Frequency Dependency and Voltage Acceleration Factor of A54SX-A/MEC Long-term Life Test and Radiation Test Results of RT

1. Frequency Dependency and Voltage Acceleration Factor of A54SX-A/MEC Long-term Life TestandRadiation Test Results of RTSX-SU Noriko YAMADA Electronic, Mechanical Components and Materials Engineering Group Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

2. Background • Antifuse failures of antifuse FPGAs, RTSX-S and SX-A, produced with MEC die (‘the MEC devices’) have been reported since the beginning of 2003. • The Industry Tiger Team (ITT) and NASA are making examinations and evaluation tests to find out the cause of the failures. • The evaluation tests showed that the failures, increases of delay time, were caused by increases of resistance of antifuses (‘antifuse degradation’). • The antifuse degradation is attributed to the problem of the structure of the MEC devices , and the manufacture recommends change to the UMC devices which are compatible with the MEC devices. • JAXA carried out evaluation tests to evaluate the risk of the MEC devices mounted in the PWB and to determine the management plan for every project. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

3. Introduction • Risk assessment of the MEC Devices • To determine the acceleration factors and the failure rates of antifuse failures. • Long-term life tests and thermal shock tests are carried out . • Reliability evaluation of the UMC Devices • The UMC device received QML authorization in October,2004, though the UMC devices have no experience in the space. • To evaluate the reliability for space applications, long-term life tests and radiation tests were performed. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

4. Test Samples Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

5. Test Vehicle (1) Block Unit • 4-input AND-OR chains : Maximum utilization of antifuses • Stable operation using an external clock circuit: Easier failure detection • R-cells driven by skewed clock: Delays detectable to less than 10nsec • Continuous monitoring of XORed outputs from the same 4 to 8 circuit blocks: Real-time detection of failures x4:32A 32SU x8:72A Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

6. Test Vehicle (2) The number of antifuses in test vehicles Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

7. Items and Conditions Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

8. MEC:Results (1): Weibull Plots • Weibull parameter (beta< 1) Infant mortality, identical to the result in the U.S. Weibull Parameters Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

9. MEC:Results (2): Failure Rate 90% Confidence Interval (Upper-Lower) • The temperature dependency of the failure rates is very small. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

10. MEC:Results (3): Activation Energy Ea=0.002eV • The activation energy (Ea) was calculated based on the Weibull plot of 72A sample. • Ea=0.002eV  PPBI (125 deg.C, 240 hours) is almost ineffective in screening for the antifuse failures. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

11. UMC:Results • Long-term Life Test (1MHz, 125 deg.C, 1000H) No Failure • The antifuse failures which were detected in the MEC devices did not take place. • The failure caused by ESD was not observed.* * The evaluation in the U.S, in which the failure by ESD occurred, demonstrated that the electrostatic tolerance of 32SU is extremely low. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

12. MEC:Voltage Acceleration Factor • Long-term life test was carried out on 72A at 3.0V following 1000H life test at 2.5V (@1MHz, 25deg.C) • 2.5V  3.0V Acceleration factor is about 50 times. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

13. MEC:Frequency Dependency (1): 32A • 33MHz1MHz Failure rates were almost the same and the frequency dependency was not observed. • The thermal migration model which Actel advocates cannot explain the results. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

14. MEC:Frequency Dependency (2): 72A • In order to avoid the influence of noise when testing at 33MHz, power supply voltage was set up 0.1V lower than when testing at 1MHz. The failure rate at 33MHz is low. • The voltage acceleration factor calculated in page12 was about 3 (0.1V). This is almost the same as that of estimated from this graph. No frequency dependency Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

15. Discussion Suggested Mechanism of acceleration of antifuse failures • JAXA considers the Percolation Model can explain the voltage acceleration. Cross section of antifuse structure of MEC devices Al • A bad conductive path is not a solid line, but a dispersion of Si and metal. • An electrical stress generates a conductive percolation path inside the bad conductive path. • An overcurrent happens, resulting in the breakdown of the path. • The resistance of the path increases. TiN Good conductive path W Plug SiO2 SiO2 Bad conductive path TiN a:Si SiO2 SiO2 SiN TiN Al Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

16. UMC:Radiation Test (1): SEE • SEU(Single Event Upset) / SEL(Single Event Latch-up) - Japan Atomic Energy Agency(JAEA), TIARA* AVF CYCLOTRON - Xe (LET=64[MeV/mg/cm2]) • No SEL was observed • No SEU was observed in the static state . • SEUs were observed in the dynamic state, andshowed frequency dependency (right figure). Frequency dependence TIARA* (Takasaki Ion Accelerators for Advanced Radiation Application) Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

17. UMC:Radiation Test (2): TID JAXA result Manufacturer’s result • Increase in power supply current was observed.  Sharp increase of Icca at 600Gy(Si)  Recovered to the initial value after annealing (100 deg.C with 168H) Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

18. Thermal Shock Test • Thermal Shock Test • Conditions: -65~150 deg.C • Test period: 1000 cycles • Test samples: MEC: 32A&72A, UMC: 32SU • No Failure was observed Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

19. Summary (1) • MEC： • Delays caused by anti-fuse degradation were observed and considered to be similar phenomena observed in evaluation conducted by ITT and NASA. • The failure rates of MEC are large, compared with general LSI for the space. • The temperature acceleration of the failure rates was too small to screen out the defective antifuses throughout PPBI (125deg.C 240hours) • The frequency dependence of the failure rates was not observed. It is new findings that the degradation mechanism can not be explained by thermal migration model which Actel advocates. • The antifuse degradation was accelerated by supply voltage: accelerated about 50 times for 2.5V to 3.0V of Vcc. This results showed the degradation mechanism would be percolation model. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application

20. Summary (2) • UMC: • The antifuse failures which were detected in the MEC devices did not take place. • The failure by ESD was not observed. • SEL was not observed. • SEU was not observed in static state but was observed in dynamic state, with frequency dependency: Xe (LET=64[MeV/mg/cm2]). • The increase in power supply current was observed during TID testing. The trend is similar to the Actel’s data. • Thermal Shock Testing • No failure was observed. Briefing: RT54SX-S, RTSX-SU, RTAX-S, and Eclipse FPGAs for Spaceborne Application