Modeling and Simulation of MISO Diversity for UHF RFID Communication

Transcription

1 Proceedings of he Federaed Conference on Compuer Science and Informaion Sysems pp ISBN Modeling and Simulaion of Diversiy for UHF RFID Communicaion Grzegorz Smieanka and Jürgen Göze 2 Informaion Processing Lab Deparmen of Elecrical Engineering and Informaion Technology, TU Dormund Universiy grzegorz.smieanka@udo.edu, 2 :juergen.goeze@udo.edu bsrac Radio Frequency Idenificaion (RFID) is used in high scaering environmens where deep fading exiss. This makes diversiy paricularly ineresing for his communicaion scenario. In his paper he poenial of using muliple ag anennas for RFID-communicaion is shown. The bi error rae (BER) and packe error rae () is presened, which include he backscaering answer of a RFID-Tag according o he EPC Class- Gen-2 proocol. The raes are regarded in combinaion wih he relevan fading channel models for RFID communicaion such as he Rician- and he Dyadic Backscaer Channel. I is shown ha he possible diversiy gain from he signal according o EPC Class- Gen-2 proocol is several db regarding he error raes. I is also shown ha his diversiy gain increases wih he correlaion of he forward and backward link and decreases wih he usage of a more robus encoding scheme and a correlaion beween he several ransmission pahs. ddiionally he performance of Muliple Inpu Single Oupu () sysem wih differen spaial and forward/backward correlaion siuaions is regarded o have a deailed view on a correlaed RFID ransmission sysem using diversiy. The performance of his model is verified, using simulaions of his propagaion sysem. Index Terms RFID, Communicaion, Diversiy Mehods, Numerical Simulaion, Rician-, Dyadic Backscaer Channel I. INTRODUCTION In RFID echnology he communicaion is done by backscaering [], where a coninuous wave (CW) is sen ou o a ag and refleced by is ransmier. The reflecion can be driven by he receive power of he ransmier. This makes he echnology very cheap as he ags can be produced in large quaniies and wih small dimensions because of he inexisen power source. Furhermore i can provide non line of sigh communicaion over a couple of meers. The primary field of research for passive RFID ags is aiming a he improvemen of reading accuracy and operaing range. To model he RFID propagaion a environmen wih a fading channel is necessary. In his scenario a fading channel wih a Rician disribuion can be used. Bu for a backscaer link his model can only provide a rough modeling accuracy. Therefore a Dyadic Backscaer Channel was inroduced in [2], where he characerisic of RFID communicaion wih forward and backward link is considered. This implies deeper fades which decrease he performance of he sysem significanly. I is shown in [3] ha for cerain consellaions he read rae is very poor even if he range beween he reader and he ag is no very large. The EPCglobal communicaion sysem [4] does no suppor any forward error correcion (FEC). This becomes a serious problem when he sysem operaes in noisy fading channels. common mehod o raise he performance in a fading channel wihou using FEC is he usage of muliple anennas. In [5] a diversiy scheme is presened, where he usage of muliple anennas clearly increases he performance of he sysem. In his paper he performance of he EPCglobal ransmission proocol in case of using wo ag anennas is invesigaed. I is shown ha due o he disadvanageous channel consellaion in RFID communicaion he possible diversiy gains for he signal according o EPCglobal proocol is several db regarding he bi- or packe error rae. I will be also shown ha he diversiy gain is sill significan when he spaial correlaion during a ransmission is high. In secion II he basics for RFID communicaion including he en- and decoding of he backscaer signal are presened. The relevan channel models are inroduced in secion III. In secion IV a ransmi diversiy scheme according o [5] is described. Simulaion resuls using his combinaion scheme for ransmission and are shown for differen channel consellaions in secion V. II. RFID COMMUNICTION SYSTEM Encoding Modulaion Demodulaion Decoding PIE SK/PSK h f n b Decoding Demodulaion Modulaion (Ã) Encoding Fig.. n f h b SK (OOK) RFID schemaic ransmission link Processing Reader-Info FM / MILLER-2/4/8 Reader Tag The communicaion sysem, which is considered in his work, is based upon he EPCglobal sandard [4] for UHF RFID ransmission. During a ransmission he reader emis a coninous wave. On his wave he informaion for a populaion of ags is pulse inerval encoded (PIE) and modulaed via ampliude- (SK) or phase shif keying (PSK). ny ag in he read range of he reader will send back his informaion by reflecing he incoming CW using a FM /$25. c 22 IEEE 83

