> mm 13-48 IEEE BMSB 2013 < 1 A Pilot Aidd Avraging Channl Estimator for DVB-T2 Spiridon Zttas, Pavlos I. Lazaridis, Zaharias D. Zaharis, Stylianos Kasampalis, and John Cosmas, Snior Mmbr, IEEE Abstract In DVB-T2 broadcasting applications, rrors occur du to channl imprfctions, rducing th systm s throughput. Th us of a channl stimator and qualizr countracts th channl inducd distortions and thus improvs th systm s Bit Error Rat (BER). In our study, a simpl-to-implmnt channl stimator is proposd. It is basd on avraging th sub-channl frquncy rsponss of succssiv OFDM symbols in tim, considring a vry slow tim-varying channl. In th frquncy domain, splin intrpolation is applid. Th proposd stimator can b usd in practical rcivrs with limitd computation powr and low mmory capacity. Indx Trms Channl stimation, DVB-T2, OFDM, scattrd pilots D I. INTRODUCTION IGITAL Vido Broadcasting Scond Gnration Trrstrial (DVB-T2) [1] was publishd in 2009 as an improvmnt to DVB-T which was publishd in 1997, and now is usd in most countris for Digital Trrstrial Tlvision (DTTV or DTT). DVB-T2 adopts many high-nd tchnologis from DVB-T and Digital Vido Broadcasting Scond Gnration Satllit (DVB-S2) [2]. Th nw standard is vry flxibl and thus it uss nw additional tchnologis, such as Multipl Physical Layr Pips (MPLPs), Alamouti coding, Constllation Rotation (CR), Extndd Intrlaving and Futur Extnsion Frams (FEF). Th combinations of QAM ordr, Guard Intrval (GI), Forward Error Corrction (FEC), and high FFT sizs up to 32k hlps th systm to achiv high bit rats up to 45.5Mbps [3]. Ths high bit rats ar suitabl for transmitting High Dfinition TV (HDTV) contnt. Radio channl imprfctions, such as attnuation, phas shifting and tim dlays, rsult in rrors, rducing in this way th systm s throughput. In addition, du to multipath propagation, th channl suffrs from distortion both in tim and frquncy domain causing in this way Intr-Symbol Intrfrnc (ISI). Th us of a channl stimator and Manuscript rcivd April 30, 2013. Spiridon Zttas, Pavlos I. Lazaridis, Stylianos Kasampalis, and John Cosmas ar with th School of Enginring and Dsign, Brunl Univrsity, London UB8 3PH, U.K. (-mail: spiridon.zttas@brunl.ac.uk; pavloslazaridis@hotmail.com; skasamp@otnt.gr; john.cosmas@brunl.ac.uk). Zaharias D. Zaharis is with th Tlcommunications Cntr, Aristotl Univrsity of Thssaloniki, 54124 Thssaloniki, Grc (-mail: zaharis@auth.gr). qualizr countracts th channl s inducd distortions and thus improvs th Bit Error Rat (BER). Th topic of channl stimation and qualization has thoroughly bn studid in past yars bcaus of its importanc. Svral mthods hav bn proposd to compnsat th channl distortion. In th DVB-T2 cas, pilots ar usd for fram, frquncy and tim synchronization, channl stimation, phas nois tracking and idntification of th transmission mod [1]. Th prformanc of a convntional Last Squar (LS) stimator and a Last Man Squar (LMS) stimator is invstigatd in [4] for diffrnt pilot arrangmnts. In [5], an analysis of th ffcts of tim-basd pilot intrpolation ovr tim-varying channls is prsntd. Using diffrnt pilot pattrns, a DVB-T2 channl stimator and an qualizr wr modld and valuatd in trms of prformanc in [6] and [7]. Also, an in-dpth study of minimizing th numbr of th