Transcript ppt
ESE370: Circuit-Level Modeling, Design, and Optimization for Digital Systems Day 32: November 24, 2010 Uncorrelated Noise Sources Ionizing, Thermal, Shot 1 Penn ESE370 Fall2010 -- DeHon Today • Ionizing Particles • Thermal Noise • Shot Noise 2 Penn ESE370 Fall2010 -- DeHon Ionizing Particles • Alpha Particles (He nucleus=He2+) • Impact with mega-electron-volts of energy (3—10MeV) 3 Penn ESE370 Fall2010 -- DeHon Ionizing Particles • Alpha Particles (He nucleus=He2+) • Can penetrate microns into Si • Creating 2×106 electron-hole pairs 4 Penn ESE370 Fall2010 -- DeHon Ionizing Particles • Alpha Particles (He nucleus=He2+) • Can be generated by decay in packaging materials – Lead common one including some fraction of radioactive isotopes 210Pb 5 Penn ESE370 Fall2010 -- DeHon Src: http://en.wikipedia.org/wiki/File:Wirebonding2.svg Comparisons • How many electrons in: – Capacitor: 1fF charges to 1V – e = 1.6×10-19 Coulombs • Recall C0 = 0.01fF, – typical load around 10-20C0 • How large a capacitor to withstand loss of 2×106 electrons? 6 Penn ESE370 Fall2010 -- DeHon Disrupt • Alpha particle will disrupt DRAM Cell • Can disrupt undriven nodes – Latch – Dynamic node – Memory bit 7 Penn ESE370 Fall2010 -- DeHon Ionizing Particles • There are other particles with different energies – Neutrons from cosmic rays • 10x energy of alpha particles 8 Penn ESE370 Fall2010 -- DeHon Particle Flux • Differs with location – Altitude • Denver vs. Philadelphia • Ground vs. aircraft at 30,000 feet • Space (outside atmosphere) – Near poles • Changes upset rate seen by chips 9 Penn ESE370 Fall2010 -- DeHon LLNSD October 2005 [Quinn and Graham/FCCM2005] Soft Error Failure Rate Projections 1 FIT = 1 Failure per 109 hours of operation LLNSD October 2005 [Heather Quinn 2005] Scaling • Charge holding memory decreases – More susceptible • Cross-sectional area of bit decreases – Each bit less likely to be hit – Hit may be large enough to take out multiple bits at a time • More bits on the chip – More targets! more likely something gets hit 12 Penn ESE370 Fall2010 -- DeHon Driven Node • What happens if the alpha particle impacts a driven node? • Will recover – Creates a glitch – May slow down node – Only a problem if latched into register 13 Penn ESE370 Fall2010 -- DeHon Driven Failure • Driven input to latch -- value failure? – When is it a problem? • Occurs at end of cycle – Right at last transition time for node • Occurs earlier – Not expecting value to have settled • Occurs later – May not propagate to latch 14 Penn ESE370 Fall2010 -- DeHon Logic Failure Rate Pfail T upset T Pstrike cycle • Probability will see increases with increasing frequency • Tupset ~ picoseconds? 15 Penn ESE370 Fall2010 -- DeHon Frequency Dependence Gill (Intel), IEEE International Reliability Physics Symposium 2009 16 Penn ESE370 Fall2010 -- DeHon SEU/bit Norm to 130nm Scaling and Error Rates Increasing Error Rates 10 2X bit/latch count increase per generation logic cache arrays 1 180 130 90 65 45 32 Technology (nm) Penn ESE370 Fall2010 -- DeHon Source: Carter/Intel 17 Thermal Noise 18 Penn ESE370 Fall2010 -- DeHon Thermal Background • Except at absolute 0 (Temperature) – Particles are moving around randomly • Thermal bath means free energy around • Electron can be borrow the thermal energy to hop over barrier – Out of an energy well, bond cite – …Out of a capacitor 19 Penn ESE370 Fall2010 -- DeHon Day 8 Doping with P • End up with extra electrons – Donor electrons • Not tightly bound to atom – Low energy to displace – Easy for these electrons to move 20 Penn ESE370 Fall2010 -- DeHon Day 8 Doped Band Gaps • Addition of donor electrons