INVESTIGATION OF CLOSE-TO-WALL WIRELESS SENSOR DEPLOYMENT USING 2D FINITE-DIFFERENCE TIME-DOMAIN MODELLING. Y. Wu and I. J.

Transcription

1 2 nd Intenational Confeence on Wieless Communications in Undegound and Confined Aeas August 25-27, 2008 Val-d O - Québec - Canada INVESTIGATION OF CLOSE-TO-WALL WIRELESS SENSOR DEPLOYMENT USING 2D FINITE-DIFFERENCE TIME-DOMAIN MODELLING Y. Wu and I. J. Wassell Compute Laboatoy, Univesity of Cambidge, Cambidge, CB3 0FD, United Kingdom yw264@cam.ac.uk, ijw24@cam.ac.uk ABSTRACT When Wieless Senso Netwoks (WSNs) ae deployed in a ailway tunnel envionment, the available deployment positions and equipment size ae esticted by egulations fom the infastuctue ownes in ode to maintain the loading gauge. As we have discoveed in the Aldwych tunnel [1], antennas mounted nea to the tunnel wall wosens the path loss (PL) pefomance at 868MHz and 2.45GHz. Howeve, no liteatue exists concening the detailed effects of close to wall antenna mounting and its impact on wieless senso deployment. Cuently, 2.4GHz is the pefeed opeating fequency, since it is employed by many of the existing commecially available low powe wieless communication systems. In this pape, the Finite-Diffeence Time-Domain (FDTD) method [2] has been used to analytically exploe the adiation pattens (RPs) of a 2.4GHz antenna mounted at diffeent distances away fom a tunnel wall constucted fom diffeent mateials, i.e., cast ion, concete and polythene. 1. INTRODUCTION Changes in geomety, stess state and mateial deteioation ae the majo causes of damage in tunnels. In geneal, the damage is difficult to assess though conventional methods. Consequently, thee has been a lot of ecent wok in the field egading the use of Wieless Senso Netwoks (WSNs) fo this pupose. WSNs ae becoming an inceasingly common tool fo monitoing and assessing the condition of aged civil infastuctue. The benefits of WSNs ove wied netwoks include the ease of deployment and lowe costs due to lack of wiing. Howeve, in pactice, the pocess of deploying WSNs pesents a new set of challenges, such as choices of senso nodes, powe havesting, netwok topology and its optimisation, data mining, secuity, adio popagation and development of common communication potocols fo diffeent systems as each deployment is stongly dependent on the individual application. These mattes ae futhe complicated by a lack of standads and a wide ange of diffeent wieless netwoking hadwae and potocols. In addition, little is known about the adio popagation in paticula envionments. Fo example in a tunnel, factos such as tansmit fequency, antenna position, tunnel diamete, building mateial and couse [3-7] can in vaious ways affect the adio popagation. Even so, thee ae still many othe factos, which need to be addessed. Hee we pesent ou investigation concening wieless senso deployment fo a close-towall scenaio using the finite-diffeent time-domain (FDTD) technique. The FDTD technique is one of the impotant simulation tools in the study of Electomagnetic popagation. Following the Pefectly Mached Layes (PMLs) poposed by Beenge [8], the absobing bounday condition was extended to addess a wide ange of situations. With espect to dispesion which occus in the FDTD method, thee ae geneally two ways to minimize its effects. The fist possibility is to use the Highe Ode FDTD technique [9] while the second option is to decease the size of the unit elements. Hee we apply the latte to ou poblem, i.e., when one twentieth of the signal wavelength (λ ) in fee space is used as the basic element dimension i.e., the unit cell size, good accuacy can be achieved in a FDTD simulation [10]. We will use this technique to investigate the close to wall antenna adiation