PERFORMANCE EVALUATION OF A SPECTRUM-SENSING TECHNIQUE FOR LDACS AND JTIDS COEXISTENCE IN L-BAND

Transcription

1 Proceedngs of SDR'1-WInnComm-Europe, 7-9 June 01 PERFORMANCE EVALUATION OF A SPECTRUM-SENSING TECHNIQUE FOR LDACS AND JTIDS COEXISTENCE IN L-BAND Gulo Bartol (Unverstà degl Stud d Frenze, Florence, Italy; gulo.bartol@unf.t); Romano Fantacc (Unverstà degl Stud d Frenze, Florence, Italy; romano.fantacc@unf.t); Dana Marabss (Unverstà degl Stud d Frenze, Florence, Italy; dana.marabss@unf.t); Luga Mccullo (Unverstà degl Stud d Frenze, Florence, Italy; luga.mccullo@unf.t); Claudo Arman (SELEX Elsag S.p.A.; claudo.arman@selexelsag.com); Roberto Merlo (SelexElsag S.p.A.; roberto.merlo@selexelsag.com). ABSTRACT Ths paper deals wth a cogntve approach able to guarantee the coexstence of new data lnk for ar-ground aeronautcal communcatons LDACS and mltary JTIDS systems. Future LDACS shall coexst wth current systems operatng n the same frequency band for ths reason coexstence ssues must be carefully nvestgated. In partcular JTIDS transmssons can affect the LDACS performance actng as dsruptve mpulse nose. JTIDS explots frequency hoppng to protect nformaton, hence ts nterference on LDACS system cannot be foreseen and avoded. In addton the bandwdth of the two sgnals results to be completely overlapped n case of collson. The dsruptve effects of JTIDS nterference on LDACS can be mtgated f the collsons can be detected and hence sutable processng technques can be actvated. Ths paper proposes a method to detect the presence of JTIDS nterference explotng an energy detecton spectrum sensng technque based on sldng wndows and packets retransmsson. The performance of the proposed method s presented n terms of mssed detecton probablty of the JTIDS nterference and error rate of the LDACS system showng a sgnfcant capablty to counteract JTIDS nterference. 1. INTRODUCTION The ncreasng demand for advanced communcaton servces n cvl avaton leads to the need for a new management and communcaton framework able to support the capacty and securty requrements of the ar transportaton system [1]. The European project SESAR (Sngle European Sky ATM Research) ams to develop and valdate a new communcaton system capable of satsfy the requrements specfed n Communcaton Operatng Concept and Requrements (COCR) document []. Snce the COCR operatonal requrements cannot be fulflled by a sngle technology, the Future Communcaton Infrastructure (FCI), consttutng the communcaton part of the framework, wll be mplemented as a system of systems, ntegratng exstng as well as new communcaton technologes. The FCI should support both dgtal voce and data communcatons. Partcular emphass s gven to the data, snce n case of falure, voce would not be able to mantan the operatons wth the same relablty. Snce the VHF frequences are congested, the communcaton system n charge of supportng the ar/ground data lnk wll operate on the aeronautcal L-Band ( MHz) and t wll be named L-Band Dgtal Communcaton System (LDACS). Ths technology s currently under development and, at the present tme, two optons have been dentfed. LDACS1 s the former opton. It s a Frequency Dvson Duplex (FDD) system explotng the OFDM (Orthogonal Frequency Dvson Multplex) technque, that s very effectve aganst the nter symbols nterference. The latter opton, LDACS, s a Tme Dvson Duplex (TDD) system utlzng the CPFSK (Contnuous- Phase Frequency-Shft Keyng) modulaton, that allows to reduce the out-of-band emssons. At the end of the SESAR program studes one of them wll be selected as the key technology for ar/ground communcatons. Ths work focuses on LDACS1. LDACS wll operate n the L-Band where several legacy systems are already present (e.g., DME, SSR, UAT, JTIDS/MIDS), hence the spectral compatblty s an mportant ssue that needs to be addressed. In partcular n ths paper we consder the coexstence between LDACS1 and the Jont Tactcal Informaton Dstrbuton System (JTIDS), also known as Mult-functonal Informaton Dstrbuton System n the NATO mplementaton. JTIDS s mltary system used for several purposes, lke dentfcaton n survellance and msson management. It explots an mpulsve sgnal and frequency hoppng, n order to make the system nterference-tolerant. 01 The Software Defned Rado Forum, Inc.-All Rghts Reserved 17

