The Design of a Spatial Diversity Model to Mitigate Narrowband and Broadband Interference in DSSS Ad Hoc Networks

Transcription

1 The Design of a Spatial Diversity Model to Mitigate Narrowband and Broadband Interferene in DSSS Ad Ho Networks Sonia Furman, and Mario Gerla University of California, Los Angeles Abstrat Spatial diversity has been gaining signifiant momentum in ellular systems due to its ability to improve radio links in mobile wireless hannels. Tehniques of spatial diversity have been expanded to build foundations for advaned antenna array tehnology, to inrease apaity in mobile wireless ommuniation systems due to radiation patterns that diret energy only in the intended diretion, thereby greatly reduing interferene. Similar benefits may be gained in DSSS ad ho networks, though work in this area has been limited. To exploit spatial proessing in future ad ho networks, aurate array designs that embed spatial harateristis of the hannel and radiation patterns are neessary to quantify the performane benefits. Therefore, in this study we designed a spatial diversity model and using simulation examined the performane benefits attained when onfigured in a DSSS ad ho network, subjet to hannels with narrowband and broadband interferene. I. INTRODUCTION Reeption in spread spetrum systems depends to a large extent on the harateristis of the interferene and signal reovery methods used at the reeiver. To improve signal robustness at the reeiver it is essential to provide solutions that have the ability to mitigate the effets of interferene in the hannel. Spatial diversity is a well-known tehnique used in spread spetrum ode division multiple aess (CDMA) wireless ommuniations systems to ompensate for hannel impairments, and results in improved radio links, network apaity and effiient spetrum utilization without the need to inrease power at the transmitter nor hannel bandwidth. In spread spetrum systems this is essential sine they are known to be suseptible to interferene and may be jammed by narrowband interferene (a ontinuous wave near the enter frequeny of the transeiver), or by a pulsed broadband signal whih in either ase does not fit the stationary additive white Gaussian noise (AWAGN) model [1]. In ontrast to the majority of researh on antenna arrays and spatial diversity that has foused on ellular systems [2], this study deals with spatial diversity in ad ho networks, whih in partiular presents a great hallenge due to its inherent arhiteture. Unlike ellular systems, ad ho networks are loal area networks (LANs) that ommuniate over a medium with no observable boundaries, are formed without pre-planning, and exist only for as long as they are needed. Though this makes them attrative for appliations where fast deployment is ritial, mehanisms of entral ontrol (as with base stations) is not appliable. In the DSSS ad ho networks, eah station uses the same 11- hip pseudo-noise (PN) spreading ode that results in a proessing gain (PG) of only 10.4dB [3]. This proessing gain undoubtedly redues the bit error rate to some extent but it may not be suffiient to attain the desired bit error rate in situations where the hannel is subjet to high interferene. For this reason we designed a spatial diversity model for simulation in an ad ho network, and ompared the performane with and without spatial diversity for hannels affeted by narrowband and broadband interferene. The results obtained from the simulation show that with spatial diversity, the throughput onsistently improved ompared to onfigurations with only an omnidiretional antenna. The approah (for the most part) in prior related work employs abstrat antenna models, and has foused on modifiations in the Medium Aess Control (MAC) protool to aommodate diretional antennas at both the transmitter and reeiver [4][5], whih differs onsiderably to the work presented in this paper. The fous of this study has been on the design of a spatial diversity model based on equal gain ombining to mitigate the effets of interferene in a DSSS ad ho network. The remainder of this paper is organized as follows: Setion II presents the spatial diversity model. Setion III desribes the system model, and details the effets of interferene in the system together with a brief overview of DSSS in ad ho networks. Setion IV desribes the network simulation and results, followed by onlusions in setion V. II. THE SPATIAL DIVERSITY MODEL Diversity is desribed by [6] as a powerful ommuniation reeiver tehnique that provides wireless link improvement by exploiting the randomness of the radio propagation in a wireless hannel. Sine the radio goes through multiple paths with different fading, one signal may be weak yet the other may be strong whih implies that by ombining or having more than one path from whih to selet, both the average as well as the instantaneous SNRs at the reeiver an be improved. There are four basi ategories that desribe spatial diversity: seletion, feedbak, maximal ratio ombining, and equal ratio ombining. Both maximal ombining ratio and the equal ratio ombining spatial diversity tehniques use omplex weights to adjust the inoming signal from eah antenna element, whih is then ombined (summed) into a signal direted to the reeiver s detetor. The primary differene between the two tehniques is that the weights assoiated with eah of the elements in the equal ratio ombining method are fixed rather than adjustable. Figure 1 illustrates the basi differene between the two diversity tehniques. The reeived signals in equal ratio ombining are o-phased to provide equal gain in all diretions.

