 6 Desred sgnal voltage (db( μv)) a) Dstrbuted ntegral boundares 4 6 Desred sgnal voltage (db( μv)) b) Equvalent blurred sgnal ponts Report BT.9-13 The ntegral calculaton aganst the dstrbuted boundares n equaton (14) s obtaned by summng up the product of the defnte ntegral aganst one of the dstrbuted boundares and the probablty that the boundary taes place. Then we wll obtan equaton (15). where: P, (15) SFN falure condtons ( μx μ y ) = Prb() z F ( Ex, E y ) dex de y dz Prb ( z) : probablty densty functon of Dfference Dstrbuton z : devaton from CNR Increase Functon (or db dfference n Fg. 1b). Equaton (15) s mpractcal to calculate, and we wll mae approxmaton of t. Tang nto account that F ( E x, E y ) n equaton (15) s a -dmensonal normal dstrbuton around a sgnal pont at ( μ x, μ y ), we wll assume here that the defnte ntegral n the equaton taes an dentcal value under ether condton of a fxed boundary and blurred sgnal ponts (Fg. 13b) or the orgnal condton of dstrbuted boundares and a fxed sgnal pont (Fg. 13a). Further assumng that Dfference Dstrbuton s a normal dstrbuton, we can use equaton (14) wth modfed standard devatons nstead of σ and σ n equaton (13). x y Snce the dstrbuton of feld strengths (or sgnal voltages) and Dfference Dstrbuton are mutually ndependent events, the modfed standard devaton s smply obtaned by equaton (16) where: x nc σ = σ + σ and σ = σ + σ (16) σ xm and σ ym : denote modfed one for σ x and xm σ y ym y nc σ nc : denotes standard devaton of Dfference Dstrbuton.

21 Rep. ITU-R BT Wth the approxmated calculaton above, two ndependent dstrbutons, expressed by σ x (or σ y ) and σ nc, are combned nto a sngle dstrbuton of σ xm (or σ ym ). The values of σ nc are obtaned by dvdng the margns lsted n Table 3 by a factor of 1.64, as the margns are gven for cumulatve percentage of 95% whch corresponds to 1.64 tmes the standard devaton. When applyng the feld strength dstrbuton of σ x = 5.5 db, the modfed standard devatons are, for example, 6. db for 64QAM-FEC7/8 and 5.53 db for 64-QAM-FEC3/4, almost neglgble changes to the orgnal value of 5.5 db. Thus the number of SFN waves gves lttle effects on SFN falure probablty, and the consderatons gven n are generally applcable regardless of the number of SFN waves. Chapter II SFN wth delays exceedng guard nterval duraton When the delay of SFN wave exceeds the guard nterval duraton, the orthogonal features among OFDM carrers are lost, resultng n a large ncrease n the requred DUR. The Japanese channel plan apples DUR = 8 db ncludng margns for several factors. It may be reasonable for the basc channel plan to nclude margns, because the basc plan should guarantee stable networs operaton. If there were enough room n the spectrum, we could apply the plan values for each of the transmsson statons. But there s not enough spectrum n the real world, hence we have to establsh the approprate values, whch are requred especally for small-scale transmsson statons. 3 In the case of SFNs wth delays exceedng guard nterval duraton Frst we wll analyse the characterstcs of OFDM sgnals under the exstence of SFN waves havng delays exceedng guard nterval duraton. Then we wll dscuss about the recever characterstcs to be specfed. We call the SFN wth delays exceedng guard nterval duraton as outer-gi SFN n short. 3.1 SFN wave wth a large delay exceedng guard nterval duraton Snce SFN sgnal wth a very large delay has no correlaton to the desred sgnal, we can treat t as random nose. The requred DUR s calculated by the followng equaton: CN = N DU Crms fx + Urms = 1 CN CN and ( N C ) where: C rms and U rms : denote r.m.s. ampltude of the desred wave and of the SFN wave respectvely N fx : denotes fxed nose of recever CN : rato of desred sgnal to unnecessary components (the same values as the reference CNR) DU : requred DUR. fx DU rms C = U rms rms (17)

