无线传感器网络节点的物理层设计_英文_钟子果.pdf

Journal of SoutheastUniversity English Edition V o l 22 No 1 pp 21 25M ar 2006 ISSN 1003 7985 Physical layer design of wireless sensor net work nodes Zhong Ziguo Hu A iqun W ang Dan D epartment ofRadio Engineering SoutheastUniversity N anjing 210096 China Abstract M ajor consideration di mensions for the physical layer design of w ireless sensor network W SN nodes is analyzed by comparing different w ireless communication approaches diverse mature standards i mportant radiofrequency RF para m eters and various m icrocontroller unit MCU solutions An i mp lementation of theW SN node is presented w ith experi mental results and a novel one processorworking at two frequencies energy saving strategy The lifeti me esti mation issue is analyzed w ith consideration to the periodical listen required by commonW SN media access control MAC algorithm s It can be concluded that the startup ti me of theRF which deter m ines the best sleep ti me ratio and the shortest backoff slot ti m e ofM AC theRF frequency and modulation methods which deter m inate the RX and TX curren t and the overall energy consumption of the dual frequencyMCU SOC syste m on chip are themost essential factors for theW SN node physical layer design Key words w ireless sensor network node physical layer radio frequency energy consumption node lifeti m e Received 2005 06 30 Foundation item The National High T echnology Research and D eve lopment Program of China 863 Program No 2003AA 143040 B iographies Z hong Z iguo 1981 male g raduate Hu A iqun corres ponding author male doctor professor aqhu seu edu cn The w ireless sensor network W SN has been shown to be a prom ising prospect in various applica tions Initial research on W SN w as mainly motivated by m ilitary applications e g s m art dust 1 funded by the Defense Advanced ResearchPro jects Agency DARPA W ith the explosive development of m icro electro m echanical system s MEM S integrated cir cuits andw ireless comm unications low cos t low power consu m ption and tiny size components are off the shelf now W SN has becomem ore available in civil applica tions e g ani mal observation 2 glaciermonitoring 2 nuclear power plantm onitoring 3 etc The design of W SN nodes presents a big cha l lenge because of the li m ited resources of each node A lot of w ork has been done in this area such as the M ote s fam ily in the smart dust project in Berkeley the P icoRadio projec talso in Berkeley the i Badge in UCLA M IT s PushPin andW iseNet in Sw itzerland M assive deploym ent of the nodes requires the cost of each node to be as low as possible the inconvenience of replacing the batteries and the relatively slow ly i m proved battery technology advocate m ini m al power consu m ption or energy extracting techniques small size is al w ays a good advantage during deploym en t In the W SN the communication data rate isusually not an es sential factor since W SN is deployed usually not for data transm ission but for event detecting In this paper W SN node physical layer design considerations are discussed in detail and an i mplemen tation is presented follow ed by test results and analysis of the node s lifeti m e esti m ation 1 Design Consideration D i mensions W ithout doub t a good W SN node design needs not only consider the physical layer but also other op ti m ized design spaces 2 such as the opti m ized MAC layer algorithm the smart route strategy and the eff i cient softw are system etc In this section w e w ill con centrate on some basic consideration di mensions for the design of the physical layer M assive m anufacture can settle the cost problem here energy is themajor factor 1 1 Architecture of the node Fig 1 show s the system architecture of a generic W SN node Fig 1 W SN node architecture The basic parts are composed of four entities The pow er supply and management mode l usually some kind of battery provides the nodes electrical energy The communication block com prises a w ireless com munication channel in the for m s of radio laser infra red etc The processing and control unit should at least include am icrocontroller a storagememory and the in terface to the sensor block The last block is a group of sensors or other deviceswhich co llect the desired data Since different applicationsemploy diverse sensors this paper w ill not go further than the sensor par t 1 2 Comm unication approaches For the W SN according to specific application scenes large quantities of nodesm ight be deployed