Multipth Fding Mesurements for Multi-Antenn Bcksctter RFID t 5.8 GHz Joshu D. Griffin 1 nd Gregory D. Durgin 2 School of Electricl nd Computer Engineering, Georgi Institute of Technology 777 Atlntic Dr., Atlnt, Georgi 30332 0250 Emil 1 : jdgriffin@ieee.org, Emil 2 : durgin@ece.gtech.edu Abstrct UHF nd microwve bcksctter RF-tg systems, including rdio frequency identifiction (RFID) nd sensor systems, experience multipth fding tht cn be more severe thn tht found in conventionl trnsmitter-to-receiver chnnel. Previous work hs shown tht multipth fding cn be reduced on the modulted-bcksctter signl received from the RF tg in non-line-of-sight (NLOS) chnnel if more thn one RF-tg ntenn is used to modulte bcksctter. This pper presents the first multipth fding mesurements for bcksctter tgs using multiple ntenns t 5.79 GHz the center of the 5.725-5.850 GHz, unlicensed industril, scientific, nd medicl (ISM) frequency bnd tht my offer relible opertion for future, miniture RF tgs. NLOS mesurement results re presented s cumultive density functions (CDF) nd fde mrgins for use in bcksctter rdio link budget nlysis nd detiled description of the custom bcksctter testbed used to tke the mesurements is provided. The mesurements show tht gins re vilble for multiple-ntenn RF tgs nd results mtch well with gins predicted using the nlytic fding distributions derived previously. I. INTRODUCTION The potentil of bcksctter rdio for use in rdio frequency identifiction (RFID) nd sensor pplictions is gret nd so re the chllenges fced in designing relible, low-cost bcksctter systems with dequte rnge. At the physicl level, the rnge nd relibility of the bcksctter rdio system is limited by the power consumption of the RF trnsponder, or RF tg; polriztion mismtch losses; object ttchment losses resulting from the impednce mismtch nd ntenn gin reduction cused by the RF tg s close proximity or ttchment to dielectric or conductive mterils; smll-scle fding loss; nd losses cused by blockges to the line-ofsight (LOS) between the reder nd RF tg. Smll-scle fding lone cn cuse significnt reductions in rnge nd relibility nd is most pronounced on the modulted-bcksctter signl received t the reder. The fding on this signl often follows product-ricin distribution resulting in deeper fdes thn those found on the signl received by the RF tg [1]. One wy to reduce fding in the bcksctter chnnel is through ntenn diversity which uses multiple ntenns t the reder nd RF tg to provide sptilly-seprted diversity The work reported in this pper ws supported in prt by the Ntionl Science Foundtion (NSF) CAREER Grnt #0546955. J. D. Griffin performed this work while PhD student t Georgi Tech nd is now with Disney Reserch Pittsburgh, 4615 Forbes Ave., Pittsburgh, PA 15213. brnches. This technique ws first explored for bcksctter rdio by Ingrm et l. [2] nd others [3], [4] hve used multiple ntenns t the reder for this purpose. Multiple ntenns cn lso be used on the RF tg nd it hs been shown tht modulting bcksctter with multiple, sptillyseprted RF-tg ntenns cn reduce smll-scle fding on the modulted-bcksctter signl [5]. However, no mesurement cmpigns hve been reported tht investigte smllscle fding with multiple-ntenn RF tgs nd only few hve studied smll-scle fding in the bcksctter chnnel Kim et l. [1] mde mesurements of the bcksctter chnnel reporting envelope cumultive distribution functions (CDFs) nd pth loss mesurements t 2.4 GHz nd Bnerjee et l. [6], [7] hve presented fding mesurements s well s sptil nd frequency diversity gin mesurements t 915 MHz. This pper presents smll-scle fding mesurements for RF tgs with one nd two ntenns t 5.79 GHz. This frequency ws chosen becuse it is in the unlicensed, 5.725-5.850 GHz industril, scientific, nd medicl (ISM) frequency bnd vilble for bcksctter rdio pplictions. This frequency bnd hs severl potentil dvntges for bcksctter rdio systems including