22 Rep. ITU-R BT SFN wth relatvely small delays (but larger than guard nterval duraton) We wll analyse the demodulated sgnals when an outer-gi SFN wave exsts. We assume the ampltude of delayed sgnal to be U and the porton of the delay exceedng GI to be τ (normalzed by FFT nterval). In Fg. 14, the demodulated sgnals are obtaned by multplyng each of OFDM carrers to the nput sgnal and then ntegratng each product over the FFT nterval. Assumng the orgnal OFDM sgnal to be expressed by equaton (18), then the demodulated component for the man sgnal (X) and that for delayed sgnal (Y) are wrtten as below: S ( j π ) = A exp t (18) o 1 1 = A ( jπt) exp( jπot) dt = A exp( jπ( o ) t) dt Ao X exp = (19) Y U τ ( jπt) exp ( jπ t) dt + A exp ( jπt) exp( jπ t) = Bexp o 1 τ o dt () FIGURE 14 OFDM demodulaton Tang nto account the guard nterval, the ntegral of the nd term n equaton () has the same result as ntegratng over vrtual nterval nstead of FFT nterval shown n Fg. 14. nd term = U = UA 1 o U A exp τ ( jπ( ) t) dt A exp ( jπ( ) t) A exp o 1 1 τ ( jπ( ) t) o dt o dt (1) τ ( jπ( ) t) dt U A exp( jπ( ) t) Y = UAo + U Bexp o o dt () The 1 st term n equaton () represents the sgnal component, the nd term nter-symbol nterference, and the 3 rd term nter-carrer nterference. Thus, the delayed sgnal, of whch delay exceeds the guard nterval, ncludes the sgnal component, and we call t as effectve nner-gi component. Ths component nterferes wth the man sgnal and produces rpples n frequency response, resultng n the requred CNR ncrease. Ths can be treated n the same way as the nner-gi SFN wave wth an excepton of nterferng duraton, whch s (1 τ) nstead of FFT nterval. The nter-symbol nterference,.e. the nd term n equaton () has no correlaton wth the sgnal τ

23 Rep. ITU-R BT component; t can be treated as random nose of whch power ncreases n proporton to τ. The expectaton of 3 rd term ncreases n proporton to τ when τ s close to zero, although the value tself vares accordng to the sgnal components A. Snce the 3 rd term equals to sgnal component when τ = 1, the expectaton of nterference component s n proporton to (1 τ) when τ s close to 1. Thus, the nter-carrer nterference,.e. the 3 rd term n equaton () can be regarded as random nose of whch power ncreases n proporton to τ(1 τ). We call these components that have no correlaton to the sgnal as effectve outer-gi component, and we can treat them n the same way as random nose. Tang nto account the above features of the nd and 3 rd terms n equaton (), we can obtan the requred DUR under outer-gi SFN envronments by resolvng equaton (3). where: Requred CNR = CN NoseComponent= τu + CN rms up ( UdB+ 1 log ( 1 τ) ) + τ(1 τ) U rms + N amp db (n real number expresson) ( τu + τ(1 τ) U + N ) = CN + CN UdB + 1 log( τ) ( ) 1 log rms rms amp up 1 (3) U rms : denotes power of delayed wave N amp : nose power of APN CN : CN up (*): reference CNR (db) CNR Increase Functon (db) UdB: power of delayed wave (db). The soluton of equaton (3) at τ = corresponds to the requred DUR for nner-gi SFN envronment. When the equaton has no soluton at τ =, the SFN falure does not taes place. Equaton (4) s a concrete example of equaton (3), and Fg. 15 shows the correspondng graphs for the equaton. Fgure 16 shows the relatonshp between delay and requred DUR, whch s called bathtub curve. Requred CNR = CN Nose Component = + α exp where, α, β and γ are shown n Table n 1.3. γ ( β ( UdB + 1 log( 1 τ) ) ) UdB 1 NampdB 1 ( τ τ ) (4)

24 Rep. ITU-R BT.9-1 FIGURE 15 Requred CNR and nterference component under outer-gi envronment FIGURE 16 Bathtub curve (calculated for APN of 35 db) 3.3 Alasng effects of scattered plot sgnals In the prevous secton, we have analysed the nterference components n the OFDM demodulaton process, n whch we assumed mplctly that the rppled frequency response was deally compensated by some means. Ths compensaton process s equvalent to regenerate each of OFDM reference carrers from the receved scattered plot sgnals (SP herenafter). As the SP sgnals are sent n every 3 OFDM carrers, the mnmum frequency spacng n the rpples measured wth the SP sgnals s lmted, that s, the maxmum delay that can be observed s lmted. If there s a delayed wave exceedng ths maxmum value, correct observaton cannot be performed, just as alasng found n general samplng process. Fgure 17 explans the above crcumstance of SP sgnal. The fgure shows the frequency response when delayed waves wth a long delay and a short delay coexst. The frequency response has fne rpples due to the long delay wave, and the OFDM carrers tae the values as shown by * mars n the fgure. On the other hand, we wll estmate the rpples to be a rough curve as shown n the