makes more metallic – Easier to conduct Semiconductor 0.045ev 1.1ev Ec ED Ev 21 Penn ESE370 Fall2010 -- DeHon Day 8 Electron Conduction 22 Penn ESE370 Fall2010 -- DeHon Thermal Background • Except at absolute 0 (Temperature) – Particles are moving around randomly • Thermal bath means free energy around • Electron can be borrow the thermal energy to hop over barrier • Are doing it all the time to give us our semiconductors 23 Penn ESE370 Fall2010 -- DeHon Rising above the Thermal Noise • Must apply more energy than background noise to – Hold electron in place – Move an electron from place to place • Charge/discharge a node with some reliability 24 Penn ESE370 Fall2010 -- DeHon Minimum Energy • Single bit gate output – Set from previous value to 0 or 1 – Reduce state space by factor of 2 – Entropy: S= k×ln(before/after)=k×ln2 – Energy=T S=kT×ln(2) • Setting a bit costs at least kT×ln(2) 25 Penn ESE370 Fall2010 -- DeHon Probability of Noise Error • This minimum energy around kT is just to have 50% probability of setting bit correctly • Probability exponential in energy – Not exactly this…but basic dependence Perror e E kT 26 Penn ESE370 Fall2010 -- DeHon Implication • To keep error rate sufficiently low – Need energy of operation (of storage) to be some multiple of kT Perror e E kT 27 Penn ESE370 Fall2010 -- DeHon Where are we today? • How does kT compare to switching 10C0 at 1V? – k=1.4×10-23 J/K – T=300K (Room Temperature) 28 Penn ESE370 Fall2010 -- DeHon Where are we today? • How does kT compare to switching 10C0 at 1V? – k=1.4×10-23 J/K – T=300K (Room Temperature) • kT=4.2×10-21 J • Eswitch=CV2 = 0.1fJ=10-16J • Eswitch~=2×104 kT 29 Penn ESE370 Fall2010 -- DeHon Scaling • Eswitch~=2×104 kT • 45nm to 4.5nm impact on Eswitch? – reduce capacitance by 10x – reduce voltage by 2x – Eswitch~=500 kT 30 Penn ESE370 Fall2010 -- DeHon Relate kT to electrons • If we arrange the capacitor to hold a single electron – What storage/switching just equals 1 kT? 31 Penn ESE370 Fall2010 -- DeHon Chip Reliability • Chip has many transistors • …And many switching events • Each of which may fail 32 Penn ESE370 Fall2010 -- DeHon Chip Upset Rates Kish, Physics Letters A 205 144—149 (2002) Penn ESE370 Fall2010 -- DeHon 33 Shot Noise 34 Penn ESE370 Fall2010 -- DeHon Shot Noise • Actual electron transport is probabilistic • Current is a statement about average rate of electron flow • For large numbers of electrons – Law of large numbers convergence to mean s ~= Sqrt(N) – Large N sqrt(N)/N small • Small percentage variation Penn ESE370 Fall2010 -- DeHon 35 Shot Noise • For small number of electrons (N) s ~= Sqrt(N) – sqrt(N)/N not so small – Higher variation – Noise in switching time 36 Penn ESE370 Fall2010 -- DeHon Electron Counts • How many electrons (N) – in 0.1fF, 1V switching event? – in 0.01fF, 0.5V switching event? s? • How many s out to only get 50% of electrons moving? 37 Penn ESE370 Fall2010 -- DeHon Will we see? • Large chips, fast clock rates many events….samples far out on curve From: http://en.wikipedia.org/wiki/File:Standard_deviation_diagram.svg 38 Penn ESE534 Spring2010 -- DeHon Gaussian Distrubution Number Sigma 1 in How many 1 2 3 4 5 6 3.2 22 370 16K 1.7 M 510M 39 Penn ESE370 Fall2010 -- DeHon Admin • Class Monday • Lab on Wednesday or Friday (TBD) – Class other day 40 Penn ESE370 Fall2010 -- DeHon Idea • Many sources cause upsets – Ionizing particles, thermal, shot noise • Tend to depend on charge – Of node, of switching even • Scaling decreases charge – Lower voltage, lower capacitance • Also increases susceptible nodes – Also increases frequency susceptibility 41 Penn ESE370 Fall2010 -- DeHon