Double Directional Channel Characterization on Board Ships

Transcription

1 Double Directional Channel Characterization on Board Ship H. Kdouh, H. Farhat, C. Broueau, G. Zaharia, G. Grunfelder, G. El Zein Intitut d Electronique et de Télécommunication de Renne Renne, France huein.kdouh@ina-renne.fr T. Tenoux, Y. Lotanlen Siradel Renne, France ttenoux@iradel.com Abtract Due to the metallic tructure of deck, bulkhead and watertight (WT) door, wirele communication are a eriou challenge in the particular environment of hip. In order to deploy reliable hipboard wirele network, wirele device (acce point, router, enor node, etc.) mut be located at trategic location enuring full radio coverage and network connectivity. Strategic location can be determined from the identification of the main propagation direction of electromagnetic (EM) wave within a hip. Thi paper preent the reult of a radio propagation meaurement campaign performed on board a hip. A dual-band Multiple-Input Multiple-Output (MIMO) channel ounder and antenna array have been ued. Meaurement data have been proceed with a claic beamforming technique and a high reolution algorithm to extract dominant path. A ray-tracing baed imulation tool ha been ued to undertand meaurement reult. Obtained reult are ued for optimal placement of radio device when deploying hipboard wirele network. Keyword-propagation; channel ounding; ray-tracing; hip I. INTRODUCTION Ship are an important part of modern ytem, which are widely ued in commercial and military purpoe. Modern hip are equipped with automatic alarm and monitoring ytem, which control and enure afety and accuracy of the whole hip operation. Current hipboard monitoring ytem ue extenive length of cable to connect everal thouand of enor to control unit. In addition to the high cot and weight due to wire intallation during hip contruction, veel repreent a complex and harh environment in which extenive length of wire are vulnerable to heat, moiture and toxic agent [1]. Hence, applying wirele technologie uch a Wirele Senor Network (WSN) to hipboard monitoring ytem may be a cot-effective and urvivable olution. Furthermore, wirele ytem are eaily and inexpenively reconfigured. Moreover, wirele communication can be ued for other application on board hip. Ferry companie, for example, aim to enhance their paenger' atifaction by equipping their boat with WiFi network. Cordle phone are very ueful for communication between crew member. However, wirele communication are a eriou challenge in the particular harh environment of hip. In fact, everal factor may limit the performance of wirele ytem on board hip [2]. Firtly, bulkhead and door are made of metal, mot often teel. Although the teel i not a perfect conductor, it can everely decreae the power of radio wave. Another limiting factor i the multipath propagation: a tranmitted EM wave could be reflected, cattered or diffracted by different object leading to everal delayed copie to the receiver. The exitence of multiple copie of the tranmitted wave may caue diturbing interference ignal at the receiver. In order to deploy reliable wirele network on board a hip, wirele device mut be located at trategic location enuring optimal radio coverage and network connectivity. Strategic location can be determined by identifying the main direction of EM wave within a hip. Few work have tudied the EM wave propagation on board hip [2-5]. However, patio-temporal characterization of EM wave propagation within thi particular environment ha not been done yet. Exiting tudie have only conidered the poibility of wirele communication without determining the direction of EM wave propagation. In [6], we have conducted Continuou Wave (CW) meaurement on board everal ferrie to verify the poibility of intra-, inter-compartment and inter-deck radio communication. Meaurement reult helped u to determine the path lo exponent in different hipboard environment and different communication configuration. However, CW meaurement cannot determine the Direction of Departure () and Direction of Arrival () of EM wave. Double directional or MIMO meaurement may be an efficient technique to etimate and by uing antenna array at both link end. In thi paper, we preent the reult of a meaurement campaign uing a dual-band MIMO channel ounder [7]. Thi meaurement ytem, developed in the IETR laboratory, wa already ued for outdoor, indoor and outdoor-to-indoor MIMO channel ounding [8]. The performed meaurement provide a patio-temporal characterization of the EM wave propagation within a hip. Collected data are analyzed uing a claic beamforming technique and a high reolution algorithm. Obtained reult are then compared to imulation reult baed on a raytracing tool. Thi comparion help u to undertand direction of propagation of EM wave in typical hipboard environment uch a engine room, parking and paenger deck. Obtained reult are ued to enure an optimal placement of radio device when deploying hipboard wirele network. The remainder of thi paper i organized a follow: Section II decribe the methodology and meaurement etup ued in thi tudy. Section III preent and analyze the obtained meaurement and imulation reult. Finally, everal concluion are drawn in Section IV. 224

