Reduced Front-End Reception Requirements for Satellite Broadcast using Interference Processing

Transcription

1 Reduced Front-End Recepton Requrements for Satellte Broadcast usng Interference Processng Enrco Casn, Gennaro Gallnaro, and Joel Grotz, Member, IEEE Abstract The problem of fxed satellte broadcast recepton s consdered. The possblty of reducng the requrements on the antenna front-end dmensons s nvestgated. Interference processng and mtgaton technques are employed to cope wth the ncreased level of adjacent system nterference at the satellte broadcast recever resultng from the less drectve antenna. A novel satellte recepton front-end antenna based on a multple nput recever s proposed to adapt the nterference processng methods to the broadcast recepton scenaro. The potental performance of the devsed scheme s thoroughly dscussed and assessed by extensve software smulatons. Index Terms Satellte Broadcast Recepton, Interference Mtgaton, MMSE, Lnear Pre-Processng, Dual Input Satellte Antenna. I. INTRODUCTION The scenaro of fxed satellte drect-to-home (DT broadcastng s n general a well consoldated busness all around the world. Customers are equpped wth parabolc or equvalent antenna (C/Ku-band and a sngle RF front-end connectng the outdoor Unt (ODU wth the dgtal recever (IDU. The standard set-up consders a 6-9 cm antenna, nstalled on the roof/wall of the customer premses and connected to the ndoor unt wth coaxal cablng. Such confguraton has been wdely accepted, at least n the past, because of the added value provded by the dgtal TV (SDTV and by the wde range of broadcastng channel selecton. A new broadcastng standard has been recently ntroduced n the satellte world communty named DVB-S []. Thanks to ts powerful FEC codng, t offers up to 3% bandwdth savng when compared to ts former and frst broadcast satellte standard for the same end-user equpment nstallaton. Alternatvely, ts hgher codng gan can be exploted to stream gh defnton dgtal TV (DTV, on the same bandwdth as the standard dgtal TV (SDTV. owever, dgtal televson and content-on-demand have been recently offered by terrestral operators through the hgh capacty of Dgtal Subscrber Lne (DSL at a prce Ths work was funded by the European Space Agency (ESA under the ESA ARTES contract Number 87/4/NL/US. J. Grotz s wth the SES ASTRA, Chateau de Betzdorf, L-685 Betzdorf, Luxembourg, (e-mal: joel.grotz@ses-astra.com. G. Gallnaro s wth Space Engneerng S.p.A., Italy (e-mal: gennaro.gallnaro@space.t and E. Casn prevously wth the European Space Agency s now wth NATO C3 Agency, Transmsson and Network Servces, Oude Waalsdorperweg 6, 597 AK Den aag, The Netherlands, (e-mal: enrco.casn@nc3a.nato.nt. compettve wth the satellte offer. To compete successfully n urban areas wth such scenaro and wth other means of broadcast recepton, the ntroducton of small and compact recepton antennas s a key commercal pont. Indeed, the optmzaton of the antenna sze has n the past been a focus for the system desgn already wth the ntroducton of C-band, [5], and later Ku-band servces, [6]. The possblty of offerng dgtal televson broadcastng servce wth a reduced front-end requrement s nvestgated n ths work, manly focusng on an outdoor unt wth a very small (about 3-4 cm parabolc dsh reflector or a small flat panel array antenna of smlar sze. owever, wth smaller antennas, many ssues arse. Frst of all the nterference from adjacent satelltes (ASI wll start becomng the domnant source of lnk degradaton actually preventng the transton to smaller antennas. Then the gan of the antenna may not be adequate to operate the lnk wth a suffcent lnk margn. The ntroducton of DVB-S standard and ts powerful FEC encoder allow very low Sgnal-to-Nose Rato (SNR and helps reducng the requrements on the fgure of mert of the front-end. The nterference from other satellte system s however stll present and only a proper dgtal processng of t can reduce ts degradaton on the lnk qualty. The present work proposes the use of a multple-nput front-end n combnaton wth subsequent nterference processng. The proposed scheme extends the work presented n [7] on mult-nput DT recepton. A dual-nput front-end s consdered n conjuncton wth coherent combnaton of the nput paths wth a lnear pre-processng flter. Subsequent teratve nterference processng s also consdered to further reduce the mpact of adjacent satellte nterference. Well establshed nterference processng methods for mult-user systems, [8]-[], are extended to the feld of broadcast recepton n a novel manner. The possblty of mplementng the proposed method n a robust and cost-effectve manner s demonstrated. The sequel of ths paper s organzed as follows: Secton II defnes the typcal reference recepton scenaro whereas secton III outlnes the proposed nterference processng technques. Smulaton results for the dfferent cases are presented n secton IV and Secton V concludes the paper wth a summary of the benefts of the proposed method.

