Single Event Effects in Avionics - solar storms

1 BBoeing oeing RRadiation adiation EEffects ffects LLabab11/4/99 Single Event Effects inSingle Event Effects inAvionicsAvionicsPresented to C-17 ProgramDecember 16, 1998Dr. Eugene Normand Chief Scientist, Boeing Radiation Effects LabBBoeing oeing RRadiation adiation EEffects ffects LLabab21/4/99 Overview of PresentationOverview of PresentationnIntroduction to Single Event Effects [SEE]nAtmospheric Neutron EnvironmentnEvidence for Neutron-Induced SEU in AvionicsnEnvironments and Effects Related to AvionicsSEUnResults of Recent Avionics SEU EvaluationBBoeing oeing RRadiation adiation EEffects ffects LLabab31/4/99 Description of Various Types ofDescription of Various Types ofSingle Event Effects in ICsSingle Event Effects in ICsnSingle Event Effects [SEE] Disturbance of an active electronic device caused by a singleenergetic particlelUpset (SEU) --change in logic state, simplest example is amemory cell in RAMlLatchup (SEL) --sharp increase in current resulting fromturning on parasitic pnpnlDamage or burnout (SEB) of power transistor or other highvoltage devicelFunctional interrupt (SEFI)

2 - malfunctions in more complexparts sometimes as lockup, hard error, etcBBoeing oeing RRadiation adiation EEffects ffects LLabab41/4/99 SEE: Which EnvironmentsSEE: Which Environmentsare Importantare ImportantnSpace Galactic Cosmic Rays (heavy ions) Trapped Belts (protons) solar Flares (protons & heavy ions)nAircraft Altitudes Neutrons and Protons (Pions)nGround Level Neutrons and ProtonsBBoeing oeing RRadiation adiation EEffects ffects LLabab51/4/99 Neutron Environment in theNeutron Environment in theAtmosphereAtmosphereNeutrons, created bycosmic ray interactionswith the O2 and N2 in the air,peak at ~60,000 ft. At 30,000 ftthe neutrons are about 1/3the peak flux, and on theground, ~1/400 of the peakflux. The peak flux is ~4neutron/cm sec. Otherparticles such as secondaryprotons and pions are alsocreated, but for SEU the neutronsare the most oeing RRadiation adiation EEffects ffects LLabab61/4/991-10 MeV Atmospheric Neutron Flux 1-10 MeV Atmospheric Neutron Flux , Simplified Boeing ModelAltitude, Simplified Boeing , Thousands of Feet1-10 MeV Neut Flux, n/cm^ FluxBBoeing oeing RRadiation adiation EEffects ffects LLabab71/4/991-10 MeV Atmospheric Neutron Flux 1-10 MeV Atmospheric Neutron Flux , Simplified Boeing ModelLatitude, Simplified Boeing , degrees1-10 MeV Neut Flux, n/cm -secBBoeing oeing RRadiation adiation EEffects ffects LLabab81/4/99 Energy Dependence of AtmosphericEnergy Dependence of AtmosphericNeutron Flux, Simplified Boeing ModelNeutron Flux, Simplified Boeing Model1E-51E-41E-31E-21E-11E+01E+01E+11E+ 21E+31E+4 Neutron Energy, MeVDiff'tial Neutron Flux, n/cm -s-MeVBoeing Model.