patten (RP) as a function of the distance to the wall ( d ) and of tunnel walls constucted fom mateials having diffeent physical constants: elative pemittivity ( ε ), elative pemeability ( µ ) and conductivity (σ ). To simplify the aguments, in ou model, we assume that: (a) a point souce is deployed on the suface of a flat wall. (b) eflections fom othe walls ae not taken into consideations when conducting ou RP calculations. The pape is oganised as the follows. The Two- Element Aay Model is biefly intoduced in Section 2. The 2D FDTD simulation setup is descibed in Section 3. In Section 4, the validation of the simulation esults is evealed and deployment ecommendations ae made. Finally, Section 5 daws ou conclusions. 2. TWO-ELEMENT ARRAY MODEL When two identical point souces (S1 and S2) in the hoizontal plane synchonously popagate adio waves, a Two-Element Aay is fomed as shown in Figue 1. Consequently, in the fa field (i.e., 1 2

2 2 nd Intenational Confeence on Wieless Communications in Undegound and Confined Aeas August 25-27, 2008 Val-d O - Québec - Canada and 1 2 ), the Aay Facto (AF) descibed in [11, 12] can be expessed as: π D θ AF = cos( cos + ), (1) λ 2 whee is the azimuth angle, D is the spacing between the two antennas, whee D= 2d, and θ is the phase diffeence between the cuents diving the antenna elements. z d d S y ( ε, µ, σ ) to the coesponding unit cells to epesent 3 walls made of cast ion ( ε = 1.0, µ = 1.0, σ = ), concete ( ε = 7.0, µ = 1.0, σ = 0.15), which ae the main lining mateials in tunnels o polythene 3 ( ε = 3.0, µ = 1.0, σ = ). The emaining unit cells in the poblem space ae assigned as ai ( ε = 1.0, µ = 1.0, σ = 0). The simulation bounday is teminated by the PMLs. The point souce is placed at the positions of inteest, i.e., at vaious spacings fom the wall, popagating a sinusoidal signal of fequency 2.40GHz (i.e., λ =12.5cm). Afte executing the FDTD iteations fo a numbe of time steps until the steady state is eached, signal stength samples ae collected. By compaing with the 2D FDTD fee space simulation esults unde the same FDTD paametes, the RP is plotted. Mateial Cast Ion S2 Figue 1: Geomety of the Two-Element Aay Model As we can ealise, the AF configuation is equivalent to the scenaio when a point souce is placed in font of a metal wall (along the diection of y in Figue 1), whee ideally the metal acts as a pefect eflecto to the point souce. In othe wods, the wave eflected by the wall can be consideed as emitted by the image antenna. Theefoe 0 in this situation it is appopiate to setθ = 180. Although the Two-Element Aay Model gives us some basic undestanding of the fom of the RPs yielded in ou investigation, it does not povide us with the additional flexibility to cope with diffeent wall mateials, which can be epesented as a set of physical constants ( ε, µ, σ ) fo each mateial. 3. FDTD SIMULATION SETUP As we have found in [13], fo a simple stuctued FDTD model, e.g., the fee space model and the plane eath model, we can always use the Modified 2D FDTD simulation instead of a full 3D simulation, povided appopiate coection factos ae applied. Fo the 2D FDTD model in this pape, we use the TM mode, which contains the field components (Ez,Hx,Hy) descibed in [14]. In the tunnel deployment, this coesponds to positioning a vetical antenna in paallel with and close to the wall and pependicula to the ails. Within the poblem space, we assign the appopiate values of Concete Polythene 0.6cm 2cm 3.125cm 6.25cm 12.5cm 25cm Distance fom the Wall (cm) Figue 2: Radiation Pattens at 2.4GHz 4. SIMULATION RESULTS AND ANALYSIS Owing to the deployment issues, the feasible values of d ae less than a few 10s of cm. In ou simulations the point souce is positioned at 7 diffeent distances away fom the wall, specifically 0.6cm 1 25 λ, 2cm 4 25 λ, 3.125cm 1 4 λ, 6.25c 1 2 λ, 12.5cm( 1λ ) and 25cm( 2λ ). Figue 2 shows the RPs plotted in the spatial domain (azimuth) fo each of the cases descibed peviously, whee the dak blue colou epesents low adiation gain and the ed colou epesents high adiation gain. The left bounday of each individual figue has the wall of