2 II. RECEPTION SCENARIO A. Typcal Broadcast Recepton Scenaro Satellte broadcast n Ku-band for fxed servces uses a shared frequency band wth adjacent satelltes. The satellte spacng s regulated and typcally depends on the frequency band. In some cases, adjacent satellte nterferers (ASI can be present on both sdes of the wanted sgnal however we consder here the common scenaro of one domnant nterferer. The coexstence of dfferent physcal layer standard, namely DVB-S [] and DVB-S [], adds another element of dssmlarty among broadcast systems. Among the many and dfferent possbltes, two reference scenaros that reflects qute well the majorty of broadcast servce n a few years tme are defned n the sequel. A survey test campagn has been conducted over a range of typcal geostatonary broadcast orbtal slots servng Europe and North Amerca. The common nterference scenaros were classfed and smlar nterference patterns can be recognzed, the two followng scenaros are among the most common and serve as the bass for the present work: Scenaro : DVB-S reference sgnal wth a sngle cochannel Interferer located at degrees spacng and wth lower power level (3-5 db. ere, 3dB reduced power level of the nterferer s assumed. It s common practce to place hgh power broadcast transmssons n the vcnty of lower power downlnk sgnals (satellte to gateway, typcally VSAT return traffc or DSNG traffc. Scenaro : DVB-S reference sgnal wth a sngle cochannel Interferer located at 3 degrees spacng and same power level. Such angular spacng s a typcal common mnmal dstance used between adjacent satellte systems offerng the same servce, e.g. broadcast servce. TABLE I TYPICAL REFERENCE LINK BUDGET FOR 6CM ANTENNA Satellte EIRP Free Space Loss Recever Nose Temperature Ran Margn (typ. 99.9% aval. Antenna Gan Antenna Gan at deg. offset Antenna Gan at 3 deg. offset Resultng C/I, Scenaro Resultng C/I, Scenaro Resultng C/N (mn. Resultng C/N, 3 MSps (mn. 53 dbw 5 db K ~3 db 35 db 9 db 9 db 5 db 6 db 88 db/z 3 db The scenaros consder only the case of a sngle strong sgnal nterferng wth the desred sgnal as ths s consdered the most common pattern. Extenson to the case of several adjacent nterferng satellte sgnals s possble but s not consdered n ths work. The nterferers are assumed dgtal sgnals that comply wth ether DVB-S [] or DVB-S []. The nterferer s assumed always co-channel but a detaled analyss wll be also performed for the case of staggered channel, when there s a sgnfcant frequency offset (up to half symbol rate among the channels. TABLE II LINK BUDGET FOR DIFFERENT ANTENNA UNITS Satellte EIRP 53 dbw Free Space Loss 5 db Recever Nose Temperature K Ran Margn (typ. 99.9% aval. ~3 db Antenna 45cm 33cm W Antenna Gan at deg. Offset 3dB 9dB 9dB Antenna Gan at 3 deg. Offset 8dB 7dB 5dB Resultng C/I, Scenaro db 9dB db Resultng C/I, Scenaro 9dB 5dB db Resultng C/N (mn. 85dB/z 8dB/z 8dB/z Resultng C/N, 3 MSps (mn. db 7dB 7dB B. Reference Lnk Budget As a baselne, the lnk budget of a recommended parabolc 6cm antenna s used. The qualty recepton requred for broadcastng servce s expressed by the target average year avalablty of 99.9%, accordng ITU ran and atmospherc model statstcs, see [3]-[5]. The typcal reference lnk budget scenaro s outlned n Table I; see [6] for ITU reference antenna pattern masks. Table II reports the lnk budget for the same satellte confguraton but wth dfferent front-end equpments, actually dfferent antennae unts. Two parabolc reflector dshes of 45 and 33 cm are consdered as well as a wavegude horn (W based antenna wth 6x8 elements, 4cm x cm evenly taped. From ths lnk budget, t s evdent that the lnk s severely domnated by the nterference of adjacent systems, especally for the smaller antennae. In order to better understand the potental of any nterference processng technques, Fg. llustrate the computed sgnal-to-nose and nterference ratos (SNIR or C/(N+I for dfferent antenna szes for scenaro and worst case recepton condtons (ncludng 3dB of ran attenuaton. The potental mprovement that can be exploted by any processng technque can be read on ths fgure as the dfference between C/N and C/(N+I. C/X Antenna Sze [m] C/(N+I C/N C/I Delta C/(N+I - C/N Fg.. Sgnal-to-nose and sgnal-to-nterference ratos for a typcal broadcast sgnal receved wth a gven parabolc DT recepton dsh sze, assumng an nterferer at 3 degrees offset and smlar power level.