3 Fit to Ms'rd NASA-Ames DataArmstrong CalculationBBoeing oeing RRadiation adiation EEffects ffects LLabab91/4/994 Independent Sources Confirm Atmospheric4 Independent Sources Confirm AtmosphericNeutrons as Cause of Avionics SEUN eutrons as Cause of Avionics SEUB asis of Neutrons as CauseAirborne Data Serving as Basis1. Variation of in-flight SEU rateswith altitude and latitude have thesame behavior as the variation ofatmospheric neutron flux withaltitude and latitude1. In-flight SEU rates from IBMexperiments (over, Seattle,northern California and Norway)and from CC-2E computer in TS-3E-3 aircraft (mainly over WestCoast)2. Agreement between themeasured in-flight upset ratesand the upset rates calculatedusing the atmospheric neutronflux and laboratory-measuredSEU cross section data2. In-flight SEU rates from IBMexperiments (over Seattle,northern California and Norway),CC-2E computer on TS-3 (WestCoast), CC-2E computer onmilitary aircraft in Europe andSRAMs in other aircraft (trans-Atlantic and worldwide)BBoeing oeing RRadiation adiation EEffects ffects LLabab101/4/994 Independent Sources Confirm Atmospheric4 Independent Sources Confirm AtmosphericNeutrons as Cause of Avionics SEU, Neutrons as Cause of Avionics SEU, Cont dCont dBasis of Neutrons as CauseAirborne Data Serving as between groundlevel SEU rates in SRAMs andDRAMs and those at aircraftaltitudes, when accounting for thedifference in atmospheric neutronflux at ground and 4E4 In-flight SEU rates in 1 and 2above and measured ground levelrates at various locations in US(Seattle, Austin, TX and Batavia,IL)4.

4 Agreement between the energydeposition spectrum measured inthe airborne CREAM detector andthe spectrum measured inBoeing's surface barrier detectorin the WNR high energy neutronbeam at Los Alamos4. Energy deposition spectra inthe CREAM detector onboardConcorde on London-Washington, , London-NewYork and Washington-MiamiroutesBBoeing oeing RRadiation adiation EEffects ffects LLabab111/4/99 Correlation of SEU Rate andCorrelation of SEU Rate andAtmosAtmos. Neutron Flux with Altitude. Neutron Flux with + +010100002000030000400005000060000700008 0000 Altitude, Ft1-10 MeV Neutron Flux and SEU Rate1-10 MeV Atm Neutron FluxSEU Rate, Up/bit-day 1E7 Memory Upset RateNeutron Flux at AltitudeBBoeing oeing RRadiation adiation EEffects ffects LLabab121/4/99 Correlation of SEU Rate andCorrelation of SEU Rate andAtmosAtmos. Neutron Flux with Latitude. Neutron Flux with Latitude0E+01E-22E-23E-24E-2304050607080 90 Latitude, DegreesSEU/Hr and n/cm -secIn-Flight SEUs, Europe & N.

5 America1-10 MeV Boeing Atmos Neut Flux ( .02)BBoeing oeing RRadiation adiation EEffects ffects LLabab131/4/99 Comparison of Measured andComparison of Measured andCalculated In-flight SEU RatesCalculated In-flight SEU RatesAircraftFlightPathAltitude,ft ( E3)SystemOperat gCondt nsSRAM#UpsetMeas dUp/bit-hrCalct dUp/bit-hrE-3 Seattle29 IBMExp VIMS E-9ER-2N. VIMS E-9E-3 Europe-129CC-2E5 VIMS VIMS cllJetlinerTrans-Altlantic~ V Calif.<25AP-1025 VIMS clJetlinerWorld-wide~33 Avionicscomputer5 oeing RRadiation adiation EEffects ffects LLabab141/4/99 Comparison of SEU Rates onComparison of SEU Rates onGround, Field Data and WNR RatesGround, Field Data and WNR RatesObserved Rate From FieldUpset Rate From WNR RAM System/SER Source of Data hrSimilar RAM at WNRUp/b-hrDRAMACPMAPS Computer Fermilab7E-13TC514400-802E-12 IBM Field Tests3E-13 MSM514400-804E-13 CRAY YMP-8, Bulk Storage 6E-13 TMS441002E-12 Nite Hawk 5800 Computer2E-12 DRAMS imple Average1E-12 Simple Average2E-12 SRAMM otorola Field SER tests2E-12 IDT712561E-12 ICDs within patients 2E-12HM656564E-12 CRAY YMP-8 Main Memory2E-12 MCM62063E-12 MCM62462E-13 SRAMS imple Average2E-12 Simple Average (4) 2E-12 Conclude that field Upset rates (large computers, RAM vendor tests, etc.)