2 w 3 w 2 Reeiver Detetor w 1 Weight Adapting (a) w 3 w 2 Reeiver Detetor w 1 (b) Fig. 1. (a) Maximal ratio ombining (b) Equal gain ombining diversity We designed a 5-element antenna array that omprises of isotropi elements, uniformly exited and equally spaed along the perimeter of a irle of radius R = 1.5λ. The 1.5 oeffiient multiplier of the wavelength (lambda) was the optimum seletion that met the diversity riteria for maintaining a multiple of wavelengths separation between the antennas to ahieve deorrelation at the reeiver where the reeived signal undergoes independent attenuation at eah antenna element. For an impinged plane wave originating at an i th soure arriving at the antenna element, the angle of arrival AOA ( θ, φ ) of the multipath omponent is desribed in terms of its azimuthal angleφ and its elevation angleθ. In our model we assume that the elevation angle is equal to π/2, and we derived the gain assoiated with the AOA in the azimuthal diretion φ for the o-phased reeived signals with onstant weights and equal gain in the diretion of the boresight. We relied on the Medium Aess Control (MAC) protool implemented in the QualNet [7] simulator libraries to perform the beam sanning in whih a single main beam is steered in the diretion of the signal of interest. The radiation patterns derived from implementing the diversity system resulted in a boresight gain of dBi; a half power beamwidth (HPBW) of 72 o degrees; and a side lobe level (SLL) of dB. The SLL parameter expresses the ratio in terms of the absolute maximum value of the largest sidelobe to the absolute maximum value of the main lobe [8]. Figure 2 shows the radiation pattern gain for two suessive beams with boresights of 144 o and 216 o, indiating the peak power gain [Watts/Watt] and its equivalent representation in db to illustrate the SLL performane Fig. 2. Radiation patterns of the spatial diversity model III. SYSTEM MODEL The primary attributes of systems that employ spread spetrum tehniques resides in their ability to rejet interferene, and provide a high degree of resistane to multipath fading and jamming. Spread spetrum signals have noise-like properties due to a pseudo-noise (PN) sequene that ontrols their waveform. By multiplying the baseband data pulses with a PN sequene at the transmitter, the message signal is spread at the modulator that results in a muh higher pulse rate (greater than 10). The message, whih is independent of the data sequene, results in a diret sequene (DS) spread spetrum (DSSS) signal. The DSSS system desribed in [3] for the physial (PHY) layer is aimed at operation in the 2.4GHz to GHz band, whih is designated as the Industrial Sientifi and Medial (ISM) band for appliations in the USA that support an RF LAN with 1Mbits/s, and a 2Mbits/s payload ommuniation apability. In the designated PN sequene is the Barker ode. Barker odes belong to a lass of PN sequenes known speifially for their sharp autoorrelation with relatively low sidelobes whih makes them well adaptable to low detetion thresholds [9]. The Barker ode is represented by an 11-hip antipodal sequene [ ] generated with a PN lok of 11MHz (f h ) in onjuntion with a baseband signal of 1MHz (f b ), resulting in a PG= 10.4 db. (PG = T b /T h, usually expressed in db). The rossorrelation at the reeiver in systems is performed with the same Barker ode that is generated loally to restore the modulated signal of interest. If the rossorrelation is performed on any other signal, suh as interferene, the result will be a small amount of wideband noise that is easily filtered in the IF setion of the reeiver. In the time domain, the delayed version of the PN sequene will have a poor orrelation with the original PN sequene and thus appear as an unorrelated user or wideband noise.