2 Unfortunately the frequency hoppng s performed on a large range of frequences spannng almost the whole L- Band, hence the probablty to have collsons between the LDACS and the JTIDS sgnals s very hgh. Snce LDACS s an OFDM system and the decodng s performed n the frequency doman, the mpulsve nose affects all the bts carred by the nterfered OFDM symbol. Ths represents an advantage untl the power of the nterference does not exceed a certan threshold [3], but t becomes very dsruptve snce leads to hgh error probablty on each OFDM subcarrer. The problem of mpulsve nterference n OFDM system s a known problem n Power Lne Communcaton (PLC), where the man-made nose, as the turnng on/off of an electrcal swtch, can deterorate the performance. Ths topc has only recently been nvestgated n wreless networks context, where the co-channel nterference can assume an mpulsve nature. The solutons avalable n the lterature deal wth non-lnear elaboratons of the nterfered sgnal through clppng or blankng schemes [4]-[6]: when the receved sgnal ampltude exceeds a certan threshold the sgnal level s set to the threshold value or to zero, respectvely. Another nterestng soluton reles on a close loop scheme based on data detecton and successve nose estmaton and reducton [7]. Unfortunately ths scheme has a hgh computatonal cost. However, none of these methods takes under consderaton any advanced technque to dstngush whch samples are affected by nterference. Coexstence between systems through spectrum sensng s a typcal topc of the Cogntve Rado networks, where an unlcensed (Secondary) system coexsts wth a lcensed (Prmary) system n a transparent manner: the Secondary user s rado dentfes the free frequency channels for secondary usage and transmts n these frequences n a nonnterferng manner. The scenaro consdered n ths paper s dfferent: sensng operaton s used to detect the presence of the JTIDS sgnal and then to perform sutable operatons at the recever sde. In partcular the proposed scheme s based on energy detecton and packets retransmsson. The energy detector s a well-known sensng algorthm that allows the detecton of nterferng sgnals wth low complexty and works well when the nterference features are not known. In our scheme sensng operaton s combned wth packets retransmsson and s performed by explotng two sgnal replcas n order to have a more effcent detecton. Furthermore the replcas are combned to mprove the sgnal detecton after an nterference blankng. The paper s organzed as follow: Secton presents the system model and Secton 3 provdes an analytcal evaluaton of the mpact of JTIDS system on LDACS performance. Secton 4 presents the proposed scheme whle n Secton 5 the numercal results are shown and n Secton 6 the conclusons are drawn.. SYSTEM MODEL In the scenaro under consderaton the LDACS and JTIDS systems operate n the same area and LDACS frequency band s part of one of the hoppng bands used by JTIDS. We consdered a system where LDACS operates wthn one of the band used by JTIDS, so, due to the frequency hoppng technque adopted by the latter, occasonally the two systems nterfere each other. In partcular, LDACS explot OFDM modulaton technque, where the data stream s dvded n many orthogonal sub-streams, referred as subcarrers, each one wth reduced rate. Ths allows to reduce the negatve effects of Inter Symbol Interference (ISI) wthout decreasng the total data rate. Accordng to the latest specfcaton n [8], the total LDACS avalable band s equal to B LDACS = khz, that s oversampled wth samplng rate f s =65 khz. Ths spectrum s dvded n N=64 subcarrers, of whch only 50 are actve. In addton, a cyclc prefx (CP) wth duraton equal to 11 samples s nserted n order to avod ISI. In the consdered system each actve subcarrer s modulated followng a QPSK scheme. The LDACS man parameters are outlned n Table 1, where t s s the samplng perod, Δf s the frequency spacng between two contguous subcarrers and T u, T g and T TOT are the useful symbol tme, the guard nterval duraton and the total OFDM symbol duraton, respectvely. B LDACS khz N 64 T u 10.4 µs f s khz CP 11 T g 17.6 µs Δf 9.76 khz