2 II. METHODOLOGY AND MEASUREMENT SETUP The dual frequency band (2.2 and 3.5 GHz) ued channel ounder tranmit a pread pectrum waveform uing a periodic M-equence. It ha an 11.9 n temporal reolution for 100 MHz ounding bandwidth. The dynamic range i 50 db for the 1023 code length. Synchronization between the tranmitter (Tx) and the receiver (Rx) i achieved with highly table 10 MHz rubidium ocillator. Different impule repone length can be choen from 1.27 to µ, depending on the ounding bandwidth and the code length. A an example, for 100 MHz bandwidth and 1023 code length, the recorded impule repone length i µ. Different type of antenna array were developed for thi ounder. At 2.2 GHz, a 4-element Uniform Linear antenna Array (ULA) and a 16-element Uniform Rectangular antenna Array (URA) are ued repectively for the Tx and the Rx to characterize the double directional channel on a 120 beamwidth in the horizontal plan. At 3.5 GHz, a 4-element Uniform Circular Array (UCA) i ued at Tx and a 16- element UCA i ued at Rx. With thi configuration, we can characterize 360 azimuthal double directional channel at both link ide. In order to improve the meaurement dynamic range, power amplifier have been directly integrated in the Tx antenna array, and low noie amplifier have been placed directly behind the Rx antenna array. We aume a quai time-invariant channel during the meaurement. Attention wa paid that no people were moving in the urrounding area. For each Rx location, everal meaurement were taken and averaged to reduce the noie effect. The collected channel data were tored on a laptop for pot-proceing. The meaurement objective wa to characterize the double directional channel impule repone h [9]. In the cae of omnidirectional antenna at Tx location r Tx and Rx location r Rx, it could be expreed a: h( rtx, rrx,,,,, ) S (1) h ( r, r,,,,, ) 1 Tx Rx where S i the number of multipath component, i the delay, Ө and Ф are the azimuth and elevation angle of and. With the plane wave aumption, the contribution of each multipath component i: h ( r Tx, r j a e. ( ). ( ). ( ) (2) where Rx (,,,, j a e, ). (, ),, ), i the multipath component complex amplitude of the component. Analyi of collected data i performed with a claic beamforming technique [10] and the high reolution Space- Alternating Generalized Expectation-maximization (SAGE) algorithm [11], in order to determine dominant path which propagate between Tx and Rx. We gave a pecial attention to the and, which are the mot important parameter to determine the direction of propagation. In order to preciely determine thee direction, a imulator of EM propagation [12] ha been ued. Thi imulator, baed on a 3D ray tracing algorithm [13], compute ucceive EM interaction thank to the Geometrical Optic (GO) or Uniform geometrical Theory of Diffraction (UTD). Thi computation i aociated with a 3D geometrical decription of the conidered cene, including the tructure and the poition of bulkhead, floor, ceiling and WT door. A maximal number of 4 reflection and one diffraction are conidered for each imulation cenario. When diffraction i involved, only one reflection can occur before and after it. III. MEASUREMENT RESULTS AND ANALYSIS Thi ection preent and analyze the obtained meaurement and imulation reult. Three typical environment have been tudied on board the Armorique ferry: the engine room, the parking and the paenger deck. The engine room of Armorique include metallic engine, generator, valve and pump arranged in a complex way. The common bulkhead of thee room are metallic and include liding metallic WT door. The parking i a big hall with metallic wall. A metallic wall located in the middle of the parking divide it into two main part. Vehicle of different type were parked in the parking (car, bue, truck, etc.) when performing meaurement. The paenger deck i compoed of paenger' cabin and corridor. In contrat with the two previou environment, corridor wall and cabin door are not fully made of metal. A. EM wave propagation within engine room In pite of the fully metallic tructure of bulkhead and WT door in the lower deck area of ferrie, CW meaurement have hown that wirele communication between adjacent room remain poible after cloing WT door [6]. A MIMO channel ounder and antenna array are ued to identify the opening allowing EM wave leakage. The tudied environment i the econd deck of Armorique, where liding metallic WT door are ued between adjacent room. Thi environment i highly metallic and confined. It generate everal propagation phenomena (reflection, diffraction, cattering). Meaurement were carried out uing the URA and ULA array at 2.2 GHz. The tudy wa limited to a 120 characterization in the azimuthal plan. It i ufficient to characterize the propagation phenomena through the bulkhead door between adjacent room. Fig. 1 how the layout of the room, location of Tx and Rx in the econd deck of Armorique. Figure 1. Tx and Rx location in the engine room. 225