3 An extension of the PHY for DSSS system alled High Rate DSSS (HR/DSSS) [3] builds on the data rate apabilities desribed above, to provide 5.5 Mbits/s and 11Mbit/s payload data rates in addition to the 1 and 2 Mbps rates. In order to provide these high rates, an 8-hip (omplex) omplementary ode keying (CCK) is employed as the modulation sheme. However, it is important to note that the hipping rate remains the same 11MHz. The system model is based on the PHY layer in [3]. The basi mehanism for transmitting pakets between the medium aess ontrol (MAC) layer and the PHY layer is provided by the physial layer onversion proedure (PLCP). The PLCP s frame format ontains all the neessary information pertaining to modulation, synhronization, data rates and CRCs. Elements of the transmitter inlude a srambler, spreader and modulator. All bits transmitted by the DSSS PHY layer are srambled by the polynomial G (z)=z -7 +z The output of the srambled data is then onvolved with the 11-bit PN Barker ode (11 MHz hip rate T ). Differential binary phase shift key (DQPSK) is used to adjust the arrier and bit timing based on the spreading performed by the Barker ode, in order to produe the in-phase (I) and quadrature (Q) omponents of the transmitted signal. Given that the information rate at the transmitter is R b = 1Mbits/s where the bit interval T b =1/ R b (1 miroseond), and that the hannel bandwidth is B =22MHz (- 11, +11MHz measured from the arrier frequeny f ), we see that the hannel bandwidth obviously is muh larger than the bit rate (B >> R b ). This is a required ondition for the spreading to take plae. By representing the information-bearing baseband signal by m(t), as shown in (1), and the PN sequene (t) in (2), the transmitted signal may then be expressed in (3). mt () = an gt( t ntb) (1) n= where: a n =+ 1 is the binary data in antipodal representation for - < n <, and g T is a retangular pulse of duration T b. t () = n pt ( nt ) (2) n= where: n = + 1 represents the binary PN sequene and p(t) is a retangular pulse of duration T, st () = Amtt ()()os( 2πft + φ ) (3) where: A is the amplitude and f, φ the arrier frequeny and phase respetively. The output of the hannel r(t) expressed in (4) omprises of the signal with proessing gain plus interferene i(t). rt () = Amtt ()()os( 2πft + φ ) + it () (4) Figure 3 depits the spread baseband signal at the transmitter and the signal after despreading at the reeiver. Fig. 3. The baseband signal after spreading at the Tx and despreading at the Rx. Upon reeption, the reeived signal r(t) is onvolved (*) with the loal PN sequene, whih despreads the informationbearing signal, and spreads the interferene through onvolving the total reeived signal from the hannel (5). rt ()* t () = Amt ()os( 2πft + φ ) + it ()* t () (5) The net effet is a redution in the interferene power by a fator equal to that of the proessing gain. We express the power spetral density of the narrowband interferene as S I, 2 where S I = P I /W, and PI = AI / 2 is the average power of the narrowband interferene. After dispreading the signal, the resulting signal is then demodulated by a orrelator or a mathed filter with a bandwidth R b [Hz]. The total power in the interferene at the output of the demodulator is then S I R b = (P I /W) R b whih an be shown through substitutions to be the same as P I /PG. Hene the power in the interfering signal is redued by the amount of proessing gain. For narrowband interferene we onsider the interferene i(t) to be sinusoidal in the bandwidth (BW) of the spread spetrum signal expressed in (6) it () = Aos( 2πft + θ ) (6) I I The signal to noise ratio plus interferene at the detetor is (SINR) d = 2E b /P I T, where the energy E b = P S T b. expressed in terms of proessing gain (through substitution) beomes: (SINR) d = 2(Ps/P I )PG. For broadband interferene we assume a random proess with a onstant power spetral density over the entire BW of the spread signal expressed in (7).