25 Rep. ITU-R BT fgure, because we can observe the response sampled only at the SP sgnals. Therefore, the carrer references to be used n the demodulaton nclude errors n phase and ampltude. FIGURE 17 Frequency response estmated from SP sgnals The average error can be estmated n the same way as the general analyss of dstortons ncluded n the sgnal recovered from the sampled values, although the error of each carrer vares dependng on the delay and ampltude of SFN waves, characterstcs of the nterpolaton flter, and so on. The average error ( A e A e ) can be wrtten as follows: outband components + = nterpolaton error components (5) Note that n analysng a samplng system, the usual way s that the orgnal functon s a tme-doman functon and the converson functon s the Fourer spectrum, whle n equaton (5), the orgnal functon s Fourer spectrum and the converson functon s the tme-doman one. Fgure 18 shows the out-band components and nterpolaton errors used n equaton (5). The fgure s the case for the FFT nterval of 1 8 μs, and the value 336 μs corresponds to 1/3 of the FFT nterval. FIGURE 18 Delay profle, out-band components and nterpolaton errors

26 4 Rep. ITU-R BT.9-1 FIGURE 19 Effects of carrer recovery errors The effects of carrer recovery errors dffers by sgnal ponts on the constellaton, as shown n Fg. 19. It s large on the sgnal ponts at outer sde of the constellaton. To smplfy the calculaton, we assume here that the r.m.s. error at every sgnal pont has the same error at the SP pont. Ths smplfcaton results that the calculated BER tends to become worse than that of actual one. 3.4 Calculaton method for requred DUR When there are many nds of nterference factors, the requred DUR can be wrtten n general form as below: where: CN : ( U ) + W ( U ) 1 = W1 + CN Requred DUR = 1log( U ) db (6) denotes the requred CNR that s determned by modulaton and FEC (such as 19. db for 64-QAM-FEC3/4) U : denotes the power of delayed sgnal W 1 (*), W (*), etc.: represent the weghtng functons whch convert the delayed sgnal nto the equvalent nose power aganst each nterference factor. The major factors that affect correct recepton are nter-symbol nterference ( nd term n equaton ()), nter-carrer nterference (3 rd term), and carrer recovery error whch we are dscussng n ths secton. Referrng to equaton (3), we apply equaton (6) as below: 1 [ CN CN ( U )] = N + τ U + τ ( τ) U + ( τ) A up rms amp rms 1 rms 1 e The nd, 3 rd and 4 th terms n the rght sde of equaton (7) correspond to nter-symbol nterference, nter-carrer nterference, and carrer recovery errors, respectvely. Note that we neglect the fxed nose assumng the receved sgnal to be enough hgh. Resolvng equaton (7) for U, the requred DUR s obtaned by the nverse number of U. (7)

27 Rep. ITU-R BT The hardware characterstcs that affect the recever performance are N amp and A e n equaton (7), and the other terms are ndependent of hardware characterstcs. Therefore, we have to specfy these two for the reference recever to be appled n networ desgn. The actual values of ampltude proportonal nose were at 35 db for the worst recevers avalable on the maret. So we have adopted ths 35 db as the specfcaton of APN. Before we specfy the nterpolaton flter characterstcs, we wll tae an overvew on the flter characterstcs. When assumng an deal LPF as the nterpolaton flter, equaton (7) s wrtten as below: 1 [ CN CN up ( U rms )] = Namp + τu rms + τ ( 1 τ) U rms ( delay < 168μs) 1 [ CN CN ( U )] = N + τu + τ ( 1 τ) U + ( 1 τ) U ( delay 168μs) up rms amp rms In ths case, the OFDM carrers can be recovered perfectly and no recovery errors tae place when the delay s wthn the Nyqust band. When the delay s out of Nyqust band, A e equals to the power of the delayed wave tself. In the case where the nterpolaton flter s not an deal LPF, the power of nterpolaton errors ( A e ) s calculated by the below: A e = DL j rms rms [( LPF( DLj )) U j ] + < 168μs DL U 168μs (8) 1 (9) where: U j : represents the ampltude of delayed waves wthn the Nyqust band ( DL < 168μs ) U : ampltude of delayed waves outsde the Nyqust band ( Fgure shows examples of calculaton results for some nterpolaton flters. DL 168μs ). FIGURE Interpolaton flter and correspondng DUR characterstcs (by equaton (9))