e i ther indoors or outdoors either deliberately or random ly either fixed or w ith mobility Therefore the w ire less communication approaches usedinthe W SN should m eet various and even rough environments Here five comm on w ireless approaches are listed and analyzed 1 Infrared Infrared isw idely used in remote control applica tions The advantages of infrared are as follow s the cost of the infrared dev ices is extrem ely low no anten na is needed so a tiny size of the node can be achieved The defects are as follow s not energy eff i cient a 10 m 20 directional trans m itter consu m es about 60 mW short communication range usually less than 1m of the diffused type and severalm eters of the directional type poor ability of traveling round or penetrating the obstacles In su m infrared isnot a good so lution for theW SN 2 Laser opticalm ethods Them erits of laser are as follow s energy efficient because of itsextremely focused energy security due to no broadcasting trans m ission no antenna is needed a small laser em itter is also cheap The shortcom ings are as fo llow s line of sight LOS is needed the nodes must be carefully placed so that the laser beam can be received it is sensible to atmospheric conditions Due to those shortcom ings laser is not attractive in the W SN 3 Inductive and capacitive coupling M ost passive RFI D syste m s use this approach Its disadvantage is s mall working distances usually only about tens of centm i eters TheW SN requiresa commun i cation range from severalmeters to hundreds of meters 4 Sound Sound or ultrasound is typically used for commu nication underw ater but not formostW SNs 5 Radio frequency RF The benefitsof RF are as follow s om n i direction trans m ission w ith ability of diffracting or penetrating the obstacles NLOS low er pow er consumption than that of infrared relatively long function range 1 100m 0 dBm The biggest challenge for RF is the antenna design To opti m ize the antenna gain the PCB antenna needs 4 in length forwhip or stub shape is the wavelength of the carrier which adds to the size of the node The advantages of RF have made it an ideal testing platfor m for theW SN Actually m ost of theW SN projects e m ployed the RF solution 1 3 Radio comm unication solutions 1 M ature short distance radio standards WPAN w ireless personal area networks IEEE 802 15G roup and WLAN w ireless local area ne t works IEEE 802 11 Group are now adays m ajor short range w ireless netw orks The 802 11 series 802 11b a g etc is developed for a w ireless Ethernet w hich does not take into consideration the energy but has a high data rate 11 to 54M bit s con sum ing m ore than 1Win active m ode and m ore than 50 mW 4 in pow er sav ing mode Therefore the WLAN dev ices are not suitable for the W SN nodes 802 15 series includes 802 15 1 2 B luetooth 802 15 3 UWB and 802 15 4 Z igBee The UWB ai m s atm ulti m edia applicationsand isnot ener gy saving Som e projects did use Bluetooth to i mple m entW SN 5 but because of its short range about 10 m and lack of extrem e pow er efficiency refer to the B luetooth chip BRF6150 in Tab 1 the B luetooth is also not a rem arkable solution Z igBee is regarded as the most suitable standard for the W SN The PHY layer of Z igBee is al m ost suitable forW SN not the best see theZ igBee chips in Tab 1 TheMAC layer is not energy efficient enoughforits traditional CS MA CA algorithm 6 2 Off the shelf components Lots of SDR short distance radio chips are available today T ab 1 lists several m arket leading chipsw ith their features 3 RF device considerations There are mainly five di m ensions needed to be considered to m ake a good cho ice for the RF par t Operation frequency The W SN is expected to operate in the 433MHz 868MH z 2 4GHz 5GHz IS M bands In genera l higher bands can prov ide high er data rates and allow for s m aller antenna size but a higher frequency needs a higher clock rate 7 w hich means higher power consu m ption and brings greater oscillator phase noise 7 w hich reduces the reception 22Zhong Z iguo HuA iqun andW ang Dan sensitivity In T ab 1 the 2 4 GH z devices generally consum em ore pow er especially in RX mode and w ith low erRX sensitivity M odulation The m ore com plex the modula tion is the more pow er is needed Constant envelope m odulations that allowhigh efficient nonlinear PA s are favored 7 Therefore FSK is a good so lution for W SN s stringent low pow er requirem en t Startup ti m e W ith regard to the MAC layer strategy the nodes