3 III. INTERFERENCE PROCESSING TECNIQUES AND RECEIVER DESIGN The proposed recepton method conssts of a two steps approach, a frst multple nput lnear-preprocessng (LPP or spatal MMSE step followed by a subsequent teratve nterference cancellaton (IIC step, see Fg.. The frst stage conssts of a lnear pre-processng block [7] that coherently combnes the receved sgnal strans usng maxmal rato combnng (MRC for both the wanted sgnal and for the man nterferng sgnal. The second stage s the subsequent use of an teratve soft-decson based nterference processng unt, based on the Sngle User Matched Flter (SUMF based Soft-IC methods proposed n [9]. It s worth to remark that both stages can be used n a standalone confguraton as they do not depend on each other. Benefts provded by each of them or by ther combnaton drve the selecton of the best nterference processng technque. A dual-port LNB and RF front end s assumed n the followng. The IDU has two tuners so that the sgnal comng from the ports antenna, even though not completely orthogonal, can be processed by nputs block. s p IIC Step LPP Step s s p y y LNB LNB predefned spacng Fg.. Baselne recever block dagram: A lnear pre-processng block (LPP followed by an teratve nterference cancellaton unt (IIC. Fg. 3 llustrates the proposed dual-nput setup wth the example of a parabolc reflector antenna wth two LNBs. The wanted sgnal s assumed on the axs of the optmal pontng drecton and the nterferer spaced at a certan offset angle on an adjacent satellte. The wanted sgnal s thus receved wth maxmal combned gan of the two LNB nputs. LNB LNB Parabolc Reflector Antenna Optmal Pontng Drecton Regon of expected adjacent satellte nterferers (ACI Fg. 3. Illustraton of the azmuth cut of the proposed dual-nput DT recepton antenna setup proposed. Assumng the presence of N adjacent satellte systems, each one transmttng a sgnal s j the receved sgnal at the two antenna ports y = [y, y ] T can be wrtten n matrx form as: y = s+n ( where s s the column vector [s, s,, s N ] T, s a two by N matrx whose elements h j represent the complex gan of antenna port toward satellte j. Fnally n s the column vector whose element n and n are the thermal nose affectng y and y respectvely. The nose covarance matrx s not dagonal as the nose on the two nput ports are correlated as the antenna beams are partally overlapped.. A. Dual Input Antenna and MMSE Pre-Processng A weghted-sum based lnear pre-processng (LPP that coherently combnes the two antenna outputs s consdered as frst canddate algorthm. Ths method has been already proposed n [7] and ts applcablty to the selected scenaro s brefly recalled hereafter for the consdered scenaros where only satellte systems are consdered. The lnear pre-processor block actually mplement a matrx flter M on the receved sgnal y to provde the vector sgnal p = [p, p ] wth a maxmzed Sgnal-to-Nose-plus-Interference Rato C/(N+I for respectvely the desred and the domnant nterferng sgnals. An MMSE flter s consdered as lnear pre-processor whch can be decomposed n spatal flters, one for the wanted sgnal and one for the domnant nterferer. Denotng each row of matrx M as m wth = or, can be computed as the vector whch mnmzes the expectaton of -m y, where s s the desred sgnal vector,.e.: s { s } m m = arg mn E m y ( m By means of smple elaboratons and P = E s s, each spatal flter can be calculated as defnng { } * ( Σ + P m = P h (3 Takng nto account that Σ + P s equal to the covarance matrx of the receved sgnal, we may also wrte: - m = P R h (4 The LPP flter calculaton s performed synchronously for both the man wanted sgnal and the nterference, whch results n an mproved recepton of both the wanted and the nterferng sgnal, ndexed = and respectvely. Note that a best possble nterferer recepton helps at the subsequent IIC step, even though the nterferer tself s not of prmary nterest. Note also that ths computaton takes nto account the correlaton between the two separate recepton paths and the combnaton of the two sgnals. The sgnal to nose and nterference rato can then be expressed n combnaton as follows: C = m R m (5 N + I R s the cross-correlaton matrx between the two separate recepton paths, takng nto account the nterferers, as outlned n [7]. It s nterestng to note that the dual nput antenna has the nherent advantage of beng robust aganst a small pontng error, as demonstrated n [7]. Ths advantage s a key feature for the practcalty of ths method.