4 PI = Si( f ) df = WI (7) 0 Figure 4 shows the probability of the bit error for an adaptive white Gaussian noise hannel with no interferene and no proessing gain, in addition to the SINR urves for both narrowband and broadband interferene in the DSSS system with a proessing gain of 11 (10.4dB). Fig. 4. BER vs. SNR: AWGN & Narrowband, Broadband Interferene IV. NETWORK SIMULATION AND RESULTS The ad ho network environment is represented by a terrain of 1600x1600m with nodes that are randomly distributed and share a wireless hannel with a data payload of 2Mbps. The experiments were performed with the assumption that the nodes are mobile by using the Random Waypoint mobility algorithm (speed 0 to 10m/s, and 0 pause time) for 50, 75, 100, 125, and 150 nodes. The total noise power inluding thermal plus interferene was used to derive the SINR at the reeiver. BER/SNR lookup tables were generated using differential quadrature pulse shift keying (DQPSK) modulation for both narrowband and broadband interferene in a system with a proessing gain of 10.4dB. The minimum threshold for the reeiver to reeive pakets was set at -81dBm and its sensitivity at -91.0dBm. The arrier frequeny used was f =2.4GHz. Traffi was generated using a onstant bit rate (CBR) generator with a ratio of 1:5 sessions per CBR. Eah paket was set at 512 bytes in length. Three performane metris are onsidered: throughput, the perentage of bits per seond (bps) that were suessfully reeived at the appliation layer; delay, the average elapsed time (in seonds) between the transmission and reeption of suessful pakets; and the maximum throughput performane improvement gain (MPG), whih is the ratio of performane with spatial diversity to that without for a given number of nodes [8]. The results obtained represent an average of 4 runs with random seeds. Figures 5a and 5b show that with narrowband interferene the average network throughput performane was 38.3% better with spatial diversity than without, whereas with broadband it was 47.62%. The highest maximum performane improvement gain (MPG) with the narrowband interferene was 2.06 and ourred with 125 nodes, whereas for the broadband interferene the MPG was 1.95 and ourred at 100 nodes. We attribute the gain in performane to the spatial diversity. In the ase where the average SNR per hannelγ is suffiiently large (greater than 10dB) as with the narrowband and broadband interferene (Fig.4), the probability of error with diversity an be approximated by (8) where L is the order of diversity [1]. P e (2L 1)! L L ( 1)!! 1 4γ L It is readily seen from (8) that for L=1 (the ase of no diversity), the resulting probability of γ an hange from an 1 inverse linear relationship of to an inverse-squared relationship for L=2, and to an inverse ube relationship for L=3, et. This redues the probability of error suffiiently to overome the adverse effets resulting from narrowband and broadband interferene ompared to an AWGN hannel. For example, for a high average SNR per hannel of 25dB the probability of error using (8), with no diversity is e-4 ompared to e-7, with a seond order diversity (L=2). It should be noted that though (8) was derived for BPSK, without loss of generality it learly demonstrates the impat of diversity assuming the average SNR per branh to be equal. The signifiane of the simulation is that while the diversity gain may be determined analytially at the output of the ombiner (Fig. 1), the results of the simulation inorporate the effets of salability, mobility, and routing in the ad ho network. In the senarios we simulated, the radio range is m, and sine all the nodes are not within range of eah other this indiates the need for multihop routing that affets the performane of diversity. The vulnerability to salability is evident in Figure 5 at the transition from 100 nodes. At low densities, (number of nodes per area), the throughput is in the ninety perentile with spatial diversity for