t s 1.6 µs T TOT 10.0 µs Table 1: LDACS system parameters JTIDS explots the TDMA (Tme Dvson Multple Access) technque [9]. The transmsson n each tme slot s composed by a mnmum of 58 pulses of duraton T p =6.4µs and spaced by T =6.6µs. The actve part of each pulse s multpled for a spreadng sequence contanng 3 chps modulated through a CPFSK scheme and representng a combnaton of 5 nformatve bts: n partcular, JTIDS specfes only one spreadng sequence and the partcular combnaton of nformatve bts ntroduces a crcular offset n that. The chp duraton s equal to T chp =00 ns and hence the sgnal bandwdth s equal to 5MHz. However, snce the JTIDS system operates n a large range of frequences, [ ] MHz, dvded n bands of 3 MHz, the sgnal s fltered to ft the bands. Each pulse s transmtted on a dfferent frequency accordng to a certan hoppng sequence. In partcular there are N ch =51 possble carrers. Snce the hoppng sequence s a classfed nformaton, n our model we assume every carrer has the same probablty to be selected, hence we consder a hoppng pattern randomly generated wth unform probablty on all frequences. Consderng that the LDACS spectrum s 18

3 sgnfcantly smaller than the JTIDS bands, for the sake of smplcty we can assume t s completely contaned n one of the these bands. Ths means that only one hoppng frequency affects the LDACS sgnal but on the whole spectrum. Fnally, snce the JTIDS samplng rate s much hgher of that of LDACS, the JTIDS sgnal s downsampled. Ths makes the nterference even harder to be detected, snce the sgnal power s spread on a longer perod. We assume both the sgnals (LDACS and JTIDS) are transmtted on fadng channels and AWGN nose s added at the recever. Fadng coeffcents are modeled as random varables wth Rcean dstrbuton, wth Rcean factor equal to K=4 db and normalzed power. 3. ANALYTICAL INTERFERENC EVALUATION The development of a new system operatng n a frequency band densely populated by legacy systems ntroduces the need of a careful evaluaton of coexstence ssue. The jont representaton of the frequency doman sgnals of dfferent systems operatng on the same band gves a qualtatve evaluaton of the mutual nterference permts to put n evdence potentally crtcal scenaros. For ths reason we gve a jont representaton of JTIDS and LDACS spectrum n nomnal band and n Out of Band and Spurous Doman. The sgnals have been generated n Tme Doman (TD) takng nto account the standards mandatory features [8],[9] that drectly affect the spectrum shape. For LDACS1, an OFDM sgnal wth 64 subcarrers has been generated n FD and transformed n TD by means of a IDFT (Inverse Dscrete Fourer Transform). Then a rased cosne wndow that ams to reduce out-of-band radatons has been appled. Later, the sgnal has been transformed n FD and fltered wth the spectral mask. For JTIDS an mpulse of 6.4µs duraton has been produced and modulated wth a spreadng sequence of 3 chps each of 00ns duraton and CPFSK modulated. Once transformed the sgnal to FD, the spectrum mask has been appled. Fgure 1 shows LDACS and JTIDS power spectra when an offset (Δf) of 5 MHz s assumed between the central frequences of the two systems. It s evdent that the JTIDS spectrum heavy nterferers wth the LDACS due to ts hgh transmsson power. It can be noted that the results are smlar even wth hgher frequency offsets (.e. translatng the LDACS spectrum). To quantfy the effect of the nterference a numercal analyss can be performed. The am s to dentfy the condtons that permt to LDACS to operate n presence of JTIDS nterference. The analyss conssts of the computaton of the nterference power level at the vctm recever and on ts comparson wth the maxmum tolerable nterference power level obtaned from a protecton crtera typcally the mnmum Carrer to Interference rato (C/I). Fgure 1 JTIDS and LDACS spectra wth a frequency offset equal to 5 MHz The analyss follows the procedure defned n the CEPT MCL (Mnmum Couplng Loss) method [10]. The nterference level s obtaned by means of a lnk budget (1) that depends on dfferent parameters, as frequency separaton and dstance: I ( d, f ) = EIRP PL( d) + Grx Lrx + OCR( f ) + DC (1) where: - EIRP s the JTIDS Equvalent Isotropcally Radated Power; - PL(d) s the free