28 6 Rep. ITU-R BT Re-consderaton on alasng effects of scattered plot sgnals We have assumed the alasng effects of SP sgnals to be n proporton to the average error power A e. However, the requred DUR of actual recevers s hgher by 1 db than that calculated by equaton (7). Ths suggests that the carrer recovery errors are affected by other factors n addton to the average error power. The recovery error for each OFDM carrer corresponds to the error voltage of that carrer but not to the r.m.s. error voltage. The error voltage dffers from carrer to carrer, and hence the BER dffers carrer by carrer. In these cases where the error voltage dffers by carrer, the total BER averaged over all carrers becomes worse compared to the case where every carrer taes the same error voltage. Ths s because the BER degradaton of the carrers wth error voltages larger than the average s more severe compared to the BER mprovement of the carrers wth smaller error voltages. Therefore, we need to add a new term n equaton (7) to express the effects due to error voltages beng dfferent from carrer to carrer. However, t s qute dffcult to derve theoretcally such effects. Consderng that the effects may dffer dependng on the delay and/or phase of the nput waves, and that the amount of effect s only 1- db, we ntroduce a new coeffcent to be consstent wth the measured results, as below: 1 ( ) ( ) ( ) CN CN U = N +τ U +τ 1 τ U τ A here, the value for the new coeffcent to be 1.5. up rms amp rms rms e Fgure 1 shows the examples calculated by equaton (3). (3) FIGURE 1 Interpolaton flter and correspondng DUR characterstcs (by equaton (3)) 3.5 Settng of FFT wndow Fgure shows the relatonshp between the FFT wndow n a recever and the requred DUR characterstcs or the bathtub curve. The fgure s the case of guard nterval of 16 μs (1/8 of FFT frame duraton). Fgure a shows the bathtub curve due only to the nterference components,.e. nter-symbol nterference and nter-carrer nterference. Fgure b represents the characterstcs of nterpolaton

29 Rep. ITU-R BT flter, whch has a unty response wthn the bandwdth of ±LPFbw and zero response outsde the Nyqust band (±168 μs). Fgure c shows the overall bathtub curve, whch ncludes all the factors. A recever adjusts ts FFT wndow n such a way that the man wave s postoned at the bottom of the bathtub curve. Otherwse, a large nter-symbol nterference taes place. Therefore, the adjustment range of the FFT wndow s lmted wthn ±GI/ (GI denotes guard nterval duraton). In the actual hardware, some margns are necessary to avod mss-adjustment, and the adjustment range of FFT wndow s lmted wthn ±(GI/ T m ) as shown n the fgure. Note that the poston of μs n Fg. c s dfferent from Fgs a and b, as t s a custom to express the delays relatve to the man wave. FIGURE FFT wndow and bathtub curve Optmum poston of FFT wndow Here we summarze how to estmate the recepton falure. 1) To splt each SFN wave nto effectve nner-gi component and effectve outer-gi component. Regardng effectve nner-gi component, we assume one equvalent SFN wave of whch power s equal to the sum of each effectve nner-gi component, that s, ( τ ) U eq = 1 U (31) Regardng effectve outer-gi component, we assume equvalent nose, whch s the sum of nter-symbol nterference; nter-carrer nterference and carrer recovery error of each SFN wave, that s, N eq = τ U + τ ( τ ) U ( 1 τ )( 1 LPF( DL )) 1 U (3)

30 8 Rep. ITU-R BT.9-1 ) To sum up all of the non-correlated nose powers, whch are fxed nose, co-channel dgtal waves (wth dfferent programs) and co-channel analogue waves n addton to equaton (3), that s, N total = N fx + N eq + CoD + CoA AtoD where, N fx denotes the fxed nose, CoD the power of a co-channel dgtal wave wth dfferent programs, CoA the power of a co-channel analogue wave, and AtoD the weghtng factor wth whch the analogue sgnal s dealt as equvalent random nose. 3) To calculate the probablty of falure occurrence, usng equatons (11) and (14), n whch the total nose power s gven by equaton (33). In ths case, the optmum poston of the FFT wndow s defned to be such one that mnmzes the value of equaton (33). 4) To apply the followng methods descrbed n paragraphs 5) and 6) nstead of the above paragraph 3), as t s not practcal to calculate equaton (11). 5) To plot the sgnal voltages of the desred wave and the equvalent nner-gi SFN wave on the -dmensonal voltage dagram, as shown n Fg. 3 (green coloured mar). To shft ths nput pont to the pont shown by brown coloured mar (left down drecton). The value of sft s gven by the followng equaton: total N fx (33) 1 log( N ) db (34) 6) To apply equaton (14) aganst the equvalent nput voltage,.e. the shfted pont n the fgure. SFN sgnal voltage (db μ V) FIGURE 3 Equvalent nput sgnal voltages : Actual nput voltage : Equvalent voltage Mod: 64-QAM FEC: 3/4 APN: 3 db Desred sgnal voltage (db μv) Report BT.9-3 The optmum poston of FFT wndow s stated n the above 3, but t s not useful to apply equaton (33) from the vew pont of computaton hours. Therefore we ntroduce the followng alternatve method.