3 2 nd Intenational Confeence on Wieless Communications in Undegound and Confined Aeas August 25-27, 2008 Val-d O - Québec - Canada diffeent mateials teminated. As we can obseve fom the plots, the RP changes damatically as the atio of d to λ is vaied. Essentially, we do not want to locate the senso too close to the wall, since fom the 1 st column in Figue 2, it can be seen that most of the powe is diected towad the opposite wall, which will give poo communication to othe senso nodes mounted along the tunnel. Fo each of the mateials shown in the ows in Figue 2, the RP tends to spead wide and wide as the spacing inceases fom 1 25 λ up to 1 λ. Howeve when 4 the distance is futhe inceased, the RP splits initially into two lobes and then moe lobes as d is inceased futhe. Between these lobes, we have zones of low gain. Looking at the 4 th, 5 th, and 6 th columns of Figue 2, concete and polythene walls exhibit less deep nulls than cast ion in the RPs howeve they have a educed gain oveall owing to enegy absoption. In Figue 3, the RPs ae pesented in the fom of pola plots (in db). It can be seen that the Two-Element Aay solution of Eqn. (1) shows a good fit with the cast ion wall FDTD simulation esults while at the same time, the RPs fo concete and polythene walls ae also compaed. Inside each figue, we define the diection fom the cente to 0 degee as the nomal to the wall and the wall is located in line with -90 degees and 90 degees. Based on ou findings, when the distance is futhe away fom the wall egadless of its mateial, the maximum gain is inceased. Again we need to ensue that intusion into the tunnel is minimized while maintaining an acceptable RP. Fo geneal wieless communication puposes, at an opeating fequency of 2.4GHz, an antenna sepaation fom the wall of at least 2~3cm is ecommended. A tunnel is a special adio popagation envionment in the sense that is a lagely linea aangement. In ode to maximize the wieless communication ange, we in geneal pefe that the main powe lobes point along the tunnel couse, which ensues that powe is tansmitted up and down the tunnel athe than bouncing between walls. We see that this can be achieved by using a spacing of 6.25cm 1 2 λ as shown in the 4th column in Figue 2 o Figue 3(d). A futhe peliminay investigation has shown that, by placing a dielectic mateial in font of cast ion wall, the RP pefomance can be impoved. As an example, we (a) (b) (c) (d) (f) (g) Figue 3. Radiation Pattens at 2.4GHz

4 2 nd Intenational Confeence on Wieless Communications in Undegound and Confined Aeas August 25-27, 2008 Val-d O - Québec - Canada added a polythene slab of 6cm in width and 1cm in thickness as illustated in Figue 4(a). Fo compaison, we show Figue 4(b) and (c) which coesponding to the 3 d and 4 th column of the Cast Ion scenaio in Figue 2. As we can obseve, Figue 4(d) diects moe enegy along the tunnel wall than does the aangement without the dielectic shown in Figue 4(b). Meanwhile it also has a less deep null in the diection (0 degee) nomal to the wall and has a smalle spacing fom the wall than that of Figue 4(c). Figue 4(e) illustates the pola plots of the coesponding RPs. In pactice, the motes opeate in the ISM band between 2.405GHz and 2.480GHz. We have also undetaken a set of simulations to detemine the stability of the RPs ove the specified ISM fequency band. Fom the esults shown in Figue 5, it can be seen that the RPs ae easonably stable, paticulaly fo the pactically impotant close to wall spacings. The lagest vaiation is obseved at 0 degees fo the lage antenna spacing shown in Figue 5(c). 