3 Two WT door, called WT1 and WT2 on Fig. 1, are located on the propagation path between Tx and Rx, which are in No Line of Sight (NLoS) configuration. Meaurement were carried out when WT1 wa opened and when it wa cloed repectively, while WT2 wa maintained cloed. Meaurement data were proceed uing the conventional beamforming. Fig. 2 and Fig. 3 preent the normalized pace-delay power graph for the two meaured cenario (radiu cale of beamformer graph i in n, and the color repreent the normalized amplitude at each angle). Thee graph repreent the patio-temporal channel repone and allow identifying the main direction of energy propagation between Tx and Rx. It can be een that the received energy i not homogeneouly ditributed on the 120 beamwidth. For the two cenario, the angular ditribution of energy are imilar. It can be een that a main beam of energy with about 20 beamwidth (between 125 and 145 ) and other beam around 180 direction are formed at the Tx ide. Moreover, a beam of 50 i formed around 0 at the Rx ide. The main beam formed at the Tx and Rx ide correpond to the energy going to WT1 and coming from WT2 repectively. Thi imilarity between the two cenario (cloed and open door) how that the energy i propagating through the WT door regardle their tatu. The other beam een at 180 at the Tx ide may be probably due to two repective reflection of EM wave on the bulkhead in front and behind the Tx, before penetration through WT1 and WT2 toward Rx. We have preented in [14] imilar meaurement for cenario where the direct path between Tx and Rx i blocked by the WT door. It reult how that radio ignal propagation i made through the opening on the edge of the metallic watertight door. B. EM wave propagation within the parking A tated before, the parking of Armorique i a big hall where all wall, ceiling and floor are totally metallic. A big bulkhead intalled in the middle divide the parking into two part (called lower part and upper part in the following). Two communication cenario were conidered. In the firt one, Tx and Rx are both located in the lower part of the parking (Line of Sight LoS configuration). In the econd cenario, Tx i located in the upper part of the parking and the Rx i located in the lower one. The middle bulkhead block the LoS between the Tx and Rx (NLoS configuration). (a) Figure 2. Beamforming reult when WT1 and WT2 are cloed: (a) At Rx (), (b) At Tx (). (b) (a) Figure 3. Beamforming reult when WT1 i opened and WT2 i cloed: (a) At Rx (), (b) At Tx (). (b) 226

4 The ULA and URA 2.2 GHz are ued in thee meaurement. The reult are proceed uing beamforming and compared to ray-tracing imulation reult. Fig. 4 how the layout of the parking, location of the Tx and Rx and ray tracing reult for the LoS configuration. A guiding effect i clearly oberved in the lower part of the parking due to the middle bulkhead; ome other path coming from the upper part can be detected. The middle bulkhead and the wall of the parking act a a waveguide between the Tx and Rx. Fig. 5 how the comparion between beamforming reult and ray-tracing imulation reult at Tx and Rx. The comparion indicate alo that main beam (in red) are formed in the lower part of the parking. A ignificant agreement between meaurement and imulation reult i found. Fig. 6 how the layout of the parking, location of the Tx and Rx and beamforming and ray-tracing reult for the NLoS configuration. In pite of the middle bulkhead, which block the direct path between Tx and Rx, EM wave are able to achieve the Rx through reflection on the parking wall and the middle bulkhead, and diffraction on the middle bulkhead edge. A in the LoS configuration, a ignificant agreement between meaurement and imulation reult i found. Thi agreement prove that the model of the parking ued for imulation can be conidered a realitic. Figure 4. Ray-tracing reult for the parking in LoS configuration. C. EM wave propagation on the paenger deck Fig. 7 how the layout of the paenger deck and the location of Tx and Rx. Rx i placed in the tairway located at the top left of thi deck. The tairway door i opened. Tx i placed at three location (Tx1, Tx2 and Tx3 in Fig.7).Tx i firtly located in a paenger cabin whoe door i cloed (location Tx1), then in a corridor in the middle of the deck (location Tx2) and finally at the bottom right corner of the deck (location Tx3). In all thee meaurement cenario, Tx and Rx are in NLoS configuration. A thi environment i not totally metallic like the parking and the engine room, no particular propagation direction can be excluded. Thu, the circular array have been ued at the Tx and Rx to determine the direction of propagation in the 360 azimuth. (a) Figure 5. Beamforming and ray-tracing imulation reult for LoS configuration in the parking: (a) at Tx (), (b) at Rx (). (b) Figure 6. Beamforming and ray tracing reult for the parking in NLoS configuration. 227