reduced ntenn size, incresed ntenn gin, reduced object ttchment losses [5], [8] nd hs been used for t lest one pssive bcksctter rdio systems [9]. In the following section, brief overview of the M L N, dydic bcksctter chnnel is given followed by detiled description of the testbed used to tke the fding mesurements. Afterwrds, the mesurement procedure nd clibrtion technique re discussed. Finlly, the mesurement results re presented in terms of envelope CDFs nd fde mrgins for use in bcksctter rdio link budgets. The mesured CDFs re compred to nlytic distributions for the M L N, dydic bcksctter chnnel nd reduction in multipth fding is shown. II. THE M L N DYADIC BACKSCATTER CHANNEL Before proceeding, it is useful to briefly outline the M L N, dydic bcksctter chnnel nd the probbility density functions (PDF) tht describe smll-scle fding under nonline-of-sight (NLOS) conditions. The M L N, dydic bcksctter chnnel is composed of forwrd link tht describes signl propgtion from the M reder trnsmitter ntenns to the L RF-tg ntenns nd bcksctter link tht describes propgtion from the L RF-tg ntenns to the N
reder receiver ntenns. This chnnel is pinhole chnnel [10] in which ech RF-tg ntenn cts s pinhole through which signls propgte. As more pinholes re dded to the chnnel, fding on the signl received t the n th reder ntenn decreses, especilly when the forwrd nd bcksctter links experience Ryleigh fding. This fding improvement, or pinhole diversity gin, rises from the fct tht the envelope PDF of the signl received t the n th reder receiver ntenn chnges shpe s the number of RF-tg ntenns is incresed. The envelope PDF f A (α, ρ) for the M L N bcksctter chnnel with Ryleigh-fding forwrd nd bcksctter links is given by the following two equtions: ( ) 1+L f A (α, ρ =0)=α L 2 Mσb σ f ( 21 L Γ(L) K 2α (1 L) ), (1) Mσb σ f where α is the chnnel envelope, Γ( ) is the gmm function, σ b nd σ f re the vrinces of the forwrd nd bcksctter links, nd K ν ( ) is modified bessel function of the second kind with order ν =1 L. The second is ( ) 1+L/2 f A (α, ρ =1)=α L/2 1 σ b σ f M ( ) 21 L/2 Γ ( α ) K L (1 L/2), (2) 2 σ b σ f M where K ν ( ) is modified bessel function of the second kind with order ν =1 L/2 nd ll other terms re s defined for (1). In these PDFs, ρ denotes link correltion, the sttisticl correltion between the fding on the forwrd nd bcksctter links which cn extend over the rnge 1 ρ 1. Detils of these topics my be found in [5], [11]. III. THE BACKSCATTER TESTBED A custom bcksctter testbed ws designed, prototyped, nd used to tke the fding mesurements reported in this pper. The following sections present the design rtionle nd overview of the testbed components. A. Testbed Overview A simplified block digrm of the bisttic bcksctter testbed, which ws composed of combintion of lbortory bench nd custom equipment, is shown in Fig. 1. The mjor components of the testbed included n Agilent E8247C signl genertor to provide the continuous wve (CW) trnsmitted signl nd locl oscilltor (LO) for the receivers; two custom, coherent, direct-conversion receivers whose signls were smpled by nlog-to-digitl conversion bords housed in personl computer; two custom RF tgs whose modultion signl ws provided by two Agilent 33250A function genertors; custom reder trnsmitter nd receiver ntenns; nd screw-drive liner positioner mde by Velmex for positioning the RF tgs. Further detils of the custom-designed equipment is provided below. 1) Reder Antenns: Linerly-polrized ptch ntenns, shown in Fig. 2, were used t the reder trnsmitter nd receiver. These ntenns were designed for 5.79 GHz on n FR4 substrte nd ech hd brodside gin of pproximtely 3.8 dbi. Custom ntenns were designed becuse RFID reder ntenns for the 5.725-5.850 GHz frequency bnd re not commonly vilble. 