Академический Документы
Профессиональный Документы
Культура Документы
ControlwithWirelessHART
Copyright2009HARTCommunicationFoundation Thisdocumentcontainscopyrightedmaterialandmaynotbereproducedinanyfashionwithoutthewritten permissionoftheHARTCommunicationFoundation. TrademarkInformation HARTisaregisteredtrademarkoftheHARTCommunicationFoundation,Austin,Texas,USA. AnyuseofthetermHARThereafterinthisdocument,orinanydocumentreferencedbythisdocument,implies theregisteredtrademark.WirelessHARTisatrademarkoftheHARTCommunicationFoundation.Allother trademarksusedinthisorreferenceddocumentsaretrademarksoftheirrespectivecompanies.Formore informationcontacttheHCFStaffattheaddressbelow. Attention:FoundationDirector HARTCommunicationFoundation 9390ResearchBoulevard SuiteI350 Austin,TX78759,USA Voice:(512)7940369 FAX:(512)7943904 http://www.hartcomm.org
Page2
ControlwithWirelessHART
Page3
ControlwithWirelessHART Fornearlytwodecades,HARTCommunicationhasbeentheprocessindustrystandardforsecure, simple,andreliableoperations.NewcapabilitiesandwirelesscommunicationintroducedwithHART version7,complimentestablishedpracticeandexpandtheuseofHARTCommunicationintonewareas andapplications. WirelessHARTwasdesignedspecificallytosupportthewiderangeofprocessindustryusecasesfrom simplemonitoringtoclosedloopcontrol.Testingandfieldtrialswithwirelessdeviceshave demonstratedthatthecommunicationaccuracy,stability,totalperformance,andreliabilitycanmeet thedemandsofindustrialprocessmonitoringandcontrolapplications. Controlapplicationrequirementsforsamplingintervals,jitter,andlatencywerespecificallyaddressed anddesignedintotheWirelessHARTtechnology.Infact,controlperformancewithWirelessHARTcanbe comparabletothatofawiredsystemusingtraditionalfieldbuses.Let'slookatsomeofthefactorsthat cancomeupinconsideringWirelessHARTforcontrolapplications. Let'slookatsomeofthefactorsthatcancomeupinconsideringWirelessHARTforcontrolapplications. Samplingintervals WirelessHARTallowssamplingintervalsthatmeettherequirementsofmostcontrolloopswhileatthe sametimeminimizingtheimpactonfielddevicesthatmaybepoweredbyabattery. Thetypicalruleofthumbisthatfeedbackcontrolshouldbeexecuted410timesfasterthantheprocess responsetime,whereresponsetimeequalstheprocesstimeconstantplusdeadtime. Becausemeasurementsystemsareoftenunsynchronizedwiththecontrolsystem,measurementvalues aretypicallysampledasmuchas210timesfasterthantheprocesscanrespond(Figure1below).
Page4
Figure2
Morefrequentcommunicationofmeasuredvaluesiscertainlypossibleandinthecaseoflinepowered devicescanworksimilartowirednetworks.Inthecaseofbatterypoweredordeviceswhereenergy conservationisimportant,WirelessHARToffersuserstheopportunitytofindanapplication'soptimum balancebetweencommunicationintervalsandbatterylife. LatencyandJitter Effectivecontrolrequirestimelyaccesstomeasurementandcontrolinformation.Asystem'sabilityto meetitscontrolperformancerequirementscanbeaffectedbothbydelays(latency)andvariation(jitter) inwhentheinformationisavailable. Insomesystems,latencyandjittercanstartwiththetimingofthemeasurementsthemselves.But WirelessHARTisatimesynchronizedprotocol,witheverydevicehavingacommonsenseoftime accurateto1millisecondacrosstheentirenetworkacapabilitynotavailableinmanyotherprotocols. Themeasurementsoftwareandcircuitryusethissenseoftimeinschedulingmeasurements,allbut eliminatingdelaysandvariationinmeasurementtiming. Page5
Page6