4 Fg. 4. Processed wndow Example n the case of two carrer processor. B. C/N and C/I performance wth dual-nput antennas We denote wth C/N the sgnal-to-thermal nose rato of the system and wth C/I the sgnal-to-nterference rato. In the followng table the C/N and C/I ratos are optmzed for the wanted sgnal, receved as p and the C/N and C/I ratos are optmzed recepton levels for the man nterferer, receved as p. Note that the recepton of p s performed specfcally for a subsequent nterference processng. To compute C/N and C/I n the Tables III and IV, the expresson (5 was used wth a reference antenna scenaro wth 3 degrees spacng offset between the recepton elements. The cross-correlaton n matrx R was computed takng nto account the antenna gan pattern. TABLE III SCENARIO, DEGREES POINTING SPACING BETWEEN DUAL INPUT ANTENNA ELEMENTS, INTERFERERS 3DB BELOW SIGNAL POWER ANT. SIZE C/N C/I C/(N+I C/N C/I C/(N+I 4 cm cm cm cm cm TABLE IV SCENARIO, 3 DEGREES POINTING SPACING BETWEEN DUAL INPUT ANTENNA ELEMENTS, INTERFERERS AT SAME SIGNAL POWER ANT. SIZE C/N C/I C/(N+I C/N C/I C/(N+I 4 cm cm cm cm cm After the MMSE pre-processng step, t appears that n scenaro performances are essentally close to nterference free. No sgnfcant mprovement can be expected by further subsequent nterference cancellaton. In scenaro, vce versa, resdual nterference after the MMSE Pre-Processor produces a more sgnfcant degradaton of the overall C/(N+I. In ths context teratve nterference mtgaton can provde some further mprovement whch can be translated n a further reducton of the DT antenna dameter. C. Iteratve Interference Cancellaton We consder scenaro n the sequel and we recall that n scenaro, nterference from the adjacent satellte s assumed to be also a broadcastng sgnal. We may thus assume that both wanted and nterferng sgnals are of the DVB-S type. We assess the possble mprovements whch may derve from teratve nterference cancellaton n such a scenaro. As mentoned the proposed approach s based on the SUMF Soft-IC methods proposed n [9] whch, although suboptmum wth respect to a Full Belef Propagaton approach s much more practcal from a complexty vewpont. In fact the Full Belef Propagaton algorthm has a complexty that ncreases exponentally wth the modulaton order and the channel memory. (Note: Channel memory, s usually gnored n theoretcal analyss because fully synchronous sgnals are consdered. Coupled wth Nyqust flterng, ths mples no ISI and the need of only consder nterference from homologous symbols from nterferng carrers. Regardng the modulaton orders consdered for the broadcast applcatons, QPSK and 8PSK are expected to be typcally used, especally when targetng small antennas (6- APSK and 3-APSK are also supported by DVB-S []. It shall be further observed that prelmnary smulatons have also shown that performance of the selected SUMF-IC method, n our context, was practcally the same as that of the Uncondtoned MMSE-IC method also ntroduced n [9] provded that an MMSE pre-processor stage s also used. Gven ts lower complexty SUMF-IC was fnally selected for ths nvestgaton. A problem n the applcaton of the well known SUMF Soft-IC method to the present case s the fact that wanted and nterferng sgnals are not synchronous. Asynchroncty s not only at the symbol clock and carrer frequency levels but also at the frame level. Iteratve Interference cancellaton algorthms explotng jont decodng and detecton requre that the FEC frames of the processed sgnals be synchronous. Ths s not the case n our applcaton. In fact, DVB-S sgnals transmtted through dfferent satelltes are totally uncoordnated. Further, wth DVB-S the FEC frame length It shall be noted that n cted lterature the reference scenaro s CDMAbased. The Sngle User Matched Flter (SUMF recever s thus the one whch correlate the receved samples wth the conjugate of the spreadng sequence. In our multbeam TDMA context, the equvalent of the spreadng sequences are the the columns of the channel matrx. In practce, we do not perform such correlaton but only consder the contrbuton of the domnant channel matrx term (.e. the domnant beam.