inW SN w ill be in a periodic listen and sleep cycle 8 in order to save pow er According to the research on ourMAC layer protoco l we conclude that the startup ti me of the component plays an essen tial ro le in theW SN Since the startup ti me is the ti m e period that the node needs to sw itch betw een sleep state and active state it deter m ines the best ti me ratio of sleep mode in the cycle the evoking ti m e needed for waking up the neighboring nodes and the m ini mal ti m e slot of theMAC backoff strategy Sleep current The node inW SN w ill be in sleep state m ost of ti m e therefore the sleep current is an i m portant parameter in the design RSSI receive signal strength indicator The RSSI signal is useful for fast carrier sense which is an indispensable part in theMAC backoff algorithm Tab 1 Radio com ponents comparison ComponentsCC1010 8051 RF CC2420 Z igBee nRF24E1 8051 RF BRF6150 Bluetoo th M C13192 Z igBee TDA 5250 Frequency GHz0 3 to 12 42 42 42 40 868 M odulationFSKO Q PSK D SSS GFSKG FSK FHSS O QPSK D SSS A SK FSK Vo ltage V2 7 to 3 62 1 to 3 61 9 to 3 61 7 to 3 62 0 to 3 42 1 to 5 5 TX current mA10 4 0 dBm 433MHz 17 4 0 dBm 13 0 dBm 2530 0 dBm 9 4 6 dBm RX current mA9 1 433MHz 19 718 250K bit s 37379 3V FSK RX sensitiv ity dBm 107 1 2 K bit s 94 90 85 92 109 Data rate K bit s 1 76 8 max 2501 000 max 723 225064 max Startup ti me s250 RX 270 TX 192 200NA 2 5 1041 2 104 Sleep current A0 2 RF only 12 8051 RF 252 59 RSSIYesY esN oNAN oY es 1 4 Processing units and sensor interfaces TheMCU in aWSN node serves as both the node controller and the data processor H ere the processing ability of theMCU is not an acute issue but since the MCU should never be turned off most of the ti me in i dle mode it occupies a big part in power consump tion Thus the power consumption in idle state is an i mportant parameter for the choice ofMCU On the oth er hand the cos t the power and the size li m itations of the node argue that theMCU can only be an 8bit or 16 bit low endm icroprocessor SOC solution theMCU chip integrates the code and data memories the ADC con verters andUART SPI GPIO interfaces There is another recently appearing solution the MCU embedded in the RF chips such as CC1010 nRF9E5 nRF24E1 etc Compared w ith the traditional solution of MCU connecting w ith an RF chip the ad vantages of this integration are as follows better system reliability due to fewer chips and fewer inter chip con nections lower power consumption for inner chip data transfer and shared clock oscillation parts smaller PCB size T ab 2 lists the comparisons of som e practical MCU solutions Tab 2 M CU comparisons MCUATm ega16L 8 bit P89LPC912 8 bit M SP430F147 16bit CC1010 8051 core nRF24E1 8051 core M axi m u m perform ance8M IPS 8MH z9M IPS 18MHz8M IPS 8MHz5M IPS 24MH z5M IPS 20MHz Flash EEPROM RAM16KB 512B 1KB1KB 0 128B32KB 256B 0 1KB 32KB 0 2KB 128B0 0 4KB 256 B ADC8 channe l 10 bitCom parator12 bitADC comparator3 channe l 10 bit9 channe l 10bit IOPWM UART SPI IIC GPIOPWM UART SPI GPIOUSART SPI GPIOP WM UART SPI GPIOPWM UART GPIO V oltage V2 7 to 5 52 4 to 3 31 8 to 3 62 7 to 3 61 9 to 3 6 Current active idle 3 8mA 1 2mA 4MHz 3V 1 1mA 0 35mA 1MH z 3V 7mA 1 5mA 12MHz 3 6V 4mA 1mA 7 373MHz 3 6V 280 A 1 6 A 1MHz 2 6V 14 8mA 12 8mA 14 74 56MHz 1 3mA 29 4 A 32KHz 3mA 2 A 16MHz 3V 1 5 Power supply andm anagement issues The power supply block consists of a battery and relative circuits Batteries can be divided into two clas ses non rechargeable and rechargeable In genera l the non rechargeable has higher energy density and lower cos t Itmay be possible to extend the lifeti me of a sen sor node by extracting energy from the environmen t for instance solar ligh t vibrations RF etc But the high cost of these techniques hinders the i mplementation Power management strategies are considered in 23Physical layer design ofw ireless sensor network nodes prior research 7 9 10 Although it is usually accom plished by the application layer there are also some methods in the physical layer to reduce power con sumption In Refs 9 10 dual processor solutions are suggested for power saving because the power consump tion of the m icroprocessor is in direct proportion to the frequency at which it works One high frequency pro cessor which serves for the RF mode l can be in sleep mode when there is no data to transfer one low fre quency processor controls the node and does some data processing However this designresults in several problems the cost of the node ascends the