2 84 PROCEEDINGS OF THE FEDCSIS. WROCŁW, 22 or Miller subcarrier encoding and an SK modulaion. For a more deailed descripion of UHF RFID communicaion we refer o he EPCglobal sandard [4]. Subsequenly only he backscaering answer of he ag o he reader is considered because of heir higher sensiiviy o noise. Neverheless he forward and backward fading value are aken ino accoun because of heir dependence during he ransmission. In figure his implies he marked area. The en- and decoding of he ag informaion is examined in he following subsecion.. En- / Decoding The EPCglobal sandard conains wo specificaions for he encoding of he backscaer daa. For boh encoding ypes one bi is spread ino a chip sequence which is defined as a sequence of several one/zero combinaions, where more chips implies a more robus bu also more ime cosly code. The mos simple ransmission mehod is he FM encoding. I is disinguished by one edge on he boundery of wo symbols and an edge change in he middle of a symbol which describes a zero bi. s a resul one bi can be FM-encoded wih wo chip values. Figure 2 shows he sae diagram for his encoding scheme. h h l l l h l h T h h l l Fig. 3. l h l h Miller encoding sae diagramm every received chip value larger han he hreshold is se o his hreshold: = for > () This hreshold is esablished for a couple of chips which reach very high ampliudes because of very deep fades. Deep fades are equivalen wih high noise, which changes he ampliude of he ransmied sysmbol drasically. This have a sligh negaive impac on he performance of he sysem. Subsequenly i is assumed ha = h. Noe ha wihou his hreshold he coding gain for some of he used encoding schemes is several db lower. III. CHNNEL MODELS The channel can be modeled by a weighing of he ranmied symbol wih a specific fading coefficien wih normalized power. This value is sen over a common ddiive-whie- Gaussian-Noise (WGN) channel. Boh channel ypes use he following channel equaion: T Fig. 2. FM encoding sae diagramm r = h s+a n (2) For he Miller-M subcarrier encoding one phase shif is done on he boundary of wo zero symbols and here is also a phase shif in he middle of one symbol [4]. In his case a phase shif implies ha he signal ampliude does no change form one period o anoher. This concep is shown in figure 3, where subcarriers are no considered. In pracice M represen he number of cycles used for he encoding of one bi, where every symbol needs = 2M T wih M {2,4,8} o be ransmied. T is he ime period of one chip. In his case M = 8 sands for he mos robus bu also for he mos ime consuming ransmission mode. For he ransmission of he encoded values he daa is sampled wih one value per chip duraion T. The decoding is done via correlaion of he noisy bi sream wih he possible symbols [6]. For his calculaion he received SK signal is shifed, so he high and he low ampliude level have he same absolu value and differen signs. The possible symbols, which are used for he correlaion correspond o ha scheme. ddiionally he hreshold is inroduced, where The ransmied signal s is muliplied wih a complex channel facor h and ransmied over a complex WGN channel. The real and imaginary par of n are boh random Gaussian numbers, where n r and n i are uncorrelaed: n r N(, σ2 2 ); n i N(, σ2 2 ) (3) where σ 2 represens he variance / power of he complex Gaussian noise [7], a is a facor represening he Signal o Noise Raio (SNR) P a = SNR db, (4) P sands for he mean signal power and SNR db is he SNR in db. The fading of he channel is considered in h. This coefficen can be seen as a facor which in- or decreases he WGN noise. Noe ha he successive approach can be easily ransformed o a sysem wih any number of ransmi-, receiveor ag-anennas as shown in [2].