5 may be dfferent for dfferent carrers []. Ths problem can be solved extendng the processng wndow to multple frames. The prncple s depcted n Fg. 4 where 3 frames per carrer are processed by each decoder before fed-back the extrnsc nfo (or APP probabltes to the MUD front-end whch wll then compute the new a pror probabltes for the next decoder teraton. It s clear that Frame k 3 may not be successful decoded because nterference on the symbols at ts end cannot be cancelled because the correspondng data from the other carrer s not yet processed. Also Frame n may not be successfully decoded f Frame k was not prevously successfully decoded. owever, Frame n and Frame k would probably be correctly decoded f the nput SNIR s n the order of that manageable wth synchronous frames. Prevously decoded frames (ether the soft data estmaton derved from the extrnsc or APP probabltes of last teraton or the hard decson data dependng f the decodng process was successful or not are also stored. In partcular ths helps the processng of the wndow of Fg. 4 because the stored data of Frame k are used for cancelng nterference from the part of Frame n. It s clear that, n order to mnmze complexty ncrease, the number of frames n the processed wndow shall be the mnmum, however further elaboratons on such ssue are out of the scope of the present work. Before presentng smulaton results t shall be underlned that n prncple Iteratve Interference Cancellaton may be adopted also n absence of the MMSE Pre-Processor. In ths regard, the Iteratve Interference Cancellaton (IIC can be used even wth a conventonal sngle-nput antenna where the MMSE spatal processng cannot be used. owever results wth sngle-nput antenna are out of the scope of the present work. It s also worth to menton that at each teraton the channel estmaton and synchronzaton parameters can be refned. In practce, the software smulator performs estmaton of the carrer phase as well as of the SNIR at each teraton, but not the symbol tmng as ths would ncrease the W complexty sgnfcantly wthout correspondng benefts. Actually also carrer phase estmaton may not be requred n most of the scenaros f some small penalzaton (few tenth of a db s accepted. IV. SIMULATION RESULTS In order to obtan more realstc results from software smulatons, a full emulaton of a real DVB-S demodulator was mplemented. In partcular, the demodulaton strategy and modem algorthms documented n [] were mplemented n our smulator. Accordng to the strategy n [], symbol tmng s frst recovered over the ncomng DVB-S sgnal, followed by frame acquston and then Carrer Frequency recovery and Carrer Phase recovery. Plot-aded frequency and phase recovery was assumed. Sgnal AGC and SNIR estmaton s also performed by explotng the plot symbols. After demodulaton the soft-demodulated data are fed to the LDPC decoder whch performs a sngle teraton to compute the extrnsc probablty of channel bts. Ths nformaton s used to compute the extrnsc probablty of channel symbols and hence of a soft estmaton of the nput symbols (by averagng each possble channel symbol accordng to ts extrnsc probablty as detaled n [9]. Only Scenaro has been consdered for the smulaton campagn beng the one that can beneft the most from Interference cancellaton technques. When the dual-ports antenna s consdered, the results were obtaned wth and wthout the use of the MMSE Lnear Pre-Processor, but always wth the SUMF Soft-Interference Cancellaton. The benefts provded by the MMSE Pre-Processor depend on the channel matrx, the Sgnal-to-Nose Rato at the nput of the pre-processor and the thermal nose correlaton. Several channel matrces were consdered for smulatons, all havng a Carrer to Interference on each port, before processng, equal to 3 db. ha (6 = hb The consdered scenaros are summarzed n the followng table: TABLE V CANNEL MATRIX PARAMETERS Case A Case B Case C h A j5 e h B - j5 e j e j e The MMSE Pre-Processor performances are expressed n terms of rato between the Symbol Energy and the total nose and nterference power densty at the output of the preprocessor. For the consdered channel matrces the performance s shown n Fg. 5 and Fg. 6. It appears that when no thermal nose correlaton s assumed, the case A matrx produces much better performance than case B matrx. Ths s due to the fact that case A matrx as opposed to case B has rows whch are closer to orthogonalty. As a matter of fact, however, case B matrx s actually closer to what can be expected by a two feed antenna whose drectvty s not enough to reject nterference (.e. the nterference s also n the man lobe of the antenna gan patterns. The level of nose correlaton n practce depends on how the front-end antenna wth the dual-nput LNB s desgned. It shall be observed that nose correlaton may also have an mpact on the resultng MMSE optmum flter and, as a consequence, on achevable performances. Ths s clearly shown n Fg. 6 where nose cancellaton (n addton to nterference cancellaton can also be acheved (see the upper curve n the same fgure. Smulatons whch follow assume uncorrelated nose on the two antenna port. Dfferent physcal layer waveforms were consdered to span most of the possbltes offered by the