2) RF Tgs: Two RF tgs were designed nd prototyped for this mesurement cmpign. The first ws single ntenn tg (STAG) nd the second ws dul ntenn tg (DTAG) whose block digrms re shown in Fig. 3. The following fctors motivted their design: Equl Comprison: For fir comprison of fding with the STAG nd DTAG, the two RF tgs were designed with the sme type of ntenn nd the ntenns of the DTAG were mde on the sme substrte. Flexibility: The STAG nd DTAG designs llowed the bckscttered signl to be modulted with n rbitrry bit sequence i.e. ny mplitude shift keying (ASK) wveform. In this reliztion of the testbed, differentil bit-sequence ws output from two Agilent 33250A function genertors; however, ny source of differentil binry signls could be used. Decoupling: Creful design ws required to decouple the closely spced ntenns of the DTAG. This ws ccomplished using orthogonl DTAG ntenns which were ech impednce mtched to the 50 Ω switch using Ansoft HFSS, 3D electromgnetic nd microwve softwre pckge. Ech RF tg used 5.79 GHz slot ntenn (the DTAG used two such ntenns) whose lod ws switched between n open nd short circuit (shorted through DC block) by microwve GAs PHEMT switch (M/A-Com MASW-007107 V2). Ech slot ntenn ws connected to the microwve switch through n ungrounded, coplnr-wveguide (CPW) trnsmission line, mtching section, nd DC block s shown in Fig. 4. The switch ws toggled using two digitl control lines referenced to third ground line. One potentil problem with the STAG nd DTAG designs ws tht the control lines from the signl source could ct s n ntenn nd contribute to the modultedbcksctter signl. To mesure the contribution from the control lines, the STAG nd DTAG ntenn ptterns were mesured with nd without their ntenns shorted. When the ntenns were shorted with copper tpe, the mesured pttern ws composed of only signls bckscttered from the control lines or other unwnted, time-vrying sctterers. All of the pttern mesurements showed tht the bcksctter modulted by the control lines ws much smller thn tht from the unshorted slot ntenns. In generl, the desired signl ws t lest 20 db greter thn tht from the control lines nd the two only becme comprble ner the nulls of the un-shorted slot ntenn ptterns. The STAG nd DTAG ntenn ptterns were dipole-like in shpe. 3) Direct-Conversion Receiver: Two custom, directconversion receivers, shown in Fig. 5, were designed nd prototyped. The receivers downconverted signls from the 5.725-
Fig. 1. A simplified block digrm of the bcksctter testbed used in this mesurement cmpign. () Fig. 2. The 5.79 GHz, linerly-polrized ptch ntenns used t the testbed trnsmitter nd receiver. 5.850 GHz ISM bnd to bsebnd in-phse (I) nd qudrture (Q) signls tht were smpled externlly. The design of the receivers ws motivted by the following: Self-interference Mitigtion: All bcksctter rdio receivers must be ble to receive the strong, unmodulted crrier i.e. self interference trnsmitted from the reder while still detecting the much weker modulted bcksctter from the RF tg. These custom receivers were ble to receive -12 dbm self-interference signl while still receiving the smll modulted-bcksctter signl. The sensitivity nd dynmic rnge of the testbed is discussed in the lst prgrph of Section IV. The custom receivers lso blocked the self-interference signl fter down-converting it to DC which prevented the signl from sturting the bsebnd mplifiction stge. Coherent reception: Coherent reception ws required becuse it is possible for the envelope of the totl received signl i.e. the unmodulted crrier plus the modultedbcksctter signl to remin constnt s the tg switches between its modultion sttes. Since the phse of the Fig. 3. Block digrms of the () STAG nd DTAG. modulted bcksctter is not fixed with respect to the phse of the crrier, there re lwys two possible tg modultion sttes tht will result in the sme envelope of the totl received signl. In such situtions, simple envelope detection would not be ble to differentite between these sttes.