3 GRZEGORZ SMIETNK, JÜRGEN GÖTZE: MODELING ND SIMULTION OF DIVERSITY FOR UHF RFID COMMUNICTION 85. Rician Channel The RFID communicaion is done in a rich scaering environmen where a line of sigh (LOS) pah is ofen available. To consider he performance for such a consellaion a Rician channel model is used. The channel facor h for his model has wo componens. The real valued line of sigh (LOS) componen and he complex none line of sigh (NLOS) componen which represens he scaering par of he signal. This par is represened by a complex Gaussian random number h ray wih he same characerisic as he random number of he Gaussian noise wih zero mean and normalized power. The relaionship beween boh componens is h = P LOS + P NLOS h ray (5) where P LOS = K K+ and P NLOS = K+ can be describe by he Rician facor K = PLOS P NLOS, which characerizes he relaionship beween he power of he LOS and he power of he NLOS componens. ccording o he sum of P LOS and P NLOS and he normalized power of h ray, he power of h is uniy, so his facor does no influence he SNR characerisic of he channel. The envelope of he random variable h follows he Rician disribuion and can be defined as a funcion of K [8]: f( h ) =2 h (+K) exp( h 2 (+K) K) I (2 h K(K +)) (6) where I is he modified Bessel funcion of he firs kind and zeroh order. For K = he Rician disribuion resuls in a Rayleigh disribuion and for K he disribuion becomes a dirac a h = which is equivalen o an WGN channel. The behavior of his disribuion for varying K is shown in figure 4. f( h ) K= K= K= is a forward and a backward link available. In his case a Rican channel can only be considered as a rough model for RFID communicaion. For a more precise descripion a Dyadic Backscaer Channel should be analysed. This model was firs described in he conex of RFID communicaion in [2]. The channel can be describe as a wo way channel wih a forward and a backward link. These fading coefficiens are h f and h b. For he overall fading per link boh componens can be muliplied o h = h f h b (7) The fading of he forward and he backward channel can be generaed as described in he previous secion. To consider he saisical dependence of he forward and he backward link he link correlaion ρ is inroduced, where ρ = represens a saisical independen forward- / backward fading and ρ = a fully correlaed forward- / backward link, which is equivalen wih h f = h b. For a channel characerisic wih a link correlaion < ρ < h f,ray and h b,ray can be obained by wo uncorrelaed Rayleigh fading coefficiens h U and h U2 wih zero mean and equal varianceσu 2 = σ2 U2 in he following way: h f,ray =h U (8) h b,ray =ρ h U + ρ 2 h U2 (9) The relaion beween ρ and he fading value is given as follows: ρ = 2Cov(Re(h f,ray),re(h b,ray )) σ f σ b = 2Cov(Im(h f,ray),im(h b,ray )) σ f σ b () where Cov(, ) is he covariance operaor and σ f and σ b are he sandard deviaions of he fading values. In he following σ f = σ b = is valid. lso Rician fading coefficens are generaed wih his assumpion. In his case he Rayleigh numbers are calculaed firs and hen ransformed ino Rician numbers [9] wih formular (5). ddiionally o he forward/backward correlaion a spaial correlaion ρ spa as generally described in [] is considered during he ransmission wih wo ag anennas. ρ spa is defined in he same way as ρ. Considering a ransmission wih one reader and wo ag anennas over a dyadic backscaer channel, he fading coefficiens for one ransmission period are describe in he following way. Fig h Envelope for a Rician disribued number h wih differen K B. Dyadic Backscaer Channel The Rician channel is an adequae model for LOS fading channels bu i is resriced o one way channels. s he RFID communicaion from he ag is done via backscaering here h f,ray =h U () h b,ray =ρ h U + ρ 2 h U2 (2) h f2,ray =ρ spa h f,ray + ρ 2 spa h U3 (3) h b2,ray =ρh f2,ray + ρ 2 ( ) ρ spa h b,ray + ρ 2 spa h U4 (4) where h U,...,h U4 represen uncorrelaed fading coefficens. n alernaively calculaion of h b2,ray, where h f2,ray