6 DVB-S standard and results are presented per every modulaton and codng scheme. A remark s n order before gong through all smulatons results. The BER and PER performance are reported as functon of the nput E S /N rato, so only thermal nose. owever the nterferer s always present and n scenaro, a C/I of 3 db s assumed. A classcal recever not employng any nterference cancellaton algorthm would consder the nterferng sgnal as a nose source. If the nterferer was Gaussan dstrbuted and wth a whte spectrum, the equvalent total Sgnal-to-Nose and Interference Raton, SNIR eqwg would be: ( ( SNIReqWG = Es N + C I The gan of the dfferent Interference cancellaton schemes for dfferent modulaton and codng scheme wll be presented n terms of SNIR eqwg.. Post MMSE Es/(No+Io (db = Input Es/No (db = cos(5 jsn(5 No Nose Correlaton cos(5 + jsn(5 Fg. 5. MMSE performance for two channel matrces wth C/I=3 db versus the nput E S /N (uncorrelated nose. Post MMSE Es/(No+Io (db Nose Correlaton Input Es/No (db = cos(5 jsn(5 cos(5 + jsn(5 (7 to whch further. db for losses due to recever mplementaton shall be added. The channel matrx of case B s consdered n ths case. It can be observed that wthout the use of nterference mtgaton technques, the SNIR eqwg s always lower than or at the most equal to 3 db, hence preventng the correct decodng of QPSK ¾. On the other hand when MMSE pre-processor and SUMF-IC algorthms are consdered, smulaton results show BER performance around -5 for E S /N of 8.5 db, hence SNIR eqwg around db, so at least. db gan s acheved when compared to pure AWGN case. Fg. 9 shows the smulaton results wthout the MMSE preprocessor and the same channel matrx. It s evdent that MMSE flter provdes a gan of almost.5 db It s to be mentoned that MMSE pre-processor actually estmates the covarance matrx from the receved symbols and then calculates and apples the estmated flter. In ths case the covarance matrx estmator operates on relatvely short number of nput samples (496 samples or 4 symbols as the smulator works wth 4 sample/symbols 3. As matter of fact, Fg. 7 shows how the performance could be mproved by usng a longer sequence for the covarance estmator. Error Probablty.E+.E-.E-.E-3.E-4.E-5 BER, Es/No = 8.4 db FER, Es/No = 8.4 db BER, Es/No = 8.3 db FER, Es/No = 8.3 db 8 Iteratons QPSK r=3/4, C/I= 3 db.e Covarance Estmator Samples Fg. 7. Performance of MMSE Pre-Processor + SUMF-IC versus the total nput sgnal covarance estmator length. Uncorrelated nose. Frequency offset of - Rs. Four samples per symbol used. All results above wth SUMF-IC have been obtaned usng 8 nterference cancellaton teratons. It shall be observed that, at each nterference cancellaton teraton, a sngle teraton of the LDPC decoder s also done 4. Fg. 6. MMSE performance for dfferent nose correlaton versus the nput E S /N. A. QPSK 3/4 Results Fg. 8 and Fg. 9 report the results for QPSK modulaton scheme wth LDPC code rate ¾ wth and wthout MMSE preprocessor channel matrx of case B. It s to be mentoned that on a pure Addtve Whte Gaussan Nose channel, such scheme would acheve a BER of -5 at a SNR value of 4 db 3 Ths does not mples that four samples / symbols are actually requred for covarance computaton. Gven the asynchroncty of nput carrers, the number of samples/symbol shall anyway satsfy the samplng theorem. 4 Operaton schedulng was sequental,.e. a decodng operaton for one of the carrers s frst done wth subsequent soft- reconstructon of the correspondng sgnal for subtracton from the other carrer. Then a decodng teraton for such last carrer s done and the cycle s repeated. Wth respect to a parallel schedulng of operatons such sequental schedulng brng a small advantage as t allows a slght reducton n the number of teratons for the same performance target.

7 Error Probablty.E+.E-.E-.E-3.E-4.E-5.E-6.E Es/No (db BER FER QPSK 3/4 8 Iteratons Fg. 8. Performance of MMSE Pre-Processor + SUMF-IC. Frequency offset between carrers - Rs. Covarance Estmator length 496 samples (four samples per symbol used n smulatons. Error Probablty.E+.E-.E-.E-3 QPSK 3/4 8 Iter. SUMF-IC.579 j.46 = j.646.e Es/No (db BER FER Fg. 9. Performance n the same condtons (.e. nput C/I as n Fg. 8 but wthout MMSE pre-processor. B. 8-PSK /3 Results 8-PSK modulaton scheme on an AWGN channel acheves a BER of -5 for SNR of 6.5 db to whch about.3 db of recever mplementaton losses shall be added. Due to the lower thermal nose level of the operaton condtons, t s expected that the effectveness of MMSE flter s hgher than n prevous QPSK cases. As a matter of fact MMSE pre-processor alone and MMSE plus SUMF-IC have been smulated and results shown n Fg. and respectvely. The channel matrx of case C was used n ths case and C/I=6 db was consdered, ths because t s expected that f hgher order modulaton schemes wants to be employed, nterference level shall be lower than QPSK. owever the value of 6 db s stll an aggressve value as t s comparable to the SNR threshold of the selected modulaton and codng scheme for pure AWGN channel. Error Probablty.E+.E-.E-.E-3 8PSK /3 C/I=6 db MMSE - NO IIC.E-4 cos( + j sn( cos( j sn(.e Es/No (db BER FER Fg.. Performance of 8-PSK rate /3 wth MMSE preprocessor only (LDPC decoder teratons equal to 8. Error Probablty.E+.E-.E-.E-3.E-4.E-5 BER FER 8PSK /3 C/I=6 db MMSE - SUMF-IC cos(.e Es/No (db j sn( cos( + j sn( Fg.. Performance of 8-PSK rate /3 wth MMSE preprocessor and SUMF-IC (8 Iteratons. It appears that the use of SUMF-IC only mproves the performance of about.4 db at least for relatvely hgh BER value 5 (.e. -4 to -5. The reason of such behavor can be understood thanks to Fg. whch shows the Sgnal-to-Nose-and-Interference Rato and the C/I at the output of the MMSE pre-processor. db cos( j sn( cos( + j sn( Input Es/No (db Output Es/(No+Io Output C/I Fg.. MMSE Pre-Processor Performance as E S /N rato at the preprocessor output, assumng a channel matrx case C and for C/I=6 db 5 As qualty of nterference cancellaton ncreases rapdly when Es/No exceed the threshold, we may expect that the performance advantage wth IIC cannot decrease at lower BER/FER target. In fact, although not shown n Fg., we were not able to fnd errors at E S /N >9. db for smulaton exceedng 3, frames.

