Single-Event Transient Pulse-Width Measurements in Advanced ...

Autonomous SET captureVanderbilt Radiation Effects Group1 st stageSchool of EngineeringTargetcombinationalcircuitLatch Latch Latch Latch LatchAutonomous orself-triggeredSET pulsemeasurementdelayCONTROL SIGNALNarasimham et al, IEEE TDMR, 2006 SET generated in custom inverter target circuitry – propagates andtriggers measurement Pulse width is proportional to number of switched states (measured inunits of inverter delays) Upsets after the trigger stage are not measured Most single latch upsets (due to direct strikes on the latch) are identifiableand SET width data not affectedMURI ReviewMay 13, 2008 4

Heavy-Ion – 130-nm & 90-nmVanderbilt Radiation Effects GroupSET pulse width (ps)2000180016001400120010008006004002000130-nmIBM 130-nm (8RF)9.7 9.7(60°)SET pulse width (ps)1600max140012001000 +1óavg800-1ó600min40031.2 31.2200(50°)LET (MeV-cm0 2 /mg)58.7 58.7(38°)9.7(130-nm)130-nm90-nm9.0(90-nm)58.7(55°)1000 160090-nm1400130-nm vs 90-nm10010Events measuredSET pulse-width (ps)31.2(130-nm)12001000800600400200030.0(90-nm)LET (MeV-cm 2 /mg)IBM 90-nm (9SF)max+1óavg-1óminSchool of Engineering1.8 3 5.7 9 20.6 30 40.7 59LET (MeV-cm 2 /mg)58.7(130-nm)59.0(90-nm)10000100010010Events measured Results indicate increasing SET pulse widths with scaling important aslatch setup/hold times decrease with scaling more SETs latched as errorMURI ReviewMay 13, 2008 6

1.2 48 58.7 75 100Vanderbilt Radiation Effects GroupSchool of Engineering-1σ+1σ -1σ+1σSET cross section per inverter (!m 2 )100.80.60.40.20.0Cross Section Comparisonmin-1óavg+1ó100240 480 720 960 1200 1440SET pulse width (ps)9.7 MeV-cm^2/mg20 MeV-cm^2/mg58.7 MeV-cm^2/mg130-nmEvents measured10.6 30 40.7 592 /mg)maxSET cross section per inverter (!m 2 )5 MeV-cm^2/mg0.620 MeV-cm^2/mg40 MeV-cm^2/mg0.40.2010001000100Events measured100 300 500 700 900 1100 1300SET pulse width (ps)90-nm Event cross section dominated by pulses between 300 ps to 700 ps in 130 nm 400 ps to 900 ps in 90 nm Increase in wider transients translates to higher error rates with scalingMURI ReviewMay 13, 2008 7

Neutron-Induced SETsVanderbilt Radiation Effects GroupDUT boards87Neutron-induced SETs in 90-nmSchool of EngineeringneutronbeamMeasured events6543210100 300 500 700 900 1100 1300SET pulse width (ps) Tested six 90-nm ICs at WNR facility at LANL Energy spectrum matches sea-level spectrum for energies from 10 to500 MeV Neutron fluence 1.33×10 11 neutrons/cm 2 Neutron SET cross-section ~ 2.5 × 10 -6 µm 2 /inverterMURI ReviewMay 13, 2008 8

Alpha-Induced SETsVanderbilt Radiation Effects GroupSchool of EngineeringMeasured events6050403020100Alpha-induced SETs in 90-nm100 200 300 400 500 600 700 800 900 1000 1100SET pulse width (ps) Alpha tests at TIusing foil of Am 241 Energy ~ 5.5 MeV Fluence ~ 4.45×10 10alphas/cm 2 Total eventsmeasured ~ 300 Alpha SET cross section ~ 6.74 × 10 -4 µm 2 /inverterMURI ReviewMay 13, 2008 9

Proton-Induced SETsVanderbilt Radiation Effects GroupProtons-induced SETs in 90-nm1200School of Engineering1000maxSET pulse width (ps)800600400+1óavg-1ó200min00 30 (WE) 60 (WE) 85 (WE) 60 (NS) Tests conducted at Indiana Univ with 200 MeV protons Proton SET cross-section ~ 2 × 10 -6 µm 2 /inverter Results indicate that proton angle of incidence does not affect SETwidth distributionsMURI ReviewMay 13, 2008 10