both narrowband and broadband interferene, while at high densities the performane delined rapidly. Interestingly, the behavior in high density was similar in the rate of degradation with and without spatial diversity though the MPG always exeeded a ratio of 1 (i.e. with spatial diversity the performane onsistently exeeded that without spatial diversity in all the senarios). The MPG for narrowband ranged from 1.06 to 2.06 and for broadband 1.12 to The dynamis of maintaining information at neighboring nodes in mobile ad ho networks results is an overhead (due to short ontrol messages neessary for routing), whih adversely affets the utilization of spatial diversity sine it is highly dependent on disrete time sampling of signal replias from the hannel (equations 1-6). With other routing protools it may be possible to obtain even better results. P e 4γ (8)

5 Fig. 5. Throughput and delay for narrowband (NB) and broadband (BB) interferene with and without spatial diversity (SD): (a) NB interferene (b) BB interferene () NB & BB wo/sd (d) NB & BB w/sd (e) Delay NB (e) Delay BB Figures 5 and 5d show the results for narrowband ompared to broadband without spatial diversity and with spatial diversity respetively. The similarity in performane is due to the small differene of 1.25 db between narrowband and broadband interferene (Fig. 4) for BERs between 10-7 and Figures 5e and 5f show a delay improvement with spatial diversity for both narrowband and broadband interferene. The average delay improvement for narrowband was 56.69% with spatial diversity and 38.98% for broadband The inrease in performane with spatial diversity is attributed to the redued delay spread inherent in the implementation of the antenna array. V. CONCLUSIONS In this study, we were able to demonstrate that with spatial diversity in the DSSS ad ho network the throughput performane inreased, and a redution in end-to-end delay was attained for hannels inundated by either narrowband or broadband interferene. This is due to the spatial diversity 5- element antenna model we designed whih ombines the reeived o-phased signals with fixed weights assoiated with eah of the elements, in order to provide equal gain in all boresight diretions. The results obtained in this study enhane the understanding of interferene in DSSS ad ho networks, spatial diversity onsiderations, and may be applied in future researh to improve the design of transeivers and network protools for even further improved link quality in advaned ad ho network appliations. REFERENCES [1] R.L. Peterson, R.E. Ziemer, and D.E. Borth, Introdution to Spread Spetrum Communiations, Englewood Cliffs, N.J., Prentie Hall, [2] R.B. Ertel, P. Cardieri, K.W. Sowerby, T. S. Rappaport, and J. H. Reed, Overview of Spatial Channel Models for Antenna Array Communiation Systems, IEEE Personal Comm., pp , Feb [3] Institute of Eletrial and Eletronis Engineers, AIEEE Standard b: Supplement to ANSI/IEEE Std Edition, Wireless LAN Medium Aess Control (MAC) and physial Layer (PHY) IEEE, In., New York, [4] Y-B. Ko, V. Shankarkumar, and N. H.Vaidyn, AMedium Aess Control Protools using Diretional Antennas in Ad Ho Networks,@ IEEE INFOCOM, pp 13-21, May [5] A. Nasipuri, S. Ye, J. You, and R.E. Hiromoto, AA MAC Protool for Mobile Ad Ho Networks Using Diretional Wireless Communiations and Networking Conferene, vol.3, pp , [6] T.S. Rappaport, Wireless Communiations Priniples and Pratie, Upper Saddle River, NJ, Prentie Hall PTR, [7] QualNet Simulator Version 3.0 User=s Manual, Salable Network Tehnologies, In., Los Angeles, CA, [8] S. Furman, and R. Bagrodia, A Comparative Study on the Effets of Spatial Diversity in Ad Ho Networks using On- Demand Routing Protools, International Conferene on Parallel Proessing, Proeedings of ICCP Workshops, pp , [9] I. Bar-David, and R. Krishnamoorthy, ABarker Code Position Modulation for High-Rate Communiation in the ISM Bands,@ Bell Labs Tehnial Journal, pp , Autumn 1996.