8 It s evdent that the C/I s already very hgh hence no further nterference cancellaton s requred. From the same Fgure, t appears that the convergence of the nterference cancellaton algorthm happens when the MMSE preprocessor output SNIR s close to 6 db whch s close to the threshold of 8-PSK /3 LDPC code. C. Results wth asynchronous frames The above results were obtaned assumng that the physcal layer frames of the two consdered carrers are quassynchronous,.e. that the frame offset between carrers s lower that 9 symbols (.e. less than the frame header. Ths allows processng a sngle frame of each carrer at the tme. When frame offset s larger than the frame header we need to use the strategy depcted before n Fg. 4. It was found that f the startng SNR s rather hgh such as correct decodng s possble and then decreasng the sgnal level, t s possble to operate at SNR level even lower than those shown n prevous fgures. owever, as soon some errors are experenced, a burst of wrong frames s produced wthout any possblty of recoverng them f the SNR rato s not ncreased agan above the decodng level threshold. For QPSK ¾ the set of results are shown n Table VI for a processng wndow sze of 3 frames, whle Table VII shows the results for a wndow sze of frames. No sgnfcant degradaton s reported for the shorter processng wndow that of course reduces the complexty requrements. Fnally t shall be noted that from smulaton results n order that the Iteratve-Interference Cancellaton (IIC algorthm s convergng t s requred that the C/(N+I of at least one of the channels s greater than approxmately the decodng threshold of the consdered MODCOD. In partcular the QPSK ¾ mnmum ntal C/(N+I shall be larger than about. db ((versus a threshold of the consdered DVB-S LDPC code of about 3.9 db. Such results should not have been fully unexpected as complant wth theoretcal research reported n the lterature. In partcular, t has been shown, [4], that MMSE-SIC s able to reach the channel capacty of the Gaussan mult-user channel f capacty achevng codes are used. Ths result mples that you should not be able to correctly decode f none of the carrers has a C/(N+I greater than the sngle user capacty threshold. TABLE VI QPSK ¾ PERFORMANCE WIT ALF-FRAME ASYNCRONY BETWEEN CARRIERS. MMSE PRE-PROCESSOR WIT SUMF-IC (8 ITERATIONS PROCESSING WINDOW SIZE EQUAL TO 3 FRAMES ES/NO [DB] INITIAL TRANSIENT # FRAMES IN ERROR AFTER TRANSIENT frames to recover 3 errors n channel # out of 38 frames per carrer frames to recover No errors n successve 53 frames per carrer. 8.4 frame to recover No errors n successve 3457 frames per carrer. TABLE VII SAME PARAMETERS AS TABLE VI WIT A WINDOW SIZE OF FRAMES E S /N O [DB] INITIAL TRANSIENT # FRAMES IN ERROR AFTER TRANSIENT frames to recover No error n successve 8 frames frames to recover No error n successve 5, smulated frames 8.4 frames to recover No error on successve, smulated frames 8.5 frame to recover No error on successve, smulated frames V. CONCLUSION A novel applcaton for known nterference processng methods was presented. The feld of satellte broadcast recepton was revewed n the perspectve of a mnmzaton on the antenna front-end requrements wth respect to nterferer rejecton. Ths leads to potentally smaller antennas that eventually allow robust nstallaton yet a proper recepton at consumer (end-user homes. Key to the proposed method s the combnaton of a dual-nput recepton antenna wth subsequent nterference processng. The use of MMSE preprocessor n combnaton wth Iteratve Interference Cancellaton algorthm s pvotal to allow an error-free recepton of the desred sgnal that otherwse would not be possble because of the nterference power level. The typcal nterference scenaros ncludng asynchrony frames were outlned and the sgnal-to-nose ratos of nterest were smulated for the proposed nterference processng methods. A realstc recever mplementaton ncludng calbraton method and correlaton matrx estmaton have been also consdered. A consderable mprovement was recorded for a sgnfcant range of modulaton and codng schemes together wth practcal range of sgnal-to-nose and nterference rato values. Results prove the applcablty of the proposed scheme for the DT broadcast recepton. REFERENCES [] ETSI EN 3 37, v.., Dgtal Vdeo Broadcastng (DVB; Second Generaton framng structure, channel codng and modulaton systems for Broadcastng, Interactve Servces, News Gatherng and broadband satellte applcatons [] ETSI EN 3 4, v.., Dgtal Vdeo Broadcastng (DVB; Framng structure, channel codng and modulaton for / Gz satellte servces [3] ITU-R Recommendaton P.676-4, Attenuaton by Atmospherc Gases [4] ITU-R Recommendaton P.68-6, Propagaton Data and Predcton Methods Requred for the Desgn of Earth-Space Telecommuncaton Systems [5] ITU-R Recommendaton P.837-, Characterstcs of Precptaton for Propagaton Modelng [6] ITU-R Recommendaton S.67-4, Satellte Antenna Radaton Pattern for Use as a Desgn Objectve n the Fxed Satellte Servce Employng Geostatonary Satelltes [7] J. Grotz, B. Ottersten, J. Krause, Applcablty of Interference Processng to DT Recepton, 9 th Internatonal Sgnal Processng for Space Communcaton Conference, ESA/ESTEC, Noordwjk, The Netherlands, September 6.