reliability of the node descends the size of the node PCB increa ses There are some devices such as CC1010 that can use two crystal clock inputs one is 32 kH z and the otherm ay be 3MHz and we can make the node with only oneMCU but work at two different clock frequen cies so as to achieve the same results asusing dual pro cessors One processor working at two frequencies is a much better solution than dual processors 2 An I mplementation of theWSN Nodes 2 1 Solution and design According to the considerations mentioned above to m ini m ize power consu mption node size and node cos t and to guarantee the reliability of the device we prefer an SOC solution For the frequency band 433 868 MHz is chosen It is best that there is no hardware or fir mwareMAC layer in the chip so that our ownMAC and route protocols later can be deployed on thisWSN node platfor m So the Z igBee devices are deserted and the existence of RSSI signal and short startup ti me are required F inally we accomplished our prototype WSN node see Fig 2 using CC1010 The node integrated a compact PCB antenna a temperature sensor and a DC DC converter The RF part of CC1010 was configured to work at 868 277MH z Fig 2 W SN node 2 2 Experi m ental results W e conducted a node to node open air LOS data transfer test onNovember 20th 2005 The resultswere the reliable communication distances bet weent wo nodes which are shown in Tab 3 Tab 3 Po int to po int communication range LO S m Data rate bit s 1 RF pow er dBm 20 100 19 200253646 9 600333749 2 400474858 600484981 3 Lifeti me Esti mation There are fourworking statesof each node and the power consumptions of each state are listed in Tab 4 The ti m e ratio in Tab 4 means the assumed ti me per cent of each state We took into consideration the per i odic listen and sleep cycles 8 required by the MAC layer algorithm so state 3 owns a ti me ratio as high as 5 and 1 The lifeti meTlifeti mecan be esti m ated as Tlifeti me Cbattery Iaverage 1 Iaverage 4 i 1 PiIi 2 where Cbatteryis the capacity of the battery mA h Iaverageis the average current consumption mA Piis the ti me ratio in the state i and Iiis the current con sumption in the state i Tab 4 Four states of the node w ith the pow er consu m ption of each state 868 277MHz M CURFState Current mA Ti m e ratio A T i m e ratio B Iavarage mALifeti m e h Idle 29 4 A 32 k H z Power down 0 2 A 1 Sleep0 029 693 598 80 906286 CR2330 Active 3 7mA 3 686MH z Power down 0 2 A 2D ata process3 710 1 ratio A 1434 CR123A Active 3 7mA 3 686MH z RX 11 9mA 3RX15 6510 2011427 CR2330 Active 3 7mA 3 686MH z TX 8 6mA 20dBm 4 TX12 30 50 1 ratio B 7138 CR123A As shown in Tab 4 the lifetm i e was calculated based on two types of batteries CR2330 L iMn coin ba t tery 260mA h max 3mA continuous curren t CR123 A L i Mn column battery 1 300mA h max 1A continuous cur rent and under different state tm i e ratios From the life tm i e estm i ation we can confir m that the most essential design factors for power saving are as follows the RF star tup tm i e which decideswhich tm i e ratio can be achieved by the MAC algorith m the RX current including MCU active current since the node has to listen to the media 24Zhong Z iguo HuA iqun andW ang Dan periodically and the sleep curren t which occupies the biggest tm i e par t The lifeti me evaluation in Re f 9 is proble m atic for its lack of considering the node s periodical RX state The mult i hop feature of the WSN requires that the node should act as both a data source and a ne t work router so each node should periodically listen to them edia preparing for data relay This error in Re f 9 leads to themuch longer butm istaken esti m ated l i feti me 13 1 years 4 Conclusion In this paper the basic di mensions for the WSN node physical layer design are discussed and an i mple mentation is presented togetherw ith the test results and lifeti me evaluation analysis To meet the energy size and cost design challen ges every aspect of the node system should be carefully considered For the physical layer the startup ti me of the RF which decides the best sleep ti me ratio and the shortest backoff slot ti m e ofMAC can be achieved The RF frequency and modulation methods which determ i nate theRX and TX curren t the dualworking frequen cyMCU SOC which contributes to the reliability and o verall energy consumptions are the threemost essential factors for this challenge References 1 K ahn JM K atz R H Em erg ing challenges mobile networ king for s mart dust J Journa l of Comm