Fig. 5. The custom, coherent, 5.725-5.850 GHz direct-conversion receiver. () Fig. 4. The () STAG nd DTAG showing the 5.79 GHz slot ntenn, CPW trnsmission line, mtching section, nd microwve microwve switch. The tgs were etched on 62-mil, FR4 substrte. IV. NLOS MEASUREMENT PROCEDURE The gol of this mesurement cmpign ws to determine the envelope distribution of the M L N, dydic bcksctter chnnel s function of RF-tg position t 5.79 GHz (λ o 5.2 cm). Fding mesurements of two chnnels re reported: Bisttic 1 1 2 Chnnel: In this chnnel, the reder trnsmitter ntenn ws seprted from the two receiver ntenns by 6.5λ o for RX 1 nd 10λ o for RX 2. The STAG ws used to modulte bcksctter. Bisttic 1 2 2 Chnnel: This chnnel used the sme testbed configurtion described bove, but the DTAG ws used to modulte the bckscttered signl. These NLOS mesurements were through-wll i.e. the bcksctter testbed reder ws locted in room 560 of the Vn Leer Building on the Georgi Institute of Technology cmpus nd the RF tg ws locted in room 558 of the sme building s shown in Fig. 6 nd Fig. 7. The LOS ws blocked by both the sheet-rock wll nd lrge, metllic sheet (ctully, metllic stripline cvity) for the purpose of creting rich scttering environment. For ech mesurement, n unmodulted, 5.79 GHz crrier ws trnsmitted nd scttered by the RF tg. The RF tg modulted the bcksctter using 31- bit, mximl-length pseudo-rndom code (m-sequence) [12] with chip rte of 1 MHz. The modulted-bcksctter signl ws received by the two direct-conversion receivers discussed previously nd the I nd Q bsebnd signls were digitized nd stored for processing. The liner positioner moved the RF tg through squre grid tht ws 30 cm 30 cm (pproximtely 6λ o 6λ o ) nd mesurement ws tken every 1 cm (pproximtely every λ o /5). This high sptil smpling rte ws necessry becuse the sptil Nyquist rte for bcksctter chnnel is twice tht of conventionl chnnel. This is shown in Fig. 8 where the totl pth length from the reder trnsmitter to the RF tg nd bck to the reder receiver r f + r b is proportionl to 2 r tg, r f + r b = 2 r tg r tx r rx. (3) Prior to ech mesurement, the testbed ws clibrted by plcing the RF tg 40.5 cm from the reder ntenns nd recording the received signl. The clibrtion mesurements were performed indoors nd wy from sctterers, were polriztion mtched, nd the tgs were close enough to the reder ntenns tht ny multipth signls were not significnt. A digrm of the clibrtion setup for the bisttic testbed is shown in Fig. 9. All STAG nd DTAG mesurements re reported reltive to their respective clibrtions. In these mesurements, it ws importnt to ensure tht noise did not dversely ffect the mesured signl, prticulrly for the smll signls mesured in deep fde. If the mesured signl ws dominted by noise, which cn be ssumed to
Fig. 6. The bisttic mesurement setup between rooms E560 nd E558 of the Vn Leer Building on the Georgi Institute of Technology min cmpus. Coherent chnnel smples were tken t 5.79 GHz s function of RF-tg position cross the ornge shded squre. The RF tgs were pproximtely 86 cm bove the floor nd chnnel smple ws recorded every 1 cm (pproximtely every λ o/5). Fig. 8. The reltionship between the trnsmitter-to-tg-to-receiver distnce, r f + r b, nd the position of the receiver, trnsmitter, nd RF tg given respectively by r rx, r tx, nd r tg. Eqution (3) shows tht r f + r b is proportionl to 2 r tg. Fig. 7. The testbed setup in room 560 (see Fig. 6) for the through-wll, NLOS mesurements. The trnsmitter nd receiver ntenns were mounted on plstic pole nd the direct-conversion receivers nd signl sources were locted on the tble. Although shown higher in this photo, the trnsmitter nd receiver ntenns were 86 cm bove the floor for the mesurements. follow Gussin rndom process, then the resulting envelope distribution would pper to be Ryleigh nd hrd to distinguish from the expected product-ryleigh distribution. Therefore, the mesured dt ws compred to the liner-scle men of the noise power which ws mesured by operting the testbed with no RF-tg modultion. Only mesurements tht were 20 db bove the liner-scle men noise power 1 re reported. If the mesured envelope is α mes = α true ± α noise, then the percent error cused by noise cn be defined s Percent Error = 100 α mes α true, (4) α true where it is ssumed tht α true > α noise. A 20 db difference between the noise power nd the true signl power results in 1 The noise power ws clculted reltive to the STAG nd DTAG clibrtions.