space path loss; - Grx e Lrx are the LDACS recever antenna gan and cable loss, respectvely; - OCR(Δf) (Off Channel Rejecton) s a term that takes nto account the ablty of the vctm recever to reject the nterferer sgnal. It depends on the power spectral densty of the nterferer and on the frequency separaton between nterferer and vctm [11]. - DC s a term that takes nto account the nterferer duty cycle. For JTIDS, t depends on the Tme Slot Duty Factor (TSDF) that represents the maxm number of slots assgned to an user n a frame. From ths analyss t s possble to determne the mnmum dstance between the nterferer and vctm equpment at whch the nterference results to be tolerable. The coexstence between the two systems s guaranteed f ths dstance s lower than the mnmum operatonal dstance: n a rough evaluaton we can consder the mnmum operatonal dstance equal to the mnmum vertcal separaton of the arcrafts: 300mt (on the ground ths dstance s lower). The results of the analytcal analyss are shown n Table n a worst case assumng the followng workng hypothess: - Parameters of the lnk budget as antenna gans and cable losses are vared accordng to the scenaro 19

4 takng nto account typcal values for ground or arcraft nstallatons; - The rato C/I s fxed to 10 db, n accordng to B- AMC (Broadband Aeronautcal Mult-carrer Communcaton) specfcatons, from whch LDACS has been derved, snce ths value s not avalable n the current LDACS specfcaton. - The value of maxmum nterference power acceptable s obtaned from C/I rato assumng the carrer power equal to recever senstvty. - The JTIDS transmsson power s fxed to 1000 W. - TSDF s set to 50% and 5% that represent the maxmum and mnmum values. Scenaro Non Interferng Dstance TSDF=50% TSDF=5% 1 Ground Staton to d > 500km d > 157km Arborne Arcraft Arborne Arcraft to d > 500km d > 500km Arborne Arcraft to 3 Arborne Arcraft to d > 500km d > 60km Ground Staton 4 Ground Staton to d > 500km d > 500km Ground Staton 5 Arcraft on the d > 6.6km d > 18.5km Ground to Arcraft on the Ground 6 Ground Staton to d > 500km d > 16km Arcraft on the Ground 7 Arcraft on the Ground to Ground Staton d > 46.5km d > 5.5km Table : Analytcal Evaluaton Results These results show that mnmum dstance that allow the spectral compatblty s consderably hgher than the mnmum operatonal dstance n all scenaros, therefore the problem of the nterference of JTIDS transmsson on LDACS1 requres the applcaton of some countermeasures to ensure the coexstence between the two systems. 4. INTERFERENCE SENSING AND MITIGATION In ths Secton the proposed sensng and mtgaton scheme s explaned. The basc dea s the retransmsson of the Packet Data Unt (PDU) when the presence of JTIDS system s detected (t can be done through a frst sensng phase). The frst copy of the packet s stored and combned wth ts retransmsson: snce JTIDS and LDACS transmssons are ndependent processes, even f both the copes of the PDU are affected by JTIDS nterference wth hgh probablty dfferent portons of the PDU are corrupted. Packet combnng s used ether to mprove nterference detecton and sgnal decodng. Interference detecton ams to detect wth sgnfcant relablty whch samples of the PDU are affected by nterference. Ths s done through the energy detector, that s a partcular spectrum sensng algorthm whch computes the energy of the receved samples durng a tme nterval called sensng perod. Ths detector s well-known n the Cogntve Rado networks, where a secondary unlcensed user (SU) looks for spectrum holes that are not used by the prmary lcensed user (PU). Free spectrum portons can be used by the SU for transmsson. Energy detector provdes the test statstc (.e., the energy of the sgnal comng from the sampler) used to decde between two bnary hypothess: the channel s free and only thermal nose s present or prmary user sgnal plus nose s present. Accuracy of energy detector s proportonal to the duraton of the sensng perod: ncreasng the number of collected samples s possble to mprove the performance. The nterference detecton proposed here s slghtly dfferent: n our system the goal s to detect whch samples are corrupted and not only f the nterference s present. It means that the sensng perod must be lmted: dfferently