usd pilots in trms of rror probability can b found in [8]. A comparativ study of channl stimation mthods is providd in [9], whr a LS stimator sms to b computationally suprior compard to a Minimum Man Squar Error (MMSE) stimator. In [10], a blind channl stimation mthod is proposd whr no pilots ar usd and thus th throughput is maximizd. In th mthod proposd hr, a simpl-to-implmnt channl stimator is proposd. It is basd on avraging th channl frquncy rsponss of succssiv OFDM symbols assuming a slow tim-varying channl. In this study, as th channl varis slowly in tim, th stimator trats th channl as constant in th tim domain for th intrval of th buffrd OFDM symbols. In th frquncy domain, th channl is assumd as slctiv. For th simulation nds, Rayligh and Rican modls hav bn usd. Th proposd stimator can b usd in practic by rcivrs that hav limitd computation powr and rducd mmory (RAM) capacity. Th rst of th papr is organizd as follows: In Sction II th architctur of DVB-T2 is dscribd in brif. In Sction III th Scattrd Pilots (SP) usd in DVB-T2 ar discussd in dtail. In Sction IV th proposd mthod is xplaind and analyzd in dpth. Th simulation rsults for diffrnt channl nvironmnts as wll as for diffrnt configurations of SP pattrns, FFT sizs, QAM ordrs and mobil spds ar drivd and analyzd in Sction V. Finally, in Sction VI th ffctivnss of th avraging channl stimator is discussd and improvmnts of th stimator for fast tim-varying
> mm 13-48 IEEE BMSB 2013 < 2 channls ar proposd. II. ARCHITECTURE OF DVB-T2 SYSTEM Th flowchart of a simplifid DVB-T2 systm is dpictd in Fig.1. Th Mod and Stram Adaptation (MSA) block is rsponsibl to form th basband fram (BBFRAME). Th BBFRAME is insrtd into Bit Intrlavd Coding and Modulation (BICM) block. Th outr codr (BCH) and th innr codr (LDPC) offr coding rats of 1/2, 3/5, 2/3, 3/4, 4/5, 5/6 and construct th forward rror corrction fram (FECFRAME). Aftr a bitwis intrlaving, ach FECFRAME is d-multiplxd into paralll cll words and mappd into constllation valus of Quadratur Amplitud Modulation (QAM). Th QAM availabl ordrs ar QPSK, 16QAM, 64QAM, and 256QAM. For robustnss improvmnt, constllation rotation is optionally providd. Ths data clls ar furthr intrlavd in tim to nsur uncorrlatd intrfrnc and distortion along ach FECFRAME. Th data clls of th FECFRAME ar thn insrtd into th Fram Buildr and OFDM gnrator. Th Fram Buildr constructs th T2-Fram by assmbling th data clls into th P2, th Normal and th Fram closing symbols. PLP 0 Tabl I dpicts th typ of pilots in ach symbol. In this papr, without loss of gnrality, w focus only in normal OFDM symbols and thrfor only SP ar considrd. TABLE I CONFIGURATION OF PILOTS PER OFDM SYMBOL TYPE Pilot Typ Symbol Fram Scattrd Continual Edg P2 Closing P1 P2 Normal Fram Closing Th IFFT block is rsponsibl to transform th data and th pilot cll information into an quivalnt signal in tim domain. Th subcarrirs of ach frquncy cll ar orthogonal to ach othr to prvnt Intr Carrir Intrfrnc (ICI). Equation (1) dscribs mathmatically th OFDM systm. N 1 nk, k s n= k= 0 s(t) = c g (t nt ) (1) with CRC8 Basband hadr insrtion Zro Padding if ndd Basband scramblr BCH/LDPC Bit Intrlavr Dmultiplxr QAM Mappr Gray mapping Fram Buildr Frquncy Intrlavr SISO/MISO Pilot