Fig. 9. The clibrtion setup for the bisttic mesurements. () ±10% envelope error. Time verging ws used to lower the noise floor of the testbed nd, hence, increse its sensitivity. Thirty-two time verges were used resulting in noise floor, clculted from the linerly-verged noise output from the two receivers, of -142 dbm. Since the mximum input power of the receiver is -12 dbm, the useful dynmic rnge of the testbed is 110 db. V. MEASUREMENT RESULTS Bcksctter-fding mesurement results re presented in terms of envelope cumultive distribution functions (CDF) to explore pinhole diversity gins nd fde mrgins for use in bcksctter rdio link budgets [8]. The mesured CDFs re compred to CDFs clculted from (1) nd (2) evluted for severl different chnnel configurtions. As mentioned previously, ll of the reported mesurements re normlized by the clibrtion mesurements. Fig. 10. The () STAG nd DTAG mesured power in db reltive to the mximum in the NLOS bcksctter chnnel. The orienttion of these figures mtches tht for the mesurement digrm shown in Fig. 6. The coloring of ech squre represents the mesured power of the chnnel t tht RF-tg position. A. NLOS Sptil Fding Plots Before delving into the NLOS envelope distributions, it is useful to exmine the chnnel smples plotted s function of RF-tg position. Ech squre in Fig. 10 shows the mesured chnnel power (in db) for n RF-tg position, reltive to the clibrtion mesurements. The power is normlized to the mximum power recorded in ech mesurement so tht the mgnitude of ech fde cn be esily seen. Both the STAG nd DTAG plots show deep, rpid fdes of up to 40 db. This high multipth chnnel consists of wves diffrcting round the metllic sheet s well s wves scttered from the drill press, bookcses, nd other smll clutter not shown in Fig. 6. For reference, the mximum power received in the STAG nd DTAG mesurements ws -82 dbm. Note tht this bsolute power vlue is not normlized by the clibrtion mesurements. B. Mesured Estimtes of the NLOS, Bcksctter-Chnnel CDFs Mesured estimtes of the NLOS, bcksctter-chnnel envelope distributions for the STAG nd DTAG re shown in Fig. 11 long with the corresponding nlytic CDFs clculted from (1) nd (2). Visul inspection revels tht the mesured estimtes of the NLOS CDF mtch the nlytic distributions very well. This observtion is confirmed using two error mesurements [1]. The first error mesurement is the liner men-squre-error (MSE), simply clculted s the verge squred difference between the mesured nd derived CDFs. The second is the logrithmic MSE,
TABLE I COMPARISON OF THE MEASURED ESTIMATES OF THE CDF FOR THE STAG AND DTAG IN THE NLOS, BISTATIC CHANNEL WITH THE 1 1 1, 1 2 1, 1 3 1, AND 1 4 1 DERIVED DISTRIBUTIONS FOR ρ =0(FROM (1)) AND ρ =1(FROM (2)). Mesured CDF Liner MSE Log MSE Mtch Mtch MMSE MMSE L ρ L ρ STAG RX 1 1.57E-4 1 0 9.99E-2 1 0 STAG RX 2 9.85E-5 1 0 1.13E-2 1 0 DTAG RX 1 7.22E-5 3 0 8.92E-2 2 0 DTAG RX 2 4.12E-5 2 0 9.62E-2 2 0 () Fig. 11. The CDF of the bisttic () STAG nd DTAG mesurements in the NLOS bcksctter chnnel long with the corresponding 1 1 1 nd 1 2 1 CDFs clculted from (1). The CDFs re plotted on xes normlized by the root of the power of ech distribution P for unbised comprisons. Log MSE = 1 N N [ [10 log 10 Fm (α i / P ) ] i=1 10 log 10 [ F (αi / P ) ]] 2, (5) where F m ( ) is the mesured estimte of the fding CDF, F ( ) is the nlyticl CDF, N is the totl number of mesured dt points, nd