1、Analytical Approach for Soft Error Rate Estimation of SRAM-Based FPGAs,Test & Reliability Group (TRG) Department of Electrical & Computer Engineering Northeastern University,Outline,Problem Statement & Motivation Soft Errors Background & Previous work Error Models in FPGAs SER Estimation Experimenta
2、l Results Summary & conclusions,Problem Statement,Estimating soft error rate in FPGAs The probability of system failure Due to soft errors For a given mapped design Mean time to manifest a corrupted conf. bit To primary outputs or Flip-flops,Motivation,Need for soft error rate estimation Exponential
3、 growth of vulnerable bits due to Moores law High cost of Error tolerant schemes To make appropriate cost/reliability trade-offs Where to put redundancy Why an analytical method? Previous work: Fault Injection Time-consuming / Incomplete / Expensive Needs physical prototype board Cannot be used in d
4、esign phases,Background: Error Definitions,Soft Errors: Intermittent malfunctions of the hardwareNot reproducible Energetic Particles Single Event Upsets (SEUs)Soft Errors (may cause) System Failure,Previous Work,Based on Fault Injection (FI) Inject fault Run several workloads Compare results with f
5、ault-free circuit Exhaustive FI is very time-consuming Candidate some locations for FI Analysis based on statistics,Previous Work (Cont.),Radiation-based fault injection Expensive & not commonly used Needs physical implementation Cannot be used during design phases Can damage prototype board Hard er
6、ror Simulation-based fault injection Bit-stream alteration Needs physical implementation Bridging errors may lead to hard errors,Outline,Problem Statement & Motivation Soft Errors Background & Previous work Error Models in FPGAs SER Estimation Experimental Results Summary & conclusions,Error Models
7、in FPGAs,Memory resources: User bits Flip-flops, RAMs, Configuration bits Mux select bits, LUT bits, User bits Transient errors Config. bits Permanent errors,Error Models in FPGAs (Cont.),ff,F1,F2,F3,F4,Configuration Memory Cell,M,M,M,M,M,M,M,LUT,BlockRAM,SEU (Bit flip),clk,E1 E2,E1 E3,E2 E3,Bit fli
8、p Transient errorCan be corrected at the next load,Virtex (Xilinx),Bit flipPermanent errorCorrected by reconfiguration,Short or open circuitCorrected by reconfiguration, Lima (DAC03),Error Models in FPGAs (Cont.),Transient errors User flip-flops, Logic gates, Block RAMs Permanent errors (all configu
9、ration bits) Routing: MUX select bits PIP: Short/Open Buffer: On/Off LUT Control/Clocking Bits,Error Models in FPGAs (Cont.),Only permanent errors considered Conf. bits comprise more than 99% of all memory elements excluding RAM blocks 95% of all memory elements including RAM blocks,Outline,Problem
10、Statement & Motivation Soft Errors Background & Previous work Error Models in FPGAs SER Estimation Experimental Results Summary & conclusions,SER Estimation,Traversing structural paths Asadi04 From fault sites to POs,SER Estimation in ASIC Designs,S(n): System failure probability (SFP) vector Si: SF
11、P given node i erroneous n: total fault sites Experiments on ISCAS89 show that: Three order of magnitude faster Compared to random-input simulation Average accuracy: 97%,FPGA vs. ASIC in SER Estimation,ASIC: transient error Only requires propagation probability FPGA: both transient & permanent error
12、s Transient errors: the same Permanent errors: needs activation as well Nodes with different error rates in FPGAs Fault sites: all nodes,SER Estimation of FPGAs: Steps,Compute permanent error rates for all nodes PRi : the permanent error rate of node i n: total number of fault sitesCompute netlist f
13、ailure probability vector Ni= failure prob. given node i erroneous System failure rate vector (S) = PR N Si = PRi Ni,How to Compute Ni?,Open & stuck-at errors: Ni = SPi PPi(0) + (1-SPi) PPi(1) = PPi PPi: Propagation prob. (the method used for ASIC) SP: Signal probability is used for activation prob.
14、Bridging wired-AND error (nets i and j): Ni = SPi(1-SPj)PPi(0) + (1-SPi) SPjPPj(0)Bridging wired-OR error (nets i and j): Ni = SPi(1-SPj)PPj(1) + (1-SPi) SPjPPi(1),How to Compute PRi?,PR(n): permanent error rate vector PRi : r f r: Raw error rate of an SRAM cell f: Number of all possible errors at n
15、ode i n: total number of fault sites PRAB= 6 r,System Failure Rate,For the first clock:For c clock cycles:The same probability is valid for the next clock cycles c: Number of clocks checking the state of the circuit After particle hit,Outline,Problem Statement & Motivation Soft Errors Background & p
16、revious work Error Models in FPGAs SER Estimation Experimental Results Summary & conclusions,Error List,Mux-open PIP open Buffer off A bit-flip in LUT Control bit-flip,Experimental Setup,Xilinx Virtex 300 (XCV300) Xilinx Design Language (XDL) Benchmark: some ISCAS89 circuits r = raw failure rate for
17、 an SRAM cell r=0.01 FIT/bit 1000 clocks executed for each SEU Platform: Sun Solaris Ultra-10 256 MB Main Memory,Results: Sensitive Bits,Number of sensitive SRAM bits for each part,Results: Manifestation Time,(Results are in terms of cycles),Mean Time To Manifest (MTTM) errors to outputs,Results: SF
18、R & Estimation Time,Number of Clock cycles: 1000 SP Time: Signal Probability computation time SFR Time: System Failure Rate computation time,System Failure Rate & Estimation Time,Summary & Conclusions,A new approach for SER estimation For SRAM-based FPGAs No physical implementation required Can be u
19、sed in early design stages Very fast simulation time Can cover all possible faults Mean Time To Manifest errors to outputs: MTTM(Control/clocking) MTTM(routing) MTTM(routing) MTTM(LUT),Appendix & Backup,Background: Soft Error Origin,The main sources in terrestrial conditions: Alpha particles & Neutr
20、ons Soft Error occurs:if hitting particles generate more than Qcrit Critical Charge (Qcrit): the minimum charge needed To flip the value stored in the cell,Exp. Increase of Soft Errors,e-Qcrit/Qs trend with technology scaling (Shivakumar , DSN 2002) Qcrit: the critical charge (depend on characterist
21、ics of the circuit) Qs: the charge collection efficiency of a particle strike on the device Particles of lower energies occur far more frequently,Background: Definitions,How to express Soft Error Rate (SER) MTBF (Mean Time Between Failures) FIT (Failure-in-Time) 1 failure in a billion hours 1 year M
22、TBF = 114,155 FIT,Background: Definitions,Failure definition: (a) Propagation of an erroneous value to at least one flip-flip or primary output or (b) Propagation of an erroneous valueto at least one primary output Definition (a) is compatible with (b) If there is no redundant flip-flop in the circu
23、it,Failure Error Rate of LUT,To reduce number of nodes LUT as a complex gate P(tx): the probability of O=tx LUT failure rate SO=AP(t0)+AP(t1)+AP(t15).r.NO = r.NO,Xilinx Virtex FPGA Model,Logic block,Switch Matrix (SM),IO Mux,CLB,IOB,Line Segments,CLB Architecture,Error Models in FPGAs (Cont.),Config. Bits: Care bits All 1s Some of 0s Dont care bits Some of 0s,Error Models: PIP Short/Open,10: causes open 01: may cause short or bridging error,Error Models (Cont.),Buffer on/off Tri-state buffers Used in IOBsLook-Up Table,