31 Rep. ITU-R BT Fgure 4 shows the relatonshp between the nput waves and FFT wndow poston. The curve named GI Mas characterstcs n the fgure s the nverse bathtub curve (opposte polarty n db), whch ndcates the maxmum allowable levels of delayed sgnals. The recever adjusts ts FFT wndow so that all delayed waves become below the mas. If a delayed sgnal exceeds the mas, the recever cannot receve the sgnal correctly. When there are many delayed sgnals of whch ampltudes are close to (stll below) the mas, the recever may fal correct recepton. So, we ntroduce a new concept of faulty power, whch s defned to be the dfference n db between the ampltude of delayed sgnal and the correspondng mas. When the sum of the faulty power s larger than db, we regard the delayed sgnals to exceed the mas as a whole. We also assume that the recever adjusts ts FFT wndow to mnmze the total faulty power, as wrtten n the followng equatons: PdB P und = UdB = 1 PdB MasdB 1 ( DL ) ( db) mnmze (35) FIGURE 4 Optmum settng of FFT wndow 3.6 Protecton ratos for analogue to dgtal nterference The effects of nterference from analogue sgnal depends not only on the recever characterstcs but also on the analogue sgnal or analogue program contents. We consder here the protecton rato aganst the worst-case analogue sgnals. We have tested several recevers on the protecton ratos, n whch we apply colour-bar wth 1% modulated stereo sgnal as the worst-case analogue sgnal. The results measured for the worst recever s as follows. for 64-QAM-FEC3/4 protecton rato of 5 db for 64-QAM-FEC7/8 protecton rato of 13 db. 3.7 Recever characterstcs to be specfed As dscussed above, many of the factors related to SFN recepton are caused by OFDM tself, and are automatcally defned by the sgnal parameters appled. The factors caused by hardware, whch should be specfed for the reference recever, are lsted n Table 4 together wth the ARIB specfcatons.

32 3 Rep. ITU-R BT.9-1 TABLE 4 Factors to be specfed Item ARIB specfcaton Remars Ampltude proportonal nose 35 db Relatve to nput sgnal level Interpolaton flter for carrer recovery Flat 16 μs ~ 16 μs Transton 168~ 16 μs and 16~168 μs FFT wndow settng margn 6 μs Protecton rato nterfered by analogue TV 5 db 64-QAM-FEC3/4 13 db 64-QAM-FEC7/8 3.8 GI Mas characterstcs of recevers on the maret Fgure 5 shows examples of guard nterval mas of recevers avalable on the maret together wth the characterstcs (denoted by ARIB n the fgure) specfed n 3.7. Recevers except for the one expressed by dot lne n the fgure exhbt better characterstcs than the ARIB specfcaton. The dot-lned recever s one that was produced before the ARIB specfcaton has been establshed. The characterstc dfference among recevers comes from the characterstcs of nterpolaton flter used for reference carrer recovery (see 3.3). FIGURE 5 GI Mas of recevers avalable on the maret

33 Rep. ITU-R BT Chapter III Adjacent channel nterference 4 Major factors affectng on nterference performance The nterference of adjacent channel and/or any other channels than the desred channel (referred to out-channel herenafter) s caused by recever tself. It s obvous that no nterference would occur f an deal BPF s nserted at the recever nput termnal. Major factors affectng the nterference performance are ntermodulaton nose generated at the frst stage amplfer, out-channel suppresson characterstcs of flters, clp and quantzaton nose of A/D converter and alasng nose n dgtal processes. We wll consder the nter-relatonshps of the factors and derve a smple equaton to express the out-channel nterference performance of a recever. 4.1 Recever confguraton Fgure 6 shows the basc recever confguraton together wth the major factors affectng the nterference performance. Although there are many types of recevers that provde dfferent confguratons such as slcon tuner, the factors to be consdered are common. For example, drect converson type recevers, whch convert the RF sgnal drectly nto the baseband usng quadrature detecton, equpped wth LPF nstead of IF Flter to reduce undesred out-channel sgnals, where the factor to be consdered s out-channel suppresson characterstcs but not the flter types. So are the other factors. Image channel suppresson characterstcs are one of the factors to be consdered wth conventonal type recevers, but wll not be treated n ths document. Recent dgtal recevers use double converson or drect converson scheme that provdes superor suppresson performance. FIGURE 6 Equvalent confguraton of a recever 4. Unnecessary components Unnecessary components n a recever are fxed nose N fx manly generated n the frst stage amplfer, ntermodulaton nose caused by non-lnear characterstcs of the amplfer, quantzaton nose of A/D converter and out-channel components beng suppressed by IF flter. Input-output characterstcs of the frst stage amplfer are generally approxmated by a 3 rd order polynomal, where the ntermodulaton nose s n proporton to the 3 rd power of the nput sgnal as below. ( D U ) 3 Intermodul aton nose +

