270 likes | 464 Views
Radiation Effects in FPGAs & SEU Modeling Methodology Ketil Røed Høgskolen i Bergen. Topics. Radiation effects in the programmable logic What is the problem / challange? Key words: SEU rate in CERN ALICE TPC FEE FPGA Experimental work Why? FPGA acceptance study
E N D
Radiation Effects in FPGAs & SEU Modeling Methodology Ketil Røed Høgskolen i Bergen
Topics • Radiation effects in the programmable logic • What is the problem / challange? • Key words: SEU rate in CERN ALICE TPC FEE FPGA • Experimental work • Why? FPGA acceptance study • How? Accelarated beam testing • Results • System design solution for radiation tolerant electronics • Simulation work • Why? Predict SEU rate • How? Monte Carlo based physics simulation • Results
Bly Bly Problemstilling: Elektronikk i strålingsmiljø • Hvordan oppfører elektronikk seg i et slikt miljø?
Field Programmable Gate Array, FPGA • Enkel sammenligning: • ASIC /Full Custom • Application Specific Integrated Circuit • FPGA • Field Programmable Gate Array
Field Programmable Gate Array, FPGA • En FPGA er laget av flere predefinerte elementer som kan konfigureres • Konfigurasjons-data lagres i SRAM => Rekonfigurerbar
0101000101 0101011000 0101011110 1010110010 Transistor Field Programmable Gate Array, FPGA
+ - - + + - - Transistor Transistor + + - - + - Transistor Og problemet var? Single Event Upset (SEU) 0101000101 0101011000 0101011110 1010110010 0101000101 0101011000 0101011110 1010100010
Utvikle testmetoder Utføre strålingstester Analyse Aksepttest: Arbeidsflyt Kartlegge Strålingsmiljø
Irradiation Test Facilities • The Svedberg Laboratory, TSL • 38 & 180 MeV protons • 90 MeV neutrons • 106-107 p/cm2s &103-104 n/cm2s • Beam spot radius 2 & 15 cm • The Oslo Cyclotron, OCL • 29 MeV protons • 106-107 p/cm2s • Beam spot radius: 2 cm
SEU Cross Section Result (Xilinx Virtex II Pro) • TSL (180 MeV p): σ = 2.14 x 10-14 cm2 / bit • OCL (29 MeV p): σ = 2.11 x 10-14 cm2 / bit • Independent experiment by Xilinx (atmospheric neutron spectra) • Rosetta+σ = 2.98 x 10-14 cm2 / bit • Scaled to the ALICE TPC radiation environment • Simulated* hadron flux 100-400 h/cm2s • Results per 216 FPGA / 4 hr unit Run • # SEU ~120 • # functional failures+ ~12 • Conservative numbers + Using an SEUPI: Single Event Upset Probability Impact = 10 Lesea et. al. The Rosetta Experiment, IEEE TRANS. ON DEVICE AND MATERIALS RELIABILITY, V 5, N3, 2005 *Fasso et. at, Radiation in the ALICE TPC detector, ALICE internal Note-TRD 2003
Alternativt Kontoret Fly (12000m) 1 FPGA CERN 216 FPGA CERN 1 FPGA OCL 1 feil / ~4000 timer 1 feil / ~400 timer 1 feil / ~100 timer 1 feil / ~0.5 timer 1 feil / ~sekunder Resultatet er i overkant av hva som kan tolereres
Systemløsning: Kontrollert rekonfigurering Fails during run
Komplisert resultat • Configuration memory is read back and bit-flips are corrected • The FPGA firmware design can be protected • Hamming bits • Triple Mode Redundancy • 1+2: Reduces the functional • failures to an acceptable level • Test flux: 106-107 p/cm2s • TPC flux: 100-400 h/cm2s • ~ factor 104 lower flux 1 2 3 1 1+2 No action Readback and Correction enabled Triple Modular Redundancy turned enabled
Ingen beskyttelse Enklere versjon Kombinasjonen har vist gode resultater i strålingstester Rekonfigurasjon + TMR % Feil under test < 5 <10 60-90 100
Next step: SEU Monte Carlo simulation Model the physical processes responsible for causing SEUs • General tools • FLUKA • GEANT4 • Dedicated tool • SEMM2 (IBM) • Not plug and play! • Utfordring: Tilpassing til vårt bruk Physics Software Tools SEU SIMULATION Semiconductor device technology
Reasearch Visit at IBM • 7 months research visit at IBM T.J. Watson Research Center, NY, USA(NFR Leiv Eiriksson mobilitetsstipend) • To learn the business of SEU modeling methodology
+ -+ - - - + - + - + - - + Modeling Methodology Describe physical processes • Inelastic and elastic Interactions • Nuclear fragmentation • Transportation Target information • Technology • Layout • Material Radiation environment • Particle type • Energy & flux Modelling Software Tool Radiation event generator Charge transport & collection Critical Charge < SEU
Gate Source Drain N+ + - N+ + - + - + - + - + - + - + - + - + - + - + - + - + - P substrate The Physics of SEU p • Basic mechanism • Charge particle depositing energy in a device through ionization • If Qdep > Qcrit => SEU • TPC radiation environment* • Energetic Hadrons (E > 20 MeV) • Nuclear Interaction • Short range recoil-ion, • Typically E < 10 MeV Si(p,p α)Mg p α Mg Qdep * Fasso et. at, Radiation in the ALICE TPC detector, ALICE internal Note-TRD 2003
FIB: Focused Ion Beam Target Device Information • Investiagated using Focused Ion Beam tool (FIB)
Target Device Information: FIB Images ~5 um Device layer (90 nm transitors) Si substrate ~850 um Cu interconnet layer (0.4 um) FIB images are used to build a geometry model of our device
Radiation environment • Run new simulations • Add FEE geometry (C++ class in Aliroot) • SEU dedicated scoring regions http://aliceinfo.cern.ch/Public/panorama/WONDERS_QGP/index.html
SEMM2 Methodology Physics Radiation environment Target information Modeling software H.Tang, SEMM-2: ”A new generation of single-event-effect modeling tools”,IBM Journal of Research and Development V5 No3 2008
Adapt to CERN FPGA case CERN Radiation field BEOL FPGA geometry FEOL FPGA geometry Xilinx Qcrit information Dedicated Post-processing H.Tang, SEMM-2: ”A new generation of single-event-effect modeling tools”,IBM Journal of Research and Development V5 No3 2008
Thin Si target Alpha Recoil ion Beam Si P,n,d Ongoing work and outlook (I) • Comparison study (Fluka, GEANT4, NUSPA) • Particle production from nuclear reactions
Ongoing work and outlook (II) • Comparison study (Fluka, GEANT4, NUSPA) • Konklusjon: Modellene er forskjellige! • Men: Ingen overraskelse
Ongoing work and outlook (III) • Comparison study (Fluka, GEANT4, NUSPA) • Preparing simulation case study of FPGA • Compare with experimental results to validate method • Improve simulations of the radiation environment • Add more detailed geometry of electronics rack *Rodbell et al , Low-energy Proton induced Single Event Upsets in 65 nm…, Presented at NSREC July 2007
Konklusjon • Eksperimentelt arbeid foreløpig ferdig og aksepttest med tilhørende løsning godkjent • Mye spennende arbeid gjenstår • Tverrfaglig innsats • Strålingsproblematikken er forventent og tilta etter hvert som teknologien nedskaleres *Rodbell et al , Low-energy Proton induced Single Event Upsets in 65 nm…, Presented at NSREC July 2007