4 86 PROCEEDINGS OF THE FEDCSIS. WROCŁW, 22 and ρ are swiched wih h b,ray and ρ spa is also possible. For correlaed links in a one way channel (like a propagaion in a Rician channel), h b and h b2 are no considered during he calculaion. s s s * s * TX TX2.8.6 Rayleigh = = h h 2 RX f( h ).4.2 n n combiner h Fig. 6. lamoui ransmission scheme wih wo ransmiers and one receiver TBLE I ENCODING SCHEME FOR TRNSMIT DIVERSITY Fig. 5. Envelope of he Dyadic Backscaer disribuion wih Rayleigh random numbers compared o a normal Rayleigh disribuion To disinguish he differences beween he Dyadic Backscaer and a one way channel he probabiliy densiy funcions (PDF s) of channels beween one reader and a ag anenna for ρ = and ρ = respecively, are described. In his case boh links in he backscaer channel have a Rayleigh disribued fading, which resuls in [2]: f( h ) = h ρ+ ( 2 2 ρ ρ+ Γ( ρ+ ) K ρ ρ+ )ρ+2 ρ+ (ρ+)σ f σ b ( ) 2 h (ρ+)σ f σ b (5) whereγ( ) is he Gamma funcion andk ρ ( ) is he modified ρ+ bessel funcion of he second kind and ρ/(ρ+) order. The PDF s are shown in figure 5 and (5). The random numbers of he backscaer signal end o have smaller values ( h < ) which is equivalen o more desrucive fading during he ransmission. Fuhermore he deep fading is more disincive for a high correlaion of forward and backward channel. IV. TRNSMIT DIVERSITY TECHNIQUE To achieve a ransmi diversiy he lamoui-scheme is applied for wo ransmi anennas as described in [5] and shown in figure 6. The encoding scheme for his ransmission is shown in able I where wo arbirary symbols are send over wo anennas in one iming period. In he nex period + T he symbols swich on he anennas and addiionally boh symbols are conjugaed and he firs symbol is invered. Wih his approach a combinaion of boh symbols can be accomplished in he receiver. TX TX2 s s +T s s For he combinaion scheme i is assumed ha he fading is consan over wo iming periods. h () =h (+T) h 2 () =h 2 (+T) (6) This assumpion and he encoding in able I leads o he following received signals r =r() = h s +h 2 s +n r =r(+t) = h s +h 2 s +n (7) To combine boh signals, r and r are weighed wih he fading coefficiens and hen added. s =h r +h 2 r s =h 2 r h r (8) Because of (8) i is assumed ha we have a perfec channel esimaion a he receiver. Therefore he resul of he calculaion can be wrien as s =γs +h n +h 2 n s =γs h n +h 2 n (9) where γ is a scaling facor which is proporional o he magniude of he fading coefficiens. fer he combinaion i is obvious ha he daa pair can be exraced. During his work he diversiy scheme is used before he signal is modulaed, so he encoding of he symbols on boh anennas is done on chip level. fer he combinaion of he signals a he receiver he decoding of he bis is done as describe in secion II. Noe

5 GRZEGORZ SMIETNK, JÜRGEN GÖTZE: MODELING ND SIMULTION OF DIVERSITY FOR UHF RFID COMMUNICTION 87 ha for his combinaion scheme a perfec synchronizaion of he ransmi anennas is assumed. This holds rue for one ag wih several anennas. Bu hese ags are no used in commercial sysems. The usage of muliple commercial ags o achieve ransmi diversiy is more complicaed because of he limiaions in hardware and he required synchronizaion beween several separaed ags. V. RFID PROPGTION MODELING The modeling of he RFID propagaion is done according o he EPCglobal proocol. In his work only he backscaer ransmission from a ag o he reader is considered, because of he higher sensiiviy o noise compared o he