5 Figure 7. Layout of the paenger deck of Armorique with the location of tranmitter (Tx1, Tx2 and Tx3) and the receiver (Rx). The propagation path parameter are extracted uing SAGE algorithm. Fig. 8 preent the normalized and reult for all meaurement poition. Black arrow indicate direction and the length of each one indicate it normalized amplitude (one tick interval correpond to 10 db). Thi overview how that for the, ome privileged propagation direction can be eaily identified. Concerning the, for the three Tx location, we notice that received energy direction are more equally ditributed on 360. Thi can be explained by the receiver location in the tairway. Due to it metallic wall, thi environment i imilar to a reverberant chamber. One can think that EM wave will arrive mainly through the entrance door of the tairway. After that, EM wave are reflected on the metallic wall of tairway and achieve the receiver from all 360 direction. An adaptation of the ray-tracing baed imulation tool wa made at 3.5 GHz, mainly concerning the dielectric propertie of material contituting the paenger deck (ued for the computation of the reflection, tranmiion and diffraction coefficient). For the ray-tracing imulation, by auming that the propagation i made through the corridor, and becaue of the current limitation of the RT imulator to compute complete path from Tx to Rx, a virtual point of reception i located in the corridor near the Tx poition. The idea i to viualize how the radio wave reach thi point from the Tx. Fig. 9 preent the comparion between the obtained from the meaurement (black arrow) and ray-tracing tool (blue line) for the Tx1 location. In thi meaurement cenario, Tx wa located in a paenger cabin. The main direction of energy propagating from the paenger cabin to Rx poition obtained from ray-tracing preent a ignificant agreement with the meaurement reult. The obtained reult validate our aumption that the propagation in thi environment mainly occur through the corridor. Thee direction how that EM wave penetrate through the thin wall of cabin (which are not metallic) before arriving to the corridor, and then they are guided to the Rx (through reflection). Note that the wall located at the left of the figure correpond to a metallic tructure. Virtual Rx Figure 9. reult for the location Tx1. Fig. 10 preent the comparion between obtained from the meaurement and ray-tracing tool for Tx3. A in the previou cenario, a ignificant agreement i oberved between the main direction flow from the meaurement and the ray-tracing imulation reult. The obtained reult alo validate our aumption that the propagation in thi environment mainly occur through the corridor (but ome wave can propagate through cabin wall). In thi configuration, reflection are oberved on the corridor wall urrounding the Tx3 location before arriving to the receiver. Figure 8. and for the three meaurement cenario on the paenger deck. Figure 10. reult for the location Tx3. 228

6 D. Engineering rule for intermediate node placement A mentioned in Section I, the main objective of thee meaurement i the optimization of wirele device placement when deploying wirele network on board a hip. Thi tudy i a part of SAPHIR project, which aim at applying WSN technologie to hipboard alarm and monitoring ytem. Senor node mut be placed in all hip room and compartment to meaure phyical parameter uch a tank level, water level, temperature, humidity, etc. and then end collected data to central control unit located in the wheelhoue or the control room. A direct communication between enor node and control unit are impoible (due to large ditance and metallic environment), intermediate node mut be located to enure the whole network connectivity. Reult of thee meaurement are ueful for placing intermediate node. Therefore, the following engineering rule are recommended. In the engine room, EM wave propagate mainly through watertight door even when they are cloed. It will be then recommended to place intermediate node in front of watertight door to enure connectivity between adjacent room. In the parking, the wall contitute a wave