pattrn g k and () t = 0, j2 π f t k, t [0,T ] lswhr k fk = f0 +, k = 0,, N 1 T s s Constllation Rotation QAM Cll Intrlavr Tim Intrlavr IFFT PARP rduction P1 Symbol Insrtion TX whr c n,k is th symbol of th kth subcarrir, N is th numbr of total subcarrirs, f k is frquncy of th k-th subcarrir and f 0 is th lowst frquncy usd. Th availabl numbrs of subcarrirs ar: 1k, 2k, 4k, 8k, 16k and 32k. A Guard Intrval (GI) is mad up as a prfix of a cyclic continuation of th usful part of th OFDM symbol usd to prvnt ISI. Th availabl GI valus ar: 1/4, 19/128, 1/8, 19/256, 1/16, 1/32, and 1/128. Fig. 1. DVB-T2 systm. Thr ar two transmission options in DVB-T2, Singl Input - Singl Output (SISO) and Multipl Inputs - Multipl Outputs (MIMO). Th MIMO will b usd as MISO implmntd with a modifid Alamouti schm [11]. Thn, rfrnc information, known as pilots, is insrtd to hlp th rcivr to compnsat for th transmission channl distortion. Thr ar various typs of pilots that appar in diffrnt typs of OFDM symbols within a DVB-T2 fram. III. SCATTERED PILOTS IN DVB-T2 DVB-T2 uss 8 diffrnt SP pattrns, namd PP1 to PP8, to compnsat for th variation of th channl in th tim and in th frquncy domain. Th position of th pilots onto a subcarrir in th OFDM symbol satisfis th following condition: k mod( D D ) = D ( mod D ) (2) x y x y whr D x dfins th sparation of pilots baring carrirs in
> mm 13-48 IEEE BMSB 2013 < 3 ach OFDM symbol, D y dfins th numbr of OFDM k K, K is th symbols forming on SP squnc, [ min max ] indx of subcarrir into th OFDM symbol, and [ 1, S] is th indx of th OFDM symbol into th T2 fram that contains S symbols in total. Th availabl valus of D x and D y givn in Tabl II thortically support fluctuations in tim and frquncy up to th Nyquist limit. Thus, th spacing in tim and frquncy domain should not xcd th rspctiv limit givn blow: whr A SP is th SP amplitud in Tabl III, r, k is th rfrnc squnc gnratd by XOR-ing a Psudo Random Binary Squnc (PRBS) with a Psudo-Numbr, m is th indx of th T2 fram, k is th frquncy indx of th carrirs, and is th indx of th OFDM symbol. Finally, th arrangmnt for two Pilot Pattrns (PP) in SISO mod for DVB-T2 including Edg pilots ar shown in Fig. 2 (a) for PP1 and (b) for PP6. 1 1 Nt, N f (3) 2f T 2 Δ fτ D max whr f D is th Dopplr frquncy, Δf is th spacing btwn subcarrirs, T is th duration of th OFDM symbol, and τ max is th multipath dlay. Th maximum GI fraction should nvr xcd 1/D x. From (3) th Dopplr limit is proportional to 1/D y. Th capacity of transmission, nglcting all othr typs 1 of pilots (cf. Tabl I), is a fraction of. From Tabl II, DxD y th ovrhad is drivd to b qual to 8.33% for PP1 and 1.04% for PP7 and PP8. It is obvious that as th ovrhad incrass, highr valus ar obtaind for th Nyquist limit. With D y = 2, PP2, PP4 and PP6 provid highr Nyquist limit for Dopplr spd, [12]. TABLE II PARAMETERS FOR SP PATTERN FORMATION Pilot Pattrn D x D y PP1 3 4 PP2 6 2 PP3 6 4 PP4 1 2 PP5 1 4 PP6 2 2 PP7 2 4 PP8 6 1 To rduc th nois on channl stimation, th pilots ar boostd considring that all th symbols hav approximatly th sam powr. Tabl III shows th amplituds of ach SP pattrn. TABLE III AMPLITUDES OF THE SCATTERED PILOTS SP pattrn Amplitud Equivalnt Boost in db PP1,PP2 4/3 2.5 PP3,PP4 7/4 4.9 