forward link. For a backscaer ransmission he daa is encoded wih a spreading sequence as described in secion II. The spreaded daa is hen SK modulaed and aferwards ransmied over he channel. s an exension o his proocol he ransmi diversiy echnique, described in secion IV, is implemened before he daa is modulaed in case of a ransmission wih wo ag anennas. Subsequenly he channel correlaion is calculaed as describe in ()-(4). To he bes of our knowledge, differen ransmi diversiy siuaions wih he BER for a RFID ransmission using he EPCglobal proocol were no invesigaed so far. For a comparison of he diversiy gain he BER is shown over he SNR for cerain channels. The SNR is he raio beween he energy of one ransmied and modulaed symbol E s and he noise power specral densiy N. lso he number of chips per symbol should be aken ino accoun bu as an SK modulaion for all following resuls is used, here is one chip mapped on one modulaion symbol and he SNR does no change in his case. s we use a spreading sequence for he encoding looking a he BER and over he raio of he bi energy E b per N would also make sense. Bu as his paper observes he ransmi diversiy and no he performance of he codes boh SNR definiions can be used. If E b /N is uilized he curves in he upcoming figures move 3,6,9 and 2 db o he righ for he FM, Miller-2, Miller-4 and Miller-8 codes respecively. I is o menion ha a ransmission gain of 3 db is considered in he communicaion wih wo anennas. Therefore, half of he power is uilized on each anenna during he ransmission. For a fair comparism of he used diversiy a channel coefficien h should also be consan for wo ransmission symbols when no diversiy is used. I will be shown in secion VI ha channel coefficiens wih differen duraions have a huge influence on he performance of he encoded bi sequences (bu no influence on a sequence which is no encoded). The funcionaliy of he channels was addiionally verified wih he resuls of [2], [] and [2]. For he i is assumed ha he maximal number of backscaer bis is ransmied according o EPCglobal. This implies he RN6 wih 6 Bis and he EPC wih 528 Bis wih wo preambles in fron of hese sequences wih 8 or 22 bis for FM or Miller subcarrier encoding respecively. I is also assumed ha a packe can be deeced correcly if a leas 95% of he preamble is accuraely deeced. n erroneous deecion for one informaion bi is equal o a packe error. BER FM SISO FM SNR [E /N ] s [db] Fig. 7. Bi error rae for differen backscaer encoding schemes and ransmission wih one (SISO) or wo () ag anennas and one reader anenna in a Rayleigh fading channel. BER FM SISO FM Miller-2 FM Miller-4 Miller SNR [E [E s /N ] s /N ][db] Fig. 8. Bi error rae for differen backscaer encoding schemes and ransmission wih one (SISO) or wo () ag anennas and one reader anenna in a Rician fading channel wih Rican Fakor K = 3.. Diversiy Behavior VI. SIMULTION RESULTS Figure 7 shows he BER for a ransmission over a one way Rayleigh channel which is used as a reference o he oher BER plos. The diversiy gain exiss for every encoding scheme and increases as he spreading of he codes decreases. This leads o a diversiy gain of 6.5 db for a BER= 4 and even 7 db for a BER= 2. This gain decrease wih a more robus encoding sequence. So, for he mos solid Miller-8 encoding a diversiy gain of5.3db and2.3db for a BER= 4 and 2 is achieved, respecively. The good performance considering he diversiy of he weaker spreading sequences is because of