6 Agree with upset rates based on measured SEU rates in WNR neutron beam combined with measured cosmic ray neutron flux at ground levelBBoeing oeing RRadiation adiation EEffects ffects LLabab151/4/99 Comparison of Ground-Level SEUC omparison of Ground-Level SEUR ates with Rates in AircraftRates with Rates in AircraftRAMV endor/ RAM Type (D or S)Measr'd WNR SEU X-Sct'n, cm /bitGr'nd level SEU Rate, Up/bit-hr, WNR-Scaled Calculated Ground SEU Rate, Up/bit-hr, BGR MethodTC514400 *MSM514400 *MCM6246 (7 RAMs) * Indicates Heavy ion SEU data from related RAMG round level SEU rate ~2E-12 Upset/bit-hrAircraft SEU rate (neutron flux ~300 times higher) is 300 2E-12, ~6E-10 Upset/bit-hr, agrees w/ in-flight SEU ratesBBoeing oeing RRadiation adiation EEffects ffects LLabab161/4/99 Correlation of Energy Deposition Spectra,Correlation of Energy Deposition Spectra,CREAM on Concorde and SBD in WNR BeamCREAM on Concorde and SBD in WNR Beam1E-191E-181E-171E-161E-151E-141E-131 E-121101001000 Energy Deposited (MeV)BGR (cm / m )WNR, 300 m SBDWNR, 10 m SBDCREAM onboard ConcordeBBoeing oeing RRadiation adiation EEffects ffects LLabab171/4/99 BREL Approach for DealingBREL Approach for Dealingwith SEE in Avionicswith SEE in AvionicsnTesting WNR (at LANL) same neutron spectrum asatmospheric neutrons 14 MeV generator (BREL)nAnalysis BGR method can utilize heavy ion data to estimateneutron-induced SEU response New BGR-based model to explain neutron-inducedlatchup (SEL)BBoeing oeing RRadiation adiation EEffects ffects LLabab181/4/99 Use of the WNR BeamUse of the WNR BeamnBREL has used WNRbeam (LANL)

7 Tomeasure SEU inRAMs, processors,SEL in gate arraysand SEB in powerMOSFETs and highvoltage diodesnThis means that theseSEEs can be inducedby neutrons in aircraftand on ground WNR Neutron Fluxes 1E+51E+61E+71E+81E+91101001000 Neutron Energy, MeVWNR flux, 30 LWNR flux, 60 RAtmos Flux 3E5 BBoeing oeing RRadiation adiation EEffects ffects LLabab191/4/99 Results of Recent SEER esults of Recent SEEE valuation of Avionics SystemEvaluation of Avionics SystemnBased on new mission computing andcommunications subsystems, total of ~20-30 SBCs SEU in main memory well protected by EDAC SEU in cache memory (protected by parity checking)will result in reconfiguration ~once every 2-3 hours SEU in unprotected devices results in reboot once every~100-200 hours (cumulative, all devices)lIncludes microprocessors, FPGAs, MBU in main memory Still large uncertainties in response of some devicesthat can only be reduced by SEU testinglMicroprocessor response based only on PowerPC 603 thatwas tested with protons for space applicationsBBoeing oeing RRadiation adiation EEffects ffects LLabab201/4/99 Perceived Current Status ofPerceived Current Status ofSEE in C-17 Avionics SystemsSEE in C-17 Avionics SystemsnFlight Controls Appears to be adequately protected through combination ofEDAC and redundancynMission Avionics More likely to have subsystems that may not be adequatelyprotected Anecdotal story (from Wright-Patt, ~4 years ago) of upsets incomputer during flight, fixed via EDAC Could check operational logs for evidence of malfunctionsduring flight that could not be detected when box tested in thelab (Could Not Duplicate, Retest OK)

8 LIn commercial Avionics experience, 20% of all CNDs are due toSEUsBBoeing oeing RRadiation adiation EEffects ffects LLabab211/4/99 How Can BREL Help?How Can BREL Help?nProvide environmental requirements section (atmosphericneutrons) to be incorporated into specs for future upgradesnAssist in carrying out SEU evaluations/reviews of futureavionics upgradesnCarry out SEU evaluations of existing systems in which SEUproblems most likely, new flat panel displaynIdentify current systems containing EDAC protection fromwhich direct evidence of SEU occurrences may be derivedfrom operational logsnIncorporate error logging features into EDAC protection toallow SEU occurrences to be monitored, thereby providingfeedback for design of future systems