Changingtheroutethedatatravelscancontributetovariationincommunicationtime(jitter).Although eachadditionalhopincreaseslatency,intypicalapplicationstheaveragedelayiswellwithinthe requirementsforcontrol.Wecanillustratethiswithanexample. Inmostcases,aWirelessHARTnetworkwillbeabletoretryafailedmessageinthenexttimeslotorthe onefollowing.Forourexamplewellassumeittakes10millisecondstoprocessamessageandassignit toanothertimeslot.PathAinFigure5couldthereforeproduceasmuchas50millisecondsoftotal latency(10ms+[10ms+10ms]+[10ms+10ms]).PathBhasthesamenumberofhopsandthusthe samecommunicationslatency.ButPathChasanadditionalhop,bringingtotalcommunicationslatency to70milliseconds.Thistimingdifferenceintroducesa20millisecondjitterinthecommunications.(In manycasestheroutingdevicewillbeabletoretryinthenextslot,whichwouldreducethetotal latenciesto30millisecondsforPathsAandBandto40millisecondsforPathC.) Experienceinhundredsofwirelessfielddeviceinstallationsshowsthatcommunicationslatencyon averageismuchlowerthaninthisexample.Inrealplantsettings,typically30%ofthedevices communicatedirectlywiththegatewayornetworkaccesspoint(10milliseconds)andabout50%are onehopaway(30milliseconds).Theremaining20%maybe34hops.Usingthesenumbersfromactual plantinstallations,theaveragelatencytimewillbeabout30milliseconds. Page7
ControlwithWirelessHART Existinginstallationsalsoshowthatnetworkreliabilityistypicallygreaterthan99%,sothelatencytime willnotvarysignificantlybetweencommunicationseffectivelyeliminatingtheeffectofjitter. Butisitfastenoughforcontrol? Communicationslatencydoesnotaffectcontrolaslongasthedelayissmallcomparedtotheprocess responsetime.Appropriateschedulingoftransmissionsacrossthetimeslotsinamacrocyclecanensure thedatareachesitsdestinationwhenneeded. Forgoodcontrolweneedtobeabletoreadthecontrolmeasurement,communicatethemeasurement toacontroller,executethecontrolfunction,andcommunicatetheoutputbacktothetargetinonehalf theprocesstimeconstant.Mostcontrolloopsare1secondormore,sofora1secondcontrolloopwe wouldneedtobeabletoperformallofthesestepswithin500milliseconds. Let'sseehowthisworkswiththeexamplecontrolloopshowninFigure6.Inthisexample,the measurementisprocessedinthefielddevice,thecontrolalgorithmrunsinthegateway,andthe actuationoccursinavalvethat'sthesame"hopdepth"fromthegatewayasthemeasurementdevice. ThetotalspanofthecycleincludestheAI,PID,AO,andcommunicationtimes. Figure6
ControlwithWirelessHART Figure7
Asthisexampleshows,500millisecondmacrocyclesareeasilyachievableevenwhenmultiplehopsare assumedinthecommunication.Thisisfastenoughfortypicalcontrolloops,whichinmostcasesare muchslowerthanourexample. (Thediagramalsoshowsthatthereisalmostnoimpacttothebandwidthofthesystem.Infact,less than12%ofavailableslotsareneededtodo10highspeedcontrolloopsinparallel.) Inthisexampleweillustratedwhatwouldhappenifthecommunicationstook70milliseconds.As mentionedearlier,however,actualplantexperienceshowsthataveragelatencytimesareabout30 milliseconds.Using30millisecondsinourcalculationsreducestheloopexecutiontimetolessthan100 millisecondsandreducesthenumberofcommunicationsinthenetwork.Ifitisimportanttoreducethe delaysintroducedbymultiplehops,additionalnetworkaccesspointscanbeused. Itisalsopossibletofurtherreducecommunicationlatencyandaddresshigherspeedcontrol applicationsbyusingpeertopeercommunicationsbetweenfielddevices.Runningthecontrol algorithminafielddeviceeliminatestheneedforwirelesshopsbetweenthatdeviceandagateway residentalgorithm.Suchanarrangementmayalsouselessbandwidth,allowingformultiplecontrol loopswithminimalimpacttooverallbandwidth.Ofcourse,usingthisstrategyisdependentonwhether thereareadditionalinteractionsbetweenthecontrolloops. Thisexampleusedwirelessnetworklayoutsthatweremorecomplexthanwhatexperiencehasfoundin actualplantenvironments.Wecouldhavealsoincorporatedmultipleaccesspointstoshorten communicationpaths,andallocatedadditionalcommunicationresourcestofurtherenhancethe effectivenessoftheWirelessHARTnetwork.AndbecauseallWirelessHARTmeasurementsincludea timestamp,wecouldhaveusedthetimestampinthecontrolalgorithmtofurtherreducetheimpactof anylatencyandjitter. Conclusion Evenwithoututilizinganyoftheseadditionalfeatures,theexampleshowsthattheoverallcontrol performanceofatypicalWirelessHARTnetworkiscomparabletothatoftraditionalwiredfieldbuses. TheWirelessHARTprotocolallowsforsecure,highlyreliable,lowlatencycontrolwithalmostnoimpact onthebandwidthandabsolutelynoimpactonprocessperformance. WirelessHARTissimple,reliable,andsecure. Page9