34 3 Rep. ITU-R BT.9-1 where: D and U denote the power of desred sgnal and of out-channel sgnals, respectvely. The ntermodulaton nose depends not only on the amplfer s non-lnear characterstcs and sgnal levels but also on the sgnal spectra and frequency spacng between desred and undesred sgnals. For detals, see Appendx. The Gan Control crcut n the fgure controls ts output sgnal to reman wthn the dynamc range of A/D converter. When out-channel sgnals exst at large levels, the crcut must reduce ts gan regardless of the desred sgnal level n order not to generate severe clppng nose, resultng n the desred sgnal beng reduced. Consequently the quantzaton nose (and accompaned clppng nose) relatve to desred sgnal s n proporton to the nput sgnal to A/D converter as well as the APN value (e.g..316 = 35 db) as expressed below. Quantzaton nose = APN ( D + U R) where R denotes suppresson rato of IF flter for out-channel sgnals. The value of R may or may not change by out-channel frequences. By summng up these noses and out-channel components, the overall unnecessary component s wrtten as below. N N eq m = N = X fx sat + APN 5 ( D + U R) + N ( D + U ) where: N m : denotes a coeffcent expressng the degree of amplfer s non-lnearty (see Appendx ) X sat : denotes nput sgnal level at whch the amplfer just saturate. Fgure 7a shows an example of unnecessary components calculated under the condton of APN of.316 (= 35 db), out-channel suppresson rato of 1 (= db), amplfer saturaton of 19 dbμv ( dbm) and desred sgnal level of 6 dbμv ( 49 dbm). The horzontal axs of the graph denotes out-channel sgnal level and the vertcal axs unnecessary component, both relatve to desred sgnal. The overall unnecessary component begns to ncrease at the out-channel sgnal level of db, whch corresponds to out-channel suppresson rato of the IF flter. The ntermodulaton nose s lower than the quantzaton nose as for ths condton. 4.3 Out-channel nterference mmunty We defne the allowable lmt of out-channel sgnal to be the level that gves threshold CNR for desred sgnal. It s obtaned by resolvng the followng equaton for U. CN threshold m ( D + U R) + N ( D + U ) 3 3 N eq (36) D D = = (37) N N + APN eq fx Fgure 7b shows the calculaton results of equaton (37) n the case of 64-QAM-FEC3/4. The dot lne shows the case of deal amplfer (wthout ntermodulaton nose) and the three curves correspond to SFN condtons,.e. no SFN wave, SFN waves of 3 db and db, for whch threshold CNRs are ncreased accordng to the CNR Increase Functon. The allowable lmt ncreases n proporton to the desred sgnal n the range lower than around 5 dbμv. In ths range, the nterference mmunty or threshold undesred-to-desred rato (UDR) taes constant values, such as 35 db for no SFN wave, 31 db for SFN of 3 db and 4 db for SFN m

35 Rep. ITU-R BT of db. Ths s because the overall unnecessary component ncreases n proporton to desred sgnals as shown n Fg. 7a. FIGURE 7 Unnecessary components and out-channel nterference characterstcs (APN = 35 db, R = db, X sat = dbm) a) Unnecessary components b) Out-channel nterference characterstcs The allowable lmt exhbts saturaton for desred sgnals hgher than around 6 dbμv where the ntermodulaton nose becomes domnant. In ths range, the threshold UDRs decrease wth desred sgnal beng hgher, such as 18 db for no SFN wave, 17 db for SFN of 3 db and 15 db for SFN of db at the desred sgnal of 8 db(μv). Fgure 8 shows another example wth a hgh out-channel suppresson rato of 5 db. In ths example, the out-channel sgnals are suppressed enough that the quantzaton nose has lttle effects on the nterference characterstcs. The threshold UDRs are hgh at low desred sgnals, such as 38 db for no SFN wave, 37 db for SFN of 3 db and 35 db for SFN of db at the desred sgnal of 5 dbμv, whle they decrease wth desred sgnal beng hgher. The saturaton n the allowable lmts can be relaxed by nsertng an attenuator at the recever nput (see Fg. 6) and by controllng t approprately n accordance wth nput sgnal levels. In ths case, the values of fxed nose N fx and amplfer saturaton level X sat n equaton (37) are replaced by those ncreased accordng to the attenuaton, as follows. N fx N fx 1 ATTdB 1 and X sat X sat 1 ATTdB 1 where: ATTdB denotes attenuaton of the attenuator n db. Fgure 9 shows examples when the attenuator s controlled. The saturaton level shown n the fgure depends on how to control the attenuator, but may be determned by other reasons than equaton (37), such as protecton of amplfer from breadown. The examples apply the maxmum nput level to be 19 dbμv or dbm. As dscussed above, the out-channel nterference characterstcs depend on the three major parameters,.e. APN, out-channel suppresson rato R and amplfer s 3 rd order dstorton coeffcent N. The characterstcs are gven on a -dmensonal dagram of desred and out-channel sgnals m

36 34 Rep. ITU-R BT.9-1 by resolvng equaton (37). It s nsuffcent to express the characterstcs by a sngle fgure of adjacent channel rejecton rato (sometmes referred to adjacent channel protecton rato). It should be noted that the out-channel nterference characterstcs depend on SFN condtons. FIGURE 8 Unnecessary components and out-channel nterference characterstcs (APN = 35 db, R = 5 db, X sat = dbm) a) Unnecessary components b) Out-channel nterference characterstcs FIGURE 9 Out-channel nterference characterstcs wth attenuator controlled a) APN = 35 db, R = db, X sat = dbm b) APN = 35 db, R = 5 db, X sat = dbm Tang these ssues nto consderaton, the out-channel nterference characterstcs of a reference recever should be specfed on the -dmensonal dagram, together wth the specfcatons for APN, out-channel suppresson rato and ntermodulaton factor (or amplfer saturaton level) so that the characterstcs can be calculated for any SFN condtons.