6 88 PROCEEDINGS OF THE FEDCSIS. WROCŁW, 22 he low coding performance during he SISO ransmission. In his case he poenial for diversiy is much larger for more fragile spreading sequences. Noe ha in a usual UHF-RFID ransmission FM and Miller-2 codes are he mos common spreading sequences in virue of a faser ransmission [4]. For a one way channel wih LOS pah (Rice channel wih K = 3 in fig. 8) he diversiy gain decreases by o 2.5 db compared o fig FMO FM SISO FM SNR [E [E ] s /N ] [db] Fig. 9. Packe error rae for a EPC lengh of 528 bis, differen backscaer encoding schemes and ransmission wih one (SISO) or wo () ag anennas and one reader anenna in a Dyadic Backscaer Channel wih an uncorrelaed forward and backward link. Boh links have a Rician disribuion wih K = FM SISO FM SISO FM FM SNR SNR [E [E s /N s /N ] ] [db] Fig.. Packe error rae for a EPC lengh of 528 bis, differen backscaer encoding schemes and ransmission wih one (SISO) or wo () ag anennas and one reader anenna in a Dyadic Backscaer Channel wih a fully correlaed forward and backward link. Boh links have a Rician disribuion wih K =.5. Looking a he resuls in a fully uncorrelaed (ρ = ρ spa = ) Dyadic Backscaer Channel in figure 9 he diversiy gain increases compared o an one way channel. This is also he case when he BER s are compared. The poenial of diversiy ransmission for RFID communicaion becomes clear because of he loss caused by he deeper fades compared o a one way channel. The gain also increases for a fully correlaed forward and backward link (ρ = ) and a uncorreaed spaial link (ρ spa = ) as shown in figure. I is very likely ha he correlaion in a real RFID ransmission is high if he reader anenna serves as a ransmier and a receiver. Noe ha he performance of he encoding schemes varies wih he emporal characerisic of he channel coefficien. Regarding he BER for SISO ransmission when he channel coefficien is no changed during wo symbol periods compared o a change afer one ransmission symbol shows significan differences in he performance of he spreading sequences. Consider a Rician channel wih K = 3 and a BER of 4 he performance wih one channel coefficen h per symbol is up o db beer compared o a ransmission wih one channel coefficien for wo symbols when a Miller-4 encoding is assumed. Using he Miller-8 and Miller-2 codes a gain of.9 db and.3 db is achieved, respecively. The FM has even advanages ( db) when using one channel coefficien for wo symbols. This behavior is similar for differen channel parameers and increase as he general performance of he encoding scheme decrease. In a fully correlaed dyadic backscaer channel he gain for a Miller-4 code and differen ransmission characerisic is nearly 3 db. There is one main reason for his behavior. channel coefficien which is valid for wo ransmission symbols makes a robus spreading sequence more fragile because of he higher influence of deeper fades (small channel values). The Miller-8 code is robus enough o reasonably compensae hese disadvanages. The Miller-2 and FM codes are very fragile owards errors, so a change in he characerisic of he channel coefficien does no make a huge difference regarding he performance of he codes. B. Spaial correlaion In he resuls of Figure 7- he wo ransmi channels for communicaion are compleely uncorrelaed. This is usually no he case during he RFID communicaion because of he close spacing of he ag anennas []. The diversiy gain decreases wih a higher correlaion of boh ransmi pahs. For maximal correlaed channels (ρ spa = ) no diversiy gain could be achieved compared o a SISO ransmission because boh channels produce he same informaion. Fig. shows he effec for a correlaed channel wih correlaed forward and backward links. Noe, ha he diversiy gain decreases by several db compared o an uncorrelaed link. Bu a diversiy gain can sill be achieved as long as he channels are no fully correlaed. For exising spaial correlaion he BER behavior during ransmission is shown in figures 2 and 3 are regarded. The general behavior is shown in figure 2 where BER for differen correlaion siuaion in a dyadic backscaer channel (K =.5) are shown for a FM encoded signal. Looking a a no fully correlaed channel (ρ,ρ spa < ) a gain beween 4 5 db can be achieved, comparing he BER of he (parially) uncorrelaed channels and a (parially) correlaion