α i / P is the i th mesured envelope normlized by the verge power of the fding distribution. The verge power of the chnnel distribution is defined s P = α 2 f 0 A (α)dα. The logrithmic MSE is useful becuse it emphsizes the difference between the mesured nd nlytic CDFs for smll envelope vlues the region of most concern for bcksctter rdio designers. These error mesurements re presented in Tble I where the STAG nd DTAG mesured CDF estimtes re compred to the CDFs derived from (1) nd (2) for the 1 1 1, 1 2 1, 1 3 1, nd 1 4 1 chnnels (8 nlytic CDFs in ll). For the STAG mesurements t RX 1 nd RX 2, the nlytic 1 1 1 CDF, clculted from (1), minimizes the men-squre error. For the DTAG mesurements, the 1 2 1 CDF from (1) does the sme, except tht the liner MSE indictes the best mtch t RX 1 is the 1 3 1 CDF from (1). In this cse, however, since the CDFs for the 1 L 1 chnnels become very similr for lrge envelopes, the logrithmic MSE is more meningful comprison nd the best mtch is likely the 1 2 1 distribution from (1). C. Pinhole Diversity in the NLOS Bcksctter Chnnel Comprison of the mesured estimtes of the STAG nd DTAG CDFs revels tht pinhole diversity gin exists in this NLOS, bisttic chnnel, s Fig. 12 shows. The pinhole diversity gin is evidenced by the fct tht the STAG CDF is higher thn the DTAG CDF for normlized envelope vlues below pproximtely 0.8. Pinhole diversity gins cn lso be seen by exmining the fde mrgins clculted from the mesured CDF estimtes. The fde mrgin is defined s [[ F 1 A Fde Mrgin =10log (Outge Probbility)] ] 2 10 (6) P where F A ( ) is the mesured CDF estimte nd P is the verge power of the distribution [8]. The outge probbility is the likelihood tht the power received t the reder receiver P R hs fded below P by n mount equl to the fde mrgin, Outge Probbility = Pr[P R P/(Fde Mrgin)] [8]. Tble II shows tht the fde mrgin required to mintin given outge probbility is reduced for the DTAG compred to the STAG t ech receiver. Furthermore, the fde mrgins clculted from the mesured distribution estimtes mtch those clculted from (1) for the 1 1 1 nd 1 2 1 chnnels well. D. Discussion The NLOS mesured CDF estimtes gree well with the M L N distributions presented in Section II. These mesurements show tht pinhole diversity gins occur nd tht (1) is ccurte for the 1 1 1 nd 1 2 1 chnnels in bisttic, NLOS chnnels with rich multipth. The greement between the derived nd mesured distribution estimtes lso indictes tht this chnnel hs very smll link correltion ρ due to the wide seprtion between the reder receiver nd trnsmitter
less often thn in indoor bcksctter chnnels becuse sctterers my be locted frther from the bcksctter rdio system; however, such fding distribution is certinly possible in n outdoor environment. () VI. CONCLUSION These 5.79 GHz mesurements show tht fdes of up to 40 db re present on the modulted-bcksctter signl received from n RF tg in the NLOS, M L N, dydic bcksctter chnnel. However, fding cn be reduced by modulting bcksctter with more thn one RF-tg ntenn pinhole diversity gin. Mesurements were presented in terms of CDFs for comprison with previously derived distributions nd fde mrgins for use in bcksctter rdio link budgets. The mesured CDF estimtes showed excellent greement with the nlytic distributions derived for the M L N, Ryleighfding chnnel. Likewise, the fde mrgins were improved by severl db for the DTAG compred to the STAG. ACKNOWLEDGMENT