37 Rep. ITU-R BT Alasng nose The out-channel components beng converted nto the frequency band near the samplng frequency of A/D converter generate alasng nose. The out-channel components are reduced by the IF flter and the frequency converted components are further reduced by the LPF accompaned wth A/D converter. Addng the alasng nose to equaton (36), the overall unnecessary component s obtaned by the followng equaton. where N N a eq = LPF = N fx 1 U a R + APN a 3 ( D + U R) + Nm( D + U ) + Na N a : denotes alasng nose ncluded n the desred sgnal U a : power of out-channel components wthn the alasng frequency band R a : suppresson rato of the IF flter aganst the alasng frequency band LPF : out-band attenuaton of the LPF. (38) Chapter IV Fadng Recommendaton ITU-R P.1546 gves the relatonshp of recepton feld strength vs. propagaton dstance for varous tme avalabltes. We assume here fadng margns to be the dfference between the feld strengths of the desgnated tme avalablty and that of 5% tme (annual mean value). The Recommendaton states that the data s vald for tme avalabltes only between 5% and 1%. For the range less than 1% of tme, we assume the so-called Kumada s law, detals of whch are descrbed n Appendx 1. Fgure 3 shows the fadng margns derved from the above. The curves n the fgure correspond to transmsson antenna heghts. Snce these fadng margns are derved from the Recommendaton, general treatments, such as nterpolaton of tme avalablty, antenna heght, etc. are to follow the Recommendaton.

38 36 Rep. ITU-R BT.9-1 FIGURE 3 Fadng margns (6 MHz) a) 1% of tme (land path) b) 1% of tme (land path) c).1% of tme (land path) d).1% of tme (land path)

39 Rep. ITU-R BT Appendx 1 Fadng margn for less than 1% of tme (Kumada s law) Fgure 31 shows examples of measurement results of feld strength comng from Korea to Japan. The analogue TV waves emtted at several ctes n Korea were contnuously measured for a oneyear perod. Fgure 31a s the results measured durng June and Fg. 31b durng December. The feld strengths present bg changes n June whle almost constant n December. In ths area, fadng phenomenon taes places very often durng May to October and t s rarely durng December to March. FIGURE 31 Measurement examples of feld strengths comng from Korea (sea path) a) Fadng season (June 1) b) Non-fadng season (December 1) Fgure 3 shows the cumulatve percentage of feld strength measured n dfferent places. Although the feld strengths correspondng to a certan percentage of tme are dfferent by propagaton dstances, frequences (channels), etc. we can see a common feature from the measurement results, as follows.

40 38 Rep. ITU-R BT.9-1 FIGURE 3 Cumulatve percentage of feld strengths comng from Korea (sea path) a) Measured at Nshno-shma b) Measured at Toyota When the feld strengths for 1% of tme and 1% of tme are to be FS 1 and FS 1 (db(μv/m)), respectvely, the feld strengths for.1% and for.1% of tme can be obtaned by the followng equatons. 1 FS1 = FS1 + ( FS1 FS1) db (39) FS 1 = FS1 + ( FS1 1) db (4) 1 FS where, FS 1 and FS 1 represent the feld strengths for.1% of tme and.1% of tme respectvely. The above smple equatons are so-called Kumada s law. An example of t s shown n Fg. 33. FIGURE 33 Example of Kumada s law