from tradtonal approach ncreasng the sensng perod does not leads to a performance mprovement, snce the energy of the pulse s spread on a longer nterval makng more dffcult to dscrmnate whch samples are affected. For ths reason we ntroduce a modfed energy detector that explots a sldng wndow whch collects the energy of a part of the receved sgnal. In addton the nterference detecton s not performed on the receved sgnal but on a sample by sample dfference between the two copes of the sgnal that have been prevously equalzed to remove the channel effect. Ths permts to reduce the false alarm probablty due to by the a hgh peak to average rato (PAPR) that characterzes the LDACS (.e. OFDM) sgnal: the dfference between the two replcas depends only on the JTIDS nterference and nose power; the LDACS sgnal fluctuaton does not affect the detecton procedure. Assumng a perfect channel equalzaton, the receved sgnals dfference on the -th sample s: r ] = r [ ] r [ ] () [ 1 where r k [] s the -th receved sample of the k-th PDU transmsson (k=1,). The n-th output of the sldng wndow energy detector s the test statstc T n : 0

5 Fgure False alarm and mss detecton probablty for the proposed method and tradtonal sensng T where W s the wndow wdth and a are the wndow weghts (weghts are selected n order to gve more mportance to central samples). If T n exceeds a certan threshold nterference s supposed to be present n the sample n-th. The wndow length depends on the duraton of the nterferng sgnal that cannot be exactly known because the pulse s fltered by the recever, however a rough estmaton of the JTIDS sgnal duraton can be calculated as L = T p /f s = 4 samples, where f s s the samplng frequency and T p the pulse duraton. We adopt a wndowng sze, W, equal to L + 1 samples. A further mprovement to reduce the false alarm probablty s obtaned by observng M consecutve test statstcs: f at least M=L-1 consecutve samples are over the threshold we assume the nterference s present. Observng the retransmssons dfference s possble to know whch samples are affected by nterference but t s not known f the nterference s ntroduced by the frst, r 1 [n], or the second r [n] copy of the receved sgnal. Indcatng as n, wth = 0,, I 1 the samples affected by the nterference, for each n the values of r 1 [n ]and r [n ] are compared: the maxmum s blanked whle the mnmum s doubled. The resultng sgnals are summed together, and the fnal sgnal can be expressed as: r1 + r[ n] r = r1 f r f n = W n+ f n n r < r 1 r < r a W = n r[ ] 1 and and n = n n = n (3) (4) Fgure 3 Mean square dstorton of the receved sgnal and the sgnal after the nterference mtgaton Demodulaton and decson are taken on r sgnal. 5. NUMERICAL RESULTS Ths secton shows the numercal results obtaned to valdate the proposed scheme by resortng to computer smulatons. We start our analyss by evaluatng the performance of our nterference detecton method. The man drawback of ths method s the hgher nose power, snce by consderng the dfference between two sgnal copes the nose power s doubled. On the other hand, t does not suffer the presence of the LDACS sgnal as tradtonal method. Fgure shows the comparson of the false alarm (fa) and mss detecton probabltes (md) between our detecton technque based on dfference between two sgnal replcas ( r) and the same sensng technque performed on one sgnal (r 1 ) when the SIR (Sgnal to Inference Rato) vares. Lookng at ths fgure we can see that by explotng the two sgnal replcas t s possble to get a sgnfcant mprovement for both false alarm and mss detecton probabltes. In order to evaluate the behavor of the mtgaton scheme, we defne the performance ndex D as the mean square dstorton of the sgnal. 1 P 1 D = r'[ p] s[ p] P p= 0 Fgure 3 presents the dstorton of the receved sgnal and the reconstructed one, when the SNR vares and for dfferent SIR values. From ths fgure we can see that the proposed scheme s able to sgnfcantly decrease the dstorton of the sgnal for the consdered SIR values. In partcular, the resdual dstorton s due to the AWGN nose. (5) 1