7 GRZEGORZ SMIETNK, JÜRGEN GÖTZE: MODELING ND SIMULTION OF DIVERSITY FOR UHF RFID COMMUNICTION 89 BER FM SISO FM SISO FM FM SNR [E s /N ] SNR [E s /N ] [db] Fig.. Bi error rae for differen backscaer encoding schemes and ransmission wih one (SISO) or wo () ag anennas and one reader anenna in a Dyadic Backscaer Channel wih a correlaion of ρ =.5 for he forward and backward link. During he ransmission he wo links are correlaed wih a facor of ρ spa =.5. Boh links have a Rician disribuion wih K =.5. of ρ,ρ spa =.8. I is remarkable ha he gain beween a high correlaed and he fully correlaed channel is clearly higher hen beween an uncorrelaed and high correlaed channel. lso i is noed ha a variaion of he spaial correlaion achieves a lower gain han a variaion of he forward/backward correlaion. This leads o wo resuls when regarding he correlaion behavior. Firs, he influence of ρ is larger han he gain loss when he spaial correlaion rho spa increases. Second, he performance of he channel decrease wih increasing spaial and forward/backward correlaion, bu he main leak of performance happens when one of he correlaion values is very high ρ,ρ spa >.9. So even when he spaial correlaion is high a diversiy gain could be achieved. In figure 3 he behavior of his correlaion characerisic is analysed regarding he FM and he more robus Miller- 8 code. In his figure he SNR for BER= 4 is shown for differen channel consellaions over a dyadic backscaer channel (K =.5). The performance decreases as he correlaion coefficiens increase. For coefficiens smaller.6 he changes of he SNR are relaively small and have a nearly linear behavior. The differences become larger when one of he correlaion coefficiens is geing high, bu he larges performance loss occurs when one of he correlaion parameers is equal o one. The performance of he Miller- 8 encoded sequence has a similar behavior regarding he performance of he code. The differences in he gain are smaller and he sep beween a high and a fully correlaed channel is no ha large. The main reason is he more robus code compared o he FM encoded sequence. SNR BER [E S / N ] 6 [db] ? =..? =.2.2? =.4.4.6? =.6.8? =.8.9? =.9.? =. FM Miller-8? =. FM? SISO =. FM SISO spa? = Miller-2.8 Miller-2? SISO = SISO. spa? = Miller-4.? SISO Miller-8 spa =. SISO? = Miller-8. SISO FM? spa =.8 FM? =.8 Miller-2? Miller-2 spa =.8? = Miller-4.? Miller-8 spa =.8? = Miller-8.? spa =.? =.8? spa =.? =.? spa = ? 9 SNR SNR [E spa [E s /N s / /NN ] ] Fig. 3. SNR values a a BER= 4 for a ransmission ploed over he spaial correlaion ρ spa for differen encoding schemes. The ransmission is done over a Dyadic Backscaer channel wih LOS componen (K =.5). BER ? =. FM? SISO FM SISO =. spa? = Miller-2.8? SISO =. spa? = Miller-4.? SISO Miller-8 spa =. SISO? = Miller-8. SISO FM? spa =.8 FM? =.8 Miller-2? Miller-2 spa =.8? = Miller-4.? Miller-8 spa =.8? = Miller-8.? =. spa? =.8? spa =.? =.? =. spa = SNR SNR [E [E s /N s / /NN ] ] SNR [E s / N ] [db] Fig. 2. Bi error rae for FM encoded signal in a Dyadic Backscaer channel wih LOS componen (K =.5), a ransmission wih wo () ag anennas and differen forward/backward and spaial correlaions. spaial correlaion of ρ spa = is equivalen o a ransmission wih one (SISO) ag anenna. VII. CONCLUSION The poenial of UHF RFID backscaering ransmission using he EPCglobal sandard was invesigaed. I was shown ha he diversiy gain beween a SISO and a communicaion is a leas db for he