5. CONCLUSIONS In this pape, we have discussed the scenaio when an antenna is deployed close to a wall. By simply alteing the sepaation distance we have seen that the RPs have damatically diffeent shapes. Accodingly, we have poposed suitable antenna spacings and a potential design that achieves RPs with suitable diectional popeties. The constuction mateial affects the oveall adiated powe since non-metallic mateials can absob significant powe. In ode to impove ou simulation esults fom the 2D FDTD method, we can use fine esolution to epesent the size of a unit cell, e.g., one fotieth of λ instead of one twentieth ofλ. Also accoding to the definition fom the two-element aay model, the RPs ae calculated via the points of inteest in the fa field. Hee in ou simulation model, the RPs ae obtained at 20λ distance fom the souce. Thus we can incease this distance to impove the appoximation as indicated in Figue 1. Howeve, both come at the cost of highe computational powe in tems of CPU execution time and memoy usage. Figue 4: Radiation Patten Impovement by Adding Dielectic Slab 6. REFERENCES [1] Y. Wu, M. Lin, and I.J. Wassell, Modified 2D Finite- (a) (b) (c) Figue 5: RP Fequency Sensitivity: (a). d = 3.125cm (b). d = 6.25cm (c). d = 25cm

5 2 nd Intenational Confeence on Wieless Communications in Undegound and Confined Aeas August 25-27, 2008 Val-d O - Québec - Canada Diffeence Time-Domain Technique fo Tunnel Path Loss Pediction, 2 nd Intenational Confeence on Wieless Communication in Undegound and Confined Aeas (To Appea), Aug [2] K.S. Yee, Numeical Solution of Initial Bounday Value Poblems involving Maxwell s Equations in Isotopic Media, IEEE Tans. Antennas Popagat., vol. 14, No. 3, pp , May [3] D. Didascalou, J. Maue, and W. Wiesbeck, Subway Tunnel Guided Electomagnetic Wave Popagation at Mobile Communications Fequencies, IEEE Tans. Antennas Popagat., vol. 49, No. 11, pp , Nov [4] Y.P. Zhang, Novel Model fo Popagation Loss Pediction in Tunnels, IEEE, Tans. Vehicula Technol., vol. 52, No. 5, pp , Sept [5] C.L. Holloway, D.A. Hill, R.A. Dalke, and G.A. Huffod, Radio Wave Popagation Chaacteistics in Lossy Cicula Waveguides Such as Tunnels, Mine Shafts, and Boeholes, IEEE, Tans. Antennas Popagat., vol. 48, No. 9, pp , Sept [6] D.G. Dudley, Wieless Popagation in Cicula Tunnels, IEEE Tans. Antennas Popagat., vol. 53, No. 1, pp , Jan [7] D.J. Cichon, and T. Kune, COST Telecommunications COST Action 231 Digital Mobile Radio Towads Futue Geneation Systems Final Repot, Chapte 4, Popagation Pediction Models, COST 231 TD(95), pp , Sept [8] J.P. Beenge, A Pefectly Matched Laye fo the Absoption of Electomagneit Waves, J. Computational Physics, vol. 114, pp , [9] D. White, M. Stowell, J. Koning, R. Rieben, A. Fishe, N. Madsen, Highe-Ode Mixed Finite Element Methods fo Time Domain Electomagmetics, UCRL-TR , Intenal Repot, Feb [10] D.M. Sullivan, Electomagnetic Simulation Using the FDTD Method, Wiley-IEEE Pess, pp.8, Jul [11] J. Dunlop and D. G. Smith, Telecommunications Engineeing (3 d Edition), Chapman & Hall, [12] C.A. Balanis, Antenna Theoy Analysis and Design, 3 d ed., Chapman & Hall, [13] Y. Wu, M. Lin, I.J. Wassell, Path Loss Estimation in 3D Envionments using a Modified 2D Finite- Diffeence Time-Domain Technique, 7th Intenational Confeence on Computation in Electomagnetics, Ap [14] A. Taflove and S.C. Hagness, Computational Electodynamics the Finite-Diffeence Time-Domain Method, 3 d ed., 2005.