6 Fgure 4 BER performance of the system wthout channel codng when nterference s not mtgated and when blankng and the proposed technque are appled Ths means that the presented mtgaton scheme s able to reject the nterference made by JTIDS and to brng an addtonal gan of 3 db due to the soft combnng of the packet s replcas. The good behavor of the proposed scheme s even more evdent n terms of BER (Bt Error Rate). When the nterference s hgh the BER gan permts to counteract the reducton of throughput ntroduced by the retransmsson assurng more relable communcatons. Fgure 4 represents the BER when the QPSK modulaton s used for dfferent SIR values and when the SNR vares. In ths fgure the proposed scheme s compared wth tradtonal blankng method and the case wthout any elaboraton. We can see that nterference leads to a floor of the BER: ths means that the performance does not mprove even when the SNR gets hgher. Even tradtonal blankng does not allow to reach satsfyng performance. On the other hand, when the proposed scheme s appled we have excellent results: the performance shows the nterference s almost completely removed and we have an addtonal gan of about 3 db due to the soft combnng of the packets. For hgh sgnal to nose rato even the performance of the proposed method reaches a floor, due to those samples affected by nterference n both the sgnal replcas. However ths drawback s overcome when the channel codng s consdered: Fgure 5 shows the same performance of the prevous case but consderng the channel codng envsaged n LDACS specfcatons. In partcular we consdered an nner convolutonal coder wth codng rate equal to 1/, an nterleaver and an outer Reed-Solomon coder. We can observe that the floor concerns only the performance of the system when mtgaton s not appled. Here the performance of the proposed scheme clearly overcome the tradtonal blankng soluton. We can note from the prevous results that performance does not depends on the consdered SIR: when the nterference power Fgure 5 BER performance of the system wth channel codng when nterference s not mtgated and when blankng and the proposed technque are appled s very hgh t s easy to be detected and removed, whle when the JTIDS sgnal strength s low t becomes more dffcult to detect but t has also a lower mpact on performance. 6. CONCLUSIONS In ths paper we proposed a novel scheme for mpulsve nterference detecton and mtgaton n the new LDACS system. The proposed method s based on nterference detecton and packet retransmsson: by explotng the dfference between the sgnal replcas affected by two ndependent nterference realzatons t s possble to mprove the nterference detecton, snce t does not depend on the useful sgnal. Furthermore, the two sgnal copes can be used to reconstruct the useful sgnal, decreasng the dstorton and the system bt error rate. 7. REFERENCES [1] "Future Aeronautcal Communcatons", ISBN , book edted by: Dr. Smon Plass, German Aerospace Center (DLR), Germany, September [] Communcatons Operatng Concept and Requrements for the Future Rado System, ICAO Std., Rev. Verson, May 007. [3] M. Budsabathon and S. Hara, Robustness of OFDM sgnal aganst temporally localzed mpulsve nose, n Proc. VTC 001 Fall Vehcular Technology Conf. IEEE VTS 54th, vol. 3, 001, pp [4] S. V. Zhdkov, Performance analyss and optmzaton of OFDM recever wth blankng nonlnearty n mpulsve nose envronment, vol. 55, no. 1, pp. 34 4, 006. [5] S. V. Zhdkov, Analyss and comparson of several smple mpulsve nose mtgaton schemes for OFDM recevers, vol. 56, no. 1, pp. 5 9, 008.

7 [6] Y.-H. Km, K.-H. Km, H.-M. Oh, K.-H. Km, and S.-C. Km, Mtgaton of effect of mpulsve nose for OFDM systems over power lne channels, n Proc. IEEE Int. Symp. Power Lne Communcatons and Its Applcatons ISPLC 008, 008, pp [7] J. Armstrong and H. A. Suraweera, Impulse nose mtgaton for OFDM usng decson drected nose estmaton, n n IEEE Int. Symp. Spread Spectrum Technques and Applcatons, 004, pp [8] Updated LDACS1 System Specfcaton, SESAR ET - Task EWA04-1 T Std , Aug. 011 [9] JTIDS System Segment Specfcaton (DCB79S4000C), Std. [10] E. R. C. E. wthn the European Conference of Postal and T. A. (CEPT), A comparson of the mnmum couplng loss method, enhanced mnmum couplng loss method, and the Monte-Carlo smulaton, Tech. Rep.,1999 [11] Recommendaton SM Unwanted emsson n spurous doman, ITU-R Std. 3