mos robus encoding scheme in an one way channel. The gain increases for a higher scaering of he channel. For a ypical RFID communicaion wih a wo way channel he gain increases because of deeper fades in his channel scenario. The highes gain is achieved wih a fully correlaed forward and backward link where he diversiy gain is over 4 db for a FM encoded se of daa a a = 2. s he ypical RFID small scale fading link has he characerisic of a Dyadic Backscaer Channel and he correlaion beween he forward and backward link can be relaively high especially when he ransmier and receiver anenna of he reader is he same. For his case he high poenial of ags wih muliple anennas in RFID sysems was demonsraed. Bu i is also o menion ha he diversiy gain decreases if he correlaion of he muliple ransmission

8 82 PROCEEDINGS OF THE FEDCSIS. WROCŁW, 22 pahs increases. Neverheless significan gains during a ransmission can be achieved when wo ransmission channels are no fully correlaed, even when heir correlaion is relaivly high. CKNOWLEDGMENT The repored R+D work was carried ou in he frame of he BMBF-Projec smarti (Smar ReUsable Transpor Iems); he smarti projec is carried ou in he frame of he Efficiency Cluser Logisic Ruhr (par of he Leading-Edge Cluser - High-Tech Sraegy for Germany). This paricular research was suppored by he BMBF (Bundesminiserium fuer Bildung und Forschung) of Federal Republic of Germany under gran ICLH (Sichere, flexible Deekion und Lokalisierung von UHF RFID-Labeln in rauer Umgebung). The responsibiliy for his publicaion is held by he auhors only. In paricular we have o hank our indusrial parners Infineon Technologies G and Deusche Pos G supporing us by he required informaion regarding he RFID es environmen and ag daa.the auhors would also like o hank he R+D Coordinaor of smarti pplicaion Scenario POST Dr. W. John (SIL GmbH (ig)) for his ongoing suppor regarding he RFID challenges of he posal logisic chain and he relaed indusrial R+D requiremens. REFERENCES [] H. Sockman, Communicaion by means of refleced power, in Proceedings of he IRE, vol. 36, no., pp , Ok. 948 [2] J.D. Griffin and G.D Durgin, Link Envelop Correlaion in he Backscaer Channel in IEEE Communicaions Leers, vol., pp , Sep. 27 [3] M. Buener, D. Weherall, Flexible Sofware Radio Transceiver for UHF RFID Experimenaion, UW CSE Technical Repor, 29 [4] EPC Radio-Frequency Ideniy Proocols, Class- Generaion-2 UHF RFID Proocol for Communicaions a 86 MHz- 96 MHz, version..9, EPC Global Jan. 25. [5] S.M. lamoui, Simple Transmi Diversiy Technique for Wireless Communicaions in IEEE Journal on Sleced reas in Communicaions, vol. 6, nr. 8, pp , July 999 [6] The Comprehensive GNU Radio rchive Nework [Online]. vailable a hps:// (accessed Nov. 2) [7] W. Zhang and M.J. Miller, Baseband Equivalens in Digial Communicaion Sysem Simulaion in IEEE Transacions on Educaions, vol. 35, issue: 4, pp , Nov. 992 [8] H. Nuszkowski, Digiale Signalüberragung im Mobilfunk, s ed, Vog Verlag, pp. 29-3, 2. [9] H. Taricco and G. Coluccia, Opimum Receiver Design for Correlaed Rician Fading MIMO Channels wih Pilo-ided Deecion in IEEE Journal on Seleced reas in Communicaions, vol. 25, issue 7, pp. 3-32, Sep. 27 [] K. Vanganuru and. nnamalai, nalysis of ransmi diversiy schemes: impac of fade disribuion, spaial correlaion and channel esimaion errors in Wireless Communicaions and Neworking, vol., pp , May 23 [] Chen He and Z. Jane Wang, Gains by a space-ime-code based signaling scheme for muliple-anenna RFID ags in 23rd Canadian Conference on Elecrical and Compuer Engineering (CCECE), pp. -4, Sep. 2 [2] J.D. Griffin and G.D Durgin, Gains For RF Tags Using Muliple nennas in IEEE Transacions on nennas and Propagaion, vol. 56, nr. 2, pp , Feb. 28