Introductio
ThedesignconceptwasdirectlyinheritedbytheU.S.AirForce’sSAGEairdefensesystemandhadahugeimpactoncommercialcomputersinthe1960s.
Background
InWorldWarII,inordertotrainbomberpilots,theUSNavyaskedtheMassachusettsInstituteofTechnology(MIT)whetheritcoulddevelopacontrolledflightsimulationComputer.Theoriginalideaofthemilitarywastousethecomputertoreflectthedatageneratedbythepilot'ssimulationoperationsonthedashboardinrealtime.Differentfromtheprevioussimulationequipment,themilitaryrequiresthatthecomputershouldbedesignedbasedonaerodynamicsandbeinfinitelyclosetotherealobjectinordertoconductvariousaviationtraining.
MITgaveapositiveanswertothemilitary.SotheNavybegantoprovidefundingfortheprojectunderthenameoftheCycloneProject,andJayWrightForresterwasselectedastheprojectleader.TheR&Dstaffinitiallydevelopedalarge-scaleanalogcomputer,whoseaccuracyandflexibilitydidnotmeetthestandards.Largeroperatingsystemshavebecomeinevitableforresearchanddevelopment,butthedifficultyisunimaginable.In1945,JerryCrawford,amemberoftheprojectteam,afterwatchingthetrialrunofEniac,proposedthatadigitalcomputercouldbeusedasasolutionfortheproject.Theadvantageisthatitispossibletoimprovetheaccuracyofsimulationbyreplacingadditionalpartswithadditionalprograms.Themainstreamviewatthattimebelievedthatcomputershadextremelyhigh-speedcomputingspeeds,andthatnomatterhowcomplexsimulationswere,theycouldtheoreticallyberealized.
Thecomputeratthetimewasonlyabletoperformbatchoperationsforonetaskatatime.Thecalculationdataisinputinadvance,andthecomputerusestheinputdatatoperformcalculations,andfinallygeneratesthecalculationresults.Butthisisnotenoughtomeettherequirementsofthecyclonesystem.Theenvisionedcyclonesystemneedstoperformcontinuousoperationsonreal-timechanginginputdata.Amongthem,speedbecomesthemostcriticalissue.Inthepast,computersprintedoutthecalculationresults,whichwasextremelytime-consuming,andthespeedofthecyclonesystemcouldnotbeincreased,whichmeantthatthecomplexityofflightsimulationwouldbegreatlylimited.
Singula
DesignandManufacturing
In1947,Forresterandotherscompletedthedesignofahigh-speedbuilt-incomputer.Computersatthattimemostlyusedthe"bitseries"modeofoperation,thatis,thecharactersofeachcodewerecalculatedonebyone,andthelengthofthecodewasusually48bitor60bit.Duetothepoorperformanceofthismode,Cycloneswitchedto"bitparallel"andinstalledacomputingpathcapableofparallelprocessingof16bits.Ignoringthestoragespeed,thecalculationspeedisequivalentto16timesthatofmainstreammodels.Almostallcomputersnowadayshaveadoptedthismode,buttheparallelscalehasbeenexpandedto32bitto64bit.
Therotaryfanfirstassignsastorageaddressforeachinstructionandthenperformscalculations.TheprincipleissimilartothatofcalculatorsusingreversePolishnotation,thatis,alloperatorsarelocatedaftertheoperands,andtheonlydifferenceisthatthereisnostackinthecyclone.Inordertoutilizethelimitedstoragespaceasmuchaspossible,theaddresswidthofthecycloneissetto11bit,and5bitsrequiredtoidentify16to32commandsareadded,totaling16bits.ItissaidthataftervonNeumannlearnedthatthewordlengthofthecyclonewasonly16bits,heimmediatelylostinterestinthecyclone.
Inthefollowingyear,thephysicalmachinebeganproduction,withatotalof175peopleparticipating,70ofwhomweretechnicians.IttookthreeyearsandcameoutonApril20,1951.ThecyclonewasfirstdevelopedattherequestoftheNavy,butitwaseventuallyusedbytheU.S.AirForce'sSAGEsystem.TheNavyhastoprovidemillionsofdollarsinfundingforthisprojecteveryyear,andhaslostinterest.In1949,theSovietUnion'sfirstatomicbombwassuccessfullytested.Inresponse,theUSAirForceeventuallytookthecycloneintoitsarms.
"magneticcore" ofthecyclone
Theinitialcalculationspeedofthecyclonewasonly20kips,andthereisstillalongwaytogobeforeitcanbeputintopracticaluse.Theproblemismainlyfocusedontheuseofthemainmemory-theWilliamstube.Tosolvetheproblem,Forrestertriedtousespiraltapeinstead,buttheeffectwasnotgood,andfinallychosemagneticcorememory.Asaresult,thecalculationspeedofthecyclonecanbedoubled(40kips).
Duetotheapplicationofmagneticcorememory,thecalculationspeedofthecyclonebecametheworld'sfirstatthattime.Theadditionspeedis8microseconds,themultiplicationis25.5microseconds,andthedivisionis57microseconds(excludingthememoryreadtime).Thereadtimeof8500microsecondswhenusingdrummemoryisonly8microsecondsafterswitchingtomagneticcorememory.
ThehighspeedofthecyclonemakesthedevelopmentoftheSAGEsystempossible,andmassproductionofAN/FSQ-7hasbegun.ThemanufacturerisIBM,notthepowerfulRCAatthetime.Sincethen,IBMhasappliedthereal-timetechnologyofthecyclonetotheSABREsystem(airticketreservationsystem).ThemassproductionofAN/FSQ-7beganin1957,andtheconstructionofsupportingfacilitiessuchasbuildings,powertransmissionfacilities,andcommunicationnetworksbeganduringthesameperiod.
LaterontheCycloneFan
AsofJune30,1959,CycloneIIhasbeenprovidingtechnicalsupportfortheSAGEsystem.Inthelate1970s,BillWolf,theoriginalprojectteammember,loaneditoutfor$1peryear.ItwaspurchasedbyKennethHarryOlsen,andthecyclonewasfinallygiventotheSmithsonianInstitution.
Thecycloneiscomposedofabout5000electrontubes.KennethHarryOlsenhastriedtomaintaintheoriginaldesignofthecycloneandturnitintoatransistor.ThisisknownasTX-0.AfterthesuccessofTX-0,OlsenplanstodevelopalargerTX-1.Duetohighexpectationsanddifficulties,onlytheminiaturizedTX-2wasfinallycompleted.TheTX-2itselfalsohadmanyproblems.Soon,OlsenwithdrewfromtheprojectandformedanotherDECcompany.ThePDP-1producedbythecompanyintegratestheadvancedconceptsofTX-0andTX-2,becomingamodelofminicomputers.