The uthors would like give specil thnks to Ryn Pirkl for his expert hrdwre nd dt processing dvice nd Joey Duvll for her ssistnce with the mesurements. REFERENCES Fig. 12. The STAG nd DTAG CDFs mesured t () RX 1 nd RX 2 plotted on the log-log scle. The CDFs re plotted on xes normlized by the root of the power of ech distribution P for unbised comprisons. TABLE II COMPARISON OF THE FADE MARGINS (IN DB) CALCULATED FROM THE MEASURED CDF ESTIMATES AND THE ANALYTICAL DISTRIBUTION GIVEN BY (1). Outge STAG Eqn. (1) DTAG Eqn. (1) Probbility RX 1/2 with L =1 RX 1/2 with L =2 0.5 3.4/4.1 4.1 2.6/2.9 2.9 0.1 14/16 15 12/12 12 0.05 18/20 20 15/14 16 0.01 24/27 28 22/23 24 0.005 29/29 32 25/26 26 ntenns. The RF-tg ntenns re lso lrgely uncorrelted becuse of the hevy multipth scttering in the chnnel. The CDF estimtes were clculted from indoor multipth fding mesurements; however, they cn be pplied to ny bcksctter chnnel in which the fding follows product- Ryleigh distribution. In other words, these distributions pply to bcksctter chnnels tht do not hve dominnt speculr wve, but re composed of mny multipth wves. Product- Ryleigh fding in outdoor bcksctter chnnels my occur [1] D. Kim, M. A. Ingrm, nd W. W. Smith, Jr., Mesurements of Smll-scle Fding nd Pth Loss for Long Rnge RF Tgs, IEEE Trnsctions on Antenns nd Propgtion, vol. 51, no. 8, pp. 1740 1749, 2003. [2] M. Ingrm, M. Demirkol, nd D. Kim, Trnsmit Diversity nd Sptil Multiplexing for RF Links Using Modulted Bcksctter, in Proceedings of the Interntionl Symposium on Signls, Systems, nd Electronics, Tokyo, Jpn, July 2001. [3] J. S. Kim, K. H. Shin, S. M. Prk, W. K. Choi, nd N. S. Seong, Polriztion nd Spce Diversity Antenn Using Inverted-F Antenns for RFID Reder Applictions, Antenns nd Wireless Propgtion Letters, vol. 5, no. 1, pp. 265 268, 2006. [4] A. Rhmti, Z. Lin, M. Hiltunen, nd R. Jn, Relibility Techniques for RFID-Bsed Object Trcking Applictions, in 37th Annul IEEE/IFIP Interntionl Conference on Dependble Systems nd Networks (DSN 07), Edinburgh, UK, 2007, pp. 113 118. [5] J. D. Griffin nd G. D. Durgin, Gins for RF Tgs Using Multiple Antenns, IEEE Trnsctions on Antenns nd Propgtion, vol. 56, no. 2, pp. 563 570, 2008. [6] S. R. Bnerjee, R. Jesme, nd R. A. Sinti, Performnce Anlysis of Short Rnge UHF Propgtion s Applicble to Pssive RFID, in 2007 IEEE Interntionl Conference on RFID, Gylord Texn Resort, Grpevine, TX, USA, Mrch 2007, pp. 30 36. [7], Investigtion of Sptil nd Frequency Diversity for Long Rnge UHF RFID, in IEEE Antenns nd Propgtion Society Interntionl Symposium, Sn Diego, CA, USA, July 2008, pp. 1 4. [8] J. D. Griffin nd G. D. Durgin, Complete Link Budgets for Bcksctter Rdio nd RFID Systems, IEEE Antenns nd Propgtion Mgzine, ccepted for April, 2009. [9] B. Strssner nd K. Chng, Pssive 5.8-GHz Rdio-Frequency Identifiction Tg for Monitoring Oil Drill Pipe, IEEE Trnsctions on Microwve Theory nd Techniques, vol. 51, no. 2, pp. 356 363, Feb. 2003. [10] D. Chizhik, G. J. Foschini, nd R. A. Vlenzuel, Cpcities of Multi- Element Trnsmit nd Receive Antenns: Correltions nd Keyholes, Electronics Letters, vol. 36, no. 13, pp. 1099 1100, 2000. [11] J. D. Griffin nd G. D. Durgin, Link Envelope Correltion in the Bcksctter Chnnel, IEEE Communictions Letters, vol. 11, no. 9, pp. 735 737, 2007. [12] R. W. Dixon, Spred Spectrum Systems with Commercil Applictions, 3rd ed. New York: Wiley Interscience, 1994.