PP5,PP6,PP7,PP8 7/3 7.4 Th modulation of th pilots dpnds on th carrir indx and th symbol numbr. Th ral and imaginary parts ar givn blow: R m,, k 1 = 2 ASP( r, k) 2 Im = 0 { c } { cm,, k} (4) Fig. 2. SP pattrns in SISO mod. (a) SP pattrn for PP1 (b) SP pattrn for PP7 IV. CHANNEL ESTIMATION Aftr srial to paralll transformation and pilot insrtion dpnding on spcific SP pattrn, th data squnc {X(k)} is transformd from a frquncy domain into a tim domain signal {x(n)} by th IFFT block. Thus, th transmittd signal is transformd by th following xprssion: N 1 k= 0 2π nk j N x(n) = IFFT{ X( k)} = X( k) (5) Thus, th rcivd signal is as dscribd blow: L 1 yn ( ) = hn (, ) xn ( ) + wn ( ) (6) l = 0 whr 0 n N 1, L is th numbr of multipath vrsions of th original signal x(n), hn (, ) is th channl impuls rspons of th n-th OFDM symbol from th -th path and w(n) is th Additiv Whit Gaussian Nois (AWGN) with zro man and varianc σ 2. At th rcivr sid, aftr cyclic prfix (CP) rmoval and FFT, it is drivd that: (, ) = FFT{ y( n) } = = X( kn, ) H( kn, ) + I( kn, ) + W( kn, ) Y k n (7)
> mm 13-48 IEEE BMSB 2013 < 4 whr: H ( kn, ) is th channl frquncy rspons, W ( k, n ) is th nois signal in th frquncy domain, and I ( kn, ) dnots th introducd Intr Carrir Intrfrnc (ICI) causd by th tim-varying channl. In this study, a vry slowly timvarying channl is assumd and thus th I ( kn, ) trm can b nglctd. If P is th vctor of th position of th pilots in th OFDM symbol, thn th stimatd channl frquncy rspons p H ( k ), k P = [ p0, p1,..., p P ] undrgos Splin intrpolation in ordr to obtain th channl rspons H ( k ), k [0, kmax ] for vry carrir of th OFDM symbol. Finally th channl qualization will b don using th Zro Forc (ZF) qualizr or th Last Squar Error (LSE) qualizr, dscribd rspctivly by th following two xprssions: Y( k) X =, k 0,1,..., Kmax H ( k) = (8) DTV, and v( m s) th mobil spd. Combining (10), (11) and (12), it is drivd that: c B = floor ( ) (13) 100vfcTu If B<1 in (13), thn w st B = 1 and actually no avraging is prformd. That is why th proposd algorithm is suitabl for low mobil spds or quivalntly for larg cohrnc tim T c whr th channl varis vry slowly in tim. Howvr, th prformanc of th proposd stimator is tstd also for fixd valus of buffr siz, in ordr to clarly dmonstrat its capabilitis and limitations. Th avraging stimator oprats for th last B rcivd OFDM symbols as follows. First, th symbols ar stord in a matrix A of siz B N. Thn a vctor avpilots is constructd as givn blow: B 1 avpilots(n) = A(i,n) (14) P n i = 1 X = H * H * H Y (9) whr P n is th numbr of pilots in th nth subcarrir. It must b notd that if th n-th subcarrir dos not carry pilots, thn P n = 1. Th avraging procss is dpictd in Fig. 3. whr (.) * dnots th conjugat of a complx numbr or a complx matrix. V. PROPOSED ESTIMATOR Assuming a slowly tim-varying channl, th stimation of p H ( k ) can b calculatd by simply avraging th valus of th last B rcivd OFDM symbols. Th valu of B is drivd as follows: Lt T B b th tim intrval btwn th last rcivd OFDM symbol S and th S (B+1) OFDM symbol. Thn: T B = B T (10) u whr T u is th duration of an OFDM symbol, [1]. In ordr to nsur that TB < Tc, whr T c is th cohrnc tim, w st: T B 1 = Tc (11) 50 Th cohrnc tim can b st qual to Tc = 0.5 fd, whr f d is th Dopplr spctrum, [13], and is xtractd by th xprssion: Fig. 3. Th avraging procss of th last B rcivd OFDM symbols for PP1. Aftr th construction of th vctor avpilots, with splin intrpolation, th channl frquncy rspons Ĥ ( n ) is calculatd and by using a LSE stimator th rcivd data symbols ar qualizd. It is important to not that th proposd stimator can b usd in conjunction with mor sophisticatd stimators, such as MMSE and Kalman filtrs, whr th knowldg of th channl is an important factor for thir good prformanc. As th avraging stimator rjcts th nois, (8) can b rwrittn as: ( ) ( ) Y k H = + W( k) (15) X k f d v = fc (12) c Sinc H Y ( k) X ( k) av =, th prvious quation can b writtn in following form: whr c is th spd of light, f = 800MHz is th uppr limit of th carrir frquncy of th analog transmittd signal in c Hˆ = Hˆ + Wˆ (16) av
> mm 13-48 IEEE BMSB 2013 < 5 whr H ˆ av is th avragd stimation of th channl, Ĥ is th convntional channl stimation and W ˆ is th nois. So, this could b usd for nois stimation and thn to pass this information to a mor sophisticatd stimator. VI. SIMULATION AND RESULTS In this sction, th prformanc of th avraging stimator is tstd undr various configurations of QAM ordr, FFT siz, channl modl, spd of th rcivr, and siz of th buffr. For convninc in th comparison of th diffrnt configurations, th bandwidth of th RF signal is st qual to 8MHz, which is th usual bandwidth in DVB-T. In ordr to us on of th highst possibl frquncis, which suffr th most of th Dopplr ffct, th cntral carrir frquncy is st to b qual to 800MHz. Two fading channl modls ar usd, th Rayligh and Rican ons, with 6 taps ach. No sourc or channl ncoding is usd and non of th availabl intrlaving schms ar usd, in ordr to focus only into th improvmnt offrd by th stimator. Th first simulation tsts th prformanc of th stimator in a pur AWGN channl with no multipath componnts. A 4- QAM modulation is usd (QPSK), th FFT siz is 4k and th buffr siz B is st qual to 20. Fig. 4 dpicts th prformanc of th proposd and th convntional stimator. It is clar that th proposd stimator ovrcoms th convntional on by about 3.5dB for th sam BER. Fig. 5. Comparison btwn th proposd and th convntional stimator with spd qual to 1 Km/h. Th sam configuration as in Tabl IV is usd blow, xcpt th mobil spd which is st qual to 3 Km/h. Th buffr siz is qual to 16 which is a suitabl valu. Thus th prformanc is similar to th prvious on and th avraging stimator still works accptably. Th rsults ar dpictd in Fig. 6. Fig. 6. Comparison btwn th proposd and th convntional stimator with spd qual to 3Km/h. Fig. 4. Comparison btwn th proposd and th convntional stimator in AWGN channl. Mor ralistic configurations ar usd blow. Th first on is displayd in Tabl IV. Th buffr siz calculation is basd on (13). Fig. 5 dpicts th improvmnt bcaus of using th avraging stimator which ffctivly rducs th nois. Finally, th mobil spd is st qual to 10 Km/h. According to (13), th buffr bcoms qual to 4 and thus dgradation is xpctd. Th curvs in Fig. 7 vidntly show th prdictd dgradation. For SNR < 20 db no improvmnt occurs. For highr SNR valus, th stimator works again as in th prvious simulations. Furthr incras in th mobil spd lads to B = 1 and thus no avraging is prformd. TABLE IV CONFIGURATION OF SIMULATION Radio Environmnt Rural Ara Radio Channl Typ Rayligh QAM-Ordr 4 FFT siz 4k Spd of Mobil (Km/h) 1 Buffr Siz 49
