Single Event Response of Embedded Power PCs in a Xilinx ......3 USC/ISI Presented by Christian...
Transcript of Single Event Response of Embedded Power PCs in a Xilinx ......3 USC/ISI Presented by Christian...
Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Single Event Response of Embedded Power PCs in a Xilinx Virtex- 4 FPGA
for a Space Application
C. Poivey1, M. Berg1, M. Friendlich1, H. Kim1, D. Petrick2, S. Stansberry3, K. LaBel2,
C. Seidleck1, A. Phan1, T. Irwin1
1 MEI Technologies2 NASA – GSFC
3 USC/ISI
2Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Outline
• Introduction• Test Set-up and Conditions• Destructive Heavy-ion Test• Heavy-ion SEE Test• Proton SEE Test
3Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Introduction-Background• Mission
– Express Logistic Carrier (ELC)– International Space Station (ISS) radiation environment
• Moderate SEE requirements• Moderate TID requirements
• Application– Embedded computer (Space cube)– Uses two XC4VFX60 FPGAs: four Power PC processors
• Each processor is allocated 50% of FPGA fabric and is consideredas an independent node.
– Four processors run independently of each other, and results are voted on in a separate radiation hardened (RH) FPGA.
• RH FPGA traps error condition and flags faulty processor node• RH FPGA restores PPC functionality (warm reset, full reboot,
reconfiguration,…)– No mitigation in FPGA design except scrubbing of
configuration memory to keep processors running as long as possible
• External scrub• Self scrub using FPGA Internal Configuration Access Port (ICAP)
4Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Introduction – Test Approach
• Device– XC4VFX60 FPGA
• Virtex-4, embeds two power PC405 cores
• 90 nm CMOS bulk commercial process
• Some radiation data available on Virtex-4, but not on versions with PC core.
• Test approach– Destructive test (go no go)
• No die thinning– SEE Tests
• Die thinning (100 µm)• Heavy-ion and protons
– TID tests
5Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Test Set-up
Hi-Lo Interface
FX60
JTAG Header
PROM
Mode Select
External Power
(Optional)
General I/O
Select MAP
Power
Power Supply Power Supply or DMM
(Optional)
PC
Interface (RS232)
Program Voltage
Regulator(PT6701)
JTAG
Header
Hi-Lo Interface
FPGA (VII-Pro)
PROM
Voltage Regulators
HSDT
Controller (PC)
GPIB (PC)
Programmer
(PC)
Figure 1. Block Diagram of FX60 Test Setup Electical Interface
For Heating Unit
Heating Unit
FX60 Daughter Board
Thermistor/IR Dector
SRAM
6Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Test Set-up
7Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
DUT Design
+ shift registers
2 versions of DUT design: with and without self-scrubbing
8Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Test Conditions
• Destructive Test– Test samples from Xilinx– Unshaved dice– DUT design without high-
speed link– No implemention with self-
scrubbing– External scrubbing at 32 MHz– PPC test program running
from BRAM– Multi-interrupt test program
sending interrupt requests to each PPC every second
– No readback of configuration memory after each run
– Heavy-ion test at MSU with 106 MeV/u Xe beam
– Tests at 80°C die temperature
• Full SEE tests– Test samples from flight lot– Shaved dice (100 µm) for
heavy-ion test– DUT design with high-speed
link– One implementation with self
scrubbing– External scrubbing at 32 MHz– PPC test programs running
from external SRAM– Multi-interrupt test program
sending interrupt requests to each PPC every 100 ms.
– Counter program – Heavy ion test TAMU with 25
MeV/u beams– Protons tests at IUCF– Readback of configuration
memory after each run– No heating (~65°C die)
9Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Destructive Test•MSU cyclotron:
• Xe beam, 106 MeV/u (14.4 GeV) after scattering foil•12” of Air between scattering foil and DUT•DUT delidded•Flux between 25 and 300 #/cm2-s
LET=19 MeVcm2/mgat normal incidence
LET=30 MeVcm2/mgat 45 degrees inclination
Active part of the die
10Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Heavy-Ion SEE Tests, Readback Errors
0.00E+00
1.00E-01
2.00E-01
3.00E-01
4.00E-01
5.00E-01
6.00E-01
7.00E-01
8.00E-01
9.00E-01
1.00E+00
1 10 100
Eff LET (MeVcm2/mg)
Xse
ctio
n (c
m2 /d
ev)
no scrubexternal scrub
11Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Heavy-Ion SEE Tests, SEFI
1.00E-06
1.00E-03
2.00E-03
3.00E-03
4.00E-03
1 10 100
LET (MeVcm2/mg)
SE
FI c
ross
sec
tion
(cm
2 /dev
)
no scrub
external scrubself scrub
12Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Protons SEE Tests, Readback Errors
0.00E+00
5.00E-08
1.00E-07
1.50E-07
2.00E-07
2.50E-07
3.00E-07
3.50E-07
4.00E-07
10 100 1000
Energy (MeV)
Xse
ctio
n (c
m2 /d
ev)
no scrubself scrub
external scrub
13Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Proton SEE Tests, SEFI
0.00E+00
5.00E-10
1.00E-09
1.50E-09
2.00E-09
2.50E-09
10 100 1000
Energy (MeV)
Xse
ctio
n (c
m2 /d
ev)
no scrubself scrubexternal scrubICAP CL
14Presented by Christian Poivey at the 2007 Single Event Effects Symposium (SEESYM), April 10, 2007 to April 12, 2007 in Long Beach, CA.
Conclusions
• Data are still under analysis– Recovery after SEFI– Readback files– Data errors
• No destructive events up to a LET of 60 MeVcm2/mg
• No SEFI requires a power cycle to recover from