9 [8] B. F. Bedas,. El Gamal, S. Kay, Iteratve Interference Cancellaton for gh Spectral Effcency Satellte Communcaton, IEEE Trans. On Comm., January [9] J. Boutros, G. Care, Iteratve Multuser Jont Decodng: Unfed Framework and Asymptotc Analyss, IEEE Trans. On Informaton theory, July. [] M. Moher, An Iteratve Multuser Decoder for Near--Capacty Communcatons, IEEE Trans. On Comm., July 998. [] G. Care, R. Müller, T. Tanaka, Iteratve Multuser Jont Decodng: Optmal Power Allocaton and Low complexty mplementaton, IEEE Trans. On Informaton Theory, Sept. 4 [] E. Casn, R. De Gaudenz, A. Gnes, DVB-S modem algorthms desgn and performance over typcal satellte channels, Int. J. Satell. Commun. Network., Vol. N. 3, May-June 4. [3] G. Gallnaro, R. Rnaldo, A. Vernucc, Increasng Throughput of Wdeband Satellte Systems Reverse-Lnk through Adjacent Channel Interference Mtgaton, AIAA, 7 [4] R. Muller, Combnng Multuser Detecton wth Codng: Promses and Problems, Conf. on Informaton Scences and Systems, Prnceton Unversty, [5] E. Son, U. Mathur, Optmum Satellte ome Recevng Antenna Sze at C-band, IEEE Trans. On Broadcastng, Vol. BC-33, No., March 987 [6] K. Johannsen, U. Mathur, TVRO Antenna Sze, Ku-band Versus C- band, IEEE Trans. On Broadcastng, Vol. BC-33, No. 3, Sept. 987 Joel Grotz (M receved the Dpl.-Ing. degree n Electrcal Engneerng from the Unversty of Karlsruhe, Germany and the Dplôme d'etudes Approfondes (DEA from the Grenoble Insttute of Technology (INPG, France n 999. e worked several years as Systems Engneer for the Socété Européenne des Satelltes (SES. e s currently workng n a jont research project nvolvng SES, the Luxembourg Insttute for Advanced Studes n Informaton Technologes (LIASIT and the Royal Insttute of Technology (KT, Stockholm, Sweden. s research nterests are manly wthn the felds of sgnal processng wth applcatons n wreless communcaton systems and n satellte communcaton systems. Gennaro Gallnaro got a degree (summa cum Laude n Electronc Engneerng from Unversty of Rome n 979. e has snce worked n the satellte communcaton feld beng manly nvolved n studes, smulaton and desgns of analog and dgtal transmsson systems, payload hardware assessment, new modulaton access technques and analog/dgtal sgnal processng technologes. e was co-recpent of the IEEE Vehcular Technology Socety Jack Neubauer Memoral Award whch recognzes the best systems paper publshed n the IEEE Transactons on Vehcular Technology. Enrco Casn was born n Fglne Valdarno, Italy, on March 3th 976. e receved the Laura Degree (summa cum Laude from the Unversty of Florence, Florence, Italy n. In he joned the ESA's Research and Technology Center (ESTEC, Noordwjk, The Netherlands. Durng -3 he was a tranee nvestgatng advanced synchronzaton technques for space communcatons and nterference mtgaton technques. Snce 3 s workng for the RF Payload and System Dvson as communcaton system engneer. s research nterests nclude the characterzaton of non-lnear satellte channel, advanced modem desgn, physcal layer system smulaton and the analyss of the satellte payloads.