> mm 13-48 IEEE BMSB 2013 < 6 componnts). For K=10 5 (vry strong lin of sight componnt) th avraging stimator outprforms th convntional on, as shown in Fig. 12. For K=10 (wak lin of sight componnt) th prformanc of th avraging stimator is vn bttr, as shown in Fig 13. Fig. 7. Comparison btwn th proposd and th convntional stimator with spd qual to 10 Km/h. Thn, th prformanc of th proposd stimator is xamind for diffrnt QAM ordrs. Th configuration of Tabl IV is usd again and only th QAM ordr is changd to 16-QAM. Th proposd stimator again succds to improv th rcption up to 3dB as shown in Fig. 8. Although th stimator prforms bttr than th non-avraging on for highr QAM ordrs with improvmnt up to 6 db, as dpictd in Fig. 9, th ovrall prformanc is not accptabl. Fig. 9. Comparison btwn th proposd and th convntional stimator with 64-QAM. Fig. 10. Comparison btwn th proposd and th convntional stimator with FFT = 1k. Fig. 8. Comparison btwn th proposd and th convntional stimator with 16-QAM. Nxt, th FFT siz is tstd. Th configuration of Tabl IV is usd again but th FFT is st qual to 1k, so th buffr siz bcoms B = 196. Th curvs of BER vs. SNR for th two invstigatd stimators ar dpictd in Fig. 10. Th supriority of th avraging stimator is obvious. Th stimators ar also xamind for FFT = 8k which corrsponds to B = 24. Fig. 11 illustrats th prformanc of th stimators and shows that whil th convntional on convrgs to a platau, th avraging stimator lads to lowr valus of BER. It is intrsting to s how th proposd stimator bhavs for anothr typ of fading channl modl. Thrfor, th configuration of Tabl IV is usd again xcpt th radio channl typ which is considrd to b Rican with crtain valus of th radio channl factor K (i.., th ratio btwn th powr in th lin of sight and th powr of th multipath Fig. 11. Comparison btwn th proposd and th convntional stimator with FFT = 8k.
> mm 13-48 IEEE BMSB 2013 < 7 nhancmnt. Th avraging stimator succds to discard th nois and thus it achivs lowr valus of BER. In th cas of PP7, th pilot arrangmnt is sparsr (s Fig. 2b) and th convntional stimator prforms vn wors. Th avraging stimator is also affctd but still prforms bttr than th convntional on. Fig. 12. Comparison btwn th proposd and th convntional stimator with Rican radio channl and K=100,000. Fig. 15. Comparison btwn th proposd and th convntional stimator for PP7. Fig. 13. Comparison btwn th proposd and th convntional stimator with Rican radio channl and K=10. VII. CONCLUSION Th prformanc of an avraging stimator is thoroughly invstigatd. Initially, th stimator is tstd in a frquncy flat channl suffring only of AWGN and th prformanc of th stimator is found to b 3.5 db bttr than a non-avraging on. Th proposd stimator is also tstd in frquncy slctiv channls and for diffrnt sizs of FFT, QAM ordr, radio channls, mobil spds, and pilot pattrns. In all cass, for vry low mobil spds, th proposd stimator givs bttr rsults than th convntional on. As xpctd, th proposd stimator fails for highr mobil spds du to th Dopplr Effct. As th spd incrass th channl rspons bcoms tim-varying and th