41 Rep. ITU-R BT Appendx Intermodulaton nose A.1 Formulas for ntermodulaton calculatons We defne the nput/output characterstcs of the frst stage amplfer as follows. 3 ( 3x ) y = x Bx and B = 1 (41) where, x sat denotes the nput level at whch the amplfer just saturates. We express the nput sgnal as follows. A ωt The ntermodulaton components are the 3 rd power of the sgnal. x 3 = = a A 3 j sat x = cos a (4) + 6 [ cos3ω t + 3cosω t] + 3 A Aj [ cos( ω ± ω j ) t + cos ω jt] A A j a A a j cos a a j ( ω ± ω ± ω ) j> > j j 4 j t a (43) The above equaton ncludes components that fall nto the frequences 3 tmes the sgnal band, and we wll obtan the ntermodulaton nose by removng these components. B IM = j> > j 3 A cos ω t + 6 A A j A j j ( ω ω ) [ cos( ω + ω ω ) t + cos( ω ω + ω ) t + cos( ω + ω + ω ) t] A A cos ω t + 3 j j j j A A j cos j t j (44) Fgure 34 shows calculaton examples of the ntermodulaton nose of an OFDM sgnal. The horzontal axs denotes UHF channels and the vertcal axs sgnal levels n dbμv/.5 MHz. The green coloured bar and orange coloured curve n the graphcs expresses the OFDM sgnal and the spectrum of ntermodulaton nose, respectvely. The fgure at the top-left corner of each graph expresses the ntermodulaton nose that falls nto the OFDM sgnal band. The amplfer used n the calculaton has a saturaton level of x sat = 19 dbμv ( dbm). Fgures 34a, 34b and 34c are for the sgnal levels of 6 dbμv, 8 dbμv and 1 dbμv, respectvely. Note that the graphcs are dsplayed by approxmately 1 db lower than the sgnal levels ndcated n the captons, as they are expressed n spectrum per.5 MHz whle the captons are expressed for the sgnal bandwdth of 5.6 MHz. It s seen from the fgure that the ntermodulaton nose ncreases at a rate of 3 db/decade.

42 4 Rep. ITU-R BT.9-1 FIGURE 34 Sgnal level vs. ntermodulaton nose a) Sgnal level = 6 dbμv b) Sgnal level = 8 dbμv c) Sgnal level = 1 dbμv A. Examples of ntermodulaton nose Fgure 35 shows examples of ntermodulaton nose for varous undesred channels. Fgures 35a to 35c show the case of adjacent channel, n whch channel # s the desred sgnal at 6 dbμv and #1 s the adjacent channel undesred sgnal. In ths case, the sde lobe of ntermodulaton components around the undesred channel overlaps wth the desred channel, resultng n a large nose on the desred sgnal. Fgures 35d to 35f show the case of nd adjacent channel (#). In ths case, the sde lobe around the undesred channel does not overlap wth the desred channel, resultng n a relatvely small nose on the desred channel. Fgures 35g to 35 show the case of 1 undesred channels (#-#31), where each undesred sgnal s reduced by 1 db n order to mantan the total power of undesred sgnals beng the same as that n the other graphs,.e. 6 dbμv, 8 dbμv and 1 dbμv. In ths case, the ntermodulaton components spread wdely over the channels, resultng n a relatvely large nose on the desred sgnal.

43 Rep. ITU-R BT FIGURE 35 Dependency of ntermodulaton nose by channels, etc. a) Adj. channel = 6 dbμv b) Adj. channel = 8 dbμv c) Adj. channel = 1 dbμv d) nd Adj. channel = 6 dbμv e) nd Adj. channel = 8 dbμv f) nd Adj. channel = 1 dbμv g) 1 channels = 6 dbμv h) 1 channels = 8 dbμv ) 1 channels = 1 dbμv A.3 Determnaton of coeffcent N m Fgure 36 shows the relatonshp between ntermodulaton nose and sgnal levels for varous condtons. The dot lne (denoted by Intermodulaton ) n the graphcs s for the case wthout undesred sgnals, correspondng to Fg. 34. The ntermodulaton nose ncreases n proporton to the 3 rd power of sgnal level (ncrease rate of 3 db/decade).

44 4 Rep. ITU-R BT.9-1 The orange coloured curve ( 1 st Adj. channel ) s for the case of adjacent channel, correspondng to Fgs 35a to 35c. The nose taes a constant value at low undesred sgnal levels, whch corresponds to the ntermodulaton nose generated solely by the desred sgnal, whle t ncreases n proporton to the 3 rd power of undesred sgnal level (3 db/decade) n the range hgher than the desred sgnal (6 dbμv). The green colour curve ( nd Adj. channel ) s for the case of nd adjacent channel, correspondng to Fgs 35d to 35f. The nose ncreases n proporton to the nd power of undesred sgnal ( db/decade), whch dffers from the other cases of the 3 rd power (3 db/decade). The dar purple coloured curve ( 1 channels ) s for the case of 1 undesred channels, correspondng to Fgs 35g to 35. The nose ncreases n proporton to the 3 rd power of undesred sgnal (3 db/decade). The ntermodulaton nose thus vares by the number of out-channel undesred sgnals and/or frequency separaton even f the out-channel sgnal power s the same. We wll choose the adjacent channel case to be appled n the nterference calculaton, because the condton generates the most severe ntermodulaton nose. In ths case, the ntermodulaton nose s calculated to be 17 db lower than the amplfer saturaton level when the nput sgnal s at the saturaton level ( x sat ). Then the value of coeffcent N m appled n equatons (36), etc. s obtaned as follows. ( 5x ) or N (db) = x (db) 17( db) N m = 1 sat m sat (45) FIGURE 36 Sgnal level vs. ntermodulaton nose