Comparison – Energetic ParticlesVanderbilt Radiation Effects Group1600Proton, Neutron, Alpha and Heavy-ioninduced SETs in 90-nmSchool of Engineering1.E+0114001.E+00SET pulse-width (ps)12001000800600400max+1óavg-1ó1.E-011.E-021.E-031.E-04Event Cross Section(!m 2 /inverter)200min1.E-050proton200 MeVneutronspectrumalpha Ne - 3 Ar - 5.7 Kr - 20.6(5.5 MeV)(LET values)Energetic particle1.E-06MURI ReviewMay 13, 2008 11

Similarity in SET DistributionsVanderbilt Radiation Effects GroupSchool of EngineeringSET pulse-width (ps)16001400120010008006004002000proton200 MeVProton, Neutron, Alpha and Heavy-ioninduced SETs in 90-nmneutronspectrummax+1óminavg-1óalpha Ne - 3 Ar - 5.7 Kr - 20.6(5.5 MeV)(LET values)Energetic particle1.E+011.E+001.E-011.E-021.E-031.E-041.E-051.E-06Event Cross Section(!m 2 /inverter) Less variation in SET pulse widths for different particle types – probablydue to carrier redistribution and charge sharing effects Preliminary simulations indicate charge spread and diffusion collection bysubsequent gate may limit SET widthsMURI ReviewMay 13, 2008 12

Other ApplicationsVanderbilt Radiation Effects GroupSchool of Engineering Test circuit can help quantify effects of process orlayout variations on single events−Circuit was used for quantifying effect of guard bands inmitigating long SETs Characterization of effect of charge spread on SETwidth distributions−−Tested with ions incident from different directions withrespect to n-well layoutInterleaved inverter chains have been designed to furtheranalyze charge spread effectMURI ReviewMay 13, 2008 13

Well Contacts and Guard BandsVanderbilt Radiation Effects GroupSchool of EngineeringWell Contacts(isolated)HDWCHDWCGBsubstratecontactConventional Layout –Isolated Well ContactsHigh Density WellContacts (HDWC)HDWC + Guard Band (GB) Strikes within n-well lead to long SETs dueto parasitic bipolar effect Mixed mode simulations using calibrated130-nm process – show effect of GB andHDWC in mitigating long SETsMURI ReviewVoltage (V)1.41.210.80.60.40.20-0.2isolated contactsHDWCHDWC+GBmixed mode TCAD simulationLET = 40 MeV-cm 2 /mg0.0E+00 5.0E-10 1.0E-09 1.5E-09 2.0E-09 2.5E-09 3.0E-09time (s)May 13, 2008 14

Characterizing Guard BandsVanderbilt Radiation Effects GroupXe at 38.5 degEvents measured60504030201002403604806007208409601080SET pulse width (ps)with GB+HDWCwith isolated contacts1200132014401560Cumulative Event Cross Section (cm 2 )1.E-051.E-061.E-07School of EngineeringSETs > 1 ns - with isolated contactsSETs > 1 ns - with GB + HDWC1.E-080 20 40 60 80 100 120Effective LET (MeV-cm 2 /mg)isolated contacts –conventional layoutpracticehigh density wellcontacts (HDWC)guard bands (GB) Events > 1ns reduced by ~72% forcircuits with GB + HDWC Reduction in frequency and max SETwidth attributed to reduction inparasitic bipolar effectMURI ReviewMay 13, 2008 15

ImpactVanderbilt Radiation Effects GroupSchool of Engineering Resolve controversy on expected pulse widths Quantify technology trends in SET pulse widths Characterize effect of RHBD structures such as multiplewell contacts and guard bands on SET pulse widths Determine effect of charge sharing on SETs Technique can also be used to measure other spurioussignals such as cross-talk pulsesMURI ReviewMay 13, 2008 16

ConclusionVanderbilt Radiation Effects GroupSchool of Engineering Autonomous SET characterization technique developedand implemented to obtain precise distributions ofheavy-ion, neutron, proton and alpha induced SETwidths in 130-nm/90-nm CMOS Width and range of dominant SETs increase with scaling Neutron and alpha particle induced SETs are of order oflegitimate logic signals – concern for commercialapplications High density well contacts and guard bands helpsreduce more than 70% of SETs longer than 1 nsMURI ReviewMay 13, 2008 17