cohrnc tim gts smallr. So, fwr OFDM symbols can b buffrd and thus th avraging procdur is lss ffctiv. Th proposd stimator can also b usd for stimation of th nois varianc. This information can b utilizd by mor accurat and complicatd stimators basd on channl statistics in ordr to mak a mor accurat channl stimation. Fig. 14. Comparison btwn th proposd and th convntional stimator for PP3. Finally, th proposd stimator is tstd for diffrnt SP pattrns. Th sam configuration of Tabl IV is usd. Th PP3 and PP7 ar tstd and th rsults ar dpictd rspctivly in Figs. 14 and 15. It is clar that th proposd stimator prforms accptably vn in cass whr th availabl pilots ar fwr than PP1 such as in PP3. Th convntional stimator rsults in a flat BER curv for SNR > 18dB du to nois REFERENCES [1] Digital Vido Broadcasting (DVB). Fram structur channl coding and modulation for a scond gnration digital trrstrial tlvision broadcasting systm (DVB-T2), ETSI EN 302 755 V1.3.1, Apr. 2012. [2] Digital Vido Broadcasting (DVB). Scond gnration framing structur, channl coding and modulation systms for Broadcasting, Intractiv Srvics, Nws Gathring and othr broadband satllit applications (DVB-S2), DVB Documnt A83, Jul. 2012. [3] Digital Vido Broadcasting, Fact Sht, 2nd Gnration Trrstrial, DVB-T2, Sp. 2012. Availabl: http://www.dvb.org/tchnology/fact_shts/dvb-t2_factsht.pdf [4] S. Colri, M. Ergn, and A. Bahai, Channl stimation tchniqus basd on pilot arrangmnt in OFDM systms, IEEE Trans. Broadcast., vol. 48, no. 3, pp. 223 229, Sp. 2002.
> mm 13-48 IEEE BMSB 2013 < 8 [5] S. Tomasin and M. Butussi, Analysis of intrpolatd channl stimation for mobil OFDM systms, IEEE Trans. Communications, vol. 58, no. 5, pp. 1578 1588, May 2010. [6] F. Salman, J. Cosmas, and Yu Zhang, Modlling and prformanc of a DVB-T2 channl stimator and qualisr for diffrnt pilot pattrns, in Proc. 2012 IEEE Intrnational Symposium on Broadband Multimdia Systms and Broadcasting (BMSB), Jun. 2012, pp. 1 6. [7] Li Fu, Songlin Sun, Xiaojun Jing, and Hai Huang, Analysis of pilot pattrns and channl stimation for DVB-T2, in Proc. 2010 2nd IEEE Intrnational Confrnc on Ntwork Infrastructur and Digital Contnt, Sp. 2010, pp. 609 613. [8] X. Cai and G. B. Giannakis, Error probability minimizing pilots for OFDM with M-PSK modulation ovr Rayligh-fading channls, IEEE Trans. Vh. Tchnol., vol. 53, no. 1, pp. 146 155, Jan. 2004. [9] Mingchao Yu and P. Sadghi, A study of pilot-assistd OFDM channl stimation mthods with improvmnts for DVB-T2, IEEE Trans. Vh. Tchnol., vol. 61, no. 5, pp. 2400 2405, Jun. 2012. [10] L. Martínz, J. Robrt, H. Mul, I. Sobrón, and M. Mndicut, Improvd robustnss for channl stimation without pilots for DVB- T2, in Proc. 2010 IEEE Intrnational Symposium on Broadband Multimdia Systms and Broadcasting (BMSB), Mar. 2010, pp.1 5. [11] S. Alamouti, A simpl transmit divrsity tchniqu for wirlss communications, IEEE Journal Slct. Commun., vol. 16, no. 8, pp. 1451 1458, Oct. 1998. [12] Digital Vido Broadcasting (DVB). Implmntation guidlins for a scond gnration digital trrstrial tlvision broadcasting systm (DVB-T2), ETSI TS 102 831 V1.2.1, Aug. 2012. [13] B. Sklar, Digital Communications, Fundamntals and Applications, 2nd Ed. Nw Jrsy, USA: Prntic Hall, 2006, pp. 971.