guide between communicating node. Intermediate node may be placed on the wall of the parking, including thoe of the middle bulkhead, to enure the network connectivity of different node in the parking. In the paenger deck, EM wave propagate mainly through corridor. Intermediate node will may be located in the ceiling of corridor (to minimize the fluctuation due to paenger' movement). IV. CONCLUSION In thi paper, a meaurement campaign wa conducted on board a ferry uing a MIMO channel ounder. The objective of thi meaurement campaign wa to determine the direction of propagation of EM wave in typical hipboard environment. In pite of the totally metallic tructure of bulkhead and WT door in the lowet deck of the ferry, obtained reult how that cloing WT door doe not block totally the radio wave propagation, which i made through the opening on the edge of the metallic WT door. Moreover, the reult how that parallel metallic bulkhead, uch a in the parking, act a a waveguide between Tx and Rx. In pite of NLoS configuration, EM wave are guided through reflection on the metallic bulkhead and diffraction on the bulkhead edge, to the Rx. Meaurement and imulation reult on the paenger deck how that EM wave propagate mainly through corridor on the paenger deck. When the Tx i located in a paenger cabin, EM wave penetrate through the thin wall of cabin before arriving to the corridor, by which they are guided to the Rx. All thee reult are ued to determine engineering rule for enor node placement. ACKNOWLEDGMENT Thi work i a part of the SAPHIR project upported by the Pôle Mer Bretagne and Region Bretagne. The author thank Marinelec Technologie and Brittany Ferrie for the opportunity to conduct thee meaurement campaign. REFERENCES [1] J. P. Lynch and K. J. Loh, A ummary review of wirele enor and enor network for tructural health monitoring, The Shock and Vibration Diget, vol. 38, no. 2, 2006, pp [2] D. R. J. Ete, T. B. Welch, A. A. Sarkady, and H. Witheel, Shipboard radio frequency propagation meaurement for wirele network, Proceeding of the IEEE Military Communication Conference (MILCOM), IEEE Communication for Network-Centric Operation: Creating the Information Force, Virginia, USA, October 2001, pp [3] B-G. Paik, S-R. Cho, B-J. Park, D. Lee, B-D. Bae, and J-H. Yun, Characteritic of wirele enor network for full-cale hip application, Journal of Marine Science and Technology, vol. 14, no. 1, January 2009, pp [4] T. Pilak, J. L. Ter Haeborg, and H. Hanneken, WLAN propagation on the bridge of veel under conideration of material propertie, Proceeding of the IEEE International Sympoium on Electromagnetic Compatibility (EMC Europe), Athen, Greece, June 2009, pp [5] A. Maricotti, M. Sai, A. Qualizza, and M. Lenardon, On the propagation of wirele ignal on board hip, Proceeding of Intrumentation and Meaurement Technology Conference (I2MTC), Autin, Texa, USA, May 2010, pp [6] H. Kdouh, C. Broueau, G. Zaharia, G. Grunfelder, and G. El Zein, Meaurement and path lo model for hipboard environment at 2.4 GHz, European Microwave Conference (EuMC), Mancheter, United Kingdom, October 2011, pp [7] H. Farhat, R. Coquer, G. Grunfelder, L. Le Coq, and G. El Zein, A Dual Band MIMO Channel Sounder at 2.2 and 3.5 GHz, Proceeding of The IEEE International Intrumentation and Meaurement Technology Conference (I2MTC 2008), pp , Victoria, Vancouver Iland, Canada, May 2008, pp [8] Y. Lotanlen, H. Farhat, T. Tenoux, A. Carcelen, G. Grunfelder, and G. El Zein, Wideband Outdoor-to-Indoor MIMO channel meaurement at 3.5 GHz, Proc. of the European Conference on Antenna and Propagation, EUCAP '09, Berlin, Germany, March 2009, pp [9] M. Steinbauer, A.F. Molih, and E. Bonek, The double directional radio propagation channel, IEEE Antenna and Propagation Magazine, vol. 43, Augut 2001, pp [10] M. S. Bartlett, An Introduction to Stochatic Proce, New York: Cambridge Univ. Pre, [11] B. H. Fleury, M. Tchudin, R. Heddergott, D. Dahlhau, and K. Ingman Pederon, Channel parameter etimation in mobile radio environment uing the SAGE algorithm, IEEE Journal on Selected Area in Communication, vol. 17, no. 3, March 1999, pp [12] Y. Lotanlen, G. Gougeon, S. Borie, and A. Sibille, A determinitic indoor UWB pace-variant multipath radio channel modelling compared to meaurement on baic configuration, Proceeding of the 1t European Conference on Antenna and Propagation, EUCAP 06, November 2006, Nice, France, pp [13] A. Corucci, P. Nepa, F. Furfari, P. Barocchi, and A. Buffi, Accuracy limit of in-room localiation uing RSSI, Antenna and Propagation Society International Sympoium, APSURSI '09, Pia, Italy, June 2009, pp [14] H. Kdouh et al., Double directional characteriation of radio wave propagation through metallic watertight door on board hip, Electronic Letter, vol. 48, no. 6, March 2012, pp